repository_name
stringclasses
316 values
func_path_in_repository
stringlengths
6
223
func_name
stringlengths
1
134
language
stringclasses
1 value
func_code_string
stringlengths
57
65.5k
func_documentation_string
stringlengths
1
46.3k
split_name
stringclasses
1 value
func_code_url
stringlengths
91
315
called_functions
listlengths
1
156
enclosing_scope
stringlengths
2
1.48M
Scifabric/pbs
helpers.py
_update_task_presenter_bundle_js
python
def _update_task_presenter_bundle_js(project): if os.path.isfile ('bundle.min.js'): with open('bundle.min.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js return if os.path.isfile ('bundle.js'): with open('bundle.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js
Append to template a distribution bundle js.
train
https://github.com/Scifabric/pbs/blob/3e5d5f3f0f5d20f740eaacc4d6e872a0c9fb8b38/helpers.py#L90-L101
null
#!/usr/bin/env python # -*- coding: utf-8 -*- # This file is part of PyBOSSA. # # PyBOSSA is free software: you can redistribute it and/or modify # it under the terms of the GNU Affero General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # PyBOSSA is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU Affero General Public License for more details. # # You should have received a copy of the GNU Affero General Public License # along with PyBOSSA. If not, see <http://www.gnu.org/licenses/>. """ Helper functions for the pbs command line client. This module exports the following methods: * find_project_by_short_name: return the project by short_name. * check_api_errors: check for API errors returned by a PyBossa server. * format_error: format error message. * format_json_task: format a CSV row into JSON. """ import re import os import csv import json import time import click import datetime from StringIO import StringIO import polib import openpyxl import itertools from requests import exceptions import requests from pbsexceptions import * import logging from watchdog.observers import Observer from watchdog.events import PatternMatchingEventHandler import calendar __all__ = ['find_project_by_short_name', 'check_api_error', 'format_error', 'format_json_task', '_create_project', '_update_project', '_add_tasks', 'create_task_info', '_delete_tasks', 'enable_auto_throttling', '_update_tasks_redundancy', '_update_project_watch', 'PbsHandler', '_update_task_presenter_bundle_js', 'row_empty', '_add_helpingmaterials', 'create_helping_material_info'] def _create_project(config): """Create a project in a PyBossa server.""" try: response = config.pbclient.create_project(config.project['name'], config.project['short_name'], config.project['description']) check_api_error(response) return ("Project: %s created!" % config.project['short_name']) except exceptions.ConnectionError: return("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_project_watch(config, task_presenter, results, long_description, tutorial): # pragma: no cover """Update a project in a loop.""" logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(message)s', datefmt='%Y-%m-%d %H:%M:%S') path = os.getcwd() event_handler = PbsHandler(config, task_presenter, results, long_description, tutorial) observer = Observer() # We only want the current folder, not sub-folders observer.schedule(event_handler, path, recursive=False) observer.start() try: while True: time.sleep(1) except KeyboardInterrupt: observer.stop() observer.join() def _update_project(config, task_presenter, results, long_description, tutorial): """Update a project.""" try: # Get project project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) # Update attributes project.name = config.project['name'] project.short_name = config.project['short_name'] project.description = config.project['description'] # Update long_description with open(long_description, 'r') as f: project.long_description = f.read() # Update task presenter with open(task_presenter, 'r') as f: project.info['task_presenter'] = f.read() _update_task_presenter_bundle_js(project) # Update results with open(results, 'r') as f: project.info['results'] = f.read() # Update tutorial with open(tutorial, 'r') as f: project.info['tutorial'] = f.read() response = config.pbclient.update_project(project) check_api_error(response) return ("Project %s updated!" % config.project['short_name']) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except ProjectNotFound: return ("Project not found! The project: %s is missing." \ " Use the flag --all=1 to search in all the server " \ % config.project['short_name']) except TaskNotFound: raise def _load_data(data_file, data_type): """Load data from CSV, JSON, Excel, ..., formats.""" raw_data = data_file.read() if data_type is None: data_type = data_file.name.split('.')[-1] # Data list to process data = [] # JSON type if data_type == 'json': data = json.loads(raw_data) return data # CSV type elif data_type == 'csv': csv_data = StringIO(raw_data) reader = csv.DictReader(csv_data, delimiter=',') for line in reader: data.append(line) return data elif data_type in ['xlsx', 'xlsm', 'xltx', 'xltm']: excel_data = StringIO(raw_data) wb = openpyxl.load_workbook(excel_data) ws = wb.active # First headers headers = [] for row in ws.iter_rows(max_row=1): for cell in row: tmp = '_'.join(cell.value.split(" ")).lower() headers.append(tmp) # Simulate DictReader for row in ws.iter_rows(row_offset=1): values = [] for cell in row: values.append(cell.value) tmp = dict(itertools.izip(headers, values)) if len(values) == len(headers) and not row_empty(values): data.append(tmp) return data # PO type elif data_type == 'po': po = polib.pofile(raw_data) for entry in po.untranslated_entries(): data.append(entry.__dict__) return data # PROPERTIES type (used in Java and Firefox extensions) elif data_type == 'properties': lines = raw_data.split('\n') for l in lines: if l: var_id, string = l.split('=') tmp = dict(var_id=var_id, string=string) data.append(tmp) return data else: return data def _add_tasks(config, tasks_file, tasks_type, priority, redundancy): """Add tasks to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(tasks_file, tasks_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # If true, warn user # if sleep: # pragma: no cover # click.secho(msg, fg='yellow') # Show progress bar with click.progressbar(data, label="Adding Tasks") as pgbar: for d in pgbar: task_info = create_task_info(d) response = config.pbclient.create_task(project_id=project.id, info=task_info, n_answers=redundancy, priority_0=priority) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data) check_api_error(response) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s tasks added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _add_helpingmaterials(config, helping_file, helping_type): """Add helping materials to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(helping_file, helping_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # Show progress bar with click.progressbar(data, label="Adding Helping Materials") as pgbar: for d in pgbar: helping_info, file_path = create_helping_material_info(d) if file_path: # Create first the media object hm = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info, file_path=file_path) check_api_error(hm) z = hm.info.copy() z.update(helping_info) hm.info = z response = config.pbclient.update_helping_material(hm) check_api_error(response) else: response = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info) check_api_error(response) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data, endpoint='/api/helpinmaterial') # If true, warn user if sleep: # pragma: no cover click.secho(msg, fg='yellow') # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s helping materials added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _delete_tasks(config, task_id, limit=100, offset=0): """Delete tasks from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.delete_task(task_id) check_api_error(response) return "Task.id = %s and its associated task_runs have been deleted" % task_id else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) while len(tasks) > 0: for t in tasks: response = config.pbclient.delete_task(t.id) check_api_error(response) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks and task_runs have been deleted" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_tasks_redundancy(config, task_id, redundancy, limit=300, offset=0): """Update tasks redundancy from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.find_tasks(project.id, id=task_id) check_api_error(response) task = response[0] task.n_answers = redundancy response = config.pbclient.update_task(task) check_api_error(response) msg = "Task.id = %s redundancy has been updated to %s" % (task_id, redundancy) return msg else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) with click.progressbar(tasks, label="Updating Tasks") as pgbar: while len(tasks) > 0: for t in pgbar: t.n_answers = redundancy response = config.pbclient.update_task(t) check_api_error(response) # Check if for the data we have to auto-throttle task update sleep, msg = enable_auto_throttling(config, tasks) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks redundancy have been updated" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def find_project_by_short_name(short_name, pbclient, all=None): """Return project by short_name.""" try: response = pbclient.find_project(short_name=short_name, all=all) check_api_error(response) if (len(response) == 0): msg = '%s not found! You can use the all=1 argument to \ search in all the server.' error = 'Project Not Found' raise ProjectNotFound(msg, error) return response[0] except exceptions.ConnectionError: raise except ProjectNotFound: raise def check_api_error(api_response): print(api_response) """Check if returned API response contains an error.""" if type(api_response) == dict and 'code' in api_response and api_response['code'] <> 200: print("Server response code: %s" % api_response['code']) print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) if type(api_response) == dict and (api_response.get('status') == 'failed'): if 'ProgrammingError' in api_response.get('exception_cls'): raise DatabaseError(message='PyBossa database error.', error=api_response) if ('DBIntegrityError' in api_response.get('exception_cls') and 'project' in api_response.get('target')): msg = 'PyBossa project already exists.' raise ProjectAlreadyExists(message=msg, error=api_response) if 'project' in api_response.get('target'): raise ProjectNotFound(message='PyBossa Project not found', error=api_response) if 'task' in api_response.get('target'): raise TaskNotFound(message='PyBossa Task not found', error=api_response) else: print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) def format_error(module, error): """Format the error for the given module.""" logging.error(module) # Beautify JSON error print error.message print json.dumps(error.error, sort_keys=True, indent=4, separators=(',', ': ')) exit(1) def create_task_info(task): """Create task_info field.""" task_info = None if task.get('info'): task_info = task['info'] else: task_info = task return task_info def create_helping_material_info(helping): """Create helping_material_info field.""" helping_info = None file_path = None if helping.get('info'): helping_info = helping['info'] else: helping_info = helping if helping_info.get('file_path'): file_path = helping_info.get('file_path') del helping_info['file_path'] return helping_info, file_path def enable_auto_throttling(config, data, limit=299, endpoint='/api/task'): "Return sleep time if more tasks than those " \ "allowed by the server are requested." # Get header from server endpoint = config.server + endpoint headers = requests.head(endpoint).headers # Get limit server_limit = int(headers.get('X-RateLimit-Remaining', 0)) limit = server_limit or limit # Get reset time reset_epoch = int(headers.get('X-RateLimit-Reset', 0)) # Compute sleep time sleep = (reset_epoch - calendar.timegm(datetime.datetime.utcnow().utctimetuple())) msg = 'Warning: %s remaining hits to the endpoint.' \ ' Auto-throttling enabled!' % limit # If we have less than 10 hits on the endpoint, sleep if limit <= 10: return (sleep, msg) else: return 0, None def format_json_task(task_info): """Format task_info into JSON if applicable.""" try: return json.loads(task_info) except: return task_info def row_empty(row): """Check if all values in row are None.""" for value in row: if value is not None: return False return True class PbsHandler(PatternMatchingEventHandler): patterns = ['*/template.html', '*/tutorial.html', '*/long_description.md', '*/results.html', '*/bundle.js', '*/bundle.min.js'] def __init__(self, config, task_presenter, results, long_description, tutorial): super(PbsHandler, self).__init__() self.config = config self.task_presenter = task_presenter self.results = results self.long_description = long_description self.tutorial = tutorial def on_modified(self, event): what = 'directory' if event.is_directory else 'file' logging.info("Modified %s: %s", what, event.src_path) res = _update_project(self.config, self.task_presenter, self.results, self.long_description, self.tutorial) logging.info(res)
Scifabric/pbs
helpers.py
_update_project
python
def _update_project(config, task_presenter, results, long_description, tutorial): try: # Get project project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) # Update attributes project.name = config.project['name'] project.short_name = config.project['short_name'] project.description = config.project['description'] # Update long_description with open(long_description, 'r') as f: project.long_description = f.read() # Update task presenter with open(task_presenter, 'r') as f: project.info['task_presenter'] = f.read() _update_task_presenter_bundle_js(project) # Update results with open(results, 'r') as f: project.info['results'] = f.read() # Update tutorial with open(tutorial, 'r') as f: project.info['tutorial'] = f.read() response = config.pbclient.update_project(project) check_api_error(response) return ("Project %s updated!" % config.project['short_name']) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except ProjectNotFound: return ("Project not found! The project: %s is missing." \ " Use the flag --all=1 to search in all the server " \ % config.project['short_name']) except TaskNotFound: raise
Update a project.
train
https://github.com/Scifabric/pbs/blob/3e5d5f3f0f5d20f740eaacc4d6e872a0c9fb8b38/helpers.py#L103-L138
[ "def find_project_by_short_name(short_name, pbclient, all=None):\n \"\"\"Return project by short_name.\"\"\"\n try:\n response = pbclient.find_project(short_name=short_name, all=all)\n check_api_error(response)\n if (len(response) == 0):\n msg = '%s not found! You can use the all=1 argument to \\\n search in all the server.'\n error = 'Project Not Found'\n raise ProjectNotFound(msg, error)\n return response[0]\n except exceptions.ConnectionError:\n raise\n except ProjectNotFound:\n raise\n", "def check_api_error(api_response):\n print(api_response)\n \"\"\"Check if returned API response contains an error.\"\"\"\n if type(api_response) == dict and 'code' in api_response and api_response['code'] <> 200:\n print(\"Server response code: %s\" % api_response['code'])\n print(\"Server response: %s\" % api_response)\n raise exceptions.HTTPError('Unexpected response', response=api_response)\n if type(api_response) == dict and (api_response.get('status') == 'failed'):\n if 'ProgrammingError' in api_response.get('exception_cls'):\n raise DatabaseError(message='PyBossa database error.',\n error=api_response)\n if ('DBIntegrityError' in api_response.get('exception_cls') and\n 'project' in api_response.get('target')):\n msg = 'PyBossa project already exists.'\n raise ProjectAlreadyExists(message=msg, error=api_response)\n if 'project' in api_response.get('target'):\n raise ProjectNotFound(message='PyBossa Project not found',\n error=api_response)\n if 'task' in api_response.get('target'):\n raise TaskNotFound(message='PyBossa Task not found',\n error=api_response)\n else:\n print(\"Server response: %s\" % api_response)\n raise exceptions.HTTPError('Unexpected response', response=api_response)\n", "def _update_task_presenter_bundle_js(project):\n \"\"\"Append to template a distribution bundle js.\"\"\"\n if os.path.isfile ('bundle.min.js'):\n with open('bundle.min.js') as f:\n js = f.read()\n project.info['task_presenter'] += \"<script>\\n%s\\n</script>\" % js\n return\n\n if os.path.isfile ('bundle.js'):\n with open('bundle.js') as f:\n js = f.read()\n project.info['task_presenter'] += \"<script>\\n%s\\n</script>\" % js\n" ]
#!/usr/bin/env python # -*- coding: utf-8 -*- # This file is part of PyBOSSA. # # PyBOSSA is free software: you can redistribute it and/or modify # it under the terms of the GNU Affero General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # PyBOSSA is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU Affero General Public License for more details. # # You should have received a copy of the GNU Affero General Public License # along with PyBOSSA. If not, see <http://www.gnu.org/licenses/>. """ Helper functions for the pbs command line client. This module exports the following methods: * find_project_by_short_name: return the project by short_name. * check_api_errors: check for API errors returned by a PyBossa server. * format_error: format error message. * format_json_task: format a CSV row into JSON. """ import re import os import csv import json import time import click import datetime from StringIO import StringIO import polib import openpyxl import itertools from requests import exceptions import requests from pbsexceptions import * import logging from watchdog.observers import Observer from watchdog.events import PatternMatchingEventHandler import calendar __all__ = ['find_project_by_short_name', 'check_api_error', 'format_error', 'format_json_task', '_create_project', '_update_project', '_add_tasks', 'create_task_info', '_delete_tasks', 'enable_auto_throttling', '_update_tasks_redundancy', '_update_project_watch', 'PbsHandler', '_update_task_presenter_bundle_js', 'row_empty', '_add_helpingmaterials', 'create_helping_material_info'] def _create_project(config): """Create a project in a PyBossa server.""" try: response = config.pbclient.create_project(config.project['name'], config.project['short_name'], config.project['description']) check_api_error(response) return ("Project: %s created!" % config.project['short_name']) except exceptions.ConnectionError: return("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_project_watch(config, task_presenter, results, long_description, tutorial): # pragma: no cover """Update a project in a loop.""" logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(message)s', datefmt='%Y-%m-%d %H:%M:%S') path = os.getcwd() event_handler = PbsHandler(config, task_presenter, results, long_description, tutorial) observer = Observer() # We only want the current folder, not sub-folders observer.schedule(event_handler, path, recursive=False) observer.start() try: while True: time.sleep(1) except KeyboardInterrupt: observer.stop() observer.join() def _update_task_presenter_bundle_js(project): """Append to template a distribution bundle js.""" if os.path.isfile ('bundle.min.js'): with open('bundle.min.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js return if os.path.isfile ('bundle.js'): with open('bundle.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js def _load_data(data_file, data_type): """Load data from CSV, JSON, Excel, ..., formats.""" raw_data = data_file.read() if data_type is None: data_type = data_file.name.split('.')[-1] # Data list to process data = [] # JSON type if data_type == 'json': data = json.loads(raw_data) return data # CSV type elif data_type == 'csv': csv_data = StringIO(raw_data) reader = csv.DictReader(csv_data, delimiter=',') for line in reader: data.append(line) return data elif data_type in ['xlsx', 'xlsm', 'xltx', 'xltm']: excel_data = StringIO(raw_data) wb = openpyxl.load_workbook(excel_data) ws = wb.active # First headers headers = [] for row in ws.iter_rows(max_row=1): for cell in row: tmp = '_'.join(cell.value.split(" ")).lower() headers.append(tmp) # Simulate DictReader for row in ws.iter_rows(row_offset=1): values = [] for cell in row: values.append(cell.value) tmp = dict(itertools.izip(headers, values)) if len(values) == len(headers) and not row_empty(values): data.append(tmp) return data # PO type elif data_type == 'po': po = polib.pofile(raw_data) for entry in po.untranslated_entries(): data.append(entry.__dict__) return data # PROPERTIES type (used in Java and Firefox extensions) elif data_type == 'properties': lines = raw_data.split('\n') for l in lines: if l: var_id, string = l.split('=') tmp = dict(var_id=var_id, string=string) data.append(tmp) return data else: return data def _add_tasks(config, tasks_file, tasks_type, priority, redundancy): """Add tasks to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(tasks_file, tasks_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # If true, warn user # if sleep: # pragma: no cover # click.secho(msg, fg='yellow') # Show progress bar with click.progressbar(data, label="Adding Tasks") as pgbar: for d in pgbar: task_info = create_task_info(d) response = config.pbclient.create_task(project_id=project.id, info=task_info, n_answers=redundancy, priority_0=priority) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data) check_api_error(response) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s tasks added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _add_helpingmaterials(config, helping_file, helping_type): """Add helping materials to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(helping_file, helping_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # Show progress bar with click.progressbar(data, label="Adding Helping Materials") as pgbar: for d in pgbar: helping_info, file_path = create_helping_material_info(d) if file_path: # Create first the media object hm = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info, file_path=file_path) check_api_error(hm) z = hm.info.copy() z.update(helping_info) hm.info = z response = config.pbclient.update_helping_material(hm) check_api_error(response) else: response = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info) check_api_error(response) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data, endpoint='/api/helpinmaterial') # If true, warn user if sleep: # pragma: no cover click.secho(msg, fg='yellow') # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s helping materials added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _delete_tasks(config, task_id, limit=100, offset=0): """Delete tasks from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.delete_task(task_id) check_api_error(response) return "Task.id = %s and its associated task_runs have been deleted" % task_id else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) while len(tasks) > 0: for t in tasks: response = config.pbclient.delete_task(t.id) check_api_error(response) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks and task_runs have been deleted" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_tasks_redundancy(config, task_id, redundancy, limit=300, offset=0): """Update tasks redundancy from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.find_tasks(project.id, id=task_id) check_api_error(response) task = response[0] task.n_answers = redundancy response = config.pbclient.update_task(task) check_api_error(response) msg = "Task.id = %s redundancy has been updated to %s" % (task_id, redundancy) return msg else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) with click.progressbar(tasks, label="Updating Tasks") as pgbar: while len(tasks) > 0: for t in pgbar: t.n_answers = redundancy response = config.pbclient.update_task(t) check_api_error(response) # Check if for the data we have to auto-throttle task update sleep, msg = enable_auto_throttling(config, tasks) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks redundancy have been updated" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def find_project_by_short_name(short_name, pbclient, all=None): """Return project by short_name.""" try: response = pbclient.find_project(short_name=short_name, all=all) check_api_error(response) if (len(response) == 0): msg = '%s not found! You can use the all=1 argument to \ search in all the server.' error = 'Project Not Found' raise ProjectNotFound(msg, error) return response[0] except exceptions.ConnectionError: raise except ProjectNotFound: raise def check_api_error(api_response): print(api_response) """Check if returned API response contains an error.""" if type(api_response) == dict and 'code' in api_response and api_response['code'] <> 200: print("Server response code: %s" % api_response['code']) print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) if type(api_response) == dict and (api_response.get('status') == 'failed'): if 'ProgrammingError' in api_response.get('exception_cls'): raise DatabaseError(message='PyBossa database error.', error=api_response) if ('DBIntegrityError' in api_response.get('exception_cls') and 'project' in api_response.get('target')): msg = 'PyBossa project already exists.' raise ProjectAlreadyExists(message=msg, error=api_response) if 'project' in api_response.get('target'): raise ProjectNotFound(message='PyBossa Project not found', error=api_response) if 'task' in api_response.get('target'): raise TaskNotFound(message='PyBossa Task not found', error=api_response) else: print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) def format_error(module, error): """Format the error for the given module.""" logging.error(module) # Beautify JSON error print error.message print json.dumps(error.error, sort_keys=True, indent=4, separators=(',', ': ')) exit(1) def create_task_info(task): """Create task_info field.""" task_info = None if task.get('info'): task_info = task['info'] else: task_info = task return task_info def create_helping_material_info(helping): """Create helping_material_info field.""" helping_info = None file_path = None if helping.get('info'): helping_info = helping['info'] else: helping_info = helping if helping_info.get('file_path'): file_path = helping_info.get('file_path') del helping_info['file_path'] return helping_info, file_path def enable_auto_throttling(config, data, limit=299, endpoint='/api/task'): "Return sleep time if more tasks than those " \ "allowed by the server are requested." # Get header from server endpoint = config.server + endpoint headers = requests.head(endpoint).headers # Get limit server_limit = int(headers.get('X-RateLimit-Remaining', 0)) limit = server_limit or limit # Get reset time reset_epoch = int(headers.get('X-RateLimit-Reset', 0)) # Compute sleep time sleep = (reset_epoch - calendar.timegm(datetime.datetime.utcnow().utctimetuple())) msg = 'Warning: %s remaining hits to the endpoint.' \ ' Auto-throttling enabled!' % limit # If we have less than 10 hits on the endpoint, sleep if limit <= 10: return (sleep, msg) else: return 0, None def format_json_task(task_info): """Format task_info into JSON if applicable.""" try: return json.loads(task_info) except: return task_info def row_empty(row): """Check if all values in row are None.""" for value in row: if value is not None: return False return True class PbsHandler(PatternMatchingEventHandler): patterns = ['*/template.html', '*/tutorial.html', '*/long_description.md', '*/results.html', '*/bundle.js', '*/bundle.min.js'] def __init__(self, config, task_presenter, results, long_description, tutorial): super(PbsHandler, self).__init__() self.config = config self.task_presenter = task_presenter self.results = results self.long_description = long_description self.tutorial = tutorial def on_modified(self, event): what = 'directory' if event.is_directory else 'file' logging.info("Modified %s: %s", what, event.src_path) res = _update_project(self.config, self.task_presenter, self.results, self.long_description, self.tutorial) logging.info(res)
Scifabric/pbs
helpers.py
_load_data
python
def _load_data(data_file, data_type): raw_data = data_file.read() if data_type is None: data_type = data_file.name.split('.')[-1] # Data list to process data = [] # JSON type if data_type == 'json': data = json.loads(raw_data) return data # CSV type elif data_type == 'csv': csv_data = StringIO(raw_data) reader = csv.DictReader(csv_data, delimiter=',') for line in reader: data.append(line) return data elif data_type in ['xlsx', 'xlsm', 'xltx', 'xltm']: excel_data = StringIO(raw_data) wb = openpyxl.load_workbook(excel_data) ws = wb.active # First headers headers = [] for row in ws.iter_rows(max_row=1): for cell in row: tmp = '_'.join(cell.value.split(" ")).lower() headers.append(tmp) # Simulate DictReader for row in ws.iter_rows(row_offset=1): values = [] for cell in row: values.append(cell.value) tmp = dict(itertools.izip(headers, values)) if len(values) == len(headers) and not row_empty(values): data.append(tmp) return data # PO type elif data_type == 'po': po = polib.pofile(raw_data) for entry in po.untranslated_entries(): data.append(entry.__dict__) return data # PROPERTIES type (used in Java and Firefox extensions) elif data_type == 'properties': lines = raw_data.split('\n') for l in lines: if l: var_id, string = l.split('=') tmp = dict(var_id=var_id, string=string) data.append(tmp) return data else: return data
Load data from CSV, JSON, Excel, ..., formats.
train
https://github.com/Scifabric/pbs/blob/3e5d5f3f0f5d20f740eaacc4d6e872a0c9fb8b38/helpers.py#L141-L194
null
#!/usr/bin/env python # -*- coding: utf-8 -*- # This file is part of PyBOSSA. # # PyBOSSA is free software: you can redistribute it and/or modify # it under the terms of the GNU Affero General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # PyBOSSA is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU Affero General Public License for more details. # # You should have received a copy of the GNU Affero General Public License # along with PyBOSSA. If not, see <http://www.gnu.org/licenses/>. """ Helper functions for the pbs command line client. This module exports the following methods: * find_project_by_short_name: return the project by short_name. * check_api_errors: check for API errors returned by a PyBossa server. * format_error: format error message. * format_json_task: format a CSV row into JSON. """ import re import os import csv import json import time import click import datetime from StringIO import StringIO import polib import openpyxl import itertools from requests import exceptions import requests from pbsexceptions import * import logging from watchdog.observers import Observer from watchdog.events import PatternMatchingEventHandler import calendar __all__ = ['find_project_by_short_name', 'check_api_error', 'format_error', 'format_json_task', '_create_project', '_update_project', '_add_tasks', 'create_task_info', '_delete_tasks', 'enable_auto_throttling', '_update_tasks_redundancy', '_update_project_watch', 'PbsHandler', '_update_task_presenter_bundle_js', 'row_empty', '_add_helpingmaterials', 'create_helping_material_info'] def _create_project(config): """Create a project in a PyBossa server.""" try: response = config.pbclient.create_project(config.project['name'], config.project['short_name'], config.project['description']) check_api_error(response) return ("Project: %s created!" % config.project['short_name']) except exceptions.ConnectionError: return("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_project_watch(config, task_presenter, results, long_description, tutorial): # pragma: no cover """Update a project in a loop.""" logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(message)s', datefmt='%Y-%m-%d %H:%M:%S') path = os.getcwd() event_handler = PbsHandler(config, task_presenter, results, long_description, tutorial) observer = Observer() # We only want the current folder, not sub-folders observer.schedule(event_handler, path, recursive=False) observer.start() try: while True: time.sleep(1) except KeyboardInterrupt: observer.stop() observer.join() def _update_task_presenter_bundle_js(project): """Append to template a distribution bundle js.""" if os.path.isfile ('bundle.min.js'): with open('bundle.min.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js return if os.path.isfile ('bundle.js'): with open('bundle.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js def _update_project(config, task_presenter, results, long_description, tutorial): """Update a project.""" try: # Get project project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) # Update attributes project.name = config.project['name'] project.short_name = config.project['short_name'] project.description = config.project['description'] # Update long_description with open(long_description, 'r') as f: project.long_description = f.read() # Update task presenter with open(task_presenter, 'r') as f: project.info['task_presenter'] = f.read() _update_task_presenter_bundle_js(project) # Update results with open(results, 'r') as f: project.info['results'] = f.read() # Update tutorial with open(tutorial, 'r') as f: project.info['tutorial'] = f.read() response = config.pbclient.update_project(project) check_api_error(response) return ("Project %s updated!" % config.project['short_name']) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except ProjectNotFound: return ("Project not found! The project: %s is missing." \ " Use the flag --all=1 to search in all the server " \ % config.project['short_name']) except TaskNotFound: raise def _add_tasks(config, tasks_file, tasks_type, priority, redundancy): """Add tasks to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(tasks_file, tasks_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # If true, warn user # if sleep: # pragma: no cover # click.secho(msg, fg='yellow') # Show progress bar with click.progressbar(data, label="Adding Tasks") as pgbar: for d in pgbar: task_info = create_task_info(d) response = config.pbclient.create_task(project_id=project.id, info=task_info, n_answers=redundancy, priority_0=priority) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data) check_api_error(response) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s tasks added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _add_helpingmaterials(config, helping_file, helping_type): """Add helping materials to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(helping_file, helping_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # Show progress bar with click.progressbar(data, label="Adding Helping Materials") as pgbar: for d in pgbar: helping_info, file_path = create_helping_material_info(d) if file_path: # Create first the media object hm = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info, file_path=file_path) check_api_error(hm) z = hm.info.copy() z.update(helping_info) hm.info = z response = config.pbclient.update_helping_material(hm) check_api_error(response) else: response = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info) check_api_error(response) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data, endpoint='/api/helpinmaterial') # If true, warn user if sleep: # pragma: no cover click.secho(msg, fg='yellow') # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s helping materials added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _delete_tasks(config, task_id, limit=100, offset=0): """Delete tasks from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.delete_task(task_id) check_api_error(response) return "Task.id = %s and its associated task_runs have been deleted" % task_id else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) while len(tasks) > 0: for t in tasks: response = config.pbclient.delete_task(t.id) check_api_error(response) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks and task_runs have been deleted" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_tasks_redundancy(config, task_id, redundancy, limit=300, offset=0): """Update tasks redundancy from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.find_tasks(project.id, id=task_id) check_api_error(response) task = response[0] task.n_answers = redundancy response = config.pbclient.update_task(task) check_api_error(response) msg = "Task.id = %s redundancy has been updated to %s" % (task_id, redundancy) return msg else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) with click.progressbar(tasks, label="Updating Tasks") as pgbar: while len(tasks) > 0: for t in pgbar: t.n_answers = redundancy response = config.pbclient.update_task(t) check_api_error(response) # Check if for the data we have to auto-throttle task update sleep, msg = enable_auto_throttling(config, tasks) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks redundancy have been updated" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def find_project_by_short_name(short_name, pbclient, all=None): """Return project by short_name.""" try: response = pbclient.find_project(short_name=short_name, all=all) check_api_error(response) if (len(response) == 0): msg = '%s not found! You can use the all=1 argument to \ search in all the server.' error = 'Project Not Found' raise ProjectNotFound(msg, error) return response[0] except exceptions.ConnectionError: raise except ProjectNotFound: raise def check_api_error(api_response): print(api_response) """Check if returned API response contains an error.""" if type(api_response) == dict and 'code' in api_response and api_response['code'] <> 200: print("Server response code: %s" % api_response['code']) print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) if type(api_response) == dict and (api_response.get('status') == 'failed'): if 'ProgrammingError' in api_response.get('exception_cls'): raise DatabaseError(message='PyBossa database error.', error=api_response) if ('DBIntegrityError' in api_response.get('exception_cls') and 'project' in api_response.get('target')): msg = 'PyBossa project already exists.' raise ProjectAlreadyExists(message=msg, error=api_response) if 'project' in api_response.get('target'): raise ProjectNotFound(message='PyBossa Project not found', error=api_response) if 'task' in api_response.get('target'): raise TaskNotFound(message='PyBossa Task not found', error=api_response) else: print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) def format_error(module, error): """Format the error for the given module.""" logging.error(module) # Beautify JSON error print error.message print json.dumps(error.error, sort_keys=True, indent=4, separators=(',', ': ')) exit(1) def create_task_info(task): """Create task_info field.""" task_info = None if task.get('info'): task_info = task['info'] else: task_info = task return task_info def create_helping_material_info(helping): """Create helping_material_info field.""" helping_info = None file_path = None if helping.get('info'): helping_info = helping['info'] else: helping_info = helping if helping_info.get('file_path'): file_path = helping_info.get('file_path') del helping_info['file_path'] return helping_info, file_path def enable_auto_throttling(config, data, limit=299, endpoint='/api/task'): "Return sleep time if more tasks than those " \ "allowed by the server are requested." # Get header from server endpoint = config.server + endpoint headers = requests.head(endpoint).headers # Get limit server_limit = int(headers.get('X-RateLimit-Remaining', 0)) limit = server_limit or limit # Get reset time reset_epoch = int(headers.get('X-RateLimit-Reset', 0)) # Compute sleep time sleep = (reset_epoch - calendar.timegm(datetime.datetime.utcnow().utctimetuple())) msg = 'Warning: %s remaining hits to the endpoint.' \ ' Auto-throttling enabled!' % limit # If we have less than 10 hits on the endpoint, sleep if limit <= 10: return (sleep, msg) else: return 0, None def format_json_task(task_info): """Format task_info into JSON if applicable.""" try: return json.loads(task_info) except: return task_info def row_empty(row): """Check if all values in row are None.""" for value in row: if value is not None: return False return True class PbsHandler(PatternMatchingEventHandler): patterns = ['*/template.html', '*/tutorial.html', '*/long_description.md', '*/results.html', '*/bundle.js', '*/bundle.min.js'] def __init__(self, config, task_presenter, results, long_description, tutorial): super(PbsHandler, self).__init__() self.config = config self.task_presenter = task_presenter self.results = results self.long_description = long_description self.tutorial = tutorial def on_modified(self, event): what = 'directory' if event.is_directory else 'file' logging.info("Modified %s: %s", what, event.src_path) res = _update_project(self.config, self.task_presenter, self.results, self.long_description, self.tutorial) logging.info(res)
Scifabric/pbs
helpers.py
_add_tasks
python
def _add_tasks(config, tasks_file, tasks_type, priority, redundancy): try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(tasks_file, tasks_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # If true, warn user # if sleep: # pragma: no cover # click.secho(msg, fg='yellow') # Show progress bar with click.progressbar(data, label="Adding Tasks") as pgbar: for d in pgbar: task_info = create_task_info(d) response = config.pbclient.create_task(project_id=project.id, info=task_info, n_answers=redundancy, priority_0=priority) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data) check_api_error(response) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s tasks added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise
Add tasks to a project.
train
https://github.com/Scifabric/pbs/blob/3e5d5f3f0f5d20f740eaacc4d6e872a0c9fb8b38/helpers.py#L197-L230
null
#!/usr/bin/env python # -*- coding: utf-8 -*- # This file is part of PyBOSSA. # # PyBOSSA is free software: you can redistribute it and/or modify # it under the terms of the GNU Affero General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # PyBOSSA is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU Affero General Public License for more details. # # You should have received a copy of the GNU Affero General Public License # along with PyBOSSA. If not, see <http://www.gnu.org/licenses/>. """ Helper functions for the pbs command line client. This module exports the following methods: * find_project_by_short_name: return the project by short_name. * check_api_errors: check for API errors returned by a PyBossa server. * format_error: format error message. * format_json_task: format a CSV row into JSON. """ import re import os import csv import json import time import click import datetime from StringIO import StringIO import polib import openpyxl import itertools from requests import exceptions import requests from pbsexceptions import * import logging from watchdog.observers import Observer from watchdog.events import PatternMatchingEventHandler import calendar __all__ = ['find_project_by_short_name', 'check_api_error', 'format_error', 'format_json_task', '_create_project', '_update_project', '_add_tasks', 'create_task_info', '_delete_tasks', 'enable_auto_throttling', '_update_tasks_redundancy', '_update_project_watch', 'PbsHandler', '_update_task_presenter_bundle_js', 'row_empty', '_add_helpingmaterials', 'create_helping_material_info'] def _create_project(config): """Create a project in a PyBossa server.""" try: response = config.pbclient.create_project(config.project['name'], config.project['short_name'], config.project['description']) check_api_error(response) return ("Project: %s created!" % config.project['short_name']) except exceptions.ConnectionError: return("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_project_watch(config, task_presenter, results, long_description, tutorial): # pragma: no cover """Update a project in a loop.""" logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(message)s', datefmt='%Y-%m-%d %H:%M:%S') path = os.getcwd() event_handler = PbsHandler(config, task_presenter, results, long_description, tutorial) observer = Observer() # We only want the current folder, not sub-folders observer.schedule(event_handler, path, recursive=False) observer.start() try: while True: time.sleep(1) except KeyboardInterrupt: observer.stop() observer.join() def _update_task_presenter_bundle_js(project): """Append to template a distribution bundle js.""" if os.path.isfile ('bundle.min.js'): with open('bundle.min.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js return if os.path.isfile ('bundle.js'): with open('bundle.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js def _update_project(config, task_presenter, results, long_description, tutorial): """Update a project.""" try: # Get project project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) # Update attributes project.name = config.project['name'] project.short_name = config.project['short_name'] project.description = config.project['description'] # Update long_description with open(long_description, 'r') as f: project.long_description = f.read() # Update task presenter with open(task_presenter, 'r') as f: project.info['task_presenter'] = f.read() _update_task_presenter_bundle_js(project) # Update results with open(results, 'r') as f: project.info['results'] = f.read() # Update tutorial with open(tutorial, 'r') as f: project.info['tutorial'] = f.read() response = config.pbclient.update_project(project) check_api_error(response) return ("Project %s updated!" % config.project['short_name']) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except ProjectNotFound: return ("Project not found! The project: %s is missing." \ " Use the flag --all=1 to search in all the server " \ % config.project['short_name']) except TaskNotFound: raise def _load_data(data_file, data_type): """Load data from CSV, JSON, Excel, ..., formats.""" raw_data = data_file.read() if data_type is None: data_type = data_file.name.split('.')[-1] # Data list to process data = [] # JSON type if data_type == 'json': data = json.loads(raw_data) return data # CSV type elif data_type == 'csv': csv_data = StringIO(raw_data) reader = csv.DictReader(csv_data, delimiter=',') for line in reader: data.append(line) return data elif data_type in ['xlsx', 'xlsm', 'xltx', 'xltm']: excel_data = StringIO(raw_data) wb = openpyxl.load_workbook(excel_data) ws = wb.active # First headers headers = [] for row in ws.iter_rows(max_row=1): for cell in row: tmp = '_'.join(cell.value.split(" ")).lower() headers.append(tmp) # Simulate DictReader for row in ws.iter_rows(row_offset=1): values = [] for cell in row: values.append(cell.value) tmp = dict(itertools.izip(headers, values)) if len(values) == len(headers) and not row_empty(values): data.append(tmp) return data # PO type elif data_type == 'po': po = polib.pofile(raw_data) for entry in po.untranslated_entries(): data.append(entry.__dict__) return data # PROPERTIES type (used in Java and Firefox extensions) elif data_type == 'properties': lines = raw_data.split('\n') for l in lines: if l: var_id, string = l.split('=') tmp = dict(var_id=var_id, string=string) data.append(tmp) return data else: return data def _add_helpingmaterials(config, helping_file, helping_type): """Add helping materials to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(helping_file, helping_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # Show progress bar with click.progressbar(data, label="Adding Helping Materials") as pgbar: for d in pgbar: helping_info, file_path = create_helping_material_info(d) if file_path: # Create first the media object hm = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info, file_path=file_path) check_api_error(hm) z = hm.info.copy() z.update(helping_info) hm.info = z response = config.pbclient.update_helping_material(hm) check_api_error(response) else: response = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info) check_api_error(response) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data, endpoint='/api/helpinmaterial') # If true, warn user if sleep: # pragma: no cover click.secho(msg, fg='yellow') # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s helping materials added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _delete_tasks(config, task_id, limit=100, offset=0): """Delete tasks from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.delete_task(task_id) check_api_error(response) return "Task.id = %s and its associated task_runs have been deleted" % task_id else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) while len(tasks) > 0: for t in tasks: response = config.pbclient.delete_task(t.id) check_api_error(response) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks and task_runs have been deleted" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_tasks_redundancy(config, task_id, redundancy, limit=300, offset=0): """Update tasks redundancy from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.find_tasks(project.id, id=task_id) check_api_error(response) task = response[0] task.n_answers = redundancy response = config.pbclient.update_task(task) check_api_error(response) msg = "Task.id = %s redundancy has been updated to %s" % (task_id, redundancy) return msg else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) with click.progressbar(tasks, label="Updating Tasks") as pgbar: while len(tasks) > 0: for t in pgbar: t.n_answers = redundancy response = config.pbclient.update_task(t) check_api_error(response) # Check if for the data we have to auto-throttle task update sleep, msg = enable_auto_throttling(config, tasks) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks redundancy have been updated" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def find_project_by_short_name(short_name, pbclient, all=None): """Return project by short_name.""" try: response = pbclient.find_project(short_name=short_name, all=all) check_api_error(response) if (len(response) == 0): msg = '%s not found! You can use the all=1 argument to \ search in all the server.' error = 'Project Not Found' raise ProjectNotFound(msg, error) return response[0] except exceptions.ConnectionError: raise except ProjectNotFound: raise def check_api_error(api_response): print(api_response) """Check if returned API response contains an error.""" if type(api_response) == dict and 'code' in api_response and api_response['code'] <> 200: print("Server response code: %s" % api_response['code']) print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) if type(api_response) == dict and (api_response.get('status') == 'failed'): if 'ProgrammingError' in api_response.get('exception_cls'): raise DatabaseError(message='PyBossa database error.', error=api_response) if ('DBIntegrityError' in api_response.get('exception_cls') and 'project' in api_response.get('target')): msg = 'PyBossa project already exists.' raise ProjectAlreadyExists(message=msg, error=api_response) if 'project' in api_response.get('target'): raise ProjectNotFound(message='PyBossa Project not found', error=api_response) if 'task' in api_response.get('target'): raise TaskNotFound(message='PyBossa Task not found', error=api_response) else: print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) def format_error(module, error): """Format the error for the given module.""" logging.error(module) # Beautify JSON error print error.message print json.dumps(error.error, sort_keys=True, indent=4, separators=(',', ': ')) exit(1) def create_task_info(task): """Create task_info field.""" task_info = None if task.get('info'): task_info = task['info'] else: task_info = task return task_info def create_helping_material_info(helping): """Create helping_material_info field.""" helping_info = None file_path = None if helping.get('info'): helping_info = helping['info'] else: helping_info = helping if helping_info.get('file_path'): file_path = helping_info.get('file_path') del helping_info['file_path'] return helping_info, file_path def enable_auto_throttling(config, data, limit=299, endpoint='/api/task'): "Return sleep time if more tasks than those " \ "allowed by the server are requested." # Get header from server endpoint = config.server + endpoint headers = requests.head(endpoint).headers # Get limit server_limit = int(headers.get('X-RateLimit-Remaining', 0)) limit = server_limit or limit # Get reset time reset_epoch = int(headers.get('X-RateLimit-Reset', 0)) # Compute sleep time sleep = (reset_epoch - calendar.timegm(datetime.datetime.utcnow().utctimetuple())) msg = 'Warning: %s remaining hits to the endpoint.' \ ' Auto-throttling enabled!' % limit # If we have less than 10 hits on the endpoint, sleep if limit <= 10: return (sleep, msg) else: return 0, None def format_json_task(task_info): """Format task_info into JSON if applicable.""" try: return json.loads(task_info) except: return task_info def row_empty(row): """Check if all values in row are None.""" for value in row: if value is not None: return False return True class PbsHandler(PatternMatchingEventHandler): patterns = ['*/template.html', '*/tutorial.html', '*/long_description.md', '*/results.html', '*/bundle.js', '*/bundle.min.js'] def __init__(self, config, task_presenter, results, long_description, tutorial): super(PbsHandler, self).__init__() self.config = config self.task_presenter = task_presenter self.results = results self.long_description = long_description self.tutorial = tutorial def on_modified(self, event): what = 'directory' if event.is_directory else 'file' logging.info("Modified %s: %s", what, event.src_path) res = _update_project(self.config, self.task_presenter, self.results, self.long_description, self.tutorial) logging.info(res)
Scifabric/pbs
helpers.py
_add_helpingmaterials
python
def _add_helpingmaterials(config, helping_file, helping_type): try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(helping_file, helping_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # Show progress bar with click.progressbar(data, label="Adding Helping Materials") as pgbar: for d in pgbar: helping_info, file_path = create_helping_material_info(d) if file_path: # Create first the media object hm = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info, file_path=file_path) check_api_error(hm) z = hm.info.copy() z.update(helping_info) hm.info = z response = config.pbclient.update_helping_material(hm) check_api_error(response) else: response = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info) check_api_error(response) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data, endpoint='/api/helpinmaterial') # If true, warn user if sleep: # pragma: no cover click.secho(msg, fg='yellow') # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s helping materials added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise
Add helping materials to a project.
train
https://github.com/Scifabric/pbs/blob/3e5d5f3f0f5d20f740eaacc4d6e872a0c9fb8b38/helpers.py#L233-L277
null
#!/usr/bin/env python # -*- coding: utf-8 -*- # This file is part of PyBOSSA. # # PyBOSSA is free software: you can redistribute it and/or modify # it under the terms of the GNU Affero General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # PyBOSSA is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU Affero General Public License for more details. # # You should have received a copy of the GNU Affero General Public License # along with PyBOSSA. If not, see <http://www.gnu.org/licenses/>. """ Helper functions for the pbs command line client. This module exports the following methods: * find_project_by_short_name: return the project by short_name. * check_api_errors: check for API errors returned by a PyBossa server. * format_error: format error message. * format_json_task: format a CSV row into JSON. """ import re import os import csv import json import time import click import datetime from StringIO import StringIO import polib import openpyxl import itertools from requests import exceptions import requests from pbsexceptions import * import logging from watchdog.observers import Observer from watchdog.events import PatternMatchingEventHandler import calendar __all__ = ['find_project_by_short_name', 'check_api_error', 'format_error', 'format_json_task', '_create_project', '_update_project', '_add_tasks', 'create_task_info', '_delete_tasks', 'enable_auto_throttling', '_update_tasks_redundancy', '_update_project_watch', 'PbsHandler', '_update_task_presenter_bundle_js', 'row_empty', '_add_helpingmaterials', 'create_helping_material_info'] def _create_project(config): """Create a project in a PyBossa server.""" try: response = config.pbclient.create_project(config.project['name'], config.project['short_name'], config.project['description']) check_api_error(response) return ("Project: %s created!" % config.project['short_name']) except exceptions.ConnectionError: return("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_project_watch(config, task_presenter, results, long_description, tutorial): # pragma: no cover """Update a project in a loop.""" logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(message)s', datefmt='%Y-%m-%d %H:%M:%S') path = os.getcwd() event_handler = PbsHandler(config, task_presenter, results, long_description, tutorial) observer = Observer() # We only want the current folder, not sub-folders observer.schedule(event_handler, path, recursive=False) observer.start() try: while True: time.sleep(1) except KeyboardInterrupt: observer.stop() observer.join() def _update_task_presenter_bundle_js(project): """Append to template a distribution bundle js.""" if os.path.isfile ('bundle.min.js'): with open('bundle.min.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js return if os.path.isfile ('bundle.js'): with open('bundle.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js def _update_project(config, task_presenter, results, long_description, tutorial): """Update a project.""" try: # Get project project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) # Update attributes project.name = config.project['name'] project.short_name = config.project['short_name'] project.description = config.project['description'] # Update long_description with open(long_description, 'r') as f: project.long_description = f.read() # Update task presenter with open(task_presenter, 'r') as f: project.info['task_presenter'] = f.read() _update_task_presenter_bundle_js(project) # Update results with open(results, 'r') as f: project.info['results'] = f.read() # Update tutorial with open(tutorial, 'r') as f: project.info['tutorial'] = f.read() response = config.pbclient.update_project(project) check_api_error(response) return ("Project %s updated!" % config.project['short_name']) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except ProjectNotFound: return ("Project not found! The project: %s is missing." \ " Use the flag --all=1 to search in all the server " \ % config.project['short_name']) except TaskNotFound: raise def _load_data(data_file, data_type): """Load data from CSV, JSON, Excel, ..., formats.""" raw_data = data_file.read() if data_type is None: data_type = data_file.name.split('.')[-1] # Data list to process data = [] # JSON type if data_type == 'json': data = json.loads(raw_data) return data # CSV type elif data_type == 'csv': csv_data = StringIO(raw_data) reader = csv.DictReader(csv_data, delimiter=',') for line in reader: data.append(line) return data elif data_type in ['xlsx', 'xlsm', 'xltx', 'xltm']: excel_data = StringIO(raw_data) wb = openpyxl.load_workbook(excel_data) ws = wb.active # First headers headers = [] for row in ws.iter_rows(max_row=1): for cell in row: tmp = '_'.join(cell.value.split(" ")).lower() headers.append(tmp) # Simulate DictReader for row in ws.iter_rows(row_offset=1): values = [] for cell in row: values.append(cell.value) tmp = dict(itertools.izip(headers, values)) if len(values) == len(headers) and not row_empty(values): data.append(tmp) return data # PO type elif data_type == 'po': po = polib.pofile(raw_data) for entry in po.untranslated_entries(): data.append(entry.__dict__) return data # PROPERTIES type (used in Java and Firefox extensions) elif data_type == 'properties': lines = raw_data.split('\n') for l in lines: if l: var_id, string = l.split('=') tmp = dict(var_id=var_id, string=string) data.append(tmp) return data else: return data def _add_tasks(config, tasks_file, tasks_type, priority, redundancy): """Add tasks to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(tasks_file, tasks_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # If true, warn user # if sleep: # pragma: no cover # click.secho(msg, fg='yellow') # Show progress bar with click.progressbar(data, label="Adding Tasks") as pgbar: for d in pgbar: task_info = create_task_info(d) response = config.pbclient.create_task(project_id=project.id, info=task_info, n_answers=redundancy, priority_0=priority) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data) check_api_error(response) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s tasks added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _delete_tasks(config, task_id, limit=100, offset=0): """Delete tasks from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.delete_task(task_id) check_api_error(response) return "Task.id = %s and its associated task_runs have been deleted" % task_id else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) while len(tasks) > 0: for t in tasks: response = config.pbclient.delete_task(t.id) check_api_error(response) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks and task_runs have been deleted" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_tasks_redundancy(config, task_id, redundancy, limit=300, offset=0): """Update tasks redundancy from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.find_tasks(project.id, id=task_id) check_api_error(response) task = response[0] task.n_answers = redundancy response = config.pbclient.update_task(task) check_api_error(response) msg = "Task.id = %s redundancy has been updated to %s" % (task_id, redundancy) return msg else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) with click.progressbar(tasks, label="Updating Tasks") as pgbar: while len(tasks) > 0: for t in pgbar: t.n_answers = redundancy response = config.pbclient.update_task(t) check_api_error(response) # Check if for the data we have to auto-throttle task update sleep, msg = enable_auto_throttling(config, tasks) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks redundancy have been updated" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def find_project_by_short_name(short_name, pbclient, all=None): """Return project by short_name.""" try: response = pbclient.find_project(short_name=short_name, all=all) check_api_error(response) if (len(response) == 0): msg = '%s not found! You can use the all=1 argument to \ search in all the server.' error = 'Project Not Found' raise ProjectNotFound(msg, error) return response[0] except exceptions.ConnectionError: raise except ProjectNotFound: raise def check_api_error(api_response): print(api_response) """Check if returned API response contains an error.""" if type(api_response) == dict and 'code' in api_response and api_response['code'] <> 200: print("Server response code: %s" % api_response['code']) print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) if type(api_response) == dict and (api_response.get('status') == 'failed'): if 'ProgrammingError' in api_response.get('exception_cls'): raise DatabaseError(message='PyBossa database error.', error=api_response) if ('DBIntegrityError' in api_response.get('exception_cls') and 'project' in api_response.get('target')): msg = 'PyBossa project already exists.' raise ProjectAlreadyExists(message=msg, error=api_response) if 'project' in api_response.get('target'): raise ProjectNotFound(message='PyBossa Project not found', error=api_response) if 'task' in api_response.get('target'): raise TaskNotFound(message='PyBossa Task not found', error=api_response) else: print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) def format_error(module, error): """Format the error for the given module.""" logging.error(module) # Beautify JSON error print error.message print json.dumps(error.error, sort_keys=True, indent=4, separators=(',', ': ')) exit(1) def create_task_info(task): """Create task_info field.""" task_info = None if task.get('info'): task_info = task['info'] else: task_info = task return task_info def create_helping_material_info(helping): """Create helping_material_info field.""" helping_info = None file_path = None if helping.get('info'): helping_info = helping['info'] else: helping_info = helping if helping_info.get('file_path'): file_path = helping_info.get('file_path') del helping_info['file_path'] return helping_info, file_path def enable_auto_throttling(config, data, limit=299, endpoint='/api/task'): "Return sleep time if more tasks than those " \ "allowed by the server are requested." # Get header from server endpoint = config.server + endpoint headers = requests.head(endpoint).headers # Get limit server_limit = int(headers.get('X-RateLimit-Remaining', 0)) limit = server_limit or limit # Get reset time reset_epoch = int(headers.get('X-RateLimit-Reset', 0)) # Compute sleep time sleep = (reset_epoch - calendar.timegm(datetime.datetime.utcnow().utctimetuple())) msg = 'Warning: %s remaining hits to the endpoint.' \ ' Auto-throttling enabled!' % limit # If we have less than 10 hits on the endpoint, sleep if limit <= 10: return (sleep, msg) else: return 0, None def format_json_task(task_info): """Format task_info into JSON if applicable.""" try: return json.loads(task_info) except: return task_info def row_empty(row): """Check if all values in row are None.""" for value in row: if value is not None: return False return True class PbsHandler(PatternMatchingEventHandler): patterns = ['*/template.html', '*/tutorial.html', '*/long_description.md', '*/results.html', '*/bundle.js', '*/bundle.min.js'] def __init__(self, config, task_presenter, results, long_description, tutorial): super(PbsHandler, self).__init__() self.config = config self.task_presenter = task_presenter self.results = results self.long_description = long_description self.tutorial = tutorial def on_modified(self, event): what = 'directory' if event.is_directory else 'file' logging.info("Modified %s: %s", what, event.src_path) res = _update_project(self.config, self.task_presenter, self.results, self.long_description, self.tutorial) logging.info(res)
Scifabric/pbs
helpers.py
_delete_tasks
python
def _delete_tasks(config, task_id, limit=100, offset=0): try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.delete_task(task_id) check_api_error(response) return "Task.id = %s and its associated task_runs have been deleted" % task_id else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) while len(tasks) > 0: for t in tasks: response = config.pbclient.delete_task(t.id) check_api_error(response) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks and task_runs have been deleted" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise
Delete tasks from a project.
train
https://github.com/Scifabric/pbs/blob/3e5d5f3f0f5d20f740eaacc4d6e872a0c9fb8b38/helpers.py#L281-L305
null
#!/usr/bin/env python # -*- coding: utf-8 -*- # This file is part of PyBOSSA. # # PyBOSSA is free software: you can redistribute it and/or modify # it under the terms of the GNU Affero General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # PyBOSSA is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU Affero General Public License for more details. # # You should have received a copy of the GNU Affero General Public License # along with PyBOSSA. If not, see <http://www.gnu.org/licenses/>. """ Helper functions for the pbs command line client. This module exports the following methods: * find_project_by_short_name: return the project by short_name. * check_api_errors: check for API errors returned by a PyBossa server. * format_error: format error message. * format_json_task: format a CSV row into JSON. """ import re import os import csv import json import time import click import datetime from StringIO import StringIO import polib import openpyxl import itertools from requests import exceptions import requests from pbsexceptions import * import logging from watchdog.observers import Observer from watchdog.events import PatternMatchingEventHandler import calendar __all__ = ['find_project_by_short_name', 'check_api_error', 'format_error', 'format_json_task', '_create_project', '_update_project', '_add_tasks', 'create_task_info', '_delete_tasks', 'enable_auto_throttling', '_update_tasks_redundancy', '_update_project_watch', 'PbsHandler', '_update_task_presenter_bundle_js', 'row_empty', '_add_helpingmaterials', 'create_helping_material_info'] def _create_project(config): """Create a project in a PyBossa server.""" try: response = config.pbclient.create_project(config.project['name'], config.project['short_name'], config.project['description']) check_api_error(response) return ("Project: %s created!" % config.project['short_name']) except exceptions.ConnectionError: return("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_project_watch(config, task_presenter, results, long_description, tutorial): # pragma: no cover """Update a project in a loop.""" logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(message)s', datefmt='%Y-%m-%d %H:%M:%S') path = os.getcwd() event_handler = PbsHandler(config, task_presenter, results, long_description, tutorial) observer = Observer() # We only want the current folder, not sub-folders observer.schedule(event_handler, path, recursive=False) observer.start() try: while True: time.sleep(1) except KeyboardInterrupt: observer.stop() observer.join() def _update_task_presenter_bundle_js(project): """Append to template a distribution bundle js.""" if os.path.isfile ('bundle.min.js'): with open('bundle.min.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js return if os.path.isfile ('bundle.js'): with open('bundle.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js def _update_project(config, task_presenter, results, long_description, tutorial): """Update a project.""" try: # Get project project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) # Update attributes project.name = config.project['name'] project.short_name = config.project['short_name'] project.description = config.project['description'] # Update long_description with open(long_description, 'r') as f: project.long_description = f.read() # Update task presenter with open(task_presenter, 'r') as f: project.info['task_presenter'] = f.read() _update_task_presenter_bundle_js(project) # Update results with open(results, 'r') as f: project.info['results'] = f.read() # Update tutorial with open(tutorial, 'r') as f: project.info['tutorial'] = f.read() response = config.pbclient.update_project(project) check_api_error(response) return ("Project %s updated!" % config.project['short_name']) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except ProjectNotFound: return ("Project not found! The project: %s is missing." \ " Use the flag --all=1 to search in all the server " \ % config.project['short_name']) except TaskNotFound: raise def _load_data(data_file, data_type): """Load data from CSV, JSON, Excel, ..., formats.""" raw_data = data_file.read() if data_type is None: data_type = data_file.name.split('.')[-1] # Data list to process data = [] # JSON type if data_type == 'json': data = json.loads(raw_data) return data # CSV type elif data_type == 'csv': csv_data = StringIO(raw_data) reader = csv.DictReader(csv_data, delimiter=',') for line in reader: data.append(line) return data elif data_type in ['xlsx', 'xlsm', 'xltx', 'xltm']: excel_data = StringIO(raw_data) wb = openpyxl.load_workbook(excel_data) ws = wb.active # First headers headers = [] for row in ws.iter_rows(max_row=1): for cell in row: tmp = '_'.join(cell.value.split(" ")).lower() headers.append(tmp) # Simulate DictReader for row in ws.iter_rows(row_offset=1): values = [] for cell in row: values.append(cell.value) tmp = dict(itertools.izip(headers, values)) if len(values) == len(headers) and not row_empty(values): data.append(tmp) return data # PO type elif data_type == 'po': po = polib.pofile(raw_data) for entry in po.untranslated_entries(): data.append(entry.__dict__) return data # PROPERTIES type (used in Java and Firefox extensions) elif data_type == 'properties': lines = raw_data.split('\n') for l in lines: if l: var_id, string = l.split('=') tmp = dict(var_id=var_id, string=string) data.append(tmp) return data else: return data def _add_tasks(config, tasks_file, tasks_type, priority, redundancy): """Add tasks to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(tasks_file, tasks_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # If true, warn user # if sleep: # pragma: no cover # click.secho(msg, fg='yellow') # Show progress bar with click.progressbar(data, label="Adding Tasks") as pgbar: for d in pgbar: task_info = create_task_info(d) response = config.pbclient.create_task(project_id=project.id, info=task_info, n_answers=redundancy, priority_0=priority) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data) check_api_error(response) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s tasks added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _add_helpingmaterials(config, helping_file, helping_type): """Add helping materials to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(helping_file, helping_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # Show progress bar with click.progressbar(data, label="Adding Helping Materials") as pgbar: for d in pgbar: helping_info, file_path = create_helping_material_info(d) if file_path: # Create first the media object hm = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info, file_path=file_path) check_api_error(hm) z = hm.info.copy() z.update(helping_info) hm.info = z response = config.pbclient.update_helping_material(hm) check_api_error(response) else: response = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info) check_api_error(response) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data, endpoint='/api/helpinmaterial') # If true, warn user if sleep: # pragma: no cover click.secho(msg, fg='yellow') # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s helping materials added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_tasks_redundancy(config, task_id, redundancy, limit=300, offset=0): """Update tasks redundancy from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.find_tasks(project.id, id=task_id) check_api_error(response) task = response[0] task.n_answers = redundancy response = config.pbclient.update_task(task) check_api_error(response) msg = "Task.id = %s redundancy has been updated to %s" % (task_id, redundancy) return msg else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) with click.progressbar(tasks, label="Updating Tasks") as pgbar: while len(tasks) > 0: for t in pgbar: t.n_answers = redundancy response = config.pbclient.update_task(t) check_api_error(response) # Check if for the data we have to auto-throttle task update sleep, msg = enable_auto_throttling(config, tasks) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks redundancy have been updated" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def find_project_by_short_name(short_name, pbclient, all=None): """Return project by short_name.""" try: response = pbclient.find_project(short_name=short_name, all=all) check_api_error(response) if (len(response) == 0): msg = '%s not found! You can use the all=1 argument to \ search in all the server.' error = 'Project Not Found' raise ProjectNotFound(msg, error) return response[0] except exceptions.ConnectionError: raise except ProjectNotFound: raise def check_api_error(api_response): print(api_response) """Check if returned API response contains an error.""" if type(api_response) == dict and 'code' in api_response and api_response['code'] <> 200: print("Server response code: %s" % api_response['code']) print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) if type(api_response) == dict and (api_response.get('status') == 'failed'): if 'ProgrammingError' in api_response.get('exception_cls'): raise DatabaseError(message='PyBossa database error.', error=api_response) if ('DBIntegrityError' in api_response.get('exception_cls') and 'project' in api_response.get('target')): msg = 'PyBossa project already exists.' raise ProjectAlreadyExists(message=msg, error=api_response) if 'project' in api_response.get('target'): raise ProjectNotFound(message='PyBossa Project not found', error=api_response) if 'task' in api_response.get('target'): raise TaskNotFound(message='PyBossa Task not found', error=api_response) else: print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) def format_error(module, error): """Format the error for the given module.""" logging.error(module) # Beautify JSON error print error.message print json.dumps(error.error, sort_keys=True, indent=4, separators=(',', ': ')) exit(1) def create_task_info(task): """Create task_info field.""" task_info = None if task.get('info'): task_info = task['info'] else: task_info = task return task_info def create_helping_material_info(helping): """Create helping_material_info field.""" helping_info = None file_path = None if helping.get('info'): helping_info = helping['info'] else: helping_info = helping if helping_info.get('file_path'): file_path = helping_info.get('file_path') del helping_info['file_path'] return helping_info, file_path def enable_auto_throttling(config, data, limit=299, endpoint='/api/task'): "Return sleep time if more tasks than those " \ "allowed by the server are requested." # Get header from server endpoint = config.server + endpoint headers = requests.head(endpoint).headers # Get limit server_limit = int(headers.get('X-RateLimit-Remaining', 0)) limit = server_limit or limit # Get reset time reset_epoch = int(headers.get('X-RateLimit-Reset', 0)) # Compute sleep time sleep = (reset_epoch - calendar.timegm(datetime.datetime.utcnow().utctimetuple())) msg = 'Warning: %s remaining hits to the endpoint.' \ ' Auto-throttling enabled!' % limit # If we have less than 10 hits on the endpoint, sleep if limit <= 10: return (sleep, msg) else: return 0, None def format_json_task(task_info): """Format task_info into JSON if applicable.""" try: return json.loads(task_info) except: return task_info def row_empty(row): """Check if all values in row are None.""" for value in row: if value is not None: return False return True class PbsHandler(PatternMatchingEventHandler): patterns = ['*/template.html', '*/tutorial.html', '*/long_description.md', '*/results.html', '*/bundle.js', '*/bundle.min.js'] def __init__(self, config, task_presenter, results, long_description, tutorial): super(PbsHandler, self).__init__() self.config = config self.task_presenter = task_presenter self.results = results self.long_description = long_description self.tutorial = tutorial def on_modified(self, event): what = 'directory' if event.is_directory else 'file' logging.info("Modified %s: %s", what, event.src_path) res = _update_project(self.config, self.task_presenter, self.results, self.long_description, self.tutorial) logging.info(res)
Scifabric/pbs
helpers.py
_update_tasks_redundancy
python
def _update_tasks_redundancy(config, task_id, redundancy, limit=300, offset=0): try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.find_tasks(project.id, id=task_id) check_api_error(response) task = response[0] task.n_answers = redundancy response = config.pbclient.update_task(task) check_api_error(response) msg = "Task.id = %s redundancy has been updated to %s" % (task_id, redundancy) return msg else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) with click.progressbar(tasks, label="Updating Tasks") as pgbar: while len(tasks) > 0: for t in pgbar: t.n_answers = redundancy response = config.pbclient.update_task(t) check_api_error(response) # Check if for the data we have to auto-throttle task update sleep, msg = enable_auto_throttling(config, tasks) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks redundancy have been updated" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise
Update tasks redundancy from a project.
train
https://github.com/Scifabric/pbs/blob/3e5d5f3f0f5d20f740eaacc4d6e872a0c9fb8b38/helpers.py#L308-L345
[ "def find_project_by_short_name(short_name, pbclient, all=None):\n \"\"\"Return project by short_name.\"\"\"\n try:\n response = pbclient.find_project(short_name=short_name, all=all)\n check_api_error(response)\n if (len(response) == 0):\n msg = '%s not found! You can use the all=1 argument to \\\n search in all the server.'\n error = 'Project Not Found'\n raise ProjectNotFound(msg, error)\n return response[0]\n except exceptions.ConnectionError:\n raise\n except ProjectNotFound:\n raise\n", "def check_api_error(api_response):\n print(api_response)\n \"\"\"Check if returned API response contains an error.\"\"\"\n if type(api_response) == dict and 'code' in api_response and api_response['code'] <> 200:\n print(\"Server response code: %s\" % api_response['code'])\n print(\"Server response: %s\" % api_response)\n raise exceptions.HTTPError('Unexpected response', response=api_response)\n if type(api_response) == dict and (api_response.get('status') == 'failed'):\n if 'ProgrammingError' in api_response.get('exception_cls'):\n raise DatabaseError(message='PyBossa database error.',\n error=api_response)\n if ('DBIntegrityError' in api_response.get('exception_cls') and\n 'project' in api_response.get('target')):\n msg = 'PyBossa project already exists.'\n raise ProjectAlreadyExists(message=msg, error=api_response)\n if 'project' in api_response.get('target'):\n raise ProjectNotFound(message='PyBossa Project not found',\n error=api_response)\n if 'task' in api_response.get('target'):\n raise TaskNotFound(message='PyBossa Task not found',\n error=api_response)\n else:\n print(\"Server response: %s\" % api_response)\n raise exceptions.HTTPError('Unexpected response', response=api_response)\n", "def enable_auto_throttling(config, data, limit=299, endpoint='/api/task'):\n \"Return sleep time if more tasks than those \" \\\n \"allowed by the server are requested.\"\n # Get header from server\n endpoint = config.server + endpoint\n headers = requests.head(endpoint).headers\n # Get limit\n server_limit = int(headers.get('X-RateLimit-Remaining', 0))\n limit = server_limit or limit\n # Get reset time\n reset_epoch = int(headers.get('X-RateLimit-Reset', 0))\n # Compute sleep time\n sleep = (reset_epoch -\n calendar.timegm(datetime.datetime.utcnow().utctimetuple()))\n msg = 'Warning: %s remaining hits to the endpoint.' \\\n ' Auto-throttling enabled!' % limit\n # If we have less than 10 hits on the endpoint, sleep\n if limit <= 10:\n return (sleep, msg)\n else:\n return 0, None\n" ]
#!/usr/bin/env python # -*- coding: utf-8 -*- # This file is part of PyBOSSA. # # PyBOSSA is free software: you can redistribute it and/or modify # it under the terms of the GNU Affero General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # PyBOSSA is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU Affero General Public License for more details. # # You should have received a copy of the GNU Affero General Public License # along with PyBOSSA. If not, see <http://www.gnu.org/licenses/>. """ Helper functions for the pbs command line client. This module exports the following methods: * find_project_by_short_name: return the project by short_name. * check_api_errors: check for API errors returned by a PyBossa server. * format_error: format error message. * format_json_task: format a CSV row into JSON. """ import re import os import csv import json import time import click import datetime from StringIO import StringIO import polib import openpyxl import itertools from requests import exceptions import requests from pbsexceptions import * import logging from watchdog.observers import Observer from watchdog.events import PatternMatchingEventHandler import calendar __all__ = ['find_project_by_short_name', 'check_api_error', 'format_error', 'format_json_task', '_create_project', '_update_project', '_add_tasks', 'create_task_info', '_delete_tasks', 'enable_auto_throttling', '_update_tasks_redundancy', '_update_project_watch', 'PbsHandler', '_update_task_presenter_bundle_js', 'row_empty', '_add_helpingmaterials', 'create_helping_material_info'] def _create_project(config): """Create a project in a PyBossa server.""" try: response = config.pbclient.create_project(config.project['name'], config.project['short_name'], config.project['description']) check_api_error(response) return ("Project: %s created!" % config.project['short_name']) except exceptions.ConnectionError: return("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_project_watch(config, task_presenter, results, long_description, tutorial): # pragma: no cover """Update a project in a loop.""" logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(message)s', datefmt='%Y-%m-%d %H:%M:%S') path = os.getcwd() event_handler = PbsHandler(config, task_presenter, results, long_description, tutorial) observer = Observer() # We only want the current folder, not sub-folders observer.schedule(event_handler, path, recursive=False) observer.start() try: while True: time.sleep(1) except KeyboardInterrupt: observer.stop() observer.join() def _update_task_presenter_bundle_js(project): """Append to template a distribution bundle js.""" if os.path.isfile ('bundle.min.js'): with open('bundle.min.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js return if os.path.isfile ('bundle.js'): with open('bundle.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js def _update_project(config, task_presenter, results, long_description, tutorial): """Update a project.""" try: # Get project project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) # Update attributes project.name = config.project['name'] project.short_name = config.project['short_name'] project.description = config.project['description'] # Update long_description with open(long_description, 'r') as f: project.long_description = f.read() # Update task presenter with open(task_presenter, 'r') as f: project.info['task_presenter'] = f.read() _update_task_presenter_bundle_js(project) # Update results with open(results, 'r') as f: project.info['results'] = f.read() # Update tutorial with open(tutorial, 'r') as f: project.info['tutorial'] = f.read() response = config.pbclient.update_project(project) check_api_error(response) return ("Project %s updated!" % config.project['short_name']) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except ProjectNotFound: return ("Project not found! The project: %s is missing." \ " Use the flag --all=1 to search in all the server " \ % config.project['short_name']) except TaskNotFound: raise def _load_data(data_file, data_type): """Load data from CSV, JSON, Excel, ..., formats.""" raw_data = data_file.read() if data_type is None: data_type = data_file.name.split('.')[-1] # Data list to process data = [] # JSON type if data_type == 'json': data = json.loads(raw_data) return data # CSV type elif data_type == 'csv': csv_data = StringIO(raw_data) reader = csv.DictReader(csv_data, delimiter=',') for line in reader: data.append(line) return data elif data_type in ['xlsx', 'xlsm', 'xltx', 'xltm']: excel_data = StringIO(raw_data) wb = openpyxl.load_workbook(excel_data) ws = wb.active # First headers headers = [] for row in ws.iter_rows(max_row=1): for cell in row: tmp = '_'.join(cell.value.split(" ")).lower() headers.append(tmp) # Simulate DictReader for row in ws.iter_rows(row_offset=1): values = [] for cell in row: values.append(cell.value) tmp = dict(itertools.izip(headers, values)) if len(values) == len(headers) and not row_empty(values): data.append(tmp) return data # PO type elif data_type == 'po': po = polib.pofile(raw_data) for entry in po.untranslated_entries(): data.append(entry.__dict__) return data # PROPERTIES type (used in Java and Firefox extensions) elif data_type == 'properties': lines = raw_data.split('\n') for l in lines: if l: var_id, string = l.split('=') tmp = dict(var_id=var_id, string=string) data.append(tmp) return data else: return data def _add_tasks(config, tasks_file, tasks_type, priority, redundancy): """Add tasks to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(tasks_file, tasks_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # If true, warn user # if sleep: # pragma: no cover # click.secho(msg, fg='yellow') # Show progress bar with click.progressbar(data, label="Adding Tasks") as pgbar: for d in pgbar: task_info = create_task_info(d) response = config.pbclient.create_task(project_id=project.id, info=task_info, n_answers=redundancy, priority_0=priority) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data) check_api_error(response) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s tasks added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _add_helpingmaterials(config, helping_file, helping_type): """Add helping materials to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(helping_file, helping_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # Show progress bar with click.progressbar(data, label="Adding Helping Materials") as pgbar: for d in pgbar: helping_info, file_path = create_helping_material_info(d) if file_path: # Create first the media object hm = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info, file_path=file_path) check_api_error(hm) z = hm.info.copy() z.update(helping_info) hm.info = z response = config.pbclient.update_helping_material(hm) check_api_error(response) else: response = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info) check_api_error(response) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data, endpoint='/api/helpinmaterial') # If true, warn user if sleep: # pragma: no cover click.secho(msg, fg='yellow') # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s helping materials added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _delete_tasks(config, task_id, limit=100, offset=0): """Delete tasks from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.delete_task(task_id) check_api_error(response) return "Task.id = %s and its associated task_runs have been deleted" % task_id else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) while len(tasks) > 0: for t in tasks: response = config.pbclient.delete_task(t.id) check_api_error(response) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks and task_runs have been deleted" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def find_project_by_short_name(short_name, pbclient, all=None): """Return project by short_name.""" try: response = pbclient.find_project(short_name=short_name, all=all) check_api_error(response) if (len(response) == 0): msg = '%s not found! You can use the all=1 argument to \ search in all the server.' error = 'Project Not Found' raise ProjectNotFound(msg, error) return response[0] except exceptions.ConnectionError: raise except ProjectNotFound: raise def check_api_error(api_response): print(api_response) """Check if returned API response contains an error.""" if type(api_response) == dict and 'code' in api_response and api_response['code'] <> 200: print("Server response code: %s" % api_response['code']) print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) if type(api_response) == dict and (api_response.get('status') == 'failed'): if 'ProgrammingError' in api_response.get('exception_cls'): raise DatabaseError(message='PyBossa database error.', error=api_response) if ('DBIntegrityError' in api_response.get('exception_cls') and 'project' in api_response.get('target')): msg = 'PyBossa project already exists.' raise ProjectAlreadyExists(message=msg, error=api_response) if 'project' in api_response.get('target'): raise ProjectNotFound(message='PyBossa Project not found', error=api_response) if 'task' in api_response.get('target'): raise TaskNotFound(message='PyBossa Task not found', error=api_response) else: print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) def format_error(module, error): """Format the error for the given module.""" logging.error(module) # Beautify JSON error print error.message print json.dumps(error.error, sort_keys=True, indent=4, separators=(',', ': ')) exit(1) def create_task_info(task): """Create task_info field.""" task_info = None if task.get('info'): task_info = task['info'] else: task_info = task return task_info def create_helping_material_info(helping): """Create helping_material_info field.""" helping_info = None file_path = None if helping.get('info'): helping_info = helping['info'] else: helping_info = helping if helping_info.get('file_path'): file_path = helping_info.get('file_path') del helping_info['file_path'] return helping_info, file_path def enable_auto_throttling(config, data, limit=299, endpoint='/api/task'): "Return sleep time if more tasks than those " \ "allowed by the server are requested." # Get header from server endpoint = config.server + endpoint headers = requests.head(endpoint).headers # Get limit server_limit = int(headers.get('X-RateLimit-Remaining', 0)) limit = server_limit or limit # Get reset time reset_epoch = int(headers.get('X-RateLimit-Reset', 0)) # Compute sleep time sleep = (reset_epoch - calendar.timegm(datetime.datetime.utcnow().utctimetuple())) msg = 'Warning: %s remaining hits to the endpoint.' \ ' Auto-throttling enabled!' % limit # If we have less than 10 hits on the endpoint, sleep if limit <= 10: return (sleep, msg) else: return 0, None def format_json_task(task_info): """Format task_info into JSON if applicable.""" try: return json.loads(task_info) except: return task_info def row_empty(row): """Check if all values in row are None.""" for value in row: if value is not None: return False return True class PbsHandler(PatternMatchingEventHandler): patterns = ['*/template.html', '*/tutorial.html', '*/long_description.md', '*/results.html', '*/bundle.js', '*/bundle.min.js'] def __init__(self, config, task_presenter, results, long_description, tutorial): super(PbsHandler, self).__init__() self.config = config self.task_presenter = task_presenter self.results = results self.long_description = long_description self.tutorial = tutorial def on_modified(self, event): what = 'directory' if event.is_directory else 'file' logging.info("Modified %s: %s", what, event.src_path) res = _update_project(self.config, self.task_presenter, self.results, self.long_description, self.tutorial) logging.info(res)
Scifabric/pbs
helpers.py
find_project_by_short_name
python
def find_project_by_short_name(short_name, pbclient, all=None): try: response = pbclient.find_project(short_name=short_name, all=all) check_api_error(response) if (len(response) == 0): msg = '%s not found! You can use the all=1 argument to \ search in all the server.' error = 'Project Not Found' raise ProjectNotFound(msg, error) return response[0] except exceptions.ConnectionError: raise except ProjectNotFound: raise
Return project by short_name.
train
https://github.com/Scifabric/pbs/blob/3e5d5f3f0f5d20f740eaacc4d6e872a0c9fb8b38/helpers.py#L348-L362
[ "def check_api_error(api_response):\n print(api_response)\n \"\"\"Check if returned API response contains an error.\"\"\"\n if type(api_response) == dict and 'code' in api_response and api_response['code'] <> 200:\n print(\"Server response code: %s\" % api_response['code'])\n print(\"Server response: %s\" % api_response)\n raise exceptions.HTTPError('Unexpected response', response=api_response)\n if type(api_response) == dict and (api_response.get('status') == 'failed'):\n if 'ProgrammingError' in api_response.get('exception_cls'):\n raise DatabaseError(message='PyBossa database error.',\n error=api_response)\n if ('DBIntegrityError' in api_response.get('exception_cls') and\n 'project' in api_response.get('target')):\n msg = 'PyBossa project already exists.'\n raise ProjectAlreadyExists(message=msg, error=api_response)\n if 'project' in api_response.get('target'):\n raise ProjectNotFound(message='PyBossa Project not found',\n error=api_response)\n if 'task' in api_response.get('target'):\n raise TaskNotFound(message='PyBossa Task not found',\n error=api_response)\n else:\n print(\"Server response: %s\" % api_response)\n raise exceptions.HTTPError('Unexpected response', response=api_response)\n" ]
#!/usr/bin/env python # -*- coding: utf-8 -*- # This file is part of PyBOSSA. # # PyBOSSA is free software: you can redistribute it and/or modify # it under the terms of the GNU Affero General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # PyBOSSA is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU Affero General Public License for more details. # # You should have received a copy of the GNU Affero General Public License # along with PyBOSSA. If not, see <http://www.gnu.org/licenses/>. """ Helper functions for the pbs command line client. This module exports the following methods: * find_project_by_short_name: return the project by short_name. * check_api_errors: check for API errors returned by a PyBossa server. * format_error: format error message. * format_json_task: format a CSV row into JSON. """ import re import os import csv import json import time import click import datetime from StringIO import StringIO import polib import openpyxl import itertools from requests import exceptions import requests from pbsexceptions import * import logging from watchdog.observers import Observer from watchdog.events import PatternMatchingEventHandler import calendar __all__ = ['find_project_by_short_name', 'check_api_error', 'format_error', 'format_json_task', '_create_project', '_update_project', '_add_tasks', 'create_task_info', '_delete_tasks', 'enable_auto_throttling', '_update_tasks_redundancy', '_update_project_watch', 'PbsHandler', '_update_task_presenter_bundle_js', 'row_empty', '_add_helpingmaterials', 'create_helping_material_info'] def _create_project(config): """Create a project in a PyBossa server.""" try: response = config.pbclient.create_project(config.project['name'], config.project['short_name'], config.project['description']) check_api_error(response) return ("Project: %s created!" % config.project['short_name']) except exceptions.ConnectionError: return("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_project_watch(config, task_presenter, results, long_description, tutorial): # pragma: no cover """Update a project in a loop.""" logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(message)s', datefmt='%Y-%m-%d %H:%M:%S') path = os.getcwd() event_handler = PbsHandler(config, task_presenter, results, long_description, tutorial) observer = Observer() # We only want the current folder, not sub-folders observer.schedule(event_handler, path, recursive=False) observer.start() try: while True: time.sleep(1) except KeyboardInterrupt: observer.stop() observer.join() def _update_task_presenter_bundle_js(project): """Append to template a distribution bundle js.""" if os.path.isfile ('bundle.min.js'): with open('bundle.min.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js return if os.path.isfile ('bundle.js'): with open('bundle.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js def _update_project(config, task_presenter, results, long_description, tutorial): """Update a project.""" try: # Get project project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) # Update attributes project.name = config.project['name'] project.short_name = config.project['short_name'] project.description = config.project['description'] # Update long_description with open(long_description, 'r') as f: project.long_description = f.read() # Update task presenter with open(task_presenter, 'r') as f: project.info['task_presenter'] = f.read() _update_task_presenter_bundle_js(project) # Update results with open(results, 'r') as f: project.info['results'] = f.read() # Update tutorial with open(tutorial, 'r') as f: project.info['tutorial'] = f.read() response = config.pbclient.update_project(project) check_api_error(response) return ("Project %s updated!" % config.project['short_name']) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except ProjectNotFound: return ("Project not found! The project: %s is missing." \ " Use the flag --all=1 to search in all the server " \ % config.project['short_name']) except TaskNotFound: raise def _load_data(data_file, data_type): """Load data from CSV, JSON, Excel, ..., formats.""" raw_data = data_file.read() if data_type is None: data_type = data_file.name.split('.')[-1] # Data list to process data = [] # JSON type if data_type == 'json': data = json.loads(raw_data) return data # CSV type elif data_type == 'csv': csv_data = StringIO(raw_data) reader = csv.DictReader(csv_data, delimiter=',') for line in reader: data.append(line) return data elif data_type in ['xlsx', 'xlsm', 'xltx', 'xltm']: excel_data = StringIO(raw_data) wb = openpyxl.load_workbook(excel_data) ws = wb.active # First headers headers = [] for row in ws.iter_rows(max_row=1): for cell in row: tmp = '_'.join(cell.value.split(" ")).lower() headers.append(tmp) # Simulate DictReader for row in ws.iter_rows(row_offset=1): values = [] for cell in row: values.append(cell.value) tmp = dict(itertools.izip(headers, values)) if len(values) == len(headers) and not row_empty(values): data.append(tmp) return data # PO type elif data_type == 'po': po = polib.pofile(raw_data) for entry in po.untranslated_entries(): data.append(entry.__dict__) return data # PROPERTIES type (used in Java and Firefox extensions) elif data_type == 'properties': lines = raw_data.split('\n') for l in lines: if l: var_id, string = l.split('=') tmp = dict(var_id=var_id, string=string) data.append(tmp) return data else: return data def _add_tasks(config, tasks_file, tasks_type, priority, redundancy): """Add tasks to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(tasks_file, tasks_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # If true, warn user # if sleep: # pragma: no cover # click.secho(msg, fg='yellow') # Show progress bar with click.progressbar(data, label="Adding Tasks") as pgbar: for d in pgbar: task_info = create_task_info(d) response = config.pbclient.create_task(project_id=project.id, info=task_info, n_answers=redundancy, priority_0=priority) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data) check_api_error(response) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s tasks added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _add_helpingmaterials(config, helping_file, helping_type): """Add helping materials to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(helping_file, helping_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # Show progress bar with click.progressbar(data, label="Adding Helping Materials") as pgbar: for d in pgbar: helping_info, file_path = create_helping_material_info(d) if file_path: # Create first the media object hm = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info, file_path=file_path) check_api_error(hm) z = hm.info.copy() z.update(helping_info) hm.info = z response = config.pbclient.update_helping_material(hm) check_api_error(response) else: response = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info) check_api_error(response) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data, endpoint='/api/helpinmaterial') # If true, warn user if sleep: # pragma: no cover click.secho(msg, fg='yellow') # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s helping materials added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _delete_tasks(config, task_id, limit=100, offset=0): """Delete tasks from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.delete_task(task_id) check_api_error(response) return "Task.id = %s and its associated task_runs have been deleted" % task_id else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) while len(tasks) > 0: for t in tasks: response = config.pbclient.delete_task(t.id) check_api_error(response) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks and task_runs have been deleted" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_tasks_redundancy(config, task_id, redundancy, limit=300, offset=0): """Update tasks redundancy from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.find_tasks(project.id, id=task_id) check_api_error(response) task = response[0] task.n_answers = redundancy response = config.pbclient.update_task(task) check_api_error(response) msg = "Task.id = %s redundancy has been updated to %s" % (task_id, redundancy) return msg else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) with click.progressbar(tasks, label="Updating Tasks") as pgbar: while len(tasks) > 0: for t in pgbar: t.n_answers = redundancy response = config.pbclient.update_task(t) check_api_error(response) # Check if for the data we have to auto-throttle task update sleep, msg = enable_auto_throttling(config, tasks) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks redundancy have been updated" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def check_api_error(api_response): print(api_response) """Check if returned API response contains an error.""" if type(api_response) == dict and 'code' in api_response and api_response['code'] <> 200: print("Server response code: %s" % api_response['code']) print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) if type(api_response) == dict and (api_response.get('status') == 'failed'): if 'ProgrammingError' in api_response.get('exception_cls'): raise DatabaseError(message='PyBossa database error.', error=api_response) if ('DBIntegrityError' in api_response.get('exception_cls') and 'project' in api_response.get('target')): msg = 'PyBossa project already exists.' raise ProjectAlreadyExists(message=msg, error=api_response) if 'project' in api_response.get('target'): raise ProjectNotFound(message='PyBossa Project not found', error=api_response) if 'task' in api_response.get('target'): raise TaskNotFound(message='PyBossa Task not found', error=api_response) else: print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) def format_error(module, error): """Format the error for the given module.""" logging.error(module) # Beautify JSON error print error.message print json.dumps(error.error, sort_keys=True, indent=4, separators=(',', ': ')) exit(1) def create_task_info(task): """Create task_info field.""" task_info = None if task.get('info'): task_info = task['info'] else: task_info = task return task_info def create_helping_material_info(helping): """Create helping_material_info field.""" helping_info = None file_path = None if helping.get('info'): helping_info = helping['info'] else: helping_info = helping if helping_info.get('file_path'): file_path = helping_info.get('file_path') del helping_info['file_path'] return helping_info, file_path def enable_auto_throttling(config, data, limit=299, endpoint='/api/task'): "Return sleep time if more tasks than those " \ "allowed by the server are requested." # Get header from server endpoint = config.server + endpoint headers = requests.head(endpoint).headers # Get limit server_limit = int(headers.get('X-RateLimit-Remaining', 0)) limit = server_limit or limit # Get reset time reset_epoch = int(headers.get('X-RateLimit-Reset', 0)) # Compute sleep time sleep = (reset_epoch - calendar.timegm(datetime.datetime.utcnow().utctimetuple())) msg = 'Warning: %s remaining hits to the endpoint.' \ ' Auto-throttling enabled!' % limit # If we have less than 10 hits on the endpoint, sleep if limit <= 10: return (sleep, msg) else: return 0, None def format_json_task(task_info): """Format task_info into JSON if applicable.""" try: return json.loads(task_info) except: return task_info def row_empty(row): """Check if all values in row are None.""" for value in row: if value is not None: return False return True class PbsHandler(PatternMatchingEventHandler): patterns = ['*/template.html', '*/tutorial.html', '*/long_description.md', '*/results.html', '*/bundle.js', '*/bundle.min.js'] def __init__(self, config, task_presenter, results, long_description, tutorial): super(PbsHandler, self).__init__() self.config = config self.task_presenter = task_presenter self.results = results self.long_description = long_description self.tutorial = tutorial def on_modified(self, event): what = 'directory' if event.is_directory else 'file' logging.info("Modified %s: %s", what, event.src_path) res = _update_project(self.config, self.task_presenter, self.results, self.long_description, self.tutorial) logging.info(res)
Scifabric/pbs
helpers.py
check_api_error
python
def check_api_error(api_response): print(api_response) if type(api_response) == dict and 'code' in api_response and api_response['code'] <> 200: print("Server response code: %s" % api_response['code']) print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) if type(api_response) == dict and (api_response.get('status') == 'failed'): if 'ProgrammingError' in api_response.get('exception_cls'): raise DatabaseError(message='PyBossa database error.', error=api_response) if ('DBIntegrityError' in api_response.get('exception_cls') and 'project' in api_response.get('target')): msg = 'PyBossa project already exists.' raise ProjectAlreadyExists(message=msg, error=api_response) if 'project' in api_response.get('target'): raise ProjectNotFound(message='PyBossa Project not found', error=api_response) if 'task' in api_response.get('target'): raise TaskNotFound(message='PyBossa Task not found', error=api_response) else: print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response)
Check if returned API response contains an error.
train
https://github.com/Scifabric/pbs/blob/3e5d5f3f0f5d20f740eaacc4d6e872a0c9fb8b38/helpers.py#L365-L388
null
#!/usr/bin/env python # -*- coding: utf-8 -*- # This file is part of PyBOSSA. # # PyBOSSA is free software: you can redistribute it and/or modify # it under the terms of the GNU Affero General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # PyBOSSA is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU Affero General Public License for more details. # # You should have received a copy of the GNU Affero General Public License # along with PyBOSSA. If not, see <http://www.gnu.org/licenses/>. """ Helper functions for the pbs command line client. This module exports the following methods: * find_project_by_short_name: return the project by short_name. * check_api_errors: check for API errors returned by a PyBossa server. * format_error: format error message. * format_json_task: format a CSV row into JSON. """ import re import os import csv import json import time import click import datetime from StringIO import StringIO import polib import openpyxl import itertools from requests import exceptions import requests from pbsexceptions import * import logging from watchdog.observers import Observer from watchdog.events import PatternMatchingEventHandler import calendar __all__ = ['find_project_by_short_name', 'check_api_error', 'format_error', 'format_json_task', '_create_project', '_update_project', '_add_tasks', 'create_task_info', '_delete_tasks', 'enable_auto_throttling', '_update_tasks_redundancy', '_update_project_watch', 'PbsHandler', '_update_task_presenter_bundle_js', 'row_empty', '_add_helpingmaterials', 'create_helping_material_info'] def _create_project(config): """Create a project in a PyBossa server.""" try: response = config.pbclient.create_project(config.project['name'], config.project['short_name'], config.project['description']) check_api_error(response) return ("Project: %s created!" % config.project['short_name']) except exceptions.ConnectionError: return("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_project_watch(config, task_presenter, results, long_description, tutorial): # pragma: no cover """Update a project in a loop.""" logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(message)s', datefmt='%Y-%m-%d %H:%M:%S') path = os.getcwd() event_handler = PbsHandler(config, task_presenter, results, long_description, tutorial) observer = Observer() # We only want the current folder, not sub-folders observer.schedule(event_handler, path, recursive=False) observer.start() try: while True: time.sleep(1) except KeyboardInterrupt: observer.stop() observer.join() def _update_task_presenter_bundle_js(project): """Append to template a distribution bundle js.""" if os.path.isfile ('bundle.min.js'): with open('bundle.min.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js return if os.path.isfile ('bundle.js'): with open('bundle.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js def _update_project(config, task_presenter, results, long_description, tutorial): """Update a project.""" try: # Get project project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) # Update attributes project.name = config.project['name'] project.short_name = config.project['short_name'] project.description = config.project['description'] # Update long_description with open(long_description, 'r') as f: project.long_description = f.read() # Update task presenter with open(task_presenter, 'r') as f: project.info['task_presenter'] = f.read() _update_task_presenter_bundle_js(project) # Update results with open(results, 'r') as f: project.info['results'] = f.read() # Update tutorial with open(tutorial, 'r') as f: project.info['tutorial'] = f.read() response = config.pbclient.update_project(project) check_api_error(response) return ("Project %s updated!" % config.project['short_name']) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except ProjectNotFound: return ("Project not found! The project: %s is missing." \ " Use the flag --all=1 to search in all the server " \ % config.project['short_name']) except TaskNotFound: raise def _load_data(data_file, data_type): """Load data from CSV, JSON, Excel, ..., formats.""" raw_data = data_file.read() if data_type is None: data_type = data_file.name.split('.')[-1] # Data list to process data = [] # JSON type if data_type == 'json': data = json.loads(raw_data) return data # CSV type elif data_type == 'csv': csv_data = StringIO(raw_data) reader = csv.DictReader(csv_data, delimiter=',') for line in reader: data.append(line) return data elif data_type in ['xlsx', 'xlsm', 'xltx', 'xltm']: excel_data = StringIO(raw_data) wb = openpyxl.load_workbook(excel_data) ws = wb.active # First headers headers = [] for row in ws.iter_rows(max_row=1): for cell in row: tmp = '_'.join(cell.value.split(" ")).lower() headers.append(tmp) # Simulate DictReader for row in ws.iter_rows(row_offset=1): values = [] for cell in row: values.append(cell.value) tmp = dict(itertools.izip(headers, values)) if len(values) == len(headers) and not row_empty(values): data.append(tmp) return data # PO type elif data_type == 'po': po = polib.pofile(raw_data) for entry in po.untranslated_entries(): data.append(entry.__dict__) return data # PROPERTIES type (used in Java and Firefox extensions) elif data_type == 'properties': lines = raw_data.split('\n') for l in lines: if l: var_id, string = l.split('=') tmp = dict(var_id=var_id, string=string) data.append(tmp) return data else: return data def _add_tasks(config, tasks_file, tasks_type, priority, redundancy): """Add tasks to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(tasks_file, tasks_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # If true, warn user # if sleep: # pragma: no cover # click.secho(msg, fg='yellow') # Show progress bar with click.progressbar(data, label="Adding Tasks") as pgbar: for d in pgbar: task_info = create_task_info(d) response = config.pbclient.create_task(project_id=project.id, info=task_info, n_answers=redundancy, priority_0=priority) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data) check_api_error(response) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s tasks added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _add_helpingmaterials(config, helping_file, helping_type): """Add helping materials to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(helping_file, helping_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # Show progress bar with click.progressbar(data, label="Adding Helping Materials") as pgbar: for d in pgbar: helping_info, file_path = create_helping_material_info(d) if file_path: # Create first the media object hm = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info, file_path=file_path) check_api_error(hm) z = hm.info.copy() z.update(helping_info) hm.info = z response = config.pbclient.update_helping_material(hm) check_api_error(response) else: response = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info) check_api_error(response) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data, endpoint='/api/helpinmaterial') # If true, warn user if sleep: # pragma: no cover click.secho(msg, fg='yellow') # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s helping materials added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _delete_tasks(config, task_id, limit=100, offset=0): """Delete tasks from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.delete_task(task_id) check_api_error(response) return "Task.id = %s and its associated task_runs have been deleted" % task_id else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) while len(tasks) > 0: for t in tasks: response = config.pbclient.delete_task(t.id) check_api_error(response) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks and task_runs have been deleted" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_tasks_redundancy(config, task_id, redundancy, limit=300, offset=0): """Update tasks redundancy from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.find_tasks(project.id, id=task_id) check_api_error(response) task = response[0] task.n_answers = redundancy response = config.pbclient.update_task(task) check_api_error(response) msg = "Task.id = %s redundancy has been updated to %s" % (task_id, redundancy) return msg else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) with click.progressbar(tasks, label="Updating Tasks") as pgbar: while len(tasks) > 0: for t in pgbar: t.n_answers = redundancy response = config.pbclient.update_task(t) check_api_error(response) # Check if for the data we have to auto-throttle task update sleep, msg = enable_auto_throttling(config, tasks) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks redundancy have been updated" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def find_project_by_short_name(short_name, pbclient, all=None): """Return project by short_name.""" try: response = pbclient.find_project(short_name=short_name, all=all) check_api_error(response) if (len(response) == 0): msg = '%s not found! You can use the all=1 argument to \ search in all the server.' error = 'Project Not Found' raise ProjectNotFound(msg, error) return response[0] except exceptions.ConnectionError: raise except ProjectNotFound: raise def format_error(module, error): """Format the error for the given module.""" logging.error(module) # Beautify JSON error print error.message print json.dumps(error.error, sort_keys=True, indent=4, separators=(',', ': ')) exit(1) def create_task_info(task): """Create task_info field.""" task_info = None if task.get('info'): task_info = task['info'] else: task_info = task return task_info def create_helping_material_info(helping): """Create helping_material_info field.""" helping_info = None file_path = None if helping.get('info'): helping_info = helping['info'] else: helping_info = helping if helping_info.get('file_path'): file_path = helping_info.get('file_path') del helping_info['file_path'] return helping_info, file_path def enable_auto_throttling(config, data, limit=299, endpoint='/api/task'): "Return sleep time if more tasks than those " \ "allowed by the server are requested." # Get header from server endpoint = config.server + endpoint headers = requests.head(endpoint).headers # Get limit server_limit = int(headers.get('X-RateLimit-Remaining', 0)) limit = server_limit or limit # Get reset time reset_epoch = int(headers.get('X-RateLimit-Reset', 0)) # Compute sleep time sleep = (reset_epoch - calendar.timegm(datetime.datetime.utcnow().utctimetuple())) msg = 'Warning: %s remaining hits to the endpoint.' \ ' Auto-throttling enabled!' % limit # If we have less than 10 hits on the endpoint, sleep if limit <= 10: return (sleep, msg) else: return 0, None def format_json_task(task_info): """Format task_info into JSON if applicable.""" try: return json.loads(task_info) except: return task_info def row_empty(row): """Check if all values in row are None.""" for value in row: if value is not None: return False return True class PbsHandler(PatternMatchingEventHandler): patterns = ['*/template.html', '*/tutorial.html', '*/long_description.md', '*/results.html', '*/bundle.js', '*/bundle.min.js'] def __init__(self, config, task_presenter, results, long_description, tutorial): super(PbsHandler, self).__init__() self.config = config self.task_presenter = task_presenter self.results = results self.long_description = long_description self.tutorial = tutorial def on_modified(self, event): what = 'directory' if event.is_directory else 'file' logging.info("Modified %s: %s", what, event.src_path) res = _update_project(self.config, self.task_presenter, self.results, self.long_description, self.tutorial) logging.info(res)
Scifabric/pbs
helpers.py
format_error
python
def format_error(module, error): logging.error(module) # Beautify JSON error print error.message print json.dumps(error.error, sort_keys=True, indent=4, separators=(',', ': ')) exit(1)
Format the error for the given module.
train
https://github.com/Scifabric/pbs/blob/3e5d5f3f0f5d20f740eaacc4d6e872a0c9fb8b38/helpers.py#L391-L397
null
#!/usr/bin/env python # -*- coding: utf-8 -*- # This file is part of PyBOSSA. # # PyBOSSA is free software: you can redistribute it and/or modify # it under the terms of the GNU Affero General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # PyBOSSA is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU Affero General Public License for more details. # # You should have received a copy of the GNU Affero General Public License # along with PyBOSSA. If not, see <http://www.gnu.org/licenses/>. """ Helper functions for the pbs command line client. This module exports the following methods: * find_project_by_short_name: return the project by short_name. * check_api_errors: check for API errors returned by a PyBossa server. * format_error: format error message. * format_json_task: format a CSV row into JSON. """ import re import os import csv import json import time import click import datetime from StringIO import StringIO import polib import openpyxl import itertools from requests import exceptions import requests from pbsexceptions import * import logging from watchdog.observers import Observer from watchdog.events import PatternMatchingEventHandler import calendar __all__ = ['find_project_by_short_name', 'check_api_error', 'format_error', 'format_json_task', '_create_project', '_update_project', '_add_tasks', 'create_task_info', '_delete_tasks', 'enable_auto_throttling', '_update_tasks_redundancy', '_update_project_watch', 'PbsHandler', '_update_task_presenter_bundle_js', 'row_empty', '_add_helpingmaterials', 'create_helping_material_info'] def _create_project(config): """Create a project in a PyBossa server.""" try: response = config.pbclient.create_project(config.project['name'], config.project['short_name'], config.project['description']) check_api_error(response) return ("Project: %s created!" % config.project['short_name']) except exceptions.ConnectionError: return("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_project_watch(config, task_presenter, results, long_description, tutorial): # pragma: no cover """Update a project in a loop.""" logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(message)s', datefmt='%Y-%m-%d %H:%M:%S') path = os.getcwd() event_handler = PbsHandler(config, task_presenter, results, long_description, tutorial) observer = Observer() # We only want the current folder, not sub-folders observer.schedule(event_handler, path, recursive=False) observer.start() try: while True: time.sleep(1) except KeyboardInterrupt: observer.stop() observer.join() def _update_task_presenter_bundle_js(project): """Append to template a distribution bundle js.""" if os.path.isfile ('bundle.min.js'): with open('bundle.min.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js return if os.path.isfile ('bundle.js'): with open('bundle.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js def _update_project(config, task_presenter, results, long_description, tutorial): """Update a project.""" try: # Get project project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) # Update attributes project.name = config.project['name'] project.short_name = config.project['short_name'] project.description = config.project['description'] # Update long_description with open(long_description, 'r') as f: project.long_description = f.read() # Update task presenter with open(task_presenter, 'r') as f: project.info['task_presenter'] = f.read() _update_task_presenter_bundle_js(project) # Update results with open(results, 'r') as f: project.info['results'] = f.read() # Update tutorial with open(tutorial, 'r') as f: project.info['tutorial'] = f.read() response = config.pbclient.update_project(project) check_api_error(response) return ("Project %s updated!" % config.project['short_name']) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except ProjectNotFound: return ("Project not found! The project: %s is missing." \ " Use the flag --all=1 to search in all the server " \ % config.project['short_name']) except TaskNotFound: raise def _load_data(data_file, data_type): """Load data from CSV, JSON, Excel, ..., formats.""" raw_data = data_file.read() if data_type is None: data_type = data_file.name.split('.')[-1] # Data list to process data = [] # JSON type if data_type == 'json': data = json.loads(raw_data) return data # CSV type elif data_type == 'csv': csv_data = StringIO(raw_data) reader = csv.DictReader(csv_data, delimiter=',') for line in reader: data.append(line) return data elif data_type in ['xlsx', 'xlsm', 'xltx', 'xltm']: excel_data = StringIO(raw_data) wb = openpyxl.load_workbook(excel_data) ws = wb.active # First headers headers = [] for row in ws.iter_rows(max_row=1): for cell in row: tmp = '_'.join(cell.value.split(" ")).lower() headers.append(tmp) # Simulate DictReader for row in ws.iter_rows(row_offset=1): values = [] for cell in row: values.append(cell.value) tmp = dict(itertools.izip(headers, values)) if len(values) == len(headers) and not row_empty(values): data.append(tmp) return data # PO type elif data_type == 'po': po = polib.pofile(raw_data) for entry in po.untranslated_entries(): data.append(entry.__dict__) return data # PROPERTIES type (used in Java and Firefox extensions) elif data_type == 'properties': lines = raw_data.split('\n') for l in lines: if l: var_id, string = l.split('=') tmp = dict(var_id=var_id, string=string) data.append(tmp) return data else: return data def _add_tasks(config, tasks_file, tasks_type, priority, redundancy): """Add tasks to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(tasks_file, tasks_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # If true, warn user # if sleep: # pragma: no cover # click.secho(msg, fg='yellow') # Show progress bar with click.progressbar(data, label="Adding Tasks") as pgbar: for d in pgbar: task_info = create_task_info(d) response = config.pbclient.create_task(project_id=project.id, info=task_info, n_answers=redundancy, priority_0=priority) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data) check_api_error(response) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s tasks added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _add_helpingmaterials(config, helping_file, helping_type): """Add helping materials to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(helping_file, helping_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # Show progress bar with click.progressbar(data, label="Adding Helping Materials") as pgbar: for d in pgbar: helping_info, file_path = create_helping_material_info(d) if file_path: # Create first the media object hm = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info, file_path=file_path) check_api_error(hm) z = hm.info.copy() z.update(helping_info) hm.info = z response = config.pbclient.update_helping_material(hm) check_api_error(response) else: response = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info) check_api_error(response) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data, endpoint='/api/helpinmaterial') # If true, warn user if sleep: # pragma: no cover click.secho(msg, fg='yellow') # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s helping materials added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _delete_tasks(config, task_id, limit=100, offset=0): """Delete tasks from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.delete_task(task_id) check_api_error(response) return "Task.id = %s and its associated task_runs have been deleted" % task_id else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) while len(tasks) > 0: for t in tasks: response = config.pbclient.delete_task(t.id) check_api_error(response) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks and task_runs have been deleted" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_tasks_redundancy(config, task_id, redundancy, limit=300, offset=0): """Update tasks redundancy from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.find_tasks(project.id, id=task_id) check_api_error(response) task = response[0] task.n_answers = redundancy response = config.pbclient.update_task(task) check_api_error(response) msg = "Task.id = %s redundancy has been updated to %s" % (task_id, redundancy) return msg else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) with click.progressbar(tasks, label="Updating Tasks") as pgbar: while len(tasks) > 0: for t in pgbar: t.n_answers = redundancy response = config.pbclient.update_task(t) check_api_error(response) # Check if for the data we have to auto-throttle task update sleep, msg = enable_auto_throttling(config, tasks) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks redundancy have been updated" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def find_project_by_short_name(short_name, pbclient, all=None): """Return project by short_name.""" try: response = pbclient.find_project(short_name=short_name, all=all) check_api_error(response) if (len(response) == 0): msg = '%s not found! You can use the all=1 argument to \ search in all the server.' error = 'Project Not Found' raise ProjectNotFound(msg, error) return response[0] except exceptions.ConnectionError: raise except ProjectNotFound: raise def check_api_error(api_response): print(api_response) """Check if returned API response contains an error.""" if type(api_response) == dict and 'code' in api_response and api_response['code'] <> 200: print("Server response code: %s" % api_response['code']) print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) if type(api_response) == dict and (api_response.get('status') == 'failed'): if 'ProgrammingError' in api_response.get('exception_cls'): raise DatabaseError(message='PyBossa database error.', error=api_response) if ('DBIntegrityError' in api_response.get('exception_cls') and 'project' in api_response.get('target')): msg = 'PyBossa project already exists.' raise ProjectAlreadyExists(message=msg, error=api_response) if 'project' in api_response.get('target'): raise ProjectNotFound(message='PyBossa Project not found', error=api_response) if 'task' in api_response.get('target'): raise TaskNotFound(message='PyBossa Task not found', error=api_response) else: print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) def create_task_info(task): """Create task_info field.""" task_info = None if task.get('info'): task_info = task['info'] else: task_info = task return task_info def create_helping_material_info(helping): """Create helping_material_info field.""" helping_info = None file_path = None if helping.get('info'): helping_info = helping['info'] else: helping_info = helping if helping_info.get('file_path'): file_path = helping_info.get('file_path') del helping_info['file_path'] return helping_info, file_path def enable_auto_throttling(config, data, limit=299, endpoint='/api/task'): "Return sleep time if more tasks than those " \ "allowed by the server are requested." # Get header from server endpoint = config.server + endpoint headers = requests.head(endpoint).headers # Get limit server_limit = int(headers.get('X-RateLimit-Remaining', 0)) limit = server_limit or limit # Get reset time reset_epoch = int(headers.get('X-RateLimit-Reset', 0)) # Compute sleep time sleep = (reset_epoch - calendar.timegm(datetime.datetime.utcnow().utctimetuple())) msg = 'Warning: %s remaining hits to the endpoint.' \ ' Auto-throttling enabled!' % limit # If we have less than 10 hits on the endpoint, sleep if limit <= 10: return (sleep, msg) else: return 0, None def format_json_task(task_info): """Format task_info into JSON if applicable.""" try: return json.loads(task_info) except: return task_info def row_empty(row): """Check if all values in row are None.""" for value in row: if value is not None: return False return True class PbsHandler(PatternMatchingEventHandler): patterns = ['*/template.html', '*/tutorial.html', '*/long_description.md', '*/results.html', '*/bundle.js', '*/bundle.min.js'] def __init__(self, config, task_presenter, results, long_description, tutorial): super(PbsHandler, self).__init__() self.config = config self.task_presenter = task_presenter self.results = results self.long_description = long_description self.tutorial = tutorial def on_modified(self, event): what = 'directory' if event.is_directory else 'file' logging.info("Modified %s: %s", what, event.src_path) res = _update_project(self.config, self.task_presenter, self.results, self.long_description, self.tutorial) logging.info(res)
Scifabric/pbs
helpers.py
create_task_info
python
def create_task_info(task): task_info = None if task.get('info'): task_info = task['info'] else: task_info = task return task_info
Create task_info field.
train
https://github.com/Scifabric/pbs/blob/3e5d5f3f0f5d20f740eaacc4d6e872a0c9fb8b38/helpers.py#L400-L407
null
#!/usr/bin/env python # -*- coding: utf-8 -*- # This file is part of PyBOSSA. # # PyBOSSA is free software: you can redistribute it and/or modify # it under the terms of the GNU Affero General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # PyBOSSA is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU Affero General Public License for more details. # # You should have received a copy of the GNU Affero General Public License # along with PyBOSSA. If not, see <http://www.gnu.org/licenses/>. """ Helper functions for the pbs command line client. This module exports the following methods: * find_project_by_short_name: return the project by short_name. * check_api_errors: check for API errors returned by a PyBossa server. * format_error: format error message. * format_json_task: format a CSV row into JSON. """ import re import os import csv import json import time import click import datetime from StringIO import StringIO import polib import openpyxl import itertools from requests import exceptions import requests from pbsexceptions import * import logging from watchdog.observers import Observer from watchdog.events import PatternMatchingEventHandler import calendar __all__ = ['find_project_by_short_name', 'check_api_error', 'format_error', 'format_json_task', '_create_project', '_update_project', '_add_tasks', 'create_task_info', '_delete_tasks', 'enable_auto_throttling', '_update_tasks_redundancy', '_update_project_watch', 'PbsHandler', '_update_task_presenter_bundle_js', 'row_empty', '_add_helpingmaterials', 'create_helping_material_info'] def _create_project(config): """Create a project in a PyBossa server.""" try: response = config.pbclient.create_project(config.project['name'], config.project['short_name'], config.project['description']) check_api_error(response) return ("Project: %s created!" % config.project['short_name']) except exceptions.ConnectionError: return("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_project_watch(config, task_presenter, results, long_description, tutorial): # pragma: no cover """Update a project in a loop.""" logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(message)s', datefmt='%Y-%m-%d %H:%M:%S') path = os.getcwd() event_handler = PbsHandler(config, task_presenter, results, long_description, tutorial) observer = Observer() # We only want the current folder, not sub-folders observer.schedule(event_handler, path, recursive=False) observer.start() try: while True: time.sleep(1) except KeyboardInterrupt: observer.stop() observer.join() def _update_task_presenter_bundle_js(project): """Append to template a distribution bundle js.""" if os.path.isfile ('bundle.min.js'): with open('bundle.min.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js return if os.path.isfile ('bundle.js'): with open('bundle.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js def _update_project(config, task_presenter, results, long_description, tutorial): """Update a project.""" try: # Get project project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) # Update attributes project.name = config.project['name'] project.short_name = config.project['short_name'] project.description = config.project['description'] # Update long_description with open(long_description, 'r') as f: project.long_description = f.read() # Update task presenter with open(task_presenter, 'r') as f: project.info['task_presenter'] = f.read() _update_task_presenter_bundle_js(project) # Update results with open(results, 'r') as f: project.info['results'] = f.read() # Update tutorial with open(tutorial, 'r') as f: project.info['tutorial'] = f.read() response = config.pbclient.update_project(project) check_api_error(response) return ("Project %s updated!" % config.project['short_name']) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except ProjectNotFound: return ("Project not found! The project: %s is missing." \ " Use the flag --all=1 to search in all the server " \ % config.project['short_name']) except TaskNotFound: raise def _load_data(data_file, data_type): """Load data from CSV, JSON, Excel, ..., formats.""" raw_data = data_file.read() if data_type is None: data_type = data_file.name.split('.')[-1] # Data list to process data = [] # JSON type if data_type == 'json': data = json.loads(raw_data) return data # CSV type elif data_type == 'csv': csv_data = StringIO(raw_data) reader = csv.DictReader(csv_data, delimiter=',') for line in reader: data.append(line) return data elif data_type in ['xlsx', 'xlsm', 'xltx', 'xltm']: excel_data = StringIO(raw_data) wb = openpyxl.load_workbook(excel_data) ws = wb.active # First headers headers = [] for row in ws.iter_rows(max_row=1): for cell in row: tmp = '_'.join(cell.value.split(" ")).lower() headers.append(tmp) # Simulate DictReader for row in ws.iter_rows(row_offset=1): values = [] for cell in row: values.append(cell.value) tmp = dict(itertools.izip(headers, values)) if len(values) == len(headers) and not row_empty(values): data.append(tmp) return data # PO type elif data_type == 'po': po = polib.pofile(raw_data) for entry in po.untranslated_entries(): data.append(entry.__dict__) return data # PROPERTIES type (used in Java and Firefox extensions) elif data_type == 'properties': lines = raw_data.split('\n') for l in lines: if l: var_id, string = l.split('=') tmp = dict(var_id=var_id, string=string) data.append(tmp) return data else: return data def _add_tasks(config, tasks_file, tasks_type, priority, redundancy): """Add tasks to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(tasks_file, tasks_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # If true, warn user # if sleep: # pragma: no cover # click.secho(msg, fg='yellow') # Show progress bar with click.progressbar(data, label="Adding Tasks") as pgbar: for d in pgbar: task_info = create_task_info(d) response = config.pbclient.create_task(project_id=project.id, info=task_info, n_answers=redundancy, priority_0=priority) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data) check_api_error(response) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s tasks added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _add_helpingmaterials(config, helping_file, helping_type): """Add helping materials to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(helping_file, helping_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # Show progress bar with click.progressbar(data, label="Adding Helping Materials") as pgbar: for d in pgbar: helping_info, file_path = create_helping_material_info(d) if file_path: # Create first the media object hm = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info, file_path=file_path) check_api_error(hm) z = hm.info.copy() z.update(helping_info) hm.info = z response = config.pbclient.update_helping_material(hm) check_api_error(response) else: response = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info) check_api_error(response) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data, endpoint='/api/helpinmaterial') # If true, warn user if sleep: # pragma: no cover click.secho(msg, fg='yellow') # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s helping materials added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _delete_tasks(config, task_id, limit=100, offset=0): """Delete tasks from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.delete_task(task_id) check_api_error(response) return "Task.id = %s and its associated task_runs have been deleted" % task_id else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) while len(tasks) > 0: for t in tasks: response = config.pbclient.delete_task(t.id) check_api_error(response) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks and task_runs have been deleted" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_tasks_redundancy(config, task_id, redundancy, limit=300, offset=0): """Update tasks redundancy from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.find_tasks(project.id, id=task_id) check_api_error(response) task = response[0] task.n_answers = redundancy response = config.pbclient.update_task(task) check_api_error(response) msg = "Task.id = %s redundancy has been updated to %s" % (task_id, redundancy) return msg else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) with click.progressbar(tasks, label="Updating Tasks") as pgbar: while len(tasks) > 0: for t in pgbar: t.n_answers = redundancy response = config.pbclient.update_task(t) check_api_error(response) # Check if for the data we have to auto-throttle task update sleep, msg = enable_auto_throttling(config, tasks) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks redundancy have been updated" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def find_project_by_short_name(short_name, pbclient, all=None): """Return project by short_name.""" try: response = pbclient.find_project(short_name=short_name, all=all) check_api_error(response) if (len(response) == 0): msg = '%s not found! You can use the all=1 argument to \ search in all the server.' error = 'Project Not Found' raise ProjectNotFound(msg, error) return response[0] except exceptions.ConnectionError: raise except ProjectNotFound: raise def check_api_error(api_response): print(api_response) """Check if returned API response contains an error.""" if type(api_response) == dict and 'code' in api_response and api_response['code'] <> 200: print("Server response code: %s" % api_response['code']) print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) if type(api_response) == dict and (api_response.get('status') == 'failed'): if 'ProgrammingError' in api_response.get('exception_cls'): raise DatabaseError(message='PyBossa database error.', error=api_response) if ('DBIntegrityError' in api_response.get('exception_cls') and 'project' in api_response.get('target')): msg = 'PyBossa project already exists.' raise ProjectAlreadyExists(message=msg, error=api_response) if 'project' in api_response.get('target'): raise ProjectNotFound(message='PyBossa Project not found', error=api_response) if 'task' in api_response.get('target'): raise TaskNotFound(message='PyBossa Task not found', error=api_response) else: print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) def format_error(module, error): """Format the error for the given module.""" logging.error(module) # Beautify JSON error print error.message print json.dumps(error.error, sort_keys=True, indent=4, separators=(',', ': ')) exit(1) def create_helping_material_info(helping): """Create helping_material_info field.""" helping_info = None file_path = None if helping.get('info'): helping_info = helping['info'] else: helping_info = helping if helping_info.get('file_path'): file_path = helping_info.get('file_path') del helping_info['file_path'] return helping_info, file_path def enable_auto_throttling(config, data, limit=299, endpoint='/api/task'): "Return sleep time if more tasks than those " \ "allowed by the server are requested." # Get header from server endpoint = config.server + endpoint headers = requests.head(endpoint).headers # Get limit server_limit = int(headers.get('X-RateLimit-Remaining', 0)) limit = server_limit or limit # Get reset time reset_epoch = int(headers.get('X-RateLimit-Reset', 0)) # Compute sleep time sleep = (reset_epoch - calendar.timegm(datetime.datetime.utcnow().utctimetuple())) msg = 'Warning: %s remaining hits to the endpoint.' \ ' Auto-throttling enabled!' % limit # If we have less than 10 hits on the endpoint, sleep if limit <= 10: return (sleep, msg) else: return 0, None def format_json_task(task_info): """Format task_info into JSON if applicable.""" try: return json.loads(task_info) except: return task_info def row_empty(row): """Check if all values in row are None.""" for value in row: if value is not None: return False return True class PbsHandler(PatternMatchingEventHandler): patterns = ['*/template.html', '*/tutorial.html', '*/long_description.md', '*/results.html', '*/bundle.js', '*/bundle.min.js'] def __init__(self, config, task_presenter, results, long_description, tutorial): super(PbsHandler, self).__init__() self.config = config self.task_presenter = task_presenter self.results = results self.long_description = long_description self.tutorial = tutorial def on_modified(self, event): what = 'directory' if event.is_directory else 'file' logging.info("Modified %s: %s", what, event.src_path) res = _update_project(self.config, self.task_presenter, self.results, self.long_description, self.tutorial) logging.info(res)
Scifabric/pbs
helpers.py
create_helping_material_info
python
def create_helping_material_info(helping): helping_info = None file_path = None if helping.get('info'): helping_info = helping['info'] else: helping_info = helping if helping_info.get('file_path'): file_path = helping_info.get('file_path') del helping_info['file_path'] return helping_info, file_path
Create helping_material_info field.
train
https://github.com/Scifabric/pbs/blob/3e5d5f3f0f5d20f740eaacc4d6e872a0c9fb8b38/helpers.py#L410-L421
null
#!/usr/bin/env python # -*- coding: utf-8 -*- # This file is part of PyBOSSA. # # PyBOSSA is free software: you can redistribute it and/or modify # it under the terms of the GNU Affero General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # PyBOSSA is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU Affero General Public License for more details. # # You should have received a copy of the GNU Affero General Public License # along with PyBOSSA. If not, see <http://www.gnu.org/licenses/>. """ Helper functions for the pbs command line client. This module exports the following methods: * find_project_by_short_name: return the project by short_name. * check_api_errors: check for API errors returned by a PyBossa server. * format_error: format error message. * format_json_task: format a CSV row into JSON. """ import re import os import csv import json import time import click import datetime from StringIO import StringIO import polib import openpyxl import itertools from requests import exceptions import requests from pbsexceptions import * import logging from watchdog.observers import Observer from watchdog.events import PatternMatchingEventHandler import calendar __all__ = ['find_project_by_short_name', 'check_api_error', 'format_error', 'format_json_task', '_create_project', '_update_project', '_add_tasks', 'create_task_info', '_delete_tasks', 'enable_auto_throttling', '_update_tasks_redundancy', '_update_project_watch', 'PbsHandler', '_update_task_presenter_bundle_js', 'row_empty', '_add_helpingmaterials', 'create_helping_material_info'] def _create_project(config): """Create a project in a PyBossa server.""" try: response = config.pbclient.create_project(config.project['name'], config.project['short_name'], config.project['description']) check_api_error(response) return ("Project: %s created!" % config.project['short_name']) except exceptions.ConnectionError: return("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_project_watch(config, task_presenter, results, long_description, tutorial): # pragma: no cover """Update a project in a loop.""" logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(message)s', datefmt='%Y-%m-%d %H:%M:%S') path = os.getcwd() event_handler = PbsHandler(config, task_presenter, results, long_description, tutorial) observer = Observer() # We only want the current folder, not sub-folders observer.schedule(event_handler, path, recursive=False) observer.start() try: while True: time.sleep(1) except KeyboardInterrupt: observer.stop() observer.join() def _update_task_presenter_bundle_js(project): """Append to template a distribution bundle js.""" if os.path.isfile ('bundle.min.js'): with open('bundle.min.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js return if os.path.isfile ('bundle.js'): with open('bundle.js') as f: js = f.read() project.info['task_presenter'] += "<script>\n%s\n</script>" % js def _update_project(config, task_presenter, results, long_description, tutorial): """Update a project.""" try: # Get project project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) # Update attributes project.name = config.project['name'] project.short_name = config.project['short_name'] project.description = config.project['description'] # Update long_description with open(long_description, 'r') as f: project.long_description = f.read() # Update task presenter with open(task_presenter, 'r') as f: project.info['task_presenter'] = f.read() _update_task_presenter_bundle_js(project) # Update results with open(results, 'r') as f: project.info['results'] = f.read() # Update tutorial with open(tutorial, 'r') as f: project.info['tutorial'] = f.read() response = config.pbclient.update_project(project) check_api_error(response) return ("Project %s updated!" % config.project['short_name']) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except ProjectNotFound: return ("Project not found! The project: %s is missing." \ " Use the flag --all=1 to search in all the server " \ % config.project['short_name']) except TaskNotFound: raise def _load_data(data_file, data_type): """Load data from CSV, JSON, Excel, ..., formats.""" raw_data = data_file.read() if data_type is None: data_type = data_file.name.split('.')[-1] # Data list to process data = [] # JSON type if data_type == 'json': data = json.loads(raw_data) return data # CSV type elif data_type == 'csv': csv_data = StringIO(raw_data) reader = csv.DictReader(csv_data, delimiter=',') for line in reader: data.append(line) return data elif data_type in ['xlsx', 'xlsm', 'xltx', 'xltm']: excel_data = StringIO(raw_data) wb = openpyxl.load_workbook(excel_data) ws = wb.active # First headers headers = [] for row in ws.iter_rows(max_row=1): for cell in row: tmp = '_'.join(cell.value.split(" ")).lower() headers.append(tmp) # Simulate DictReader for row in ws.iter_rows(row_offset=1): values = [] for cell in row: values.append(cell.value) tmp = dict(itertools.izip(headers, values)) if len(values) == len(headers) and not row_empty(values): data.append(tmp) return data # PO type elif data_type == 'po': po = polib.pofile(raw_data) for entry in po.untranslated_entries(): data.append(entry.__dict__) return data # PROPERTIES type (used in Java and Firefox extensions) elif data_type == 'properties': lines = raw_data.split('\n') for l in lines: if l: var_id, string = l.split('=') tmp = dict(var_id=var_id, string=string) data.append(tmp) return data else: return data def _add_tasks(config, tasks_file, tasks_type, priority, redundancy): """Add tasks to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(tasks_file, tasks_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # If true, warn user # if sleep: # pragma: no cover # click.secho(msg, fg='yellow') # Show progress bar with click.progressbar(data, label="Adding Tasks") as pgbar: for d in pgbar: task_info = create_task_info(d) response = config.pbclient.create_task(project_id=project.id, info=task_info, n_answers=redundancy, priority_0=priority) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data) check_api_error(response) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s tasks added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _add_helpingmaterials(config, helping_file, helping_type): """Add helping materials to a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) data = _load_data(helping_file, helping_type) if len(data) == 0: return ("Unknown format for the tasks file. Use json, csv, po or " "properties.") # Show progress bar with click.progressbar(data, label="Adding Helping Materials") as pgbar: for d in pgbar: helping_info, file_path = create_helping_material_info(d) if file_path: # Create first the media object hm = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info, file_path=file_path) check_api_error(hm) z = hm.info.copy() z.update(helping_info) hm.info = z response = config.pbclient.update_helping_material(hm) check_api_error(response) else: response = config.pbclient.create_helpingmaterial(project_id=project.id, info=helping_info) check_api_error(response) # Check if for the data we have to auto-throttle task creation sleep, msg = enable_auto_throttling(config, data, endpoint='/api/helpinmaterial') # If true, warn user if sleep: # pragma: no cover click.secho(msg, fg='yellow') # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) return ("%s helping materials added to project: %s" % (len(data), config.project['short_name'])) except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _delete_tasks(config, task_id, limit=100, offset=0): """Delete tasks from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.delete_task(task_id) check_api_error(response) return "Task.id = %s and its associated task_runs have been deleted" % task_id else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) while len(tasks) > 0: for t in tasks: response = config.pbclient.delete_task(t.id) check_api_error(response) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks and task_runs have been deleted" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def _update_tasks_redundancy(config, task_id, redundancy, limit=300, offset=0): """Update tasks redundancy from a project.""" try: project = find_project_by_short_name(config.project['short_name'], config.pbclient, config.all) if task_id: response = config.pbclient.find_tasks(project.id, id=task_id) check_api_error(response) task = response[0] task.n_answers = redundancy response = config.pbclient.update_task(task) check_api_error(response) msg = "Task.id = %s redundancy has been updated to %s" % (task_id, redundancy) return msg else: limit = limit offset = offset tasks = config.pbclient.get_tasks(project.id, limit, offset) with click.progressbar(tasks, label="Updating Tasks") as pgbar: while len(tasks) > 0: for t in pgbar: t.n_answers = redundancy response = config.pbclient.update_task(t) check_api_error(response) # Check if for the data we have to auto-throttle task update sleep, msg = enable_auto_throttling(config, tasks) # If auto-throttling enabled, sleep for sleep seconds if sleep: # pragma: no cover time.sleep(sleep) offset += limit tasks = config.pbclient.get_tasks(project.id, limit, offset) return "All tasks redundancy have been updated" except exceptions.ConnectionError: return ("Connection Error! The server %s is not responding" % config.server) except (ProjectNotFound, TaskNotFound): raise def find_project_by_short_name(short_name, pbclient, all=None): """Return project by short_name.""" try: response = pbclient.find_project(short_name=short_name, all=all) check_api_error(response) if (len(response) == 0): msg = '%s not found! You can use the all=1 argument to \ search in all the server.' error = 'Project Not Found' raise ProjectNotFound(msg, error) return response[0] except exceptions.ConnectionError: raise except ProjectNotFound: raise def check_api_error(api_response): print(api_response) """Check if returned API response contains an error.""" if type(api_response) == dict and 'code' in api_response and api_response['code'] <> 200: print("Server response code: %s" % api_response['code']) print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) if type(api_response) == dict and (api_response.get('status') == 'failed'): if 'ProgrammingError' in api_response.get('exception_cls'): raise DatabaseError(message='PyBossa database error.', error=api_response) if ('DBIntegrityError' in api_response.get('exception_cls') and 'project' in api_response.get('target')): msg = 'PyBossa project already exists.' raise ProjectAlreadyExists(message=msg, error=api_response) if 'project' in api_response.get('target'): raise ProjectNotFound(message='PyBossa Project not found', error=api_response) if 'task' in api_response.get('target'): raise TaskNotFound(message='PyBossa Task not found', error=api_response) else: print("Server response: %s" % api_response) raise exceptions.HTTPError('Unexpected response', response=api_response) def format_error(module, error): """Format the error for the given module.""" logging.error(module) # Beautify JSON error print error.message print json.dumps(error.error, sort_keys=True, indent=4, separators=(',', ': ')) exit(1) def create_task_info(task): """Create task_info field.""" task_info = None if task.get('info'): task_info = task['info'] else: task_info = task return task_info def enable_auto_throttling(config, data, limit=299, endpoint='/api/task'): "Return sleep time if more tasks than those " \ "allowed by the server are requested." # Get header from server endpoint = config.server + endpoint headers = requests.head(endpoint).headers # Get limit server_limit = int(headers.get('X-RateLimit-Remaining', 0)) limit = server_limit or limit # Get reset time reset_epoch = int(headers.get('X-RateLimit-Reset', 0)) # Compute sleep time sleep = (reset_epoch - calendar.timegm(datetime.datetime.utcnow().utctimetuple())) msg = 'Warning: %s remaining hits to the endpoint.' \ ' Auto-throttling enabled!' % limit # If we have less than 10 hits on the endpoint, sleep if limit <= 10: return (sleep, msg) else: return 0, None def format_json_task(task_info): """Format task_info into JSON if applicable.""" try: return json.loads(task_info) except: return task_info def row_empty(row): """Check if all values in row are None.""" for value in row: if value is not None: return False return True class PbsHandler(PatternMatchingEventHandler): patterns = ['*/template.html', '*/tutorial.html', '*/long_description.md', '*/results.html', '*/bundle.js', '*/bundle.min.js'] def __init__(self, config, task_presenter, results, long_description, tutorial): super(PbsHandler, self).__init__() self.config = config self.task_presenter = task_presenter self.results = results self.long_description = long_description self.tutorial = tutorial def on_modified(self, event): what = 'directory' if event.is_directory else 'file' logging.info("Modified %s: %s", what, event.src_path) res = _update_project(self.config, self.task_presenter, self.results, self.long_description, self.tutorial) logging.info(res)
cenkalti/kuyruk
kuyruk/worker.py
Worker.run
python
def run(self) -> None: if self._logging_level: logging.basicConfig( level=getattr(logging, self._logging_level.upper()), format="%(levelname).1s %(name)s.%(funcName)s:%(lineno)d - %(message)s") signal.signal(signal.SIGINT, self._handle_sigint) signal.signal(signal.SIGTERM, self._handle_sigterm) if platform.system() != 'Windows': # These features will not be available on Windows, but that is OK. # Read this issue for more details: # https://github.com/cenkalti/kuyruk/issues/54 signal.signal(signal.SIGHUP, self._handle_sighup) signal.signal(signal.SIGUSR1, self._handle_sigusr1) signal.signal(signal.SIGUSR2, self._handle_sigusr2) self._started_at = os.times().elapsed for t in self._threads: t.start() try: signals.worker_start.send(self.kuyruk, worker=self) self._consume_messages() signals.worker_shutdown.send(self.kuyruk, worker=self) finally: self.shutdown_pending.set() for t in self._threads: t.join() logger.debug("End run worker")
Runs the worker and consumes messages from RabbitMQ. Returns only after `shutdown()` is called.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/worker.py#L81-L115
[ "def _consume_messages(self) -> None:\n with self.kuyruk.channel() as ch:\n # Set prefetch count to 1. If we don't set this, RabbitMQ keeps\n # sending messages while we are already working on a message.\n ch.basic_qos(0, 1, True)\n\n self._declare_queues(ch)\n self._consume_queues(ch)\n logger.info('Consumer started')\n self._main_loop(ch)\n" ]
class Worker: """Consumes tasks from queues and runs them. :param app: An instance of :class:`~kuyruk.Kuyruk` :param args: Command line arguments """ def __init__(self, app: Kuyruk, args: argparse.Namespace) -> None: self.kuyruk = app if not args.queues: args.queues = ['kuyruk'] def add_host(queue: str) -> str: if queue.endswith('.localhost'): queue = queue.rsplit('.localhost')[0] return "%s.%s" % (queue, self._hostname) else: return queue self._hostname = socket.gethostname() self.queues = [add_host(q) for q in args.queues] self._tasks = {} # type: Dict[Tuple[str, str], Task] self.shutdown_pending = threading.Event() self.consuming = False self.current_task = None # type: Task self.current_args = None # type: Tuple self.current_kwargs = None # type: Dict[str, Any] self._started_at = None # type: float self._pid = os.getpid() self._logging_level = app.config.WORKER_LOGGING_LEVEL if args.logging_level is not None: self._logging_level = args.logging_level self._max_run_time = app.config.WORKER_MAX_RUN_TIME if args.max_run_time is not None: self._max_run_time = args.max_run_time self._max_load = app.config.WORKER_MAX_LOAD if args.max_load is not None: self._max_load = args.max_load if self._max_load == -1: self._max_load == multiprocessing.cpu_count() self._threads = [] # type: List[threading.Thread] if self._max_load: self._threads.append(threading.Thread(target=self._watch_load)) if self._max_run_time: self._threads.append(threading.Thread(target=self._shutdown_timer)) signals.worker_init.send(self.kuyruk, worker=self) def _consume_messages(self) -> None: with self.kuyruk.channel() as ch: # Set prefetch count to 1. If we don't set this, RabbitMQ keeps # sending messages while we are already working on a message. ch.basic_qos(0, 1, True) self._declare_queues(ch) self._consume_queues(ch) logger.info('Consumer started') self._main_loop(ch) def _main_loop(self, ch: amqp.Channel) -> None: while not self.shutdown_pending.is_set(): self._pause_or_resume(ch) try: ch.connection.heartbeat_tick() ch.connection.drain_events(timeout=1) except socket.timeout: pass def _consumer_tag(self, queue: str) -> str: return "%s:%s@%s" % (queue, self._pid, self._hostname) def _declare_queues(self, ch: amqp.Channel) -> None: for queue in self.queues: logger.debug("queue_declare: %s", queue) ch.queue_declare(queue=queue, durable=True, auto_delete=False) def _pause_or_resume(self, channel: amqp.Channel) -> None: if not self._max_load: return try: load = self._current_load except AttributeError: should_pause = False else: should_pause = load > self._max_load if should_pause and self.consuming: logger.warning('Load is above the treshold (%.2f/%s), ' 'pausing consumer', load, self._max_load) self._cancel_queues(channel) elif not should_pause and not self.consuming: logger.warning('Load is below the treshold (%.2f/%s), ' 'resuming consumer', load, self._max_load) self._consume_queues(channel) def _consume_queues(self, ch: amqp.Channel) -> None: self.consuming = True for queue in self.queues: logger.debug("basic_consume: %s", queue) ch.basic_consume(queue=queue, consumer_tag=self._consumer_tag(queue), callback=self._process_message) def _cancel_queues(self, ch: amqp.Channel) -> None: self.consuming = False for queue in self.queues: logger.debug("basic_cancel: %s", queue) ch.basic_cancel(self._consumer_tag(queue)) def _process_message(self, message: amqp.Message) -> None: """Processes the message received from the queue.""" if self.shutdown_pending.is_set(): return try: if isinstance(message.body, bytes): message.body = message.body.decode() description = json.loads(message.body) except Exception: logger.error("Cannot decode message. Dropping. Message: %r", message.body) traceback.print_exc() message.channel.basic_reject(message.delivery_tag, requeue=False) else: logger.info("Processing task: %r", description) self._process_description(message, description) def _process_description(self, message: amqp.Message, description: Dict[str, Any]) -> None: try: task = self._import_task(description['module'], description['function']) args, kwargs = description['args'], description['kwargs'] except Exception: logger.error('Cannot import task') exc_info = sys.exc_info() signals.worker_failure.send(self.kuyruk, description=description, exc_info=exc_info, worker=self) message.channel.basic_reject(message.delivery_tag, requeue=False) else: self._process_task(message, description, task, args, kwargs) def _import_task(self, module: str, function: str) -> Task: if (module, function) in self._tasks: return self._tasks[(module, function)] task = importer.import_object(module, function) self._tasks[(module, function)] = task return task def _process_task( self, message: amqp.Message, description: Dict[str, Any], task: Task, args: Tuple, kwargs: Dict[str, Any], ) -> None: queue = message.delivery_info['routing_key'] reply_to = message.properties.get('reply_to') try: result = self._run_task(message.channel.connection, task, args, kwargs) except Reject: logger.warning('Task is rejected') message.channel.basic_reject(message.delivery_tag, requeue=True) except Discard: logger.warning('Task is discarded') message.channel.basic_reject(message.delivery_tag, requeue=False) if reply_to: exc_info = sys.exc_info() self._send_reply(reply_to, message.channel, None, exc_info) except HeartbeatError as e: logger.error('Error while sending heartbeat') exc_info = e.exc_info logger.error(''.join(traceback.format_exception(*exc_info))) signals.worker_failure.send( self.kuyruk, description=description, task=task, args=args, kwargs=kwargs, exc_info=exc_info, worker=self, queue=queue) raise except Exception: logger.error('Task raised an exception') exc_info = sys.exc_info() logger.error(''.join(traceback.format_exception(*exc_info))) signals.worker_failure.send( self.kuyruk, description=description, task=task, args=args, kwargs=kwargs, exc_info=exc_info, worker=self, queue=queue) message.channel.basic_reject(message.delivery_tag, requeue=False) if reply_to: self._send_reply(reply_to, message.channel, None, exc_info) else: logger.info('Task is successful') message.channel.basic_ack(message.delivery_tag) if reply_to: self._send_reply(reply_to, message.channel, result, None) finally: logger.debug("Task is processed") def _run_task(self, connection: amqp.Connection, task: Task, args: Tuple, kwargs: Dict[str, Any]) -> Any: hb = Heartbeat(connection, self._on_heartbeat_error) hb.start() self.current_task = task self.current_args = args self.current_kwargs = kwargs try: return self._apply_task(task, args, kwargs) finally: self.current_task = None self.current_args = None self.current_kwargs = None hb.stop() def _on_heartbeat_error(self, exc_info: ExcInfoType) -> None: self._heartbeat_exc_info = exc_info os.kill(os.getpid(), signal.SIGHUP) @staticmethod def _apply_task(task: Task, args: Tuple, kwargs: Dict[str, Any]) -> Any: """Logs the time spent while running the task.""" if args is None: args = () if kwargs is None: kwargs = {} start = monotonic() try: return task.apply(*args, **kwargs) finally: delta = monotonic() - start logger.info("%s finished in %i seconds." % (task.name, delta)) def _send_reply( self, reply_to: str, channel: amqp.Channel, result: Any, exc_info: ExcInfoType, ) -> None: logger.debug("Sending reply result=%r", result) reply = {'result': result} if exc_info: reply['exception'] = self._exc_info_dict(exc_info) try: body = json.dumps(reply) except Exception as e: logger.error('Cannot serialize result as JSON: %s', e) exc_info = sys.exc_info() reply = {'result': None, 'exception': self._exc_info_dict(exc_info)} body = json.dumps(reply) msg = amqp.Message(body=body) channel.basic_publish(msg, exchange="", routing_key=reply_to) @staticmethod def _exc_info_dict(exc_info: ExcInfoType) -> Dict[str, str]: type_, val, tb = exc_info return { 'type': '%s.%s' % (type_.__module__, type_.__name__), 'value': str(val), 'traceback': ''.join(traceback.format_tb(tb)), } def _watch_load(self) -> None: """Pause consuming messages if lood goes above the allowed limit.""" while not self.shutdown_pending.wait(1): self._current_load = os.getloadavg()[0] @property def uptime(self) -> float: if not self._started_at: return 0 return os.times().elapsed - self._started_at def _shutdown_timer(self) -> None: """Counts down from MAX_WORKER_RUN_TIME. When it reaches zero sutdown gracefully. """ remaining = self._max_run_time - self.uptime if not self.shutdown_pending.wait(remaining): logger.warning('Run time reached zero') self.shutdown() def shutdown(self) -> None: """Exits after the current task is finished.""" logger.warning("Shutdown requested") self.shutdown_pending.set() def _handle_sigint(self, signum: int, frame: Any) -> None: """Shutdown after processing current task.""" logger.warning("Catched SIGINT") self.shutdown() def _handle_sigterm(self, signum: int, frame: Any) -> None: """Shutdown after processing current task.""" logger.warning("Catched SIGTERM") self.shutdown() def _handle_sighup(self, signum: int, frame: Any) -> None: """Used internally to fail the task when connection to RabbitMQ is lost during the execution of the task. """ logger.warning("Catched SIGHUP") exc_info = self._heartbeat_exc_info self._heartbeat_exc_info = None # Format exception info to see in tools like Sentry. formatted_exception = ''.join(traceback.format_exception(*exc_info)) # noqa raise HeartbeatError(exc_info) @staticmethod def _handle_sigusr1(signum: int, frame: Any) -> None: """Print stacktrace.""" print('=' * 70) print(''.join(traceback.format_stack())) print('-' * 70) def _handle_sigusr2(self, signum: int, frame: Any) -> None: """Drop current task.""" logger.warning("Catched SIGUSR2") if self.current_task: logger.warning("Dropping current task...") raise Discard def drop_task(self) -> None: os.kill(os.getpid(), signal.SIGUSR2)
cenkalti/kuyruk
kuyruk/worker.py
Worker._process_message
python
def _process_message(self, message: amqp.Message) -> None: if self.shutdown_pending.is_set(): return try: if isinstance(message.body, bytes): message.body = message.body.decode() description = json.loads(message.body) except Exception: logger.error("Cannot decode message. Dropping. Message: %r", message.body) traceback.print_exc() message.channel.basic_reject(message.delivery_tag, requeue=False) else: logger.info("Processing task: %r", description) self._process_description(message, description)
Processes the message received from the queue.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/worker.py#L175-L190
null
class Worker: """Consumes tasks from queues and runs them. :param app: An instance of :class:`~kuyruk.Kuyruk` :param args: Command line arguments """ def __init__(self, app: Kuyruk, args: argparse.Namespace) -> None: self.kuyruk = app if not args.queues: args.queues = ['kuyruk'] def add_host(queue: str) -> str: if queue.endswith('.localhost'): queue = queue.rsplit('.localhost')[0] return "%s.%s" % (queue, self._hostname) else: return queue self._hostname = socket.gethostname() self.queues = [add_host(q) for q in args.queues] self._tasks = {} # type: Dict[Tuple[str, str], Task] self.shutdown_pending = threading.Event() self.consuming = False self.current_task = None # type: Task self.current_args = None # type: Tuple self.current_kwargs = None # type: Dict[str, Any] self._started_at = None # type: float self._pid = os.getpid() self._logging_level = app.config.WORKER_LOGGING_LEVEL if args.logging_level is not None: self._logging_level = args.logging_level self._max_run_time = app.config.WORKER_MAX_RUN_TIME if args.max_run_time is not None: self._max_run_time = args.max_run_time self._max_load = app.config.WORKER_MAX_LOAD if args.max_load is not None: self._max_load = args.max_load if self._max_load == -1: self._max_load == multiprocessing.cpu_count() self._threads = [] # type: List[threading.Thread] if self._max_load: self._threads.append(threading.Thread(target=self._watch_load)) if self._max_run_time: self._threads.append(threading.Thread(target=self._shutdown_timer)) signals.worker_init.send(self.kuyruk, worker=self) def run(self) -> None: """Runs the worker and consumes messages from RabbitMQ. Returns only after `shutdown()` is called. """ if self._logging_level: logging.basicConfig( level=getattr(logging, self._logging_level.upper()), format="%(levelname).1s %(name)s.%(funcName)s:%(lineno)d - %(message)s") signal.signal(signal.SIGINT, self._handle_sigint) signal.signal(signal.SIGTERM, self._handle_sigterm) if platform.system() != 'Windows': # These features will not be available on Windows, but that is OK. # Read this issue for more details: # https://github.com/cenkalti/kuyruk/issues/54 signal.signal(signal.SIGHUP, self._handle_sighup) signal.signal(signal.SIGUSR1, self._handle_sigusr1) signal.signal(signal.SIGUSR2, self._handle_sigusr2) self._started_at = os.times().elapsed for t in self._threads: t.start() try: signals.worker_start.send(self.kuyruk, worker=self) self._consume_messages() signals.worker_shutdown.send(self.kuyruk, worker=self) finally: self.shutdown_pending.set() for t in self._threads: t.join() logger.debug("End run worker") def _consume_messages(self) -> None: with self.kuyruk.channel() as ch: # Set prefetch count to 1. If we don't set this, RabbitMQ keeps # sending messages while we are already working on a message. ch.basic_qos(0, 1, True) self._declare_queues(ch) self._consume_queues(ch) logger.info('Consumer started') self._main_loop(ch) def _main_loop(self, ch: amqp.Channel) -> None: while not self.shutdown_pending.is_set(): self._pause_or_resume(ch) try: ch.connection.heartbeat_tick() ch.connection.drain_events(timeout=1) except socket.timeout: pass def _consumer_tag(self, queue: str) -> str: return "%s:%s@%s" % (queue, self._pid, self._hostname) def _declare_queues(self, ch: amqp.Channel) -> None: for queue in self.queues: logger.debug("queue_declare: %s", queue) ch.queue_declare(queue=queue, durable=True, auto_delete=False) def _pause_or_resume(self, channel: amqp.Channel) -> None: if not self._max_load: return try: load = self._current_load except AttributeError: should_pause = False else: should_pause = load > self._max_load if should_pause and self.consuming: logger.warning('Load is above the treshold (%.2f/%s), ' 'pausing consumer', load, self._max_load) self._cancel_queues(channel) elif not should_pause and not self.consuming: logger.warning('Load is below the treshold (%.2f/%s), ' 'resuming consumer', load, self._max_load) self._consume_queues(channel) def _consume_queues(self, ch: amqp.Channel) -> None: self.consuming = True for queue in self.queues: logger.debug("basic_consume: %s", queue) ch.basic_consume(queue=queue, consumer_tag=self._consumer_tag(queue), callback=self._process_message) def _cancel_queues(self, ch: amqp.Channel) -> None: self.consuming = False for queue in self.queues: logger.debug("basic_cancel: %s", queue) ch.basic_cancel(self._consumer_tag(queue)) def _process_description(self, message: amqp.Message, description: Dict[str, Any]) -> None: try: task = self._import_task(description['module'], description['function']) args, kwargs = description['args'], description['kwargs'] except Exception: logger.error('Cannot import task') exc_info = sys.exc_info() signals.worker_failure.send(self.kuyruk, description=description, exc_info=exc_info, worker=self) message.channel.basic_reject(message.delivery_tag, requeue=False) else: self._process_task(message, description, task, args, kwargs) def _import_task(self, module: str, function: str) -> Task: if (module, function) in self._tasks: return self._tasks[(module, function)] task = importer.import_object(module, function) self._tasks[(module, function)] = task return task def _process_task( self, message: amqp.Message, description: Dict[str, Any], task: Task, args: Tuple, kwargs: Dict[str, Any], ) -> None: queue = message.delivery_info['routing_key'] reply_to = message.properties.get('reply_to') try: result = self._run_task(message.channel.connection, task, args, kwargs) except Reject: logger.warning('Task is rejected') message.channel.basic_reject(message.delivery_tag, requeue=True) except Discard: logger.warning('Task is discarded') message.channel.basic_reject(message.delivery_tag, requeue=False) if reply_to: exc_info = sys.exc_info() self._send_reply(reply_to, message.channel, None, exc_info) except HeartbeatError as e: logger.error('Error while sending heartbeat') exc_info = e.exc_info logger.error(''.join(traceback.format_exception(*exc_info))) signals.worker_failure.send( self.kuyruk, description=description, task=task, args=args, kwargs=kwargs, exc_info=exc_info, worker=self, queue=queue) raise except Exception: logger.error('Task raised an exception') exc_info = sys.exc_info() logger.error(''.join(traceback.format_exception(*exc_info))) signals.worker_failure.send( self.kuyruk, description=description, task=task, args=args, kwargs=kwargs, exc_info=exc_info, worker=self, queue=queue) message.channel.basic_reject(message.delivery_tag, requeue=False) if reply_to: self._send_reply(reply_to, message.channel, None, exc_info) else: logger.info('Task is successful') message.channel.basic_ack(message.delivery_tag) if reply_to: self._send_reply(reply_to, message.channel, result, None) finally: logger.debug("Task is processed") def _run_task(self, connection: amqp.Connection, task: Task, args: Tuple, kwargs: Dict[str, Any]) -> Any: hb = Heartbeat(connection, self._on_heartbeat_error) hb.start() self.current_task = task self.current_args = args self.current_kwargs = kwargs try: return self._apply_task(task, args, kwargs) finally: self.current_task = None self.current_args = None self.current_kwargs = None hb.stop() def _on_heartbeat_error(self, exc_info: ExcInfoType) -> None: self._heartbeat_exc_info = exc_info os.kill(os.getpid(), signal.SIGHUP) @staticmethod def _apply_task(task: Task, args: Tuple, kwargs: Dict[str, Any]) -> Any: """Logs the time spent while running the task.""" if args is None: args = () if kwargs is None: kwargs = {} start = monotonic() try: return task.apply(*args, **kwargs) finally: delta = monotonic() - start logger.info("%s finished in %i seconds." % (task.name, delta)) def _send_reply( self, reply_to: str, channel: amqp.Channel, result: Any, exc_info: ExcInfoType, ) -> None: logger.debug("Sending reply result=%r", result) reply = {'result': result} if exc_info: reply['exception'] = self._exc_info_dict(exc_info) try: body = json.dumps(reply) except Exception as e: logger.error('Cannot serialize result as JSON: %s', e) exc_info = sys.exc_info() reply = {'result': None, 'exception': self._exc_info_dict(exc_info)} body = json.dumps(reply) msg = amqp.Message(body=body) channel.basic_publish(msg, exchange="", routing_key=reply_to) @staticmethod def _exc_info_dict(exc_info: ExcInfoType) -> Dict[str, str]: type_, val, tb = exc_info return { 'type': '%s.%s' % (type_.__module__, type_.__name__), 'value': str(val), 'traceback': ''.join(traceback.format_tb(tb)), } def _watch_load(self) -> None: """Pause consuming messages if lood goes above the allowed limit.""" while not self.shutdown_pending.wait(1): self._current_load = os.getloadavg()[0] @property def uptime(self) -> float: if not self._started_at: return 0 return os.times().elapsed - self._started_at def _shutdown_timer(self) -> None: """Counts down from MAX_WORKER_RUN_TIME. When it reaches zero sutdown gracefully. """ remaining = self._max_run_time - self.uptime if not self.shutdown_pending.wait(remaining): logger.warning('Run time reached zero') self.shutdown() def shutdown(self) -> None: """Exits after the current task is finished.""" logger.warning("Shutdown requested") self.shutdown_pending.set() def _handle_sigint(self, signum: int, frame: Any) -> None: """Shutdown after processing current task.""" logger.warning("Catched SIGINT") self.shutdown() def _handle_sigterm(self, signum: int, frame: Any) -> None: """Shutdown after processing current task.""" logger.warning("Catched SIGTERM") self.shutdown() def _handle_sighup(self, signum: int, frame: Any) -> None: """Used internally to fail the task when connection to RabbitMQ is lost during the execution of the task. """ logger.warning("Catched SIGHUP") exc_info = self._heartbeat_exc_info self._heartbeat_exc_info = None # Format exception info to see in tools like Sentry. formatted_exception = ''.join(traceback.format_exception(*exc_info)) # noqa raise HeartbeatError(exc_info) @staticmethod def _handle_sigusr1(signum: int, frame: Any) -> None: """Print stacktrace.""" print('=' * 70) print(''.join(traceback.format_stack())) print('-' * 70) def _handle_sigusr2(self, signum: int, frame: Any) -> None: """Drop current task.""" logger.warning("Catched SIGUSR2") if self.current_task: logger.warning("Dropping current task...") raise Discard def drop_task(self) -> None: os.kill(os.getpid(), signal.SIGUSR2)
cenkalti/kuyruk
kuyruk/worker.py
Worker._apply_task
python
def _apply_task(task: Task, args: Tuple, kwargs: Dict[str, Any]) -> Any: if args is None: args = () if kwargs is None: kwargs = {} start = monotonic() try: return task.apply(*args, **kwargs) finally: delta = monotonic() - start logger.info("%s finished in %i seconds." % (task.name, delta))
Logs the time spent while running the task.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/worker.py#L292-L304
null
class Worker: """Consumes tasks from queues and runs them. :param app: An instance of :class:`~kuyruk.Kuyruk` :param args: Command line arguments """ def __init__(self, app: Kuyruk, args: argparse.Namespace) -> None: self.kuyruk = app if not args.queues: args.queues = ['kuyruk'] def add_host(queue: str) -> str: if queue.endswith('.localhost'): queue = queue.rsplit('.localhost')[0] return "%s.%s" % (queue, self._hostname) else: return queue self._hostname = socket.gethostname() self.queues = [add_host(q) for q in args.queues] self._tasks = {} # type: Dict[Tuple[str, str], Task] self.shutdown_pending = threading.Event() self.consuming = False self.current_task = None # type: Task self.current_args = None # type: Tuple self.current_kwargs = None # type: Dict[str, Any] self._started_at = None # type: float self._pid = os.getpid() self._logging_level = app.config.WORKER_LOGGING_LEVEL if args.logging_level is not None: self._logging_level = args.logging_level self._max_run_time = app.config.WORKER_MAX_RUN_TIME if args.max_run_time is not None: self._max_run_time = args.max_run_time self._max_load = app.config.WORKER_MAX_LOAD if args.max_load is not None: self._max_load = args.max_load if self._max_load == -1: self._max_load == multiprocessing.cpu_count() self._threads = [] # type: List[threading.Thread] if self._max_load: self._threads.append(threading.Thread(target=self._watch_load)) if self._max_run_time: self._threads.append(threading.Thread(target=self._shutdown_timer)) signals.worker_init.send(self.kuyruk, worker=self) def run(self) -> None: """Runs the worker and consumes messages from RabbitMQ. Returns only after `shutdown()` is called. """ if self._logging_level: logging.basicConfig( level=getattr(logging, self._logging_level.upper()), format="%(levelname).1s %(name)s.%(funcName)s:%(lineno)d - %(message)s") signal.signal(signal.SIGINT, self._handle_sigint) signal.signal(signal.SIGTERM, self._handle_sigterm) if platform.system() != 'Windows': # These features will not be available on Windows, but that is OK. # Read this issue for more details: # https://github.com/cenkalti/kuyruk/issues/54 signal.signal(signal.SIGHUP, self._handle_sighup) signal.signal(signal.SIGUSR1, self._handle_sigusr1) signal.signal(signal.SIGUSR2, self._handle_sigusr2) self._started_at = os.times().elapsed for t in self._threads: t.start() try: signals.worker_start.send(self.kuyruk, worker=self) self._consume_messages() signals.worker_shutdown.send(self.kuyruk, worker=self) finally: self.shutdown_pending.set() for t in self._threads: t.join() logger.debug("End run worker") def _consume_messages(self) -> None: with self.kuyruk.channel() as ch: # Set prefetch count to 1. If we don't set this, RabbitMQ keeps # sending messages while we are already working on a message. ch.basic_qos(0, 1, True) self._declare_queues(ch) self._consume_queues(ch) logger.info('Consumer started') self._main_loop(ch) def _main_loop(self, ch: amqp.Channel) -> None: while not self.shutdown_pending.is_set(): self._pause_or_resume(ch) try: ch.connection.heartbeat_tick() ch.connection.drain_events(timeout=1) except socket.timeout: pass def _consumer_tag(self, queue: str) -> str: return "%s:%s@%s" % (queue, self._pid, self._hostname) def _declare_queues(self, ch: amqp.Channel) -> None: for queue in self.queues: logger.debug("queue_declare: %s", queue) ch.queue_declare(queue=queue, durable=True, auto_delete=False) def _pause_or_resume(self, channel: amqp.Channel) -> None: if not self._max_load: return try: load = self._current_load except AttributeError: should_pause = False else: should_pause = load > self._max_load if should_pause and self.consuming: logger.warning('Load is above the treshold (%.2f/%s), ' 'pausing consumer', load, self._max_load) self._cancel_queues(channel) elif not should_pause and not self.consuming: logger.warning('Load is below the treshold (%.2f/%s), ' 'resuming consumer', load, self._max_load) self._consume_queues(channel) def _consume_queues(self, ch: amqp.Channel) -> None: self.consuming = True for queue in self.queues: logger.debug("basic_consume: %s", queue) ch.basic_consume(queue=queue, consumer_tag=self._consumer_tag(queue), callback=self._process_message) def _cancel_queues(self, ch: amqp.Channel) -> None: self.consuming = False for queue in self.queues: logger.debug("basic_cancel: %s", queue) ch.basic_cancel(self._consumer_tag(queue)) def _process_message(self, message: amqp.Message) -> None: """Processes the message received from the queue.""" if self.shutdown_pending.is_set(): return try: if isinstance(message.body, bytes): message.body = message.body.decode() description = json.loads(message.body) except Exception: logger.error("Cannot decode message. Dropping. Message: %r", message.body) traceback.print_exc() message.channel.basic_reject(message.delivery_tag, requeue=False) else: logger.info("Processing task: %r", description) self._process_description(message, description) def _process_description(self, message: amqp.Message, description: Dict[str, Any]) -> None: try: task = self._import_task(description['module'], description['function']) args, kwargs = description['args'], description['kwargs'] except Exception: logger.error('Cannot import task') exc_info = sys.exc_info() signals.worker_failure.send(self.kuyruk, description=description, exc_info=exc_info, worker=self) message.channel.basic_reject(message.delivery_tag, requeue=False) else: self._process_task(message, description, task, args, kwargs) def _import_task(self, module: str, function: str) -> Task: if (module, function) in self._tasks: return self._tasks[(module, function)] task = importer.import_object(module, function) self._tasks[(module, function)] = task return task def _process_task( self, message: amqp.Message, description: Dict[str, Any], task: Task, args: Tuple, kwargs: Dict[str, Any], ) -> None: queue = message.delivery_info['routing_key'] reply_to = message.properties.get('reply_to') try: result = self._run_task(message.channel.connection, task, args, kwargs) except Reject: logger.warning('Task is rejected') message.channel.basic_reject(message.delivery_tag, requeue=True) except Discard: logger.warning('Task is discarded') message.channel.basic_reject(message.delivery_tag, requeue=False) if reply_to: exc_info = sys.exc_info() self._send_reply(reply_to, message.channel, None, exc_info) except HeartbeatError as e: logger.error('Error while sending heartbeat') exc_info = e.exc_info logger.error(''.join(traceback.format_exception(*exc_info))) signals.worker_failure.send( self.kuyruk, description=description, task=task, args=args, kwargs=kwargs, exc_info=exc_info, worker=self, queue=queue) raise except Exception: logger.error('Task raised an exception') exc_info = sys.exc_info() logger.error(''.join(traceback.format_exception(*exc_info))) signals.worker_failure.send( self.kuyruk, description=description, task=task, args=args, kwargs=kwargs, exc_info=exc_info, worker=self, queue=queue) message.channel.basic_reject(message.delivery_tag, requeue=False) if reply_to: self._send_reply(reply_to, message.channel, None, exc_info) else: logger.info('Task is successful') message.channel.basic_ack(message.delivery_tag) if reply_to: self._send_reply(reply_to, message.channel, result, None) finally: logger.debug("Task is processed") def _run_task(self, connection: amqp.Connection, task: Task, args: Tuple, kwargs: Dict[str, Any]) -> Any: hb = Heartbeat(connection, self._on_heartbeat_error) hb.start() self.current_task = task self.current_args = args self.current_kwargs = kwargs try: return self._apply_task(task, args, kwargs) finally: self.current_task = None self.current_args = None self.current_kwargs = None hb.stop() def _on_heartbeat_error(self, exc_info: ExcInfoType) -> None: self._heartbeat_exc_info = exc_info os.kill(os.getpid(), signal.SIGHUP) @staticmethod def _send_reply( self, reply_to: str, channel: amqp.Channel, result: Any, exc_info: ExcInfoType, ) -> None: logger.debug("Sending reply result=%r", result) reply = {'result': result} if exc_info: reply['exception'] = self._exc_info_dict(exc_info) try: body = json.dumps(reply) except Exception as e: logger.error('Cannot serialize result as JSON: %s', e) exc_info = sys.exc_info() reply = {'result': None, 'exception': self._exc_info_dict(exc_info)} body = json.dumps(reply) msg = amqp.Message(body=body) channel.basic_publish(msg, exchange="", routing_key=reply_to) @staticmethod def _exc_info_dict(exc_info: ExcInfoType) -> Dict[str, str]: type_, val, tb = exc_info return { 'type': '%s.%s' % (type_.__module__, type_.__name__), 'value': str(val), 'traceback': ''.join(traceback.format_tb(tb)), } def _watch_load(self) -> None: """Pause consuming messages if lood goes above the allowed limit.""" while not self.shutdown_pending.wait(1): self._current_load = os.getloadavg()[0] @property def uptime(self) -> float: if not self._started_at: return 0 return os.times().elapsed - self._started_at def _shutdown_timer(self) -> None: """Counts down from MAX_WORKER_RUN_TIME. When it reaches zero sutdown gracefully. """ remaining = self._max_run_time - self.uptime if not self.shutdown_pending.wait(remaining): logger.warning('Run time reached zero') self.shutdown() def shutdown(self) -> None: """Exits after the current task is finished.""" logger.warning("Shutdown requested") self.shutdown_pending.set() def _handle_sigint(self, signum: int, frame: Any) -> None: """Shutdown after processing current task.""" logger.warning("Catched SIGINT") self.shutdown() def _handle_sigterm(self, signum: int, frame: Any) -> None: """Shutdown after processing current task.""" logger.warning("Catched SIGTERM") self.shutdown() def _handle_sighup(self, signum: int, frame: Any) -> None: """Used internally to fail the task when connection to RabbitMQ is lost during the execution of the task. """ logger.warning("Catched SIGHUP") exc_info = self._heartbeat_exc_info self._heartbeat_exc_info = None # Format exception info to see in tools like Sentry. formatted_exception = ''.join(traceback.format_exception(*exc_info)) # noqa raise HeartbeatError(exc_info) @staticmethod def _handle_sigusr1(signum: int, frame: Any) -> None: """Print stacktrace.""" print('=' * 70) print(''.join(traceback.format_stack())) print('-' * 70) def _handle_sigusr2(self, signum: int, frame: Any) -> None: """Drop current task.""" logger.warning("Catched SIGUSR2") if self.current_task: logger.warning("Dropping current task...") raise Discard def drop_task(self) -> None: os.kill(os.getpid(), signal.SIGUSR2)
cenkalti/kuyruk
kuyruk/worker.py
Worker._shutdown_timer
python
def _shutdown_timer(self) -> None: remaining = self._max_run_time - self.uptime if not self.shutdown_pending.wait(remaining): logger.warning('Run time reached zero') self.shutdown()
Counts down from MAX_WORKER_RUN_TIME. When it reaches zero sutdown gracefully.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/worker.py#L351-L359
null
class Worker: """Consumes tasks from queues and runs them. :param app: An instance of :class:`~kuyruk.Kuyruk` :param args: Command line arguments """ def __init__(self, app: Kuyruk, args: argparse.Namespace) -> None: self.kuyruk = app if not args.queues: args.queues = ['kuyruk'] def add_host(queue: str) -> str: if queue.endswith('.localhost'): queue = queue.rsplit('.localhost')[0] return "%s.%s" % (queue, self._hostname) else: return queue self._hostname = socket.gethostname() self.queues = [add_host(q) for q in args.queues] self._tasks = {} # type: Dict[Tuple[str, str], Task] self.shutdown_pending = threading.Event() self.consuming = False self.current_task = None # type: Task self.current_args = None # type: Tuple self.current_kwargs = None # type: Dict[str, Any] self._started_at = None # type: float self._pid = os.getpid() self._logging_level = app.config.WORKER_LOGGING_LEVEL if args.logging_level is not None: self._logging_level = args.logging_level self._max_run_time = app.config.WORKER_MAX_RUN_TIME if args.max_run_time is not None: self._max_run_time = args.max_run_time self._max_load = app.config.WORKER_MAX_LOAD if args.max_load is not None: self._max_load = args.max_load if self._max_load == -1: self._max_load == multiprocessing.cpu_count() self._threads = [] # type: List[threading.Thread] if self._max_load: self._threads.append(threading.Thread(target=self._watch_load)) if self._max_run_time: self._threads.append(threading.Thread(target=self._shutdown_timer)) signals.worker_init.send(self.kuyruk, worker=self) def run(self) -> None: """Runs the worker and consumes messages from RabbitMQ. Returns only after `shutdown()` is called. """ if self._logging_level: logging.basicConfig( level=getattr(logging, self._logging_level.upper()), format="%(levelname).1s %(name)s.%(funcName)s:%(lineno)d - %(message)s") signal.signal(signal.SIGINT, self._handle_sigint) signal.signal(signal.SIGTERM, self._handle_sigterm) if platform.system() != 'Windows': # These features will not be available on Windows, but that is OK. # Read this issue for more details: # https://github.com/cenkalti/kuyruk/issues/54 signal.signal(signal.SIGHUP, self._handle_sighup) signal.signal(signal.SIGUSR1, self._handle_sigusr1) signal.signal(signal.SIGUSR2, self._handle_sigusr2) self._started_at = os.times().elapsed for t in self._threads: t.start() try: signals.worker_start.send(self.kuyruk, worker=self) self._consume_messages() signals.worker_shutdown.send(self.kuyruk, worker=self) finally: self.shutdown_pending.set() for t in self._threads: t.join() logger.debug("End run worker") def _consume_messages(self) -> None: with self.kuyruk.channel() as ch: # Set prefetch count to 1. If we don't set this, RabbitMQ keeps # sending messages while we are already working on a message. ch.basic_qos(0, 1, True) self._declare_queues(ch) self._consume_queues(ch) logger.info('Consumer started') self._main_loop(ch) def _main_loop(self, ch: amqp.Channel) -> None: while not self.shutdown_pending.is_set(): self._pause_or_resume(ch) try: ch.connection.heartbeat_tick() ch.connection.drain_events(timeout=1) except socket.timeout: pass def _consumer_tag(self, queue: str) -> str: return "%s:%s@%s" % (queue, self._pid, self._hostname) def _declare_queues(self, ch: amqp.Channel) -> None: for queue in self.queues: logger.debug("queue_declare: %s", queue) ch.queue_declare(queue=queue, durable=True, auto_delete=False) def _pause_or_resume(self, channel: amqp.Channel) -> None: if not self._max_load: return try: load = self._current_load except AttributeError: should_pause = False else: should_pause = load > self._max_load if should_pause and self.consuming: logger.warning('Load is above the treshold (%.2f/%s), ' 'pausing consumer', load, self._max_load) self._cancel_queues(channel) elif not should_pause and not self.consuming: logger.warning('Load is below the treshold (%.2f/%s), ' 'resuming consumer', load, self._max_load) self._consume_queues(channel) def _consume_queues(self, ch: amqp.Channel) -> None: self.consuming = True for queue in self.queues: logger.debug("basic_consume: %s", queue) ch.basic_consume(queue=queue, consumer_tag=self._consumer_tag(queue), callback=self._process_message) def _cancel_queues(self, ch: amqp.Channel) -> None: self.consuming = False for queue in self.queues: logger.debug("basic_cancel: %s", queue) ch.basic_cancel(self._consumer_tag(queue)) def _process_message(self, message: amqp.Message) -> None: """Processes the message received from the queue.""" if self.shutdown_pending.is_set(): return try: if isinstance(message.body, bytes): message.body = message.body.decode() description = json.loads(message.body) except Exception: logger.error("Cannot decode message. Dropping. Message: %r", message.body) traceback.print_exc() message.channel.basic_reject(message.delivery_tag, requeue=False) else: logger.info("Processing task: %r", description) self._process_description(message, description) def _process_description(self, message: amqp.Message, description: Dict[str, Any]) -> None: try: task = self._import_task(description['module'], description['function']) args, kwargs = description['args'], description['kwargs'] except Exception: logger.error('Cannot import task') exc_info = sys.exc_info() signals.worker_failure.send(self.kuyruk, description=description, exc_info=exc_info, worker=self) message.channel.basic_reject(message.delivery_tag, requeue=False) else: self._process_task(message, description, task, args, kwargs) def _import_task(self, module: str, function: str) -> Task: if (module, function) in self._tasks: return self._tasks[(module, function)] task = importer.import_object(module, function) self._tasks[(module, function)] = task return task def _process_task( self, message: amqp.Message, description: Dict[str, Any], task: Task, args: Tuple, kwargs: Dict[str, Any], ) -> None: queue = message.delivery_info['routing_key'] reply_to = message.properties.get('reply_to') try: result = self._run_task(message.channel.connection, task, args, kwargs) except Reject: logger.warning('Task is rejected') message.channel.basic_reject(message.delivery_tag, requeue=True) except Discard: logger.warning('Task is discarded') message.channel.basic_reject(message.delivery_tag, requeue=False) if reply_to: exc_info = sys.exc_info() self._send_reply(reply_to, message.channel, None, exc_info) except HeartbeatError as e: logger.error('Error while sending heartbeat') exc_info = e.exc_info logger.error(''.join(traceback.format_exception(*exc_info))) signals.worker_failure.send( self.kuyruk, description=description, task=task, args=args, kwargs=kwargs, exc_info=exc_info, worker=self, queue=queue) raise except Exception: logger.error('Task raised an exception') exc_info = sys.exc_info() logger.error(''.join(traceback.format_exception(*exc_info))) signals.worker_failure.send( self.kuyruk, description=description, task=task, args=args, kwargs=kwargs, exc_info=exc_info, worker=self, queue=queue) message.channel.basic_reject(message.delivery_tag, requeue=False) if reply_to: self._send_reply(reply_to, message.channel, None, exc_info) else: logger.info('Task is successful') message.channel.basic_ack(message.delivery_tag) if reply_to: self._send_reply(reply_to, message.channel, result, None) finally: logger.debug("Task is processed") def _run_task(self, connection: amqp.Connection, task: Task, args: Tuple, kwargs: Dict[str, Any]) -> Any: hb = Heartbeat(connection, self._on_heartbeat_error) hb.start() self.current_task = task self.current_args = args self.current_kwargs = kwargs try: return self._apply_task(task, args, kwargs) finally: self.current_task = None self.current_args = None self.current_kwargs = None hb.stop() def _on_heartbeat_error(self, exc_info: ExcInfoType) -> None: self._heartbeat_exc_info = exc_info os.kill(os.getpid(), signal.SIGHUP) @staticmethod def _apply_task(task: Task, args: Tuple, kwargs: Dict[str, Any]) -> Any: """Logs the time spent while running the task.""" if args is None: args = () if kwargs is None: kwargs = {} start = monotonic() try: return task.apply(*args, **kwargs) finally: delta = monotonic() - start logger.info("%s finished in %i seconds." % (task.name, delta)) def _send_reply( self, reply_to: str, channel: amqp.Channel, result: Any, exc_info: ExcInfoType, ) -> None: logger.debug("Sending reply result=%r", result) reply = {'result': result} if exc_info: reply['exception'] = self._exc_info_dict(exc_info) try: body = json.dumps(reply) except Exception as e: logger.error('Cannot serialize result as JSON: %s', e) exc_info = sys.exc_info() reply = {'result': None, 'exception': self._exc_info_dict(exc_info)} body = json.dumps(reply) msg = amqp.Message(body=body) channel.basic_publish(msg, exchange="", routing_key=reply_to) @staticmethod def _exc_info_dict(exc_info: ExcInfoType) -> Dict[str, str]: type_, val, tb = exc_info return { 'type': '%s.%s' % (type_.__module__, type_.__name__), 'value': str(val), 'traceback': ''.join(traceback.format_tb(tb)), } def _watch_load(self) -> None: """Pause consuming messages if lood goes above the allowed limit.""" while not self.shutdown_pending.wait(1): self._current_load = os.getloadavg()[0] @property def uptime(self) -> float: if not self._started_at: return 0 return os.times().elapsed - self._started_at def shutdown(self) -> None: """Exits after the current task is finished.""" logger.warning("Shutdown requested") self.shutdown_pending.set() def _handle_sigint(self, signum: int, frame: Any) -> None: """Shutdown after processing current task.""" logger.warning("Catched SIGINT") self.shutdown() def _handle_sigterm(self, signum: int, frame: Any) -> None: """Shutdown after processing current task.""" logger.warning("Catched SIGTERM") self.shutdown() def _handle_sighup(self, signum: int, frame: Any) -> None: """Used internally to fail the task when connection to RabbitMQ is lost during the execution of the task. """ logger.warning("Catched SIGHUP") exc_info = self._heartbeat_exc_info self._heartbeat_exc_info = None # Format exception info to see in tools like Sentry. formatted_exception = ''.join(traceback.format_exception(*exc_info)) # noqa raise HeartbeatError(exc_info) @staticmethod def _handle_sigusr1(signum: int, frame: Any) -> None: """Print stacktrace.""" print('=' * 70) print(''.join(traceback.format_stack())) print('-' * 70) def _handle_sigusr2(self, signum: int, frame: Any) -> None: """Drop current task.""" logger.warning("Catched SIGUSR2") if self.current_task: logger.warning("Dropping current task...") raise Discard def drop_task(self) -> None: os.kill(os.getpid(), signal.SIGUSR2)
cenkalti/kuyruk
kuyruk/worker.py
Worker._handle_sigint
python
def _handle_sigint(self, signum: int, frame: Any) -> None: logger.warning("Catched SIGINT") self.shutdown()
Shutdown after processing current task.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/worker.py#L366-L369
null
class Worker: """Consumes tasks from queues and runs them. :param app: An instance of :class:`~kuyruk.Kuyruk` :param args: Command line arguments """ def __init__(self, app: Kuyruk, args: argparse.Namespace) -> None: self.kuyruk = app if not args.queues: args.queues = ['kuyruk'] def add_host(queue: str) -> str: if queue.endswith('.localhost'): queue = queue.rsplit('.localhost')[0] return "%s.%s" % (queue, self._hostname) else: return queue self._hostname = socket.gethostname() self.queues = [add_host(q) for q in args.queues] self._tasks = {} # type: Dict[Tuple[str, str], Task] self.shutdown_pending = threading.Event() self.consuming = False self.current_task = None # type: Task self.current_args = None # type: Tuple self.current_kwargs = None # type: Dict[str, Any] self._started_at = None # type: float self._pid = os.getpid() self._logging_level = app.config.WORKER_LOGGING_LEVEL if args.logging_level is not None: self._logging_level = args.logging_level self._max_run_time = app.config.WORKER_MAX_RUN_TIME if args.max_run_time is not None: self._max_run_time = args.max_run_time self._max_load = app.config.WORKER_MAX_LOAD if args.max_load is not None: self._max_load = args.max_load if self._max_load == -1: self._max_load == multiprocessing.cpu_count() self._threads = [] # type: List[threading.Thread] if self._max_load: self._threads.append(threading.Thread(target=self._watch_load)) if self._max_run_time: self._threads.append(threading.Thread(target=self._shutdown_timer)) signals.worker_init.send(self.kuyruk, worker=self) def run(self) -> None: """Runs the worker and consumes messages from RabbitMQ. Returns only after `shutdown()` is called. """ if self._logging_level: logging.basicConfig( level=getattr(logging, self._logging_level.upper()), format="%(levelname).1s %(name)s.%(funcName)s:%(lineno)d - %(message)s") signal.signal(signal.SIGINT, self._handle_sigint) signal.signal(signal.SIGTERM, self._handle_sigterm) if platform.system() != 'Windows': # These features will not be available on Windows, but that is OK. # Read this issue for more details: # https://github.com/cenkalti/kuyruk/issues/54 signal.signal(signal.SIGHUP, self._handle_sighup) signal.signal(signal.SIGUSR1, self._handle_sigusr1) signal.signal(signal.SIGUSR2, self._handle_sigusr2) self._started_at = os.times().elapsed for t in self._threads: t.start() try: signals.worker_start.send(self.kuyruk, worker=self) self._consume_messages() signals.worker_shutdown.send(self.kuyruk, worker=self) finally: self.shutdown_pending.set() for t in self._threads: t.join() logger.debug("End run worker") def _consume_messages(self) -> None: with self.kuyruk.channel() as ch: # Set prefetch count to 1. If we don't set this, RabbitMQ keeps # sending messages while we are already working on a message. ch.basic_qos(0, 1, True) self._declare_queues(ch) self._consume_queues(ch) logger.info('Consumer started') self._main_loop(ch) def _main_loop(self, ch: amqp.Channel) -> None: while not self.shutdown_pending.is_set(): self._pause_or_resume(ch) try: ch.connection.heartbeat_tick() ch.connection.drain_events(timeout=1) except socket.timeout: pass def _consumer_tag(self, queue: str) -> str: return "%s:%s@%s" % (queue, self._pid, self._hostname) def _declare_queues(self, ch: amqp.Channel) -> None: for queue in self.queues: logger.debug("queue_declare: %s", queue) ch.queue_declare(queue=queue, durable=True, auto_delete=False) def _pause_or_resume(self, channel: amqp.Channel) -> None: if not self._max_load: return try: load = self._current_load except AttributeError: should_pause = False else: should_pause = load > self._max_load if should_pause and self.consuming: logger.warning('Load is above the treshold (%.2f/%s), ' 'pausing consumer', load, self._max_load) self._cancel_queues(channel) elif not should_pause and not self.consuming: logger.warning('Load is below the treshold (%.2f/%s), ' 'resuming consumer', load, self._max_load) self._consume_queues(channel) def _consume_queues(self, ch: amqp.Channel) -> None: self.consuming = True for queue in self.queues: logger.debug("basic_consume: %s", queue) ch.basic_consume(queue=queue, consumer_tag=self._consumer_tag(queue), callback=self._process_message) def _cancel_queues(self, ch: amqp.Channel) -> None: self.consuming = False for queue in self.queues: logger.debug("basic_cancel: %s", queue) ch.basic_cancel(self._consumer_tag(queue)) def _process_message(self, message: amqp.Message) -> None: """Processes the message received from the queue.""" if self.shutdown_pending.is_set(): return try: if isinstance(message.body, bytes): message.body = message.body.decode() description = json.loads(message.body) except Exception: logger.error("Cannot decode message. Dropping. Message: %r", message.body) traceback.print_exc() message.channel.basic_reject(message.delivery_tag, requeue=False) else: logger.info("Processing task: %r", description) self._process_description(message, description) def _process_description(self, message: amqp.Message, description: Dict[str, Any]) -> None: try: task = self._import_task(description['module'], description['function']) args, kwargs = description['args'], description['kwargs'] except Exception: logger.error('Cannot import task') exc_info = sys.exc_info() signals.worker_failure.send(self.kuyruk, description=description, exc_info=exc_info, worker=self) message.channel.basic_reject(message.delivery_tag, requeue=False) else: self._process_task(message, description, task, args, kwargs) def _import_task(self, module: str, function: str) -> Task: if (module, function) in self._tasks: return self._tasks[(module, function)] task = importer.import_object(module, function) self._tasks[(module, function)] = task return task def _process_task( self, message: amqp.Message, description: Dict[str, Any], task: Task, args: Tuple, kwargs: Dict[str, Any], ) -> None: queue = message.delivery_info['routing_key'] reply_to = message.properties.get('reply_to') try: result = self._run_task(message.channel.connection, task, args, kwargs) except Reject: logger.warning('Task is rejected') message.channel.basic_reject(message.delivery_tag, requeue=True) except Discard: logger.warning('Task is discarded') message.channel.basic_reject(message.delivery_tag, requeue=False) if reply_to: exc_info = sys.exc_info() self._send_reply(reply_to, message.channel, None, exc_info) except HeartbeatError as e: logger.error('Error while sending heartbeat') exc_info = e.exc_info logger.error(''.join(traceback.format_exception(*exc_info))) signals.worker_failure.send( self.kuyruk, description=description, task=task, args=args, kwargs=kwargs, exc_info=exc_info, worker=self, queue=queue) raise except Exception: logger.error('Task raised an exception') exc_info = sys.exc_info() logger.error(''.join(traceback.format_exception(*exc_info))) signals.worker_failure.send( self.kuyruk, description=description, task=task, args=args, kwargs=kwargs, exc_info=exc_info, worker=self, queue=queue) message.channel.basic_reject(message.delivery_tag, requeue=False) if reply_to: self._send_reply(reply_to, message.channel, None, exc_info) else: logger.info('Task is successful') message.channel.basic_ack(message.delivery_tag) if reply_to: self._send_reply(reply_to, message.channel, result, None) finally: logger.debug("Task is processed") def _run_task(self, connection: amqp.Connection, task: Task, args: Tuple, kwargs: Dict[str, Any]) -> Any: hb = Heartbeat(connection, self._on_heartbeat_error) hb.start() self.current_task = task self.current_args = args self.current_kwargs = kwargs try: return self._apply_task(task, args, kwargs) finally: self.current_task = None self.current_args = None self.current_kwargs = None hb.stop() def _on_heartbeat_error(self, exc_info: ExcInfoType) -> None: self._heartbeat_exc_info = exc_info os.kill(os.getpid(), signal.SIGHUP) @staticmethod def _apply_task(task: Task, args: Tuple, kwargs: Dict[str, Any]) -> Any: """Logs the time spent while running the task.""" if args is None: args = () if kwargs is None: kwargs = {} start = monotonic() try: return task.apply(*args, **kwargs) finally: delta = monotonic() - start logger.info("%s finished in %i seconds." % (task.name, delta)) def _send_reply( self, reply_to: str, channel: amqp.Channel, result: Any, exc_info: ExcInfoType, ) -> None: logger.debug("Sending reply result=%r", result) reply = {'result': result} if exc_info: reply['exception'] = self._exc_info_dict(exc_info) try: body = json.dumps(reply) except Exception as e: logger.error('Cannot serialize result as JSON: %s', e) exc_info = sys.exc_info() reply = {'result': None, 'exception': self._exc_info_dict(exc_info)} body = json.dumps(reply) msg = amqp.Message(body=body) channel.basic_publish(msg, exchange="", routing_key=reply_to) @staticmethod def _exc_info_dict(exc_info: ExcInfoType) -> Dict[str, str]: type_, val, tb = exc_info return { 'type': '%s.%s' % (type_.__module__, type_.__name__), 'value': str(val), 'traceback': ''.join(traceback.format_tb(tb)), } def _watch_load(self) -> None: """Pause consuming messages if lood goes above the allowed limit.""" while not self.shutdown_pending.wait(1): self._current_load = os.getloadavg()[0] @property def uptime(self) -> float: if not self._started_at: return 0 return os.times().elapsed - self._started_at def _shutdown_timer(self) -> None: """Counts down from MAX_WORKER_RUN_TIME. When it reaches zero sutdown gracefully. """ remaining = self._max_run_time - self.uptime if not self.shutdown_pending.wait(remaining): logger.warning('Run time reached zero') self.shutdown() def shutdown(self) -> None: """Exits after the current task is finished.""" logger.warning("Shutdown requested") self.shutdown_pending.set() def _handle_sigterm(self, signum: int, frame: Any) -> None: """Shutdown after processing current task.""" logger.warning("Catched SIGTERM") self.shutdown() def _handle_sighup(self, signum: int, frame: Any) -> None: """Used internally to fail the task when connection to RabbitMQ is lost during the execution of the task. """ logger.warning("Catched SIGHUP") exc_info = self._heartbeat_exc_info self._heartbeat_exc_info = None # Format exception info to see in tools like Sentry. formatted_exception = ''.join(traceback.format_exception(*exc_info)) # noqa raise HeartbeatError(exc_info) @staticmethod def _handle_sigusr1(signum: int, frame: Any) -> None: """Print stacktrace.""" print('=' * 70) print(''.join(traceback.format_stack())) print('-' * 70) def _handle_sigusr2(self, signum: int, frame: Any) -> None: """Drop current task.""" logger.warning("Catched SIGUSR2") if self.current_task: logger.warning("Dropping current task...") raise Discard def drop_task(self) -> None: os.kill(os.getpid(), signal.SIGUSR2)
cenkalti/kuyruk
kuyruk/worker.py
Worker._handle_sighup
python
def _handle_sighup(self, signum: int, frame: Any) -> None: logger.warning("Catched SIGHUP") exc_info = self._heartbeat_exc_info self._heartbeat_exc_info = None # Format exception info to see in tools like Sentry. formatted_exception = ''.join(traceback.format_exception(*exc_info)) # noqa raise HeartbeatError(exc_info)
Used internally to fail the task when connection to RabbitMQ is lost during the execution of the task.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/worker.py#L376-L386
null
class Worker: """Consumes tasks from queues and runs them. :param app: An instance of :class:`~kuyruk.Kuyruk` :param args: Command line arguments """ def __init__(self, app: Kuyruk, args: argparse.Namespace) -> None: self.kuyruk = app if not args.queues: args.queues = ['kuyruk'] def add_host(queue: str) -> str: if queue.endswith('.localhost'): queue = queue.rsplit('.localhost')[0] return "%s.%s" % (queue, self._hostname) else: return queue self._hostname = socket.gethostname() self.queues = [add_host(q) for q in args.queues] self._tasks = {} # type: Dict[Tuple[str, str], Task] self.shutdown_pending = threading.Event() self.consuming = False self.current_task = None # type: Task self.current_args = None # type: Tuple self.current_kwargs = None # type: Dict[str, Any] self._started_at = None # type: float self._pid = os.getpid() self._logging_level = app.config.WORKER_LOGGING_LEVEL if args.logging_level is not None: self._logging_level = args.logging_level self._max_run_time = app.config.WORKER_MAX_RUN_TIME if args.max_run_time is not None: self._max_run_time = args.max_run_time self._max_load = app.config.WORKER_MAX_LOAD if args.max_load is not None: self._max_load = args.max_load if self._max_load == -1: self._max_load == multiprocessing.cpu_count() self._threads = [] # type: List[threading.Thread] if self._max_load: self._threads.append(threading.Thread(target=self._watch_load)) if self._max_run_time: self._threads.append(threading.Thread(target=self._shutdown_timer)) signals.worker_init.send(self.kuyruk, worker=self) def run(self) -> None: """Runs the worker and consumes messages from RabbitMQ. Returns only after `shutdown()` is called. """ if self._logging_level: logging.basicConfig( level=getattr(logging, self._logging_level.upper()), format="%(levelname).1s %(name)s.%(funcName)s:%(lineno)d - %(message)s") signal.signal(signal.SIGINT, self._handle_sigint) signal.signal(signal.SIGTERM, self._handle_sigterm) if platform.system() != 'Windows': # These features will not be available on Windows, but that is OK. # Read this issue for more details: # https://github.com/cenkalti/kuyruk/issues/54 signal.signal(signal.SIGHUP, self._handle_sighup) signal.signal(signal.SIGUSR1, self._handle_sigusr1) signal.signal(signal.SIGUSR2, self._handle_sigusr2) self._started_at = os.times().elapsed for t in self._threads: t.start() try: signals.worker_start.send(self.kuyruk, worker=self) self._consume_messages() signals.worker_shutdown.send(self.kuyruk, worker=self) finally: self.shutdown_pending.set() for t in self._threads: t.join() logger.debug("End run worker") def _consume_messages(self) -> None: with self.kuyruk.channel() as ch: # Set prefetch count to 1. If we don't set this, RabbitMQ keeps # sending messages while we are already working on a message. ch.basic_qos(0, 1, True) self._declare_queues(ch) self._consume_queues(ch) logger.info('Consumer started') self._main_loop(ch) def _main_loop(self, ch: amqp.Channel) -> None: while not self.shutdown_pending.is_set(): self._pause_or_resume(ch) try: ch.connection.heartbeat_tick() ch.connection.drain_events(timeout=1) except socket.timeout: pass def _consumer_tag(self, queue: str) -> str: return "%s:%s@%s" % (queue, self._pid, self._hostname) def _declare_queues(self, ch: amqp.Channel) -> None: for queue in self.queues: logger.debug("queue_declare: %s", queue) ch.queue_declare(queue=queue, durable=True, auto_delete=False) def _pause_or_resume(self, channel: amqp.Channel) -> None: if not self._max_load: return try: load = self._current_load except AttributeError: should_pause = False else: should_pause = load > self._max_load if should_pause and self.consuming: logger.warning('Load is above the treshold (%.2f/%s), ' 'pausing consumer', load, self._max_load) self._cancel_queues(channel) elif not should_pause and not self.consuming: logger.warning('Load is below the treshold (%.2f/%s), ' 'resuming consumer', load, self._max_load) self._consume_queues(channel) def _consume_queues(self, ch: amqp.Channel) -> None: self.consuming = True for queue in self.queues: logger.debug("basic_consume: %s", queue) ch.basic_consume(queue=queue, consumer_tag=self._consumer_tag(queue), callback=self._process_message) def _cancel_queues(self, ch: amqp.Channel) -> None: self.consuming = False for queue in self.queues: logger.debug("basic_cancel: %s", queue) ch.basic_cancel(self._consumer_tag(queue)) def _process_message(self, message: amqp.Message) -> None: """Processes the message received from the queue.""" if self.shutdown_pending.is_set(): return try: if isinstance(message.body, bytes): message.body = message.body.decode() description = json.loads(message.body) except Exception: logger.error("Cannot decode message. Dropping. Message: %r", message.body) traceback.print_exc() message.channel.basic_reject(message.delivery_tag, requeue=False) else: logger.info("Processing task: %r", description) self._process_description(message, description) def _process_description(self, message: amqp.Message, description: Dict[str, Any]) -> None: try: task = self._import_task(description['module'], description['function']) args, kwargs = description['args'], description['kwargs'] except Exception: logger.error('Cannot import task') exc_info = sys.exc_info() signals.worker_failure.send(self.kuyruk, description=description, exc_info=exc_info, worker=self) message.channel.basic_reject(message.delivery_tag, requeue=False) else: self._process_task(message, description, task, args, kwargs) def _import_task(self, module: str, function: str) -> Task: if (module, function) in self._tasks: return self._tasks[(module, function)] task = importer.import_object(module, function) self._tasks[(module, function)] = task return task def _process_task( self, message: amqp.Message, description: Dict[str, Any], task: Task, args: Tuple, kwargs: Dict[str, Any], ) -> None: queue = message.delivery_info['routing_key'] reply_to = message.properties.get('reply_to') try: result = self._run_task(message.channel.connection, task, args, kwargs) except Reject: logger.warning('Task is rejected') message.channel.basic_reject(message.delivery_tag, requeue=True) except Discard: logger.warning('Task is discarded') message.channel.basic_reject(message.delivery_tag, requeue=False) if reply_to: exc_info = sys.exc_info() self._send_reply(reply_to, message.channel, None, exc_info) except HeartbeatError as e: logger.error('Error while sending heartbeat') exc_info = e.exc_info logger.error(''.join(traceback.format_exception(*exc_info))) signals.worker_failure.send( self.kuyruk, description=description, task=task, args=args, kwargs=kwargs, exc_info=exc_info, worker=self, queue=queue) raise except Exception: logger.error('Task raised an exception') exc_info = sys.exc_info() logger.error(''.join(traceback.format_exception(*exc_info))) signals.worker_failure.send( self.kuyruk, description=description, task=task, args=args, kwargs=kwargs, exc_info=exc_info, worker=self, queue=queue) message.channel.basic_reject(message.delivery_tag, requeue=False) if reply_to: self._send_reply(reply_to, message.channel, None, exc_info) else: logger.info('Task is successful') message.channel.basic_ack(message.delivery_tag) if reply_to: self._send_reply(reply_to, message.channel, result, None) finally: logger.debug("Task is processed") def _run_task(self, connection: amqp.Connection, task: Task, args: Tuple, kwargs: Dict[str, Any]) -> Any: hb = Heartbeat(connection, self._on_heartbeat_error) hb.start() self.current_task = task self.current_args = args self.current_kwargs = kwargs try: return self._apply_task(task, args, kwargs) finally: self.current_task = None self.current_args = None self.current_kwargs = None hb.stop() def _on_heartbeat_error(self, exc_info: ExcInfoType) -> None: self._heartbeat_exc_info = exc_info os.kill(os.getpid(), signal.SIGHUP) @staticmethod def _apply_task(task: Task, args: Tuple, kwargs: Dict[str, Any]) -> Any: """Logs the time spent while running the task.""" if args is None: args = () if kwargs is None: kwargs = {} start = monotonic() try: return task.apply(*args, **kwargs) finally: delta = monotonic() - start logger.info("%s finished in %i seconds." % (task.name, delta)) def _send_reply( self, reply_to: str, channel: amqp.Channel, result: Any, exc_info: ExcInfoType, ) -> None: logger.debug("Sending reply result=%r", result) reply = {'result': result} if exc_info: reply['exception'] = self._exc_info_dict(exc_info) try: body = json.dumps(reply) except Exception as e: logger.error('Cannot serialize result as JSON: %s', e) exc_info = sys.exc_info() reply = {'result': None, 'exception': self._exc_info_dict(exc_info)} body = json.dumps(reply) msg = amqp.Message(body=body) channel.basic_publish(msg, exchange="", routing_key=reply_to) @staticmethod def _exc_info_dict(exc_info: ExcInfoType) -> Dict[str, str]: type_, val, tb = exc_info return { 'type': '%s.%s' % (type_.__module__, type_.__name__), 'value': str(val), 'traceback': ''.join(traceback.format_tb(tb)), } def _watch_load(self) -> None: """Pause consuming messages if lood goes above the allowed limit.""" while not self.shutdown_pending.wait(1): self._current_load = os.getloadavg()[0] @property def uptime(self) -> float: if not self._started_at: return 0 return os.times().elapsed - self._started_at def _shutdown_timer(self) -> None: """Counts down from MAX_WORKER_RUN_TIME. When it reaches zero sutdown gracefully. """ remaining = self._max_run_time - self.uptime if not self.shutdown_pending.wait(remaining): logger.warning('Run time reached zero') self.shutdown() def shutdown(self) -> None: """Exits after the current task is finished.""" logger.warning("Shutdown requested") self.shutdown_pending.set() def _handle_sigint(self, signum: int, frame: Any) -> None: """Shutdown after processing current task.""" logger.warning("Catched SIGINT") self.shutdown() def _handle_sigterm(self, signum: int, frame: Any) -> None: """Shutdown after processing current task.""" logger.warning("Catched SIGTERM") self.shutdown() @staticmethod def _handle_sigusr1(signum: int, frame: Any) -> None: """Print stacktrace.""" print('=' * 70) print(''.join(traceback.format_stack())) print('-' * 70) def _handle_sigusr2(self, signum: int, frame: Any) -> None: """Drop current task.""" logger.warning("Catched SIGUSR2") if self.current_task: logger.warning("Dropping current task...") raise Discard def drop_task(self) -> None: os.kill(os.getpid(), signal.SIGUSR2)
cenkalti/kuyruk
kuyruk/worker.py
Worker._handle_sigusr1
python
def _handle_sigusr1(signum: int, frame: Any) -> None: print('=' * 70) print(''.join(traceback.format_stack())) print('-' * 70)
Print stacktrace.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/worker.py#L389-L393
null
class Worker: """Consumes tasks from queues and runs them. :param app: An instance of :class:`~kuyruk.Kuyruk` :param args: Command line arguments """ def __init__(self, app: Kuyruk, args: argparse.Namespace) -> None: self.kuyruk = app if not args.queues: args.queues = ['kuyruk'] def add_host(queue: str) -> str: if queue.endswith('.localhost'): queue = queue.rsplit('.localhost')[0] return "%s.%s" % (queue, self._hostname) else: return queue self._hostname = socket.gethostname() self.queues = [add_host(q) for q in args.queues] self._tasks = {} # type: Dict[Tuple[str, str], Task] self.shutdown_pending = threading.Event() self.consuming = False self.current_task = None # type: Task self.current_args = None # type: Tuple self.current_kwargs = None # type: Dict[str, Any] self._started_at = None # type: float self._pid = os.getpid() self._logging_level = app.config.WORKER_LOGGING_LEVEL if args.logging_level is not None: self._logging_level = args.logging_level self._max_run_time = app.config.WORKER_MAX_RUN_TIME if args.max_run_time is not None: self._max_run_time = args.max_run_time self._max_load = app.config.WORKER_MAX_LOAD if args.max_load is not None: self._max_load = args.max_load if self._max_load == -1: self._max_load == multiprocessing.cpu_count() self._threads = [] # type: List[threading.Thread] if self._max_load: self._threads.append(threading.Thread(target=self._watch_load)) if self._max_run_time: self._threads.append(threading.Thread(target=self._shutdown_timer)) signals.worker_init.send(self.kuyruk, worker=self) def run(self) -> None: """Runs the worker and consumes messages from RabbitMQ. Returns only after `shutdown()` is called. """ if self._logging_level: logging.basicConfig( level=getattr(logging, self._logging_level.upper()), format="%(levelname).1s %(name)s.%(funcName)s:%(lineno)d - %(message)s") signal.signal(signal.SIGINT, self._handle_sigint) signal.signal(signal.SIGTERM, self._handle_sigterm) if platform.system() != 'Windows': # These features will not be available on Windows, but that is OK. # Read this issue for more details: # https://github.com/cenkalti/kuyruk/issues/54 signal.signal(signal.SIGHUP, self._handle_sighup) signal.signal(signal.SIGUSR1, self._handle_sigusr1) signal.signal(signal.SIGUSR2, self._handle_sigusr2) self._started_at = os.times().elapsed for t in self._threads: t.start() try: signals.worker_start.send(self.kuyruk, worker=self) self._consume_messages() signals.worker_shutdown.send(self.kuyruk, worker=self) finally: self.shutdown_pending.set() for t in self._threads: t.join() logger.debug("End run worker") def _consume_messages(self) -> None: with self.kuyruk.channel() as ch: # Set prefetch count to 1. If we don't set this, RabbitMQ keeps # sending messages while we are already working on a message. ch.basic_qos(0, 1, True) self._declare_queues(ch) self._consume_queues(ch) logger.info('Consumer started') self._main_loop(ch) def _main_loop(self, ch: amqp.Channel) -> None: while not self.shutdown_pending.is_set(): self._pause_or_resume(ch) try: ch.connection.heartbeat_tick() ch.connection.drain_events(timeout=1) except socket.timeout: pass def _consumer_tag(self, queue: str) -> str: return "%s:%s@%s" % (queue, self._pid, self._hostname) def _declare_queues(self, ch: amqp.Channel) -> None: for queue in self.queues: logger.debug("queue_declare: %s", queue) ch.queue_declare(queue=queue, durable=True, auto_delete=False) def _pause_or_resume(self, channel: amqp.Channel) -> None: if not self._max_load: return try: load = self._current_load except AttributeError: should_pause = False else: should_pause = load > self._max_load if should_pause and self.consuming: logger.warning('Load is above the treshold (%.2f/%s), ' 'pausing consumer', load, self._max_load) self._cancel_queues(channel) elif not should_pause and not self.consuming: logger.warning('Load is below the treshold (%.2f/%s), ' 'resuming consumer', load, self._max_load) self._consume_queues(channel) def _consume_queues(self, ch: amqp.Channel) -> None: self.consuming = True for queue in self.queues: logger.debug("basic_consume: %s", queue) ch.basic_consume(queue=queue, consumer_tag=self._consumer_tag(queue), callback=self._process_message) def _cancel_queues(self, ch: amqp.Channel) -> None: self.consuming = False for queue in self.queues: logger.debug("basic_cancel: %s", queue) ch.basic_cancel(self._consumer_tag(queue)) def _process_message(self, message: amqp.Message) -> None: """Processes the message received from the queue.""" if self.shutdown_pending.is_set(): return try: if isinstance(message.body, bytes): message.body = message.body.decode() description = json.loads(message.body) except Exception: logger.error("Cannot decode message. Dropping. Message: %r", message.body) traceback.print_exc() message.channel.basic_reject(message.delivery_tag, requeue=False) else: logger.info("Processing task: %r", description) self._process_description(message, description) def _process_description(self, message: amqp.Message, description: Dict[str, Any]) -> None: try: task = self._import_task(description['module'], description['function']) args, kwargs = description['args'], description['kwargs'] except Exception: logger.error('Cannot import task') exc_info = sys.exc_info() signals.worker_failure.send(self.kuyruk, description=description, exc_info=exc_info, worker=self) message.channel.basic_reject(message.delivery_tag, requeue=False) else: self._process_task(message, description, task, args, kwargs) def _import_task(self, module: str, function: str) -> Task: if (module, function) in self._tasks: return self._tasks[(module, function)] task = importer.import_object(module, function) self._tasks[(module, function)] = task return task def _process_task( self, message: amqp.Message, description: Dict[str, Any], task: Task, args: Tuple, kwargs: Dict[str, Any], ) -> None: queue = message.delivery_info['routing_key'] reply_to = message.properties.get('reply_to') try: result = self._run_task(message.channel.connection, task, args, kwargs) except Reject: logger.warning('Task is rejected') message.channel.basic_reject(message.delivery_tag, requeue=True) except Discard: logger.warning('Task is discarded') message.channel.basic_reject(message.delivery_tag, requeue=False) if reply_to: exc_info = sys.exc_info() self._send_reply(reply_to, message.channel, None, exc_info) except HeartbeatError as e: logger.error('Error while sending heartbeat') exc_info = e.exc_info logger.error(''.join(traceback.format_exception(*exc_info))) signals.worker_failure.send( self.kuyruk, description=description, task=task, args=args, kwargs=kwargs, exc_info=exc_info, worker=self, queue=queue) raise except Exception: logger.error('Task raised an exception') exc_info = sys.exc_info() logger.error(''.join(traceback.format_exception(*exc_info))) signals.worker_failure.send( self.kuyruk, description=description, task=task, args=args, kwargs=kwargs, exc_info=exc_info, worker=self, queue=queue) message.channel.basic_reject(message.delivery_tag, requeue=False) if reply_to: self._send_reply(reply_to, message.channel, None, exc_info) else: logger.info('Task is successful') message.channel.basic_ack(message.delivery_tag) if reply_to: self._send_reply(reply_to, message.channel, result, None) finally: logger.debug("Task is processed") def _run_task(self, connection: amqp.Connection, task: Task, args: Tuple, kwargs: Dict[str, Any]) -> Any: hb = Heartbeat(connection, self._on_heartbeat_error) hb.start() self.current_task = task self.current_args = args self.current_kwargs = kwargs try: return self._apply_task(task, args, kwargs) finally: self.current_task = None self.current_args = None self.current_kwargs = None hb.stop() def _on_heartbeat_error(self, exc_info: ExcInfoType) -> None: self._heartbeat_exc_info = exc_info os.kill(os.getpid(), signal.SIGHUP) @staticmethod def _apply_task(task: Task, args: Tuple, kwargs: Dict[str, Any]) -> Any: """Logs the time spent while running the task.""" if args is None: args = () if kwargs is None: kwargs = {} start = monotonic() try: return task.apply(*args, **kwargs) finally: delta = monotonic() - start logger.info("%s finished in %i seconds." % (task.name, delta)) def _send_reply( self, reply_to: str, channel: amqp.Channel, result: Any, exc_info: ExcInfoType, ) -> None: logger.debug("Sending reply result=%r", result) reply = {'result': result} if exc_info: reply['exception'] = self._exc_info_dict(exc_info) try: body = json.dumps(reply) except Exception as e: logger.error('Cannot serialize result as JSON: %s', e) exc_info = sys.exc_info() reply = {'result': None, 'exception': self._exc_info_dict(exc_info)} body = json.dumps(reply) msg = amqp.Message(body=body) channel.basic_publish(msg, exchange="", routing_key=reply_to) @staticmethod def _exc_info_dict(exc_info: ExcInfoType) -> Dict[str, str]: type_, val, tb = exc_info return { 'type': '%s.%s' % (type_.__module__, type_.__name__), 'value': str(val), 'traceback': ''.join(traceback.format_tb(tb)), } def _watch_load(self) -> None: """Pause consuming messages if lood goes above the allowed limit.""" while not self.shutdown_pending.wait(1): self._current_load = os.getloadavg()[0] @property def uptime(self) -> float: if not self._started_at: return 0 return os.times().elapsed - self._started_at def _shutdown_timer(self) -> None: """Counts down from MAX_WORKER_RUN_TIME. When it reaches zero sutdown gracefully. """ remaining = self._max_run_time - self.uptime if not self.shutdown_pending.wait(remaining): logger.warning('Run time reached zero') self.shutdown() def shutdown(self) -> None: """Exits after the current task is finished.""" logger.warning("Shutdown requested") self.shutdown_pending.set() def _handle_sigint(self, signum: int, frame: Any) -> None: """Shutdown after processing current task.""" logger.warning("Catched SIGINT") self.shutdown() def _handle_sigterm(self, signum: int, frame: Any) -> None: """Shutdown after processing current task.""" logger.warning("Catched SIGTERM") self.shutdown() def _handle_sighup(self, signum: int, frame: Any) -> None: """Used internally to fail the task when connection to RabbitMQ is lost during the execution of the task. """ logger.warning("Catched SIGHUP") exc_info = self._heartbeat_exc_info self._heartbeat_exc_info = None # Format exception info to see in tools like Sentry. formatted_exception = ''.join(traceback.format_exception(*exc_info)) # noqa raise HeartbeatError(exc_info) @staticmethod def _handle_sigusr2(self, signum: int, frame: Any) -> None: """Drop current task.""" logger.warning("Catched SIGUSR2") if self.current_task: logger.warning("Dropping current task...") raise Discard def drop_task(self) -> None: os.kill(os.getpid(), signal.SIGUSR2)
cenkalti/kuyruk
kuyruk/worker.py
Worker._handle_sigusr2
python
def _handle_sigusr2(self, signum: int, frame: Any) -> None: logger.warning("Catched SIGUSR2") if self.current_task: logger.warning("Dropping current task...") raise Discard
Drop current task.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/worker.py#L395-L400
null
class Worker: """Consumes tasks from queues and runs them. :param app: An instance of :class:`~kuyruk.Kuyruk` :param args: Command line arguments """ def __init__(self, app: Kuyruk, args: argparse.Namespace) -> None: self.kuyruk = app if not args.queues: args.queues = ['kuyruk'] def add_host(queue: str) -> str: if queue.endswith('.localhost'): queue = queue.rsplit('.localhost')[0] return "%s.%s" % (queue, self._hostname) else: return queue self._hostname = socket.gethostname() self.queues = [add_host(q) for q in args.queues] self._tasks = {} # type: Dict[Tuple[str, str], Task] self.shutdown_pending = threading.Event() self.consuming = False self.current_task = None # type: Task self.current_args = None # type: Tuple self.current_kwargs = None # type: Dict[str, Any] self._started_at = None # type: float self._pid = os.getpid() self._logging_level = app.config.WORKER_LOGGING_LEVEL if args.logging_level is not None: self._logging_level = args.logging_level self._max_run_time = app.config.WORKER_MAX_RUN_TIME if args.max_run_time is not None: self._max_run_time = args.max_run_time self._max_load = app.config.WORKER_MAX_LOAD if args.max_load is not None: self._max_load = args.max_load if self._max_load == -1: self._max_load == multiprocessing.cpu_count() self._threads = [] # type: List[threading.Thread] if self._max_load: self._threads.append(threading.Thread(target=self._watch_load)) if self._max_run_time: self._threads.append(threading.Thread(target=self._shutdown_timer)) signals.worker_init.send(self.kuyruk, worker=self) def run(self) -> None: """Runs the worker and consumes messages from RabbitMQ. Returns only after `shutdown()` is called. """ if self._logging_level: logging.basicConfig( level=getattr(logging, self._logging_level.upper()), format="%(levelname).1s %(name)s.%(funcName)s:%(lineno)d - %(message)s") signal.signal(signal.SIGINT, self._handle_sigint) signal.signal(signal.SIGTERM, self._handle_sigterm) if platform.system() != 'Windows': # These features will not be available on Windows, but that is OK. # Read this issue for more details: # https://github.com/cenkalti/kuyruk/issues/54 signal.signal(signal.SIGHUP, self._handle_sighup) signal.signal(signal.SIGUSR1, self._handle_sigusr1) signal.signal(signal.SIGUSR2, self._handle_sigusr2) self._started_at = os.times().elapsed for t in self._threads: t.start() try: signals.worker_start.send(self.kuyruk, worker=self) self._consume_messages() signals.worker_shutdown.send(self.kuyruk, worker=self) finally: self.shutdown_pending.set() for t in self._threads: t.join() logger.debug("End run worker") def _consume_messages(self) -> None: with self.kuyruk.channel() as ch: # Set prefetch count to 1. If we don't set this, RabbitMQ keeps # sending messages while we are already working on a message. ch.basic_qos(0, 1, True) self._declare_queues(ch) self._consume_queues(ch) logger.info('Consumer started') self._main_loop(ch) def _main_loop(self, ch: amqp.Channel) -> None: while not self.shutdown_pending.is_set(): self._pause_or_resume(ch) try: ch.connection.heartbeat_tick() ch.connection.drain_events(timeout=1) except socket.timeout: pass def _consumer_tag(self, queue: str) -> str: return "%s:%s@%s" % (queue, self._pid, self._hostname) def _declare_queues(self, ch: amqp.Channel) -> None: for queue in self.queues: logger.debug("queue_declare: %s", queue) ch.queue_declare(queue=queue, durable=True, auto_delete=False) def _pause_or_resume(self, channel: amqp.Channel) -> None: if not self._max_load: return try: load = self._current_load except AttributeError: should_pause = False else: should_pause = load > self._max_load if should_pause and self.consuming: logger.warning('Load is above the treshold (%.2f/%s), ' 'pausing consumer', load, self._max_load) self._cancel_queues(channel) elif not should_pause and not self.consuming: logger.warning('Load is below the treshold (%.2f/%s), ' 'resuming consumer', load, self._max_load) self._consume_queues(channel) def _consume_queues(self, ch: amqp.Channel) -> None: self.consuming = True for queue in self.queues: logger.debug("basic_consume: %s", queue) ch.basic_consume(queue=queue, consumer_tag=self._consumer_tag(queue), callback=self._process_message) def _cancel_queues(self, ch: amqp.Channel) -> None: self.consuming = False for queue in self.queues: logger.debug("basic_cancel: %s", queue) ch.basic_cancel(self._consumer_tag(queue)) def _process_message(self, message: amqp.Message) -> None: """Processes the message received from the queue.""" if self.shutdown_pending.is_set(): return try: if isinstance(message.body, bytes): message.body = message.body.decode() description = json.loads(message.body) except Exception: logger.error("Cannot decode message. Dropping. Message: %r", message.body) traceback.print_exc() message.channel.basic_reject(message.delivery_tag, requeue=False) else: logger.info("Processing task: %r", description) self._process_description(message, description) def _process_description(self, message: amqp.Message, description: Dict[str, Any]) -> None: try: task = self._import_task(description['module'], description['function']) args, kwargs = description['args'], description['kwargs'] except Exception: logger.error('Cannot import task') exc_info = sys.exc_info() signals.worker_failure.send(self.kuyruk, description=description, exc_info=exc_info, worker=self) message.channel.basic_reject(message.delivery_tag, requeue=False) else: self._process_task(message, description, task, args, kwargs) def _import_task(self, module: str, function: str) -> Task: if (module, function) in self._tasks: return self._tasks[(module, function)] task = importer.import_object(module, function) self._tasks[(module, function)] = task return task def _process_task( self, message: amqp.Message, description: Dict[str, Any], task: Task, args: Tuple, kwargs: Dict[str, Any], ) -> None: queue = message.delivery_info['routing_key'] reply_to = message.properties.get('reply_to') try: result = self._run_task(message.channel.connection, task, args, kwargs) except Reject: logger.warning('Task is rejected') message.channel.basic_reject(message.delivery_tag, requeue=True) except Discard: logger.warning('Task is discarded') message.channel.basic_reject(message.delivery_tag, requeue=False) if reply_to: exc_info = sys.exc_info() self._send_reply(reply_to, message.channel, None, exc_info) except HeartbeatError as e: logger.error('Error while sending heartbeat') exc_info = e.exc_info logger.error(''.join(traceback.format_exception(*exc_info))) signals.worker_failure.send( self.kuyruk, description=description, task=task, args=args, kwargs=kwargs, exc_info=exc_info, worker=self, queue=queue) raise except Exception: logger.error('Task raised an exception') exc_info = sys.exc_info() logger.error(''.join(traceback.format_exception(*exc_info))) signals.worker_failure.send( self.kuyruk, description=description, task=task, args=args, kwargs=kwargs, exc_info=exc_info, worker=self, queue=queue) message.channel.basic_reject(message.delivery_tag, requeue=False) if reply_to: self._send_reply(reply_to, message.channel, None, exc_info) else: logger.info('Task is successful') message.channel.basic_ack(message.delivery_tag) if reply_to: self._send_reply(reply_to, message.channel, result, None) finally: logger.debug("Task is processed") def _run_task(self, connection: amqp.Connection, task: Task, args: Tuple, kwargs: Dict[str, Any]) -> Any: hb = Heartbeat(connection, self._on_heartbeat_error) hb.start() self.current_task = task self.current_args = args self.current_kwargs = kwargs try: return self._apply_task(task, args, kwargs) finally: self.current_task = None self.current_args = None self.current_kwargs = None hb.stop() def _on_heartbeat_error(self, exc_info: ExcInfoType) -> None: self._heartbeat_exc_info = exc_info os.kill(os.getpid(), signal.SIGHUP) @staticmethod def _apply_task(task: Task, args: Tuple, kwargs: Dict[str, Any]) -> Any: """Logs the time spent while running the task.""" if args is None: args = () if kwargs is None: kwargs = {} start = monotonic() try: return task.apply(*args, **kwargs) finally: delta = monotonic() - start logger.info("%s finished in %i seconds." % (task.name, delta)) def _send_reply( self, reply_to: str, channel: amqp.Channel, result: Any, exc_info: ExcInfoType, ) -> None: logger.debug("Sending reply result=%r", result) reply = {'result': result} if exc_info: reply['exception'] = self._exc_info_dict(exc_info) try: body = json.dumps(reply) except Exception as e: logger.error('Cannot serialize result as JSON: %s', e) exc_info = sys.exc_info() reply = {'result': None, 'exception': self._exc_info_dict(exc_info)} body = json.dumps(reply) msg = amqp.Message(body=body) channel.basic_publish(msg, exchange="", routing_key=reply_to) @staticmethod def _exc_info_dict(exc_info: ExcInfoType) -> Dict[str, str]: type_, val, tb = exc_info return { 'type': '%s.%s' % (type_.__module__, type_.__name__), 'value': str(val), 'traceback': ''.join(traceback.format_tb(tb)), } def _watch_load(self) -> None: """Pause consuming messages if lood goes above the allowed limit.""" while not self.shutdown_pending.wait(1): self._current_load = os.getloadavg()[0] @property def uptime(self) -> float: if not self._started_at: return 0 return os.times().elapsed - self._started_at def _shutdown_timer(self) -> None: """Counts down from MAX_WORKER_RUN_TIME. When it reaches zero sutdown gracefully. """ remaining = self._max_run_time - self.uptime if not self.shutdown_pending.wait(remaining): logger.warning('Run time reached zero') self.shutdown() def shutdown(self) -> None: """Exits after the current task is finished.""" logger.warning("Shutdown requested") self.shutdown_pending.set() def _handle_sigint(self, signum: int, frame: Any) -> None: """Shutdown after processing current task.""" logger.warning("Catched SIGINT") self.shutdown() def _handle_sigterm(self, signum: int, frame: Any) -> None: """Shutdown after processing current task.""" logger.warning("Catched SIGTERM") self.shutdown() def _handle_sighup(self, signum: int, frame: Any) -> None: """Used internally to fail the task when connection to RabbitMQ is lost during the execution of the task. """ logger.warning("Catched SIGHUP") exc_info = self._heartbeat_exc_info self._heartbeat_exc_info = None # Format exception info to see in tools like Sentry. formatted_exception = ''.join(traceback.format_exception(*exc_info)) # noqa raise HeartbeatError(exc_info) @staticmethod def _handle_sigusr1(signum: int, frame: Any) -> None: """Print stacktrace.""" print('=' * 70) print(''.join(traceback.format_stack())) print('-' * 70) def drop_task(self) -> None: os.kill(os.getpid(), signal.SIGUSR2)
cenkalti/kuyruk
kuyruk/task.py
Task.send_to_queue
python
def send_to_queue( self, args: Tuple=(), kwargs: Dict[str, Any]={}, host: str=None, wait_result: Union[int, float]=None, message_ttl: Union[int, float]=None, ) -> Any: if self.kuyruk.config.EAGER: # Run the task in current process result = self.apply(*args, **kwargs) return result if wait_result else None logger.debug("Task.send_to_queue args=%r, kwargs=%r", args, kwargs) queue = self._queue_for_host(host) description = self._get_description(args, kwargs) self._send_signal(signals.task_presend, args=args, kwargs=kwargs, description=description) body = json.dumps(description) msg = amqp.Message(body=body) if wait_result: # Use direct reply-to feature from RabbitMQ: # https://www.rabbitmq.com/direct-reply-to.html msg.properties['reply_to'] = 'amq.rabbitmq.reply-to' if message_ttl: msg.properties['expiration'] = str(int(message_ttl * 1000)) with self.kuyruk.channel() as ch: if wait_result: result = Result(ch.connection) ch.basic_consume(queue='amq.rabbitmq.reply-to', no_ack=True, callback=result.process_message) ch.queue_declare(queue=queue, durable=True, auto_delete=False) ch.basic_publish(msg, exchange="", routing_key=queue) self._send_signal(signals.task_postsend, args=args, kwargs=kwargs, description=description) if wait_result: return result.wait(wait_result)
Sends a message to the queue. A worker will run the task's function when it receives the message. :param args: Arguments that will be passed to task on execution. :param kwargs: Keyword arguments that will be passed to task on execution. :param host: Send this task to specific host. ``host`` will be appended to the queue name. If ``host`` is "localhost", hostname of the server will be appended to the queue name. :param wait_result: Wait for result from worker for ``wait_result`` seconds. If timeout occurs, :class:`~kuyruk.exceptions.ResultTimeout` is raised. If excecption occurs in worker, :class:`~kuyruk.exceptions.RemoteException` is raised. :param message_ttl: If set, message will be destroyed in queue after ``message_ttl`` seconds. :return: Result from worker if ``wait_result`` is set, else :const:`None`.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/task.py#L69-L130
[ "def wait(self, timeout: Union[float, int]) -> None:\n logger.debug(\"Waiting for task result\")\n\n start = monotonic()\n while True:\n if self.exception:\n raise RemoteException(self.exception['type'],\n self.exception['value'],\n self.exception['traceback'])\n if self.result:\n return self.result\n\n if monotonic() - start > timeout:\n raise ResultTimeout\n\n try:\n self._connection.heartbeat_tick()\n self._connection.drain_events(timeout=1)\n except socket.timeout:\n pass\n", "def _queue_for_host(self, host: str) -> str:\n if not host:\n return self.queue\n if host == 'localhost':\n host = socket.gethostname()\n return \"%s.%s\" % (self.queue, host)\n", "def _get_description(self, args: Tuple, kwargs: Dict[str, Any]) -> Dict[str, Any]:\n \"\"\"Return the dictionary to be sent to the queue.\"\"\"\n return {\n 'id': uuid1().hex,\n 'args': args,\n 'kwargs': kwargs,\n 'module': self._module_name,\n 'function': self.f.__name__,\n 'sender_hostname': socket.gethostname(),\n 'sender_pid': os.getpid(),\n 'sender_cmd': ' '.join(sys.argv),\n 'sender_timestamp': datetime.utcnow().isoformat()[:19],\n }\n", "def _send_signal(self, sig: Signal, **data: Any) -> None:\n sig.send(self.kuyruk, task=self, **data)\n", "def apply(self, *args: Any, **kwargs: Any) -> Any:\n \"\"\"Called by workers to run the wrapped function.\n You may call it yourself if you want to run the task in current process\n without sending to the queue.\n\n If task has a `retry` property it will be retried on failure.\n\n If task has a `max_run_time` property the task will not be allowed to\n run more than that.\n \"\"\"\n def send_signal(sig: Signal, **extra: Any) -> None:\n self._send_signal(sig, args=args, kwargs=kwargs, **extra)\n\n logger.debug(\"Applying %r, args=%r, kwargs=%r\", self, args, kwargs)\n\n send_signal(signals.task_preapply)\n try:\n tries = 1 + self.retry\n while 1:\n tries -= 1\n send_signal(signals.task_prerun)\n try:\n with time_limit(self.max_run_time or 0):\n return self.f(*args, **kwargs)\n except Exception:\n send_signal(signals.task_error, exc_info=sys.exc_info())\n if tries <= 0:\n raise\n else:\n break\n finally:\n send_signal(signals.task_postrun)\n except Exception:\n send_signal(signals.task_failure, exc_info=sys.exc_info())\n raise\n else:\n send_signal(signals.task_success)\n finally:\n send_signal(signals.task_postapply)\n" ]
class Task: """Calling a :class:`~kuyruk.Task` object serializes the task to JSON and sends it to the queue. :param retry: Retry this times before give up. The failed task will be retried in the same worker. :param max_run_time: Maximum allowed time in seconds for task to complete. """ def __init__(self, f: Callable, kuyruk: 'Kuyruk', queue: str, retry: int=0, max_run_time: int=None) -> None: self.f = f self.kuyruk = kuyruk self.queue = queue self.retry = retry self.max_run_time = max_run_time self._send_signal(signals.task_init) def __repr__(self) -> str: return "<Task of %r>" % self.name def __call__(self, *args: Tuple, **kwargs: Any) -> None: """When a function is wrapped with a task decorator it will be converted to a Task object. By overriding __call__ method we are sending this task to queue instead of invoking the function without changing the client code. """ logger.debug("Task.__call__ args=%r, kwargs=%r", args, kwargs) self.send_to_queue(args, kwargs) def subtask(self, args: Tuple=(), kwargs: Dict[str, Any]={}, host: str=None) -> SubTask: return SubTask(self, args, kwargs, host) def _queue_for_host(self, host: str) -> str: if not host: return self.queue if host == 'localhost': host = socket.gethostname() return "%s.%s" % (self.queue, host) def _get_description(self, args: Tuple, kwargs: Dict[str, Any]) -> Dict[str, Any]: """Return the dictionary to be sent to the queue.""" return { 'id': uuid1().hex, 'args': args, 'kwargs': kwargs, 'module': self._module_name, 'function': self.f.__name__, 'sender_hostname': socket.gethostname(), 'sender_pid': os.getpid(), 'sender_cmd': ' '.join(sys.argv), 'sender_timestamp': datetime.utcnow().isoformat()[:19], } def _send_signal(self, sig: Signal, **data: Any) -> None: sig.send(self.kuyruk, task=self, **data) def apply(self, *args: Any, **kwargs: Any) -> Any: """Called by workers to run the wrapped function. You may call it yourself if you want to run the task in current process without sending to the queue. If task has a `retry` property it will be retried on failure. If task has a `max_run_time` property the task will not be allowed to run more than that. """ def send_signal(sig: Signal, **extra: Any) -> None: self._send_signal(sig, args=args, kwargs=kwargs, **extra) logger.debug("Applying %r, args=%r, kwargs=%r", self, args, kwargs) send_signal(signals.task_preapply) try: tries = 1 + self.retry while 1: tries -= 1 send_signal(signals.task_prerun) try: with time_limit(self.max_run_time or 0): return self.f(*args, **kwargs) except Exception: send_signal(signals.task_error, exc_info=sys.exc_info()) if tries <= 0: raise else: break finally: send_signal(signals.task_postrun) except Exception: send_signal(signals.task_failure, exc_info=sys.exc_info()) raise else: send_signal(signals.task_success) finally: send_signal(signals.task_postapply) @property def name(self) -> str: """Full path to the task in the form of `<module>.<function>`. Workers find and import tasks by this path. """ return "%s:%s" % (self._module_name, self.f.__name__) @property def _module_name(self) -> str: """Module name of the wrapped function.""" name = self.f.__module__ if name == '__main__': return importer.main_module_name() return name
cenkalti/kuyruk
kuyruk/task.py
Task._get_description
python
def _get_description(self, args: Tuple, kwargs: Dict[str, Any]) -> Dict[str, Any]: return { 'id': uuid1().hex, 'args': args, 'kwargs': kwargs, 'module': self._module_name, 'function': self.f.__name__, 'sender_hostname': socket.gethostname(), 'sender_pid': os.getpid(), 'sender_cmd': ' '.join(sys.argv), 'sender_timestamp': datetime.utcnow().isoformat()[:19], }
Return the dictionary to be sent to the queue.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/task.py#L139-L151
null
class Task: """Calling a :class:`~kuyruk.Task` object serializes the task to JSON and sends it to the queue. :param retry: Retry this times before give up. The failed task will be retried in the same worker. :param max_run_time: Maximum allowed time in seconds for task to complete. """ def __init__(self, f: Callable, kuyruk: 'Kuyruk', queue: str, retry: int=0, max_run_time: int=None) -> None: self.f = f self.kuyruk = kuyruk self.queue = queue self.retry = retry self.max_run_time = max_run_time self._send_signal(signals.task_init) def __repr__(self) -> str: return "<Task of %r>" % self.name def __call__(self, *args: Tuple, **kwargs: Any) -> None: """When a function is wrapped with a task decorator it will be converted to a Task object. By overriding __call__ method we are sending this task to queue instead of invoking the function without changing the client code. """ logger.debug("Task.__call__ args=%r, kwargs=%r", args, kwargs) self.send_to_queue(args, kwargs) def subtask(self, args: Tuple=(), kwargs: Dict[str, Any]={}, host: str=None) -> SubTask: return SubTask(self, args, kwargs, host) def send_to_queue( self, args: Tuple=(), kwargs: Dict[str, Any]={}, host: str=None, wait_result: Union[int, float]=None, message_ttl: Union[int, float]=None, ) -> Any: """ Sends a message to the queue. A worker will run the task's function when it receives the message. :param args: Arguments that will be passed to task on execution. :param kwargs: Keyword arguments that will be passed to task on execution. :param host: Send this task to specific host. ``host`` will be appended to the queue name. If ``host`` is "localhost", hostname of the server will be appended to the queue name. :param wait_result: Wait for result from worker for ``wait_result`` seconds. If timeout occurs, :class:`~kuyruk.exceptions.ResultTimeout` is raised. If excecption occurs in worker, :class:`~kuyruk.exceptions.RemoteException` is raised. :param message_ttl: If set, message will be destroyed in queue after ``message_ttl`` seconds. :return: Result from worker if ``wait_result`` is set, else :const:`None`. """ if self.kuyruk.config.EAGER: # Run the task in current process result = self.apply(*args, **kwargs) return result if wait_result else None logger.debug("Task.send_to_queue args=%r, kwargs=%r", args, kwargs) queue = self._queue_for_host(host) description = self._get_description(args, kwargs) self._send_signal(signals.task_presend, args=args, kwargs=kwargs, description=description) body = json.dumps(description) msg = amqp.Message(body=body) if wait_result: # Use direct reply-to feature from RabbitMQ: # https://www.rabbitmq.com/direct-reply-to.html msg.properties['reply_to'] = 'amq.rabbitmq.reply-to' if message_ttl: msg.properties['expiration'] = str(int(message_ttl * 1000)) with self.kuyruk.channel() as ch: if wait_result: result = Result(ch.connection) ch.basic_consume(queue='amq.rabbitmq.reply-to', no_ack=True, callback=result.process_message) ch.queue_declare(queue=queue, durable=True, auto_delete=False) ch.basic_publish(msg, exchange="", routing_key=queue) self._send_signal(signals.task_postsend, args=args, kwargs=kwargs, description=description) if wait_result: return result.wait(wait_result) def _queue_for_host(self, host: str) -> str: if not host: return self.queue if host == 'localhost': host = socket.gethostname() return "%s.%s" % (self.queue, host) def _send_signal(self, sig: Signal, **data: Any) -> None: sig.send(self.kuyruk, task=self, **data) def apply(self, *args: Any, **kwargs: Any) -> Any: """Called by workers to run the wrapped function. You may call it yourself if you want to run the task in current process without sending to the queue. If task has a `retry` property it will be retried on failure. If task has a `max_run_time` property the task will not be allowed to run more than that. """ def send_signal(sig: Signal, **extra: Any) -> None: self._send_signal(sig, args=args, kwargs=kwargs, **extra) logger.debug("Applying %r, args=%r, kwargs=%r", self, args, kwargs) send_signal(signals.task_preapply) try: tries = 1 + self.retry while 1: tries -= 1 send_signal(signals.task_prerun) try: with time_limit(self.max_run_time or 0): return self.f(*args, **kwargs) except Exception: send_signal(signals.task_error, exc_info=sys.exc_info()) if tries <= 0: raise else: break finally: send_signal(signals.task_postrun) except Exception: send_signal(signals.task_failure, exc_info=sys.exc_info()) raise else: send_signal(signals.task_success) finally: send_signal(signals.task_postapply) @property def name(self) -> str: """Full path to the task in the form of `<module>.<function>`. Workers find and import tasks by this path. """ return "%s:%s" % (self._module_name, self.f.__name__) @property def _module_name(self) -> str: """Module name of the wrapped function.""" name = self.f.__module__ if name == '__main__': return importer.main_module_name() return name
cenkalti/kuyruk
kuyruk/task.py
Task.apply
python
def apply(self, *args: Any, **kwargs: Any) -> Any: def send_signal(sig: Signal, **extra: Any) -> None: self._send_signal(sig, args=args, kwargs=kwargs, **extra) logger.debug("Applying %r, args=%r, kwargs=%r", self, args, kwargs) send_signal(signals.task_preapply) try: tries = 1 + self.retry while 1: tries -= 1 send_signal(signals.task_prerun) try: with time_limit(self.max_run_time or 0): return self.f(*args, **kwargs) except Exception: send_signal(signals.task_error, exc_info=sys.exc_info()) if tries <= 0: raise else: break finally: send_signal(signals.task_postrun) except Exception: send_signal(signals.task_failure, exc_info=sys.exc_info()) raise else: send_signal(signals.task_success) finally: send_signal(signals.task_postapply)
Called by workers to run the wrapped function. You may call it yourself if you want to run the task in current process without sending to the queue. If task has a `retry` property it will be retried on failure. If task has a `max_run_time` property the task will not be allowed to run more than that.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/task.py#L156-L194
[ "def send_signal(sig: Signal, **extra: Any) -> None:\n self._send_signal(sig, args=args, kwargs=kwargs, **extra)\n" ]
class Task: """Calling a :class:`~kuyruk.Task` object serializes the task to JSON and sends it to the queue. :param retry: Retry this times before give up. The failed task will be retried in the same worker. :param max_run_time: Maximum allowed time in seconds for task to complete. """ def __init__(self, f: Callable, kuyruk: 'Kuyruk', queue: str, retry: int=0, max_run_time: int=None) -> None: self.f = f self.kuyruk = kuyruk self.queue = queue self.retry = retry self.max_run_time = max_run_time self._send_signal(signals.task_init) def __repr__(self) -> str: return "<Task of %r>" % self.name def __call__(self, *args: Tuple, **kwargs: Any) -> None: """When a function is wrapped with a task decorator it will be converted to a Task object. By overriding __call__ method we are sending this task to queue instead of invoking the function without changing the client code. """ logger.debug("Task.__call__ args=%r, kwargs=%r", args, kwargs) self.send_to_queue(args, kwargs) def subtask(self, args: Tuple=(), kwargs: Dict[str, Any]={}, host: str=None) -> SubTask: return SubTask(self, args, kwargs, host) def send_to_queue( self, args: Tuple=(), kwargs: Dict[str, Any]={}, host: str=None, wait_result: Union[int, float]=None, message_ttl: Union[int, float]=None, ) -> Any: """ Sends a message to the queue. A worker will run the task's function when it receives the message. :param args: Arguments that will be passed to task on execution. :param kwargs: Keyword arguments that will be passed to task on execution. :param host: Send this task to specific host. ``host`` will be appended to the queue name. If ``host`` is "localhost", hostname of the server will be appended to the queue name. :param wait_result: Wait for result from worker for ``wait_result`` seconds. If timeout occurs, :class:`~kuyruk.exceptions.ResultTimeout` is raised. If excecption occurs in worker, :class:`~kuyruk.exceptions.RemoteException` is raised. :param message_ttl: If set, message will be destroyed in queue after ``message_ttl`` seconds. :return: Result from worker if ``wait_result`` is set, else :const:`None`. """ if self.kuyruk.config.EAGER: # Run the task in current process result = self.apply(*args, **kwargs) return result if wait_result else None logger.debug("Task.send_to_queue args=%r, kwargs=%r", args, kwargs) queue = self._queue_for_host(host) description = self._get_description(args, kwargs) self._send_signal(signals.task_presend, args=args, kwargs=kwargs, description=description) body = json.dumps(description) msg = amqp.Message(body=body) if wait_result: # Use direct reply-to feature from RabbitMQ: # https://www.rabbitmq.com/direct-reply-to.html msg.properties['reply_to'] = 'amq.rabbitmq.reply-to' if message_ttl: msg.properties['expiration'] = str(int(message_ttl * 1000)) with self.kuyruk.channel() as ch: if wait_result: result = Result(ch.connection) ch.basic_consume(queue='amq.rabbitmq.reply-to', no_ack=True, callback=result.process_message) ch.queue_declare(queue=queue, durable=True, auto_delete=False) ch.basic_publish(msg, exchange="", routing_key=queue) self._send_signal(signals.task_postsend, args=args, kwargs=kwargs, description=description) if wait_result: return result.wait(wait_result) def _queue_for_host(self, host: str) -> str: if not host: return self.queue if host == 'localhost': host = socket.gethostname() return "%s.%s" % (self.queue, host) def _get_description(self, args: Tuple, kwargs: Dict[str, Any]) -> Dict[str, Any]: """Return the dictionary to be sent to the queue.""" return { 'id': uuid1().hex, 'args': args, 'kwargs': kwargs, 'module': self._module_name, 'function': self.f.__name__, 'sender_hostname': socket.gethostname(), 'sender_pid': os.getpid(), 'sender_cmd': ' '.join(sys.argv), 'sender_timestamp': datetime.utcnow().isoformat()[:19], } def _send_signal(self, sig: Signal, **data: Any) -> None: sig.send(self.kuyruk, task=self, **data) @property def name(self) -> str: """Full path to the task in the form of `<module>.<function>`. Workers find and import tasks by this path. """ return "%s:%s" % (self._module_name, self.f.__name__) @property def _module_name(self) -> str: """Module name of the wrapped function.""" name = self.f.__module__ if name == '__main__': return importer.main_module_name() return name
cenkalti/kuyruk
kuyruk/task.py
Task._module_name
python
def _module_name(self) -> str: name = self.f.__module__ if name == '__main__': return importer.main_module_name() return name
Module name of the wrapped function.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/task.py#L205-L210
null
class Task: """Calling a :class:`~kuyruk.Task` object serializes the task to JSON and sends it to the queue. :param retry: Retry this times before give up. The failed task will be retried in the same worker. :param max_run_time: Maximum allowed time in seconds for task to complete. """ def __init__(self, f: Callable, kuyruk: 'Kuyruk', queue: str, retry: int=0, max_run_time: int=None) -> None: self.f = f self.kuyruk = kuyruk self.queue = queue self.retry = retry self.max_run_time = max_run_time self._send_signal(signals.task_init) def __repr__(self) -> str: return "<Task of %r>" % self.name def __call__(self, *args: Tuple, **kwargs: Any) -> None: """When a function is wrapped with a task decorator it will be converted to a Task object. By overriding __call__ method we are sending this task to queue instead of invoking the function without changing the client code. """ logger.debug("Task.__call__ args=%r, kwargs=%r", args, kwargs) self.send_to_queue(args, kwargs) def subtask(self, args: Tuple=(), kwargs: Dict[str, Any]={}, host: str=None) -> SubTask: return SubTask(self, args, kwargs, host) def send_to_queue( self, args: Tuple=(), kwargs: Dict[str, Any]={}, host: str=None, wait_result: Union[int, float]=None, message_ttl: Union[int, float]=None, ) -> Any: """ Sends a message to the queue. A worker will run the task's function when it receives the message. :param args: Arguments that will be passed to task on execution. :param kwargs: Keyword arguments that will be passed to task on execution. :param host: Send this task to specific host. ``host`` will be appended to the queue name. If ``host`` is "localhost", hostname of the server will be appended to the queue name. :param wait_result: Wait for result from worker for ``wait_result`` seconds. If timeout occurs, :class:`~kuyruk.exceptions.ResultTimeout` is raised. If excecption occurs in worker, :class:`~kuyruk.exceptions.RemoteException` is raised. :param message_ttl: If set, message will be destroyed in queue after ``message_ttl`` seconds. :return: Result from worker if ``wait_result`` is set, else :const:`None`. """ if self.kuyruk.config.EAGER: # Run the task in current process result = self.apply(*args, **kwargs) return result if wait_result else None logger.debug("Task.send_to_queue args=%r, kwargs=%r", args, kwargs) queue = self._queue_for_host(host) description = self._get_description(args, kwargs) self._send_signal(signals.task_presend, args=args, kwargs=kwargs, description=description) body = json.dumps(description) msg = amqp.Message(body=body) if wait_result: # Use direct reply-to feature from RabbitMQ: # https://www.rabbitmq.com/direct-reply-to.html msg.properties['reply_to'] = 'amq.rabbitmq.reply-to' if message_ttl: msg.properties['expiration'] = str(int(message_ttl * 1000)) with self.kuyruk.channel() as ch: if wait_result: result = Result(ch.connection) ch.basic_consume(queue='amq.rabbitmq.reply-to', no_ack=True, callback=result.process_message) ch.queue_declare(queue=queue, durable=True, auto_delete=False) ch.basic_publish(msg, exchange="", routing_key=queue) self._send_signal(signals.task_postsend, args=args, kwargs=kwargs, description=description) if wait_result: return result.wait(wait_result) def _queue_for_host(self, host: str) -> str: if not host: return self.queue if host == 'localhost': host = socket.gethostname() return "%s.%s" % (self.queue, host) def _get_description(self, args: Tuple, kwargs: Dict[str, Any]) -> Dict[str, Any]: """Return the dictionary to be sent to the queue.""" return { 'id': uuid1().hex, 'args': args, 'kwargs': kwargs, 'module': self._module_name, 'function': self.f.__name__, 'sender_hostname': socket.gethostname(), 'sender_pid': os.getpid(), 'sender_cmd': ' '.join(sys.argv), 'sender_timestamp': datetime.utcnow().isoformat()[:19], } def _send_signal(self, sig: Signal, **data: Any) -> None: sig.send(self.kuyruk, task=self, **data) def apply(self, *args: Any, **kwargs: Any) -> Any: """Called by workers to run the wrapped function. You may call it yourself if you want to run the task in current process without sending to the queue. If task has a `retry` property it will be retried on failure. If task has a `max_run_time` property the task will not be allowed to run more than that. """ def send_signal(sig: Signal, **extra: Any) -> None: self._send_signal(sig, args=args, kwargs=kwargs, **extra) logger.debug("Applying %r, args=%r, kwargs=%r", self, args, kwargs) send_signal(signals.task_preapply) try: tries = 1 + self.retry while 1: tries -= 1 send_signal(signals.task_prerun) try: with time_limit(self.max_run_time or 0): return self.f(*args, **kwargs) except Exception: send_signal(signals.task_error, exc_info=sys.exc_info()) if tries <= 0: raise else: break finally: send_signal(signals.task_postrun) except Exception: send_signal(signals.task_failure, exc_info=sys.exc_info()) raise else: send_signal(signals.task_success) finally: send_signal(signals.task_postapply) @property def name(self) -> str: """Full path to the task in the form of `<module>.<function>`. Workers find and import tasks by this path. """ return "%s:%s" % (self._module_name, self.f.__name__) @property
cenkalti/kuyruk
kuyruk/config.py
Config.from_object
python
def from_object(self, obj: Union[str, Any]) -> None: if isinstance(obj, str): obj = importer.import_object_str(obj) for key in dir(obj): if key.isupper(): value = getattr(obj, key) self._setattr(key, value) logger.info("Config is loaded from object: %r", obj)
Load values from an object.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/config.py#L62-L72
[ "def import_object_str(s: str) -> Any:\n module, obj = s.rsplit('.', 1)\n return import_object(module, obj)\n", "def _setattr(self, key: str, value: Any) -> None:\n if not hasattr(self.__class__, key):\n raise ValueError(\"Unknown config key: %s\" % key)\n\n setattr(self, key, value)\n" ]
class Config: """Kuyruk configuration object. Default values are defined as class attributes. Additional attributes may be added by extensions. """ # Connection Options #################### RABBIT_HOST = 'localhost' """RabbitMQ host.""" RABBIT_PORT = 5672 """RabbitMQ port.""" RABBIT_VIRTUAL_HOST = '/' """RabbitMQ virtual host.""" RABBIT_USER = 'guest' """RabbitMQ user.""" RABBIT_PASSWORD = 'guest' """RabbitMQ password.""" RABBIT_HEARTBEAT = 60 RABBIT_CONNECT_TIMEOUT = 5 RABBIT_READ_TIMEOUT = 5 RABBIT_WRITE_TIMEOUT = 5 TCP_USER_TIMEOUT = 60 # Instance Options ################## EAGER = False """Run tasks in the process without sending to queue. Useful in tests.""" # Worker Options ################ WORKER_MAX_LOAD = None """Pause consuming queue when the load goes above this level.""" WORKER_MAX_RUN_TIME = None """Gracefully shutdown worker after running this seconds.""" WORKER_LOGGING_LEVEL = 'INFO' """Logging level of root logger.""" def from_dict(self, d: Dict[str, Any]) -> None: """Load values from a dict.""" for key, value in d.items(): if key.isupper(): self._setattr(key, value) logger.info("Config is loaded from dict: %r", d) def from_pymodule(self, name: str) -> None: module = importer.import_module(name) for key, value in module.__dict__.items(): if (key.isupper() and not isinstance(value, types.ModuleType)): self._setattr(key, value) logger.info("Config is loaded from module: %s", name) def from_pyfile(self, filename: str) -> None: """Load values from a Python file.""" globals_ = {} # type: Dict[str, Any] locals_ = {} # type: Dict[str, Any] with open(filename, "rb") as f: exec(compile(f.read(), filename, 'exec'), globals_, locals_) for key, value in locals_.items(): if (key.isupper() and not isinstance(value, types.ModuleType)): self._setattr(key, value) logger.info("Config is loaded from file: %s", filename) def from_env_vars(self) -> None: """Load values from environment variables. Keys must start with `KUYRUK_`.""" for key, value in os.environ.items(): if key.startswith('KUYRUK_'): key = key[7:] if hasattr(Config, key): try: value = ast.literal_eval(value) except (ValueError, SyntaxError): pass self._setattr(key, value) def _setattr(self, key: str, value: Any) -> None: if not hasattr(self.__class__, key): raise ValueError("Unknown config key: %s" % key) setattr(self, key, value)
cenkalti/kuyruk
kuyruk/config.py
Config.from_dict
python
def from_dict(self, d: Dict[str, Any]) -> None: for key, value in d.items(): if key.isupper(): self._setattr(key, value) logger.info("Config is loaded from dict: %r", d)
Load values from a dict.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/config.py#L74-L80
[ "def _setattr(self, key: str, value: Any) -> None:\n if not hasattr(self.__class__, key):\n raise ValueError(\"Unknown config key: %s\" % key)\n\n setattr(self, key, value)\n" ]
class Config: """Kuyruk configuration object. Default values are defined as class attributes. Additional attributes may be added by extensions. """ # Connection Options #################### RABBIT_HOST = 'localhost' """RabbitMQ host.""" RABBIT_PORT = 5672 """RabbitMQ port.""" RABBIT_VIRTUAL_HOST = '/' """RabbitMQ virtual host.""" RABBIT_USER = 'guest' """RabbitMQ user.""" RABBIT_PASSWORD = 'guest' """RabbitMQ password.""" RABBIT_HEARTBEAT = 60 RABBIT_CONNECT_TIMEOUT = 5 RABBIT_READ_TIMEOUT = 5 RABBIT_WRITE_TIMEOUT = 5 TCP_USER_TIMEOUT = 60 # Instance Options ################## EAGER = False """Run tasks in the process without sending to queue. Useful in tests.""" # Worker Options ################ WORKER_MAX_LOAD = None """Pause consuming queue when the load goes above this level.""" WORKER_MAX_RUN_TIME = None """Gracefully shutdown worker after running this seconds.""" WORKER_LOGGING_LEVEL = 'INFO' """Logging level of root logger.""" def from_object(self, obj: Union[str, Any]) -> None: """Load values from an object.""" if isinstance(obj, str): obj = importer.import_object_str(obj) for key in dir(obj): if key.isupper(): value = getattr(obj, key) self._setattr(key, value) logger.info("Config is loaded from object: %r", obj) def from_pymodule(self, name: str) -> None: module = importer.import_module(name) for key, value in module.__dict__.items(): if (key.isupper() and not isinstance(value, types.ModuleType)): self._setattr(key, value) logger.info("Config is loaded from module: %s", name) def from_pyfile(self, filename: str) -> None: """Load values from a Python file.""" globals_ = {} # type: Dict[str, Any] locals_ = {} # type: Dict[str, Any] with open(filename, "rb") as f: exec(compile(f.read(), filename, 'exec'), globals_, locals_) for key, value in locals_.items(): if (key.isupper() and not isinstance(value, types.ModuleType)): self._setattr(key, value) logger.info("Config is loaded from file: %s", filename) def from_env_vars(self) -> None: """Load values from environment variables. Keys must start with `KUYRUK_`.""" for key, value in os.environ.items(): if key.startswith('KUYRUK_'): key = key[7:] if hasattr(Config, key): try: value = ast.literal_eval(value) except (ValueError, SyntaxError): pass self._setattr(key, value) def _setattr(self, key: str, value: Any) -> None: if not hasattr(self.__class__, key): raise ValueError("Unknown config key: %s" % key) setattr(self, key, value)
cenkalti/kuyruk
kuyruk/config.py
Config.from_pyfile
python
def from_pyfile(self, filename: str) -> None: globals_ = {} # type: Dict[str, Any] locals_ = {} # type: Dict[str, Any] with open(filename, "rb") as f: exec(compile(f.read(), filename, 'exec'), globals_, locals_) for key, value in locals_.items(): if (key.isupper() and not isinstance(value, types.ModuleType)): self._setattr(key, value) logger.info("Config is loaded from file: %s", filename)
Load values from a Python file.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/config.py#L90-L101
[ "def _setattr(self, key: str, value: Any) -> None:\n if not hasattr(self.__class__, key):\n raise ValueError(\"Unknown config key: %s\" % key)\n\n setattr(self, key, value)\n" ]
class Config: """Kuyruk configuration object. Default values are defined as class attributes. Additional attributes may be added by extensions. """ # Connection Options #################### RABBIT_HOST = 'localhost' """RabbitMQ host.""" RABBIT_PORT = 5672 """RabbitMQ port.""" RABBIT_VIRTUAL_HOST = '/' """RabbitMQ virtual host.""" RABBIT_USER = 'guest' """RabbitMQ user.""" RABBIT_PASSWORD = 'guest' """RabbitMQ password.""" RABBIT_HEARTBEAT = 60 RABBIT_CONNECT_TIMEOUT = 5 RABBIT_READ_TIMEOUT = 5 RABBIT_WRITE_TIMEOUT = 5 TCP_USER_TIMEOUT = 60 # Instance Options ################## EAGER = False """Run tasks in the process without sending to queue. Useful in tests.""" # Worker Options ################ WORKER_MAX_LOAD = None """Pause consuming queue when the load goes above this level.""" WORKER_MAX_RUN_TIME = None """Gracefully shutdown worker after running this seconds.""" WORKER_LOGGING_LEVEL = 'INFO' """Logging level of root logger.""" def from_object(self, obj: Union[str, Any]) -> None: """Load values from an object.""" if isinstance(obj, str): obj = importer.import_object_str(obj) for key in dir(obj): if key.isupper(): value = getattr(obj, key) self._setattr(key, value) logger.info("Config is loaded from object: %r", obj) def from_dict(self, d: Dict[str, Any]) -> None: """Load values from a dict.""" for key, value in d.items(): if key.isupper(): self._setattr(key, value) logger.info("Config is loaded from dict: %r", d) def from_pymodule(self, name: str) -> None: module = importer.import_module(name) for key, value in module.__dict__.items(): if (key.isupper() and not isinstance(value, types.ModuleType)): self._setattr(key, value) logger.info("Config is loaded from module: %s", name) def from_env_vars(self) -> None: """Load values from environment variables. Keys must start with `KUYRUK_`.""" for key, value in os.environ.items(): if key.startswith('KUYRUK_'): key = key[7:] if hasattr(Config, key): try: value = ast.literal_eval(value) except (ValueError, SyntaxError): pass self._setattr(key, value) def _setattr(self, key: str, value: Any) -> None: if not hasattr(self.__class__, key): raise ValueError("Unknown config key: %s" % key) setattr(self, key, value)
cenkalti/kuyruk
kuyruk/config.py
Config.from_env_vars
python
def from_env_vars(self) -> None: for key, value in os.environ.items(): if key.startswith('KUYRUK_'): key = key[7:] if hasattr(Config, key): try: value = ast.literal_eval(value) except (ValueError, SyntaxError): pass self._setattr(key, value)
Load values from environment variables. Keys must start with `KUYRUK_`.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/config.py#L103-L115
[ "def _setattr(self, key: str, value: Any) -> None:\n if not hasattr(self.__class__, key):\n raise ValueError(\"Unknown config key: %s\" % key)\n\n setattr(self, key, value)\n" ]
class Config: """Kuyruk configuration object. Default values are defined as class attributes. Additional attributes may be added by extensions. """ # Connection Options #################### RABBIT_HOST = 'localhost' """RabbitMQ host.""" RABBIT_PORT = 5672 """RabbitMQ port.""" RABBIT_VIRTUAL_HOST = '/' """RabbitMQ virtual host.""" RABBIT_USER = 'guest' """RabbitMQ user.""" RABBIT_PASSWORD = 'guest' """RabbitMQ password.""" RABBIT_HEARTBEAT = 60 RABBIT_CONNECT_TIMEOUT = 5 RABBIT_READ_TIMEOUT = 5 RABBIT_WRITE_TIMEOUT = 5 TCP_USER_TIMEOUT = 60 # Instance Options ################## EAGER = False """Run tasks in the process without sending to queue. Useful in tests.""" # Worker Options ################ WORKER_MAX_LOAD = None """Pause consuming queue when the load goes above this level.""" WORKER_MAX_RUN_TIME = None """Gracefully shutdown worker after running this seconds.""" WORKER_LOGGING_LEVEL = 'INFO' """Logging level of root logger.""" def from_object(self, obj: Union[str, Any]) -> None: """Load values from an object.""" if isinstance(obj, str): obj = importer.import_object_str(obj) for key in dir(obj): if key.isupper(): value = getattr(obj, key) self._setattr(key, value) logger.info("Config is loaded from object: %r", obj) def from_dict(self, d: Dict[str, Any]) -> None: """Load values from a dict.""" for key, value in d.items(): if key.isupper(): self._setattr(key, value) logger.info("Config is loaded from dict: %r", d) def from_pymodule(self, name: str) -> None: module = importer.import_module(name) for key, value in module.__dict__.items(): if (key.isupper() and not isinstance(value, types.ModuleType)): self._setattr(key, value) logger.info("Config is loaded from module: %s", name) def from_pyfile(self, filename: str) -> None: """Load values from a Python file.""" globals_ = {} # type: Dict[str, Any] locals_ = {} # type: Dict[str, Any] with open(filename, "rb") as f: exec(compile(f.read(), filename, 'exec'), globals_, locals_) for key, value in locals_.items(): if (key.isupper() and not isinstance(value, types.ModuleType)): self._setattr(key, value) logger.info("Config is loaded from file: %s", filename) def _setattr(self, key: str, value: Any) -> None: if not hasattr(self.__class__, key): raise ValueError("Unknown config key: %s" % key) setattr(self, key, value)
cenkalti/kuyruk
kuyruk/kuyruk.py
Kuyruk.task
python
def task(self, queue: str = 'kuyruk', **kwargs: Any) -> Callable: def wrapper(f: Callable) -> Task: return Task(f, self, queue, **kwargs) return wrapper
Wrap functions with this decorator to convert them to *tasks*. After wrapping, calling the function will send a message to a queue instead of running the function. :param queue: Queue name for the tasks. :param kwargs: Keyword arguments will be passed to :class:`~kuyruk.Task` constructor. :return: Callable :class:`~kuyruk.Task` object wrapping the original function.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/kuyruk.py#L37-L53
null
class Kuyruk: """ Provides :func:`~kuyruk.Kuyruk.task` decorator to convert a function into a :class:`~kuyruk.Task`. Provides :func:`~kuyruk.Kuyruk.channel` context manager for opening a new channel on the connection. Connection is opened when the first channel is created. :param config: Must be an instance of :class:`~kuyruk.Config`. If ``None``, default config is used. See :class:`~kuyruk.Config` for default values. """ def __init__(self, config: Config = None) -> None: if config is None: config = Config() self.config = config self.extensions = {} # type: Dict[str, Any] @contextmanager def channel(self) -> Iterator[amqp.Channel]: """Returns a new channel from a new connection as a context manager.""" with self.connection() as conn: ch = conn.channel() logger.info('Opened new channel') with _safe_close(ch): yield ch @contextmanager def connection(self) -> Iterator[amqp.Connection]: """Returns a new connection as a context manager.""" TCP_USER_TIMEOUT = 18 # constant is available on Python 3.6+. socket_settings = {TCP_USER_TIMEOUT: self.config.TCP_USER_TIMEOUT} if sys.platform.startswith('darwin'): del socket_settings[TCP_USER_TIMEOUT] conn = amqp.Connection( host="%s:%s" % (self.config.RABBIT_HOST, self.config.RABBIT_PORT), userid=self.config.RABBIT_USER, password=self.config.RABBIT_PASSWORD, virtual_host=self.config.RABBIT_VIRTUAL_HOST, connect_timeout=self.config.RABBIT_CONNECT_TIMEOUT, read_timeout=self.config.RABBIT_READ_TIMEOUT, write_timeout=self.config.RABBIT_WRITE_TIMEOUT, socket_settings=socket_settings, heartbeat=self.config.RABBIT_HEARTBEAT, ) conn.connect() logger.info('Connected to RabbitMQ') with _safe_close(conn): yield conn def send_tasks_to_queue(self, subtasks: List[SubTask]) -> None: if self.config.EAGER: for subtask in subtasks: subtask.task.apply(*subtask.args, **subtask.kwargs) return declared_queues = set() # type: Set[str] with self.channel() as ch: for subtask in subtasks: queue = subtask.task._queue_for_host(subtask.host) if queue not in declared_queues: ch.queue_declare(queue=queue, durable=True, auto_delete=False) declared_queues.add(queue) description = subtask.task._get_description(subtask.args, subtask.kwargs) subtask.task._send_signal(signals.task_presend, args=subtask.args, kwargs=subtask.kwargs, description=description) body = json.dumps(description) msg = amqp.Message(body=body) ch.basic_publish(msg, exchange="", routing_key=queue) subtask.task._send_signal(signals.task_postsend, args=subtask.args, kwargs=subtask.kwargs, description=description)
cenkalti/kuyruk
kuyruk/kuyruk.py
Kuyruk.channel
python
def channel(self) -> Iterator[amqp.Channel]: with self.connection() as conn: ch = conn.channel() logger.info('Opened new channel') with _safe_close(ch): yield ch
Returns a new channel from a new connection as a context manager.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/kuyruk.py#L56-L62
null
class Kuyruk: """ Provides :func:`~kuyruk.Kuyruk.task` decorator to convert a function into a :class:`~kuyruk.Task`. Provides :func:`~kuyruk.Kuyruk.channel` context manager for opening a new channel on the connection. Connection is opened when the first channel is created. :param config: Must be an instance of :class:`~kuyruk.Config`. If ``None``, default config is used. See :class:`~kuyruk.Config` for default values. """ def __init__(self, config: Config = None) -> None: if config is None: config = Config() self.config = config self.extensions = {} # type: Dict[str, Any] def task(self, queue: str = 'kuyruk', **kwargs: Any) -> Callable: """ Wrap functions with this decorator to convert them to *tasks*. After wrapping, calling the function will send a message to a queue instead of running the function. :param queue: Queue name for the tasks. :param kwargs: Keyword arguments will be passed to :class:`~kuyruk.Task` constructor. :return: Callable :class:`~kuyruk.Task` object wrapping the original function. """ def wrapper(f: Callable) -> Task: return Task(f, self, queue, **kwargs) return wrapper @contextmanager @contextmanager def connection(self) -> Iterator[amqp.Connection]: """Returns a new connection as a context manager.""" TCP_USER_TIMEOUT = 18 # constant is available on Python 3.6+. socket_settings = {TCP_USER_TIMEOUT: self.config.TCP_USER_TIMEOUT} if sys.platform.startswith('darwin'): del socket_settings[TCP_USER_TIMEOUT] conn = amqp.Connection( host="%s:%s" % (self.config.RABBIT_HOST, self.config.RABBIT_PORT), userid=self.config.RABBIT_USER, password=self.config.RABBIT_PASSWORD, virtual_host=self.config.RABBIT_VIRTUAL_HOST, connect_timeout=self.config.RABBIT_CONNECT_TIMEOUT, read_timeout=self.config.RABBIT_READ_TIMEOUT, write_timeout=self.config.RABBIT_WRITE_TIMEOUT, socket_settings=socket_settings, heartbeat=self.config.RABBIT_HEARTBEAT, ) conn.connect() logger.info('Connected to RabbitMQ') with _safe_close(conn): yield conn def send_tasks_to_queue(self, subtasks: List[SubTask]) -> None: if self.config.EAGER: for subtask in subtasks: subtask.task.apply(*subtask.args, **subtask.kwargs) return declared_queues = set() # type: Set[str] with self.channel() as ch: for subtask in subtasks: queue = subtask.task._queue_for_host(subtask.host) if queue not in declared_queues: ch.queue_declare(queue=queue, durable=True, auto_delete=False) declared_queues.add(queue) description = subtask.task._get_description(subtask.args, subtask.kwargs) subtask.task._send_signal(signals.task_presend, args=subtask.args, kwargs=subtask.kwargs, description=description) body = json.dumps(description) msg = amqp.Message(body=body) ch.basic_publish(msg, exchange="", routing_key=queue) subtask.task._send_signal(signals.task_postsend, args=subtask.args, kwargs=subtask.kwargs, description=description)
cenkalti/kuyruk
kuyruk/kuyruk.py
Kuyruk.connection
python
def connection(self) -> Iterator[amqp.Connection]: TCP_USER_TIMEOUT = 18 # constant is available on Python 3.6+. socket_settings = {TCP_USER_TIMEOUT: self.config.TCP_USER_TIMEOUT} if sys.platform.startswith('darwin'): del socket_settings[TCP_USER_TIMEOUT] conn = amqp.Connection( host="%s:%s" % (self.config.RABBIT_HOST, self.config.RABBIT_PORT), userid=self.config.RABBIT_USER, password=self.config.RABBIT_PASSWORD, virtual_host=self.config.RABBIT_VIRTUAL_HOST, connect_timeout=self.config.RABBIT_CONNECT_TIMEOUT, read_timeout=self.config.RABBIT_READ_TIMEOUT, write_timeout=self.config.RABBIT_WRITE_TIMEOUT, socket_settings=socket_settings, heartbeat=self.config.RABBIT_HEARTBEAT, ) conn.connect() logger.info('Connected to RabbitMQ') with _safe_close(conn): yield conn
Returns a new connection as a context manager.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/kuyruk.py#L65-L87
null
class Kuyruk: """ Provides :func:`~kuyruk.Kuyruk.task` decorator to convert a function into a :class:`~kuyruk.Task`. Provides :func:`~kuyruk.Kuyruk.channel` context manager for opening a new channel on the connection. Connection is opened when the first channel is created. :param config: Must be an instance of :class:`~kuyruk.Config`. If ``None``, default config is used. See :class:`~kuyruk.Config` for default values. """ def __init__(self, config: Config = None) -> None: if config is None: config = Config() self.config = config self.extensions = {} # type: Dict[str, Any] def task(self, queue: str = 'kuyruk', **kwargs: Any) -> Callable: """ Wrap functions with this decorator to convert them to *tasks*. After wrapping, calling the function will send a message to a queue instead of running the function. :param queue: Queue name for the tasks. :param kwargs: Keyword arguments will be passed to :class:`~kuyruk.Task` constructor. :return: Callable :class:`~kuyruk.Task` object wrapping the original function. """ def wrapper(f: Callable) -> Task: return Task(f, self, queue, **kwargs) return wrapper @contextmanager def channel(self) -> Iterator[amqp.Channel]: """Returns a new channel from a new connection as a context manager.""" with self.connection() as conn: ch = conn.channel() logger.info('Opened new channel') with _safe_close(ch): yield ch @contextmanager def send_tasks_to_queue(self, subtasks: List[SubTask]) -> None: if self.config.EAGER: for subtask in subtasks: subtask.task.apply(*subtask.args, **subtask.kwargs) return declared_queues = set() # type: Set[str] with self.channel() as ch: for subtask in subtasks: queue = subtask.task._queue_for_host(subtask.host) if queue not in declared_queues: ch.queue_declare(queue=queue, durable=True, auto_delete=False) declared_queues.add(queue) description = subtask.task._get_description(subtask.args, subtask.kwargs) subtask.task._send_signal(signals.task_presend, args=subtask.args, kwargs=subtask.kwargs, description=description) body = json.dumps(description) msg = amqp.Message(body=body) ch.basic_publish(msg, exchange="", routing_key=queue) subtask.task._send_signal(signals.task_postsend, args=subtask.args, kwargs=subtask.kwargs, description=description)
cenkalti/kuyruk
kuyruk/importer.py
import_module
python
def import_module(name: str) -> ModuleType: logger.debug("Importing module: %s", name) if name == main_module_name(): return main_module return importlib.import_module(name)
Import module by it's name from following places in order: - main module - current working directory - Python path
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/importer.py#L13-L24
[ "def main_module_name() -> str:\n \"\"\"Returns main module and module name pair.\"\"\"\n if not hasattr(main_module, '__file__'):\n # running from interactive shell\n return None\n\n main_filename = os.path.basename(main_module.__file__)\n module_name, ext = os.path.splitext(main_filename)\n return module_name\n" ]
import os import sys import logging import importlib from typing import Any from types import ModuleType logger = logging.getLogger(__name__) main_module = sys.modules['__main__'] def import_object(module_name: str, object_name: str) -> Any: module = import_module(module_name) try: return getattr(module, object_name) except AttributeError as e: raise ImportError(e) def import_object_str(s: str) -> Any: module, obj = s.rsplit('.', 1) return import_object(module, obj) def main_module_name() -> str: """Returns main module and module name pair.""" if not hasattr(main_module, '__file__'): # running from interactive shell return None main_filename = os.path.basename(main_module.__file__) module_name, ext = os.path.splitext(main_filename) return module_name
cenkalti/kuyruk
kuyruk/importer.py
main_module_name
python
def main_module_name() -> str: if not hasattr(main_module, '__file__'): # running from interactive shell return None main_filename = os.path.basename(main_module.__file__) module_name, ext = os.path.splitext(main_filename) return module_name
Returns main module and module name pair.
train
https://github.com/cenkalti/kuyruk/blob/c99d66be9d8fb077610f2fa883d5a1d268b42f04/kuyruk/importer.py#L40-L48
null
import os import sys import logging import importlib from typing import Any from types import ModuleType logger = logging.getLogger(__name__) main_module = sys.modules['__main__'] def import_module(name: str) -> ModuleType: """Import module by it's name from following places in order: - main module - current working directory - Python path """ logger.debug("Importing module: %s", name) if name == main_module_name(): return main_module return importlib.import_module(name) def import_object(module_name: str, object_name: str) -> Any: module = import_module(module_name) try: return getattr(module, object_name) except AttributeError as e: raise ImportError(e) def import_object_str(s: str) -> Any: module, obj = s.rsplit('.', 1) return import_object(module, obj)
lmjohns3/downhill
examples/rosenbrock.py
build
python
def build(algo, init): '''Build and return an optimizer for the rosenbrock function. In downhill, an optimizer can be constructed using the build() top-level function. This function requires several Theano quantities such as the loss being optimized and the parameters to update during optimization. ''' x = theano.shared(np.array(init, FLOAT), name='x') n = 0.1 * RandomStreams().normal((len(init) - 1, )) monitors = [] if len(init) == 2: # this gives us access to the x and y locations during optimization. monitors.extend([('x', x[:-1].sum()), ('y', x[1:].sum())]) return downhill.build( algo, loss=(n + 100 * (x[1:] - x[:-1] ** 2) ** 2 + (1 - x[:-1]) ** 2).sum(), params=[x], monitors=monitors, monitor_gradients=True)
Build and return an optimizer for the rosenbrock function. In downhill, an optimizer can be constructed using the build() top-level function. This function requires several Theano quantities such as the loss being optimized and the parameters to update during optimization.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/examples/rosenbrock.py#L15-L33
[ "def build(algo, loss, params=None, inputs=None, updates=(), monitors=(),\n monitor_gradients=False):\n '''Construct an optimizer by name.\n\n Parameters\n ----------\n algo : str\n The name of the optimization algorithm to build.\n loss : Theano expression\n Loss function to minimize. This must be a scalar-valued expression.\n params : list of Theano variables, optional\n Symbolic variables to adjust to minimize the loss. If not given, these\n will be computed automatically by walking the computation graph.\n inputs : list of Theano variables, optional\n Symbolic variables required to compute the loss. If not given, these\n will be computed automatically by walking the computation graph.\n updates : list of update pairs, optional\n A list of pairs providing updates for the internal of the loss\n computation. Normally this is empty, but it can be provided if the loss,\n for example, requires an update to an internal random number generator.\n monitors : dict or sequence of (str, Theano expression) tuples, optional\n Additional values to monitor during optimization. These must be provided\n as either a sequence of (name, expression) tuples, or as a dictionary\n mapping string names to Theano expressions.\n monitor_gradients : bool, optional\n If True, add monitors to log the norms of the parameter gradients during\n optimization. Defaults to False.\n\n Returns\n -------\n optimizer : :class:`Optimizer`\n An optimizer instance.\n '''\n return Optimizer.build(algo, loss, params, inputs,\n updates=updates, monitors=monitors,\n monitor_gradients=monitor_gradients)\n" ]
'''Helper functions for rosenbrock optimization examples.''' import downhill import numpy as np import theano from theano.sandbox.rng_mrg import MRG_RandomStreams as RandomStreams COLORS = ('#d62728 #1f77b4 #2ca02c #9467bd #ff7f0e ' '#e377c2 #8c564b #bcbd22 #7f7f7f #17becf').split() FLOAT = 'df'[theano.config.floatX == 'float32'] def build_and_trace(algo, init, limit=100, **kwargs): '''Run an optimizer on the rosenbrock function. Return xs, ys, and losses. In downhill, optimization algorithms can be iterated over to progressively minimize the loss. At each iteration, the optimizer yields a dictionary of monitor values that were computed during that iteration. Here we build an optimizer and then run it for a fixed number of iterations. ''' kw = dict(min_improvement=0, patience=0, max_gradient_norm=100) kw.update(kwargs) xs, ys, loss = [], [], [] for tm, _ in build(algo, init).iterate([[]], **kw): if len(init) == 2: xs.append(tm['x']) ys.append(tm['y']) loss.append(tm['loss']) if len(loss) == limit: break # Return the optimization up to any failure of patience. return xs[:-9], ys[:-9], loss[-9] def test(algos, n=10, init=[-1.1, 0], limit=100): '''Run several optimizers for comparison. Each optimizer is run a fixed number of times with random hyperparameter values, and the results are yielded back to the caller (often stored in a dictionary). Returns ------- results : sequence of (key, value) pairs A sequence of results from running tests. Each result contains a "key" that describes the test run and a "value" that contains the results from the run. The key is a tuple containing (a) the algorithm, (b) the learning rate, (c) the momentum, (d) the RMS halflife, and (e) the RMS regularizer. The value is a tuple containing the (a) x-values and (b) y-values during the optimization, and (c) the loss value. (The x- and y-value are only non-empty for 2D experiments.) ''' for algo in algos: for _ in range(n): mu = max(0, np.random.uniform(0, 2) - 1) rate = np.exp(np.random.uniform(-8, -1)) half = int(np.exp(np.random.uniform(0, 4))) reg = np.exp(np.random.uniform(-12, 0)) yield (algo, rate, mu, half, reg), build_and_trace( algo, init, limit, momentum=mu, learning_rate=rate, rms_halflife=half, rms_regularizer=reg)
lmjohns3/downhill
examples/rosenbrock.py
build_and_trace
python
def build_and_trace(algo, init, limit=100, **kwargs): '''Run an optimizer on the rosenbrock function. Return xs, ys, and losses. In downhill, optimization algorithms can be iterated over to progressively minimize the loss. At each iteration, the optimizer yields a dictionary of monitor values that were computed during that iteration. Here we build an optimizer and then run it for a fixed number of iterations. ''' kw = dict(min_improvement=0, patience=0, max_gradient_norm=100) kw.update(kwargs) xs, ys, loss = [], [], [] for tm, _ in build(algo, init).iterate([[]], **kw): if len(init) == 2: xs.append(tm['x']) ys.append(tm['y']) loss.append(tm['loss']) if len(loss) == limit: break # Return the optimization up to any failure of patience. return xs[:-9], ys[:-9], loss[-9]
Run an optimizer on the rosenbrock function. Return xs, ys, and losses. In downhill, optimization algorithms can be iterated over to progressively minimize the loss. At each iteration, the optimizer yields a dictionary of monitor values that were computed during that iteration. Here we build an optimizer and then run it for a fixed number of iterations.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/examples/rosenbrock.py#L36-L55
[ "def build(algo, init):\n '''Build and return an optimizer for the rosenbrock function.\n\n In downhill, an optimizer can be constructed using the build() top-level\n function. This function requires several Theano quantities such as the loss\n being optimized and the parameters to update during optimization.\n '''\n x = theano.shared(np.array(init, FLOAT), name='x')\n n = 0.1 * RandomStreams().normal((len(init) - 1, ))\n monitors = []\n if len(init) == 2:\n # this gives us access to the x and y locations during optimization.\n monitors.extend([('x', x[:-1].sum()), ('y', x[1:].sum())])\n return downhill.build(\n algo,\n loss=(n + 100 * (x[1:] - x[:-1] ** 2) ** 2 + (1 - x[:-1]) ** 2).sum(),\n params=[x],\n monitors=monitors,\n monitor_gradients=True)\n" ]
'''Helper functions for rosenbrock optimization examples.''' import downhill import numpy as np import theano from theano.sandbox.rng_mrg import MRG_RandomStreams as RandomStreams COLORS = ('#d62728 #1f77b4 #2ca02c #9467bd #ff7f0e ' '#e377c2 #8c564b #bcbd22 #7f7f7f #17becf').split() FLOAT = 'df'[theano.config.floatX == 'float32'] def build(algo, init): '''Build and return an optimizer for the rosenbrock function. In downhill, an optimizer can be constructed using the build() top-level function. This function requires several Theano quantities such as the loss being optimized and the parameters to update during optimization. ''' x = theano.shared(np.array(init, FLOAT), name='x') n = 0.1 * RandomStreams().normal((len(init) - 1, )) monitors = [] if len(init) == 2: # this gives us access to the x and y locations during optimization. monitors.extend([('x', x[:-1].sum()), ('y', x[1:].sum())]) return downhill.build( algo, loss=(n + 100 * (x[1:] - x[:-1] ** 2) ** 2 + (1 - x[:-1]) ** 2).sum(), params=[x], monitors=monitors, monitor_gradients=True) def test(algos, n=10, init=[-1.1, 0], limit=100): '''Run several optimizers for comparison. Each optimizer is run a fixed number of times with random hyperparameter values, and the results are yielded back to the caller (often stored in a dictionary). Returns ------- results : sequence of (key, value) pairs A sequence of results from running tests. Each result contains a "key" that describes the test run and a "value" that contains the results from the run. The key is a tuple containing (a) the algorithm, (b) the learning rate, (c) the momentum, (d) the RMS halflife, and (e) the RMS regularizer. The value is a tuple containing the (a) x-values and (b) y-values during the optimization, and (c) the loss value. (The x- and y-value are only non-empty for 2D experiments.) ''' for algo in algos: for _ in range(n): mu = max(0, np.random.uniform(0, 2) - 1) rate = np.exp(np.random.uniform(-8, -1)) half = int(np.exp(np.random.uniform(0, 4))) reg = np.exp(np.random.uniform(-12, 0)) yield (algo, rate, mu, half, reg), build_and_trace( algo, init, limit, momentum=mu, learning_rate=rate, rms_halflife=half, rms_regularizer=reg)
lmjohns3/downhill
downhill/__init__.py
minimize
python
def minimize(loss, train, valid=None, params=None, inputs=None, algo='rmsprop', updates=(), monitors=(), monitor_gradients=False, batch_size=32, train_batches=None, valid_batches=None, **kwargs): '''Minimize a loss function with respect to some symbolic parameters. Additional keyword arguments are passed to the underlying :class:`Optimizer <downhill.base.Optimizer>` instance. Parameters ---------- loss : Theano expression Loss function to minimize. This must be a scalar-valued expression. train : :class:`Dataset <downhill.dataset.Dataset>`, ndarray, or callable Dataset to use for computing gradient updates. valid : :class:`Dataset <downhill.dataset.Dataset>`, ndarray, or callable, optional Dataset to use for validating the minimization process. The training dataset is used if this is not provided. params : list of Theano variables, optional Symbolic variables to adjust to minimize the loss. If not given, these will be computed automatically by walking the computation graph. inputs : list of Theano variables, optional Symbolic variables required to compute the loss. If not given, these will be computed automatically by walking the computation graph. algo : str, optional Name of the minimization algorithm to use. Must be one of the strings that can be passed to :func:`build`. Defaults to ``'rmsprop'``. updates : list of update pairs, optional A list of pairs providing updates for the internal of the loss computation. Normally this is empty, but it can be provided if the loss, for example, requires an update to an internal random number generator. monitors : dict or sequence of (str, Theano expression) tuples, optional Additional values to monitor during optimization. These must be provided as either a sequence of (name, expression) tuples, or as a dictionary mapping string names to Theano expressions. monitor_gradients : bool, optional If True, add monitors to log the norms of the parameter gradients during optimization. Defaults to False. batch_size : int, optional Size of batches provided by datasets. Defaults to 32. train_batches : int, optional Number of batches of training data to iterate over during one pass of optimization. Defaults to None, which uses the entire training dataset. valid_batches : int, optional Number of batches of validation data to iterate over during one pass of validation. Defaults to None, which uses the entire validation dataset. Returns ------- train_monitors : dict A dictionary mapping monitor names to monitor values. This dictionary will always contain the ``'loss'`` key, giving the value of the loss evaluated on the training dataset. valid_monitors : dict A dictionary mapping monitor names to monitor values, evaluated on the validation dataset. This dictionary will always contain the ``'loss'`` key, giving the value of the loss function. Because validation is not always computed after every optimization update, these monitor values may be "stale"; however, they will always contain the most recently computed values. ''' if not isinstance(train, Dataset): train = Dataset( train, name='train', batch_size=batch_size, iteration_size=train_batches, ) if valid is not None and not isinstance(valid, Dataset): valid = Dataset( valid, name='valid', batch_size=batch_size, iteration_size=valid_batches, ) return build( algo, loss=loss, params=params, inputs=inputs, updates=updates, monitors=monitors, monitor_gradients=monitor_gradients, ).minimize(train, valid, **kwargs)
Minimize a loss function with respect to some symbolic parameters. Additional keyword arguments are passed to the underlying :class:`Optimizer <downhill.base.Optimizer>` instance. Parameters ---------- loss : Theano expression Loss function to minimize. This must be a scalar-valued expression. train : :class:`Dataset <downhill.dataset.Dataset>`, ndarray, or callable Dataset to use for computing gradient updates. valid : :class:`Dataset <downhill.dataset.Dataset>`, ndarray, or callable, optional Dataset to use for validating the minimization process. The training dataset is used if this is not provided. params : list of Theano variables, optional Symbolic variables to adjust to minimize the loss. If not given, these will be computed automatically by walking the computation graph. inputs : list of Theano variables, optional Symbolic variables required to compute the loss. If not given, these will be computed automatically by walking the computation graph. algo : str, optional Name of the minimization algorithm to use. Must be one of the strings that can be passed to :func:`build`. Defaults to ``'rmsprop'``. updates : list of update pairs, optional A list of pairs providing updates for the internal of the loss computation. Normally this is empty, but it can be provided if the loss, for example, requires an update to an internal random number generator. monitors : dict or sequence of (str, Theano expression) tuples, optional Additional values to monitor during optimization. These must be provided as either a sequence of (name, expression) tuples, or as a dictionary mapping string names to Theano expressions. monitor_gradients : bool, optional If True, add monitors to log the norms of the parameter gradients during optimization. Defaults to False. batch_size : int, optional Size of batches provided by datasets. Defaults to 32. train_batches : int, optional Number of batches of training data to iterate over during one pass of optimization. Defaults to None, which uses the entire training dataset. valid_batches : int, optional Number of batches of validation data to iterate over during one pass of validation. Defaults to None, which uses the entire validation dataset. Returns ------- train_monitors : dict A dictionary mapping monitor names to monitor values. This dictionary will always contain the ``'loss'`` key, giving the value of the loss evaluated on the training dataset. valid_monitors : dict A dictionary mapping monitor names to monitor values, evaluated on the validation dataset. This dictionary will always contain the ``'loss'`` key, giving the value of the loss function. Because validation is not always computed after every optimization update, these monitor values may be "stale"; however, they will always contain the most recently computed values.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/__init__.py#L9-L91
[ "def build(algo, loss, params=None, inputs=None, updates=(), monitors=(),\n monitor_gradients=False):\n '''Construct an optimizer by name.\n\n Parameters\n ----------\n algo : str\n The name of the optimization algorithm to build.\n loss : Theano expression\n Loss function to minimize. This must be a scalar-valued expression.\n params : list of Theano variables, optional\n Symbolic variables to adjust to minimize the loss. If not given, these\n will be computed automatically by walking the computation graph.\n inputs : list of Theano variables, optional\n Symbolic variables required to compute the loss. If not given, these\n will be computed automatically by walking the computation graph.\n updates : list of update pairs, optional\n A list of pairs providing updates for the internal of the loss\n computation. Normally this is empty, but it can be provided if the loss,\n for example, requires an update to an internal random number generator.\n monitors : dict or sequence of (str, Theano expression) tuples, optional\n Additional values to monitor during optimization. These must be provided\n as either a sequence of (name, expression) tuples, or as a dictionary\n mapping string names to Theano expressions.\n monitor_gradients : bool, optional\n If True, add monitors to log the norms of the parameter gradients during\n optimization. Defaults to False.\n\n Returns\n -------\n optimizer : :class:`Optimizer`\n An optimizer instance.\n '''\n return Optimizer.build(algo, loss, params, inputs,\n updates=updates, monitors=monitors,\n monitor_gradients=monitor_gradients)\n" ]
from .adaptive import * from .base import build, Optimizer from .dataset import Dataset from .first_order import * __version__ = '0.5.0pre'
lmjohns3/downhill
examples/rosenbrock-2d.py
make_label
python
def make_label(loss, key): '''Create a legend label for an optimization run.''' algo, rate, mu, half, reg = key slots, args = ['{:.3f}', '{}', 'm={:.3f}'], [loss, algo, mu] if algo in 'SGD NAG RMSProp Adam ESGD'.split(): slots.append('lr={:.2e}') args.append(rate) if algo in 'RMSProp ADADELTA ESGD'.split(): slots.append('rmsh={}') args.append(half) slots.append('rmsr={:.2e}') args.append(reg) return ' '.join(slots).format(*args)
Create a legend label for an optimization run.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/examples/rosenbrock-2d.py#L25-L37
null
'''Optimization example using the two-dimensional Rosenbrock "banana" function. This example trains up several optimization algorithms and displays the performance of each algorithm across several different (randomly-chosen) hyperparameter settings. This example is meant to show how different optimization algorithms perform when given the same optimization problem. Many of the algorithms' performances are strongly dependent on the values of various hyperparameters, such as the learning rate and momentum values. ''' import matplotlib.pyplot as plt import numpy as np import rosenbrock def by_loss(item): '''Helper for sorting optimization runs by their final loss value.''' label, (xs, ys, loss) = item return loss # Here we run a number of rosenbrock optimization algorithms and measure their # performance. Below we plot the results. algos = 'SGD NAG RMSProp RProp Adam ADADELTA ESGD'.split() results = ((make_label(loss, key), xs, ys) for key, (xs, ys, loss) in sorted(rosenbrock.test(algos), key=by_loss)) _, ax = plt.subplots(1, 1) for color, (label, xs, ys) in zip(rosenbrock.COLORS, results): ax.plot(xs, ys, 'o-', color=color, label=label, alpha=0.8, lw=2, markersize=5, mew=1, mec=color, mfc='none') # make a contour plot of the rosenbrock function surface. X, Y = np.meshgrid(np.linspace(-1.3, 1.3, 31), np.linspace(-0.9, 1.7, 31)) Z = 100 * (Y - X ** 2) ** 2 + (1 - X) ** 2 ax.plot([1], [1], 'x', mew=3, markersize=10, color='#111111') ax.contourf(X, Y, Z, np.logspace(-1, 3, 31), cmap='gray_r') ax.set_xlim(-1.3, 1.3) ax.set_ylim(-0.9, 1.7) plt.legend(loc='lower right') plt.show()
lmjohns3/downhill
downhill/dataset.py
Dataset.iterate
python
def iterate(self, shuffle=True): '''Iterate over batches in the dataset. This method generates ``iteration_size`` batches from the dataset and then returns. Parameters ---------- shuffle : bool, optional Shuffle the batches in this dataset if the iteration reaches the end of the batch list. Defaults to True. Yields ------ batches : data batches A sequence of batches---often from a training, validation, or test dataset. ''' for _ in range(self.iteration_size): if self._callable is not None: yield self._callable() else: yield self._next_batch(shuffle)
Iterate over batches in the dataset. This method generates ``iteration_size`` batches from the dataset and then returns. Parameters ---------- shuffle : bool, optional Shuffle the batches in this dataset if the iteration reaches the end of the batch list. Defaults to True. Yields ------ batches : data batches A sequence of batches---often from a training, validation, or test dataset.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/dataset.py#L183-L205
[ "def _next_batch(self, shuffle=True):\n batch = [x.iloc[i] if hasattr(x, 'iloc') else x[i]\n for x, i in zip(self._inputs, self._slices[self._index])]\n self._index += 1\n if self._index >= len(self._slices):\n if shuffle:\n self.shuffle()\n self._index = 0\n return batch\n" ]
class Dataset: '''This class handles batching and shuffling a dataset. In ``downhill``, losses are optimized using sets of data collected from the problem that generated the loss. During optimization, data are grouped into "mini-batches"---that is, chunks that are larger than 1 sample and smaller than the entire set of samples; typically the size of a mini-batch is between 10 and 100, but the specific setting can be varied depending on your model, hardware, dataset, and so forth. These mini-batches must be presented to the optimization algorithm in pseudo-random order to match the underlying stochasticity assumptions of many optimization algorithms. This class handles the process of grouping data into mini-batches as well as iterating and shuffling these mini-batches dynamically as the dataset is consumed by the optimization algorithm. For many tasks, a dataset is obtained as a large block of sample data, which in Python is normally assembled as a ``numpy`` ndarray. To use this class on such a dataset, just pass in a list or tuple containing ``numpy`` arrays; the number of these arrays must match the number of inputs that your loss computation requires. There are some cases when a suitable set of training data would be prohibitively expensive to assemble in memory as a single ``numpy`` array. To handle these cases, this class can also handle a dataset that is provided via a Python callable. For more information on using callables to provide data to your model, see :ref:`data-using-callables`. Parameters ---------- inputs : callable or list of ndarray/sparse matrix/DataFrame/theano shared var One or more sets of data. If this parameter is callable, then mini-batches will be obtained by calling the callable with no arguments; the callable is expected to return a tuple of ndarray-like objects that will be suitable for optimizing the loss at hand. If this parameter is a list (or a tuple), it must contain array-like objects: ``numpy.ndarray``, ``scipy.sparse.csc_matrix``, ``scipy.sparse.csr_matrix``, ``pandas.DataFrame`` or ``theano.shared``. These are assumed to contain data for computing the loss, so the length of this tuple or list should match the number of inputs required by the loss computation. If multiple arrays are provided, their lengths along the axis given by the ``axis`` parameter (defaults to 0) must match. name : str, optional A string that is used to describe this dataset. Usually something like 'test' or 'train'. batch_size : int, optional The size of the mini-batches to create from the data sequences. If this is negative or zero, all data in the dataset will be used in one batch. Defaults to 32. This parameter has no effect if ``inputs`` is callable. iteration_size : int, optional The number of batches to yield for each call to iterate(). Defaults to the length of the data divided by batch_size. If the dataset is a callable, then the number is len(callable). If callable has no length, then the number is set to 100. axis : int, optional The axis along which to split the data arrays, if the first parameter is given as one or more ndarrays. If not provided, defaults to 0. rng : :class:`numpy.random.RandomState` or int, optional A random number generator, or an integer seed for a random number generator. If not provided, the random number generator will be created with an automatically chosen seed. ''' _count = 0 def __init__(self, inputs, name=None, batch_size=32, iteration_size=None, axis=0, rng=None): self.name = name or 'dataset{}'.format(Dataset._count) Dataset._count += 1 self.batch_size = batch_size self.iteration_size = iteration_size self.rng = rng if rng is None or isinstance(rng, int): self.rng = np.random.RandomState(rng) self._inputs = None self._slices = None self._callable = None if isinstance(inputs, collections.Callable): self._init_callable(inputs) else: self._init_arrays(inputs, axis) def _init_callable(self, inputs): self._callable = inputs if not self.iteration_size: try: self.iteration_size = len(inputs) except (TypeError, AttributeError): # has no len self.iteration_size = 100 util.log('{0.name}: {0.iteration_size} mini-batches from callable', self) def _init_arrays(self, inputs, axis=0): if not isinstance(inputs, (tuple, list)): inputs = (inputs, ) shapes = [] self._inputs = [] for i, x in enumerate(inputs): self._inputs.append(x) if isinstance(x, np.ndarray): shapes.append(x.shape) continue if isinstance(x, theano.compile.SharedVariable): shapes.append(x.get_value(borrow=True).shape) continue if 'pandas.' in str(type(x)): # hacky but prevents a global import import pandas as pd if isinstance(x, (pd.Series, pd.DataFrame)): shapes.append(x.shape) continue if 'scipy.sparse.' in str(type(x)): # same here import scipy.sparse as ss if isinstance(x, (ss.csr.csr_matrix, ss.csc.csc_matrix)): shapes.append(x.shape) continue raise ValueError( 'input {} (type {}) must be numpy.array, theano.shared, ' 'or pandas.{{Series,DataFrame}}'.format(i, type(x))) L = shapes[0][axis] assert all(L == s[axis] for s in shapes), \ 'shapes do not match along axis {}: {}'.format( axis, '; '.join(str(s) for s in shapes)) B = L if self.batch_size <= 0 else self.batch_size self._index = 0 self._slices = [] for i in range(0, L, B): where = [] for shape in shapes: slices = [slice(None) for _ in shape] slices[axis] = slice(i, min(L, i + B)) where.append(tuple(slices)) self._slices.append(where) self.shuffle() if not self.iteration_size: self.iteration_size = len(self._slices) util.log('{0.name}: {0.iteration_size} of {1} mini-batches from {2}', self, len(self._slices), '; '.join(str(s) for s in shapes)) def __iter__(self): return self.iterate(True) def shuffle(self): '''Shuffle the batches in the dataset. If this dataset was constructed using a callable, this method has no effect. ''' if self._slices is not None: self.rng.shuffle(self._slices) def _next_batch(self, shuffle=True): batch = [x.iloc[i] if hasattr(x, 'iloc') else x[i] for x, i in zip(self._inputs, self._slices[self._index])] self._index += 1 if self._index >= len(self._slices): if shuffle: self.shuffle() self._index = 0 return batch
lmjohns3/downhill
downhill/util.py
shared_like
python
def shared_like(param, suffix, init=0): '''Create a Theano shared variable like an existing parameter. Parameters ---------- param : Theano variable Theano variable to use for shape information. suffix : str Suffix to append to the parameter's name for the new variable. init : float or ndarray, optional Initial value of the shared variable. Defaults to 0. Returns ------- shared : Theano shared variable A new shared variable with the same shape and data type as ``param``. ''' return theano.shared(np.zeros_like(param.get_value()) + init, name='{}_{}'.format(param.name, suffix), broadcastable=param.broadcastable)
Create a Theano shared variable like an existing parameter. Parameters ---------- param : Theano variable Theano variable to use for shape information. suffix : str Suffix to append to the parameter's name for the new variable. init : float or ndarray, optional Initial value of the shared variable. Defaults to 0. Returns ------- shared : Theano shared variable A new shared variable with the same shape and data type as ``param``.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/util.py#L30-L49
null
# -*- coding: utf-8 -*- '''A module of utility functions and other goodies.''' import click import datetime import inspect import numpy as np import theano import theano.tensor as TT class Registrar(type): '''A metaclass that builds a registry of its subclasses.''' def __init__(cls, name, bases, dct): if not hasattr(cls, '_registry'): cls._registry = {} else: cls._registry[name.lower()] = cls super(Registrar, cls).__init__(name, bases, dct) def build(cls, key, *args, **kwargs): return cls._registry[key.lower()](*args, **kwargs) def is_registered(cls, key): return key.lower() in cls._registry def as_float(x): '''Cast a floating point value to a Theano ``floatX`` symbol. Parameters ---------- x : float, ndarray, or Theano expression Some quantity to cast to floating point. Returns ------- x : Theano expression A symbolic variable cast as a ``floatX`` value. ''' return TT.cast(x, theano.config.floatX) def find_inputs_and_params(node): '''Walk a computation graph and extract root variables. Parameters ---------- node : Theano expression A symbolic Theano expression to walk. Returns ------- inputs : list Theano variables A list of candidate inputs for this graph. Inputs are nodes in the graph with no parents that are not shared and are not constants. params : list of Theano shared variables A list of candidate parameters for this graph. Parameters are nodes in the graph that are shared variables. ''' queue, seen, inputs, params = [node], set(), set(), set() while queue: node = queue.pop() seen.add(node) queue.extend(p for p in node.get_parents() if p not in seen) if not node.get_parents(): if isinstance(node, theano.compile.SharedVariable): params.add(node) elif not isinstance(node, TT.Constant): inputs.add(node) return list(inputs), list(params) _detailed_callsite = False def enable_detailed_callsite_logging(): '''Enable detailed callsite logging.''' global _detailed_callsite _detailed_callsite = True def log(msg, *args, **kwargs): '''Log a message to the console. Parameters ---------- msg : str A string to display on the console. This can contain {}-style formatting commands; the remaining positional and keyword arguments will be used to fill them in. ''' now = datetime.datetime.now() module = 'downhill' if _detailed_callsite: caller = inspect.stack()[1] parts = caller.filename.replace('.py', '').split('/') module = '{}:{}'.format( '.'.join(parts[parts.index('downhill')+1:]), caller.lineno) click.echo(' '.join(( click.style(now.strftime('%Y%m%d'), fg='blue'), click.style(now.strftime('%H%M%S'), fg='cyan'), click.style(module, fg='magenta'), msg.format(*args, **kwargs), ))) def log_param(name, value): '''Log a parameter value to the console. Parameters ---------- name : str Name of the parameter being logged. value : any Value of the parameter being logged. ''' log('setting {} = {}', click.style(str(name)), click.style(str(value), fg='yellow'))
lmjohns3/downhill
downhill/util.py
find_inputs_and_params
python
def find_inputs_and_params(node): '''Walk a computation graph and extract root variables. Parameters ---------- node : Theano expression A symbolic Theano expression to walk. Returns ------- inputs : list Theano variables A list of candidate inputs for this graph. Inputs are nodes in the graph with no parents that are not shared and are not constants. params : list of Theano shared variables A list of candidate parameters for this graph. Parameters are nodes in the graph that are shared variables. ''' queue, seen, inputs, params = [node], set(), set(), set() while queue: node = queue.pop() seen.add(node) queue.extend(p for p in node.get_parents() if p not in seen) if not node.get_parents(): if isinstance(node, theano.compile.SharedVariable): params.add(node) elif not isinstance(node, TT.Constant): inputs.add(node) return list(inputs), list(params)
Walk a computation graph and extract root variables. Parameters ---------- node : Theano expression A symbolic Theano expression to walk. Returns ------- inputs : list Theano variables A list of candidate inputs for this graph. Inputs are nodes in the graph with no parents that are not shared and are not constants. params : list of Theano shared variables A list of candidate parameters for this graph. Parameters are nodes in the graph that are shared variables.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/util.py#L68-L95
null
# -*- coding: utf-8 -*- '''A module of utility functions and other goodies.''' import click import datetime import inspect import numpy as np import theano import theano.tensor as TT class Registrar(type): '''A metaclass that builds a registry of its subclasses.''' def __init__(cls, name, bases, dct): if not hasattr(cls, '_registry'): cls._registry = {} else: cls._registry[name.lower()] = cls super(Registrar, cls).__init__(name, bases, dct) def build(cls, key, *args, **kwargs): return cls._registry[key.lower()](*args, **kwargs) def is_registered(cls, key): return key.lower() in cls._registry def shared_like(param, suffix, init=0): '''Create a Theano shared variable like an existing parameter. Parameters ---------- param : Theano variable Theano variable to use for shape information. suffix : str Suffix to append to the parameter's name for the new variable. init : float or ndarray, optional Initial value of the shared variable. Defaults to 0. Returns ------- shared : Theano shared variable A new shared variable with the same shape and data type as ``param``. ''' return theano.shared(np.zeros_like(param.get_value()) + init, name='{}_{}'.format(param.name, suffix), broadcastable=param.broadcastable) def as_float(x): '''Cast a floating point value to a Theano ``floatX`` symbol. Parameters ---------- x : float, ndarray, or Theano expression Some quantity to cast to floating point. Returns ------- x : Theano expression A symbolic variable cast as a ``floatX`` value. ''' return TT.cast(x, theano.config.floatX) _detailed_callsite = False def enable_detailed_callsite_logging(): '''Enable detailed callsite logging.''' global _detailed_callsite _detailed_callsite = True def log(msg, *args, **kwargs): '''Log a message to the console. Parameters ---------- msg : str A string to display on the console. This can contain {}-style formatting commands; the remaining positional and keyword arguments will be used to fill them in. ''' now = datetime.datetime.now() module = 'downhill' if _detailed_callsite: caller = inspect.stack()[1] parts = caller.filename.replace('.py', '').split('/') module = '{}:{}'.format( '.'.join(parts[parts.index('downhill')+1:]), caller.lineno) click.echo(' '.join(( click.style(now.strftime('%Y%m%d'), fg='blue'), click.style(now.strftime('%H%M%S'), fg='cyan'), click.style(module, fg='magenta'), msg.format(*args, **kwargs), ))) def log_param(name, value): '''Log a parameter value to the console. Parameters ---------- name : str Name of the parameter being logged. value : any Value of the parameter being logged. ''' log('setting {} = {}', click.style(str(name)), click.style(str(value), fg='yellow'))
lmjohns3/downhill
downhill/util.py
log
python
def log(msg, *args, **kwargs): '''Log a message to the console. Parameters ---------- msg : str A string to display on the console. This can contain {}-style formatting commands; the remaining positional and keyword arguments will be used to fill them in. ''' now = datetime.datetime.now() module = 'downhill' if _detailed_callsite: caller = inspect.stack()[1] parts = caller.filename.replace('.py', '').split('/') module = '{}:{}'.format( '.'.join(parts[parts.index('downhill')+1:]), caller.lineno) click.echo(' '.join(( click.style(now.strftime('%Y%m%d'), fg='blue'), click.style(now.strftime('%H%M%S'), fg='cyan'), click.style(module, fg='magenta'), msg.format(*args, **kwargs), )))
Log a message to the console. Parameters ---------- msg : str A string to display on the console. This can contain {}-style formatting commands; the remaining positional and keyword arguments will be used to fill them in.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/util.py#L107-L129
null
# -*- coding: utf-8 -*- '''A module of utility functions and other goodies.''' import click import datetime import inspect import numpy as np import theano import theano.tensor as TT class Registrar(type): '''A metaclass that builds a registry of its subclasses.''' def __init__(cls, name, bases, dct): if not hasattr(cls, '_registry'): cls._registry = {} else: cls._registry[name.lower()] = cls super(Registrar, cls).__init__(name, bases, dct) def build(cls, key, *args, **kwargs): return cls._registry[key.lower()](*args, **kwargs) def is_registered(cls, key): return key.lower() in cls._registry def shared_like(param, suffix, init=0): '''Create a Theano shared variable like an existing parameter. Parameters ---------- param : Theano variable Theano variable to use for shape information. suffix : str Suffix to append to the parameter's name for the new variable. init : float or ndarray, optional Initial value of the shared variable. Defaults to 0. Returns ------- shared : Theano shared variable A new shared variable with the same shape and data type as ``param``. ''' return theano.shared(np.zeros_like(param.get_value()) + init, name='{}_{}'.format(param.name, suffix), broadcastable=param.broadcastable) def as_float(x): '''Cast a floating point value to a Theano ``floatX`` symbol. Parameters ---------- x : float, ndarray, or Theano expression Some quantity to cast to floating point. Returns ------- x : Theano expression A symbolic variable cast as a ``floatX`` value. ''' return TT.cast(x, theano.config.floatX) def find_inputs_and_params(node): '''Walk a computation graph and extract root variables. Parameters ---------- node : Theano expression A symbolic Theano expression to walk. Returns ------- inputs : list Theano variables A list of candidate inputs for this graph. Inputs are nodes in the graph with no parents that are not shared and are not constants. params : list of Theano shared variables A list of candidate parameters for this graph. Parameters are nodes in the graph that are shared variables. ''' queue, seen, inputs, params = [node], set(), set(), set() while queue: node = queue.pop() seen.add(node) queue.extend(p for p in node.get_parents() if p not in seen) if not node.get_parents(): if isinstance(node, theano.compile.SharedVariable): params.add(node) elif not isinstance(node, TT.Constant): inputs.add(node) return list(inputs), list(params) _detailed_callsite = False def enable_detailed_callsite_logging(): '''Enable detailed callsite logging.''' global _detailed_callsite _detailed_callsite = True def log_param(name, value): '''Log a parameter value to the console. Parameters ---------- name : str Name of the parameter being logged. value : any Value of the parameter being logged. ''' log('setting {} = {}', click.style(str(name)), click.style(str(value), fg='yellow'))
lmjohns3/downhill
downhill/util.py
log_param
python
def log_param(name, value): '''Log a parameter value to the console. Parameters ---------- name : str Name of the parameter being logged. value : any Value of the parameter being logged. ''' log('setting {} = {}', click.style(str(name)), click.style(str(value), fg='yellow'))
Log a parameter value to the console. Parameters ---------- name : str Name of the parameter being logged. value : any Value of the parameter being logged.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/util.py#L132-L143
[ "def log(msg, *args, **kwargs):\n '''Log a message to the console.\n\n Parameters\n ----------\n msg : str\n A string to display on the console. This can contain {}-style\n formatting commands; the remaining positional and keyword arguments\n will be used to fill them in.\n '''\n now = datetime.datetime.now()\n module = 'downhill'\n if _detailed_callsite:\n caller = inspect.stack()[1]\n parts = caller.filename.replace('.py', '').split('/')\n module = '{}:{}'.format(\n '.'.join(parts[parts.index('downhill')+1:]), caller.lineno)\n click.echo(' '.join((\n click.style(now.strftime('%Y%m%d'), fg='blue'),\n click.style(now.strftime('%H%M%S'), fg='cyan'),\n click.style(module, fg='magenta'),\n msg.format(*args, **kwargs),\n )))\n" ]
# -*- coding: utf-8 -*- '''A module of utility functions and other goodies.''' import click import datetime import inspect import numpy as np import theano import theano.tensor as TT class Registrar(type): '''A metaclass that builds a registry of its subclasses.''' def __init__(cls, name, bases, dct): if not hasattr(cls, '_registry'): cls._registry = {} else: cls._registry[name.lower()] = cls super(Registrar, cls).__init__(name, bases, dct) def build(cls, key, *args, **kwargs): return cls._registry[key.lower()](*args, **kwargs) def is_registered(cls, key): return key.lower() in cls._registry def shared_like(param, suffix, init=0): '''Create a Theano shared variable like an existing parameter. Parameters ---------- param : Theano variable Theano variable to use for shape information. suffix : str Suffix to append to the parameter's name for the new variable. init : float or ndarray, optional Initial value of the shared variable. Defaults to 0. Returns ------- shared : Theano shared variable A new shared variable with the same shape and data type as ``param``. ''' return theano.shared(np.zeros_like(param.get_value()) + init, name='{}_{}'.format(param.name, suffix), broadcastable=param.broadcastable) def as_float(x): '''Cast a floating point value to a Theano ``floatX`` symbol. Parameters ---------- x : float, ndarray, or Theano expression Some quantity to cast to floating point. Returns ------- x : Theano expression A symbolic variable cast as a ``floatX`` value. ''' return TT.cast(x, theano.config.floatX) def find_inputs_and_params(node): '''Walk a computation graph and extract root variables. Parameters ---------- node : Theano expression A symbolic Theano expression to walk. Returns ------- inputs : list Theano variables A list of candidate inputs for this graph. Inputs are nodes in the graph with no parents that are not shared and are not constants. params : list of Theano shared variables A list of candidate parameters for this graph. Parameters are nodes in the graph that are shared variables. ''' queue, seen, inputs, params = [node], set(), set(), set() while queue: node = queue.pop() seen.add(node) queue.extend(p for p in node.get_parents() if p not in seen) if not node.get_parents(): if isinstance(node, theano.compile.SharedVariable): params.add(node) elif not isinstance(node, TT.Constant): inputs.add(node) return list(inputs), list(params) _detailed_callsite = False def enable_detailed_callsite_logging(): '''Enable detailed callsite logging.''' global _detailed_callsite _detailed_callsite = True def log(msg, *args, **kwargs): '''Log a message to the console. Parameters ---------- msg : str A string to display on the console. This can contain {}-style formatting commands; the remaining positional and keyword arguments will be used to fill them in. ''' now = datetime.datetime.now() module = 'downhill' if _detailed_callsite: caller = inspect.stack()[1] parts = caller.filename.replace('.py', '').split('/') module = '{}:{}'.format( '.'.join(parts[parts.index('downhill')+1:]), caller.lineno) click.echo(' '.join(( click.style(now.strftime('%Y%m%d'), fg='blue'), click.style(now.strftime('%H%M%S'), fg='cyan'), click.style(module, fg='magenta'), msg.format(*args, **kwargs), )))
lmjohns3/downhill
examples/mnist-sparse-factorization.py
load_mnist
python
def load_mnist(): '''Load the MNIST digits dataset.''' mnist = skdata.mnist.dataset.MNIST() mnist.meta # trigger download if needed. def arr(n, dtype): arr = mnist.arrays[n] return arr.reshape((len(arr), -1)).astype(dtype) train_images = arr('train_images', np.float32) / 128 - 1 train_labels = arr('train_labels', np.uint8) return ((train_images[:50000], train_labels[:50000, 0]), (train_images[50000:], train_labels[50000:, 0]))
Load the MNIST digits dataset.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/examples/mnist-sparse-factorization.py#L11-L22
[ "def arr(n, dtype):\n arr = mnist.arrays[n]\n return arr.reshape((len(arr), -1)).astype(dtype)\n" ]
import downhill import matplotlib.pyplot as plt import numpy as np import skdata.mnist import theano import theano.tensor as TT FLOAT = 'df'[theano.config.floatX == 'float32'] def plot_images(imgs, loc=111, title=None, channels=1): '''Plot an array of images. We assume that we are given a matrix of data whose shape is (n*n, s*s*c) -- that is, there are n^2 images along the first axis of the array, and each image is c squares measuring s pixels on a side. Each row of the input will be plotted as a sub-region within a single image array containing an n x n grid of images. ''' n = int(np.sqrt(len(imgs))) assert n * n == len(imgs), 'images array must contain a square number of rows!' s = int(np.sqrt(len(imgs[0]) / channels)) assert s * s == len(imgs[0]) / channels, 'images must be square!' img = np.zeros((s * n, s * n, channels), dtype=imgs[0].dtype) for i, pix in enumerate(imgs): r, c = divmod(i, n) img[r * s:(r+1) * s, c * s:(c+1) * s] = pix.reshape((s, s, channels)) img -= img.min() img /= img.max() ax = plt.gcf().add_subplot(loc) ax.xaxis.set_visible(False) ax.yaxis.set_visible(False) ax.set_frame_on(False) ax.imshow(img.squeeze(), cmap=plt.cm.gray) if title: ax.set_title(title) (t_images, t_labels), (v_images, v_labels) = load_mnist() # construct training/validation sets consisting of the fours. train = t_images[t_labels == 4] valid = v_images[v_labels == 4] N = 20 K = 20 B = 784 x = TT.matrix('x') u = theano.shared(np.random.randn(N * N, K * K).astype(FLOAT), name='u') v = theano.shared(np.random.randn(K * K, B).astype(FLOAT), name='v') err = TT.sqr(x - TT.dot(u, v)).mean() downhill.minimize( loss=err + 100 * (0.01 * abs(u).mean() + (v * v).mean()), params=[u, v], inputs=[x], train=train, valid=valid, batch_size=N * N, monitor_gradients=True, monitors=[ ('err', err), ('u<-0.5', (u < -0.5).mean()), ('u<-0.1', (u < -0.1).mean()), ('u<0.1', (u < 0.1).mean()), ('u<0.5', (u < 0.5).mean()), ], algo='sgd', max_gradient_clip=1, learning_rate=0.5, momentum=0.9, patience=3, min_improvement=0.1, ) plot_images(v.get_value(), 121) plot_images(np.dot(u.get_value(), v.get_value()), 122) plt.show()
lmjohns3/downhill
downhill/base.py
build
python
def build(algo, loss, params=None, inputs=None, updates=(), monitors=(), monitor_gradients=False): '''Construct an optimizer by name. Parameters ---------- algo : str The name of the optimization algorithm to build. loss : Theano expression Loss function to minimize. This must be a scalar-valued expression. params : list of Theano variables, optional Symbolic variables to adjust to minimize the loss. If not given, these will be computed automatically by walking the computation graph. inputs : list of Theano variables, optional Symbolic variables required to compute the loss. If not given, these will be computed automatically by walking the computation graph. updates : list of update pairs, optional A list of pairs providing updates for the internal of the loss computation. Normally this is empty, but it can be provided if the loss, for example, requires an update to an internal random number generator. monitors : dict or sequence of (str, Theano expression) tuples, optional Additional values to monitor during optimization. These must be provided as either a sequence of (name, expression) tuples, or as a dictionary mapping string names to Theano expressions. monitor_gradients : bool, optional If True, add monitors to log the norms of the parameter gradients during optimization. Defaults to False. Returns ------- optimizer : :class:`Optimizer` An optimizer instance. ''' return Optimizer.build(algo, loss, params, inputs, updates=updates, monitors=monitors, monitor_gradients=monitor_gradients)
Construct an optimizer by name. Parameters ---------- algo : str The name of the optimization algorithm to build. loss : Theano expression Loss function to minimize. This must be a scalar-valued expression. params : list of Theano variables, optional Symbolic variables to adjust to minimize the loss. If not given, these will be computed automatically by walking the computation graph. inputs : list of Theano variables, optional Symbolic variables required to compute the loss. If not given, these will be computed automatically by walking the computation graph. updates : list of update pairs, optional A list of pairs providing updates for the internal of the loss computation. Normally this is empty, but it can be provided if the loss, for example, requires an update to an internal random number generator. monitors : dict or sequence of (str, Theano expression) tuples, optional Additional values to monitor during optimization. These must be provided as either a sequence of (name, expression) tuples, or as a dictionary mapping string names to Theano expressions. monitor_gradients : bool, optional If True, add monitors to log the norms of the parameter gradients during optimization. Defaults to False. Returns ------- optimizer : :class:`Optimizer` An optimizer instance.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/base.py#L15-L50
null
# -*- coding: utf-8 -*- '''This module defines a base class for optimization techniques.''' import click import collections import numpy as np import theano import theano.tensor as TT import warnings from . import util class Optimizer(util.Registrar(str('Base'), (), {})): '''An optimizer computes gradient updates to iteratively optimize a loss. Attributes ---------- patience : int, optional Number of validation "failures" that we are willing to tolerate before stopping the optimization process. A validation failure happens whenever the loss on the validation dataset decreases by less than ``min_improvement`` (relative) over the previous best validation loss. Defaults to 5. validate_every : int, optional Evaluate the loss on the validation dataset after making this many passes over the training data. Defaults to 10. min_improvement : float, optional Insist that the validation loss must improve by this relative amount before considering that the optimization has made progress. The optimization process halts when ``patience`` validations have failed to make this relative improvement. Defaults to 0; set to a larger value (e.g., 0.01 for 1% improvement) to halt the optimization process sooner. max_gradient_norm : float, optional Rescale each parameter's gradient so that it has at most this L2 norm. Set to 0 (the default) to disable norm rescaling. If ``max_gradient_elem`` is also specified, then this has no effect. max_gradient_elem : float, optional Perform elementwise clipping on the magnitude of gradient values. Set to 0 (the default) to disable. If elementwise clipping is enabled, norm rescaling (via ``max_gradient_norm``) will have no effect. Deprecated synonyms of this parameter are "max_gradient_clip" and "gradient_clip". learning_rate : float, optional Many SGD-based optimization algorithms require a learning rate hyperparameter that scales the gradient step. Defaults to 1e-4. momentum : float, optional Apply momentum to the parameter updates for this optimizer, with the given strength. Typically this value ranges from 0 (no momentum) to :math:`1 - \epsilon` (large momentum). Defaults to 0. nesterov : bool, optional If True, and ``momentum`` is nonzero, apply Nesterov-style momentum to parameter updates for this optimizer. If False, and ``momentum`` is nonzero, "regular" momentum is applied. Has no effect if ``momentum`` is zero. See :class:`NAG <downhill.NAG>` for a description of Nesterov momentum. Parameters ---------- loss : Theano expression Loss function to minimize. This must be a scalar-valued expression. params : list of Theano variables, optional Symbolic variables to adjust to minimize the loss. If not given, these will be computed automatically by walking the computation graph. inputs : list of Theano variables, optional Symbolic variables required to compute the loss. If not given, these will be computed automatically by walking the computation graph. updates : list of update pairs, optional A list of pairs providing updates for the internals of the loss computation. Normally this is empty, but it can be provided if the loss, for example, requires an update to an internal random number generator. monitors : sequence of (str, Theano expression) tuples, optional Additional values to monitor during optimization. These must be provided as a sequence of (name, expression) tuples. monitor_gradients : bool, optional If True, add monitors to log the norms of the parameter gradients during optimization. Defaults to False. ''' def __init__(self, loss, params=None, inputs=None, updates=(), monitors=(), monitor_gradients=False): inputs_, params_ = util.find_inputs_and_params(loss) self._loss = loss self._params = params or params_ self._inputs = inputs or inputs_ self._updates = updates self._shapes = [p.get_value(borrow=True).shape for p in self._params] self._counts = [np.prod(s) for s in self._shapes] self._starts = np.cumsum([0] + self._counts)[:-1] self._dtype = self._params[0].get_value().dtype self._curr_iter = 0 self._best_iter = 0 self._best_loss = 1e100 self._best_params = [p.get_value().copy() for p in self._params] self._monitor_exprs = [self._loss] self._monitor_names = ['loss'] for name, monitor in monitors: self._monitor_names.append(name) self._monitor_exprs.append(monitor) if monitor_gradients: unnamed = 0 for p, g in zip(self._params, TT.grad(self._loss, self._params)): name = p.name if not name: name = 'unnamed{}'.format(unnamed) unnamed += 1 util.log('"{}" unnamed, will be "{}" internally'.format(p, name)) self._monitor_names.append('grad({})'.format(name)) self._monitor_exprs.append((g * g).sum()) def _compile(self, **kwargs): '''Compile the Theano functions for evaluating and updating our model. ''' util.log('compiling evaluation function') self.f_eval = theano.function(self._inputs, self._monitor_exprs, updates=self._updates, name='evaluation') label = self.__class__.__name__ util.log('compiling {} optimizer'.format(click.style(label, fg='red'))) updates = list(self._updates) + list(self.get_updates(**kwargs)) self.f_step = theano.function(self._inputs, self._monitor_exprs, updates=updates, name=label) def get_updates(self, **kwargs): '''Get parameter update expressions for performing optimization. Keyword arguments can be applied here to set any of the global optimizer attributes. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' self._prepare(**kwargs) for param, grad in self._differentiate(): for var, update in self._get_updates_for(param, grad): # For auxiliary variables, updates are meant to replace the # existing variable value. if var != param: yield var, update continue # If momentum is disabled, just apply the parameter delta. if self.momentum == 0: yield var, param - update continue # Momentum is enabled, so we keep track of velocity here. vel_tm1 = util.shared_like(param, 'vel') vel_t = util.as_float(self.momentum) * vel_tm1 - update if self.nesterov: # see http://arxiv.org/pdf/1212.0901v2.pdf (eq 7) and # https://github.com/lisa-lab/pylearn2/pull/136#issuecomment-10381617 mom_sqr = util.as_float(self.momentum ** 2) mom_inc = util.as_float(1 + self.momentum) vel_t = mom_sqr * vel_tm1 - mom_inc * update yield vel_tm1, vel_t yield param, param + vel_t def _get_updates_for(self, param, grad): '''Generate some update pairs for the given model parameter. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' raise NotImplementedError def _differentiate(self, params=None): '''Return a sequence of gradients for our parameters. If this optimizer has been configured with a gradient norm limit, or with elementwise gradient clipping, this method applies the appropriate rescaling and clipping operations before returning the gradient. Parameters ---------- params : list of Theano variables, optional Return the gradient with respect to these parameters. Defaults to all parameters that the optimizer knows about. Yields ------ pairs : (param, grad) tuples Generates a sequence of tuples representing each of the parameters requested and the corresponding Theano gradient expressions. ''' if params is None: params = self._params for param, grad in zip(params, TT.grad(self._loss, params)): if self.max_gradient_elem > 0: limit = util.as_float(self.max_gradient_elem) yield param, TT.clip(grad, -limit, limit) elif self.max_gradient_norm > 0: norm = TT.sqrt((grad * grad).sum()) limit = util.as_float(self.max_gradient_norm) yield param, grad * TT.minimum(1, limit / norm) else: yield param, grad def set_params(self, targets=None): '''Set the values of the parameters to the given target values. Parameters ---------- targets : sequence of ndarray, optional Arrays for setting the parameters of our model. If this is not provided, the current best parameters for this optimizer will be used. ''' if not isinstance(targets, (list, tuple)): targets = self._best_params for param, target in zip(self._params, targets): param.set_value(target) def _log(self, monitors, iteration, label='', suffix=''): '''Log the state of the optimizer on the console. Parameters ---------- monitors : OrderedDict A dictionary of monitor names mapped to values. These names and values are what is being logged. iteration : int Optimization iteration that we are logging. label : str, optional A label for the name of the optimizer creating the log line. Defaults to the name of the current class. suffix : str, optional A suffix to add to the end of the log line, if any. ''' label = label or self.__class__.__name__ fields = (('{}={:.6f}').format(k, v) for k, v in monitors.items()) util.log('{} {} {}{}'.format(label, iteration, ' '.join(fields), suffix)) def evaluate(self, dataset): '''Evaluate the current model parameters on a dataset. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A set of data to use for evaluating the model. Returns ------- monitors : OrderedDict A dictionary mapping monitor names to values. Monitors are quantities of interest during optimization---for example, loss function, accuracy, or whatever the optimization task requires. ''' if dataset is None: values = [self.f_eval()] else: values = [self.f_eval(*x) for x in dataset] monitors = zip(self._monitor_names, np.mean(values, axis=0)) return collections.OrderedDict(monitors) def _test_patience(self, monitors): '''Test whether our patience with optimization has elapsed. Parameters ---------- monitors : dict A dictionary mapping monitor names to values. The 'loss' key from this dictionary will be used to evaluate optimization progress. Returns ------- elapsed : bool True iff our patience has elapsed and the model is no longer improving. ''' self._curr_iter += 1 marker = '' loss = monitors['loss'] if self._best_loss - loss > self._best_loss * self.min_improvement: self._best_loss = loss self._best_iter = self._curr_iter self._best_params = [p.get_value().copy() for p in self._params] marker = ' *' self._log(monitors, self._curr_iter - 1, 'validation', marker) return self._curr_iter - self._best_iter > self.patience def _prepare(self, **kwargs): '''Set up properties for optimization. This method can be overridden by base classes to provide parameters that are specific to a particular optimization technique (e.g., setting up a learning rate value). ''' self.learning_rate = util.as_float(kwargs.pop('learning_rate', 1e-4)) self.momentum = kwargs.pop('momentum', 0) self.nesterov = kwargs.pop('nesterov', False) self.patience = kwargs.get('patience', 5) self.validate_every = kwargs.pop('validate_every', 10) self.min_improvement = kwargs.pop('min_improvement', 0) self.max_gradient_norm = kwargs.pop('max_gradient_norm', 0) self.max_gradient_elem = kwargs.pop('max_gradient_elem', 0) util.log_param('patience', self.patience) util.log_param('validate_every', self.validate_every) util.log_param('min_improvement', self.min_improvement) util.log_param('max_gradient_norm', self.max_gradient_norm) util.log_param('max_gradient_elem', self.max_gradient_elem) util.log_param('learning_rate', self.learning_rate) util.log_param('momentum', self.momentum) util.log_param('nesterov', self.nesterov) def iterate(self, train=None, valid=None, max_updates=None, **kwargs): r'''Optimize a loss iteratively using a training and validation dataset. This method yields a series of monitor values to the caller. After every optimization epoch, a pair of monitor dictionaries is generated: one evaluated on the training dataset during the epoch, and another evaluated on the validation dataset at the most recent validation epoch. The validation monitors might not be updated during every optimization iteration; in this case, the most recent validation monitors will be yielded along with the training monitors. Additional keyword arguments supplied here will set the global optimizer attributes. Parameters ---------- train : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of training data for computing updates to model parameters. valid : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of validation data for computing monitor values and determining when the loss has stopped improving. Defaults to the training data. max_updates : int, optional If specified, halt optimization after this many gradient updates have been processed. If not provided, uses early stopping to decide when to halt. Yields ------ train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' self._compile(**kwargs) if valid is None: valid = train iteration = 0 training = validation = None while max_updates is None or iteration < max_updates: if not iteration % self.validate_every: try: validation = self.evaluate(valid) except KeyboardInterrupt: util.log('interrupted!') break if self._test_patience(validation): util.log('patience elapsed!') break try: training = self._step(train) except KeyboardInterrupt: util.log('interrupted!') break iteration += 1 self._log(training, iteration) yield training, validation self.set_params('best') def minimize(self, *args, **kwargs): '''Optimize our loss exhaustively. This method is a thin wrapper over the :func:`iterate` method. It simply exhausts the iterative optimization process and returns the final monitor values. Returns ------- train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' monitors = None for monitors in self.iterate(*args, **kwargs): pass return monitors def _step(self, dataset): '''Advance the state of the optimizer by one step. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A dataset for optimizing the model. Returns ------- train_monitors : dict A dictionary mapping monitor names to values. ''' if dataset is None: values = [self.f_step()] else: values = [self.f_step(*x) for x in dataset] return collections.OrderedDict( zip(self._monitor_names, np.mean(values, axis=0)))
lmjohns3/downhill
downhill/base.py
Optimizer._compile
python
def _compile(self, **kwargs): '''Compile the Theano functions for evaluating and updating our model. ''' util.log('compiling evaluation function') self.f_eval = theano.function(self._inputs, self._monitor_exprs, updates=self._updates, name='evaluation') label = self.__class__.__name__ util.log('compiling {} optimizer'.format(click.style(label, fg='red'))) updates = list(self._updates) + list(self.get_updates(**kwargs)) self.f_step = theano.function(self._inputs, self._monitor_exprs, updates=updates, name=label)
Compile the Theano functions for evaluating and updating our model.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/base.py#L153-L167
null
class Optimizer(util.Registrar(str('Base'), (), {})): '''An optimizer computes gradient updates to iteratively optimize a loss. Attributes ---------- patience : int, optional Number of validation "failures" that we are willing to tolerate before stopping the optimization process. A validation failure happens whenever the loss on the validation dataset decreases by less than ``min_improvement`` (relative) over the previous best validation loss. Defaults to 5. validate_every : int, optional Evaluate the loss on the validation dataset after making this many passes over the training data. Defaults to 10. min_improvement : float, optional Insist that the validation loss must improve by this relative amount before considering that the optimization has made progress. The optimization process halts when ``patience`` validations have failed to make this relative improvement. Defaults to 0; set to a larger value (e.g., 0.01 for 1% improvement) to halt the optimization process sooner. max_gradient_norm : float, optional Rescale each parameter's gradient so that it has at most this L2 norm. Set to 0 (the default) to disable norm rescaling. If ``max_gradient_elem`` is also specified, then this has no effect. max_gradient_elem : float, optional Perform elementwise clipping on the magnitude of gradient values. Set to 0 (the default) to disable. If elementwise clipping is enabled, norm rescaling (via ``max_gradient_norm``) will have no effect. Deprecated synonyms of this parameter are "max_gradient_clip" and "gradient_clip". learning_rate : float, optional Many SGD-based optimization algorithms require a learning rate hyperparameter that scales the gradient step. Defaults to 1e-4. momentum : float, optional Apply momentum to the parameter updates for this optimizer, with the given strength. Typically this value ranges from 0 (no momentum) to :math:`1 - \epsilon` (large momentum). Defaults to 0. nesterov : bool, optional If True, and ``momentum`` is nonzero, apply Nesterov-style momentum to parameter updates for this optimizer. If False, and ``momentum`` is nonzero, "regular" momentum is applied. Has no effect if ``momentum`` is zero. See :class:`NAG <downhill.NAG>` for a description of Nesterov momentum. Parameters ---------- loss : Theano expression Loss function to minimize. This must be a scalar-valued expression. params : list of Theano variables, optional Symbolic variables to adjust to minimize the loss. If not given, these will be computed automatically by walking the computation graph. inputs : list of Theano variables, optional Symbolic variables required to compute the loss. If not given, these will be computed automatically by walking the computation graph. updates : list of update pairs, optional A list of pairs providing updates for the internals of the loss computation. Normally this is empty, but it can be provided if the loss, for example, requires an update to an internal random number generator. monitors : sequence of (str, Theano expression) tuples, optional Additional values to monitor during optimization. These must be provided as a sequence of (name, expression) tuples. monitor_gradients : bool, optional If True, add monitors to log the norms of the parameter gradients during optimization. Defaults to False. ''' def __init__(self, loss, params=None, inputs=None, updates=(), monitors=(), monitor_gradients=False): inputs_, params_ = util.find_inputs_and_params(loss) self._loss = loss self._params = params or params_ self._inputs = inputs or inputs_ self._updates = updates self._shapes = [p.get_value(borrow=True).shape for p in self._params] self._counts = [np.prod(s) for s in self._shapes] self._starts = np.cumsum([0] + self._counts)[:-1] self._dtype = self._params[0].get_value().dtype self._curr_iter = 0 self._best_iter = 0 self._best_loss = 1e100 self._best_params = [p.get_value().copy() for p in self._params] self._monitor_exprs = [self._loss] self._monitor_names = ['loss'] for name, monitor in monitors: self._monitor_names.append(name) self._monitor_exprs.append(monitor) if monitor_gradients: unnamed = 0 for p, g in zip(self._params, TT.grad(self._loss, self._params)): name = p.name if not name: name = 'unnamed{}'.format(unnamed) unnamed += 1 util.log('"{}" unnamed, will be "{}" internally'.format(p, name)) self._monitor_names.append('grad({})'.format(name)) self._monitor_exprs.append((g * g).sum()) def get_updates(self, **kwargs): '''Get parameter update expressions for performing optimization. Keyword arguments can be applied here to set any of the global optimizer attributes. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' self._prepare(**kwargs) for param, grad in self._differentiate(): for var, update in self._get_updates_for(param, grad): # For auxiliary variables, updates are meant to replace the # existing variable value. if var != param: yield var, update continue # If momentum is disabled, just apply the parameter delta. if self.momentum == 0: yield var, param - update continue # Momentum is enabled, so we keep track of velocity here. vel_tm1 = util.shared_like(param, 'vel') vel_t = util.as_float(self.momentum) * vel_tm1 - update if self.nesterov: # see http://arxiv.org/pdf/1212.0901v2.pdf (eq 7) and # https://github.com/lisa-lab/pylearn2/pull/136#issuecomment-10381617 mom_sqr = util.as_float(self.momentum ** 2) mom_inc = util.as_float(1 + self.momentum) vel_t = mom_sqr * vel_tm1 - mom_inc * update yield vel_tm1, vel_t yield param, param + vel_t def _get_updates_for(self, param, grad): '''Generate some update pairs for the given model parameter. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' raise NotImplementedError def _differentiate(self, params=None): '''Return a sequence of gradients for our parameters. If this optimizer has been configured with a gradient norm limit, or with elementwise gradient clipping, this method applies the appropriate rescaling and clipping operations before returning the gradient. Parameters ---------- params : list of Theano variables, optional Return the gradient with respect to these parameters. Defaults to all parameters that the optimizer knows about. Yields ------ pairs : (param, grad) tuples Generates a sequence of tuples representing each of the parameters requested and the corresponding Theano gradient expressions. ''' if params is None: params = self._params for param, grad in zip(params, TT.grad(self._loss, params)): if self.max_gradient_elem > 0: limit = util.as_float(self.max_gradient_elem) yield param, TT.clip(grad, -limit, limit) elif self.max_gradient_norm > 0: norm = TT.sqrt((grad * grad).sum()) limit = util.as_float(self.max_gradient_norm) yield param, grad * TT.minimum(1, limit / norm) else: yield param, grad def set_params(self, targets=None): '''Set the values of the parameters to the given target values. Parameters ---------- targets : sequence of ndarray, optional Arrays for setting the parameters of our model. If this is not provided, the current best parameters for this optimizer will be used. ''' if not isinstance(targets, (list, tuple)): targets = self._best_params for param, target in zip(self._params, targets): param.set_value(target) def _log(self, monitors, iteration, label='', suffix=''): '''Log the state of the optimizer on the console. Parameters ---------- monitors : OrderedDict A dictionary of monitor names mapped to values. These names and values are what is being logged. iteration : int Optimization iteration that we are logging. label : str, optional A label for the name of the optimizer creating the log line. Defaults to the name of the current class. suffix : str, optional A suffix to add to the end of the log line, if any. ''' label = label or self.__class__.__name__ fields = (('{}={:.6f}').format(k, v) for k, v in monitors.items()) util.log('{} {} {}{}'.format(label, iteration, ' '.join(fields), suffix)) def evaluate(self, dataset): '''Evaluate the current model parameters on a dataset. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A set of data to use for evaluating the model. Returns ------- monitors : OrderedDict A dictionary mapping monitor names to values. Monitors are quantities of interest during optimization---for example, loss function, accuracy, or whatever the optimization task requires. ''' if dataset is None: values = [self.f_eval()] else: values = [self.f_eval(*x) for x in dataset] monitors = zip(self._monitor_names, np.mean(values, axis=0)) return collections.OrderedDict(monitors) def _test_patience(self, monitors): '''Test whether our patience with optimization has elapsed. Parameters ---------- monitors : dict A dictionary mapping monitor names to values. The 'loss' key from this dictionary will be used to evaluate optimization progress. Returns ------- elapsed : bool True iff our patience has elapsed and the model is no longer improving. ''' self._curr_iter += 1 marker = '' loss = monitors['loss'] if self._best_loss - loss > self._best_loss * self.min_improvement: self._best_loss = loss self._best_iter = self._curr_iter self._best_params = [p.get_value().copy() for p in self._params] marker = ' *' self._log(monitors, self._curr_iter - 1, 'validation', marker) return self._curr_iter - self._best_iter > self.patience def _prepare(self, **kwargs): '''Set up properties for optimization. This method can be overridden by base classes to provide parameters that are specific to a particular optimization technique (e.g., setting up a learning rate value). ''' self.learning_rate = util.as_float(kwargs.pop('learning_rate', 1e-4)) self.momentum = kwargs.pop('momentum', 0) self.nesterov = kwargs.pop('nesterov', False) self.patience = kwargs.get('patience', 5) self.validate_every = kwargs.pop('validate_every', 10) self.min_improvement = kwargs.pop('min_improvement', 0) self.max_gradient_norm = kwargs.pop('max_gradient_norm', 0) self.max_gradient_elem = kwargs.pop('max_gradient_elem', 0) util.log_param('patience', self.patience) util.log_param('validate_every', self.validate_every) util.log_param('min_improvement', self.min_improvement) util.log_param('max_gradient_norm', self.max_gradient_norm) util.log_param('max_gradient_elem', self.max_gradient_elem) util.log_param('learning_rate', self.learning_rate) util.log_param('momentum', self.momentum) util.log_param('nesterov', self.nesterov) def iterate(self, train=None, valid=None, max_updates=None, **kwargs): r'''Optimize a loss iteratively using a training and validation dataset. This method yields a series of monitor values to the caller. After every optimization epoch, a pair of monitor dictionaries is generated: one evaluated on the training dataset during the epoch, and another evaluated on the validation dataset at the most recent validation epoch. The validation monitors might not be updated during every optimization iteration; in this case, the most recent validation monitors will be yielded along with the training monitors. Additional keyword arguments supplied here will set the global optimizer attributes. Parameters ---------- train : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of training data for computing updates to model parameters. valid : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of validation data for computing monitor values and determining when the loss has stopped improving. Defaults to the training data. max_updates : int, optional If specified, halt optimization after this many gradient updates have been processed. If not provided, uses early stopping to decide when to halt. Yields ------ train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' self._compile(**kwargs) if valid is None: valid = train iteration = 0 training = validation = None while max_updates is None or iteration < max_updates: if not iteration % self.validate_every: try: validation = self.evaluate(valid) except KeyboardInterrupt: util.log('interrupted!') break if self._test_patience(validation): util.log('patience elapsed!') break try: training = self._step(train) except KeyboardInterrupt: util.log('interrupted!') break iteration += 1 self._log(training, iteration) yield training, validation self.set_params('best') def minimize(self, *args, **kwargs): '''Optimize our loss exhaustively. This method is a thin wrapper over the :func:`iterate` method. It simply exhausts the iterative optimization process and returns the final monitor values. Returns ------- train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' monitors = None for monitors in self.iterate(*args, **kwargs): pass return monitors def _step(self, dataset): '''Advance the state of the optimizer by one step. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A dataset for optimizing the model. Returns ------- train_monitors : dict A dictionary mapping monitor names to values. ''' if dataset is None: values = [self.f_step()] else: values = [self.f_step(*x) for x in dataset] return collections.OrderedDict( zip(self._monitor_names, np.mean(values, axis=0)))
lmjohns3/downhill
downhill/base.py
Optimizer.get_updates
python
def get_updates(self, **kwargs): '''Get parameter update expressions for performing optimization. Keyword arguments can be applied here to set any of the global optimizer attributes. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' self._prepare(**kwargs) for param, grad in self._differentiate(): for var, update in self._get_updates_for(param, grad): # For auxiliary variables, updates are meant to replace the # existing variable value. if var != param: yield var, update continue # If momentum is disabled, just apply the parameter delta. if self.momentum == 0: yield var, param - update continue # Momentum is enabled, so we keep track of velocity here. vel_tm1 = util.shared_like(param, 'vel') vel_t = util.as_float(self.momentum) * vel_tm1 - update if self.nesterov: # see http://arxiv.org/pdf/1212.0901v2.pdf (eq 7) and # https://github.com/lisa-lab/pylearn2/pull/136#issuecomment-10381617 mom_sqr = util.as_float(self.momentum ** 2) mom_inc = util.as_float(1 + self.momentum) vel_t = mom_sqr * vel_tm1 - mom_inc * update yield vel_tm1, vel_t yield param, param + vel_t
Get parameter update expressions for performing optimization. Keyword arguments can be applied here to set any of the global optimizer attributes. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/base.py#L169-L202
[ "def as_float(x):\n '''Cast a floating point value to a Theano ``floatX`` symbol.\n\n Parameters\n ----------\n x : float, ndarray, or Theano expression\n Some quantity to cast to floating point.\n\n Returns\n -------\n x : Theano expression\n A symbolic variable cast as a ``floatX`` value.\n '''\n return TT.cast(x, theano.config.floatX)\n", "def shared_like(param, suffix, init=0):\n '''Create a Theano shared variable like an existing parameter.\n\n Parameters\n ----------\n param : Theano variable\n Theano variable to use for shape information.\n suffix : str\n Suffix to append to the parameter's name for the new variable.\n init : float or ndarray, optional\n Initial value of the shared variable. Defaults to 0.\n\n Returns\n -------\n shared : Theano shared variable\n A new shared variable with the same shape and data type as ``param``.\n '''\n return theano.shared(np.zeros_like(param.get_value()) + init,\n name='{}_{}'.format(param.name, suffix),\n broadcastable=param.broadcastable)\n" ]
class Optimizer(util.Registrar(str('Base'), (), {})): '''An optimizer computes gradient updates to iteratively optimize a loss. Attributes ---------- patience : int, optional Number of validation "failures" that we are willing to tolerate before stopping the optimization process. A validation failure happens whenever the loss on the validation dataset decreases by less than ``min_improvement`` (relative) over the previous best validation loss. Defaults to 5. validate_every : int, optional Evaluate the loss on the validation dataset after making this many passes over the training data. Defaults to 10. min_improvement : float, optional Insist that the validation loss must improve by this relative amount before considering that the optimization has made progress. The optimization process halts when ``patience`` validations have failed to make this relative improvement. Defaults to 0; set to a larger value (e.g., 0.01 for 1% improvement) to halt the optimization process sooner. max_gradient_norm : float, optional Rescale each parameter's gradient so that it has at most this L2 norm. Set to 0 (the default) to disable norm rescaling. If ``max_gradient_elem`` is also specified, then this has no effect. max_gradient_elem : float, optional Perform elementwise clipping on the magnitude of gradient values. Set to 0 (the default) to disable. If elementwise clipping is enabled, norm rescaling (via ``max_gradient_norm``) will have no effect. Deprecated synonyms of this parameter are "max_gradient_clip" and "gradient_clip". learning_rate : float, optional Many SGD-based optimization algorithms require a learning rate hyperparameter that scales the gradient step. Defaults to 1e-4. momentum : float, optional Apply momentum to the parameter updates for this optimizer, with the given strength. Typically this value ranges from 0 (no momentum) to :math:`1 - \epsilon` (large momentum). Defaults to 0. nesterov : bool, optional If True, and ``momentum`` is nonzero, apply Nesterov-style momentum to parameter updates for this optimizer. If False, and ``momentum`` is nonzero, "regular" momentum is applied. Has no effect if ``momentum`` is zero. See :class:`NAG <downhill.NAG>` for a description of Nesterov momentum. Parameters ---------- loss : Theano expression Loss function to minimize. This must be a scalar-valued expression. params : list of Theano variables, optional Symbolic variables to adjust to minimize the loss. If not given, these will be computed automatically by walking the computation graph. inputs : list of Theano variables, optional Symbolic variables required to compute the loss. If not given, these will be computed automatically by walking the computation graph. updates : list of update pairs, optional A list of pairs providing updates for the internals of the loss computation. Normally this is empty, but it can be provided if the loss, for example, requires an update to an internal random number generator. monitors : sequence of (str, Theano expression) tuples, optional Additional values to monitor during optimization. These must be provided as a sequence of (name, expression) tuples. monitor_gradients : bool, optional If True, add monitors to log the norms of the parameter gradients during optimization. Defaults to False. ''' def __init__(self, loss, params=None, inputs=None, updates=(), monitors=(), monitor_gradients=False): inputs_, params_ = util.find_inputs_and_params(loss) self._loss = loss self._params = params or params_ self._inputs = inputs or inputs_ self._updates = updates self._shapes = [p.get_value(borrow=True).shape for p in self._params] self._counts = [np.prod(s) for s in self._shapes] self._starts = np.cumsum([0] + self._counts)[:-1] self._dtype = self._params[0].get_value().dtype self._curr_iter = 0 self._best_iter = 0 self._best_loss = 1e100 self._best_params = [p.get_value().copy() for p in self._params] self._monitor_exprs = [self._loss] self._monitor_names = ['loss'] for name, monitor in monitors: self._monitor_names.append(name) self._monitor_exprs.append(monitor) if monitor_gradients: unnamed = 0 for p, g in zip(self._params, TT.grad(self._loss, self._params)): name = p.name if not name: name = 'unnamed{}'.format(unnamed) unnamed += 1 util.log('"{}" unnamed, will be "{}" internally'.format(p, name)) self._monitor_names.append('grad({})'.format(name)) self._monitor_exprs.append((g * g).sum()) def _compile(self, **kwargs): '''Compile the Theano functions for evaluating and updating our model. ''' util.log('compiling evaluation function') self.f_eval = theano.function(self._inputs, self._monitor_exprs, updates=self._updates, name='evaluation') label = self.__class__.__name__ util.log('compiling {} optimizer'.format(click.style(label, fg='red'))) updates = list(self._updates) + list(self.get_updates(**kwargs)) self.f_step = theano.function(self._inputs, self._monitor_exprs, updates=updates, name=label) def _get_updates_for(self, param, grad): '''Generate some update pairs for the given model parameter. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' raise NotImplementedError def _differentiate(self, params=None): '''Return a sequence of gradients for our parameters. If this optimizer has been configured with a gradient norm limit, or with elementwise gradient clipping, this method applies the appropriate rescaling and clipping operations before returning the gradient. Parameters ---------- params : list of Theano variables, optional Return the gradient with respect to these parameters. Defaults to all parameters that the optimizer knows about. Yields ------ pairs : (param, grad) tuples Generates a sequence of tuples representing each of the parameters requested and the corresponding Theano gradient expressions. ''' if params is None: params = self._params for param, grad in zip(params, TT.grad(self._loss, params)): if self.max_gradient_elem > 0: limit = util.as_float(self.max_gradient_elem) yield param, TT.clip(grad, -limit, limit) elif self.max_gradient_norm > 0: norm = TT.sqrt((grad * grad).sum()) limit = util.as_float(self.max_gradient_norm) yield param, grad * TT.minimum(1, limit / norm) else: yield param, grad def set_params(self, targets=None): '''Set the values of the parameters to the given target values. Parameters ---------- targets : sequence of ndarray, optional Arrays for setting the parameters of our model. If this is not provided, the current best parameters for this optimizer will be used. ''' if not isinstance(targets, (list, tuple)): targets = self._best_params for param, target in zip(self._params, targets): param.set_value(target) def _log(self, monitors, iteration, label='', suffix=''): '''Log the state of the optimizer on the console. Parameters ---------- monitors : OrderedDict A dictionary of monitor names mapped to values. These names and values are what is being logged. iteration : int Optimization iteration that we are logging. label : str, optional A label for the name of the optimizer creating the log line. Defaults to the name of the current class. suffix : str, optional A suffix to add to the end of the log line, if any. ''' label = label or self.__class__.__name__ fields = (('{}={:.6f}').format(k, v) for k, v in monitors.items()) util.log('{} {} {}{}'.format(label, iteration, ' '.join(fields), suffix)) def evaluate(self, dataset): '''Evaluate the current model parameters on a dataset. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A set of data to use for evaluating the model. Returns ------- monitors : OrderedDict A dictionary mapping monitor names to values. Monitors are quantities of interest during optimization---for example, loss function, accuracy, or whatever the optimization task requires. ''' if dataset is None: values = [self.f_eval()] else: values = [self.f_eval(*x) for x in dataset] monitors = zip(self._monitor_names, np.mean(values, axis=0)) return collections.OrderedDict(monitors) def _test_patience(self, monitors): '''Test whether our patience with optimization has elapsed. Parameters ---------- monitors : dict A dictionary mapping monitor names to values. The 'loss' key from this dictionary will be used to evaluate optimization progress. Returns ------- elapsed : bool True iff our patience has elapsed and the model is no longer improving. ''' self._curr_iter += 1 marker = '' loss = monitors['loss'] if self._best_loss - loss > self._best_loss * self.min_improvement: self._best_loss = loss self._best_iter = self._curr_iter self._best_params = [p.get_value().copy() for p in self._params] marker = ' *' self._log(monitors, self._curr_iter - 1, 'validation', marker) return self._curr_iter - self._best_iter > self.patience def _prepare(self, **kwargs): '''Set up properties for optimization. This method can be overridden by base classes to provide parameters that are specific to a particular optimization technique (e.g., setting up a learning rate value). ''' self.learning_rate = util.as_float(kwargs.pop('learning_rate', 1e-4)) self.momentum = kwargs.pop('momentum', 0) self.nesterov = kwargs.pop('nesterov', False) self.patience = kwargs.get('patience', 5) self.validate_every = kwargs.pop('validate_every', 10) self.min_improvement = kwargs.pop('min_improvement', 0) self.max_gradient_norm = kwargs.pop('max_gradient_norm', 0) self.max_gradient_elem = kwargs.pop('max_gradient_elem', 0) util.log_param('patience', self.patience) util.log_param('validate_every', self.validate_every) util.log_param('min_improvement', self.min_improvement) util.log_param('max_gradient_norm', self.max_gradient_norm) util.log_param('max_gradient_elem', self.max_gradient_elem) util.log_param('learning_rate', self.learning_rate) util.log_param('momentum', self.momentum) util.log_param('nesterov', self.nesterov) def iterate(self, train=None, valid=None, max_updates=None, **kwargs): r'''Optimize a loss iteratively using a training and validation dataset. This method yields a series of monitor values to the caller. After every optimization epoch, a pair of monitor dictionaries is generated: one evaluated on the training dataset during the epoch, and another evaluated on the validation dataset at the most recent validation epoch. The validation monitors might not be updated during every optimization iteration; in this case, the most recent validation monitors will be yielded along with the training monitors. Additional keyword arguments supplied here will set the global optimizer attributes. Parameters ---------- train : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of training data for computing updates to model parameters. valid : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of validation data for computing monitor values and determining when the loss has stopped improving. Defaults to the training data. max_updates : int, optional If specified, halt optimization after this many gradient updates have been processed. If not provided, uses early stopping to decide when to halt. Yields ------ train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' self._compile(**kwargs) if valid is None: valid = train iteration = 0 training = validation = None while max_updates is None or iteration < max_updates: if not iteration % self.validate_every: try: validation = self.evaluate(valid) except KeyboardInterrupt: util.log('interrupted!') break if self._test_patience(validation): util.log('patience elapsed!') break try: training = self._step(train) except KeyboardInterrupt: util.log('interrupted!') break iteration += 1 self._log(training, iteration) yield training, validation self.set_params('best') def minimize(self, *args, **kwargs): '''Optimize our loss exhaustively. This method is a thin wrapper over the :func:`iterate` method. It simply exhausts the iterative optimization process and returns the final monitor values. Returns ------- train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' monitors = None for monitors in self.iterate(*args, **kwargs): pass return monitors def _step(self, dataset): '''Advance the state of the optimizer by one step. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A dataset for optimizing the model. Returns ------- train_monitors : dict A dictionary mapping monitor names to values. ''' if dataset is None: values = [self.f_step()] else: values = [self.f_step(*x) for x in dataset] return collections.OrderedDict( zip(self._monitor_names, np.mean(values, axis=0)))
lmjohns3/downhill
downhill/base.py
Optimizer._differentiate
python
def _differentiate(self, params=None): '''Return a sequence of gradients for our parameters. If this optimizer has been configured with a gradient norm limit, or with elementwise gradient clipping, this method applies the appropriate rescaling and clipping operations before returning the gradient. Parameters ---------- params : list of Theano variables, optional Return the gradient with respect to these parameters. Defaults to all parameters that the optimizer knows about. Yields ------ pairs : (param, grad) tuples Generates a sequence of tuples representing each of the parameters requested and the corresponding Theano gradient expressions. ''' if params is None: params = self._params for param, grad in zip(params, TT.grad(self._loss, params)): if self.max_gradient_elem > 0: limit = util.as_float(self.max_gradient_elem) yield param, TT.clip(grad, -limit, limit) elif self.max_gradient_norm > 0: norm = TT.sqrt((grad * grad).sum()) limit = util.as_float(self.max_gradient_norm) yield param, grad * TT.minimum(1, limit / norm) else: yield param, grad
Return a sequence of gradients for our parameters. If this optimizer has been configured with a gradient norm limit, or with elementwise gradient clipping, this method applies the appropriate rescaling and clipping operations before returning the gradient. Parameters ---------- params : list of Theano variables, optional Return the gradient with respect to these parameters. Defaults to all parameters that the optimizer knows about. Yields ------ pairs : (param, grad) tuples Generates a sequence of tuples representing each of the parameters requested and the corresponding Theano gradient expressions.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/base.py#L214-L244
null
class Optimizer(util.Registrar(str('Base'), (), {})): '''An optimizer computes gradient updates to iteratively optimize a loss. Attributes ---------- patience : int, optional Number of validation "failures" that we are willing to tolerate before stopping the optimization process. A validation failure happens whenever the loss on the validation dataset decreases by less than ``min_improvement`` (relative) over the previous best validation loss. Defaults to 5. validate_every : int, optional Evaluate the loss on the validation dataset after making this many passes over the training data. Defaults to 10. min_improvement : float, optional Insist that the validation loss must improve by this relative amount before considering that the optimization has made progress. The optimization process halts when ``patience`` validations have failed to make this relative improvement. Defaults to 0; set to a larger value (e.g., 0.01 for 1% improvement) to halt the optimization process sooner. max_gradient_norm : float, optional Rescale each parameter's gradient so that it has at most this L2 norm. Set to 0 (the default) to disable norm rescaling. If ``max_gradient_elem`` is also specified, then this has no effect. max_gradient_elem : float, optional Perform elementwise clipping on the magnitude of gradient values. Set to 0 (the default) to disable. If elementwise clipping is enabled, norm rescaling (via ``max_gradient_norm``) will have no effect. Deprecated synonyms of this parameter are "max_gradient_clip" and "gradient_clip". learning_rate : float, optional Many SGD-based optimization algorithms require a learning rate hyperparameter that scales the gradient step. Defaults to 1e-4. momentum : float, optional Apply momentum to the parameter updates for this optimizer, with the given strength. Typically this value ranges from 0 (no momentum) to :math:`1 - \epsilon` (large momentum). Defaults to 0. nesterov : bool, optional If True, and ``momentum`` is nonzero, apply Nesterov-style momentum to parameter updates for this optimizer. If False, and ``momentum`` is nonzero, "regular" momentum is applied. Has no effect if ``momentum`` is zero. See :class:`NAG <downhill.NAG>` for a description of Nesterov momentum. Parameters ---------- loss : Theano expression Loss function to minimize. This must be a scalar-valued expression. params : list of Theano variables, optional Symbolic variables to adjust to minimize the loss. If not given, these will be computed automatically by walking the computation graph. inputs : list of Theano variables, optional Symbolic variables required to compute the loss. If not given, these will be computed automatically by walking the computation graph. updates : list of update pairs, optional A list of pairs providing updates for the internals of the loss computation. Normally this is empty, but it can be provided if the loss, for example, requires an update to an internal random number generator. monitors : sequence of (str, Theano expression) tuples, optional Additional values to monitor during optimization. These must be provided as a sequence of (name, expression) tuples. monitor_gradients : bool, optional If True, add monitors to log the norms of the parameter gradients during optimization. Defaults to False. ''' def __init__(self, loss, params=None, inputs=None, updates=(), monitors=(), monitor_gradients=False): inputs_, params_ = util.find_inputs_and_params(loss) self._loss = loss self._params = params or params_ self._inputs = inputs or inputs_ self._updates = updates self._shapes = [p.get_value(borrow=True).shape for p in self._params] self._counts = [np.prod(s) for s in self._shapes] self._starts = np.cumsum([0] + self._counts)[:-1] self._dtype = self._params[0].get_value().dtype self._curr_iter = 0 self._best_iter = 0 self._best_loss = 1e100 self._best_params = [p.get_value().copy() for p in self._params] self._monitor_exprs = [self._loss] self._monitor_names = ['loss'] for name, monitor in monitors: self._monitor_names.append(name) self._monitor_exprs.append(monitor) if monitor_gradients: unnamed = 0 for p, g in zip(self._params, TT.grad(self._loss, self._params)): name = p.name if not name: name = 'unnamed{}'.format(unnamed) unnamed += 1 util.log('"{}" unnamed, will be "{}" internally'.format(p, name)) self._monitor_names.append('grad({})'.format(name)) self._monitor_exprs.append((g * g).sum()) def _compile(self, **kwargs): '''Compile the Theano functions for evaluating and updating our model. ''' util.log('compiling evaluation function') self.f_eval = theano.function(self._inputs, self._monitor_exprs, updates=self._updates, name='evaluation') label = self.__class__.__name__ util.log('compiling {} optimizer'.format(click.style(label, fg='red'))) updates = list(self._updates) + list(self.get_updates(**kwargs)) self.f_step = theano.function(self._inputs, self._monitor_exprs, updates=updates, name=label) def get_updates(self, **kwargs): '''Get parameter update expressions for performing optimization. Keyword arguments can be applied here to set any of the global optimizer attributes. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' self._prepare(**kwargs) for param, grad in self._differentiate(): for var, update in self._get_updates_for(param, grad): # For auxiliary variables, updates are meant to replace the # existing variable value. if var != param: yield var, update continue # If momentum is disabled, just apply the parameter delta. if self.momentum == 0: yield var, param - update continue # Momentum is enabled, so we keep track of velocity here. vel_tm1 = util.shared_like(param, 'vel') vel_t = util.as_float(self.momentum) * vel_tm1 - update if self.nesterov: # see http://arxiv.org/pdf/1212.0901v2.pdf (eq 7) and # https://github.com/lisa-lab/pylearn2/pull/136#issuecomment-10381617 mom_sqr = util.as_float(self.momentum ** 2) mom_inc = util.as_float(1 + self.momentum) vel_t = mom_sqr * vel_tm1 - mom_inc * update yield vel_tm1, vel_t yield param, param + vel_t def _get_updates_for(self, param, grad): '''Generate some update pairs for the given model parameter. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' raise NotImplementedError def set_params(self, targets=None): '''Set the values of the parameters to the given target values. Parameters ---------- targets : sequence of ndarray, optional Arrays for setting the parameters of our model. If this is not provided, the current best parameters for this optimizer will be used. ''' if not isinstance(targets, (list, tuple)): targets = self._best_params for param, target in zip(self._params, targets): param.set_value(target) def _log(self, monitors, iteration, label='', suffix=''): '''Log the state of the optimizer on the console. Parameters ---------- monitors : OrderedDict A dictionary of monitor names mapped to values. These names and values are what is being logged. iteration : int Optimization iteration that we are logging. label : str, optional A label for the name of the optimizer creating the log line. Defaults to the name of the current class. suffix : str, optional A suffix to add to the end of the log line, if any. ''' label = label or self.__class__.__name__ fields = (('{}={:.6f}').format(k, v) for k, v in monitors.items()) util.log('{} {} {}{}'.format(label, iteration, ' '.join(fields), suffix)) def evaluate(self, dataset): '''Evaluate the current model parameters on a dataset. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A set of data to use for evaluating the model. Returns ------- monitors : OrderedDict A dictionary mapping monitor names to values. Monitors are quantities of interest during optimization---for example, loss function, accuracy, or whatever the optimization task requires. ''' if dataset is None: values = [self.f_eval()] else: values = [self.f_eval(*x) for x in dataset] monitors = zip(self._monitor_names, np.mean(values, axis=0)) return collections.OrderedDict(monitors) def _test_patience(self, monitors): '''Test whether our patience with optimization has elapsed. Parameters ---------- monitors : dict A dictionary mapping monitor names to values. The 'loss' key from this dictionary will be used to evaluate optimization progress. Returns ------- elapsed : bool True iff our patience has elapsed and the model is no longer improving. ''' self._curr_iter += 1 marker = '' loss = monitors['loss'] if self._best_loss - loss > self._best_loss * self.min_improvement: self._best_loss = loss self._best_iter = self._curr_iter self._best_params = [p.get_value().copy() for p in self._params] marker = ' *' self._log(monitors, self._curr_iter - 1, 'validation', marker) return self._curr_iter - self._best_iter > self.patience def _prepare(self, **kwargs): '''Set up properties for optimization. This method can be overridden by base classes to provide parameters that are specific to a particular optimization technique (e.g., setting up a learning rate value). ''' self.learning_rate = util.as_float(kwargs.pop('learning_rate', 1e-4)) self.momentum = kwargs.pop('momentum', 0) self.nesterov = kwargs.pop('nesterov', False) self.patience = kwargs.get('patience', 5) self.validate_every = kwargs.pop('validate_every', 10) self.min_improvement = kwargs.pop('min_improvement', 0) self.max_gradient_norm = kwargs.pop('max_gradient_norm', 0) self.max_gradient_elem = kwargs.pop('max_gradient_elem', 0) util.log_param('patience', self.patience) util.log_param('validate_every', self.validate_every) util.log_param('min_improvement', self.min_improvement) util.log_param('max_gradient_norm', self.max_gradient_norm) util.log_param('max_gradient_elem', self.max_gradient_elem) util.log_param('learning_rate', self.learning_rate) util.log_param('momentum', self.momentum) util.log_param('nesterov', self.nesterov) def iterate(self, train=None, valid=None, max_updates=None, **kwargs): r'''Optimize a loss iteratively using a training and validation dataset. This method yields a series of monitor values to the caller. After every optimization epoch, a pair of monitor dictionaries is generated: one evaluated on the training dataset during the epoch, and another evaluated on the validation dataset at the most recent validation epoch. The validation monitors might not be updated during every optimization iteration; in this case, the most recent validation monitors will be yielded along with the training monitors. Additional keyword arguments supplied here will set the global optimizer attributes. Parameters ---------- train : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of training data for computing updates to model parameters. valid : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of validation data for computing monitor values and determining when the loss has stopped improving. Defaults to the training data. max_updates : int, optional If specified, halt optimization after this many gradient updates have been processed. If not provided, uses early stopping to decide when to halt. Yields ------ train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' self._compile(**kwargs) if valid is None: valid = train iteration = 0 training = validation = None while max_updates is None or iteration < max_updates: if not iteration % self.validate_every: try: validation = self.evaluate(valid) except KeyboardInterrupt: util.log('interrupted!') break if self._test_patience(validation): util.log('patience elapsed!') break try: training = self._step(train) except KeyboardInterrupt: util.log('interrupted!') break iteration += 1 self._log(training, iteration) yield training, validation self.set_params('best') def minimize(self, *args, **kwargs): '''Optimize our loss exhaustively. This method is a thin wrapper over the :func:`iterate` method. It simply exhausts the iterative optimization process and returns the final monitor values. Returns ------- train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' monitors = None for monitors in self.iterate(*args, **kwargs): pass return monitors def _step(self, dataset): '''Advance the state of the optimizer by one step. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A dataset for optimizing the model. Returns ------- train_monitors : dict A dictionary mapping monitor names to values. ''' if dataset is None: values = [self.f_step()] else: values = [self.f_step(*x) for x in dataset] return collections.OrderedDict( zip(self._monitor_names, np.mean(values, axis=0)))
lmjohns3/downhill
downhill/base.py
Optimizer.set_params
python
def set_params(self, targets=None): '''Set the values of the parameters to the given target values. Parameters ---------- targets : sequence of ndarray, optional Arrays for setting the parameters of our model. If this is not provided, the current best parameters for this optimizer will be used. ''' if not isinstance(targets, (list, tuple)): targets = self._best_params for param, target in zip(self._params, targets): param.set_value(target)
Set the values of the parameters to the given target values. Parameters ---------- targets : sequence of ndarray, optional Arrays for setting the parameters of our model. If this is not provided, the current best parameters for this optimizer will be used.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/base.py#L246-L259
null
class Optimizer(util.Registrar(str('Base'), (), {})): '''An optimizer computes gradient updates to iteratively optimize a loss. Attributes ---------- patience : int, optional Number of validation "failures" that we are willing to tolerate before stopping the optimization process. A validation failure happens whenever the loss on the validation dataset decreases by less than ``min_improvement`` (relative) over the previous best validation loss. Defaults to 5. validate_every : int, optional Evaluate the loss on the validation dataset after making this many passes over the training data. Defaults to 10. min_improvement : float, optional Insist that the validation loss must improve by this relative amount before considering that the optimization has made progress. The optimization process halts when ``patience`` validations have failed to make this relative improvement. Defaults to 0; set to a larger value (e.g., 0.01 for 1% improvement) to halt the optimization process sooner. max_gradient_norm : float, optional Rescale each parameter's gradient so that it has at most this L2 norm. Set to 0 (the default) to disable norm rescaling. If ``max_gradient_elem`` is also specified, then this has no effect. max_gradient_elem : float, optional Perform elementwise clipping on the magnitude of gradient values. Set to 0 (the default) to disable. If elementwise clipping is enabled, norm rescaling (via ``max_gradient_norm``) will have no effect. Deprecated synonyms of this parameter are "max_gradient_clip" and "gradient_clip". learning_rate : float, optional Many SGD-based optimization algorithms require a learning rate hyperparameter that scales the gradient step. Defaults to 1e-4. momentum : float, optional Apply momentum to the parameter updates for this optimizer, with the given strength. Typically this value ranges from 0 (no momentum) to :math:`1 - \epsilon` (large momentum). Defaults to 0. nesterov : bool, optional If True, and ``momentum`` is nonzero, apply Nesterov-style momentum to parameter updates for this optimizer. If False, and ``momentum`` is nonzero, "regular" momentum is applied. Has no effect if ``momentum`` is zero. See :class:`NAG <downhill.NAG>` for a description of Nesterov momentum. Parameters ---------- loss : Theano expression Loss function to minimize. This must be a scalar-valued expression. params : list of Theano variables, optional Symbolic variables to adjust to minimize the loss. If not given, these will be computed automatically by walking the computation graph. inputs : list of Theano variables, optional Symbolic variables required to compute the loss. If not given, these will be computed automatically by walking the computation graph. updates : list of update pairs, optional A list of pairs providing updates for the internals of the loss computation. Normally this is empty, but it can be provided if the loss, for example, requires an update to an internal random number generator. monitors : sequence of (str, Theano expression) tuples, optional Additional values to monitor during optimization. These must be provided as a sequence of (name, expression) tuples. monitor_gradients : bool, optional If True, add monitors to log the norms of the parameter gradients during optimization. Defaults to False. ''' def __init__(self, loss, params=None, inputs=None, updates=(), monitors=(), monitor_gradients=False): inputs_, params_ = util.find_inputs_and_params(loss) self._loss = loss self._params = params or params_ self._inputs = inputs or inputs_ self._updates = updates self._shapes = [p.get_value(borrow=True).shape for p in self._params] self._counts = [np.prod(s) for s in self._shapes] self._starts = np.cumsum([0] + self._counts)[:-1] self._dtype = self._params[0].get_value().dtype self._curr_iter = 0 self._best_iter = 0 self._best_loss = 1e100 self._best_params = [p.get_value().copy() for p in self._params] self._monitor_exprs = [self._loss] self._monitor_names = ['loss'] for name, monitor in monitors: self._monitor_names.append(name) self._monitor_exprs.append(monitor) if monitor_gradients: unnamed = 0 for p, g in zip(self._params, TT.grad(self._loss, self._params)): name = p.name if not name: name = 'unnamed{}'.format(unnamed) unnamed += 1 util.log('"{}" unnamed, will be "{}" internally'.format(p, name)) self._monitor_names.append('grad({})'.format(name)) self._monitor_exprs.append((g * g).sum()) def _compile(self, **kwargs): '''Compile the Theano functions for evaluating and updating our model. ''' util.log('compiling evaluation function') self.f_eval = theano.function(self._inputs, self._monitor_exprs, updates=self._updates, name='evaluation') label = self.__class__.__name__ util.log('compiling {} optimizer'.format(click.style(label, fg='red'))) updates = list(self._updates) + list(self.get_updates(**kwargs)) self.f_step = theano.function(self._inputs, self._monitor_exprs, updates=updates, name=label) def get_updates(self, **kwargs): '''Get parameter update expressions for performing optimization. Keyword arguments can be applied here to set any of the global optimizer attributes. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' self._prepare(**kwargs) for param, grad in self._differentiate(): for var, update in self._get_updates_for(param, grad): # For auxiliary variables, updates are meant to replace the # existing variable value. if var != param: yield var, update continue # If momentum is disabled, just apply the parameter delta. if self.momentum == 0: yield var, param - update continue # Momentum is enabled, so we keep track of velocity here. vel_tm1 = util.shared_like(param, 'vel') vel_t = util.as_float(self.momentum) * vel_tm1 - update if self.nesterov: # see http://arxiv.org/pdf/1212.0901v2.pdf (eq 7) and # https://github.com/lisa-lab/pylearn2/pull/136#issuecomment-10381617 mom_sqr = util.as_float(self.momentum ** 2) mom_inc = util.as_float(1 + self.momentum) vel_t = mom_sqr * vel_tm1 - mom_inc * update yield vel_tm1, vel_t yield param, param + vel_t def _get_updates_for(self, param, grad): '''Generate some update pairs for the given model parameter. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' raise NotImplementedError def _differentiate(self, params=None): '''Return a sequence of gradients for our parameters. If this optimizer has been configured with a gradient norm limit, or with elementwise gradient clipping, this method applies the appropriate rescaling and clipping operations before returning the gradient. Parameters ---------- params : list of Theano variables, optional Return the gradient with respect to these parameters. Defaults to all parameters that the optimizer knows about. Yields ------ pairs : (param, grad) tuples Generates a sequence of tuples representing each of the parameters requested and the corresponding Theano gradient expressions. ''' if params is None: params = self._params for param, grad in zip(params, TT.grad(self._loss, params)): if self.max_gradient_elem > 0: limit = util.as_float(self.max_gradient_elem) yield param, TT.clip(grad, -limit, limit) elif self.max_gradient_norm > 0: norm = TT.sqrt((grad * grad).sum()) limit = util.as_float(self.max_gradient_norm) yield param, grad * TT.minimum(1, limit / norm) else: yield param, grad def _log(self, monitors, iteration, label='', suffix=''): '''Log the state of the optimizer on the console. Parameters ---------- monitors : OrderedDict A dictionary of monitor names mapped to values. These names and values are what is being logged. iteration : int Optimization iteration that we are logging. label : str, optional A label for the name of the optimizer creating the log line. Defaults to the name of the current class. suffix : str, optional A suffix to add to the end of the log line, if any. ''' label = label or self.__class__.__name__ fields = (('{}={:.6f}').format(k, v) for k, v in monitors.items()) util.log('{} {} {}{}'.format(label, iteration, ' '.join(fields), suffix)) def evaluate(self, dataset): '''Evaluate the current model parameters on a dataset. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A set of data to use for evaluating the model. Returns ------- monitors : OrderedDict A dictionary mapping monitor names to values. Monitors are quantities of interest during optimization---for example, loss function, accuracy, or whatever the optimization task requires. ''' if dataset is None: values = [self.f_eval()] else: values = [self.f_eval(*x) for x in dataset] monitors = zip(self._monitor_names, np.mean(values, axis=0)) return collections.OrderedDict(monitors) def _test_patience(self, monitors): '''Test whether our patience with optimization has elapsed. Parameters ---------- monitors : dict A dictionary mapping monitor names to values. The 'loss' key from this dictionary will be used to evaluate optimization progress. Returns ------- elapsed : bool True iff our patience has elapsed and the model is no longer improving. ''' self._curr_iter += 1 marker = '' loss = monitors['loss'] if self._best_loss - loss > self._best_loss * self.min_improvement: self._best_loss = loss self._best_iter = self._curr_iter self._best_params = [p.get_value().copy() for p in self._params] marker = ' *' self._log(monitors, self._curr_iter - 1, 'validation', marker) return self._curr_iter - self._best_iter > self.patience def _prepare(self, **kwargs): '''Set up properties for optimization. This method can be overridden by base classes to provide parameters that are specific to a particular optimization technique (e.g., setting up a learning rate value). ''' self.learning_rate = util.as_float(kwargs.pop('learning_rate', 1e-4)) self.momentum = kwargs.pop('momentum', 0) self.nesterov = kwargs.pop('nesterov', False) self.patience = kwargs.get('patience', 5) self.validate_every = kwargs.pop('validate_every', 10) self.min_improvement = kwargs.pop('min_improvement', 0) self.max_gradient_norm = kwargs.pop('max_gradient_norm', 0) self.max_gradient_elem = kwargs.pop('max_gradient_elem', 0) util.log_param('patience', self.patience) util.log_param('validate_every', self.validate_every) util.log_param('min_improvement', self.min_improvement) util.log_param('max_gradient_norm', self.max_gradient_norm) util.log_param('max_gradient_elem', self.max_gradient_elem) util.log_param('learning_rate', self.learning_rate) util.log_param('momentum', self.momentum) util.log_param('nesterov', self.nesterov) def iterate(self, train=None, valid=None, max_updates=None, **kwargs): r'''Optimize a loss iteratively using a training and validation dataset. This method yields a series of monitor values to the caller. After every optimization epoch, a pair of monitor dictionaries is generated: one evaluated on the training dataset during the epoch, and another evaluated on the validation dataset at the most recent validation epoch. The validation monitors might not be updated during every optimization iteration; in this case, the most recent validation monitors will be yielded along with the training monitors. Additional keyword arguments supplied here will set the global optimizer attributes. Parameters ---------- train : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of training data for computing updates to model parameters. valid : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of validation data for computing monitor values and determining when the loss has stopped improving. Defaults to the training data. max_updates : int, optional If specified, halt optimization after this many gradient updates have been processed. If not provided, uses early stopping to decide when to halt. Yields ------ train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' self._compile(**kwargs) if valid is None: valid = train iteration = 0 training = validation = None while max_updates is None or iteration < max_updates: if not iteration % self.validate_every: try: validation = self.evaluate(valid) except KeyboardInterrupt: util.log('interrupted!') break if self._test_patience(validation): util.log('patience elapsed!') break try: training = self._step(train) except KeyboardInterrupt: util.log('interrupted!') break iteration += 1 self._log(training, iteration) yield training, validation self.set_params('best') def minimize(self, *args, **kwargs): '''Optimize our loss exhaustively. This method is a thin wrapper over the :func:`iterate` method. It simply exhausts the iterative optimization process and returns the final monitor values. Returns ------- train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' monitors = None for monitors in self.iterate(*args, **kwargs): pass return monitors def _step(self, dataset): '''Advance the state of the optimizer by one step. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A dataset for optimizing the model. Returns ------- train_monitors : dict A dictionary mapping monitor names to values. ''' if dataset is None: values = [self.f_step()] else: values = [self.f_step(*x) for x in dataset] return collections.OrderedDict( zip(self._monitor_names, np.mean(values, axis=0)))
lmjohns3/downhill
downhill/base.py
Optimizer._log
python
def _log(self, monitors, iteration, label='', suffix=''): '''Log the state of the optimizer on the console. Parameters ---------- monitors : OrderedDict A dictionary of monitor names mapped to values. These names and values are what is being logged. iteration : int Optimization iteration that we are logging. label : str, optional A label for the name of the optimizer creating the log line. Defaults to the name of the current class. suffix : str, optional A suffix to add to the end of the log line, if any. ''' label = label or self.__class__.__name__ fields = (('{}={:.6f}').format(k, v) for k, v in monitors.items()) util.log('{} {} {}{}'.format(label, iteration, ' '.join(fields), suffix))
Log the state of the optimizer on the console. Parameters ---------- monitors : OrderedDict A dictionary of monitor names mapped to values. These names and values are what is being logged. iteration : int Optimization iteration that we are logging. label : str, optional A label for the name of the optimizer creating the log line. Defaults to the name of the current class. suffix : str, optional A suffix to add to the end of the log line, if any.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/base.py#L261-L279
null
class Optimizer(util.Registrar(str('Base'), (), {})): '''An optimizer computes gradient updates to iteratively optimize a loss. Attributes ---------- patience : int, optional Number of validation "failures" that we are willing to tolerate before stopping the optimization process. A validation failure happens whenever the loss on the validation dataset decreases by less than ``min_improvement`` (relative) over the previous best validation loss. Defaults to 5. validate_every : int, optional Evaluate the loss on the validation dataset after making this many passes over the training data. Defaults to 10. min_improvement : float, optional Insist that the validation loss must improve by this relative amount before considering that the optimization has made progress. The optimization process halts when ``patience`` validations have failed to make this relative improvement. Defaults to 0; set to a larger value (e.g., 0.01 for 1% improvement) to halt the optimization process sooner. max_gradient_norm : float, optional Rescale each parameter's gradient so that it has at most this L2 norm. Set to 0 (the default) to disable norm rescaling. If ``max_gradient_elem`` is also specified, then this has no effect. max_gradient_elem : float, optional Perform elementwise clipping on the magnitude of gradient values. Set to 0 (the default) to disable. If elementwise clipping is enabled, norm rescaling (via ``max_gradient_norm``) will have no effect. Deprecated synonyms of this parameter are "max_gradient_clip" and "gradient_clip". learning_rate : float, optional Many SGD-based optimization algorithms require a learning rate hyperparameter that scales the gradient step. Defaults to 1e-4. momentum : float, optional Apply momentum to the parameter updates for this optimizer, with the given strength. Typically this value ranges from 0 (no momentum) to :math:`1 - \epsilon` (large momentum). Defaults to 0. nesterov : bool, optional If True, and ``momentum`` is nonzero, apply Nesterov-style momentum to parameter updates for this optimizer. If False, and ``momentum`` is nonzero, "regular" momentum is applied. Has no effect if ``momentum`` is zero. See :class:`NAG <downhill.NAG>` for a description of Nesterov momentum. Parameters ---------- loss : Theano expression Loss function to minimize. This must be a scalar-valued expression. params : list of Theano variables, optional Symbolic variables to adjust to minimize the loss. If not given, these will be computed automatically by walking the computation graph. inputs : list of Theano variables, optional Symbolic variables required to compute the loss. If not given, these will be computed automatically by walking the computation graph. updates : list of update pairs, optional A list of pairs providing updates for the internals of the loss computation. Normally this is empty, but it can be provided if the loss, for example, requires an update to an internal random number generator. monitors : sequence of (str, Theano expression) tuples, optional Additional values to monitor during optimization. These must be provided as a sequence of (name, expression) tuples. monitor_gradients : bool, optional If True, add monitors to log the norms of the parameter gradients during optimization. Defaults to False. ''' def __init__(self, loss, params=None, inputs=None, updates=(), monitors=(), monitor_gradients=False): inputs_, params_ = util.find_inputs_and_params(loss) self._loss = loss self._params = params or params_ self._inputs = inputs or inputs_ self._updates = updates self._shapes = [p.get_value(borrow=True).shape for p in self._params] self._counts = [np.prod(s) for s in self._shapes] self._starts = np.cumsum([0] + self._counts)[:-1] self._dtype = self._params[0].get_value().dtype self._curr_iter = 0 self._best_iter = 0 self._best_loss = 1e100 self._best_params = [p.get_value().copy() for p in self._params] self._monitor_exprs = [self._loss] self._monitor_names = ['loss'] for name, monitor in monitors: self._monitor_names.append(name) self._monitor_exprs.append(monitor) if monitor_gradients: unnamed = 0 for p, g in zip(self._params, TT.grad(self._loss, self._params)): name = p.name if not name: name = 'unnamed{}'.format(unnamed) unnamed += 1 util.log('"{}" unnamed, will be "{}" internally'.format(p, name)) self._monitor_names.append('grad({})'.format(name)) self._monitor_exprs.append((g * g).sum()) def _compile(self, **kwargs): '''Compile the Theano functions for evaluating and updating our model. ''' util.log('compiling evaluation function') self.f_eval = theano.function(self._inputs, self._monitor_exprs, updates=self._updates, name='evaluation') label = self.__class__.__name__ util.log('compiling {} optimizer'.format(click.style(label, fg='red'))) updates = list(self._updates) + list(self.get_updates(**kwargs)) self.f_step = theano.function(self._inputs, self._monitor_exprs, updates=updates, name=label) def get_updates(self, **kwargs): '''Get parameter update expressions for performing optimization. Keyword arguments can be applied here to set any of the global optimizer attributes. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' self._prepare(**kwargs) for param, grad in self._differentiate(): for var, update in self._get_updates_for(param, grad): # For auxiliary variables, updates are meant to replace the # existing variable value. if var != param: yield var, update continue # If momentum is disabled, just apply the parameter delta. if self.momentum == 0: yield var, param - update continue # Momentum is enabled, so we keep track of velocity here. vel_tm1 = util.shared_like(param, 'vel') vel_t = util.as_float(self.momentum) * vel_tm1 - update if self.nesterov: # see http://arxiv.org/pdf/1212.0901v2.pdf (eq 7) and # https://github.com/lisa-lab/pylearn2/pull/136#issuecomment-10381617 mom_sqr = util.as_float(self.momentum ** 2) mom_inc = util.as_float(1 + self.momentum) vel_t = mom_sqr * vel_tm1 - mom_inc * update yield vel_tm1, vel_t yield param, param + vel_t def _get_updates_for(self, param, grad): '''Generate some update pairs for the given model parameter. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' raise NotImplementedError def _differentiate(self, params=None): '''Return a sequence of gradients for our parameters. If this optimizer has been configured with a gradient norm limit, or with elementwise gradient clipping, this method applies the appropriate rescaling and clipping operations before returning the gradient. Parameters ---------- params : list of Theano variables, optional Return the gradient with respect to these parameters. Defaults to all parameters that the optimizer knows about. Yields ------ pairs : (param, grad) tuples Generates a sequence of tuples representing each of the parameters requested and the corresponding Theano gradient expressions. ''' if params is None: params = self._params for param, grad in zip(params, TT.grad(self._loss, params)): if self.max_gradient_elem > 0: limit = util.as_float(self.max_gradient_elem) yield param, TT.clip(grad, -limit, limit) elif self.max_gradient_norm > 0: norm = TT.sqrt((grad * grad).sum()) limit = util.as_float(self.max_gradient_norm) yield param, grad * TT.minimum(1, limit / norm) else: yield param, grad def set_params(self, targets=None): '''Set the values of the parameters to the given target values. Parameters ---------- targets : sequence of ndarray, optional Arrays for setting the parameters of our model. If this is not provided, the current best parameters for this optimizer will be used. ''' if not isinstance(targets, (list, tuple)): targets = self._best_params for param, target in zip(self._params, targets): param.set_value(target) def evaluate(self, dataset): '''Evaluate the current model parameters on a dataset. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A set of data to use for evaluating the model. Returns ------- monitors : OrderedDict A dictionary mapping monitor names to values. Monitors are quantities of interest during optimization---for example, loss function, accuracy, or whatever the optimization task requires. ''' if dataset is None: values = [self.f_eval()] else: values = [self.f_eval(*x) for x in dataset] monitors = zip(self._monitor_names, np.mean(values, axis=0)) return collections.OrderedDict(monitors) def _test_patience(self, monitors): '''Test whether our patience with optimization has elapsed. Parameters ---------- monitors : dict A dictionary mapping monitor names to values. The 'loss' key from this dictionary will be used to evaluate optimization progress. Returns ------- elapsed : bool True iff our patience has elapsed and the model is no longer improving. ''' self._curr_iter += 1 marker = '' loss = monitors['loss'] if self._best_loss - loss > self._best_loss * self.min_improvement: self._best_loss = loss self._best_iter = self._curr_iter self._best_params = [p.get_value().copy() for p in self._params] marker = ' *' self._log(monitors, self._curr_iter - 1, 'validation', marker) return self._curr_iter - self._best_iter > self.patience def _prepare(self, **kwargs): '''Set up properties for optimization. This method can be overridden by base classes to provide parameters that are specific to a particular optimization technique (e.g., setting up a learning rate value). ''' self.learning_rate = util.as_float(kwargs.pop('learning_rate', 1e-4)) self.momentum = kwargs.pop('momentum', 0) self.nesterov = kwargs.pop('nesterov', False) self.patience = kwargs.get('patience', 5) self.validate_every = kwargs.pop('validate_every', 10) self.min_improvement = kwargs.pop('min_improvement', 0) self.max_gradient_norm = kwargs.pop('max_gradient_norm', 0) self.max_gradient_elem = kwargs.pop('max_gradient_elem', 0) util.log_param('patience', self.patience) util.log_param('validate_every', self.validate_every) util.log_param('min_improvement', self.min_improvement) util.log_param('max_gradient_norm', self.max_gradient_norm) util.log_param('max_gradient_elem', self.max_gradient_elem) util.log_param('learning_rate', self.learning_rate) util.log_param('momentum', self.momentum) util.log_param('nesterov', self.nesterov) def iterate(self, train=None, valid=None, max_updates=None, **kwargs): r'''Optimize a loss iteratively using a training and validation dataset. This method yields a series of monitor values to the caller. After every optimization epoch, a pair of monitor dictionaries is generated: one evaluated on the training dataset during the epoch, and another evaluated on the validation dataset at the most recent validation epoch. The validation monitors might not be updated during every optimization iteration; in this case, the most recent validation monitors will be yielded along with the training monitors. Additional keyword arguments supplied here will set the global optimizer attributes. Parameters ---------- train : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of training data for computing updates to model parameters. valid : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of validation data for computing monitor values and determining when the loss has stopped improving. Defaults to the training data. max_updates : int, optional If specified, halt optimization after this many gradient updates have been processed. If not provided, uses early stopping to decide when to halt. Yields ------ train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' self._compile(**kwargs) if valid is None: valid = train iteration = 0 training = validation = None while max_updates is None or iteration < max_updates: if not iteration % self.validate_every: try: validation = self.evaluate(valid) except KeyboardInterrupt: util.log('interrupted!') break if self._test_patience(validation): util.log('patience elapsed!') break try: training = self._step(train) except KeyboardInterrupt: util.log('interrupted!') break iteration += 1 self._log(training, iteration) yield training, validation self.set_params('best') def minimize(self, *args, **kwargs): '''Optimize our loss exhaustively. This method is a thin wrapper over the :func:`iterate` method. It simply exhausts the iterative optimization process and returns the final monitor values. Returns ------- train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' monitors = None for monitors in self.iterate(*args, **kwargs): pass return monitors def _step(self, dataset): '''Advance the state of the optimizer by one step. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A dataset for optimizing the model. Returns ------- train_monitors : dict A dictionary mapping monitor names to values. ''' if dataset is None: values = [self.f_step()] else: values = [self.f_step(*x) for x in dataset] return collections.OrderedDict( zip(self._monitor_names, np.mean(values, axis=0)))
lmjohns3/downhill
downhill/base.py
Optimizer.evaluate
python
def evaluate(self, dataset): '''Evaluate the current model parameters on a dataset. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A set of data to use for evaluating the model. Returns ------- monitors : OrderedDict A dictionary mapping monitor names to values. Monitors are quantities of interest during optimization---for example, loss function, accuracy, or whatever the optimization task requires. ''' if dataset is None: values = [self.f_eval()] else: values = [self.f_eval(*x) for x in dataset] monitors = zip(self._monitor_names, np.mean(values, axis=0)) return collections.OrderedDict(monitors)
Evaluate the current model parameters on a dataset. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A set of data to use for evaluating the model. Returns ------- monitors : OrderedDict A dictionary mapping monitor names to values. Monitors are quantities of interest during optimization---for example, loss function, accuracy, or whatever the optimization task requires.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/base.py#L281-L301
null
class Optimizer(util.Registrar(str('Base'), (), {})): '''An optimizer computes gradient updates to iteratively optimize a loss. Attributes ---------- patience : int, optional Number of validation "failures" that we are willing to tolerate before stopping the optimization process. A validation failure happens whenever the loss on the validation dataset decreases by less than ``min_improvement`` (relative) over the previous best validation loss. Defaults to 5. validate_every : int, optional Evaluate the loss on the validation dataset after making this many passes over the training data. Defaults to 10. min_improvement : float, optional Insist that the validation loss must improve by this relative amount before considering that the optimization has made progress. The optimization process halts when ``patience`` validations have failed to make this relative improvement. Defaults to 0; set to a larger value (e.g., 0.01 for 1% improvement) to halt the optimization process sooner. max_gradient_norm : float, optional Rescale each parameter's gradient so that it has at most this L2 norm. Set to 0 (the default) to disable norm rescaling. If ``max_gradient_elem`` is also specified, then this has no effect. max_gradient_elem : float, optional Perform elementwise clipping on the magnitude of gradient values. Set to 0 (the default) to disable. If elementwise clipping is enabled, norm rescaling (via ``max_gradient_norm``) will have no effect. Deprecated synonyms of this parameter are "max_gradient_clip" and "gradient_clip". learning_rate : float, optional Many SGD-based optimization algorithms require a learning rate hyperparameter that scales the gradient step. Defaults to 1e-4. momentum : float, optional Apply momentum to the parameter updates for this optimizer, with the given strength. Typically this value ranges from 0 (no momentum) to :math:`1 - \epsilon` (large momentum). Defaults to 0. nesterov : bool, optional If True, and ``momentum`` is nonzero, apply Nesterov-style momentum to parameter updates for this optimizer. If False, and ``momentum`` is nonzero, "regular" momentum is applied. Has no effect if ``momentum`` is zero. See :class:`NAG <downhill.NAG>` for a description of Nesterov momentum. Parameters ---------- loss : Theano expression Loss function to minimize. This must be a scalar-valued expression. params : list of Theano variables, optional Symbolic variables to adjust to minimize the loss. If not given, these will be computed automatically by walking the computation graph. inputs : list of Theano variables, optional Symbolic variables required to compute the loss. If not given, these will be computed automatically by walking the computation graph. updates : list of update pairs, optional A list of pairs providing updates for the internals of the loss computation. Normally this is empty, but it can be provided if the loss, for example, requires an update to an internal random number generator. monitors : sequence of (str, Theano expression) tuples, optional Additional values to monitor during optimization. These must be provided as a sequence of (name, expression) tuples. monitor_gradients : bool, optional If True, add monitors to log the norms of the parameter gradients during optimization. Defaults to False. ''' def __init__(self, loss, params=None, inputs=None, updates=(), monitors=(), monitor_gradients=False): inputs_, params_ = util.find_inputs_and_params(loss) self._loss = loss self._params = params or params_ self._inputs = inputs or inputs_ self._updates = updates self._shapes = [p.get_value(borrow=True).shape for p in self._params] self._counts = [np.prod(s) for s in self._shapes] self._starts = np.cumsum([0] + self._counts)[:-1] self._dtype = self._params[0].get_value().dtype self._curr_iter = 0 self._best_iter = 0 self._best_loss = 1e100 self._best_params = [p.get_value().copy() for p in self._params] self._monitor_exprs = [self._loss] self._monitor_names = ['loss'] for name, monitor in monitors: self._monitor_names.append(name) self._monitor_exprs.append(monitor) if monitor_gradients: unnamed = 0 for p, g in zip(self._params, TT.grad(self._loss, self._params)): name = p.name if not name: name = 'unnamed{}'.format(unnamed) unnamed += 1 util.log('"{}" unnamed, will be "{}" internally'.format(p, name)) self._monitor_names.append('grad({})'.format(name)) self._monitor_exprs.append((g * g).sum()) def _compile(self, **kwargs): '''Compile the Theano functions for evaluating and updating our model. ''' util.log('compiling evaluation function') self.f_eval = theano.function(self._inputs, self._monitor_exprs, updates=self._updates, name='evaluation') label = self.__class__.__name__ util.log('compiling {} optimizer'.format(click.style(label, fg='red'))) updates = list(self._updates) + list(self.get_updates(**kwargs)) self.f_step = theano.function(self._inputs, self._monitor_exprs, updates=updates, name=label) def get_updates(self, **kwargs): '''Get parameter update expressions for performing optimization. Keyword arguments can be applied here to set any of the global optimizer attributes. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' self._prepare(**kwargs) for param, grad in self._differentiate(): for var, update in self._get_updates_for(param, grad): # For auxiliary variables, updates are meant to replace the # existing variable value. if var != param: yield var, update continue # If momentum is disabled, just apply the parameter delta. if self.momentum == 0: yield var, param - update continue # Momentum is enabled, so we keep track of velocity here. vel_tm1 = util.shared_like(param, 'vel') vel_t = util.as_float(self.momentum) * vel_tm1 - update if self.nesterov: # see http://arxiv.org/pdf/1212.0901v2.pdf (eq 7) and # https://github.com/lisa-lab/pylearn2/pull/136#issuecomment-10381617 mom_sqr = util.as_float(self.momentum ** 2) mom_inc = util.as_float(1 + self.momentum) vel_t = mom_sqr * vel_tm1 - mom_inc * update yield vel_tm1, vel_t yield param, param + vel_t def _get_updates_for(self, param, grad): '''Generate some update pairs for the given model parameter. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' raise NotImplementedError def _differentiate(self, params=None): '''Return a sequence of gradients for our parameters. If this optimizer has been configured with a gradient norm limit, or with elementwise gradient clipping, this method applies the appropriate rescaling and clipping operations before returning the gradient. Parameters ---------- params : list of Theano variables, optional Return the gradient with respect to these parameters. Defaults to all parameters that the optimizer knows about. Yields ------ pairs : (param, grad) tuples Generates a sequence of tuples representing each of the parameters requested and the corresponding Theano gradient expressions. ''' if params is None: params = self._params for param, grad in zip(params, TT.grad(self._loss, params)): if self.max_gradient_elem > 0: limit = util.as_float(self.max_gradient_elem) yield param, TT.clip(grad, -limit, limit) elif self.max_gradient_norm > 0: norm = TT.sqrt((grad * grad).sum()) limit = util.as_float(self.max_gradient_norm) yield param, grad * TT.minimum(1, limit / norm) else: yield param, grad def set_params(self, targets=None): '''Set the values of the parameters to the given target values. Parameters ---------- targets : sequence of ndarray, optional Arrays for setting the parameters of our model. If this is not provided, the current best parameters for this optimizer will be used. ''' if not isinstance(targets, (list, tuple)): targets = self._best_params for param, target in zip(self._params, targets): param.set_value(target) def _log(self, monitors, iteration, label='', suffix=''): '''Log the state of the optimizer on the console. Parameters ---------- monitors : OrderedDict A dictionary of monitor names mapped to values. These names and values are what is being logged. iteration : int Optimization iteration that we are logging. label : str, optional A label for the name of the optimizer creating the log line. Defaults to the name of the current class. suffix : str, optional A suffix to add to the end of the log line, if any. ''' label = label or self.__class__.__name__ fields = (('{}={:.6f}').format(k, v) for k, v in monitors.items()) util.log('{} {} {}{}'.format(label, iteration, ' '.join(fields), suffix)) def _test_patience(self, monitors): '''Test whether our patience with optimization has elapsed. Parameters ---------- monitors : dict A dictionary mapping monitor names to values. The 'loss' key from this dictionary will be used to evaluate optimization progress. Returns ------- elapsed : bool True iff our patience has elapsed and the model is no longer improving. ''' self._curr_iter += 1 marker = '' loss = monitors['loss'] if self._best_loss - loss > self._best_loss * self.min_improvement: self._best_loss = loss self._best_iter = self._curr_iter self._best_params = [p.get_value().copy() for p in self._params] marker = ' *' self._log(monitors, self._curr_iter - 1, 'validation', marker) return self._curr_iter - self._best_iter > self.patience def _prepare(self, **kwargs): '''Set up properties for optimization. This method can be overridden by base classes to provide parameters that are specific to a particular optimization technique (e.g., setting up a learning rate value). ''' self.learning_rate = util.as_float(kwargs.pop('learning_rate', 1e-4)) self.momentum = kwargs.pop('momentum', 0) self.nesterov = kwargs.pop('nesterov', False) self.patience = kwargs.get('patience', 5) self.validate_every = kwargs.pop('validate_every', 10) self.min_improvement = kwargs.pop('min_improvement', 0) self.max_gradient_norm = kwargs.pop('max_gradient_norm', 0) self.max_gradient_elem = kwargs.pop('max_gradient_elem', 0) util.log_param('patience', self.patience) util.log_param('validate_every', self.validate_every) util.log_param('min_improvement', self.min_improvement) util.log_param('max_gradient_norm', self.max_gradient_norm) util.log_param('max_gradient_elem', self.max_gradient_elem) util.log_param('learning_rate', self.learning_rate) util.log_param('momentum', self.momentum) util.log_param('nesterov', self.nesterov) def iterate(self, train=None, valid=None, max_updates=None, **kwargs): r'''Optimize a loss iteratively using a training and validation dataset. This method yields a series of monitor values to the caller. After every optimization epoch, a pair of monitor dictionaries is generated: one evaluated on the training dataset during the epoch, and another evaluated on the validation dataset at the most recent validation epoch. The validation monitors might not be updated during every optimization iteration; in this case, the most recent validation monitors will be yielded along with the training monitors. Additional keyword arguments supplied here will set the global optimizer attributes. Parameters ---------- train : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of training data for computing updates to model parameters. valid : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of validation data for computing monitor values and determining when the loss has stopped improving. Defaults to the training data. max_updates : int, optional If specified, halt optimization after this many gradient updates have been processed. If not provided, uses early stopping to decide when to halt. Yields ------ train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' self._compile(**kwargs) if valid is None: valid = train iteration = 0 training = validation = None while max_updates is None or iteration < max_updates: if not iteration % self.validate_every: try: validation = self.evaluate(valid) except KeyboardInterrupt: util.log('interrupted!') break if self._test_patience(validation): util.log('patience elapsed!') break try: training = self._step(train) except KeyboardInterrupt: util.log('interrupted!') break iteration += 1 self._log(training, iteration) yield training, validation self.set_params('best') def minimize(self, *args, **kwargs): '''Optimize our loss exhaustively. This method is a thin wrapper over the :func:`iterate` method. It simply exhausts the iterative optimization process and returns the final monitor values. Returns ------- train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' monitors = None for monitors in self.iterate(*args, **kwargs): pass return monitors def _step(self, dataset): '''Advance the state of the optimizer by one step. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A dataset for optimizing the model. Returns ------- train_monitors : dict A dictionary mapping monitor names to values. ''' if dataset is None: values = [self.f_step()] else: values = [self.f_step(*x) for x in dataset] return collections.OrderedDict( zip(self._monitor_names, np.mean(values, axis=0)))
lmjohns3/downhill
downhill/base.py
Optimizer._prepare
python
def _prepare(self, **kwargs): '''Set up properties for optimization. This method can be overridden by base classes to provide parameters that are specific to a particular optimization technique (e.g., setting up a learning rate value). ''' self.learning_rate = util.as_float(kwargs.pop('learning_rate', 1e-4)) self.momentum = kwargs.pop('momentum', 0) self.nesterov = kwargs.pop('nesterov', False) self.patience = kwargs.get('patience', 5) self.validate_every = kwargs.pop('validate_every', 10) self.min_improvement = kwargs.pop('min_improvement', 0) self.max_gradient_norm = kwargs.pop('max_gradient_norm', 0) self.max_gradient_elem = kwargs.pop('max_gradient_elem', 0) util.log_param('patience', self.patience) util.log_param('validate_every', self.validate_every) util.log_param('min_improvement', self.min_improvement) util.log_param('max_gradient_norm', self.max_gradient_norm) util.log_param('max_gradient_elem', self.max_gradient_elem) util.log_param('learning_rate', self.learning_rate) util.log_param('momentum', self.momentum) util.log_param('nesterov', self.nesterov)
Set up properties for optimization. This method can be overridden by base classes to provide parameters that are specific to a particular optimization technique (e.g., setting up a learning rate value).
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/base.py#L329-L352
null
class Optimizer(util.Registrar(str('Base'), (), {})): '''An optimizer computes gradient updates to iteratively optimize a loss. Attributes ---------- patience : int, optional Number of validation "failures" that we are willing to tolerate before stopping the optimization process. A validation failure happens whenever the loss on the validation dataset decreases by less than ``min_improvement`` (relative) over the previous best validation loss. Defaults to 5. validate_every : int, optional Evaluate the loss on the validation dataset after making this many passes over the training data. Defaults to 10. min_improvement : float, optional Insist that the validation loss must improve by this relative amount before considering that the optimization has made progress. The optimization process halts when ``patience`` validations have failed to make this relative improvement. Defaults to 0; set to a larger value (e.g., 0.01 for 1% improvement) to halt the optimization process sooner. max_gradient_norm : float, optional Rescale each parameter's gradient so that it has at most this L2 norm. Set to 0 (the default) to disable norm rescaling. If ``max_gradient_elem`` is also specified, then this has no effect. max_gradient_elem : float, optional Perform elementwise clipping on the magnitude of gradient values. Set to 0 (the default) to disable. If elementwise clipping is enabled, norm rescaling (via ``max_gradient_norm``) will have no effect. Deprecated synonyms of this parameter are "max_gradient_clip" and "gradient_clip". learning_rate : float, optional Many SGD-based optimization algorithms require a learning rate hyperparameter that scales the gradient step. Defaults to 1e-4. momentum : float, optional Apply momentum to the parameter updates for this optimizer, with the given strength. Typically this value ranges from 0 (no momentum) to :math:`1 - \epsilon` (large momentum). Defaults to 0. nesterov : bool, optional If True, and ``momentum`` is nonzero, apply Nesterov-style momentum to parameter updates for this optimizer. If False, and ``momentum`` is nonzero, "regular" momentum is applied. Has no effect if ``momentum`` is zero. See :class:`NAG <downhill.NAG>` for a description of Nesterov momentum. Parameters ---------- loss : Theano expression Loss function to minimize. This must be a scalar-valued expression. params : list of Theano variables, optional Symbolic variables to adjust to minimize the loss. If not given, these will be computed automatically by walking the computation graph. inputs : list of Theano variables, optional Symbolic variables required to compute the loss. If not given, these will be computed automatically by walking the computation graph. updates : list of update pairs, optional A list of pairs providing updates for the internals of the loss computation. Normally this is empty, but it can be provided if the loss, for example, requires an update to an internal random number generator. monitors : sequence of (str, Theano expression) tuples, optional Additional values to monitor during optimization. These must be provided as a sequence of (name, expression) tuples. monitor_gradients : bool, optional If True, add monitors to log the norms of the parameter gradients during optimization. Defaults to False. ''' def __init__(self, loss, params=None, inputs=None, updates=(), monitors=(), monitor_gradients=False): inputs_, params_ = util.find_inputs_and_params(loss) self._loss = loss self._params = params or params_ self._inputs = inputs or inputs_ self._updates = updates self._shapes = [p.get_value(borrow=True).shape for p in self._params] self._counts = [np.prod(s) for s in self._shapes] self._starts = np.cumsum([0] + self._counts)[:-1] self._dtype = self._params[0].get_value().dtype self._curr_iter = 0 self._best_iter = 0 self._best_loss = 1e100 self._best_params = [p.get_value().copy() for p in self._params] self._monitor_exprs = [self._loss] self._monitor_names = ['loss'] for name, monitor in monitors: self._monitor_names.append(name) self._monitor_exprs.append(monitor) if monitor_gradients: unnamed = 0 for p, g in zip(self._params, TT.grad(self._loss, self._params)): name = p.name if not name: name = 'unnamed{}'.format(unnamed) unnamed += 1 util.log('"{}" unnamed, will be "{}" internally'.format(p, name)) self._monitor_names.append('grad({})'.format(name)) self._monitor_exprs.append((g * g).sum()) def _compile(self, **kwargs): '''Compile the Theano functions for evaluating and updating our model. ''' util.log('compiling evaluation function') self.f_eval = theano.function(self._inputs, self._monitor_exprs, updates=self._updates, name='evaluation') label = self.__class__.__name__ util.log('compiling {} optimizer'.format(click.style(label, fg='red'))) updates = list(self._updates) + list(self.get_updates(**kwargs)) self.f_step = theano.function(self._inputs, self._monitor_exprs, updates=updates, name=label) def get_updates(self, **kwargs): '''Get parameter update expressions for performing optimization. Keyword arguments can be applied here to set any of the global optimizer attributes. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' self._prepare(**kwargs) for param, grad in self._differentiate(): for var, update in self._get_updates_for(param, grad): # For auxiliary variables, updates are meant to replace the # existing variable value. if var != param: yield var, update continue # If momentum is disabled, just apply the parameter delta. if self.momentum == 0: yield var, param - update continue # Momentum is enabled, so we keep track of velocity here. vel_tm1 = util.shared_like(param, 'vel') vel_t = util.as_float(self.momentum) * vel_tm1 - update if self.nesterov: # see http://arxiv.org/pdf/1212.0901v2.pdf (eq 7) and # https://github.com/lisa-lab/pylearn2/pull/136#issuecomment-10381617 mom_sqr = util.as_float(self.momentum ** 2) mom_inc = util.as_float(1 + self.momentum) vel_t = mom_sqr * vel_tm1 - mom_inc * update yield vel_tm1, vel_t yield param, param + vel_t def _get_updates_for(self, param, grad): '''Generate some update pairs for the given model parameter. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' raise NotImplementedError def _differentiate(self, params=None): '''Return a sequence of gradients for our parameters. If this optimizer has been configured with a gradient norm limit, or with elementwise gradient clipping, this method applies the appropriate rescaling and clipping operations before returning the gradient. Parameters ---------- params : list of Theano variables, optional Return the gradient with respect to these parameters. Defaults to all parameters that the optimizer knows about. Yields ------ pairs : (param, grad) tuples Generates a sequence of tuples representing each of the parameters requested and the corresponding Theano gradient expressions. ''' if params is None: params = self._params for param, grad in zip(params, TT.grad(self._loss, params)): if self.max_gradient_elem > 0: limit = util.as_float(self.max_gradient_elem) yield param, TT.clip(grad, -limit, limit) elif self.max_gradient_norm > 0: norm = TT.sqrt((grad * grad).sum()) limit = util.as_float(self.max_gradient_norm) yield param, grad * TT.minimum(1, limit / norm) else: yield param, grad def set_params(self, targets=None): '''Set the values of the parameters to the given target values. Parameters ---------- targets : sequence of ndarray, optional Arrays for setting the parameters of our model. If this is not provided, the current best parameters for this optimizer will be used. ''' if not isinstance(targets, (list, tuple)): targets = self._best_params for param, target in zip(self._params, targets): param.set_value(target) def _log(self, monitors, iteration, label='', suffix=''): '''Log the state of the optimizer on the console. Parameters ---------- monitors : OrderedDict A dictionary of monitor names mapped to values. These names and values are what is being logged. iteration : int Optimization iteration that we are logging. label : str, optional A label for the name of the optimizer creating the log line. Defaults to the name of the current class. suffix : str, optional A suffix to add to the end of the log line, if any. ''' label = label or self.__class__.__name__ fields = (('{}={:.6f}').format(k, v) for k, v in monitors.items()) util.log('{} {} {}{}'.format(label, iteration, ' '.join(fields), suffix)) def evaluate(self, dataset): '''Evaluate the current model parameters on a dataset. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A set of data to use for evaluating the model. Returns ------- monitors : OrderedDict A dictionary mapping monitor names to values. Monitors are quantities of interest during optimization---for example, loss function, accuracy, or whatever the optimization task requires. ''' if dataset is None: values = [self.f_eval()] else: values = [self.f_eval(*x) for x in dataset] monitors = zip(self._monitor_names, np.mean(values, axis=0)) return collections.OrderedDict(monitors) def _test_patience(self, monitors): '''Test whether our patience with optimization has elapsed. Parameters ---------- monitors : dict A dictionary mapping monitor names to values. The 'loss' key from this dictionary will be used to evaluate optimization progress. Returns ------- elapsed : bool True iff our patience has elapsed and the model is no longer improving. ''' self._curr_iter += 1 marker = '' loss = monitors['loss'] if self._best_loss - loss > self._best_loss * self.min_improvement: self._best_loss = loss self._best_iter = self._curr_iter self._best_params = [p.get_value().copy() for p in self._params] marker = ' *' self._log(monitors, self._curr_iter - 1, 'validation', marker) return self._curr_iter - self._best_iter > self.patience def iterate(self, train=None, valid=None, max_updates=None, **kwargs): r'''Optimize a loss iteratively using a training and validation dataset. This method yields a series of monitor values to the caller. After every optimization epoch, a pair of monitor dictionaries is generated: one evaluated on the training dataset during the epoch, and another evaluated on the validation dataset at the most recent validation epoch. The validation monitors might not be updated during every optimization iteration; in this case, the most recent validation monitors will be yielded along with the training monitors. Additional keyword arguments supplied here will set the global optimizer attributes. Parameters ---------- train : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of training data for computing updates to model parameters. valid : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of validation data for computing monitor values and determining when the loss has stopped improving. Defaults to the training data. max_updates : int, optional If specified, halt optimization after this many gradient updates have been processed. If not provided, uses early stopping to decide when to halt. Yields ------ train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' self._compile(**kwargs) if valid is None: valid = train iteration = 0 training = validation = None while max_updates is None or iteration < max_updates: if not iteration % self.validate_every: try: validation = self.evaluate(valid) except KeyboardInterrupt: util.log('interrupted!') break if self._test_patience(validation): util.log('patience elapsed!') break try: training = self._step(train) except KeyboardInterrupt: util.log('interrupted!') break iteration += 1 self._log(training, iteration) yield training, validation self.set_params('best') def minimize(self, *args, **kwargs): '''Optimize our loss exhaustively. This method is a thin wrapper over the :func:`iterate` method. It simply exhausts the iterative optimization process and returns the final monitor values. Returns ------- train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' monitors = None for monitors in self.iterate(*args, **kwargs): pass return monitors def _step(self, dataset): '''Advance the state of the optimizer by one step. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A dataset for optimizing the model. Returns ------- train_monitors : dict A dictionary mapping monitor names to values. ''' if dataset is None: values = [self.f_step()] else: values = [self.f_step(*x) for x in dataset] return collections.OrderedDict( zip(self._monitor_names, np.mean(values, axis=0)))
lmjohns3/downhill
downhill/base.py
Optimizer.iterate
python
def iterate(self, train=None, valid=None, max_updates=None, **kwargs): r'''Optimize a loss iteratively using a training and validation dataset. This method yields a series of monitor values to the caller. After every optimization epoch, a pair of monitor dictionaries is generated: one evaluated on the training dataset during the epoch, and another evaluated on the validation dataset at the most recent validation epoch. The validation monitors might not be updated during every optimization iteration; in this case, the most recent validation monitors will be yielded along with the training monitors. Additional keyword arguments supplied here will set the global optimizer attributes. Parameters ---------- train : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of training data for computing updates to model parameters. valid : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of validation data for computing monitor values and determining when the loss has stopped improving. Defaults to the training data. max_updates : int, optional If specified, halt optimization after this many gradient updates have been processed. If not provided, uses early stopping to decide when to halt. Yields ------ train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' self._compile(**kwargs) if valid is None: valid = train iteration = 0 training = validation = None while max_updates is None or iteration < max_updates: if not iteration % self.validate_every: try: validation = self.evaluate(valid) except KeyboardInterrupt: util.log('interrupted!') break if self._test_patience(validation): util.log('patience elapsed!') break try: training = self._step(train) except KeyboardInterrupt: util.log('interrupted!') break iteration += 1 self._log(training, iteration) yield training, validation self.set_params('best')
r'''Optimize a loss iteratively using a training and validation dataset. This method yields a series of monitor values to the caller. After every optimization epoch, a pair of monitor dictionaries is generated: one evaluated on the training dataset during the epoch, and another evaluated on the validation dataset at the most recent validation epoch. The validation monitors might not be updated during every optimization iteration; in this case, the most recent validation monitors will be yielded along with the training monitors. Additional keyword arguments supplied here will set the global optimizer attributes. Parameters ---------- train : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of training data for computing updates to model parameters. valid : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of validation data for computing monitor values and determining when the loss has stopped improving. Defaults to the training data. max_updates : int, optional If specified, halt optimization after this many gradient updates have been processed. If not provided, uses early stopping to decide when to halt. Yields ------ train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/base.py#L354-L415
[ "def log(msg, *args, **kwargs):\n '''Log a message to the console.\n\n Parameters\n ----------\n msg : str\n A string to display on the console. This can contain {}-style\n formatting commands; the remaining positional and keyword arguments\n will be used to fill them in.\n '''\n now = datetime.datetime.now()\n module = 'downhill'\n if _detailed_callsite:\n caller = inspect.stack()[1]\n parts = caller.filename.replace('.py', '').split('/')\n module = '{}:{}'.format(\n '.'.join(parts[parts.index('downhill')+1:]), caller.lineno)\n click.echo(' '.join((\n click.style(now.strftime('%Y%m%d'), fg='blue'),\n click.style(now.strftime('%H%M%S'), fg='cyan'),\n click.style(module, fg='magenta'),\n msg.format(*args, **kwargs),\n )))\n" ]
class Optimizer(util.Registrar(str('Base'), (), {})): '''An optimizer computes gradient updates to iteratively optimize a loss. Attributes ---------- patience : int, optional Number of validation "failures" that we are willing to tolerate before stopping the optimization process. A validation failure happens whenever the loss on the validation dataset decreases by less than ``min_improvement`` (relative) over the previous best validation loss. Defaults to 5. validate_every : int, optional Evaluate the loss on the validation dataset after making this many passes over the training data. Defaults to 10. min_improvement : float, optional Insist that the validation loss must improve by this relative amount before considering that the optimization has made progress. The optimization process halts when ``patience`` validations have failed to make this relative improvement. Defaults to 0; set to a larger value (e.g., 0.01 for 1% improvement) to halt the optimization process sooner. max_gradient_norm : float, optional Rescale each parameter's gradient so that it has at most this L2 norm. Set to 0 (the default) to disable norm rescaling. If ``max_gradient_elem`` is also specified, then this has no effect. max_gradient_elem : float, optional Perform elementwise clipping on the magnitude of gradient values. Set to 0 (the default) to disable. If elementwise clipping is enabled, norm rescaling (via ``max_gradient_norm``) will have no effect. Deprecated synonyms of this parameter are "max_gradient_clip" and "gradient_clip". learning_rate : float, optional Many SGD-based optimization algorithms require a learning rate hyperparameter that scales the gradient step. Defaults to 1e-4. momentum : float, optional Apply momentum to the parameter updates for this optimizer, with the given strength. Typically this value ranges from 0 (no momentum) to :math:`1 - \epsilon` (large momentum). Defaults to 0. nesterov : bool, optional If True, and ``momentum`` is nonzero, apply Nesterov-style momentum to parameter updates for this optimizer. If False, and ``momentum`` is nonzero, "regular" momentum is applied. Has no effect if ``momentum`` is zero. See :class:`NAG <downhill.NAG>` for a description of Nesterov momentum. Parameters ---------- loss : Theano expression Loss function to minimize. This must be a scalar-valued expression. params : list of Theano variables, optional Symbolic variables to adjust to minimize the loss. If not given, these will be computed automatically by walking the computation graph. inputs : list of Theano variables, optional Symbolic variables required to compute the loss. If not given, these will be computed automatically by walking the computation graph. updates : list of update pairs, optional A list of pairs providing updates for the internals of the loss computation. Normally this is empty, but it can be provided if the loss, for example, requires an update to an internal random number generator. monitors : sequence of (str, Theano expression) tuples, optional Additional values to monitor during optimization. These must be provided as a sequence of (name, expression) tuples. monitor_gradients : bool, optional If True, add monitors to log the norms of the parameter gradients during optimization. Defaults to False. ''' def __init__(self, loss, params=None, inputs=None, updates=(), monitors=(), monitor_gradients=False): inputs_, params_ = util.find_inputs_and_params(loss) self._loss = loss self._params = params or params_ self._inputs = inputs or inputs_ self._updates = updates self._shapes = [p.get_value(borrow=True).shape for p in self._params] self._counts = [np.prod(s) for s in self._shapes] self._starts = np.cumsum([0] + self._counts)[:-1] self._dtype = self._params[0].get_value().dtype self._curr_iter = 0 self._best_iter = 0 self._best_loss = 1e100 self._best_params = [p.get_value().copy() for p in self._params] self._monitor_exprs = [self._loss] self._monitor_names = ['loss'] for name, monitor in monitors: self._monitor_names.append(name) self._monitor_exprs.append(monitor) if monitor_gradients: unnamed = 0 for p, g in zip(self._params, TT.grad(self._loss, self._params)): name = p.name if not name: name = 'unnamed{}'.format(unnamed) unnamed += 1 util.log('"{}" unnamed, will be "{}" internally'.format(p, name)) self._monitor_names.append('grad({})'.format(name)) self._monitor_exprs.append((g * g).sum()) def _compile(self, **kwargs): '''Compile the Theano functions for evaluating and updating our model. ''' util.log('compiling evaluation function') self.f_eval = theano.function(self._inputs, self._monitor_exprs, updates=self._updates, name='evaluation') label = self.__class__.__name__ util.log('compiling {} optimizer'.format(click.style(label, fg='red'))) updates = list(self._updates) + list(self.get_updates(**kwargs)) self.f_step = theano.function(self._inputs, self._monitor_exprs, updates=updates, name=label) def get_updates(self, **kwargs): '''Get parameter update expressions for performing optimization. Keyword arguments can be applied here to set any of the global optimizer attributes. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' self._prepare(**kwargs) for param, grad in self._differentiate(): for var, update in self._get_updates_for(param, grad): # For auxiliary variables, updates are meant to replace the # existing variable value. if var != param: yield var, update continue # If momentum is disabled, just apply the parameter delta. if self.momentum == 0: yield var, param - update continue # Momentum is enabled, so we keep track of velocity here. vel_tm1 = util.shared_like(param, 'vel') vel_t = util.as_float(self.momentum) * vel_tm1 - update if self.nesterov: # see http://arxiv.org/pdf/1212.0901v2.pdf (eq 7) and # https://github.com/lisa-lab/pylearn2/pull/136#issuecomment-10381617 mom_sqr = util.as_float(self.momentum ** 2) mom_inc = util.as_float(1 + self.momentum) vel_t = mom_sqr * vel_tm1 - mom_inc * update yield vel_tm1, vel_t yield param, param + vel_t def _get_updates_for(self, param, grad): '''Generate some update pairs for the given model parameter. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' raise NotImplementedError def _differentiate(self, params=None): '''Return a sequence of gradients for our parameters. If this optimizer has been configured with a gradient norm limit, or with elementwise gradient clipping, this method applies the appropriate rescaling and clipping operations before returning the gradient. Parameters ---------- params : list of Theano variables, optional Return the gradient with respect to these parameters. Defaults to all parameters that the optimizer knows about. Yields ------ pairs : (param, grad) tuples Generates a sequence of tuples representing each of the parameters requested and the corresponding Theano gradient expressions. ''' if params is None: params = self._params for param, grad in zip(params, TT.grad(self._loss, params)): if self.max_gradient_elem > 0: limit = util.as_float(self.max_gradient_elem) yield param, TT.clip(grad, -limit, limit) elif self.max_gradient_norm > 0: norm = TT.sqrt((grad * grad).sum()) limit = util.as_float(self.max_gradient_norm) yield param, grad * TT.minimum(1, limit / norm) else: yield param, grad def set_params(self, targets=None): '''Set the values of the parameters to the given target values. Parameters ---------- targets : sequence of ndarray, optional Arrays for setting the parameters of our model. If this is not provided, the current best parameters for this optimizer will be used. ''' if not isinstance(targets, (list, tuple)): targets = self._best_params for param, target in zip(self._params, targets): param.set_value(target) def _log(self, monitors, iteration, label='', suffix=''): '''Log the state of the optimizer on the console. Parameters ---------- monitors : OrderedDict A dictionary of monitor names mapped to values. These names and values are what is being logged. iteration : int Optimization iteration that we are logging. label : str, optional A label for the name of the optimizer creating the log line. Defaults to the name of the current class. suffix : str, optional A suffix to add to the end of the log line, if any. ''' label = label or self.__class__.__name__ fields = (('{}={:.6f}').format(k, v) for k, v in monitors.items()) util.log('{} {} {}{}'.format(label, iteration, ' '.join(fields), suffix)) def evaluate(self, dataset): '''Evaluate the current model parameters on a dataset. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A set of data to use for evaluating the model. Returns ------- monitors : OrderedDict A dictionary mapping monitor names to values. Monitors are quantities of interest during optimization---for example, loss function, accuracy, or whatever the optimization task requires. ''' if dataset is None: values = [self.f_eval()] else: values = [self.f_eval(*x) for x in dataset] monitors = zip(self._monitor_names, np.mean(values, axis=0)) return collections.OrderedDict(monitors) def _test_patience(self, monitors): '''Test whether our patience with optimization has elapsed. Parameters ---------- monitors : dict A dictionary mapping monitor names to values. The 'loss' key from this dictionary will be used to evaluate optimization progress. Returns ------- elapsed : bool True iff our patience has elapsed and the model is no longer improving. ''' self._curr_iter += 1 marker = '' loss = monitors['loss'] if self._best_loss - loss > self._best_loss * self.min_improvement: self._best_loss = loss self._best_iter = self._curr_iter self._best_params = [p.get_value().copy() for p in self._params] marker = ' *' self._log(monitors, self._curr_iter - 1, 'validation', marker) return self._curr_iter - self._best_iter > self.patience def _prepare(self, **kwargs): '''Set up properties for optimization. This method can be overridden by base classes to provide parameters that are specific to a particular optimization technique (e.g., setting up a learning rate value). ''' self.learning_rate = util.as_float(kwargs.pop('learning_rate', 1e-4)) self.momentum = kwargs.pop('momentum', 0) self.nesterov = kwargs.pop('nesterov', False) self.patience = kwargs.get('patience', 5) self.validate_every = kwargs.pop('validate_every', 10) self.min_improvement = kwargs.pop('min_improvement', 0) self.max_gradient_norm = kwargs.pop('max_gradient_norm', 0) self.max_gradient_elem = kwargs.pop('max_gradient_elem', 0) util.log_param('patience', self.patience) util.log_param('validate_every', self.validate_every) util.log_param('min_improvement', self.min_improvement) util.log_param('max_gradient_norm', self.max_gradient_norm) util.log_param('max_gradient_elem', self.max_gradient_elem) util.log_param('learning_rate', self.learning_rate) util.log_param('momentum', self.momentum) util.log_param('nesterov', self.nesterov) def minimize(self, *args, **kwargs): '''Optimize our loss exhaustively. This method is a thin wrapper over the :func:`iterate` method. It simply exhausts the iterative optimization process and returns the final monitor values. Returns ------- train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' monitors = None for monitors in self.iterate(*args, **kwargs): pass return monitors def _step(self, dataset): '''Advance the state of the optimizer by one step. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A dataset for optimizing the model. Returns ------- train_monitors : dict A dictionary mapping monitor names to values. ''' if dataset is None: values = [self.f_step()] else: values = [self.f_step(*x) for x in dataset] return collections.OrderedDict( zip(self._monitor_names, np.mean(values, axis=0)))
lmjohns3/downhill
downhill/base.py
Optimizer.minimize
python
def minimize(self, *args, **kwargs): '''Optimize our loss exhaustively. This method is a thin wrapper over the :func:`iterate` method. It simply exhausts the iterative optimization process and returns the final monitor values. Returns ------- train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' monitors = None for monitors in self.iterate(*args, **kwargs): pass return monitors
Optimize our loss exhaustively. This method is a thin wrapper over the :func:`iterate` method. It simply exhausts the iterative optimization process and returns the final monitor values. Returns ------- train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/base.py#L417-L436
null
class Optimizer(util.Registrar(str('Base'), (), {})): '''An optimizer computes gradient updates to iteratively optimize a loss. Attributes ---------- patience : int, optional Number of validation "failures" that we are willing to tolerate before stopping the optimization process. A validation failure happens whenever the loss on the validation dataset decreases by less than ``min_improvement`` (relative) over the previous best validation loss. Defaults to 5. validate_every : int, optional Evaluate the loss on the validation dataset after making this many passes over the training data. Defaults to 10. min_improvement : float, optional Insist that the validation loss must improve by this relative amount before considering that the optimization has made progress. The optimization process halts when ``patience`` validations have failed to make this relative improvement. Defaults to 0; set to a larger value (e.g., 0.01 for 1% improvement) to halt the optimization process sooner. max_gradient_norm : float, optional Rescale each parameter's gradient so that it has at most this L2 norm. Set to 0 (the default) to disable norm rescaling. If ``max_gradient_elem`` is also specified, then this has no effect. max_gradient_elem : float, optional Perform elementwise clipping on the magnitude of gradient values. Set to 0 (the default) to disable. If elementwise clipping is enabled, norm rescaling (via ``max_gradient_norm``) will have no effect. Deprecated synonyms of this parameter are "max_gradient_clip" and "gradient_clip". learning_rate : float, optional Many SGD-based optimization algorithms require a learning rate hyperparameter that scales the gradient step. Defaults to 1e-4. momentum : float, optional Apply momentum to the parameter updates for this optimizer, with the given strength. Typically this value ranges from 0 (no momentum) to :math:`1 - \epsilon` (large momentum). Defaults to 0. nesterov : bool, optional If True, and ``momentum`` is nonzero, apply Nesterov-style momentum to parameter updates for this optimizer. If False, and ``momentum`` is nonzero, "regular" momentum is applied. Has no effect if ``momentum`` is zero. See :class:`NAG <downhill.NAG>` for a description of Nesterov momentum. Parameters ---------- loss : Theano expression Loss function to minimize. This must be a scalar-valued expression. params : list of Theano variables, optional Symbolic variables to adjust to minimize the loss. If not given, these will be computed automatically by walking the computation graph. inputs : list of Theano variables, optional Symbolic variables required to compute the loss. If not given, these will be computed automatically by walking the computation graph. updates : list of update pairs, optional A list of pairs providing updates for the internals of the loss computation. Normally this is empty, but it can be provided if the loss, for example, requires an update to an internal random number generator. monitors : sequence of (str, Theano expression) tuples, optional Additional values to monitor during optimization. These must be provided as a sequence of (name, expression) tuples. monitor_gradients : bool, optional If True, add monitors to log the norms of the parameter gradients during optimization. Defaults to False. ''' def __init__(self, loss, params=None, inputs=None, updates=(), monitors=(), monitor_gradients=False): inputs_, params_ = util.find_inputs_and_params(loss) self._loss = loss self._params = params or params_ self._inputs = inputs or inputs_ self._updates = updates self._shapes = [p.get_value(borrow=True).shape for p in self._params] self._counts = [np.prod(s) for s in self._shapes] self._starts = np.cumsum([0] + self._counts)[:-1] self._dtype = self._params[0].get_value().dtype self._curr_iter = 0 self._best_iter = 0 self._best_loss = 1e100 self._best_params = [p.get_value().copy() for p in self._params] self._monitor_exprs = [self._loss] self._monitor_names = ['loss'] for name, monitor in monitors: self._monitor_names.append(name) self._monitor_exprs.append(monitor) if monitor_gradients: unnamed = 0 for p, g in zip(self._params, TT.grad(self._loss, self._params)): name = p.name if not name: name = 'unnamed{}'.format(unnamed) unnamed += 1 util.log('"{}" unnamed, will be "{}" internally'.format(p, name)) self._monitor_names.append('grad({})'.format(name)) self._monitor_exprs.append((g * g).sum()) def _compile(self, **kwargs): '''Compile the Theano functions for evaluating and updating our model. ''' util.log('compiling evaluation function') self.f_eval = theano.function(self._inputs, self._monitor_exprs, updates=self._updates, name='evaluation') label = self.__class__.__name__ util.log('compiling {} optimizer'.format(click.style(label, fg='red'))) updates = list(self._updates) + list(self.get_updates(**kwargs)) self.f_step = theano.function(self._inputs, self._monitor_exprs, updates=updates, name=label) def get_updates(self, **kwargs): '''Get parameter update expressions for performing optimization. Keyword arguments can be applied here to set any of the global optimizer attributes. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' self._prepare(**kwargs) for param, grad in self._differentiate(): for var, update in self._get_updates_for(param, grad): # For auxiliary variables, updates are meant to replace the # existing variable value. if var != param: yield var, update continue # If momentum is disabled, just apply the parameter delta. if self.momentum == 0: yield var, param - update continue # Momentum is enabled, so we keep track of velocity here. vel_tm1 = util.shared_like(param, 'vel') vel_t = util.as_float(self.momentum) * vel_tm1 - update if self.nesterov: # see http://arxiv.org/pdf/1212.0901v2.pdf (eq 7) and # https://github.com/lisa-lab/pylearn2/pull/136#issuecomment-10381617 mom_sqr = util.as_float(self.momentum ** 2) mom_inc = util.as_float(1 + self.momentum) vel_t = mom_sqr * vel_tm1 - mom_inc * update yield vel_tm1, vel_t yield param, param + vel_t def _get_updates_for(self, param, grad): '''Generate some update pairs for the given model parameter. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' raise NotImplementedError def _differentiate(self, params=None): '''Return a sequence of gradients for our parameters. If this optimizer has been configured with a gradient norm limit, or with elementwise gradient clipping, this method applies the appropriate rescaling and clipping operations before returning the gradient. Parameters ---------- params : list of Theano variables, optional Return the gradient with respect to these parameters. Defaults to all parameters that the optimizer knows about. Yields ------ pairs : (param, grad) tuples Generates a sequence of tuples representing each of the parameters requested and the corresponding Theano gradient expressions. ''' if params is None: params = self._params for param, grad in zip(params, TT.grad(self._loss, params)): if self.max_gradient_elem > 0: limit = util.as_float(self.max_gradient_elem) yield param, TT.clip(grad, -limit, limit) elif self.max_gradient_norm > 0: norm = TT.sqrt((grad * grad).sum()) limit = util.as_float(self.max_gradient_norm) yield param, grad * TT.minimum(1, limit / norm) else: yield param, grad def set_params(self, targets=None): '''Set the values of the parameters to the given target values. Parameters ---------- targets : sequence of ndarray, optional Arrays for setting the parameters of our model. If this is not provided, the current best parameters for this optimizer will be used. ''' if not isinstance(targets, (list, tuple)): targets = self._best_params for param, target in zip(self._params, targets): param.set_value(target) def _log(self, monitors, iteration, label='', suffix=''): '''Log the state of the optimizer on the console. Parameters ---------- monitors : OrderedDict A dictionary of monitor names mapped to values. These names and values are what is being logged. iteration : int Optimization iteration that we are logging. label : str, optional A label for the name of the optimizer creating the log line. Defaults to the name of the current class. suffix : str, optional A suffix to add to the end of the log line, if any. ''' label = label or self.__class__.__name__ fields = (('{}={:.6f}').format(k, v) for k, v in monitors.items()) util.log('{} {} {}{}'.format(label, iteration, ' '.join(fields), suffix)) def evaluate(self, dataset): '''Evaluate the current model parameters on a dataset. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A set of data to use for evaluating the model. Returns ------- monitors : OrderedDict A dictionary mapping monitor names to values. Monitors are quantities of interest during optimization---for example, loss function, accuracy, or whatever the optimization task requires. ''' if dataset is None: values = [self.f_eval()] else: values = [self.f_eval(*x) for x in dataset] monitors = zip(self._monitor_names, np.mean(values, axis=0)) return collections.OrderedDict(monitors) def _test_patience(self, monitors): '''Test whether our patience with optimization has elapsed. Parameters ---------- monitors : dict A dictionary mapping monitor names to values. The 'loss' key from this dictionary will be used to evaluate optimization progress. Returns ------- elapsed : bool True iff our patience has elapsed and the model is no longer improving. ''' self._curr_iter += 1 marker = '' loss = monitors['loss'] if self._best_loss - loss > self._best_loss * self.min_improvement: self._best_loss = loss self._best_iter = self._curr_iter self._best_params = [p.get_value().copy() for p in self._params] marker = ' *' self._log(monitors, self._curr_iter - 1, 'validation', marker) return self._curr_iter - self._best_iter > self.patience def _prepare(self, **kwargs): '''Set up properties for optimization. This method can be overridden by base classes to provide parameters that are specific to a particular optimization technique (e.g., setting up a learning rate value). ''' self.learning_rate = util.as_float(kwargs.pop('learning_rate', 1e-4)) self.momentum = kwargs.pop('momentum', 0) self.nesterov = kwargs.pop('nesterov', False) self.patience = kwargs.get('patience', 5) self.validate_every = kwargs.pop('validate_every', 10) self.min_improvement = kwargs.pop('min_improvement', 0) self.max_gradient_norm = kwargs.pop('max_gradient_norm', 0) self.max_gradient_elem = kwargs.pop('max_gradient_elem', 0) util.log_param('patience', self.patience) util.log_param('validate_every', self.validate_every) util.log_param('min_improvement', self.min_improvement) util.log_param('max_gradient_norm', self.max_gradient_norm) util.log_param('max_gradient_elem', self.max_gradient_elem) util.log_param('learning_rate', self.learning_rate) util.log_param('momentum', self.momentum) util.log_param('nesterov', self.nesterov) def iterate(self, train=None, valid=None, max_updates=None, **kwargs): r'''Optimize a loss iteratively using a training and validation dataset. This method yields a series of monitor values to the caller. After every optimization epoch, a pair of monitor dictionaries is generated: one evaluated on the training dataset during the epoch, and another evaluated on the validation dataset at the most recent validation epoch. The validation monitors might not be updated during every optimization iteration; in this case, the most recent validation monitors will be yielded along with the training monitors. Additional keyword arguments supplied here will set the global optimizer attributes. Parameters ---------- train : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of training data for computing updates to model parameters. valid : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of validation data for computing monitor values and determining when the loss has stopped improving. Defaults to the training data. max_updates : int, optional If specified, halt optimization after this many gradient updates have been processed. If not provided, uses early stopping to decide when to halt. Yields ------ train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' self._compile(**kwargs) if valid is None: valid = train iteration = 0 training = validation = None while max_updates is None or iteration < max_updates: if not iteration % self.validate_every: try: validation = self.evaluate(valid) except KeyboardInterrupt: util.log('interrupted!') break if self._test_patience(validation): util.log('patience elapsed!') break try: training = self._step(train) except KeyboardInterrupt: util.log('interrupted!') break iteration += 1 self._log(training, iteration) yield training, validation self.set_params('best') def _step(self, dataset): '''Advance the state of the optimizer by one step. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A dataset for optimizing the model. Returns ------- train_monitors : dict A dictionary mapping monitor names to values. ''' if dataset is None: values = [self.f_step()] else: values = [self.f_step(*x) for x in dataset] return collections.OrderedDict( zip(self._monitor_names, np.mean(values, axis=0)))
lmjohns3/downhill
downhill/base.py
Optimizer._step
python
def _step(self, dataset): '''Advance the state of the optimizer by one step. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A dataset for optimizing the model. Returns ------- train_monitors : dict A dictionary mapping monitor names to values. ''' if dataset is None: values = [self.f_step()] else: values = [self.f_step(*x) for x in dataset] return collections.OrderedDict( zip(self._monitor_names, np.mean(values, axis=0)))
Advance the state of the optimizer by one step. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A dataset for optimizing the model. Returns ------- train_monitors : dict A dictionary mapping monitor names to values.
train
https://github.com/lmjohns3/downhill/blob/42111ab03b5e6fa47b7bf7c7cb5caa402f10ce6d/downhill/base.py#L438-L456
null
class Optimizer(util.Registrar(str('Base'), (), {})): '''An optimizer computes gradient updates to iteratively optimize a loss. Attributes ---------- patience : int, optional Number of validation "failures" that we are willing to tolerate before stopping the optimization process. A validation failure happens whenever the loss on the validation dataset decreases by less than ``min_improvement`` (relative) over the previous best validation loss. Defaults to 5. validate_every : int, optional Evaluate the loss on the validation dataset after making this many passes over the training data. Defaults to 10. min_improvement : float, optional Insist that the validation loss must improve by this relative amount before considering that the optimization has made progress. The optimization process halts when ``patience`` validations have failed to make this relative improvement. Defaults to 0; set to a larger value (e.g., 0.01 for 1% improvement) to halt the optimization process sooner. max_gradient_norm : float, optional Rescale each parameter's gradient so that it has at most this L2 norm. Set to 0 (the default) to disable norm rescaling. If ``max_gradient_elem`` is also specified, then this has no effect. max_gradient_elem : float, optional Perform elementwise clipping on the magnitude of gradient values. Set to 0 (the default) to disable. If elementwise clipping is enabled, norm rescaling (via ``max_gradient_norm``) will have no effect. Deprecated synonyms of this parameter are "max_gradient_clip" and "gradient_clip". learning_rate : float, optional Many SGD-based optimization algorithms require a learning rate hyperparameter that scales the gradient step. Defaults to 1e-4. momentum : float, optional Apply momentum to the parameter updates for this optimizer, with the given strength. Typically this value ranges from 0 (no momentum) to :math:`1 - \epsilon` (large momentum). Defaults to 0. nesterov : bool, optional If True, and ``momentum`` is nonzero, apply Nesterov-style momentum to parameter updates for this optimizer. If False, and ``momentum`` is nonzero, "regular" momentum is applied. Has no effect if ``momentum`` is zero. See :class:`NAG <downhill.NAG>` for a description of Nesterov momentum. Parameters ---------- loss : Theano expression Loss function to minimize. This must be a scalar-valued expression. params : list of Theano variables, optional Symbolic variables to adjust to minimize the loss. If not given, these will be computed automatically by walking the computation graph. inputs : list of Theano variables, optional Symbolic variables required to compute the loss. If not given, these will be computed automatically by walking the computation graph. updates : list of update pairs, optional A list of pairs providing updates for the internals of the loss computation. Normally this is empty, but it can be provided if the loss, for example, requires an update to an internal random number generator. monitors : sequence of (str, Theano expression) tuples, optional Additional values to monitor during optimization. These must be provided as a sequence of (name, expression) tuples. monitor_gradients : bool, optional If True, add monitors to log the norms of the parameter gradients during optimization. Defaults to False. ''' def __init__(self, loss, params=None, inputs=None, updates=(), monitors=(), monitor_gradients=False): inputs_, params_ = util.find_inputs_and_params(loss) self._loss = loss self._params = params or params_ self._inputs = inputs or inputs_ self._updates = updates self._shapes = [p.get_value(borrow=True).shape for p in self._params] self._counts = [np.prod(s) for s in self._shapes] self._starts = np.cumsum([0] + self._counts)[:-1] self._dtype = self._params[0].get_value().dtype self._curr_iter = 0 self._best_iter = 0 self._best_loss = 1e100 self._best_params = [p.get_value().copy() for p in self._params] self._monitor_exprs = [self._loss] self._monitor_names = ['loss'] for name, monitor in monitors: self._monitor_names.append(name) self._monitor_exprs.append(monitor) if monitor_gradients: unnamed = 0 for p, g in zip(self._params, TT.grad(self._loss, self._params)): name = p.name if not name: name = 'unnamed{}'.format(unnamed) unnamed += 1 util.log('"{}" unnamed, will be "{}" internally'.format(p, name)) self._monitor_names.append('grad({})'.format(name)) self._monitor_exprs.append((g * g).sum()) def _compile(self, **kwargs): '''Compile the Theano functions for evaluating and updating our model. ''' util.log('compiling evaluation function') self.f_eval = theano.function(self._inputs, self._monitor_exprs, updates=self._updates, name='evaluation') label = self.__class__.__name__ util.log('compiling {} optimizer'.format(click.style(label, fg='red'))) updates = list(self._updates) + list(self.get_updates(**kwargs)) self.f_step = theano.function(self._inputs, self._monitor_exprs, updates=updates, name=label) def get_updates(self, **kwargs): '''Get parameter update expressions for performing optimization. Keyword arguments can be applied here to set any of the global optimizer attributes. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' self._prepare(**kwargs) for param, grad in self._differentiate(): for var, update in self._get_updates_for(param, grad): # For auxiliary variables, updates are meant to replace the # existing variable value. if var != param: yield var, update continue # If momentum is disabled, just apply the parameter delta. if self.momentum == 0: yield var, param - update continue # Momentum is enabled, so we keep track of velocity here. vel_tm1 = util.shared_like(param, 'vel') vel_t = util.as_float(self.momentum) * vel_tm1 - update if self.nesterov: # see http://arxiv.org/pdf/1212.0901v2.pdf (eq 7) and # https://github.com/lisa-lab/pylearn2/pull/136#issuecomment-10381617 mom_sqr = util.as_float(self.momentum ** 2) mom_inc = util.as_float(1 + self.momentum) vel_t = mom_sqr * vel_tm1 - mom_inc * update yield vel_tm1, vel_t yield param, param + vel_t def _get_updates_for(self, param, grad): '''Generate some update pairs for the given model parameter. Yields ------ updates : (parameter, expression) tuples A sequence of parameter updates to be applied during optimization. ''' raise NotImplementedError def _differentiate(self, params=None): '''Return a sequence of gradients for our parameters. If this optimizer has been configured with a gradient norm limit, or with elementwise gradient clipping, this method applies the appropriate rescaling and clipping operations before returning the gradient. Parameters ---------- params : list of Theano variables, optional Return the gradient with respect to these parameters. Defaults to all parameters that the optimizer knows about. Yields ------ pairs : (param, grad) tuples Generates a sequence of tuples representing each of the parameters requested and the corresponding Theano gradient expressions. ''' if params is None: params = self._params for param, grad in zip(params, TT.grad(self._loss, params)): if self.max_gradient_elem > 0: limit = util.as_float(self.max_gradient_elem) yield param, TT.clip(grad, -limit, limit) elif self.max_gradient_norm > 0: norm = TT.sqrt((grad * grad).sum()) limit = util.as_float(self.max_gradient_norm) yield param, grad * TT.minimum(1, limit / norm) else: yield param, grad def set_params(self, targets=None): '''Set the values of the parameters to the given target values. Parameters ---------- targets : sequence of ndarray, optional Arrays for setting the parameters of our model. If this is not provided, the current best parameters for this optimizer will be used. ''' if not isinstance(targets, (list, tuple)): targets = self._best_params for param, target in zip(self._params, targets): param.set_value(target) def _log(self, monitors, iteration, label='', suffix=''): '''Log the state of the optimizer on the console. Parameters ---------- monitors : OrderedDict A dictionary of monitor names mapped to values. These names and values are what is being logged. iteration : int Optimization iteration that we are logging. label : str, optional A label for the name of the optimizer creating the log line. Defaults to the name of the current class. suffix : str, optional A suffix to add to the end of the log line, if any. ''' label = label or self.__class__.__name__ fields = (('{}={:.6f}').format(k, v) for k, v in monitors.items()) util.log('{} {} {}{}'.format(label, iteration, ' '.join(fields), suffix)) def evaluate(self, dataset): '''Evaluate the current model parameters on a dataset. Parameters ---------- dataset : :class:`Dataset <downhill.dataset.Dataset>` A set of data to use for evaluating the model. Returns ------- monitors : OrderedDict A dictionary mapping monitor names to values. Monitors are quantities of interest during optimization---for example, loss function, accuracy, or whatever the optimization task requires. ''' if dataset is None: values = [self.f_eval()] else: values = [self.f_eval(*x) for x in dataset] monitors = zip(self._monitor_names, np.mean(values, axis=0)) return collections.OrderedDict(monitors) def _test_patience(self, monitors): '''Test whether our patience with optimization has elapsed. Parameters ---------- monitors : dict A dictionary mapping monitor names to values. The 'loss' key from this dictionary will be used to evaluate optimization progress. Returns ------- elapsed : bool True iff our patience has elapsed and the model is no longer improving. ''' self._curr_iter += 1 marker = '' loss = monitors['loss'] if self._best_loss - loss > self._best_loss * self.min_improvement: self._best_loss = loss self._best_iter = self._curr_iter self._best_params = [p.get_value().copy() for p in self._params] marker = ' *' self._log(monitors, self._curr_iter - 1, 'validation', marker) return self._curr_iter - self._best_iter > self.patience def _prepare(self, **kwargs): '''Set up properties for optimization. This method can be overridden by base classes to provide parameters that are specific to a particular optimization technique (e.g., setting up a learning rate value). ''' self.learning_rate = util.as_float(kwargs.pop('learning_rate', 1e-4)) self.momentum = kwargs.pop('momentum', 0) self.nesterov = kwargs.pop('nesterov', False) self.patience = kwargs.get('patience', 5) self.validate_every = kwargs.pop('validate_every', 10) self.min_improvement = kwargs.pop('min_improvement', 0) self.max_gradient_norm = kwargs.pop('max_gradient_norm', 0) self.max_gradient_elem = kwargs.pop('max_gradient_elem', 0) util.log_param('patience', self.patience) util.log_param('validate_every', self.validate_every) util.log_param('min_improvement', self.min_improvement) util.log_param('max_gradient_norm', self.max_gradient_norm) util.log_param('max_gradient_elem', self.max_gradient_elem) util.log_param('learning_rate', self.learning_rate) util.log_param('momentum', self.momentum) util.log_param('nesterov', self.nesterov) def iterate(self, train=None, valid=None, max_updates=None, **kwargs): r'''Optimize a loss iteratively using a training and validation dataset. This method yields a series of monitor values to the caller. After every optimization epoch, a pair of monitor dictionaries is generated: one evaluated on the training dataset during the epoch, and another evaluated on the validation dataset at the most recent validation epoch. The validation monitors might not be updated during every optimization iteration; in this case, the most recent validation monitors will be yielded along with the training monitors. Additional keyword arguments supplied here will set the global optimizer attributes. Parameters ---------- train : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of training data for computing updates to model parameters. valid : sequence or :class:`Dataset <downhill.dataset.Dataset>` A set of validation data for computing monitor values and determining when the loss has stopped improving. Defaults to the training data. max_updates : int, optional If specified, halt optimization after this many gradient updates have been processed. If not provided, uses early stopping to decide when to halt. Yields ------ train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' self._compile(**kwargs) if valid is None: valid = train iteration = 0 training = validation = None while max_updates is None or iteration < max_updates: if not iteration % self.validate_every: try: validation = self.evaluate(valid) except KeyboardInterrupt: util.log('interrupted!') break if self._test_patience(validation): util.log('patience elapsed!') break try: training = self._step(train) except KeyboardInterrupt: util.log('interrupted!') break iteration += 1 self._log(training, iteration) yield training, validation self.set_params('best') def minimize(self, *args, **kwargs): '''Optimize our loss exhaustively. This method is a thin wrapper over the :func:`iterate` method. It simply exhausts the iterative optimization process and returns the final monitor values. Returns ------- train_monitors : dict A dictionary mapping monitor names to values, evaluated on the training dataset. valid_monitors : dict A dictionary containing monitor values evaluated on the validation dataset. ''' monitors = None for monitors in self.iterate(*args, **kwargs): pass return monitors
HdrHistogram/HdrHistogram_py
hdrh/codec.py
HdrPayload.init_counts
python
def init_counts(self, counts_len): '''Called after instantiating with a compressed payload Params: counts_len counts size to use based on decoded settings in the header ''' assert self._data and counts_len and self.counts_len == 0 self.counts_len = counts_len self._init_counts() results = decode(self._data, payload_header_size, addressof(self.counts), counts_len, self.word_size) # no longer needed self._data = None return results
Called after instantiating with a compressed payload Params: counts_len counts size to use based on decoded settings in the header
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/codec.py#L156-L169
[ "def _init_counts(self):\n self.counts = (self.counter_ctype * self.counts_len)()\n" ]
class HdrPayload(): '''A class that wraps the ctypes big endian struct that will hold the histogram wire format content (including the counters). ''' def __init__(self, word_size, counts_len=0, compressed_payload=None): '''Two ways to use this class: - for an empty histogram (pass counts_len>0 and compressed_payload=None) - for a decoded histogram (counts_len=0 and compressed_payload!=None) Params: word_size counter size in bytes (2,4,8 byte counters are supported) counts_len number of counters to allocate ignored if a compressed payload is provided (not None) compressed_payload (string) a payload in zlib compressed form, decompress and decode the payload header. Caller must then invoke init_counts to pass in counts_len so that the counts array can be updated from the decoded varint buffer None if no compressed payload is to be associated to this instance ''' self.word_size = word_size self.counts_len = counts_len self._data = None try: # ctype counter type self.counter_ctype = payload_counter_ctype[word_size] except IndexError: raise ValueError('Invalid word size') if not self.counter_ctype: raise ValueError('Invalid word size') if compressed_payload: self._decompress(compressed_payload) elif counts_len: self.payload = PayloadHeader() self.payload.cookie = get_encoding_cookie() self._init_counts() else: raise RuntimeError('counts_len cannot be zero') def _init_counts(self): self.counts = (self.counter_ctype * self.counts_len)() def get_counts(self): return self.counts def _decompress(self, compressed_payload): '''Decompress a compressed payload into this payload wrapper. Note that the decompressed buffer is saved in self._data and the counts array is not yet allocated. Args: compressed_payload (string) a payload in zlib compressed form Exception: HdrCookieException: the compressed payload has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class HdrHistogramSettingsException: mismatch in the significant figures, lowest and highest trackable value ''' # make sure this instance is pristine if self._data: raise RuntimeError('Cannot decompress to an instance with payload') # Here it is important to keep a reference to the decompressed # string so that it does not get garbage collected self._data = zlib.decompress(compressed_payload) len_data = len(self._data) counts_size = len_data - payload_header_size if payload_header_size > counts_size > MAX_COUNTS_SIZE: raise HdrLengthException('Invalid size:' + str(len_data)) # copy the first bytes for the header self.payload = PayloadHeader.from_buffer_copy(self._data) cookie = self.payload.cookie if get_cookie_base(cookie) != V2_ENCODING_COOKIE_BASE: raise HdrCookieException('Invalid cookie: %x' % cookie) word_size = get_word_size_in_bytes_from_cookie(cookie) if word_size != V2_MAX_WORD_SIZE_IN_BYTES: raise HdrCookieException('Invalid V2 cookie: %x' % cookie) def compress(self, counts_limit): '''Compress this payload instance Args: counts_limit how many counters should be encoded starting from index 0 (can be 0), Return: the compressed payload (python string) ''' if self.payload: # worst case varint encoded length is when each counter is at the maximum value # in this case 1 more byte per counter is needed due to the more bits varint_len = counts_limit * (self.word_size + 1) # allocate enough space to fit the header and the varint string encode_buf = (c_byte * (payload_header_size + varint_len))() # encode past the payload header varint_len = encode(addressof(self.counts), counts_limit, self.word_size, addressof(encode_buf) + payload_header_size, varint_len) # copy the header after updating the varint stream length self.payload.payload_len = varint_len ctypes.memmove(addressof(encode_buf), addressof(self.payload), payload_header_size) cdata = zlib.compress(ctypes.string_at(encode_buf, payload_header_size + varint_len)) return cdata # can't compress if no payload raise RuntimeError('No payload to compress') def dump(self, label=None): if label: print('Payload Dump ' + label) print(' payload cookie: %x' % (self.payload.cookie)) print(' payload_len: %d' % (self.payload.payload_len)) print(' counts_len: %d' % (self.counts_len)) dump_payload(self.get_counts(), self.counts_len)
HdrHistogram/HdrHistogram_py
hdrh/codec.py
HdrPayload._decompress
python
def _decompress(self, compressed_payload): '''Decompress a compressed payload into this payload wrapper. Note that the decompressed buffer is saved in self._data and the counts array is not yet allocated. Args: compressed_payload (string) a payload in zlib compressed form Exception: HdrCookieException: the compressed payload has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class HdrHistogramSettingsException: mismatch in the significant figures, lowest and highest trackable value ''' # make sure this instance is pristine if self._data: raise RuntimeError('Cannot decompress to an instance with payload') # Here it is important to keep a reference to the decompressed # string so that it does not get garbage collected self._data = zlib.decompress(compressed_payload) len_data = len(self._data) counts_size = len_data - payload_header_size if payload_header_size > counts_size > MAX_COUNTS_SIZE: raise HdrLengthException('Invalid size:' + str(len_data)) # copy the first bytes for the header self.payload = PayloadHeader.from_buffer_copy(self._data) cookie = self.payload.cookie if get_cookie_base(cookie) != V2_ENCODING_COOKIE_BASE: raise HdrCookieException('Invalid cookie: %x' % cookie) word_size = get_word_size_in_bytes_from_cookie(cookie) if word_size != V2_MAX_WORD_SIZE_IN_BYTES: raise HdrCookieException('Invalid V2 cookie: %x' % cookie)
Decompress a compressed payload into this payload wrapper. Note that the decompressed buffer is saved in self._data and the counts array is not yet allocated. Args: compressed_payload (string) a payload in zlib compressed form Exception: HdrCookieException: the compressed payload has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class HdrHistogramSettingsException: mismatch in the significant figures, lowest and highest trackable value
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/codec.py#L174-L211
[ "def get_cookie_base(cookie):\n return cookie & ~0xf0\n", "def get_word_size_in_bytes_from_cookie(cookie):\n if (get_cookie_base(cookie) == V2_ENCODING_COOKIE_BASE) or \\\n (get_cookie_base(cookie) == V2_COMPRESSION_COOKIE_BASE):\n return V2_MAX_WORD_SIZE_IN_BYTES\n return (cookie & 0xf0) >> 4\n" ]
class HdrPayload(): '''A class that wraps the ctypes big endian struct that will hold the histogram wire format content (including the counters). ''' def __init__(self, word_size, counts_len=0, compressed_payload=None): '''Two ways to use this class: - for an empty histogram (pass counts_len>0 and compressed_payload=None) - for a decoded histogram (counts_len=0 and compressed_payload!=None) Params: word_size counter size in bytes (2,4,8 byte counters are supported) counts_len number of counters to allocate ignored if a compressed payload is provided (not None) compressed_payload (string) a payload in zlib compressed form, decompress and decode the payload header. Caller must then invoke init_counts to pass in counts_len so that the counts array can be updated from the decoded varint buffer None if no compressed payload is to be associated to this instance ''' self.word_size = word_size self.counts_len = counts_len self._data = None try: # ctype counter type self.counter_ctype = payload_counter_ctype[word_size] except IndexError: raise ValueError('Invalid word size') if not self.counter_ctype: raise ValueError('Invalid word size') if compressed_payload: self._decompress(compressed_payload) elif counts_len: self.payload = PayloadHeader() self.payload.cookie = get_encoding_cookie() self._init_counts() else: raise RuntimeError('counts_len cannot be zero') def _init_counts(self): self.counts = (self.counter_ctype * self.counts_len)() def init_counts(self, counts_len): '''Called after instantiating with a compressed payload Params: counts_len counts size to use based on decoded settings in the header ''' assert self._data and counts_len and self.counts_len == 0 self.counts_len = counts_len self._init_counts() results = decode(self._data, payload_header_size, addressof(self.counts), counts_len, self.word_size) # no longer needed self._data = None return results def get_counts(self): return self.counts def compress(self, counts_limit): '''Compress this payload instance Args: counts_limit how many counters should be encoded starting from index 0 (can be 0), Return: the compressed payload (python string) ''' if self.payload: # worst case varint encoded length is when each counter is at the maximum value # in this case 1 more byte per counter is needed due to the more bits varint_len = counts_limit * (self.word_size + 1) # allocate enough space to fit the header and the varint string encode_buf = (c_byte * (payload_header_size + varint_len))() # encode past the payload header varint_len = encode(addressof(self.counts), counts_limit, self.word_size, addressof(encode_buf) + payload_header_size, varint_len) # copy the header after updating the varint stream length self.payload.payload_len = varint_len ctypes.memmove(addressof(encode_buf), addressof(self.payload), payload_header_size) cdata = zlib.compress(ctypes.string_at(encode_buf, payload_header_size + varint_len)) return cdata # can't compress if no payload raise RuntimeError('No payload to compress') def dump(self, label=None): if label: print('Payload Dump ' + label) print(' payload cookie: %x' % (self.payload.cookie)) print(' payload_len: %d' % (self.payload.payload_len)) print(' counts_len: %d' % (self.counts_len)) dump_payload(self.get_counts(), self.counts_len)
HdrHistogram/HdrHistogram_py
hdrh/codec.py
HdrPayload.compress
python
def compress(self, counts_limit): '''Compress this payload instance Args: counts_limit how many counters should be encoded starting from index 0 (can be 0), Return: the compressed payload (python string) ''' if self.payload: # worst case varint encoded length is when each counter is at the maximum value # in this case 1 more byte per counter is needed due to the more bits varint_len = counts_limit * (self.word_size + 1) # allocate enough space to fit the header and the varint string encode_buf = (c_byte * (payload_header_size + varint_len))() # encode past the payload header varint_len = encode(addressof(self.counts), counts_limit, self.word_size, addressof(encode_buf) + payload_header_size, varint_len) # copy the header after updating the varint stream length self.payload.payload_len = varint_len ctypes.memmove(addressof(encode_buf), addressof(self.payload), payload_header_size) cdata = zlib.compress(ctypes.string_at(encode_buf, payload_header_size + varint_len)) return cdata # can't compress if no payload raise RuntimeError('No payload to compress')
Compress this payload instance Args: counts_limit how many counters should be encoded starting from index 0 (can be 0), Return: the compressed payload (python string)
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/codec.py#L213-L241
null
class HdrPayload(): '''A class that wraps the ctypes big endian struct that will hold the histogram wire format content (including the counters). ''' def __init__(self, word_size, counts_len=0, compressed_payload=None): '''Two ways to use this class: - for an empty histogram (pass counts_len>0 and compressed_payload=None) - for a decoded histogram (counts_len=0 and compressed_payload!=None) Params: word_size counter size in bytes (2,4,8 byte counters are supported) counts_len number of counters to allocate ignored if a compressed payload is provided (not None) compressed_payload (string) a payload in zlib compressed form, decompress and decode the payload header. Caller must then invoke init_counts to pass in counts_len so that the counts array can be updated from the decoded varint buffer None if no compressed payload is to be associated to this instance ''' self.word_size = word_size self.counts_len = counts_len self._data = None try: # ctype counter type self.counter_ctype = payload_counter_ctype[word_size] except IndexError: raise ValueError('Invalid word size') if not self.counter_ctype: raise ValueError('Invalid word size') if compressed_payload: self._decompress(compressed_payload) elif counts_len: self.payload = PayloadHeader() self.payload.cookie = get_encoding_cookie() self._init_counts() else: raise RuntimeError('counts_len cannot be zero') def _init_counts(self): self.counts = (self.counter_ctype * self.counts_len)() def init_counts(self, counts_len): '''Called after instantiating with a compressed payload Params: counts_len counts size to use based on decoded settings in the header ''' assert self._data and counts_len and self.counts_len == 0 self.counts_len = counts_len self._init_counts() results = decode(self._data, payload_header_size, addressof(self.counts), counts_len, self.word_size) # no longer needed self._data = None return results def get_counts(self): return self.counts def _decompress(self, compressed_payload): '''Decompress a compressed payload into this payload wrapper. Note that the decompressed buffer is saved in self._data and the counts array is not yet allocated. Args: compressed_payload (string) a payload in zlib compressed form Exception: HdrCookieException: the compressed payload has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class HdrHistogramSettingsException: mismatch in the significant figures, lowest and highest trackable value ''' # make sure this instance is pristine if self._data: raise RuntimeError('Cannot decompress to an instance with payload') # Here it is important to keep a reference to the decompressed # string so that it does not get garbage collected self._data = zlib.decompress(compressed_payload) len_data = len(self._data) counts_size = len_data - payload_header_size if payload_header_size > counts_size > MAX_COUNTS_SIZE: raise HdrLengthException('Invalid size:' + str(len_data)) # copy the first bytes for the header self.payload = PayloadHeader.from_buffer_copy(self._data) cookie = self.payload.cookie if get_cookie_base(cookie) != V2_ENCODING_COOKIE_BASE: raise HdrCookieException('Invalid cookie: %x' % cookie) word_size = get_word_size_in_bytes_from_cookie(cookie) if word_size != V2_MAX_WORD_SIZE_IN_BYTES: raise HdrCookieException('Invalid V2 cookie: %x' % cookie) def dump(self, label=None): if label: print('Payload Dump ' + label) print(' payload cookie: %x' % (self.payload.cookie)) print(' payload_len: %d' % (self.payload.payload_len)) print(' counts_len: %d' % (self.counts_len)) dump_payload(self.get_counts(), self.counts_len)
HdrHistogram/HdrHistogram_py
hdrh/codec.py
HdrHistogramEncoder.encode
python
def encode(self): '''Compress the associated encodable payload, prepend the header then encode with base64 if requested Returns: the b64 encoded wire encoding of the histogram (as a string) or the compressed payload (as a string, if b64 wrappinb is disabled) ''' # only compress the first non zero buckets # if histogram is empty we do not encode any counter if self.histogram.total_count: relevant_length = \ self.histogram.get_counts_array_index(self.histogram.max_value) + 1 else: relevant_length = 0 cpayload = self.payload.compress(relevant_length) if self.b64_wrap: self.header.length = len(cpayload) header_str = ctypes.string_at(addressof(self.header), ext_header_size) return base64.b64encode(header_str + cpayload) return cpayload
Compress the associated encodable payload, prepend the header then encode with base64 if requested Returns: the b64 encoded wire encoding of the histogram (as a string) or the compressed payload (as a string, if b64 wrappinb is disabled)
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/codec.py#L290-L310
null
class HdrHistogramEncoder(): '''An encoder class for histograms, only supports V1 encoding. ''' def __init__(self, histogram, b64_wrap=True, hdr_payload=None): '''Histogram encoder Args: histogram the histogram to encode/decode into b64_wrap determines if the base64 wrapper is enabled or not hdr_payload if None will create a new HdrPayload instance for this encoder, else will reuse the passed Hdrayload instance (useful after decoding one and to associate it to a new histogram) word_size counters size in bytes (2, 4 or 8) Exceptions: ValueError if the word_size value is unsupported ''' self.histogram = histogram if not hdr_payload: self.payload = HdrPayload(histogram.word_size, histogram.counts_len) payload = self.payload.payload # those values never change across encodings payload.normalizing_index_offset = 0 payload.conversion_ratio_bits = 1 payload.significant_figures = histogram.significant_figures payload.lowest_trackable_value = histogram.lowest_trackable_value payload.highest_trackable_value = histogram.highest_trackable_value else: self.payload = hdr_payload self.b64_wrap = b64_wrap self.header = ExternalHeader() self.header.cookie = get_compression_cookie() def get_counts(self): '''Retrieve the counts array that can be used to store live counters and that can be encoded with minimal copies using encode() ''' return self.payload.get_counts() @staticmethod def decode(encoded_histogram, b64_wrap=True): '''Decode a wire histogram encoding into a read-only Hdr Payload instance Args: encoded_histogram a string containing the wire encoding of a histogram such as one returned from encode() Returns: an hdr_payload instance with all the decoded/uncompressed fields Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class HdrHistogramSettingsException: mismatch in the significant figures, lowest and highest trackable value zlib.error: in case of zlib decompression error ''' if b64_wrap: b64decode = base64.b64decode(encoded_histogram) # this string has 2 parts in it: the header (raw) and the payload (compressed) b64dec_len = len(b64decode) if b64dec_len < ext_header_size: raise HdrLengthException('Base64 decoded message too short') header = ExternalHeader.from_buffer_copy(b64decode) if get_cookie_base(header.cookie) != V2_COMPRESSION_COOKIE_BASE: raise HdrCookieException() if header.length != b64dec_len - ext_header_size: raise HdrLengthException('Decoded length=%d buffer length=%d' % (header.length, b64dec_len - ext_header_size)) # this will result in a copy of the compressed payload part # could not find a way to do otherwise since zlib.decompress() # expects a string (and does not like a buffer or a memoryview object) cpayload = b64decode[ext_header_size:] else: cpayload = encoded_histogram hdr_payload = HdrPayload(8, compressed_payload=cpayload) return hdr_payload def add(self, other_encoder): add_array(addressof(self.get_counts()), addressof(other_encoder.get_counts()), self.histogram.counts_len, self.histogram.word_size)
HdrHistogram/HdrHistogram_py
hdrh/codec.py
HdrHistogramEncoder.decode
python
def decode(encoded_histogram, b64_wrap=True): '''Decode a wire histogram encoding into a read-only Hdr Payload instance Args: encoded_histogram a string containing the wire encoding of a histogram such as one returned from encode() Returns: an hdr_payload instance with all the decoded/uncompressed fields Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class HdrHistogramSettingsException: mismatch in the significant figures, lowest and highest trackable value zlib.error: in case of zlib decompression error ''' if b64_wrap: b64decode = base64.b64decode(encoded_histogram) # this string has 2 parts in it: the header (raw) and the payload (compressed) b64dec_len = len(b64decode) if b64dec_len < ext_header_size: raise HdrLengthException('Base64 decoded message too short') header = ExternalHeader.from_buffer_copy(b64decode) if get_cookie_base(header.cookie) != V2_COMPRESSION_COOKIE_BASE: raise HdrCookieException() if header.length != b64dec_len - ext_header_size: raise HdrLengthException('Decoded length=%d buffer length=%d' % (header.length, b64dec_len - ext_header_size)) # this will result in a copy of the compressed payload part # could not find a way to do otherwise since zlib.decompress() # expects a string (and does not like a buffer or a memoryview object) cpayload = b64decode[ext_header_size:] else: cpayload = encoded_histogram hdr_payload = HdrPayload(8, compressed_payload=cpayload) return hdr_payload
Decode a wire histogram encoding into a read-only Hdr Payload instance Args: encoded_histogram a string containing the wire encoding of a histogram such as one returned from encode() Returns: an hdr_payload instance with all the decoded/uncompressed fields Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class HdrHistogramSettingsException: mismatch in the significant figures, lowest and highest trackable value zlib.error: in case of zlib decompression error
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/codec.py#L313-L356
null
class HdrHistogramEncoder(): '''An encoder class for histograms, only supports V1 encoding. ''' def __init__(self, histogram, b64_wrap=True, hdr_payload=None): '''Histogram encoder Args: histogram the histogram to encode/decode into b64_wrap determines if the base64 wrapper is enabled or not hdr_payload if None will create a new HdrPayload instance for this encoder, else will reuse the passed Hdrayload instance (useful after decoding one and to associate it to a new histogram) word_size counters size in bytes (2, 4 or 8) Exceptions: ValueError if the word_size value is unsupported ''' self.histogram = histogram if not hdr_payload: self.payload = HdrPayload(histogram.word_size, histogram.counts_len) payload = self.payload.payload # those values never change across encodings payload.normalizing_index_offset = 0 payload.conversion_ratio_bits = 1 payload.significant_figures = histogram.significant_figures payload.lowest_trackable_value = histogram.lowest_trackable_value payload.highest_trackable_value = histogram.highest_trackable_value else: self.payload = hdr_payload self.b64_wrap = b64_wrap self.header = ExternalHeader() self.header.cookie = get_compression_cookie() def get_counts(self): '''Retrieve the counts array that can be used to store live counters and that can be encoded with minimal copies using encode() ''' return self.payload.get_counts() def encode(self): '''Compress the associated encodable payload, prepend the header then encode with base64 if requested Returns: the b64 encoded wire encoding of the histogram (as a string) or the compressed payload (as a string, if b64 wrappinb is disabled) ''' # only compress the first non zero buckets # if histogram is empty we do not encode any counter if self.histogram.total_count: relevant_length = \ self.histogram.get_counts_array_index(self.histogram.max_value) + 1 else: relevant_length = 0 cpayload = self.payload.compress(relevant_length) if self.b64_wrap: self.header.length = len(cpayload) header_str = ctypes.string_at(addressof(self.header), ext_header_size) return base64.b64encode(header_str + cpayload) return cpayload @staticmethod def add(self, other_encoder): add_array(addressof(self.get_counts()), addressof(other_encoder.get_counts()), self.histogram.counts_len, self.histogram.word_size)
HdrHistogram/HdrHistogram_py
hdrh/histogram.py
HdrHistogram.record_value
python
def record_value(self, value, count=1): '''Record a new value into the histogram Args: value: the value to record (must be in the valid range) count: incremental count (defaults to 1) ''' if value < 0: return False counts_index = self._counts_index_for(value) if (counts_index < 0) or (self.counts_len <= counts_index): return False self.counts[counts_index] += count self.total_count += count self.min_value = min(self.min_value, value) self.max_value = max(self.max_value, value) return True
Record a new value into the histogram Args: value: the value to record (must be in the valid range) count: incremental count (defaults to 1)
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/histogram.py#L175-L191
[ "def _counts_index_for(self, value):\n bucket_index = self._get_bucket_index(value)\n sub_bucket_index = self._get_sub_bucket_index(value, bucket_index)\n return self._counts_index(bucket_index, sub_bucket_index)\n" ]
class HdrHistogram(): '''This class supports the recording and analyzing of sampled data value counts across a configurable integer value range with configurable value precision within the range. Value precision is expressed as the number of significant digits in the value recording, and provides control over value quantization behavior across the value range and the subsequent value resolution at any given level. For example, a Histogram could be configured to track the counts of observed integer values between 0 and 3,600,000,000 while maintaining a value precision of 3 significant digits across that range. Value quantization within the range will thus be no larger than 1/1,000th (or 0.1%) of any value. This example Histogram could be used to track and analyze the counts of observed response times ranging between 1 microsecond and 1 hour in magnitude, while maintaining a value resolution of 1 microsecond up to 1 millisecond, a resolution of 1 millisecond (or better) up to one second, and a resolution of 1 second (or better) up to 1,000 seconds. At it's maximum tracked value (1 hour), it would still maintain a resolution of 3.6 seconds (or better). ''' def __init__(self, lowest_trackable_value, highest_trackable_value, significant_figures, word_size=8, b64_wrap=True, hdr_payload=None): '''Create a new histogram with given arguments Params: lowest_trackable_value The lowest value that can be discerned (distinguished from 0) by the histogram. Must be a positive integer that is >= 1. May be internally rounded down to nearest power of 2. highest_trackable_value The highest value to be tracked by the histogram. Must be a positive integer that is >= (2 * lowest_trackable_value). significant_figures The number of significant decimal digits to which the histogram will maintain value resolution and separation. Must be a non-negative integer between 0 and 5. word_size size of counters in bytes, only 2, 4, 8-byte counters are supported (default is 8-byte or 64-bit counters) b64_wrap specifies if the encoding of this histogram should use base64 wrapping (only useful if you need to encode the histogram to save somewhere or send over the wire. By default base64 encoding is assumed hdr_payload only used for associating an existing payload created from decoding an encoded histograme Exceptions: ValueError if the word_size value is unsupported if significant_figures is invalid ''' if significant_figures < 1 or significant_figures > 5: raise ValueError('Invalid significant_figures') self.lowest_trackable_value = lowest_trackable_value self.highest_trackable_value = highest_trackable_value self.significant_figures = significant_figures self.unit_magnitude = int(math.floor(math.log(lowest_trackable_value) / math.log(2))) largest_value_single_unit_res = 2 * math.pow(10, significant_figures) subb_count_mag = int(math.ceil(math.log(largest_value_single_unit_res) / math.log(2))) self.sub_bucket_half_count_magnitude = subb_count_mag - 1 if subb_count_mag > 1 else 0 self.sub_bucket_count = int(math.pow(2, self.sub_bucket_half_count_magnitude + 1)) self.sub_bucket_half_count = self.sub_bucket_count // 2 self.sub_bucket_mask = (self.sub_bucket_count - 1) << self.unit_magnitude self.bucket_count = get_bucket_count(highest_trackable_value, self.sub_bucket_count, self.unit_magnitude) self.min_value = sys.maxsize self.max_value = 0 self.total_count = 0 self.counts_len = (self.bucket_count + 1) * (self.sub_bucket_count // 2) self.word_size = word_size if hdr_payload: payload = hdr_payload.payload self.int_to_double_conversion_ratio = payload.conversion_ratio_bits results = hdr_payload.init_counts(self.counts_len) if results['total']: self.set_internal_tacking_values(results['min_nonzero_index'], results['max_nonzero_index'], results['total']) else: self.int_to_double_conversion_ratio = 1.0 # to encode this histogram into a compressed/base64 format ready # to be exported self.b64_wrap = b64_wrap self.encoder = HdrHistogramEncoder(self, b64_wrap, hdr_payload) # the counters reside directly in the payload object # allocated by the encoder # so that compression for wire transfer can be done without copy self.counts = self.encoder.get_counts() self.start_time_stamp_msec = 0 self.end_time_stamp_msec = 0 def _clz(self, value): """calculate the leading zeros, equivalent to C __builtin_clzll() value in hex: value = 1 clz = 63 value = 2 clz = 62 value = 4 clz = 61 value = 1000 clz = 51 value = 1000000 clz = 39 """ return 63 - (len(bin(value)) - 3) def _get_bucket_index(self, value): # smallest power of 2 containing value pow2ceiling = 64 - self._clz(int(value) | self.sub_bucket_mask) return int(pow2ceiling - self.unit_magnitude - (self.sub_bucket_half_count_magnitude + 1)) def _get_sub_bucket_index(self, value, bucket_index): return int(value) >> (bucket_index + self.unit_magnitude) def _counts_index(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ): bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - subBucketHalfCount) + bucketBaseIndex return bucket_base_index + offset_in_bucket def _counts_index_for(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) return self._counts_index(bucket_index, sub_bucket_index) # pylint: disable=inconsistent-return-statements def record_corrected_value(self, value, expected_interval, count=1): '''Record a new value into the histogram and correct for coordinated omission if needed Args: value: the value to record (must be in the valid range) expected_interval: the expected interval between 2 value samples count: incremental count (defaults to 1) ''' while True: if not self.record_value(value, count): return False if value <= expected_interval or expected_interval <= 0: return True value -= expected_interval def get_count_at_index(self, index): if index >= self.counts_len: raise IndexError() # some decoded (read-only) histograms may have truncated # counts arrays, we return zero for any index that is passed the array if index >= self.encoder.payload.counts_len: return 0 return self.counts[index] def get_count_at_sub_bucket(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ) bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # (sub_bucket_index - subBucketHalfCount) + bucketBaseIndex counts_index = bucket_base_index + offset_in_bucket return self.counts[counts_index] def get_value_from_sub_bucket(self, bucket_index, sub_bucket_index): return sub_bucket_index << (bucket_index + self.unit_magnitude) def get_value_from_index(self, index): bucket_index = (index >> self.sub_bucket_half_count_magnitude) - 1 sub_bucket_index = (index & (self.sub_bucket_half_count - 1)) + \ self.sub_bucket_half_count if bucket_index < 0: sub_bucket_index -= self.sub_bucket_half_count bucket_index = 0 return self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) def get_lowest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) return lowest_equivalent_value def get_highest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 size_of_equivalent_value_range = 1 << (self.unit_magnitude + bucket_index) next_non_equivalent_value = lowest_equivalent_value + size_of_equivalent_value_range return next_non_equivalent_value - 1 def get_target_count_at_percentile(self, percentile): requested_percentile = min(percentile, 100.0) count_at_percentile = int(((requested_percentile * self.total_count / 100)) + 0.5) return max(count_at_percentile, 1) def get_value_at_percentile(self, percentile): '''Get the value for a given percentile Args: percentile: a float in [0.0..100.0] Returns: the value for the given percentile ''' count_at_percentile = self.get_target_count_at_percentile(percentile) total = 0 for index in range(self.counts_len): total += self.get_count_at_index(index) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: return self.get_highest_equivalent_value(value_at_index) return self.get_lowest_equivalent_value(value_at_index) return 0 def get_percentile_to_value_dict(self, percentile_list): '''A faster alternative to query values for a list of percentiles. Args: percentile_list: a list of percentiles in any order, dups will be ignored each element in the list must be a float value in [0.0 .. 100.0] Returns: a dict of percentile values indexed by the percentile ''' result = {} total = 0 percentile_list_index = 0 count_at_percentile = 0 # remove dups and sort percentile_list = list(set(percentile_list)) percentile_list.sort() for index in range(self.counts_len): total += self.get_count_at_index(index) while True: # recalculate target based on next requested percentile if not count_at_percentile: if percentile_list_index == len(percentile_list): return result percentile = percentile_list[percentile_list_index] percentile_list_index += 1 if percentile > 100: return result count_at_percentile = self.get_target_count_at_percentile(percentile) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: result[percentile] = self.get_highest_equivalent_value(value_at_index) else: result[percentile] = self.get_lowest_equivalent_value(value_at_index) count_at_percentile = 0 else: break return result def get_total_count(self): return self.total_count def get_count_at_value(self, value): counts_index = self._counts_index_for(value) return self.counts[counts_index] def values_are_equivalent(self, val1, val2): '''Check whether 2 values are equivalent (meaning they are in the same bucket/range) Returns: true if the 2 values are equivalent ''' return self.get_lowest_equivalent_value(val1) == self.get_lowest_equivalent_value(val2) def get_max_value(self): if self.max_value == 0: return 0 return self.get_highest_equivalent_value(self.max_value) def get_min_value(self): if self.counts[0] > 0 or self.total_count == 0: return 0 if sys.maxsize == self.min_value: return sys.maxsize return self.get_lowest_equivalent_value(self.min_value) def _hdr_size_of_equiv_value_range(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 return 1 << (self.unit_magnitude + bucket_index) def _hdr_median_equiv_value(self, value): return self.get_lowest_equivalent_value(value) + \ (self._hdr_size_of_equiv_value_range(value) >> 1) def get_mean_value(self): if not self.total_count: return 0.0 total = 0 itr = self.get_recorded_iterator() for item in itr: total += itr.count_at_this_value * self._hdr_median_equiv_value(item.value_iterated_to) return float(total) / self.total_count def get_stddev(self): if not self.total_count: return 0.0 mean = self.get_mean_value() geometric_dev_total = 0.0 for item in self.get_recorded_iterator(): dev = (self._hdr_median_equiv_value(item.value_iterated_to) * 1.0) - mean geometric_dev_total += (dev * dev) * item.count_added_in_this_iter_step return math.sqrt(geometric_dev_total / self.total_count) def reset(self): '''Reset the histogram to a pristine state ''' for index in range(self.counts_len): self.counts[index] = 0 self.total_count = 0 self.min_value = sys.maxsize self.max_value = 0 self.start_time_stamp_msec = sys.maxsize self.end_time_stamp_msec = 0 def __iter__(self): '''Returns the recorded iterator if iter(self) is called ''' return RecordedIterator(self) def get_all_values_iterator(self): return AllValuesIterator(self) def get_recorded_iterator(self): return RecordedIterator(self) def get_percentile_iterator(self, ticks_per_half_distance): return PercentileIterator(self, ticks_per_half_distance) def get_linear_iterator(self, value_units_per_bucket): return LinearIterator(self, value_units_per_bucket) def get_log_iterator(self, value_units_first_bucket, log_base): return LogIterator(self, value_units_first_bucket, log_base) def encode(self): '''Encode this histogram Return: a string containing the base64 encoded compressed histogram (V1 format) ''' return self.encoder.encode() def adjust_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) self.max_value = max(self.max_value, max_value) if min_non_zero_index >= 0: min_value = self.get_value_from_index(min_non_zero_index) self.min_value = min(self.min_value, min_value) self.total_count += total_added def set_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: self.max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) if min_non_zero_index >= 0: self.min_value = self.get_value_from_index(min_non_zero_index) self.total_count = total_added def get_counts_array_index(self, value): '''Return the index in the counts array for a given value ''' if value < 0: raise ValueError("Histogram recorded value cannot be negative.") bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) # Calculate the index for the first entry in the bucket: bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # The following is the equivalent of ((bucket_index + 1) * sub_bucket_half_count) # Calculate the offset in the bucket (can be negative for first bucket): offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - sub_bucket_half_count) + bucket_base_index return bucket_base_index + offset_in_bucket def get_start_time_stamp(self): return self.start_time_stamp_msec def set_start_time_stamp(self, time_stamp_msec): '''Set the start time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.start_time_stamp_msec = time_stamp_msec def get_end_time_stamp(self): return self.end_time_stamp_msec def set_end_time_stamp(self, time_stamp_msec): '''Set the end time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.end_time_stamp_msec = time_stamp_msec def add(self, other_hist): highest_recordable_value = \ self.get_highest_equivalent_value(self.get_value_from_index(self.counts_len - 1)) if highest_recordable_value < other_hist.get_max_value(): raise IndexError("The other histogram includes values that do not fit %d < %d" % (highest_recordable_value, other_hist.get_max_value())) if (self.bucket_count == other_hist.bucket_count) and \ (self.sub_bucket_count == other_hist.sub_bucket_count) and \ (self.unit_magnitude == other_hist.unit_magnitude) and \ (self.word_size == other_hist.word_size): # do an in-place addition of one array to another self.encoder.add(other_hist.encoder) self.total_count += other_hist.get_total_count() self.max_value = max(self.max_value, other_hist.get_max_value()) self.min_value = min(self.get_min_value(), other_hist.get_min_value()) else: # Arrays are not a direct match, so we can't just stream through and add them. # Instead, go through the array and add each non-zero value found at it's proper value: for index in range(other_hist.counts_len): other_count = other_hist.get_count_at_index(index) if other_count > 0: self.record_value(other_hist.get_value_from_index(index), other_count) self.start_time_stamp_msec = \ min(self.start_time_stamp_msec, other_hist.start_time_stamp_msec) self.end_time_stamp_msec = \ max(self.end_time_stamp_msec, other_hist.end_time_stamp_msec) def decode_and_add(self, encoded_histogram): '''Decode an encoded histogram and add it to this histogram Args: encoded_histogram (string) an encoded histogram following the V1 format, such as one returned by the encode() method Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' other_hist = HdrHistogram.decode(encoded_histogram, self.b64_wrap) self.add(other_hist) @staticmethod def decode(encoded_histogram, b64_wrap=True): '''Decode an encoded histogram and return a new histogram instance that has been initialized with the decoded content Return: a new histogram instance representing the decoded content Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' hdr_payload = HdrHistogramEncoder.decode(encoded_histogram, b64_wrap) payload = hdr_payload.payload histogram = HdrHistogram(payload.lowest_trackable_value, payload.highest_trackable_value, payload.significant_figures, hdr_payload=hdr_payload) return histogram def get_word_size(self): return self.word_size def output_percentile_distribution(self, out_file, output_value_unit_scaling_ratio, ticks_per_half_distance=5): out_file.write(b'%12s %14s %10s %14s\n\n' % (b'Value', b'Percentile', b'TotalCount', b'1/(1-Percentile)')) percentile_format = '%12.{}f %2.12f %10d %14.2f\n'.format(self.significant_figures) last_line_percentile_format = '%12.{}f %2.12f %10d\n'.format(self.significant_figures) for iter_value in self.get_percentile_iterator(ticks_per_half_distance): value = iter_value.value_iterated_to / output_value_unit_scaling_ratio percentile = iter_value.percentile_level_iterated_to / 100 total_count = iter_value.total_count_to_this_value if iter_value.percentile_level_iterated_to != 100: other = 1 / (1 - iter_value.percentile_level_iterated_to / 100) out_file.write(percentile_format.encode() % (value, percentile, total_count, other)) else: out_file.write(last_line_percentile_format.encode() % (value, percentile, total_count)) mean = self.get_mean_value() / output_value_unit_scaling_ratio stddev = self.get_stddev() out_file.write('#[Mean = %12.{0}f, StdDeviation = %12.{0}f]\n'. format(self.significant_figures).encode() % (mean, stddev)) max = self.get_max_value() / output_value_unit_scaling_ratio total = self.get_total_count() out_file.write('#[Max = %12.{0}f, TotalCount = %12.{0}f]\n'.format( self.significant_figures).encode() % (max, total)) out_file.write(b'#[Buckets = %12d, SubBuckets = %12d]\n' % ( self.bucket_count, self.sub_bucket_count))
HdrHistogram/HdrHistogram_py
hdrh/histogram.py
HdrHistogram.record_corrected_value
python
def record_corrected_value(self, value, expected_interval, count=1): '''Record a new value into the histogram and correct for coordinated omission if needed Args: value: the value to record (must be in the valid range) expected_interval: the expected interval between 2 value samples count: incremental count (defaults to 1) ''' while True: if not self.record_value(value, count): return False if value <= expected_interval or expected_interval <= 0: return True value -= expected_interval
Record a new value into the histogram and correct for coordinated omission if needed Args: value: the value to record (must be in the valid range) expected_interval: the expected interval between 2 value samples count: incremental count (defaults to 1)
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/histogram.py#L194-L208
[ "def record_value(self, value, count=1):\n '''Record a new value into the histogram\n\n Args:\n value: the value to record (must be in the valid range)\n count: incremental count (defaults to 1)\n '''\n if value < 0:\n return False\n counts_index = self._counts_index_for(value)\n if (counts_index < 0) or (self.counts_len <= counts_index):\n return False\n self.counts[counts_index] += count\n self.total_count += count\n self.min_value = min(self.min_value, value)\n self.max_value = max(self.max_value, value)\n return True\n" ]
class HdrHistogram(): '''This class supports the recording and analyzing of sampled data value counts across a configurable integer value range with configurable value precision within the range. Value precision is expressed as the number of significant digits in the value recording, and provides control over value quantization behavior across the value range and the subsequent value resolution at any given level. For example, a Histogram could be configured to track the counts of observed integer values between 0 and 3,600,000,000 while maintaining a value precision of 3 significant digits across that range. Value quantization within the range will thus be no larger than 1/1,000th (or 0.1%) of any value. This example Histogram could be used to track and analyze the counts of observed response times ranging between 1 microsecond and 1 hour in magnitude, while maintaining a value resolution of 1 microsecond up to 1 millisecond, a resolution of 1 millisecond (or better) up to one second, and a resolution of 1 second (or better) up to 1,000 seconds. At it's maximum tracked value (1 hour), it would still maintain a resolution of 3.6 seconds (or better). ''' def __init__(self, lowest_trackable_value, highest_trackable_value, significant_figures, word_size=8, b64_wrap=True, hdr_payload=None): '''Create a new histogram with given arguments Params: lowest_trackable_value The lowest value that can be discerned (distinguished from 0) by the histogram. Must be a positive integer that is >= 1. May be internally rounded down to nearest power of 2. highest_trackable_value The highest value to be tracked by the histogram. Must be a positive integer that is >= (2 * lowest_trackable_value). significant_figures The number of significant decimal digits to which the histogram will maintain value resolution and separation. Must be a non-negative integer between 0 and 5. word_size size of counters in bytes, only 2, 4, 8-byte counters are supported (default is 8-byte or 64-bit counters) b64_wrap specifies if the encoding of this histogram should use base64 wrapping (only useful if you need to encode the histogram to save somewhere or send over the wire. By default base64 encoding is assumed hdr_payload only used for associating an existing payload created from decoding an encoded histograme Exceptions: ValueError if the word_size value is unsupported if significant_figures is invalid ''' if significant_figures < 1 or significant_figures > 5: raise ValueError('Invalid significant_figures') self.lowest_trackable_value = lowest_trackable_value self.highest_trackable_value = highest_trackable_value self.significant_figures = significant_figures self.unit_magnitude = int(math.floor(math.log(lowest_trackable_value) / math.log(2))) largest_value_single_unit_res = 2 * math.pow(10, significant_figures) subb_count_mag = int(math.ceil(math.log(largest_value_single_unit_res) / math.log(2))) self.sub_bucket_half_count_magnitude = subb_count_mag - 1 if subb_count_mag > 1 else 0 self.sub_bucket_count = int(math.pow(2, self.sub_bucket_half_count_magnitude + 1)) self.sub_bucket_half_count = self.sub_bucket_count // 2 self.sub_bucket_mask = (self.sub_bucket_count - 1) << self.unit_magnitude self.bucket_count = get_bucket_count(highest_trackable_value, self.sub_bucket_count, self.unit_magnitude) self.min_value = sys.maxsize self.max_value = 0 self.total_count = 0 self.counts_len = (self.bucket_count + 1) * (self.sub_bucket_count // 2) self.word_size = word_size if hdr_payload: payload = hdr_payload.payload self.int_to_double_conversion_ratio = payload.conversion_ratio_bits results = hdr_payload.init_counts(self.counts_len) if results['total']: self.set_internal_tacking_values(results['min_nonzero_index'], results['max_nonzero_index'], results['total']) else: self.int_to_double_conversion_ratio = 1.0 # to encode this histogram into a compressed/base64 format ready # to be exported self.b64_wrap = b64_wrap self.encoder = HdrHistogramEncoder(self, b64_wrap, hdr_payload) # the counters reside directly in the payload object # allocated by the encoder # so that compression for wire transfer can be done without copy self.counts = self.encoder.get_counts() self.start_time_stamp_msec = 0 self.end_time_stamp_msec = 0 def _clz(self, value): """calculate the leading zeros, equivalent to C __builtin_clzll() value in hex: value = 1 clz = 63 value = 2 clz = 62 value = 4 clz = 61 value = 1000 clz = 51 value = 1000000 clz = 39 """ return 63 - (len(bin(value)) - 3) def _get_bucket_index(self, value): # smallest power of 2 containing value pow2ceiling = 64 - self._clz(int(value) | self.sub_bucket_mask) return int(pow2ceiling - self.unit_magnitude - (self.sub_bucket_half_count_magnitude + 1)) def _get_sub_bucket_index(self, value, bucket_index): return int(value) >> (bucket_index + self.unit_magnitude) def _counts_index(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ): bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - subBucketHalfCount) + bucketBaseIndex return bucket_base_index + offset_in_bucket def _counts_index_for(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) return self._counts_index(bucket_index, sub_bucket_index) def record_value(self, value, count=1): '''Record a new value into the histogram Args: value: the value to record (must be in the valid range) count: incremental count (defaults to 1) ''' if value < 0: return False counts_index = self._counts_index_for(value) if (counts_index < 0) or (self.counts_len <= counts_index): return False self.counts[counts_index] += count self.total_count += count self.min_value = min(self.min_value, value) self.max_value = max(self.max_value, value) return True # pylint: disable=inconsistent-return-statements def get_count_at_index(self, index): if index >= self.counts_len: raise IndexError() # some decoded (read-only) histograms may have truncated # counts arrays, we return zero for any index that is passed the array if index >= self.encoder.payload.counts_len: return 0 return self.counts[index] def get_count_at_sub_bucket(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ) bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # (sub_bucket_index - subBucketHalfCount) + bucketBaseIndex counts_index = bucket_base_index + offset_in_bucket return self.counts[counts_index] def get_value_from_sub_bucket(self, bucket_index, sub_bucket_index): return sub_bucket_index << (bucket_index + self.unit_magnitude) def get_value_from_index(self, index): bucket_index = (index >> self.sub_bucket_half_count_magnitude) - 1 sub_bucket_index = (index & (self.sub_bucket_half_count - 1)) + \ self.sub_bucket_half_count if bucket_index < 0: sub_bucket_index -= self.sub_bucket_half_count bucket_index = 0 return self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) def get_lowest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) return lowest_equivalent_value def get_highest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 size_of_equivalent_value_range = 1 << (self.unit_magnitude + bucket_index) next_non_equivalent_value = lowest_equivalent_value + size_of_equivalent_value_range return next_non_equivalent_value - 1 def get_target_count_at_percentile(self, percentile): requested_percentile = min(percentile, 100.0) count_at_percentile = int(((requested_percentile * self.total_count / 100)) + 0.5) return max(count_at_percentile, 1) def get_value_at_percentile(self, percentile): '''Get the value for a given percentile Args: percentile: a float in [0.0..100.0] Returns: the value for the given percentile ''' count_at_percentile = self.get_target_count_at_percentile(percentile) total = 0 for index in range(self.counts_len): total += self.get_count_at_index(index) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: return self.get_highest_equivalent_value(value_at_index) return self.get_lowest_equivalent_value(value_at_index) return 0 def get_percentile_to_value_dict(self, percentile_list): '''A faster alternative to query values for a list of percentiles. Args: percentile_list: a list of percentiles in any order, dups will be ignored each element in the list must be a float value in [0.0 .. 100.0] Returns: a dict of percentile values indexed by the percentile ''' result = {} total = 0 percentile_list_index = 0 count_at_percentile = 0 # remove dups and sort percentile_list = list(set(percentile_list)) percentile_list.sort() for index in range(self.counts_len): total += self.get_count_at_index(index) while True: # recalculate target based on next requested percentile if not count_at_percentile: if percentile_list_index == len(percentile_list): return result percentile = percentile_list[percentile_list_index] percentile_list_index += 1 if percentile > 100: return result count_at_percentile = self.get_target_count_at_percentile(percentile) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: result[percentile] = self.get_highest_equivalent_value(value_at_index) else: result[percentile] = self.get_lowest_equivalent_value(value_at_index) count_at_percentile = 0 else: break return result def get_total_count(self): return self.total_count def get_count_at_value(self, value): counts_index = self._counts_index_for(value) return self.counts[counts_index] def values_are_equivalent(self, val1, val2): '''Check whether 2 values are equivalent (meaning they are in the same bucket/range) Returns: true if the 2 values are equivalent ''' return self.get_lowest_equivalent_value(val1) == self.get_lowest_equivalent_value(val2) def get_max_value(self): if self.max_value == 0: return 0 return self.get_highest_equivalent_value(self.max_value) def get_min_value(self): if self.counts[0] > 0 or self.total_count == 0: return 0 if sys.maxsize == self.min_value: return sys.maxsize return self.get_lowest_equivalent_value(self.min_value) def _hdr_size_of_equiv_value_range(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 return 1 << (self.unit_magnitude + bucket_index) def _hdr_median_equiv_value(self, value): return self.get_lowest_equivalent_value(value) + \ (self._hdr_size_of_equiv_value_range(value) >> 1) def get_mean_value(self): if not self.total_count: return 0.0 total = 0 itr = self.get_recorded_iterator() for item in itr: total += itr.count_at_this_value * self._hdr_median_equiv_value(item.value_iterated_to) return float(total) / self.total_count def get_stddev(self): if not self.total_count: return 0.0 mean = self.get_mean_value() geometric_dev_total = 0.0 for item in self.get_recorded_iterator(): dev = (self._hdr_median_equiv_value(item.value_iterated_to) * 1.0) - mean geometric_dev_total += (dev * dev) * item.count_added_in_this_iter_step return math.sqrt(geometric_dev_total / self.total_count) def reset(self): '''Reset the histogram to a pristine state ''' for index in range(self.counts_len): self.counts[index] = 0 self.total_count = 0 self.min_value = sys.maxsize self.max_value = 0 self.start_time_stamp_msec = sys.maxsize self.end_time_stamp_msec = 0 def __iter__(self): '''Returns the recorded iterator if iter(self) is called ''' return RecordedIterator(self) def get_all_values_iterator(self): return AllValuesIterator(self) def get_recorded_iterator(self): return RecordedIterator(self) def get_percentile_iterator(self, ticks_per_half_distance): return PercentileIterator(self, ticks_per_half_distance) def get_linear_iterator(self, value_units_per_bucket): return LinearIterator(self, value_units_per_bucket) def get_log_iterator(self, value_units_first_bucket, log_base): return LogIterator(self, value_units_first_bucket, log_base) def encode(self): '''Encode this histogram Return: a string containing the base64 encoded compressed histogram (V1 format) ''' return self.encoder.encode() def adjust_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) self.max_value = max(self.max_value, max_value) if min_non_zero_index >= 0: min_value = self.get_value_from_index(min_non_zero_index) self.min_value = min(self.min_value, min_value) self.total_count += total_added def set_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: self.max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) if min_non_zero_index >= 0: self.min_value = self.get_value_from_index(min_non_zero_index) self.total_count = total_added def get_counts_array_index(self, value): '''Return the index in the counts array for a given value ''' if value < 0: raise ValueError("Histogram recorded value cannot be negative.") bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) # Calculate the index for the first entry in the bucket: bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # The following is the equivalent of ((bucket_index + 1) * sub_bucket_half_count) # Calculate the offset in the bucket (can be negative for first bucket): offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - sub_bucket_half_count) + bucket_base_index return bucket_base_index + offset_in_bucket def get_start_time_stamp(self): return self.start_time_stamp_msec def set_start_time_stamp(self, time_stamp_msec): '''Set the start time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.start_time_stamp_msec = time_stamp_msec def get_end_time_stamp(self): return self.end_time_stamp_msec def set_end_time_stamp(self, time_stamp_msec): '''Set the end time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.end_time_stamp_msec = time_stamp_msec def add(self, other_hist): highest_recordable_value = \ self.get_highest_equivalent_value(self.get_value_from_index(self.counts_len - 1)) if highest_recordable_value < other_hist.get_max_value(): raise IndexError("The other histogram includes values that do not fit %d < %d" % (highest_recordable_value, other_hist.get_max_value())) if (self.bucket_count == other_hist.bucket_count) and \ (self.sub_bucket_count == other_hist.sub_bucket_count) and \ (self.unit_magnitude == other_hist.unit_magnitude) and \ (self.word_size == other_hist.word_size): # do an in-place addition of one array to another self.encoder.add(other_hist.encoder) self.total_count += other_hist.get_total_count() self.max_value = max(self.max_value, other_hist.get_max_value()) self.min_value = min(self.get_min_value(), other_hist.get_min_value()) else: # Arrays are not a direct match, so we can't just stream through and add them. # Instead, go through the array and add each non-zero value found at it's proper value: for index in range(other_hist.counts_len): other_count = other_hist.get_count_at_index(index) if other_count > 0: self.record_value(other_hist.get_value_from_index(index), other_count) self.start_time_stamp_msec = \ min(self.start_time_stamp_msec, other_hist.start_time_stamp_msec) self.end_time_stamp_msec = \ max(self.end_time_stamp_msec, other_hist.end_time_stamp_msec) def decode_and_add(self, encoded_histogram): '''Decode an encoded histogram and add it to this histogram Args: encoded_histogram (string) an encoded histogram following the V1 format, such as one returned by the encode() method Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' other_hist = HdrHistogram.decode(encoded_histogram, self.b64_wrap) self.add(other_hist) @staticmethod def decode(encoded_histogram, b64_wrap=True): '''Decode an encoded histogram and return a new histogram instance that has been initialized with the decoded content Return: a new histogram instance representing the decoded content Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' hdr_payload = HdrHistogramEncoder.decode(encoded_histogram, b64_wrap) payload = hdr_payload.payload histogram = HdrHistogram(payload.lowest_trackable_value, payload.highest_trackable_value, payload.significant_figures, hdr_payload=hdr_payload) return histogram def get_word_size(self): return self.word_size def output_percentile_distribution(self, out_file, output_value_unit_scaling_ratio, ticks_per_half_distance=5): out_file.write(b'%12s %14s %10s %14s\n\n' % (b'Value', b'Percentile', b'TotalCount', b'1/(1-Percentile)')) percentile_format = '%12.{}f %2.12f %10d %14.2f\n'.format(self.significant_figures) last_line_percentile_format = '%12.{}f %2.12f %10d\n'.format(self.significant_figures) for iter_value in self.get_percentile_iterator(ticks_per_half_distance): value = iter_value.value_iterated_to / output_value_unit_scaling_ratio percentile = iter_value.percentile_level_iterated_to / 100 total_count = iter_value.total_count_to_this_value if iter_value.percentile_level_iterated_to != 100: other = 1 / (1 - iter_value.percentile_level_iterated_to / 100) out_file.write(percentile_format.encode() % (value, percentile, total_count, other)) else: out_file.write(last_line_percentile_format.encode() % (value, percentile, total_count)) mean = self.get_mean_value() / output_value_unit_scaling_ratio stddev = self.get_stddev() out_file.write('#[Mean = %12.{0}f, StdDeviation = %12.{0}f]\n'. format(self.significant_figures).encode() % (mean, stddev)) max = self.get_max_value() / output_value_unit_scaling_ratio total = self.get_total_count() out_file.write('#[Max = %12.{0}f, TotalCount = %12.{0}f]\n'.format( self.significant_figures).encode() % (max, total)) out_file.write(b'#[Buckets = %12d, SubBuckets = %12d]\n' % ( self.bucket_count, self.sub_bucket_count))
HdrHistogram/HdrHistogram_py
hdrh/histogram.py
HdrHistogram.get_value_at_percentile
python
def get_value_at_percentile(self, percentile): '''Get the value for a given percentile Args: percentile: a float in [0.0..100.0] Returns: the value for the given percentile ''' count_at_percentile = self.get_target_count_at_percentile(percentile) total = 0 for index in range(self.counts_len): total += self.get_count_at_index(index) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: return self.get_highest_equivalent_value(value_at_index) return self.get_lowest_equivalent_value(value_at_index) return 0
Get the value for a given percentile Args: percentile: a float in [0.0..100.0] Returns: the value for the given percentile
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/histogram.py#L268-L285
[ "def get_count_at_index(self, index):\n if index >= self.counts_len:\n raise IndexError()\n # some decoded (read-only) histograms may have truncated\n # counts arrays, we return zero for any index that is passed the array\n if index >= self.encoder.payload.counts_len:\n return 0\n return self.counts[index]\n", "def get_value_from_index(self, index):\n bucket_index = (index >> self.sub_bucket_half_count_magnitude) - 1\n sub_bucket_index = (index & (self.sub_bucket_half_count - 1)) + \\\n self.sub_bucket_half_count\n if bucket_index < 0:\n sub_bucket_index -= self.sub_bucket_half_count\n bucket_index = 0\n return self.get_value_from_sub_bucket(bucket_index, sub_bucket_index)\n", "def get_lowest_equivalent_value(self, value):\n bucket_index = self._get_bucket_index(value)\n sub_bucket_index = self._get_sub_bucket_index(value, bucket_index)\n\n lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index,\n sub_bucket_index)\n return lowest_equivalent_value\n", "def get_highest_equivalent_value(self, value):\n bucket_index = self._get_bucket_index(value)\n sub_bucket_index = self._get_sub_bucket_index(value, bucket_index)\n\n lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index,\n sub_bucket_index)\n if sub_bucket_index >= self.sub_bucket_count:\n bucket_index += 1\n size_of_equivalent_value_range = 1 << (self.unit_magnitude + bucket_index)\n next_non_equivalent_value = lowest_equivalent_value + size_of_equivalent_value_range\n\n return next_non_equivalent_value - 1\n", "def get_target_count_at_percentile(self, percentile):\n requested_percentile = min(percentile, 100.0)\n count_at_percentile = int(((requested_percentile * self.total_count / 100)) + 0.5)\n return max(count_at_percentile, 1)\n" ]
class HdrHistogram(): '''This class supports the recording and analyzing of sampled data value counts across a configurable integer value range with configurable value precision within the range. Value precision is expressed as the number of significant digits in the value recording, and provides control over value quantization behavior across the value range and the subsequent value resolution at any given level. For example, a Histogram could be configured to track the counts of observed integer values between 0 and 3,600,000,000 while maintaining a value precision of 3 significant digits across that range. Value quantization within the range will thus be no larger than 1/1,000th (or 0.1%) of any value. This example Histogram could be used to track and analyze the counts of observed response times ranging between 1 microsecond and 1 hour in magnitude, while maintaining a value resolution of 1 microsecond up to 1 millisecond, a resolution of 1 millisecond (or better) up to one second, and a resolution of 1 second (or better) up to 1,000 seconds. At it's maximum tracked value (1 hour), it would still maintain a resolution of 3.6 seconds (or better). ''' def __init__(self, lowest_trackable_value, highest_trackable_value, significant_figures, word_size=8, b64_wrap=True, hdr_payload=None): '''Create a new histogram with given arguments Params: lowest_trackable_value The lowest value that can be discerned (distinguished from 0) by the histogram. Must be a positive integer that is >= 1. May be internally rounded down to nearest power of 2. highest_trackable_value The highest value to be tracked by the histogram. Must be a positive integer that is >= (2 * lowest_trackable_value). significant_figures The number of significant decimal digits to which the histogram will maintain value resolution and separation. Must be a non-negative integer between 0 and 5. word_size size of counters in bytes, only 2, 4, 8-byte counters are supported (default is 8-byte or 64-bit counters) b64_wrap specifies if the encoding of this histogram should use base64 wrapping (only useful if you need to encode the histogram to save somewhere or send over the wire. By default base64 encoding is assumed hdr_payload only used for associating an existing payload created from decoding an encoded histograme Exceptions: ValueError if the word_size value is unsupported if significant_figures is invalid ''' if significant_figures < 1 or significant_figures > 5: raise ValueError('Invalid significant_figures') self.lowest_trackable_value = lowest_trackable_value self.highest_trackable_value = highest_trackable_value self.significant_figures = significant_figures self.unit_magnitude = int(math.floor(math.log(lowest_trackable_value) / math.log(2))) largest_value_single_unit_res = 2 * math.pow(10, significant_figures) subb_count_mag = int(math.ceil(math.log(largest_value_single_unit_res) / math.log(2))) self.sub_bucket_half_count_magnitude = subb_count_mag - 1 if subb_count_mag > 1 else 0 self.sub_bucket_count = int(math.pow(2, self.sub_bucket_half_count_magnitude + 1)) self.sub_bucket_half_count = self.sub_bucket_count // 2 self.sub_bucket_mask = (self.sub_bucket_count - 1) << self.unit_magnitude self.bucket_count = get_bucket_count(highest_trackable_value, self.sub_bucket_count, self.unit_magnitude) self.min_value = sys.maxsize self.max_value = 0 self.total_count = 0 self.counts_len = (self.bucket_count + 1) * (self.sub_bucket_count // 2) self.word_size = word_size if hdr_payload: payload = hdr_payload.payload self.int_to_double_conversion_ratio = payload.conversion_ratio_bits results = hdr_payload.init_counts(self.counts_len) if results['total']: self.set_internal_tacking_values(results['min_nonzero_index'], results['max_nonzero_index'], results['total']) else: self.int_to_double_conversion_ratio = 1.0 # to encode this histogram into a compressed/base64 format ready # to be exported self.b64_wrap = b64_wrap self.encoder = HdrHistogramEncoder(self, b64_wrap, hdr_payload) # the counters reside directly in the payload object # allocated by the encoder # so that compression for wire transfer can be done without copy self.counts = self.encoder.get_counts() self.start_time_stamp_msec = 0 self.end_time_stamp_msec = 0 def _clz(self, value): """calculate the leading zeros, equivalent to C __builtin_clzll() value in hex: value = 1 clz = 63 value = 2 clz = 62 value = 4 clz = 61 value = 1000 clz = 51 value = 1000000 clz = 39 """ return 63 - (len(bin(value)) - 3) def _get_bucket_index(self, value): # smallest power of 2 containing value pow2ceiling = 64 - self._clz(int(value) | self.sub_bucket_mask) return int(pow2ceiling - self.unit_magnitude - (self.sub_bucket_half_count_magnitude + 1)) def _get_sub_bucket_index(self, value, bucket_index): return int(value) >> (bucket_index + self.unit_magnitude) def _counts_index(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ): bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - subBucketHalfCount) + bucketBaseIndex return bucket_base_index + offset_in_bucket def _counts_index_for(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) return self._counts_index(bucket_index, sub_bucket_index) def record_value(self, value, count=1): '''Record a new value into the histogram Args: value: the value to record (must be in the valid range) count: incremental count (defaults to 1) ''' if value < 0: return False counts_index = self._counts_index_for(value) if (counts_index < 0) or (self.counts_len <= counts_index): return False self.counts[counts_index] += count self.total_count += count self.min_value = min(self.min_value, value) self.max_value = max(self.max_value, value) return True # pylint: disable=inconsistent-return-statements def record_corrected_value(self, value, expected_interval, count=1): '''Record a new value into the histogram and correct for coordinated omission if needed Args: value: the value to record (must be in the valid range) expected_interval: the expected interval between 2 value samples count: incremental count (defaults to 1) ''' while True: if not self.record_value(value, count): return False if value <= expected_interval or expected_interval <= 0: return True value -= expected_interval def get_count_at_index(self, index): if index >= self.counts_len: raise IndexError() # some decoded (read-only) histograms may have truncated # counts arrays, we return zero for any index that is passed the array if index >= self.encoder.payload.counts_len: return 0 return self.counts[index] def get_count_at_sub_bucket(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ) bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # (sub_bucket_index - subBucketHalfCount) + bucketBaseIndex counts_index = bucket_base_index + offset_in_bucket return self.counts[counts_index] def get_value_from_sub_bucket(self, bucket_index, sub_bucket_index): return sub_bucket_index << (bucket_index + self.unit_magnitude) def get_value_from_index(self, index): bucket_index = (index >> self.sub_bucket_half_count_magnitude) - 1 sub_bucket_index = (index & (self.sub_bucket_half_count - 1)) + \ self.sub_bucket_half_count if bucket_index < 0: sub_bucket_index -= self.sub_bucket_half_count bucket_index = 0 return self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) def get_lowest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) return lowest_equivalent_value def get_highest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 size_of_equivalent_value_range = 1 << (self.unit_magnitude + bucket_index) next_non_equivalent_value = lowest_equivalent_value + size_of_equivalent_value_range return next_non_equivalent_value - 1 def get_target_count_at_percentile(self, percentile): requested_percentile = min(percentile, 100.0) count_at_percentile = int(((requested_percentile * self.total_count / 100)) + 0.5) return max(count_at_percentile, 1) def get_percentile_to_value_dict(self, percentile_list): '''A faster alternative to query values for a list of percentiles. Args: percentile_list: a list of percentiles in any order, dups will be ignored each element in the list must be a float value in [0.0 .. 100.0] Returns: a dict of percentile values indexed by the percentile ''' result = {} total = 0 percentile_list_index = 0 count_at_percentile = 0 # remove dups and sort percentile_list = list(set(percentile_list)) percentile_list.sort() for index in range(self.counts_len): total += self.get_count_at_index(index) while True: # recalculate target based on next requested percentile if not count_at_percentile: if percentile_list_index == len(percentile_list): return result percentile = percentile_list[percentile_list_index] percentile_list_index += 1 if percentile > 100: return result count_at_percentile = self.get_target_count_at_percentile(percentile) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: result[percentile] = self.get_highest_equivalent_value(value_at_index) else: result[percentile] = self.get_lowest_equivalent_value(value_at_index) count_at_percentile = 0 else: break return result def get_total_count(self): return self.total_count def get_count_at_value(self, value): counts_index = self._counts_index_for(value) return self.counts[counts_index] def values_are_equivalent(self, val1, val2): '''Check whether 2 values are equivalent (meaning they are in the same bucket/range) Returns: true if the 2 values are equivalent ''' return self.get_lowest_equivalent_value(val1) == self.get_lowest_equivalent_value(val2) def get_max_value(self): if self.max_value == 0: return 0 return self.get_highest_equivalent_value(self.max_value) def get_min_value(self): if self.counts[0] > 0 or self.total_count == 0: return 0 if sys.maxsize == self.min_value: return sys.maxsize return self.get_lowest_equivalent_value(self.min_value) def _hdr_size_of_equiv_value_range(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 return 1 << (self.unit_magnitude + bucket_index) def _hdr_median_equiv_value(self, value): return self.get_lowest_equivalent_value(value) + \ (self._hdr_size_of_equiv_value_range(value) >> 1) def get_mean_value(self): if not self.total_count: return 0.0 total = 0 itr = self.get_recorded_iterator() for item in itr: total += itr.count_at_this_value * self._hdr_median_equiv_value(item.value_iterated_to) return float(total) / self.total_count def get_stddev(self): if not self.total_count: return 0.0 mean = self.get_mean_value() geometric_dev_total = 0.0 for item in self.get_recorded_iterator(): dev = (self._hdr_median_equiv_value(item.value_iterated_to) * 1.0) - mean geometric_dev_total += (dev * dev) * item.count_added_in_this_iter_step return math.sqrt(geometric_dev_total / self.total_count) def reset(self): '''Reset the histogram to a pristine state ''' for index in range(self.counts_len): self.counts[index] = 0 self.total_count = 0 self.min_value = sys.maxsize self.max_value = 0 self.start_time_stamp_msec = sys.maxsize self.end_time_stamp_msec = 0 def __iter__(self): '''Returns the recorded iterator if iter(self) is called ''' return RecordedIterator(self) def get_all_values_iterator(self): return AllValuesIterator(self) def get_recorded_iterator(self): return RecordedIterator(self) def get_percentile_iterator(self, ticks_per_half_distance): return PercentileIterator(self, ticks_per_half_distance) def get_linear_iterator(self, value_units_per_bucket): return LinearIterator(self, value_units_per_bucket) def get_log_iterator(self, value_units_first_bucket, log_base): return LogIterator(self, value_units_first_bucket, log_base) def encode(self): '''Encode this histogram Return: a string containing the base64 encoded compressed histogram (V1 format) ''' return self.encoder.encode() def adjust_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) self.max_value = max(self.max_value, max_value) if min_non_zero_index >= 0: min_value = self.get_value_from_index(min_non_zero_index) self.min_value = min(self.min_value, min_value) self.total_count += total_added def set_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: self.max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) if min_non_zero_index >= 0: self.min_value = self.get_value_from_index(min_non_zero_index) self.total_count = total_added def get_counts_array_index(self, value): '''Return the index in the counts array for a given value ''' if value < 0: raise ValueError("Histogram recorded value cannot be negative.") bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) # Calculate the index for the first entry in the bucket: bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # The following is the equivalent of ((bucket_index + 1) * sub_bucket_half_count) # Calculate the offset in the bucket (can be negative for first bucket): offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - sub_bucket_half_count) + bucket_base_index return bucket_base_index + offset_in_bucket def get_start_time_stamp(self): return self.start_time_stamp_msec def set_start_time_stamp(self, time_stamp_msec): '''Set the start time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.start_time_stamp_msec = time_stamp_msec def get_end_time_stamp(self): return self.end_time_stamp_msec def set_end_time_stamp(self, time_stamp_msec): '''Set the end time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.end_time_stamp_msec = time_stamp_msec def add(self, other_hist): highest_recordable_value = \ self.get_highest_equivalent_value(self.get_value_from_index(self.counts_len - 1)) if highest_recordable_value < other_hist.get_max_value(): raise IndexError("The other histogram includes values that do not fit %d < %d" % (highest_recordable_value, other_hist.get_max_value())) if (self.bucket_count == other_hist.bucket_count) and \ (self.sub_bucket_count == other_hist.sub_bucket_count) and \ (self.unit_magnitude == other_hist.unit_magnitude) and \ (self.word_size == other_hist.word_size): # do an in-place addition of one array to another self.encoder.add(other_hist.encoder) self.total_count += other_hist.get_total_count() self.max_value = max(self.max_value, other_hist.get_max_value()) self.min_value = min(self.get_min_value(), other_hist.get_min_value()) else: # Arrays are not a direct match, so we can't just stream through and add them. # Instead, go through the array and add each non-zero value found at it's proper value: for index in range(other_hist.counts_len): other_count = other_hist.get_count_at_index(index) if other_count > 0: self.record_value(other_hist.get_value_from_index(index), other_count) self.start_time_stamp_msec = \ min(self.start_time_stamp_msec, other_hist.start_time_stamp_msec) self.end_time_stamp_msec = \ max(self.end_time_stamp_msec, other_hist.end_time_stamp_msec) def decode_and_add(self, encoded_histogram): '''Decode an encoded histogram and add it to this histogram Args: encoded_histogram (string) an encoded histogram following the V1 format, such as one returned by the encode() method Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' other_hist = HdrHistogram.decode(encoded_histogram, self.b64_wrap) self.add(other_hist) @staticmethod def decode(encoded_histogram, b64_wrap=True): '''Decode an encoded histogram and return a new histogram instance that has been initialized with the decoded content Return: a new histogram instance representing the decoded content Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' hdr_payload = HdrHistogramEncoder.decode(encoded_histogram, b64_wrap) payload = hdr_payload.payload histogram = HdrHistogram(payload.lowest_trackable_value, payload.highest_trackable_value, payload.significant_figures, hdr_payload=hdr_payload) return histogram def get_word_size(self): return self.word_size def output_percentile_distribution(self, out_file, output_value_unit_scaling_ratio, ticks_per_half_distance=5): out_file.write(b'%12s %14s %10s %14s\n\n' % (b'Value', b'Percentile', b'TotalCount', b'1/(1-Percentile)')) percentile_format = '%12.{}f %2.12f %10d %14.2f\n'.format(self.significant_figures) last_line_percentile_format = '%12.{}f %2.12f %10d\n'.format(self.significant_figures) for iter_value in self.get_percentile_iterator(ticks_per_half_distance): value = iter_value.value_iterated_to / output_value_unit_scaling_ratio percentile = iter_value.percentile_level_iterated_to / 100 total_count = iter_value.total_count_to_this_value if iter_value.percentile_level_iterated_to != 100: other = 1 / (1 - iter_value.percentile_level_iterated_to / 100) out_file.write(percentile_format.encode() % (value, percentile, total_count, other)) else: out_file.write(last_line_percentile_format.encode() % (value, percentile, total_count)) mean = self.get_mean_value() / output_value_unit_scaling_ratio stddev = self.get_stddev() out_file.write('#[Mean = %12.{0}f, StdDeviation = %12.{0}f]\n'. format(self.significant_figures).encode() % (mean, stddev)) max = self.get_max_value() / output_value_unit_scaling_ratio total = self.get_total_count() out_file.write('#[Max = %12.{0}f, TotalCount = %12.{0}f]\n'.format( self.significant_figures).encode() % (max, total)) out_file.write(b'#[Buckets = %12d, SubBuckets = %12d]\n' % ( self.bucket_count, self.sub_bucket_count))
HdrHistogram/HdrHistogram_py
hdrh/histogram.py
HdrHistogram.get_percentile_to_value_dict
python
def get_percentile_to_value_dict(self, percentile_list): '''A faster alternative to query values for a list of percentiles. Args: percentile_list: a list of percentiles in any order, dups will be ignored each element in the list must be a float value in [0.0 .. 100.0] Returns: a dict of percentile values indexed by the percentile ''' result = {} total = 0 percentile_list_index = 0 count_at_percentile = 0 # remove dups and sort percentile_list = list(set(percentile_list)) percentile_list.sort() for index in range(self.counts_len): total += self.get_count_at_index(index) while True: # recalculate target based on next requested percentile if not count_at_percentile: if percentile_list_index == len(percentile_list): return result percentile = percentile_list[percentile_list_index] percentile_list_index += 1 if percentile > 100: return result count_at_percentile = self.get_target_count_at_percentile(percentile) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: result[percentile] = self.get_highest_equivalent_value(value_at_index) else: result[percentile] = self.get_lowest_equivalent_value(value_at_index) count_at_percentile = 0 else: break return result
A faster alternative to query values for a list of percentiles. Args: percentile_list: a list of percentiles in any order, dups will be ignored each element in the list must be a float value in [0.0 .. 100.0] Returns: a dict of percentile values indexed by the percentile
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/histogram.py#L287-L326
[ "def get_count_at_index(self, index):\n if index >= self.counts_len:\n raise IndexError()\n # some decoded (read-only) histograms may have truncated\n # counts arrays, we return zero for any index that is passed the array\n if index >= self.encoder.payload.counts_len:\n return 0\n return self.counts[index]\n", "def get_value_from_index(self, index):\n bucket_index = (index >> self.sub_bucket_half_count_magnitude) - 1\n sub_bucket_index = (index & (self.sub_bucket_half_count - 1)) + \\\n self.sub_bucket_half_count\n if bucket_index < 0:\n sub_bucket_index -= self.sub_bucket_half_count\n bucket_index = 0\n return self.get_value_from_sub_bucket(bucket_index, sub_bucket_index)\n", "def get_lowest_equivalent_value(self, value):\n bucket_index = self._get_bucket_index(value)\n sub_bucket_index = self._get_sub_bucket_index(value, bucket_index)\n\n lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index,\n sub_bucket_index)\n return lowest_equivalent_value\n", "def get_highest_equivalent_value(self, value):\n bucket_index = self._get_bucket_index(value)\n sub_bucket_index = self._get_sub_bucket_index(value, bucket_index)\n\n lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index,\n sub_bucket_index)\n if sub_bucket_index >= self.sub_bucket_count:\n bucket_index += 1\n size_of_equivalent_value_range = 1 << (self.unit_magnitude + bucket_index)\n next_non_equivalent_value = lowest_equivalent_value + size_of_equivalent_value_range\n\n return next_non_equivalent_value - 1\n", "def get_target_count_at_percentile(self, percentile):\n requested_percentile = min(percentile, 100.0)\n count_at_percentile = int(((requested_percentile * self.total_count / 100)) + 0.5)\n return max(count_at_percentile, 1)\n" ]
class HdrHistogram(): '''This class supports the recording and analyzing of sampled data value counts across a configurable integer value range with configurable value precision within the range. Value precision is expressed as the number of significant digits in the value recording, and provides control over value quantization behavior across the value range and the subsequent value resolution at any given level. For example, a Histogram could be configured to track the counts of observed integer values between 0 and 3,600,000,000 while maintaining a value precision of 3 significant digits across that range. Value quantization within the range will thus be no larger than 1/1,000th (or 0.1%) of any value. This example Histogram could be used to track and analyze the counts of observed response times ranging between 1 microsecond and 1 hour in magnitude, while maintaining a value resolution of 1 microsecond up to 1 millisecond, a resolution of 1 millisecond (or better) up to one second, and a resolution of 1 second (or better) up to 1,000 seconds. At it's maximum tracked value (1 hour), it would still maintain a resolution of 3.6 seconds (or better). ''' def __init__(self, lowest_trackable_value, highest_trackable_value, significant_figures, word_size=8, b64_wrap=True, hdr_payload=None): '''Create a new histogram with given arguments Params: lowest_trackable_value The lowest value that can be discerned (distinguished from 0) by the histogram. Must be a positive integer that is >= 1. May be internally rounded down to nearest power of 2. highest_trackable_value The highest value to be tracked by the histogram. Must be a positive integer that is >= (2 * lowest_trackable_value). significant_figures The number of significant decimal digits to which the histogram will maintain value resolution and separation. Must be a non-negative integer between 0 and 5. word_size size of counters in bytes, only 2, 4, 8-byte counters are supported (default is 8-byte or 64-bit counters) b64_wrap specifies if the encoding of this histogram should use base64 wrapping (only useful if you need to encode the histogram to save somewhere or send over the wire. By default base64 encoding is assumed hdr_payload only used for associating an existing payload created from decoding an encoded histograme Exceptions: ValueError if the word_size value is unsupported if significant_figures is invalid ''' if significant_figures < 1 or significant_figures > 5: raise ValueError('Invalid significant_figures') self.lowest_trackable_value = lowest_trackable_value self.highest_trackable_value = highest_trackable_value self.significant_figures = significant_figures self.unit_magnitude = int(math.floor(math.log(lowest_trackable_value) / math.log(2))) largest_value_single_unit_res = 2 * math.pow(10, significant_figures) subb_count_mag = int(math.ceil(math.log(largest_value_single_unit_res) / math.log(2))) self.sub_bucket_half_count_magnitude = subb_count_mag - 1 if subb_count_mag > 1 else 0 self.sub_bucket_count = int(math.pow(2, self.sub_bucket_half_count_magnitude + 1)) self.sub_bucket_half_count = self.sub_bucket_count // 2 self.sub_bucket_mask = (self.sub_bucket_count - 1) << self.unit_magnitude self.bucket_count = get_bucket_count(highest_trackable_value, self.sub_bucket_count, self.unit_magnitude) self.min_value = sys.maxsize self.max_value = 0 self.total_count = 0 self.counts_len = (self.bucket_count + 1) * (self.sub_bucket_count // 2) self.word_size = word_size if hdr_payload: payload = hdr_payload.payload self.int_to_double_conversion_ratio = payload.conversion_ratio_bits results = hdr_payload.init_counts(self.counts_len) if results['total']: self.set_internal_tacking_values(results['min_nonzero_index'], results['max_nonzero_index'], results['total']) else: self.int_to_double_conversion_ratio = 1.0 # to encode this histogram into a compressed/base64 format ready # to be exported self.b64_wrap = b64_wrap self.encoder = HdrHistogramEncoder(self, b64_wrap, hdr_payload) # the counters reside directly in the payload object # allocated by the encoder # so that compression for wire transfer can be done without copy self.counts = self.encoder.get_counts() self.start_time_stamp_msec = 0 self.end_time_stamp_msec = 0 def _clz(self, value): """calculate the leading zeros, equivalent to C __builtin_clzll() value in hex: value = 1 clz = 63 value = 2 clz = 62 value = 4 clz = 61 value = 1000 clz = 51 value = 1000000 clz = 39 """ return 63 - (len(bin(value)) - 3) def _get_bucket_index(self, value): # smallest power of 2 containing value pow2ceiling = 64 - self._clz(int(value) | self.sub_bucket_mask) return int(pow2ceiling - self.unit_magnitude - (self.sub_bucket_half_count_magnitude + 1)) def _get_sub_bucket_index(self, value, bucket_index): return int(value) >> (bucket_index + self.unit_magnitude) def _counts_index(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ): bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - subBucketHalfCount) + bucketBaseIndex return bucket_base_index + offset_in_bucket def _counts_index_for(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) return self._counts_index(bucket_index, sub_bucket_index) def record_value(self, value, count=1): '''Record a new value into the histogram Args: value: the value to record (must be in the valid range) count: incremental count (defaults to 1) ''' if value < 0: return False counts_index = self._counts_index_for(value) if (counts_index < 0) or (self.counts_len <= counts_index): return False self.counts[counts_index] += count self.total_count += count self.min_value = min(self.min_value, value) self.max_value = max(self.max_value, value) return True # pylint: disable=inconsistent-return-statements def record_corrected_value(self, value, expected_interval, count=1): '''Record a new value into the histogram and correct for coordinated omission if needed Args: value: the value to record (must be in the valid range) expected_interval: the expected interval between 2 value samples count: incremental count (defaults to 1) ''' while True: if not self.record_value(value, count): return False if value <= expected_interval or expected_interval <= 0: return True value -= expected_interval def get_count_at_index(self, index): if index >= self.counts_len: raise IndexError() # some decoded (read-only) histograms may have truncated # counts arrays, we return zero for any index that is passed the array if index >= self.encoder.payload.counts_len: return 0 return self.counts[index] def get_count_at_sub_bucket(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ) bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # (sub_bucket_index - subBucketHalfCount) + bucketBaseIndex counts_index = bucket_base_index + offset_in_bucket return self.counts[counts_index] def get_value_from_sub_bucket(self, bucket_index, sub_bucket_index): return sub_bucket_index << (bucket_index + self.unit_magnitude) def get_value_from_index(self, index): bucket_index = (index >> self.sub_bucket_half_count_magnitude) - 1 sub_bucket_index = (index & (self.sub_bucket_half_count - 1)) + \ self.sub_bucket_half_count if bucket_index < 0: sub_bucket_index -= self.sub_bucket_half_count bucket_index = 0 return self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) def get_lowest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) return lowest_equivalent_value def get_highest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 size_of_equivalent_value_range = 1 << (self.unit_magnitude + bucket_index) next_non_equivalent_value = lowest_equivalent_value + size_of_equivalent_value_range return next_non_equivalent_value - 1 def get_target_count_at_percentile(self, percentile): requested_percentile = min(percentile, 100.0) count_at_percentile = int(((requested_percentile * self.total_count / 100)) + 0.5) return max(count_at_percentile, 1) def get_value_at_percentile(self, percentile): '''Get the value for a given percentile Args: percentile: a float in [0.0..100.0] Returns: the value for the given percentile ''' count_at_percentile = self.get_target_count_at_percentile(percentile) total = 0 for index in range(self.counts_len): total += self.get_count_at_index(index) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: return self.get_highest_equivalent_value(value_at_index) return self.get_lowest_equivalent_value(value_at_index) return 0 def get_total_count(self): return self.total_count def get_count_at_value(self, value): counts_index = self._counts_index_for(value) return self.counts[counts_index] def values_are_equivalent(self, val1, val2): '''Check whether 2 values are equivalent (meaning they are in the same bucket/range) Returns: true if the 2 values are equivalent ''' return self.get_lowest_equivalent_value(val1) == self.get_lowest_equivalent_value(val2) def get_max_value(self): if self.max_value == 0: return 0 return self.get_highest_equivalent_value(self.max_value) def get_min_value(self): if self.counts[0] > 0 or self.total_count == 0: return 0 if sys.maxsize == self.min_value: return sys.maxsize return self.get_lowest_equivalent_value(self.min_value) def _hdr_size_of_equiv_value_range(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 return 1 << (self.unit_magnitude + bucket_index) def _hdr_median_equiv_value(self, value): return self.get_lowest_equivalent_value(value) + \ (self._hdr_size_of_equiv_value_range(value) >> 1) def get_mean_value(self): if not self.total_count: return 0.0 total = 0 itr = self.get_recorded_iterator() for item in itr: total += itr.count_at_this_value * self._hdr_median_equiv_value(item.value_iterated_to) return float(total) / self.total_count def get_stddev(self): if not self.total_count: return 0.0 mean = self.get_mean_value() geometric_dev_total = 0.0 for item in self.get_recorded_iterator(): dev = (self._hdr_median_equiv_value(item.value_iterated_to) * 1.0) - mean geometric_dev_total += (dev * dev) * item.count_added_in_this_iter_step return math.sqrt(geometric_dev_total / self.total_count) def reset(self): '''Reset the histogram to a pristine state ''' for index in range(self.counts_len): self.counts[index] = 0 self.total_count = 0 self.min_value = sys.maxsize self.max_value = 0 self.start_time_stamp_msec = sys.maxsize self.end_time_stamp_msec = 0 def __iter__(self): '''Returns the recorded iterator if iter(self) is called ''' return RecordedIterator(self) def get_all_values_iterator(self): return AllValuesIterator(self) def get_recorded_iterator(self): return RecordedIterator(self) def get_percentile_iterator(self, ticks_per_half_distance): return PercentileIterator(self, ticks_per_half_distance) def get_linear_iterator(self, value_units_per_bucket): return LinearIterator(self, value_units_per_bucket) def get_log_iterator(self, value_units_first_bucket, log_base): return LogIterator(self, value_units_first_bucket, log_base) def encode(self): '''Encode this histogram Return: a string containing the base64 encoded compressed histogram (V1 format) ''' return self.encoder.encode() def adjust_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) self.max_value = max(self.max_value, max_value) if min_non_zero_index >= 0: min_value = self.get_value_from_index(min_non_zero_index) self.min_value = min(self.min_value, min_value) self.total_count += total_added def set_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: self.max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) if min_non_zero_index >= 0: self.min_value = self.get_value_from_index(min_non_zero_index) self.total_count = total_added def get_counts_array_index(self, value): '''Return the index in the counts array for a given value ''' if value < 0: raise ValueError("Histogram recorded value cannot be negative.") bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) # Calculate the index for the first entry in the bucket: bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # The following is the equivalent of ((bucket_index + 1) * sub_bucket_half_count) # Calculate the offset in the bucket (can be negative for first bucket): offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - sub_bucket_half_count) + bucket_base_index return bucket_base_index + offset_in_bucket def get_start_time_stamp(self): return self.start_time_stamp_msec def set_start_time_stamp(self, time_stamp_msec): '''Set the start time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.start_time_stamp_msec = time_stamp_msec def get_end_time_stamp(self): return self.end_time_stamp_msec def set_end_time_stamp(self, time_stamp_msec): '''Set the end time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.end_time_stamp_msec = time_stamp_msec def add(self, other_hist): highest_recordable_value = \ self.get_highest_equivalent_value(self.get_value_from_index(self.counts_len - 1)) if highest_recordable_value < other_hist.get_max_value(): raise IndexError("The other histogram includes values that do not fit %d < %d" % (highest_recordable_value, other_hist.get_max_value())) if (self.bucket_count == other_hist.bucket_count) and \ (self.sub_bucket_count == other_hist.sub_bucket_count) and \ (self.unit_magnitude == other_hist.unit_magnitude) and \ (self.word_size == other_hist.word_size): # do an in-place addition of one array to another self.encoder.add(other_hist.encoder) self.total_count += other_hist.get_total_count() self.max_value = max(self.max_value, other_hist.get_max_value()) self.min_value = min(self.get_min_value(), other_hist.get_min_value()) else: # Arrays are not a direct match, so we can't just stream through and add them. # Instead, go through the array and add each non-zero value found at it's proper value: for index in range(other_hist.counts_len): other_count = other_hist.get_count_at_index(index) if other_count > 0: self.record_value(other_hist.get_value_from_index(index), other_count) self.start_time_stamp_msec = \ min(self.start_time_stamp_msec, other_hist.start_time_stamp_msec) self.end_time_stamp_msec = \ max(self.end_time_stamp_msec, other_hist.end_time_stamp_msec) def decode_and_add(self, encoded_histogram): '''Decode an encoded histogram and add it to this histogram Args: encoded_histogram (string) an encoded histogram following the V1 format, such as one returned by the encode() method Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' other_hist = HdrHistogram.decode(encoded_histogram, self.b64_wrap) self.add(other_hist) @staticmethod def decode(encoded_histogram, b64_wrap=True): '''Decode an encoded histogram and return a new histogram instance that has been initialized with the decoded content Return: a new histogram instance representing the decoded content Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' hdr_payload = HdrHistogramEncoder.decode(encoded_histogram, b64_wrap) payload = hdr_payload.payload histogram = HdrHistogram(payload.lowest_trackable_value, payload.highest_trackable_value, payload.significant_figures, hdr_payload=hdr_payload) return histogram def get_word_size(self): return self.word_size def output_percentile_distribution(self, out_file, output_value_unit_scaling_ratio, ticks_per_half_distance=5): out_file.write(b'%12s %14s %10s %14s\n\n' % (b'Value', b'Percentile', b'TotalCount', b'1/(1-Percentile)')) percentile_format = '%12.{}f %2.12f %10d %14.2f\n'.format(self.significant_figures) last_line_percentile_format = '%12.{}f %2.12f %10d\n'.format(self.significant_figures) for iter_value in self.get_percentile_iterator(ticks_per_half_distance): value = iter_value.value_iterated_to / output_value_unit_scaling_ratio percentile = iter_value.percentile_level_iterated_to / 100 total_count = iter_value.total_count_to_this_value if iter_value.percentile_level_iterated_to != 100: other = 1 / (1 - iter_value.percentile_level_iterated_to / 100) out_file.write(percentile_format.encode() % (value, percentile, total_count, other)) else: out_file.write(last_line_percentile_format.encode() % (value, percentile, total_count)) mean = self.get_mean_value() / output_value_unit_scaling_ratio stddev = self.get_stddev() out_file.write('#[Mean = %12.{0}f, StdDeviation = %12.{0}f]\n'. format(self.significant_figures).encode() % (mean, stddev)) max = self.get_max_value() / output_value_unit_scaling_ratio total = self.get_total_count() out_file.write('#[Max = %12.{0}f, TotalCount = %12.{0}f]\n'.format( self.significant_figures).encode() % (max, total)) out_file.write(b'#[Buckets = %12d, SubBuckets = %12d]\n' % ( self.bucket_count, self.sub_bucket_count))
HdrHistogram/HdrHistogram_py
hdrh/histogram.py
HdrHistogram.values_are_equivalent
python
def values_are_equivalent(self, val1, val2): '''Check whether 2 values are equivalent (meaning they are in the same bucket/range) Returns: true if the 2 values are equivalent ''' return self.get_lowest_equivalent_value(val1) == self.get_lowest_equivalent_value(val2)
Check whether 2 values are equivalent (meaning they are in the same bucket/range) Returns: true if the 2 values are equivalent
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/histogram.py#L335-L342
[ "def get_lowest_equivalent_value(self, value):\n bucket_index = self._get_bucket_index(value)\n sub_bucket_index = self._get_sub_bucket_index(value, bucket_index)\n\n lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index,\n sub_bucket_index)\n return lowest_equivalent_value\n" ]
class HdrHistogram(): '''This class supports the recording and analyzing of sampled data value counts across a configurable integer value range with configurable value precision within the range. Value precision is expressed as the number of significant digits in the value recording, and provides control over value quantization behavior across the value range and the subsequent value resolution at any given level. For example, a Histogram could be configured to track the counts of observed integer values between 0 and 3,600,000,000 while maintaining a value precision of 3 significant digits across that range. Value quantization within the range will thus be no larger than 1/1,000th (or 0.1%) of any value. This example Histogram could be used to track and analyze the counts of observed response times ranging between 1 microsecond and 1 hour in magnitude, while maintaining a value resolution of 1 microsecond up to 1 millisecond, a resolution of 1 millisecond (or better) up to one second, and a resolution of 1 second (or better) up to 1,000 seconds. At it's maximum tracked value (1 hour), it would still maintain a resolution of 3.6 seconds (or better). ''' def __init__(self, lowest_trackable_value, highest_trackable_value, significant_figures, word_size=8, b64_wrap=True, hdr_payload=None): '''Create a new histogram with given arguments Params: lowest_trackable_value The lowest value that can be discerned (distinguished from 0) by the histogram. Must be a positive integer that is >= 1. May be internally rounded down to nearest power of 2. highest_trackable_value The highest value to be tracked by the histogram. Must be a positive integer that is >= (2 * lowest_trackable_value). significant_figures The number of significant decimal digits to which the histogram will maintain value resolution and separation. Must be a non-negative integer between 0 and 5. word_size size of counters in bytes, only 2, 4, 8-byte counters are supported (default is 8-byte or 64-bit counters) b64_wrap specifies if the encoding of this histogram should use base64 wrapping (only useful if you need to encode the histogram to save somewhere or send over the wire. By default base64 encoding is assumed hdr_payload only used for associating an existing payload created from decoding an encoded histograme Exceptions: ValueError if the word_size value is unsupported if significant_figures is invalid ''' if significant_figures < 1 or significant_figures > 5: raise ValueError('Invalid significant_figures') self.lowest_trackable_value = lowest_trackable_value self.highest_trackable_value = highest_trackable_value self.significant_figures = significant_figures self.unit_magnitude = int(math.floor(math.log(lowest_trackable_value) / math.log(2))) largest_value_single_unit_res = 2 * math.pow(10, significant_figures) subb_count_mag = int(math.ceil(math.log(largest_value_single_unit_res) / math.log(2))) self.sub_bucket_half_count_magnitude = subb_count_mag - 1 if subb_count_mag > 1 else 0 self.sub_bucket_count = int(math.pow(2, self.sub_bucket_half_count_magnitude + 1)) self.sub_bucket_half_count = self.sub_bucket_count // 2 self.sub_bucket_mask = (self.sub_bucket_count - 1) << self.unit_magnitude self.bucket_count = get_bucket_count(highest_trackable_value, self.sub_bucket_count, self.unit_magnitude) self.min_value = sys.maxsize self.max_value = 0 self.total_count = 0 self.counts_len = (self.bucket_count + 1) * (self.sub_bucket_count // 2) self.word_size = word_size if hdr_payload: payload = hdr_payload.payload self.int_to_double_conversion_ratio = payload.conversion_ratio_bits results = hdr_payload.init_counts(self.counts_len) if results['total']: self.set_internal_tacking_values(results['min_nonzero_index'], results['max_nonzero_index'], results['total']) else: self.int_to_double_conversion_ratio = 1.0 # to encode this histogram into a compressed/base64 format ready # to be exported self.b64_wrap = b64_wrap self.encoder = HdrHistogramEncoder(self, b64_wrap, hdr_payload) # the counters reside directly in the payload object # allocated by the encoder # so that compression for wire transfer can be done without copy self.counts = self.encoder.get_counts() self.start_time_stamp_msec = 0 self.end_time_stamp_msec = 0 def _clz(self, value): """calculate the leading zeros, equivalent to C __builtin_clzll() value in hex: value = 1 clz = 63 value = 2 clz = 62 value = 4 clz = 61 value = 1000 clz = 51 value = 1000000 clz = 39 """ return 63 - (len(bin(value)) - 3) def _get_bucket_index(self, value): # smallest power of 2 containing value pow2ceiling = 64 - self._clz(int(value) | self.sub_bucket_mask) return int(pow2ceiling - self.unit_magnitude - (self.sub_bucket_half_count_magnitude + 1)) def _get_sub_bucket_index(self, value, bucket_index): return int(value) >> (bucket_index + self.unit_magnitude) def _counts_index(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ): bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - subBucketHalfCount) + bucketBaseIndex return bucket_base_index + offset_in_bucket def _counts_index_for(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) return self._counts_index(bucket_index, sub_bucket_index) def record_value(self, value, count=1): '''Record a new value into the histogram Args: value: the value to record (must be in the valid range) count: incremental count (defaults to 1) ''' if value < 0: return False counts_index = self._counts_index_for(value) if (counts_index < 0) or (self.counts_len <= counts_index): return False self.counts[counts_index] += count self.total_count += count self.min_value = min(self.min_value, value) self.max_value = max(self.max_value, value) return True # pylint: disable=inconsistent-return-statements def record_corrected_value(self, value, expected_interval, count=1): '''Record a new value into the histogram and correct for coordinated omission if needed Args: value: the value to record (must be in the valid range) expected_interval: the expected interval between 2 value samples count: incremental count (defaults to 1) ''' while True: if not self.record_value(value, count): return False if value <= expected_interval or expected_interval <= 0: return True value -= expected_interval def get_count_at_index(self, index): if index >= self.counts_len: raise IndexError() # some decoded (read-only) histograms may have truncated # counts arrays, we return zero for any index that is passed the array if index >= self.encoder.payload.counts_len: return 0 return self.counts[index] def get_count_at_sub_bucket(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ) bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # (sub_bucket_index - subBucketHalfCount) + bucketBaseIndex counts_index = bucket_base_index + offset_in_bucket return self.counts[counts_index] def get_value_from_sub_bucket(self, bucket_index, sub_bucket_index): return sub_bucket_index << (bucket_index + self.unit_magnitude) def get_value_from_index(self, index): bucket_index = (index >> self.sub_bucket_half_count_magnitude) - 1 sub_bucket_index = (index & (self.sub_bucket_half_count - 1)) + \ self.sub_bucket_half_count if bucket_index < 0: sub_bucket_index -= self.sub_bucket_half_count bucket_index = 0 return self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) def get_lowest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) return lowest_equivalent_value def get_highest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 size_of_equivalent_value_range = 1 << (self.unit_magnitude + bucket_index) next_non_equivalent_value = lowest_equivalent_value + size_of_equivalent_value_range return next_non_equivalent_value - 1 def get_target_count_at_percentile(self, percentile): requested_percentile = min(percentile, 100.0) count_at_percentile = int(((requested_percentile * self.total_count / 100)) + 0.5) return max(count_at_percentile, 1) def get_value_at_percentile(self, percentile): '''Get the value for a given percentile Args: percentile: a float in [0.0..100.0] Returns: the value for the given percentile ''' count_at_percentile = self.get_target_count_at_percentile(percentile) total = 0 for index in range(self.counts_len): total += self.get_count_at_index(index) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: return self.get_highest_equivalent_value(value_at_index) return self.get_lowest_equivalent_value(value_at_index) return 0 def get_percentile_to_value_dict(self, percentile_list): '''A faster alternative to query values for a list of percentiles. Args: percentile_list: a list of percentiles in any order, dups will be ignored each element in the list must be a float value in [0.0 .. 100.0] Returns: a dict of percentile values indexed by the percentile ''' result = {} total = 0 percentile_list_index = 0 count_at_percentile = 0 # remove dups and sort percentile_list = list(set(percentile_list)) percentile_list.sort() for index in range(self.counts_len): total += self.get_count_at_index(index) while True: # recalculate target based on next requested percentile if not count_at_percentile: if percentile_list_index == len(percentile_list): return result percentile = percentile_list[percentile_list_index] percentile_list_index += 1 if percentile > 100: return result count_at_percentile = self.get_target_count_at_percentile(percentile) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: result[percentile] = self.get_highest_equivalent_value(value_at_index) else: result[percentile] = self.get_lowest_equivalent_value(value_at_index) count_at_percentile = 0 else: break return result def get_total_count(self): return self.total_count def get_count_at_value(self, value): counts_index = self._counts_index_for(value) return self.counts[counts_index] def get_max_value(self): if self.max_value == 0: return 0 return self.get_highest_equivalent_value(self.max_value) def get_min_value(self): if self.counts[0] > 0 or self.total_count == 0: return 0 if sys.maxsize == self.min_value: return sys.maxsize return self.get_lowest_equivalent_value(self.min_value) def _hdr_size_of_equiv_value_range(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 return 1 << (self.unit_magnitude + bucket_index) def _hdr_median_equiv_value(self, value): return self.get_lowest_equivalent_value(value) + \ (self._hdr_size_of_equiv_value_range(value) >> 1) def get_mean_value(self): if not self.total_count: return 0.0 total = 0 itr = self.get_recorded_iterator() for item in itr: total += itr.count_at_this_value * self._hdr_median_equiv_value(item.value_iterated_to) return float(total) / self.total_count def get_stddev(self): if not self.total_count: return 0.0 mean = self.get_mean_value() geometric_dev_total = 0.0 for item in self.get_recorded_iterator(): dev = (self._hdr_median_equiv_value(item.value_iterated_to) * 1.0) - mean geometric_dev_total += (dev * dev) * item.count_added_in_this_iter_step return math.sqrt(geometric_dev_total / self.total_count) def reset(self): '''Reset the histogram to a pristine state ''' for index in range(self.counts_len): self.counts[index] = 0 self.total_count = 0 self.min_value = sys.maxsize self.max_value = 0 self.start_time_stamp_msec = sys.maxsize self.end_time_stamp_msec = 0 def __iter__(self): '''Returns the recorded iterator if iter(self) is called ''' return RecordedIterator(self) def get_all_values_iterator(self): return AllValuesIterator(self) def get_recorded_iterator(self): return RecordedIterator(self) def get_percentile_iterator(self, ticks_per_half_distance): return PercentileIterator(self, ticks_per_half_distance) def get_linear_iterator(self, value_units_per_bucket): return LinearIterator(self, value_units_per_bucket) def get_log_iterator(self, value_units_first_bucket, log_base): return LogIterator(self, value_units_first_bucket, log_base) def encode(self): '''Encode this histogram Return: a string containing the base64 encoded compressed histogram (V1 format) ''' return self.encoder.encode() def adjust_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) self.max_value = max(self.max_value, max_value) if min_non_zero_index >= 0: min_value = self.get_value_from_index(min_non_zero_index) self.min_value = min(self.min_value, min_value) self.total_count += total_added def set_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: self.max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) if min_non_zero_index >= 0: self.min_value = self.get_value_from_index(min_non_zero_index) self.total_count = total_added def get_counts_array_index(self, value): '''Return the index in the counts array for a given value ''' if value < 0: raise ValueError("Histogram recorded value cannot be negative.") bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) # Calculate the index for the first entry in the bucket: bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # The following is the equivalent of ((bucket_index + 1) * sub_bucket_half_count) # Calculate the offset in the bucket (can be negative for first bucket): offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - sub_bucket_half_count) + bucket_base_index return bucket_base_index + offset_in_bucket def get_start_time_stamp(self): return self.start_time_stamp_msec def set_start_time_stamp(self, time_stamp_msec): '''Set the start time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.start_time_stamp_msec = time_stamp_msec def get_end_time_stamp(self): return self.end_time_stamp_msec def set_end_time_stamp(self, time_stamp_msec): '''Set the end time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.end_time_stamp_msec = time_stamp_msec def add(self, other_hist): highest_recordable_value = \ self.get_highest_equivalent_value(self.get_value_from_index(self.counts_len - 1)) if highest_recordable_value < other_hist.get_max_value(): raise IndexError("The other histogram includes values that do not fit %d < %d" % (highest_recordable_value, other_hist.get_max_value())) if (self.bucket_count == other_hist.bucket_count) and \ (self.sub_bucket_count == other_hist.sub_bucket_count) and \ (self.unit_magnitude == other_hist.unit_magnitude) and \ (self.word_size == other_hist.word_size): # do an in-place addition of one array to another self.encoder.add(other_hist.encoder) self.total_count += other_hist.get_total_count() self.max_value = max(self.max_value, other_hist.get_max_value()) self.min_value = min(self.get_min_value(), other_hist.get_min_value()) else: # Arrays are not a direct match, so we can't just stream through and add them. # Instead, go through the array and add each non-zero value found at it's proper value: for index in range(other_hist.counts_len): other_count = other_hist.get_count_at_index(index) if other_count > 0: self.record_value(other_hist.get_value_from_index(index), other_count) self.start_time_stamp_msec = \ min(self.start_time_stamp_msec, other_hist.start_time_stamp_msec) self.end_time_stamp_msec = \ max(self.end_time_stamp_msec, other_hist.end_time_stamp_msec) def decode_and_add(self, encoded_histogram): '''Decode an encoded histogram and add it to this histogram Args: encoded_histogram (string) an encoded histogram following the V1 format, such as one returned by the encode() method Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' other_hist = HdrHistogram.decode(encoded_histogram, self.b64_wrap) self.add(other_hist) @staticmethod def decode(encoded_histogram, b64_wrap=True): '''Decode an encoded histogram and return a new histogram instance that has been initialized with the decoded content Return: a new histogram instance representing the decoded content Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' hdr_payload = HdrHistogramEncoder.decode(encoded_histogram, b64_wrap) payload = hdr_payload.payload histogram = HdrHistogram(payload.lowest_trackable_value, payload.highest_trackable_value, payload.significant_figures, hdr_payload=hdr_payload) return histogram def get_word_size(self): return self.word_size def output_percentile_distribution(self, out_file, output_value_unit_scaling_ratio, ticks_per_half_distance=5): out_file.write(b'%12s %14s %10s %14s\n\n' % (b'Value', b'Percentile', b'TotalCount', b'1/(1-Percentile)')) percentile_format = '%12.{}f %2.12f %10d %14.2f\n'.format(self.significant_figures) last_line_percentile_format = '%12.{}f %2.12f %10d\n'.format(self.significant_figures) for iter_value in self.get_percentile_iterator(ticks_per_half_distance): value = iter_value.value_iterated_to / output_value_unit_scaling_ratio percentile = iter_value.percentile_level_iterated_to / 100 total_count = iter_value.total_count_to_this_value if iter_value.percentile_level_iterated_to != 100: other = 1 / (1 - iter_value.percentile_level_iterated_to / 100) out_file.write(percentile_format.encode() % (value, percentile, total_count, other)) else: out_file.write(last_line_percentile_format.encode() % (value, percentile, total_count)) mean = self.get_mean_value() / output_value_unit_scaling_ratio stddev = self.get_stddev() out_file.write('#[Mean = %12.{0}f, StdDeviation = %12.{0}f]\n'. format(self.significant_figures).encode() % (mean, stddev)) max = self.get_max_value() / output_value_unit_scaling_ratio total = self.get_total_count() out_file.write('#[Max = %12.{0}f, TotalCount = %12.{0}f]\n'.format( self.significant_figures).encode() % (max, total)) out_file.write(b'#[Buckets = %12d, SubBuckets = %12d]\n' % ( self.bucket_count, self.sub_bucket_count))
HdrHistogram/HdrHistogram_py
hdrh/histogram.py
HdrHistogram.reset
python
def reset(self): '''Reset the histogram to a pristine state ''' for index in range(self.counts_len): self.counts[index] = 0 self.total_count = 0 self.min_value = sys.maxsize self.max_value = 0 self.start_time_stamp_msec = sys.maxsize self.end_time_stamp_msec = 0
Reset the histogram to a pristine state
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/histogram.py#L386-L395
null
class HdrHistogram(): '''This class supports the recording and analyzing of sampled data value counts across a configurable integer value range with configurable value precision within the range. Value precision is expressed as the number of significant digits in the value recording, and provides control over value quantization behavior across the value range and the subsequent value resolution at any given level. For example, a Histogram could be configured to track the counts of observed integer values between 0 and 3,600,000,000 while maintaining a value precision of 3 significant digits across that range. Value quantization within the range will thus be no larger than 1/1,000th (or 0.1%) of any value. This example Histogram could be used to track and analyze the counts of observed response times ranging between 1 microsecond and 1 hour in magnitude, while maintaining a value resolution of 1 microsecond up to 1 millisecond, a resolution of 1 millisecond (or better) up to one second, and a resolution of 1 second (or better) up to 1,000 seconds. At it's maximum tracked value (1 hour), it would still maintain a resolution of 3.6 seconds (or better). ''' def __init__(self, lowest_trackable_value, highest_trackable_value, significant_figures, word_size=8, b64_wrap=True, hdr_payload=None): '''Create a new histogram with given arguments Params: lowest_trackable_value The lowest value that can be discerned (distinguished from 0) by the histogram. Must be a positive integer that is >= 1. May be internally rounded down to nearest power of 2. highest_trackable_value The highest value to be tracked by the histogram. Must be a positive integer that is >= (2 * lowest_trackable_value). significant_figures The number of significant decimal digits to which the histogram will maintain value resolution and separation. Must be a non-negative integer between 0 and 5. word_size size of counters in bytes, only 2, 4, 8-byte counters are supported (default is 8-byte or 64-bit counters) b64_wrap specifies if the encoding of this histogram should use base64 wrapping (only useful if you need to encode the histogram to save somewhere or send over the wire. By default base64 encoding is assumed hdr_payload only used for associating an existing payload created from decoding an encoded histograme Exceptions: ValueError if the word_size value is unsupported if significant_figures is invalid ''' if significant_figures < 1 or significant_figures > 5: raise ValueError('Invalid significant_figures') self.lowest_trackable_value = lowest_trackable_value self.highest_trackable_value = highest_trackable_value self.significant_figures = significant_figures self.unit_magnitude = int(math.floor(math.log(lowest_trackable_value) / math.log(2))) largest_value_single_unit_res = 2 * math.pow(10, significant_figures) subb_count_mag = int(math.ceil(math.log(largest_value_single_unit_res) / math.log(2))) self.sub_bucket_half_count_magnitude = subb_count_mag - 1 if subb_count_mag > 1 else 0 self.sub_bucket_count = int(math.pow(2, self.sub_bucket_half_count_magnitude + 1)) self.sub_bucket_half_count = self.sub_bucket_count // 2 self.sub_bucket_mask = (self.sub_bucket_count - 1) << self.unit_magnitude self.bucket_count = get_bucket_count(highest_trackable_value, self.sub_bucket_count, self.unit_magnitude) self.min_value = sys.maxsize self.max_value = 0 self.total_count = 0 self.counts_len = (self.bucket_count + 1) * (self.sub_bucket_count // 2) self.word_size = word_size if hdr_payload: payload = hdr_payload.payload self.int_to_double_conversion_ratio = payload.conversion_ratio_bits results = hdr_payload.init_counts(self.counts_len) if results['total']: self.set_internal_tacking_values(results['min_nonzero_index'], results['max_nonzero_index'], results['total']) else: self.int_to_double_conversion_ratio = 1.0 # to encode this histogram into a compressed/base64 format ready # to be exported self.b64_wrap = b64_wrap self.encoder = HdrHistogramEncoder(self, b64_wrap, hdr_payload) # the counters reside directly in the payload object # allocated by the encoder # so that compression for wire transfer can be done without copy self.counts = self.encoder.get_counts() self.start_time_stamp_msec = 0 self.end_time_stamp_msec = 0 def _clz(self, value): """calculate the leading zeros, equivalent to C __builtin_clzll() value in hex: value = 1 clz = 63 value = 2 clz = 62 value = 4 clz = 61 value = 1000 clz = 51 value = 1000000 clz = 39 """ return 63 - (len(bin(value)) - 3) def _get_bucket_index(self, value): # smallest power of 2 containing value pow2ceiling = 64 - self._clz(int(value) | self.sub_bucket_mask) return int(pow2ceiling - self.unit_magnitude - (self.sub_bucket_half_count_magnitude + 1)) def _get_sub_bucket_index(self, value, bucket_index): return int(value) >> (bucket_index + self.unit_magnitude) def _counts_index(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ): bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - subBucketHalfCount) + bucketBaseIndex return bucket_base_index + offset_in_bucket def _counts_index_for(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) return self._counts_index(bucket_index, sub_bucket_index) def record_value(self, value, count=1): '''Record a new value into the histogram Args: value: the value to record (must be in the valid range) count: incremental count (defaults to 1) ''' if value < 0: return False counts_index = self._counts_index_for(value) if (counts_index < 0) or (self.counts_len <= counts_index): return False self.counts[counts_index] += count self.total_count += count self.min_value = min(self.min_value, value) self.max_value = max(self.max_value, value) return True # pylint: disable=inconsistent-return-statements def record_corrected_value(self, value, expected_interval, count=1): '''Record a new value into the histogram and correct for coordinated omission if needed Args: value: the value to record (must be in the valid range) expected_interval: the expected interval between 2 value samples count: incremental count (defaults to 1) ''' while True: if not self.record_value(value, count): return False if value <= expected_interval or expected_interval <= 0: return True value -= expected_interval def get_count_at_index(self, index): if index >= self.counts_len: raise IndexError() # some decoded (read-only) histograms may have truncated # counts arrays, we return zero for any index that is passed the array if index >= self.encoder.payload.counts_len: return 0 return self.counts[index] def get_count_at_sub_bucket(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ) bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # (sub_bucket_index - subBucketHalfCount) + bucketBaseIndex counts_index = bucket_base_index + offset_in_bucket return self.counts[counts_index] def get_value_from_sub_bucket(self, bucket_index, sub_bucket_index): return sub_bucket_index << (bucket_index + self.unit_magnitude) def get_value_from_index(self, index): bucket_index = (index >> self.sub_bucket_half_count_magnitude) - 1 sub_bucket_index = (index & (self.sub_bucket_half_count - 1)) + \ self.sub_bucket_half_count if bucket_index < 0: sub_bucket_index -= self.sub_bucket_half_count bucket_index = 0 return self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) def get_lowest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) return lowest_equivalent_value def get_highest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 size_of_equivalent_value_range = 1 << (self.unit_magnitude + bucket_index) next_non_equivalent_value = lowest_equivalent_value + size_of_equivalent_value_range return next_non_equivalent_value - 1 def get_target_count_at_percentile(self, percentile): requested_percentile = min(percentile, 100.0) count_at_percentile = int(((requested_percentile * self.total_count / 100)) + 0.5) return max(count_at_percentile, 1) def get_value_at_percentile(self, percentile): '''Get the value for a given percentile Args: percentile: a float in [0.0..100.0] Returns: the value for the given percentile ''' count_at_percentile = self.get_target_count_at_percentile(percentile) total = 0 for index in range(self.counts_len): total += self.get_count_at_index(index) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: return self.get_highest_equivalent_value(value_at_index) return self.get_lowest_equivalent_value(value_at_index) return 0 def get_percentile_to_value_dict(self, percentile_list): '''A faster alternative to query values for a list of percentiles. Args: percentile_list: a list of percentiles in any order, dups will be ignored each element in the list must be a float value in [0.0 .. 100.0] Returns: a dict of percentile values indexed by the percentile ''' result = {} total = 0 percentile_list_index = 0 count_at_percentile = 0 # remove dups and sort percentile_list = list(set(percentile_list)) percentile_list.sort() for index in range(self.counts_len): total += self.get_count_at_index(index) while True: # recalculate target based on next requested percentile if not count_at_percentile: if percentile_list_index == len(percentile_list): return result percentile = percentile_list[percentile_list_index] percentile_list_index += 1 if percentile > 100: return result count_at_percentile = self.get_target_count_at_percentile(percentile) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: result[percentile] = self.get_highest_equivalent_value(value_at_index) else: result[percentile] = self.get_lowest_equivalent_value(value_at_index) count_at_percentile = 0 else: break return result def get_total_count(self): return self.total_count def get_count_at_value(self, value): counts_index = self._counts_index_for(value) return self.counts[counts_index] def values_are_equivalent(self, val1, val2): '''Check whether 2 values are equivalent (meaning they are in the same bucket/range) Returns: true if the 2 values are equivalent ''' return self.get_lowest_equivalent_value(val1) == self.get_lowest_equivalent_value(val2) def get_max_value(self): if self.max_value == 0: return 0 return self.get_highest_equivalent_value(self.max_value) def get_min_value(self): if self.counts[0] > 0 or self.total_count == 0: return 0 if sys.maxsize == self.min_value: return sys.maxsize return self.get_lowest_equivalent_value(self.min_value) def _hdr_size_of_equiv_value_range(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 return 1 << (self.unit_magnitude + bucket_index) def _hdr_median_equiv_value(self, value): return self.get_lowest_equivalent_value(value) + \ (self._hdr_size_of_equiv_value_range(value) >> 1) def get_mean_value(self): if not self.total_count: return 0.0 total = 0 itr = self.get_recorded_iterator() for item in itr: total += itr.count_at_this_value * self._hdr_median_equiv_value(item.value_iterated_to) return float(total) / self.total_count def get_stddev(self): if not self.total_count: return 0.0 mean = self.get_mean_value() geometric_dev_total = 0.0 for item in self.get_recorded_iterator(): dev = (self._hdr_median_equiv_value(item.value_iterated_to) * 1.0) - mean geometric_dev_total += (dev * dev) * item.count_added_in_this_iter_step return math.sqrt(geometric_dev_total / self.total_count) def __iter__(self): '''Returns the recorded iterator if iter(self) is called ''' return RecordedIterator(self) def get_all_values_iterator(self): return AllValuesIterator(self) def get_recorded_iterator(self): return RecordedIterator(self) def get_percentile_iterator(self, ticks_per_half_distance): return PercentileIterator(self, ticks_per_half_distance) def get_linear_iterator(self, value_units_per_bucket): return LinearIterator(self, value_units_per_bucket) def get_log_iterator(self, value_units_first_bucket, log_base): return LogIterator(self, value_units_first_bucket, log_base) def encode(self): '''Encode this histogram Return: a string containing the base64 encoded compressed histogram (V1 format) ''' return self.encoder.encode() def adjust_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) self.max_value = max(self.max_value, max_value) if min_non_zero_index >= 0: min_value = self.get_value_from_index(min_non_zero_index) self.min_value = min(self.min_value, min_value) self.total_count += total_added def set_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: self.max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) if min_non_zero_index >= 0: self.min_value = self.get_value_from_index(min_non_zero_index) self.total_count = total_added def get_counts_array_index(self, value): '''Return the index in the counts array for a given value ''' if value < 0: raise ValueError("Histogram recorded value cannot be negative.") bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) # Calculate the index for the first entry in the bucket: bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # The following is the equivalent of ((bucket_index + 1) * sub_bucket_half_count) # Calculate the offset in the bucket (can be negative for first bucket): offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - sub_bucket_half_count) + bucket_base_index return bucket_base_index + offset_in_bucket def get_start_time_stamp(self): return self.start_time_stamp_msec def set_start_time_stamp(self, time_stamp_msec): '''Set the start time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.start_time_stamp_msec = time_stamp_msec def get_end_time_stamp(self): return self.end_time_stamp_msec def set_end_time_stamp(self, time_stamp_msec): '''Set the end time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.end_time_stamp_msec = time_stamp_msec def add(self, other_hist): highest_recordable_value = \ self.get_highest_equivalent_value(self.get_value_from_index(self.counts_len - 1)) if highest_recordable_value < other_hist.get_max_value(): raise IndexError("The other histogram includes values that do not fit %d < %d" % (highest_recordable_value, other_hist.get_max_value())) if (self.bucket_count == other_hist.bucket_count) and \ (self.sub_bucket_count == other_hist.sub_bucket_count) and \ (self.unit_magnitude == other_hist.unit_magnitude) and \ (self.word_size == other_hist.word_size): # do an in-place addition of one array to another self.encoder.add(other_hist.encoder) self.total_count += other_hist.get_total_count() self.max_value = max(self.max_value, other_hist.get_max_value()) self.min_value = min(self.get_min_value(), other_hist.get_min_value()) else: # Arrays are not a direct match, so we can't just stream through and add them. # Instead, go through the array and add each non-zero value found at it's proper value: for index in range(other_hist.counts_len): other_count = other_hist.get_count_at_index(index) if other_count > 0: self.record_value(other_hist.get_value_from_index(index), other_count) self.start_time_stamp_msec = \ min(self.start_time_stamp_msec, other_hist.start_time_stamp_msec) self.end_time_stamp_msec = \ max(self.end_time_stamp_msec, other_hist.end_time_stamp_msec) def decode_and_add(self, encoded_histogram): '''Decode an encoded histogram and add it to this histogram Args: encoded_histogram (string) an encoded histogram following the V1 format, such as one returned by the encode() method Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' other_hist = HdrHistogram.decode(encoded_histogram, self.b64_wrap) self.add(other_hist) @staticmethod def decode(encoded_histogram, b64_wrap=True): '''Decode an encoded histogram and return a new histogram instance that has been initialized with the decoded content Return: a new histogram instance representing the decoded content Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' hdr_payload = HdrHistogramEncoder.decode(encoded_histogram, b64_wrap) payload = hdr_payload.payload histogram = HdrHistogram(payload.lowest_trackable_value, payload.highest_trackable_value, payload.significant_figures, hdr_payload=hdr_payload) return histogram def get_word_size(self): return self.word_size def output_percentile_distribution(self, out_file, output_value_unit_scaling_ratio, ticks_per_half_distance=5): out_file.write(b'%12s %14s %10s %14s\n\n' % (b'Value', b'Percentile', b'TotalCount', b'1/(1-Percentile)')) percentile_format = '%12.{}f %2.12f %10d %14.2f\n'.format(self.significant_figures) last_line_percentile_format = '%12.{}f %2.12f %10d\n'.format(self.significant_figures) for iter_value in self.get_percentile_iterator(ticks_per_half_distance): value = iter_value.value_iterated_to / output_value_unit_scaling_ratio percentile = iter_value.percentile_level_iterated_to / 100 total_count = iter_value.total_count_to_this_value if iter_value.percentile_level_iterated_to != 100: other = 1 / (1 - iter_value.percentile_level_iterated_to / 100) out_file.write(percentile_format.encode() % (value, percentile, total_count, other)) else: out_file.write(last_line_percentile_format.encode() % (value, percentile, total_count)) mean = self.get_mean_value() / output_value_unit_scaling_ratio stddev = self.get_stddev() out_file.write('#[Mean = %12.{0}f, StdDeviation = %12.{0}f]\n'. format(self.significant_figures).encode() % (mean, stddev)) max = self.get_max_value() / output_value_unit_scaling_ratio total = self.get_total_count() out_file.write('#[Max = %12.{0}f, TotalCount = %12.{0}f]\n'.format( self.significant_figures).encode() % (max, total)) out_file.write(b'#[Buckets = %12d, SubBuckets = %12d]\n' % ( self.bucket_count, self.sub_bucket_count))
HdrHistogram/HdrHistogram_py
hdrh/histogram.py
HdrHistogram.adjust_internal_tacking_values
python
def adjust_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) self.max_value = max(self.max_value, max_value) if min_non_zero_index >= 0: min_value = self.get_value_from_index(min_non_zero_index) self.min_value = min(self.min_value, min_value) self.total_count += total_added
Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none)
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/histogram.py#L424-L441
[ "def get_value_from_index(self, index):\n bucket_index = (index >> self.sub_bucket_half_count_magnitude) - 1\n sub_bucket_index = (index & (self.sub_bucket_half_count - 1)) + \\\n self.sub_bucket_half_count\n if bucket_index < 0:\n sub_bucket_index -= self.sub_bucket_half_count\n bucket_index = 0\n return self.get_value_from_sub_bucket(bucket_index, sub_bucket_index)\n", "def get_highest_equivalent_value(self, value):\n bucket_index = self._get_bucket_index(value)\n sub_bucket_index = self._get_sub_bucket_index(value, bucket_index)\n\n lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index,\n sub_bucket_index)\n if sub_bucket_index >= self.sub_bucket_count:\n bucket_index += 1\n size_of_equivalent_value_range = 1 << (self.unit_magnitude + bucket_index)\n next_non_equivalent_value = lowest_equivalent_value + size_of_equivalent_value_range\n\n return next_non_equivalent_value - 1\n" ]
class HdrHistogram(): '''This class supports the recording and analyzing of sampled data value counts across a configurable integer value range with configurable value precision within the range. Value precision is expressed as the number of significant digits in the value recording, and provides control over value quantization behavior across the value range and the subsequent value resolution at any given level. For example, a Histogram could be configured to track the counts of observed integer values between 0 and 3,600,000,000 while maintaining a value precision of 3 significant digits across that range. Value quantization within the range will thus be no larger than 1/1,000th (or 0.1%) of any value. This example Histogram could be used to track and analyze the counts of observed response times ranging between 1 microsecond and 1 hour in magnitude, while maintaining a value resolution of 1 microsecond up to 1 millisecond, a resolution of 1 millisecond (or better) up to one second, and a resolution of 1 second (or better) up to 1,000 seconds. At it's maximum tracked value (1 hour), it would still maintain a resolution of 3.6 seconds (or better). ''' def __init__(self, lowest_trackable_value, highest_trackable_value, significant_figures, word_size=8, b64_wrap=True, hdr_payload=None): '''Create a new histogram with given arguments Params: lowest_trackable_value The lowest value that can be discerned (distinguished from 0) by the histogram. Must be a positive integer that is >= 1. May be internally rounded down to nearest power of 2. highest_trackable_value The highest value to be tracked by the histogram. Must be a positive integer that is >= (2 * lowest_trackable_value). significant_figures The number of significant decimal digits to which the histogram will maintain value resolution and separation. Must be a non-negative integer between 0 and 5. word_size size of counters in bytes, only 2, 4, 8-byte counters are supported (default is 8-byte or 64-bit counters) b64_wrap specifies if the encoding of this histogram should use base64 wrapping (only useful if you need to encode the histogram to save somewhere or send over the wire. By default base64 encoding is assumed hdr_payload only used for associating an existing payload created from decoding an encoded histograme Exceptions: ValueError if the word_size value is unsupported if significant_figures is invalid ''' if significant_figures < 1 or significant_figures > 5: raise ValueError('Invalid significant_figures') self.lowest_trackable_value = lowest_trackable_value self.highest_trackable_value = highest_trackable_value self.significant_figures = significant_figures self.unit_magnitude = int(math.floor(math.log(lowest_trackable_value) / math.log(2))) largest_value_single_unit_res = 2 * math.pow(10, significant_figures) subb_count_mag = int(math.ceil(math.log(largest_value_single_unit_res) / math.log(2))) self.sub_bucket_half_count_magnitude = subb_count_mag - 1 if subb_count_mag > 1 else 0 self.sub_bucket_count = int(math.pow(2, self.sub_bucket_half_count_magnitude + 1)) self.sub_bucket_half_count = self.sub_bucket_count // 2 self.sub_bucket_mask = (self.sub_bucket_count - 1) << self.unit_magnitude self.bucket_count = get_bucket_count(highest_trackable_value, self.sub_bucket_count, self.unit_magnitude) self.min_value = sys.maxsize self.max_value = 0 self.total_count = 0 self.counts_len = (self.bucket_count + 1) * (self.sub_bucket_count // 2) self.word_size = word_size if hdr_payload: payload = hdr_payload.payload self.int_to_double_conversion_ratio = payload.conversion_ratio_bits results = hdr_payload.init_counts(self.counts_len) if results['total']: self.set_internal_tacking_values(results['min_nonzero_index'], results['max_nonzero_index'], results['total']) else: self.int_to_double_conversion_ratio = 1.0 # to encode this histogram into a compressed/base64 format ready # to be exported self.b64_wrap = b64_wrap self.encoder = HdrHistogramEncoder(self, b64_wrap, hdr_payload) # the counters reside directly in the payload object # allocated by the encoder # so that compression for wire transfer can be done without copy self.counts = self.encoder.get_counts() self.start_time_stamp_msec = 0 self.end_time_stamp_msec = 0 def _clz(self, value): """calculate the leading zeros, equivalent to C __builtin_clzll() value in hex: value = 1 clz = 63 value = 2 clz = 62 value = 4 clz = 61 value = 1000 clz = 51 value = 1000000 clz = 39 """ return 63 - (len(bin(value)) - 3) def _get_bucket_index(self, value): # smallest power of 2 containing value pow2ceiling = 64 - self._clz(int(value) | self.sub_bucket_mask) return int(pow2ceiling - self.unit_magnitude - (self.sub_bucket_half_count_magnitude + 1)) def _get_sub_bucket_index(self, value, bucket_index): return int(value) >> (bucket_index + self.unit_magnitude) def _counts_index(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ): bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - subBucketHalfCount) + bucketBaseIndex return bucket_base_index + offset_in_bucket def _counts_index_for(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) return self._counts_index(bucket_index, sub_bucket_index) def record_value(self, value, count=1): '''Record a new value into the histogram Args: value: the value to record (must be in the valid range) count: incremental count (defaults to 1) ''' if value < 0: return False counts_index = self._counts_index_for(value) if (counts_index < 0) or (self.counts_len <= counts_index): return False self.counts[counts_index] += count self.total_count += count self.min_value = min(self.min_value, value) self.max_value = max(self.max_value, value) return True # pylint: disable=inconsistent-return-statements def record_corrected_value(self, value, expected_interval, count=1): '''Record a new value into the histogram and correct for coordinated omission if needed Args: value: the value to record (must be in the valid range) expected_interval: the expected interval between 2 value samples count: incremental count (defaults to 1) ''' while True: if not self.record_value(value, count): return False if value <= expected_interval or expected_interval <= 0: return True value -= expected_interval def get_count_at_index(self, index): if index >= self.counts_len: raise IndexError() # some decoded (read-only) histograms may have truncated # counts arrays, we return zero for any index that is passed the array if index >= self.encoder.payload.counts_len: return 0 return self.counts[index] def get_count_at_sub_bucket(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ) bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # (sub_bucket_index - subBucketHalfCount) + bucketBaseIndex counts_index = bucket_base_index + offset_in_bucket return self.counts[counts_index] def get_value_from_sub_bucket(self, bucket_index, sub_bucket_index): return sub_bucket_index << (bucket_index + self.unit_magnitude) def get_value_from_index(self, index): bucket_index = (index >> self.sub_bucket_half_count_magnitude) - 1 sub_bucket_index = (index & (self.sub_bucket_half_count - 1)) + \ self.sub_bucket_half_count if bucket_index < 0: sub_bucket_index -= self.sub_bucket_half_count bucket_index = 0 return self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) def get_lowest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) return lowest_equivalent_value def get_highest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 size_of_equivalent_value_range = 1 << (self.unit_magnitude + bucket_index) next_non_equivalent_value = lowest_equivalent_value + size_of_equivalent_value_range return next_non_equivalent_value - 1 def get_target_count_at_percentile(self, percentile): requested_percentile = min(percentile, 100.0) count_at_percentile = int(((requested_percentile * self.total_count / 100)) + 0.5) return max(count_at_percentile, 1) def get_value_at_percentile(self, percentile): '''Get the value for a given percentile Args: percentile: a float in [0.0..100.0] Returns: the value for the given percentile ''' count_at_percentile = self.get_target_count_at_percentile(percentile) total = 0 for index in range(self.counts_len): total += self.get_count_at_index(index) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: return self.get_highest_equivalent_value(value_at_index) return self.get_lowest_equivalent_value(value_at_index) return 0 def get_percentile_to_value_dict(self, percentile_list): '''A faster alternative to query values for a list of percentiles. Args: percentile_list: a list of percentiles in any order, dups will be ignored each element in the list must be a float value in [0.0 .. 100.0] Returns: a dict of percentile values indexed by the percentile ''' result = {} total = 0 percentile_list_index = 0 count_at_percentile = 0 # remove dups and sort percentile_list = list(set(percentile_list)) percentile_list.sort() for index in range(self.counts_len): total += self.get_count_at_index(index) while True: # recalculate target based on next requested percentile if not count_at_percentile: if percentile_list_index == len(percentile_list): return result percentile = percentile_list[percentile_list_index] percentile_list_index += 1 if percentile > 100: return result count_at_percentile = self.get_target_count_at_percentile(percentile) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: result[percentile] = self.get_highest_equivalent_value(value_at_index) else: result[percentile] = self.get_lowest_equivalent_value(value_at_index) count_at_percentile = 0 else: break return result def get_total_count(self): return self.total_count def get_count_at_value(self, value): counts_index = self._counts_index_for(value) return self.counts[counts_index] def values_are_equivalent(self, val1, val2): '''Check whether 2 values are equivalent (meaning they are in the same bucket/range) Returns: true if the 2 values are equivalent ''' return self.get_lowest_equivalent_value(val1) == self.get_lowest_equivalent_value(val2) def get_max_value(self): if self.max_value == 0: return 0 return self.get_highest_equivalent_value(self.max_value) def get_min_value(self): if self.counts[0] > 0 or self.total_count == 0: return 0 if sys.maxsize == self.min_value: return sys.maxsize return self.get_lowest_equivalent_value(self.min_value) def _hdr_size_of_equiv_value_range(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 return 1 << (self.unit_magnitude + bucket_index) def _hdr_median_equiv_value(self, value): return self.get_lowest_equivalent_value(value) + \ (self._hdr_size_of_equiv_value_range(value) >> 1) def get_mean_value(self): if not self.total_count: return 0.0 total = 0 itr = self.get_recorded_iterator() for item in itr: total += itr.count_at_this_value * self._hdr_median_equiv_value(item.value_iterated_to) return float(total) / self.total_count def get_stddev(self): if not self.total_count: return 0.0 mean = self.get_mean_value() geometric_dev_total = 0.0 for item in self.get_recorded_iterator(): dev = (self._hdr_median_equiv_value(item.value_iterated_to) * 1.0) - mean geometric_dev_total += (dev * dev) * item.count_added_in_this_iter_step return math.sqrt(geometric_dev_total / self.total_count) def reset(self): '''Reset the histogram to a pristine state ''' for index in range(self.counts_len): self.counts[index] = 0 self.total_count = 0 self.min_value = sys.maxsize self.max_value = 0 self.start_time_stamp_msec = sys.maxsize self.end_time_stamp_msec = 0 def __iter__(self): '''Returns the recorded iterator if iter(self) is called ''' return RecordedIterator(self) def get_all_values_iterator(self): return AllValuesIterator(self) def get_recorded_iterator(self): return RecordedIterator(self) def get_percentile_iterator(self, ticks_per_half_distance): return PercentileIterator(self, ticks_per_half_distance) def get_linear_iterator(self, value_units_per_bucket): return LinearIterator(self, value_units_per_bucket) def get_log_iterator(self, value_units_first_bucket, log_base): return LogIterator(self, value_units_first_bucket, log_base) def encode(self): '''Encode this histogram Return: a string containing the base64 encoded compressed histogram (V1 format) ''' return self.encoder.encode() def set_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: self.max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) if min_non_zero_index >= 0: self.min_value = self.get_value_from_index(min_non_zero_index) self.total_count = total_added def get_counts_array_index(self, value): '''Return the index in the counts array for a given value ''' if value < 0: raise ValueError("Histogram recorded value cannot be negative.") bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) # Calculate the index for the first entry in the bucket: bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # The following is the equivalent of ((bucket_index + 1) * sub_bucket_half_count) # Calculate the offset in the bucket (can be negative for first bucket): offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - sub_bucket_half_count) + bucket_base_index return bucket_base_index + offset_in_bucket def get_start_time_stamp(self): return self.start_time_stamp_msec def set_start_time_stamp(self, time_stamp_msec): '''Set the start time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.start_time_stamp_msec = time_stamp_msec def get_end_time_stamp(self): return self.end_time_stamp_msec def set_end_time_stamp(self, time_stamp_msec): '''Set the end time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.end_time_stamp_msec = time_stamp_msec def add(self, other_hist): highest_recordable_value = \ self.get_highest_equivalent_value(self.get_value_from_index(self.counts_len - 1)) if highest_recordable_value < other_hist.get_max_value(): raise IndexError("The other histogram includes values that do not fit %d < %d" % (highest_recordable_value, other_hist.get_max_value())) if (self.bucket_count == other_hist.bucket_count) and \ (self.sub_bucket_count == other_hist.sub_bucket_count) and \ (self.unit_magnitude == other_hist.unit_magnitude) and \ (self.word_size == other_hist.word_size): # do an in-place addition of one array to another self.encoder.add(other_hist.encoder) self.total_count += other_hist.get_total_count() self.max_value = max(self.max_value, other_hist.get_max_value()) self.min_value = min(self.get_min_value(), other_hist.get_min_value()) else: # Arrays are not a direct match, so we can't just stream through and add them. # Instead, go through the array and add each non-zero value found at it's proper value: for index in range(other_hist.counts_len): other_count = other_hist.get_count_at_index(index) if other_count > 0: self.record_value(other_hist.get_value_from_index(index), other_count) self.start_time_stamp_msec = \ min(self.start_time_stamp_msec, other_hist.start_time_stamp_msec) self.end_time_stamp_msec = \ max(self.end_time_stamp_msec, other_hist.end_time_stamp_msec) def decode_and_add(self, encoded_histogram): '''Decode an encoded histogram and add it to this histogram Args: encoded_histogram (string) an encoded histogram following the V1 format, such as one returned by the encode() method Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' other_hist = HdrHistogram.decode(encoded_histogram, self.b64_wrap) self.add(other_hist) @staticmethod def decode(encoded_histogram, b64_wrap=True): '''Decode an encoded histogram and return a new histogram instance that has been initialized with the decoded content Return: a new histogram instance representing the decoded content Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' hdr_payload = HdrHistogramEncoder.decode(encoded_histogram, b64_wrap) payload = hdr_payload.payload histogram = HdrHistogram(payload.lowest_trackable_value, payload.highest_trackable_value, payload.significant_figures, hdr_payload=hdr_payload) return histogram def get_word_size(self): return self.word_size def output_percentile_distribution(self, out_file, output_value_unit_scaling_ratio, ticks_per_half_distance=5): out_file.write(b'%12s %14s %10s %14s\n\n' % (b'Value', b'Percentile', b'TotalCount', b'1/(1-Percentile)')) percentile_format = '%12.{}f %2.12f %10d %14.2f\n'.format(self.significant_figures) last_line_percentile_format = '%12.{}f %2.12f %10d\n'.format(self.significant_figures) for iter_value in self.get_percentile_iterator(ticks_per_half_distance): value = iter_value.value_iterated_to / output_value_unit_scaling_ratio percentile = iter_value.percentile_level_iterated_to / 100 total_count = iter_value.total_count_to_this_value if iter_value.percentile_level_iterated_to != 100: other = 1 / (1 - iter_value.percentile_level_iterated_to / 100) out_file.write(percentile_format.encode() % (value, percentile, total_count, other)) else: out_file.write(last_line_percentile_format.encode() % (value, percentile, total_count)) mean = self.get_mean_value() / output_value_unit_scaling_ratio stddev = self.get_stddev() out_file.write('#[Mean = %12.{0}f, StdDeviation = %12.{0}f]\n'. format(self.significant_figures).encode() % (mean, stddev)) max = self.get_max_value() / output_value_unit_scaling_ratio total = self.get_total_count() out_file.write('#[Max = %12.{0}f, TotalCount = %12.{0}f]\n'.format( self.significant_figures).encode() % (max, total)) out_file.write(b'#[Buckets = %12d, SubBuckets = %12d]\n' % ( self.bucket_count, self.sub_bucket_count))
HdrHistogram/HdrHistogram_py
hdrh/histogram.py
HdrHistogram.set_internal_tacking_values
python
def set_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: self.max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) if min_non_zero_index >= 0: self.min_value = self.get_value_from_index(min_non_zero_index) self.total_count = total_added
Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none)
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/histogram.py#L443-L458
[ "def get_value_from_index(self, index):\n bucket_index = (index >> self.sub_bucket_half_count_magnitude) - 1\n sub_bucket_index = (index & (self.sub_bucket_half_count - 1)) + \\\n self.sub_bucket_half_count\n if bucket_index < 0:\n sub_bucket_index -= self.sub_bucket_half_count\n bucket_index = 0\n return self.get_value_from_sub_bucket(bucket_index, sub_bucket_index)\n", "def get_highest_equivalent_value(self, value):\n bucket_index = self._get_bucket_index(value)\n sub_bucket_index = self._get_sub_bucket_index(value, bucket_index)\n\n lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index,\n sub_bucket_index)\n if sub_bucket_index >= self.sub_bucket_count:\n bucket_index += 1\n size_of_equivalent_value_range = 1 << (self.unit_magnitude + bucket_index)\n next_non_equivalent_value = lowest_equivalent_value + size_of_equivalent_value_range\n\n return next_non_equivalent_value - 1\n" ]
class HdrHistogram(): '''This class supports the recording and analyzing of sampled data value counts across a configurable integer value range with configurable value precision within the range. Value precision is expressed as the number of significant digits in the value recording, and provides control over value quantization behavior across the value range and the subsequent value resolution at any given level. For example, a Histogram could be configured to track the counts of observed integer values between 0 and 3,600,000,000 while maintaining a value precision of 3 significant digits across that range. Value quantization within the range will thus be no larger than 1/1,000th (or 0.1%) of any value. This example Histogram could be used to track and analyze the counts of observed response times ranging between 1 microsecond and 1 hour in magnitude, while maintaining a value resolution of 1 microsecond up to 1 millisecond, a resolution of 1 millisecond (or better) up to one second, and a resolution of 1 second (or better) up to 1,000 seconds. At it's maximum tracked value (1 hour), it would still maintain a resolution of 3.6 seconds (or better). ''' def __init__(self, lowest_trackable_value, highest_trackable_value, significant_figures, word_size=8, b64_wrap=True, hdr_payload=None): '''Create a new histogram with given arguments Params: lowest_trackable_value The lowest value that can be discerned (distinguished from 0) by the histogram. Must be a positive integer that is >= 1. May be internally rounded down to nearest power of 2. highest_trackable_value The highest value to be tracked by the histogram. Must be a positive integer that is >= (2 * lowest_trackable_value). significant_figures The number of significant decimal digits to which the histogram will maintain value resolution and separation. Must be a non-negative integer between 0 and 5. word_size size of counters in bytes, only 2, 4, 8-byte counters are supported (default is 8-byte or 64-bit counters) b64_wrap specifies if the encoding of this histogram should use base64 wrapping (only useful if you need to encode the histogram to save somewhere or send over the wire. By default base64 encoding is assumed hdr_payload only used for associating an existing payload created from decoding an encoded histograme Exceptions: ValueError if the word_size value is unsupported if significant_figures is invalid ''' if significant_figures < 1 or significant_figures > 5: raise ValueError('Invalid significant_figures') self.lowest_trackable_value = lowest_trackable_value self.highest_trackable_value = highest_trackable_value self.significant_figures = significant_figures self.unit_magnitude = int(math.floor(math.log(lowest_trackable_value) / math.log(2))) largest_value_single_unit_res = 2 * math.pow(10, significant_figures) subb_count_mag = int(math.ceil(math.log(largest_value_single_unit_res) / math.log(2))) self.sub_bucket_half_count_magnitude = subb_count_mag - 1 if subb_count_mag > 1 else 0 self.sub_bucket_count = int(math.pow(2, self.sub_bucket_half_count_magnitude + 1)) self.sub_bucket_half_count = self.sub_bucket_count // 2 self.sub_bucket_mask = (self.sub_bucket_count - 1) << self.unit_magnitude self.bucket_count = get_bucket_count(highest_trackable_value, self.sub_bucket_count, self.unit_magnitude) self.min_value = sys.maxsize self.max_value = 0 self.total_count = 0 self.counts_len = (self.bucket_count + 1) * (self.sub_bucket_count // 2) self.word_size = word_size if hdr_payload: payload = hdr_payload.payload self.int_to_double_conversion_ratio = payload.conversion_ratio_bits results = hdr_payload.init_counts(self.counts_len) if results['total']: self.set_internal_tacking_values(results['min_nonzero_index'], results['max_nonzero_index'], results['total']) else: self.int_to_double_conversion_ratio = 1.0 # to encode this histogram into a compressed/base64 format ready # to be exported self.b64_wrap = b64_wrap self.encoder = HdrHistogramEncoder(self, b64_wrap, hdr_payload) # the counters reside directly in the payload object # allocated by the encoder # so that compression for wire transfer can be done without copy self.counts = self.encoder.get_counts() self.start_time_stamp_msec = 0 self.end_time_stamp_msec = 0 def _clz(self, value): """calculate the leading zeros, equivalent to C __builtin_clzll() value in hex: value = 1 clz = 63 value = 2 clz = 62 value = 4 clz = 61 value = 1000 clz = 51 value = 1000000 clz = 39 """ return 63 - (len(bin(value)) - 3) def _get_bucket_index(self, value): # smallest power of 2 containing value pow2ceiling = 64 - self._clz(int(value) | self.sub_bucket_mask) return int(pow2ceiling - self.unit_magnitude - (self.sub_bucket_half_count_magnitude + 1)) def _get_sub_bucket_index(self, value, bucket_index): return int(value) >> (bucket_index + self.unit_magnitude) def _counts_index(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ): bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - subBucketHalfCount) + bucketBaseIndex return bucket_base_index + offset_in_bucket def _counts_index_for(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) return self._counts_index(bucket_index, sub_bucket_index) def record_value(self, value, count=1): '''Record a new value into the histogram Args: value: the value to record (must be in the valid range) count: incremental count (defaults to 1) ''' if value < 0: return False counts_index = self._counts_index_for(value) if (counts_index < 0) or (self.counts_len <= counts_index): return False self.counts[counts_index] += count self.total_count += count self.min_value = min(self.min_value, value) self.max_value = max(self.max_value, value) return True # pylint: disable=inconsistent-return-statements def record_corrected_value(self, value, expected_interval, count=1): '''Record a new value into the histogram and correct for coordinated omission if needed Args: value: the value to record (must be in the valid range) expected_interval: the expected interval between 2 value samples count: incremental count (defaults to 1) ''' while True: if not self.record_value(value, count): return False if value <= expected_interval or expected_interval <= 0: return True value -= expected_interval def get_count_at_index(self, index): if index >= self.counts_len: raise IndexError() # some decoded (read-only) histograms may have truncated # counts arrays, we return zero for any index that is passed the array if index >= self.encoder.payload.counts_len: return 0 return self.counts[index] def get_count_at_sub_bucket(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ) bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # (sub_bucket_index - subBucketHalfCount) + bucketBaseIndex counts_index = bucket_base_index + offset_in_bucket return self.counts[counts_index] def get_value_from_sub_bucket(self, bucket_index, sub_bucket_index): return sub_bucket_index << (bucket_index + self.unit_magnitude) def get_value_from_index(self, index): bucket_index = (index >> self.sub_bucket_half_count_magnitude) - 1 sub_bucket_index = (index & (self.sub_bucket_half_count - 1)) + \ self.sub_bucket_half_count if bucket_index < 0: sub_bucket_index -= self.sub_bucket_half_count bucket_index = 0 return self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) def get_lowest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) return lowest_equivalent_value def get_highest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 size_of_equivalent_value_range = 1 << (self.unit_magnitude + bucket_index) next_non_equivalent_value = lowest_equivalent_value + size_of_equivalent_value_range return next_non_equivalent_value - 1 def get_target_count_at_percentile(self, percentile): requested_percentile = min(percentile, 100.0) count_at_percentile = int(((requested_percentile * self.total_count / 100)) + 0.5) return max(count_at_percentile, 1) def get_value_at_percentile(self, percentile): '''Get the value for a given percentile Args: percentile: a float in [0.0..100.0] Returns: the value for the given percentile ''' count_at_percentile = self.get_target_count_at_percentile(percentile) total = 0 for index in range(self.counts_len): total += self.get_count_at_index(index) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: return self.get_highest_equivalent_value(value_at_index) return self.get_lowest_equivalent_value(value_at_index) return 0 def get_percentile_to_value_dict(self, percentile_list): '''A faster alternative to query values for a list of percentiles. Args: percentile_list: a list of percentiles in any order, dups will be ignored each element in the list must be a float value in [0.0 .. 100.0] Returns: a dict of percentile values indexed by the percentile ''' result = {} total = 0 percentile_list_index = 0 count_at_percentile = 0 # remove dups and sort percentile_list = list(set(percentile_list)) percentile_list.sort() for index in range(self.counts_len): total += self.get_count_at_index(index) while True: # recalculate target based on next requested percentile if not count_at_percentile: if percentile_list_index == len(percentile_list): return result percentile = percentile_list[percentile_list_index] percentile_list_index += 1 if percentile > 100: return result count_at_percentile = self.get_target_count_at_percentile(percentile) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: result[percentile] = self.get_highest_equivalent_value(value_at_index) else: result[percentile] = self.get_lowest_equivalent_value(value_at_index) count_at_percentile = 0 else: break return result def get_total_count(self): return self.total_count def get_count_at_value(self, value): counts_index = self._counts_index_for(value) return self.counts[counts_index] def values_are_equivalent(self, val1, val2): '''Check whether 2 values are equivalent (meaning they are in the same bucket/range) Returns: true if the 2 values are equivalent ''' return self.get_lowest_equivalent_value(val1) == self.get_lowest_equivalent_value(val2) def get_max_value(self): if self.max_value == 0: return 0 return self.get_highest_equivalent_value(self.max_value) def get_min_value(self): if self.counts[0] > 0 or self.total_count == 0: return 0 if sys.maxsize == self.min_value: return sys.maxsize return self.get_lowest_equivalent_value(self.min_value) def _hdr_size_of_equiv_value_range(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 return 1 << (self.unit_magnitude + bucket_index) def _hdr_median_equiv_value(self, value): return self.get_lowest_equivalent_value(value) + \ (self._hdr_size_of_equiv_value_range(value) >> 1) def get_mean_value(self): if not self.total_count: return 0.0 total = 0 itr = self.get_recorded_iterator() for item in itr: total += itr.count_at_this_value * self._hdr_median_equiv_value(item.value_iterated_to) return float(total) / self.total_count def get_stddev(self): if not self.total_count: return 0.0 mean = self.get_mean_value() geometric_dev_total = 0.0 for item in self.get_recorded_iterator(): dev = (self._hdr_median_equiv_value(item.value_iterated_to) * 1.0) - mean geometric_dev_total += (dev * dev) * item.count_added_in_this_iter_step return math.sqrt(geometric_dev_total / self.total_count) def reset(self): '''Reset the histogram to a pristine state ''' for index in range(self.counts_len): self.counts[index] = 0 self.total_count = 0 self.min_value = sys.maxsize self.max_value = 0 self.start_time_stamp_msec = sys.maxsize self.end_time_stamp_msec = 0 def __iter__(self): '''Returns the recorded iterator if iter(self) is called ''' return RecordedIterator(self) def get_all_values_iterator(self): return AllValuesIterator(self) def get_recorded_iterator(self): return RecordedIterator(self) def get_percentile_iterator(self, ticks_per_half_distance): return PercentileIterator(self, ticks_per_half_distance) def get_linear_iterator(self, value_units_per_bucket): return LinearIterator(self, value_units_per_bucket) def get_log_iterator(self, value_units_first_bucket, log_base): return LogIterator(self, value_units_first_bucket, log_base) def encode(self): '''Encode this histogram Return: a string containing the base64 encoded compressed histogram (V1 format) ''' return self.encoder.encode() def adjust_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) self.max_value = max(self.max_value, max_value) if min_non_zero_index >= 0: min_value = self.get_value_from_index(min_non_zero_index) self.min_value = min(self.min_value, min_value) self.total_count += total_added def get_counts_array_index(self, value): '''Return the index in the counts array for a given value ''' if value < 0: raise ValueError("Histogram recorded value cannot be negative.") bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) # Calculate the index for the first entry in the bucket: bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # The following is the equivalent of ((bucket_index + 1) * sub_bucket_half_count) # Calculate the offset in the bucket (can be negative for first bucket): offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - sub_bucket_half_count) + bucket_base_index return bucket_base_index + offset_in_bucket def get_start_time_stamp(self): return self.start_time_stamp_msec def set_start_time_stamp(self, time_stamp_msec): '''Set the start time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.start_time_stamp_msec = time_stamp_msec def get_end_time_stamp(self): return self.end_time_stamp_msec def set_end_time_stamp(self, time_stamp_msec): '''Set the end time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.end_time_stamp_msec = time_stamp_msec def add(self, other_hist): highest_recordable_value = \ self.get_highest_equivalent_value(self.get_value_from_index(self.counts_len - 1)) if highest_recordable_value < other_hist.get_max_value(): raise IndexError("The other histogram includes values that do not fit %d < %d" % (highest_recordable_value, other_hist.get_max_value())) if (self.bucket_count == other_hist.bucket_count) and \ (self.sub_bucket_count == other_hist.sub_bucket_count) and \ (self.unit_magnitude == other_hist.unit_magnitude) and \ (self.word_size == other_hist.word_size): # do an in-place addition of one array to another self.encoder.add(other_hist.encoder) self.total_count += other_hist.get_total_count() self.max_value = max(self.max_value, other_hist.get_max_value()) self.min_value = min(self.get_min_value(), other_hist.get_min_value()) else: # Arrays are not a direct match, so we can't just stream through and add them. # Instead, go through the array and add each non-zero value found at it's proper value: for index in range(other_hist.counts_len): other_count = other_hist.get_count_at_index(index) if other_count > 0: self.record_value(other_hist.get_value_from_index(index), other_count) self.start_time_stamp_msec = \ min(self.start_time_stamp_msec, other_hist.start_time_stamp_msec) self.end_time_stamp_msec = \ max(self.end_time_stamp_msec, other_hist.end_time_stamp_msec) def decode_and_add(self, encoded_histogram): '''Decode an encoded histogram and add it to this histogram Args: encoded_histogram (string) an encoded histogram following the V1 format, such as one returned by the encode() method Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' other_hist = HdrHistogram.decode(encoded_histogram, self.b64_wrap) self.add(other_hist) @staticmethod def decode(encoded_histogram, b64_wrap=True): '''Decode an encoded histogram and return a new histogram instance that has been initialized with the decoded content Return: a new histogram instance representing the decoded content Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' hdr_payload = HdrHistogramEncoder.decode(encoded_histogram, b64_wrap) payload = hdr_payload.payload histogram = HdrHistogram(payload.lowest_trackable_value, payload.highest_trackable_value, payload.significant_figures, hdr_payload=hdr_payload) return histogram def get_word_size(self): return self.word_size def output_percentile_distribution(self, out_file, output_value_unit_scaling_ratio, ticks_per_half_distance=5): out_file.write(b'%12s %14s %10s %14s\n\n' % (b'Value', b'Percentile', b'TotalCount', b'1/(1-Percentile)')) percentile_format = '%12.{}f %2.12f %10d %14.2f\n'.format(self.significant_figures) last_line_percentile_format = '%12.{}f %2.12f %10d\n'.format(self.significant_figures) for iter_value in self.get_percentile_iterator(ticks_per_half_distance): value = iter_value.value_iterated_to / output_value_unit_scaling_ratio percentile = iter_value.percentile_level_iterated_to / 100 total_count = iter_value.total_count_to_this_value if iter_value.percentile_level_iterated_to != 100: other = 1 / (1 - iter_value.percentile_level_iterated_to / 100) out_file.write(percentile_format.encode() % (value, percentile, total_count, other)) else: out_file.write(last_line_percentile_format.encode() % (value, percentile, total_count)) mean = self.get_mean_value() / output_value_unit_scaling_ratio stddev = self.get_stddev() out_file.write('#[Mean = %12.{0}f, StdDeviation = %12.{0}f]\n'. format(self.significant_figures).encode() % (mean, stddev)) max = self.get_max_value() / output_value_unit_scaling_ratio total = self.get_total_count() out_file.write('#[Max = %12.{0}f, TotalCount = %12.{0}f]\n'.format( self.significant_figures).encode() % (max, total)) out_file.write(b'#[Buckets = %12d, SubBuckets = %12d]\n' % ( self.bucket_count, self.sub_bucket_count))
HdrHistogram/HdrHistogram_py
hdrh/histogram.py
HdrHistogram.get_counts_array_index
python
def get_counts_array_index(self, value): '''Return the index in the counts array for a given value ''' if value < 0: raise ValueError("Histogram recorded value cannot be negative.") bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) # Calculate the index for the first entry in the bucket: bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # The following is the equivalent of ((bucket_index + 1) * sub_bucket_half_count) # Calculate the offset in the bucket (can be negative for first bucket): offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - sub_bucket_half_count) + bucket_base_index return bucket_base_index + offset_in_bucket
Return the index in the counts array for a given value
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/histogram.py#L460-L475
[ "def _get_bucket_index(self, value):\n # smallest power of 2 containing value\n pow2ceiling = 64 - self._clz(int(value) | self.sub_bucket_mask)\n return int(pow2ceiling - self.unit_magnitude -\n (self.sub_bucket_half_count_magnitude + 1))\n", "def _get_sub_bucket_index(self, value, bucket_index):\n return int(value) >> (bucket_index + self.unit_magnitude)\n" ]
class HdrHistogram(): '''This class supports the recording and analyzing of sampled data value counts across a configurable integer value range with configurable value precision within the range. Value precision is expressed as the number of significant digits in the value recording, and provides control over value quantization behavior across the value range and the subsequent value resolution at any given level. For example, a Histogram could be configured to track the counts of observed integer values between 0 and 3,600,000,000 while maintaining a value precision of 3 significant digits across that range. Value quantization within the range will thus be no larger than 1/1,000th (or 0.1%) of any value. This example Histogram could be used to track and analyze the counts of observed response times ranging between 1 microsecond and 1 hour in magnitude, while maintaining a value resolution of 1 microsecond up to 1 millisecond, a resolution of 1 millisecond (or better) up to one second, and a resolution of 1 second (or better) up to 1,000 seconds. At it's maximum tracked value (1 hour), it would still maintain a resolution of 3.6 seconds (or better). ''' def __init__(self, lowest_trackable_value, highest_trackable_value, significant_figures, word_size=8, b64_wrap=True, hdr_payload=None): '''Create a new histogram with given arguments Params: lowest_trackable_value The lowest value that can be discerned (distinguished from 0) by the histogram. Must be a positive integer that is >= 1. May be internally rounded down to nearest power of 2. highest_trackable_value The highest value to be tracked by the histogram. Must be a positive integer that is >= (2 * lowest_trackable_value). significant_figures The number of significant decimal digits to which the histogram will maintain value resolution and separation. Must be a non-negative integer between 0 and 5. word_size size of counters in bytes, only 2, 4, 8-byte counters are supported (default is 8-byte or 64-bit counters) b64_wrap specifies if the encoding of this histogram should use base64 wrapping (only useful if you need to encode the histogram to save somewhere or send over the wire. By default base64 encoding is assumed hdr_payload only used for associating an existing payload created from decoding an encoded histograme Exceptions: ValueError if the word_size value is unsupported if significant_figures is invalid ''' if significant_figures < 1 or significant_figures > 5: raise ValueError('Invalid significant_figures') self.lowest_trackable_value = lowest_trackable_value self.highest_trackable_value = highest_trackable_value self.significant_figures = significant_figures self.unit_magnitude = int(math.floor(math.log(lowest_trackable_value) / math.log(2))) largest_value_single_unit_res = 2 * math.pow(10, significant_figures) subb_count_mag = int(math.ceil(math.log(largest_value_single_unit_res) / math.log(2))) self.sub_bucket_half_count_magnitude = subb_count_mag - 1 if subb_count_mag > 1 else 0 self.sub_bucket_count = int(math.pow(2, self.sub_bucket_half_count_magnitude + 1)) self.sub_bucket_half_count = self.sub_bucket_count // 2 self.sub_bucket_mask = (self.sub_bucket_count - 1) << self.unit_magnitude self.bucket_count = get_bucket_count(highest_trackable_value, self.sub_bucket_count, self.unit_magnitude) self.min_value = sys.maxsize self.max_value = 0 self.total_count = 0 self.counts_len = (self.bucket_count + 1) * (self.sub_bucket_count // 2) self.word_size = word_size if hdr_payload: payload = hdr_payload.payload self.int_to_double_conversion_ratio = payload.conversion_ratio_bits results = hdr_payload.init_counts(self.counts_len) if results['total']: self.set_internal_tacking_values(results['min_nonzero_index'], results['max_nonzero_index'], results['total']) else: self.int_to_double_conversion_ratio = 1.0 # to encode this histogram into a compressed/base64 format ready # to be exported self.b64_wrap = b64_wrap self.encoder = HdrHistogramEncoder(self, b64_wrap, hdr_payload) # the counters reside directly in the payload object # allocated by the encoder # so that compression for wire transfer can be done without copy self.counts = self.encoder.get_counts() self.start_time_stamp_msec = 0 self.end_time_stamp_msec = 0 def _clz(self, value): """calculate the leading zeros, equivalent to C __builtin_clzll() value in hex: value = 1 clz = 63 value = 2 clz = 62 value = 4 clz = 61 value = 1000 clz = 51 value = 1000000 clz = 39 """ return 63 - (len(bin(value)) - 3) def _get_bucket_index(self, value): # smallest power of 2 containing value pow2ceiling = 64 - self._clz(int(value) | self.sub_bucket_mask) return int(pow2ceiling - self.unit_magnitude - (self.sub_bucket_half_count_magnitude + 1)) def _get_sub_bucket_index(self, value, bucket_index): return int(value) >> (bucket_index + self.unit_magnitude) def _counts_index(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ): bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - subBucketHalfCount) + bucketBaseIndex return bucket_base_index + offset_in_bucket def _counts_index_for(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) return self._counts_index(bucket_index, sub_bucket_index) def record_value(self, value, count=1): '''Record a new value into the histogram Args: value: the value to record (must be in the valid range) count: incremental count (defaults to 1) ''' if value < 0: return False counts_index = self._counts_index_for(value) if (counts_index < 0) or (self.counts_len <= counts_index): return False self.counts[counts_index] += count self.total_count += count self.min_value = min(self.min_value, value) self.max_value = max(self.max_value, value) return True # pylint: disable=inconsistent-return-statements def record_corrected_value(self, value, expected_interval, count=1): '''Record a new value into the histogram and correct for coordinated omission if needed Args: value: the value to record (must be in the valid range) expected_interval: the expected interval between 2 value samples count: incremental count (defaults to 1) ''' while True: if not self.record_value(value, count): return False if value <= expected_interval or expected_interval <= 0: return True value -= expected_interval def get_count_at_index(self, index): if index >= self.counts_len: raise IndexError() # some decoded (read-only) histograms may have truncated # counts arrays, we return zero for any index that is passed the array if index >= self.encoder.payload.counts_len: return 0 return self.counts[index] def get_count_at_sub_bucket(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ) bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # (sub_bucket_index - subBucketHalfCount) + bucketBaseIndex counts_index = bucket_base_index + offset_in_bucket return self.counts[counts_index] def get_value_from_sub_bucket(self, bucket_index, sub_bucket_index): return sub_bucket_index << (bucket_index + self.unit_magnitude) def get_value_from_index(self, index): bucket_index = (index >> self.sub_bucket_half_count_magnitude) - 1 sub_bucket_index = (index & (self.sub_bucket_half_count - 1)) + \ self.sub_bucket_half_count if bucket_index < 0: sub_bucket_index -= self.sub_bucket_half_count bucket_index = 0 return self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) def get_lowest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) return lowest_equivalent_value def get_highest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 size_of_equivalent_value_range = 1 << (self.unit_magnitude + bucket_index) next_non_equivalent_value = lowest_equivalent_value + size_of_equivalent_value_range return next_non_equivalent_value - 1 def get_target_count_at_percentile(self, percentile): requested_percentile = min(percentile, 100.0) count_at_percentile = int(((requested_percentile * self.total_count / 100)) + 0.5) return max(count_at_percentile, 1) def get_value_at_percentile(self, percentile): '''Get the value for a given percentile Args: percentile: a float in [0.0..100.0] Returns: the value for the given percentile ''' count_at_percentile = self.get_target_count_at_percentile(percentile) total = 0 for index in range(self.counts_len): total += self.get_count_at_index(index) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: return self.get_highest_equivalent_value(value_at_index) return self.get_lowest_equivalent_value(value_at_index) return 0 def get_percentile_to_value_dict(self, percentile_list): '''A faster alternative to query values for a list of percentiles. Args: percentile_list: a list of percentiles in any order, dups will be ignored each element in the list must be a float value in [0.0 .. 100.0] Returns: a dict of percentile values indexed by the percentile ''' result = {} total = 0 percentile_list_index = 0 count_at_percentile = 0 # remove dups and sort percentile_list = list(set(percentile_list)) percentile_list.sort() for index in range(self.counts_len): total += self.get_count_at_index(index) while True: # recalculate target based on next requested percentile if not count_at_percentile: if percentile_list_index == len(percentile_list): return result percentile = percentile_list[percentile_list_index] percentile_list_index += 1 if percentile > 100: return result count_at_percentile = self.get_target_count_at_percentile(percentile) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: result[percentile] = self.get_highest_equivalent_value(value_at_index) else: result[percentile] = self.get_lowest_equivalent_value(value_at_index) count_at_percentile = 0 else: break return result def get_total_count(self): return self.total_count def get_count_at_value(self, value): counts_index = self._counts_index_for(value) return self.counts[counts_index] def values_are_equivalent(self, val1, val2): '''Check whether 2 values are equivalent (meaning they are in the same bucket/range) Returns: true if the 2 values are equivalent ''' return self.get_lowest_equivalent_value(val1) == self.get_lowest_equivalent_value(val2) def get_max_value(self): if self.max_value == 0: return 0 return self.get_highest_equivalent_value(self.max_value) def get_min_value(self): if self.counts[0] > 0 or self.total_count == 0: return 0 if sys.maxsize == self.min_value: return sys.maxsize return self.get_lowest_equivalent_value(self.min_value) def _hdr_size_of_equiv_value_range(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 return 1 << (self.unit_magnitude + bucket_index) def _hdr_median_equiv_value(self, value): return self.get_lowest_equivalent_value(value) + \ (self._hdr_size_of_equiv_value_range(value) >> 1) def get_mean_value(self): if not self.total_count: return 0.0 total = 0 itr = self.get_recorded_iterator() for item in itr: total += itr.count_at_this_value * self._hdr_median_equiv_value(item.value_iterated_to) return float(total) / self.total_count def get_stddev(self): if not self.total_count: return 0.0 mean = self.get_mean_value() geometric_dev_total = 0.0 for item in self.get_recorded_iterator(): dev = (self._hdr_median_equiv_value(item.value_iterated_to) * 1.0) - mean geometric_dev_total += (dev * dev) * item.count_added_in_this_iter_step return math.sqrt(geometric_dev_total / self.total_count) def reset(self): '''Reset the histogram to a pristine state ''' for index in range(self.counts_len): self.counts[index] = 0 self.total_count = 0 self.min_value = sys.maxsize self.max_value = 0 self.start_time_stamp_msec = sys.maxsize self.end_time_stamp_msec = 0 def __iter__(self): '''Returns the recorded iterator if iter(self) is called ''' return RecordedIterator(self) def get_all_values_iterator(self): return AllValuesIterator(self) def get_recorded_iterator(self): return RecordedIterator(self) def get_percentile_iterator(self, ticks_per_half_distance): return PercentileIterator(self, ticks_per_half_distance) def get_linear_iterator(self, value_units_per_bucket): return LinearIterator(self, value_units_per_bucket) def get_log_iterator(self, value_units_first_bucket, log_base): return LogIterator(self, value_units_first_bucket, log_base) def encode(self): '''Encode this histogram Return: a string containing the base64 encoded compressed histogram (V1 format) ''' return self.encoder.encode() def adjust_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) self.max_value = max(self.max_value, max_value) if min_non_zero_index >= 0: min_value = self.get_value_from_index(min_non_zero_index) self.min_value = min(self.min_value, min_value) self.total_count += total_added def set_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: self.max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) if min_non_zero_index >= 0: self.min_value = self.get_value_from_index(min_non_zero_index) self.total_count = total_added def get_start_time_stamp(self): return self.start_time_stamp_msec def set_start_time_stamp(self, time_stamp_msec): '''Set the start time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.start_time_stamp_msec = time_stamp_msec def get_end_time_stamp(self): return self.end_time_stamp_msec def set_end_time_stamp(self, time_stamp_msec): '''Set the end time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.end_time_stamp_msec = time_stamp_msec def add(self, other_hist): highest_recordable_value = \ self.get_highest_equivalent_value(self.get_value_from_index(self.counts_len - 1)) if highest_recordable_value < other_hist.get_max_value(): raise IndexError("The other histogram includes values that do not fit %d < %d" % (highest_recordable_value, other_hist.get_max_value())) if (self.bucket_count == other_hist.bucket_count) and \ (self.sub_bucket_count == other_hist.sub_bucket_count) and \ (self.unit_magnitude == other_hist.unit_magnitude) and \ (self.word_size == other_hist.word_size): # do an in-place addition of one array to another self.encoder.add(other_hist.encoder) self.total_count += other_hist.get_total_count() self.max_value = max(self.max_value, other_hist.get_max_value()) self.min_value = min(self.get_min_value(), other_hist.get_min_value()) else: # Arrays are not a direct match, so we can't just stream through and add them. # Instead, go through the array and add each non-zero value found at it's proper value: for index in range(other_hist.counts_len): other_count = other_hist.get_count_at_index(index) if other_count > 0: self.record_value(other_hist.get_value_from_index(index), other_count) self.start_time_stamp_msec = \ min(self.start_time_stamp_msec, other_hist.start_time_stamp_msec) self.end_time_stamp_msec = \ max(self.end_time_stamp_msec, other_hist.end_time_stamp_msec) def decode_and_add(self, encoded_histogram): '''Decode an encoded histogram and add it to this histogram Args: encoded_histogram (string) an encoded histogram following the V1 format, such as one returned by the encode() method Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' other_hist = HdrHistogram.decode(encoded_histogram, self.b64_wrap) self.add(other_hist) @staticmethod def decode(encoded_histogram, b64_wrap=True): '''Decode an encoded histogram and return a new histogram instance that has been initialized with the decoded content Return: a new histogram instance representing the decoded content Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' hdr_payload = HdrHistogramEncoder.decode(encoded_histogram, b64_wrap) payload = hdr_payload.payload histogram = HdrHistogram(payload.lowest_trackable_value, payload.highest_trackable_value, payload.significant_figures, hdr_payload=hdr_payload) return histogram def get_word_size(self): return self.word_size def output_percentile_distribution(self, out_file, output_value_unit_scaling_ratio, ticks_per_half_distance=5): out_file.write(b'%12s %14s %10s %14s\n\n' % (b'Value', b'Percentile', b'TotalCount', b'1/(1-Percentile)')) percentile_format = '%12.{}f %2.12f %10d %14.2f\n'.format(self.significant_figures) last_line_percentile_format = '%12.{}f %2.12f %10d\n'.format(self.significant_figures) for iter_value in self.get_percentile_iterator(ticks_per_half_distance): value = iter_value.value_iterated_to / output_value_unit_scaling_ratio percentile = iter_value.percentile_level_iterated_to / 100 total_count = iter_value.total_count_to_this_value if iter_value.percentile_level_iterated_to != 100: other = 1 / (1 - iter_value.percentile_level_iterated_to / 100) out_file.write(percentile_format.encode() % (value, percentile, total_count, other)) else: out_file.write(last_line_percentile_format.encode() % (value, percentile, total_count)) mean = self.get_mean_value() / output_value_unit_scaling_ratio stddev = self.get_stddev() out_file.write('#[Mean = %12.{0}f, StdDeviation = %12.{0}f]\n'. format(self.significant_figures).encode() % (mean, stddev)) max = self.get_max_value() / output_value_unit_scaling_ratio total = self.get_total_count() out_file.write('#[Max = %12.{0}f, TotalCount = %12.{0}f]\n'.format( self.significant_figures).encode() % (max, total)) out_file.write(b'#[Buckets = %12d, SubBuckets = %12d]\n' % ( self.bucket_count, self.sub_bucket_count))
HdrHistogram/HdrHistogram_py
hdrh/histogram.py
HdrHistogram.decode_and_add
python
def decode_and_add(self, encoded_histogram): '''Decode an encoded histogram and add it to this histogram Args: encoded_histogram (string) an encoded histogram following the V1 format, such as one returned by the encode() method Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' other_hist = HdrHistogram.decode(encoded_histogram, self.b64_wrap) self.add(other_hist)
Decode an encoded histogram and add it to this histogram Args: encoded_histogram (string) an encoded histogram following the V1 format, such as one returned by the encode() method Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/histogram.py#L530-L547
[ "def add(self, other_hist):\n highest_recordable_value = \\\n self.get_highest_equivalent_value(self.get_value_from_index(self.counts_len - 1))\n if highest_recordable_value < other_hist.get_max_value():\n raise IndexError(\"The other histogram includes values that do not fit %d < %d\" %\n (highest_recordable_value, other_hist.get_max_value()))\n\n if (self.bucket_count == other_hist.bucket_count) and \\\n (self.sub_bucket_count == other_hist.sub_bucket_count) and \\\n (self.unit_magnitude == other_hist.unit_magnitude) and \\\n (self.word_size == other_hist.word_size):\n\n # do an in-place addition of one array to another\n self.encoder.add(other_hist.encoder)\n\n self.total_count += other_hist.get_total_count()\n self.max_value = max(self.max_value, other_hist.get_max_value())\n self.min_value = min(self.get_min_value(), other_hist.get_min_value())\n else:\n # Arrays are not a direct match, so we can't just stream through and add them.\n # Instead, go through the array and add each non-zero value found at it's proper value:\n for index in range(other_hist.counts_len):\n other_count = other_hist.get_count_at_index(index)\n if other_count > 0:\n self.record_value(other_hist.get_value_from_index(index), other_count)\n\n self.start_time_stamp_msec = \\\n min(self.start_time_stamp_msec, other_hist.start_time_stamp_msec)\n self.end_time_stamp_msec = \\\n max(self.end_time_stamp_msec, other_hist.end_time_stamp_msec)\n", "def decode(encoded_histogram, b64_wrap=True):\n '''Decode an encoded histogram and return a new histogram instance that\n has been initialized with the decoded content\n Return:\n a new histogram instance representing the decoded content\n Exception:\n TypeError in case of base64 decode error\n HdrCookieException:\n the main header has an invalid cookie\n the compressed payload header has an invalid cookie\n HdrLengthException:\n the decompressed size is too small for the HdrPayload structure\n or is not aligned or is too large for the passed payload class\n zlib.error:\n in case of zlib decompression error\n '''\n hdr_payload = HdrHistogramEncoder.decode(encoded_histogram, b64_wrap)\n payload = hdr_payload.payload\n histogram = HdrHistogram(payload.lowest_trackable_value,\n payload.highest_trackable_value,\n payload.significant_figures,\n hdr_payload=hdr_payload)\n return histogram\n" ]
class HdrHistogram(): '''This class supports the recording and analyzing of sampled data value counts across a configurable integer value range with configurable value precision within the range. Value precision is expressed as the number of significant digits in the value recording, and provides control over value quantization behavior across the value range and the subsequent value resolution at any given level. For example, a Histogram could be configured to track the counts of observed integer values between 0 and 3,600,000,000 while maintaining a value precision of 3 significant digits across that range. Value quantization within the range will thus be no larger than 1/1,000th (or 0.1%) of any value. This example Histogram could be used to track and analyze the counts of observed response times ranging between 1 microsecond and 1 hour in magnitude, while maintaining a value resolution of 1 microsecond up to 1 millisecond, a resolution of 1 millisecond (or better) up to one second, and a resolution of 1 second (or better) up to 1,000 seconds. At it's maximum tracked value (1 hour), it would still maintain a resolution of 3.6 seconds (or better). ''' def __init__(self, lowest_trackable_value, highest_trackable_value, significant_figures, word_size=8, b64_wrap=True, hdr_payload=None): '''Create a new histogram with given arguments Params: lowest_trackable_value The lowest value that can be discerned (distinguished from 0) by the histogram. Must be a positive integer that is >= 1. May be internally rounded down to nearest power of 2. highest_trackable_value The highest value to be tracked by the histogram. Must be a positive integer that is >= (2 * lowest_trackable_value). significant_figures The number of significant decimal digits to which the histogram will maintain value resolution and separation. Must be a non-negative integer between 0 and 5. word_size size of counters in bytes, only 2, 4, 8-byte counters are supported (default is 8-byte or 64-bit counters) b64_wrap specifies if the encoding of this histogram should use base64 wrapping (only useful if you need to encode the histogram to save somewhere or send over the wire. By default base64 encoding is assumed hdr_payload only used for associating an existing payload created from decoding an encoded histograme Exceptions: ValueError if the word_size value is unsupported if significant_figures is invalid ''' if significant_figures < 1 or significant_figures > 5: raise ValueError('Invalid significant_figures') self.lowest_trackable_value = lowest_trackable_value self.highest_trackable_value = highest_trackable_value self.significant_figures = significant_figures self.unit_magnitude = int(math.floor(math.log(lowest_trackable_value) / math.log(2))) largest_value_single_unit_res = 2 * math.pow(10, significant_figures) subb_count_mag = int(math.ceil(math.log(largest_value_single_unit_res) / math.log(2))) self.sub_bucket_half_count_magnitude = subb_count_mag - 1 if subb_count_mag > 1 else 0 self.sub_bucket_count = int(math.pow(2, self.sub_bucket_half_count_magnitude + 1)) self.sub_bucket_half_count = self.sub_bucket_count // 2 self.sub_bucket_mask = (self.sub_bucket_count - 1) << self.unit_magnitude self.bucket_count = get_bucket_count(highest_trackable_value, self.sub_bucket_count, self.unit_magnitude) self.min_value = sys.maxsize self.max_value = 0 self.total_count = 0 self.counts_len = (self.bucket_count + 1) * (self.sub_bucket_count // 2) self.word_size = word_size if hdr_payload: payload = hdr_payload.payload self.int_to_double_conversion_ratio = payload.conversion_ratio_bits results = hdr_payload.init_counts(self.counts_len) if results['total']: self.set_internal_tacking_values(results['min_nonzero_index'], results['max_nonzero_index'], results['total']) else: self.int_to_double_conversion_ratio = 1.0 # to encode this histogram into a compressed/base64 format ready # to be exported self.b64_wrap = b64_wrap self.encoder = HdrHistogramEncoder(self, b64_wrap, hdr_payload) # the counters reside directly in the payload object # allocated by the encoder # so that compression for wire transfer can be done without copy self.counts = self.encoder.get_counts() self.start_time_stamp_msec = 0 self.end_time_stamp_msec = 0 def _clz(self, value): """calculate the leading zeros, equivalent to C __builtin_clzll() value in hex: value = 1 clz = 63 value = 2 clz = 62 value = 4 clz = 61 value = 1000 clz = 51 value = 1000000 clz = 39 """ return 63 - (len(bin(value)) - 3) def _get_bucket_index(self, value): # smallest power of 2 containing value pow2ceiling = 64 - self._clz(int(value) | self.sub_bucket_mask) return int(pow2ceiling - self.unit_magnitude - (self.sub_bucket_half_count_magnitude + 1)) def _get_sub_bucket_index(self, value, bucket_index): return int(value) >> (bucket_index + self.unit_magnitude) def _counts_index(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ): bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - subBucketHalfCount) + bucketBaseIndex return bucket_base_index + offset_in_bucket def _counts_index_for(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) return self._counts_index(bucket_index, sub_bucket_index) def record_value(self, value, count=1): '''Record a new value into the histogram Args: value: the value to record (must be in the valid range) count: incremental count (defaults to 1) ''' if value < 0: return False counts_index = self._counts_index_for(value) if (counts_index < 0) or (self.counts_len <= counts_index): return False self.counts[counts_index] += count self.total_count += count self.min_value = min(self.min_value, value) self.max_value = max(self.max_value, value) return True # pylint: disable=inconsistent-return-statements def record_corrected_value(self, value, expected_interval, count=1): '''Record a new value into the histogram and correct for coordinated omission if needed Args: value: the value to record (must be in the valid range) expected_interval: the expected interval between 2 value samples count: incremental count (defaults to 1) ''' while True: if not self.record_value(value, count): return False if value <= expected_interval or expected_interval <= 0: return True value -= expected_interval def get_count_at_index(self, index): if index >= self.counts_len: raise IndexError() # some decoded (read-only) histograms may have truncated # counts arrays, we return zero for any index that is passed the array if index >= self.encoder.payload.counts_len: return 0 return self.counts[index] def get_count_at_sub_bucket(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ) bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # (sub_bucket_index - subBucketHalfCount) + bucketBaseIndex counts_index = bucket_base_index + offset_in_bucket return self.counts[counts_index] def get_value_from_sub_bucket(self, bucket_index, sub_bucket_index): return sub_bucket_index << (bucket_index + self.unit_magnitude) def get_value_from_index(self, index): bucket_index = (index >> self.sub_bucket_half_count_magnitude) - 1 sub_bucket_index = (index & (self.sub_bucket_half_count - 1)) + \ self.sub_bucket_half_count if bucket_index < 0: sub_bucket_index -= self.sub_bucket_half_count bucket_index = 0 return self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) def get_lowest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) return lowest_equivalent_value def get_highest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 size_of_equivalent_value_range = 1 << (self.unit_magnitude + bucket_index) next_non_equivalent_value = lowest_equivalent_value + size_of_equivalent_value_range return next_non_equivalent_value - 1 def get_target_count_at_percentile(self, percentile): requested_percentile = min(percentile, 100.0) count_at_percentile = int(((requested_percentile * self.total_count / 100)) + 0.5) return max(count_at_percentile, 1) def get_value_at_percentile(self, percentile): '''Get the value for a given percentile Args: percentile: a float in [0.0..100.0] Returns: the value for the given percentile ''' count_at_percentile = self.get_target_count_at_percentile(percentile) total = 0 for index in range(self.counts_len): total += self.get_count_at_index(index) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: return self.get_highest_equivalent_value(value_at_index) return self.get_lowest_equivalent_value(value_at_index) return 0 def get_percentile_to_value_dict(self, percentile_list): '''A faster alternative to query values for a list of percentiles. Args: percentile_list: a list of percentiles in any order, dups will be ignored each element in the list must be a float value in [0.0 .. 100.0] Returns: a dict of percentile values indexed by the percentile ''' result = {} total = 0 percentile_list_index = 0 count_at_percentile = 0 # remove dups and sort percentile_list = list(set(percentile_list)) percentile_list.sort() for index in range(self.counts_len): total += self.get_count_at_index(index) while True: # recalculate target based on next requested percentile if not count_at_percentile: if percentile_list_index == len(percentile_list): return result percentile = percentile_list[percentile_list_index] percentile_list_index += 1 if percentile > 100: return result count_at_percentile = self.get_target_count_at_percentile(percentile) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: result[percentile] = self.get_highest_equivalent_value(value_at_index) else: result[percentile] = self.get_lowest_equivalent_value(value_at_index) count_at_percentile = 0 else: break return result def get_total_count(self): return self.total_count def get_count_at_value(self, value): counts_index = self._counts_index_for(value) return self.counts[counts_index] def values_are_equivalent(self, val1, val2): '''Check whether 2 values are equivalent (meaning they are in the same bucket/range) Returns: true if the 2 values are equivalent ''' return self.get_lowest_equivalent_value(val1) == self.get_lowest_equivalent_value(val2) def get_max_value(self): if self.max_value == 0: return 0 return self.get_highest_equivalent_value(self.max_value) def get_min_value(self): if self.counts[0] > 0 or self.total_count == 0: return 0 if sys.maxsize == self.min_value: return sys.maxsize return self.get_lowest_equivalent_value(self.min_value) def _hdr_size_of_equiv_value_range(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 return 1 << (self.unit_magnitude + bucket_index) def _hdr_median_equiv_value(self, value): return self.get_lowest_equivalent_value(value) + \ (self._hdr_size_of_equiv_value_range(value) >> 1) def get_mean_value(self): if not self.total_count: return 0.0 total = 0 itr = self.get_recorded_iterator() for item in itr: total += itr.count_at_this_value * self._hdr_median_equiv_value(item.value_iterated_to) return float(total) / self.total_count def get_stddev(self): if not self.total_count: return 0.0 mean = self.get_mean_value() geometric_dev_total = 0.0 for item in self.get_recorded_iterator(): dev = (self._hdr_median_equiv_value(item.value_iterated_to) * 1.0) - mean geometric_dev_total += (dev * dev) * item.count_added_in_this_iter_step return math.sqrt(geometric_dev_total / self.total_count) def reset(self): '''Reset the histogram to a pristine state ''' for index in range(self.counts_len): self.counts[index] = 0 self.total_count = 0 self.min_value = sys.maxsize self.max_value = 0 self.start_time_stamp_msec = sys.maxsize self.end_time_stamp_msec = 0 def __iter__(self): '''Returns the recorded iterator if iter(self) is called ''' return RecordedIterator(self) def get_all_values_iterator(self): return AllValuesIterator(self) def get_recorded_iterator(self): return RecordedIterator(self) def get_percentile_iterator(self, ticks_per_half_distance): return PercentileIterator(self, ticks_per_half_distance) def get_linear_iterator(self, value_units_per_bucket): return LinearIterator(self, value_units_per_bucket) def get_log_iterator(self, value_units_first_bucket, log_base): return LogIterator(self, value_units_first_bucket, log_base) def encode(self): '''Encode this histogram Return: a string containing the base64 encoded compressed histogram (V1 format) ''' return self.encoder.encode() def adjust_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) self.max_value = max(self.max_value, max_value) if min_non_zero_index >= 0: min_value = self.get_value_from_index(min_non_zero_index) self.min_value = min(self.min_value, min_value) self.total_count += total_added def set_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: self.max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) if min_non_zero_index >= 0: self.min_value = self.get_value_from_index(min_non_zero_index) self.total_count = total_added def get_counts_array_index(self, value): '''Return the index in the counts array for a given value ''' if value < 0: raise ValueError("Histogram recorded value cannot be negative.") bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) # Calculate the index for the first entry in the bucket: bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # The following is the equivalent of ((bucket_index + 1) * sub_bucket_half_count) # Calculate the offset in the bucket (can be negative for first bucket): offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - sub_bucket_half_count) + bucket_base_index return bucket_base_index + offset_in_bucket def get_start_time_stamp(self): return self.start_time_stamp_msec def set_start_time_stamp(self, time_stamp_msec): '''Set the start time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.start_time_stamp_msec = time_stamp_msec def get_end_time_stamp(self): return self.end_time_stamp_msec def set_end_time_stamp(self, time_stamp_msec): '''Set the end time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.end_time_stamp_msec = time_stamp_msec def add(self, other_hist): highest_recordable_value = \ self.get_highest_equivalent_value(self.get_value_from_index(self.counts_len - 1)) if highest_recordable_value < other_hist.get_max_value(): raise IndexError("The other histogram includes values that do not fit %d < %d" % (highest_recordable_value, other_hist.get_max_value())) if (self.bucket_count == other_hist.bucket_count) and \ (self.sub_bucket_count == other_hist.sub_bucket_count) and \ (self.unit_magnitude == other_hist.unit_magnitude) and \ (self.word_size == other_hist.word_size): # do an in-place addition of one array to another self.encoder.add(other_hist.encoder) self.total_count += other_hist.get_total_count() self.max_value = max(self.max_value, other_hist.get_max_value()) self.min_value = min(self.get_min_value(), other_hist.get_min_value()) else: # Arrays are not a direct match, so we can't just stream through and add them. # Instead, go through the array and add each non-zero value found at it's proper value: for index in range(other_hist.counts_len): other_count = other_hist.get_count_at_index(index) if other_count > 0: self.record_value(other_hist.get_value_from_index(index), other_count) self.start_time_stamp_msec = \ min(self.start_time_stamp_msec, other_hist.start_time_stamp_msec) self.end_time_stamp_msec = \ max(self.end_time_stamp_msec, other_hist.end_time_stamp_msec) @staticmethod def decode(encoded_histogram, b64_wrap=True): '''Decode an encoded histogram and return a new histogram instance that has been initialized with the decoded content Return: a new histogram instance representing the decoded content Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' hdr_payload = HdrHistogramEncoder.decode(encoded_histogram, b64_wrap) payload = hdr_payload.payload histogram = HdrHistogram(payload.lowest_trackable_value, payload.highest_trackable_value, payload.significant_figures, hdr_payload=hdr_payload) return histogram def get_word_size(self): return self.word_size def output_percentile_distribution(self, out_file, output_value_unit_scaling_ratio, ticks_per_half_distance=5): out_file.write(b'%12s %14s %10s %14s\n\n' % (b'Value', b'Percentile', b'TotalCount', b'1/(1-Percentile)')) percentile_format = '%12.{}f %2.12f %10d %14.2f\n'.format(self.significant_figures) last_line_percentile_format = '%12.{}f %2.12f %10d\n'.format(self.significant_figures) for iter_value in self.get_percentile_iterator(ticks_per_half_distance): value = iter_value.value_iterated_to / output_value_unit_scaling_ratio percentile = iter_value.percentile_level_iterated_to / 100 total_count = iter_value.total_count_to_this_value if iter_value.percentile_level_iterated_to != 100: other = 1 / (1 - iter_value.percentile_level_iterated_to / 100) out_file.write(percentile_format.encode() % (value, percentile, total_count, other)) else: out_file.write(last_line_percentile_format.encode() % (value, percentile, total_count)) mean = self.get_mean_value() / output_value_unit_scaling_ratio stddev = self.get_stddev() out_file.write('#[Mean = %12.{0}f, StdDeviation = %12.{0}f]\n'. format(self.significant_figures).encode() % (mean, stddev)) max = self.get_max_value() / output_value_unit_scaling_ratio total = self.get_total_count() out_file.write('#[Max = %12.{0}f, TotalCount = %12.{0}f]\n'.format( self.significant_figures).encode() % (max, total)) out_file.write(b'#[Buckets = %12d, SubBuckets = %12d]\n' % ( self.bucket_count, self.sub_bucket_count))
HdrHistogram/HdrHistogram_py
hdrh/histogram.py
HdrHistogram.decode
python
def decode(encoded_histogram, b64_wrap=True): '''Decode an encoded histogram and return a new histogram instance that has been initialized with the decoded content Return: a new histogram instance representing the decoded content Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' hdr_payload = HdrHistogramEncoder.decode(encoded_histogram, b64_wrap) payload = hdr_payload.payload histogram = HdrHistogram(payload.lowest_trackable_value, payload.highest_trackable_value, payload.significant_figures, hdr_payload=hdr_payload) return histogram
Decode an encoded histogram and return a new histogram instance that has been initialized with the decoded content Return: a new histogram instance representing the decoded content Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/histogram.py#L550-L572
[ "def decode(encoded_histogram, b64_wrap=True):\n '''Decode a wire histogram encoding into a read-only Hdr Payload instance\n Args:\n encoded_histogram a string containing the wire encoding of a histogram\n such as one returned from encode()\n Returns:\n an hdr_payload instance with all the decoded/uncompressed fields\n\n Exception:\n TypeError in case of base64 decode error\n HdrCookieException:\n the main header has an invalid cookie\n the compressed payload header has an invalid cookie\n HdrLengthException:\n the decompressed size is too small for the HdrPayload structure\n or is not aligned or is too large for the passed payload class\n HdrHistogramSettingsException:\n mismatch in the significant figures, lowest and highest\n trackable value\n zlib.error:\n in case of zlib decompression error\n '''\n if b64_wrap:\n b64decode = base64.b64decode(encoded_histogram)\n # this string has 2 parts in it: the header (raw) and the payload (compressed)\n b64dec_len = len(b64decode)\n\n if b64dec_len < ext_header_size:\n raise HdrLengthException('Base64 decoded message too short')\n\n header = ExternalHeader.from_buffer_copy(b64decode)\n if get_cookie_base(header.cookie) != V2_COMPRESSION_COOKIE_BASE:\n raise HdrCookieException()\n if header.length != b64dec_len - ext_header_size:\n raise HdrLengthException('Decoded length=%d buffer length=%d' %\n (header.length, b64dec_len - ext_header_size))\n # this will result in a copy of the compressed payload part\n # could not find a way to do otherwise since zlib.decompress()\n # expects a string (and does not like a buffer or a memoryview object)\n cpayload = b64decode[ext_header_size:]\n else:\n cpayload = encoded_histogram\n hdr_payload = HdrPayload(8, compressed_payload=cpayload)\n return hdr_payload\n" ]
class HdrHistogram(): '''This class supports the recording and analyzing of sampled data value counts across a configurable integer value range with configurable value precision within the range. Value precision is expressed as the number of significant digits in the value recording, and provides control over value quantization behavior across the value range and the subsequent value resolution at any given level. For example, a Histogram could be configured to track the counts of observed integer values between 0 and 3,600,000,000 while maintaining a value precision of 3 significant digits across that range. Value quantization within the range will thus be no larger than 1/1,000th (or 0.1%) of any value. This example Histogram could be used to track and analyze the counts of observed response times ranging between 1 microsecond and 1 hour in magnitude, while maintaining a value resolution of 1 microsecond up to 1 millisecond, a resolution of 1 millisecond (or better) up to one second, and a resolution of 1 second (or better) up to 1,000 seconds. At it's maximum tracked value (1 hour), it would still maintain a resolution of 3.6 seconds (or better). ''' def __init__(self, lowest_trackable_value, highest_trackable_value, significant_figures, word_size=8, b64_wrap=True, hdr_payload=None): '''Create a new histogram with given arguments Params: lowest_trackable_value The lowest value that can be discerned (distinguished from 0) by the histogram. Must be a positive integer that is >= 1. May be internally rounded down to nearest power of 2. highest_trackable_value The highest value to be tracked by the histogram. Must be a positive integer that is >= (2 * lowest_trackable_value). significant_figures The number of significant decimal digits to which the histogram will maintain value resolution and separation. Must be a non-negative integer between 0 and 5. word_size size of counters in bytes, only 2, 4, 8-byte counters are supported (default is 8-byte or 64-bit counters) b64_wrap specifies if the encoding of this histogram should use base64 wrapping (only useful if you need to encode the histogram to save somewhere or send over the wire. By default base64 encoding is assumed hdr_payload only used for associating an existing payload created from decoding an encoded histograme Exceptions: ValueError if the word_size value is unsupported if significant_figures is invalid ''' if significant_figures < 1 or significant_figures > 5: raise ValueError('Invalid significant_figures') self.lowest_trackable_value = lowest_trackable_value self.highest_trackable_value = highest_trackable_value self.significant_figures = significant_figures self.unit_magnitude = int(math.floor(math.log(lowest_trackable_value) / math.log(2))) largest_value_single_unit_res = 2 * math.pow(10, significant_figures) subb_count_mag = int(math.ceil(math.log(largest_value_single_unit_res) / math.log(2))) self.sub_bucket_half_count_magnitude = subb_count_mag - 1 if subb_count_mag > 1 else 0 self.sub_bucket_count = int(math.pow(2, self.sub_bucket_half_count_magnitude + 1)) self.sub_bucket_half_count = self.sub_bucket_count // 2 self.sub_bucket_mask = (self.sub_bucket_count - 1) << self.unit_magnitude self.bucket_count = get_bucket_count(highest_trackable_value, self.sub_bucket_count, self.unit_magnitude) self.min_value = sys.maxsize self.max_value = 0 self.total_count = 0 self.counts_len = (self.bucket_count + 1) * (self.sub_bucket_count // 2) self.word_size = word_size if hdr_payload: payload = hdr_payload.payload self.int_to_double_conversion_ratio = payload.conversion_ratio_bits results = hdr_payload.init_counts(self.counts_len) if results['total']: self.set_internal_tacking_values(results['min_nonzero_index'], results['max_nonzero_index'], results['total']) else: self.int_to_double_conversion_ratio = 1.0 # to encode this histogram into a compressed/base64 format ready # to be exported self.b64_wrap = b64_wrap self.encoder = HdrHistogramEncoder(self, b64_wrap, hdr_payload) # the counters reside directly in the payload object # allocated by the encoder # so that compression for wire transfer can be done without copy self.counts = self.encoder.get_counts() self.start_time_stamp_msec = 0 self.end_time_stamp_msec = 0 def _clz(self, value): """calculate the leading zeros, equivalent to C __builtin_clzll() value in hex: value = 1 clz = 63 value = 2 clz = 62 value = 4 clz = 61 value = 1000 clz = 51 value = 1000000 clz = 39 """ return 63 - (len(bin(value)) - 3) def _get_bucket_index(self, value): # smallest power of 2 containing value pow2ceiling = 64 - self._clz(int(value) | self.sub_bucket_mask) return int(pow2ceiling - self.unit_magnitude - (self.sub_bucket_half_count_magnitude + 1)) def _get_sub_bucket_index(self, value, bucket_index): return int(value) >> (bucket_index + self.unit_magnitude) def _counts_index(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ): bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - subBucketHalfCount) + bucketBaseIndex return bucket_base_index + offset_in_bucket def _counts_index_for(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) return self._counts_index(bucket_index, sub_bucket_index) def record_value(self, value, count=1): '''Record a new value into the histogram Args: value: the value to record (must be in the valid range) count: incremental count (defaults to 1) ''' if value < 0: return False counts_index = self._counts_index_for(value) if (counts_index < 0) or (self.counts_len <= counts_index): return False self.counts[counts_index] += count self.total_count += count self.min_value = min(self.min_value, value) self.max_value = max(self.max_value, value) return True # pylint: disable=inconsistent-return-statements def record_corrected_value(self, value, expected_interval, count=1): '''Record a new value into the histogram and correct for coordinated omission if needed Args: value: the value to record (must be in the valid range) expected_interval: the expected interval between 2 value samples count: incremental count (defaults to 1) ''' while True: if not self.record_value(value, count): return False if value <= expected_interval or expected_interval <= 0: return True value -= expected_interval def get_count_at_index(self, index): if index >= self.counts_len: raise IndexError() # some decoded (read-only) histograms may have truncated # counts arrays, we return zero for any index that is passed the array if index >= self.encoder.payload.counts_len: return 0 return self.counts[index] def get_count_at_sub_bucket(self, bucket_index, sub_bucket_index): # Calculate the index for the first entry in the bucket: # (The following is the equivalent of ((bucket_index + 1) * subBucketHalfCount) ) bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # Calculate the offset in the bucket: offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # (sub_bucket_index - subBucketHalfCount) + bucketBaseIndex counts_index = bucket_base_index + offset_in_bucket return self.counts[counts_index] def get_value_from_sub_bucket(self, bucket_index, sub_bucket_index): return sub_bucket_index << (bucket_index + self.unit_magnitude) def get_value_from_index(self, index): bucket_index = (index >> self.sub_bucket_half_count_magnitude) - 1 sub_bucket_index = (index & (self.sub_bucket_half_count - 1)) + \ self.sub_bucket_half_count if bucket_index < 0: sub_bucket_index -= self.sub_bucket_half_count bucket_index = 0 return self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) def get_lowest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) return lowest_equivalent_value def get_highest_equivalent_value(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) lowest_equivalent_value = self.get_value_from_sub_bucket(bucket_index, sub_bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 size_of_equivalent_value_range = 1 << (self.unit_magnitude + bucket_index) next_non_equivalent_value = lowest_equivalent_value + size_of_equivalent_value_range return next_non_equivalent_value - 1 def get_target_count_at_percentile(self, percentile): requested_percentile = min(percentile, 100.0) count_at_percentile = int(((requested_percentile * self.total_count / 100)) + 0.5) return max(count_at_percentile, 1) def get_value_at_percentile(self, percentile): '''Get the value for a given percentile Args: percentile: a float in [0.0..100.0] Returns: the value for the given percentile ''' count_at_percentile = self.get_target_count_at_percentile(percentile) total = 0 for index in range(self.counts_len): total += self.get_count_at_index(index) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: return self.get_highest_equivalent_value(value_at_index) return self.get_lowest_equivalent_value(value_at_index) return 0 def get_percentile_to_value_dict(self, percentile_list): '''A faster alternative to query values for a list of percentiles. Args: percentile_list: a list of percentiles in any order, dups will be ignored each element in the list must be a float value in [0.0 .. 100.0] Returns: a dict of percentile values indexed by the percentile ''' result = {} total = 0 percentile_list_index = 0 count_at_percentile = 0 # remove dups and sort percentile_list = list(set(percentile_list)) percentile_list.sort() for index in range(self.counts_len): total += self.get_count_at_index(index) while True: # recalculate target based on next requested percentile if not count_at_percentile: if percentile_list_index == len(percentile_list): return result percentile = percentile_list[percentile_list_index] percentile_list_index += 1 if percentile > 100: return result count_at_percentile = self.get_target_count_at_percentile(percentile) if total >= count_at_percentile: value_at_index = self.get_value_from_index(index) if percentile: result[percentile] = self.get_highest_equivalent_value(value_at_index) else: result[percentile] = self.get_lowest_equivalent_value(value_at_index) count_at_percentile = 0 else: break return result def get_total_count(self): return self.total_count def get_count_at_value(self, value): counts_index = self._counts_index_for(value) return self.counts[counts_index] def values_are_equivalent(self, val1, val2): '''Check whether 2 values are equivalent (meaning they are in the same bucket/range) Returns: true if the 2 values are equivalent ''' return self.get_lowest_equivalent_value(val1) == self.get_lowest_equivalent_value(val2) def get_max_value(self): if self.max_value == 0: return 0 return self.get_highest_equivalent_value(self.max_value) def get_min_value(self): if self.counts[0] > 0 or self.total_count == 0: return 0 if sys.maxsize == self.min_value: return sys.maxsize return self.get_lowest_equivalent_value(self.min_value) def _hdr_size_of_equiv_value_range(self, value): bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) if sub_bucket_index >= self.sub_bucket_count: bucket_index += 1 return 1 << (self.unit_magnitude + bucket_index) def _hdr_median_equiv_value(self, value): return self.get_lowest_equivalent_value(value) + \ (self._hdr_size_of_equiv_value_range(value) >> 1) def get_mean_value(self): if not self.total_count: return 0.0 total = 0 itr = self.get_recorded_iterator() for item in itr: total += itr.count_at_this_value * self._hdr_median_equiv_value(item.value_iterated_to) return float(total) / self.total_count def get_stddev(self): if not self.total_count: return 0.0 mean = self.get_mean_value() geometric_dev_total = 0.0 for item in self.get_recorded_iterator(): dev = (self._hdr_median_equiv_value(item.value_iterated_to) * 1.0) - mean geometric_dev_total += (dev * dev) * item.count_added_in_this_iter_step return math.sqrt(geometric_dev_total / self.total_count) def reset(self): '''Reset the histogram to a pristine state ''' for index in range(self.counts_len): self.counts[index] = 0 self.total_count = 0 self.min_value = sys.maxsize self.max_value = 0 self.start_time_stamp_msec = sys.maxsize self.end_time_stamp_msec = 0 def __iter__(self): '''Returns the recorded iterator if iter(self) is called ''' return RecordedIterator(self) def get_all_values_iterator(self): return AllValuesIterator(self) def get_recorded_iterator(self): return RecordedIterator(self) def get_percentile_iterator(self, ticks_per_half_distance): return PercentileIterator(self, ticks_per_half_distance) def get_linear_iterator(self, value_units_per_bucket): return LinearIterator(self, value_units_per_bucket) def get_log_iterator(self, value_units_first_bucket, log_base): return LogIterator(self, value_units_first_bucket, log_base) def encode(self): '''Encode this histogram Return: a string containing the base64 encoded compressed histogram (V1 format) ''' return self.encoder.encode() def adjust_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) self.max_value = max(self.max_value, max_value) if min_non_zero_index >= 0: min_value = self.get_value_from_index(min_non_zero_index) self.min_value = min(self.min_value, min_value) self.total_count += total_added def set_internal_tacking_values(self, min_non_zero_index, max_index, total_added): '''Called during decoding and add to adjust the new min/max value and total count Args: min_non_zero_index min nonzero index of all added counts (-1 if none) max_index max index of all added counts (-1 if none) ''' if max_index >= 0: self.max_value = self.get_highest_equivalent_value(self.get_value_from_index(max_index)) if min_non_zero_index >= 0: self.min_value = self.get_value_from_index(min_non_zero_index) self.total_count = total_added def get_counts_array_index(self, value): '''Return the index in the counts array for a given value ''' if value < 0: raise ValueError("Histogram recorded value cannot be negative.") bucket_index = self._get_bucket_index(value) sub_bucket_index = self._get_sub_bucket_index(value, bucket_index) # Calculate the index for the first entry in the bucket: bucket_base_index = (bucket_index + 1) << self.sub_bucket_half_count_magnitude # The following is the equivalent of ((bucket_index + 1) * sub_bucket_half_count) # Calculate the offset in the bucket (can be negative for first bucket): offset_in_bucket = sub_bucket_index - self.sub_bucket_half_count # The following is the equivalent of # ((sub_bucket_index - sub_bucket_half_count) + bucket_base_index return bucket_base_index + offset_in_bucket def get_start_time_stamp(self): return self.start_time_stamp_msec def set_start_time_stamp(self, time_stamp_msec): '''Set the start time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.start_time_stamp_msec = time_stamp_msec def get_end_time_stamp(self): return self.end_time_stamp_msec def set_end_time_stamp(self, time_stamp_msec): '''Set the end time stamp value associated with this histogram to a given value. Params: time_stamp_msec the value to set the time stamp to, [by convention] in msec since the epoch. ''' self.end_time_stamp_msec = time_stamp_msec def add(self, other_hist): highest_recordable_value = \ self.get_highest_equivalent_value(self.get_value_from_index(self.counts_len - 1)) if highest_recordable_value < other_hist.get_max_value(): raise IndexError("The other histogram includes values that do not fit %d < %d" % (highest_recordable_value, other_hist.get_max_value())) if (self.bucket_count == other_hist.bucket_count) and \ (self.sub_bucket_count == other_hist.sub_bucket_count) and \ (self.unit_magnitude == other_hist.unit_magnitude) and \ (self.word_size == other_hist.word_size): # do an in-place addition of one array to another self.encoder.add(other_hist.encoder) self.total_count += other_hist.get_total_count() self.max_value = max(self.max_value, other_hist.get_max_value()) self.min_value = min(self.get_min_value(), other_hist.get_min_value()) else: # Arrays are not a direct match, so we can't just stream through and add them. # Instead, go through the array and add each non-zero value found at it's proper value: for index in range(other_hist.counts_len): other_count = other_hist.get_count_at_index(index) if other_count > 0: self.record_value(other_hist.get_value_from_index(index), other_count) self.start_time_stamp_msec = \ min(self.start_time_stamp_msec, other_hist.start_time_stamp_msec) self.end_time_stamp_msec = \ max(self.end_time_stamp_msec, other_hist.end_time_stamp_msec) def decode_and_add(self, encoded_histogram): '''Decode an encoded histogram and add it to this histogram Args: encoded_histogram (string) an encoded histogram following the V1 format, such as one returned by the encode() method Exception: TypeError in case of base64 decode error HdrCookieException: the main header has an invalid cookie the compressed payload header has an invalid cookie HdrLengthException: the decompressed size is too small for the HdrPayload structure or is not aligned or is too large for the passed payload class zlib.error: in case of zlib decompression error ''' other_hist = HdrHistogram.decode(encoded_histogram, self.b64_wrap) self.add(other_hist) @staticmethod def get_word_size(self): return self.word_size def output_percentile_distribution(self, out_file, output_value_unit_scaling_ratio, ticks_per_half_distance=5): out_file.write(b'%12s %14s %10s %14s\n\n' % (b'Value', b'Percentile', b'TotalCount', b'1/(1-Percentile)')) percentile_format = '%12.{}f %2.12f %10d %14.2f\n'.format(self.significant_figures) last_line_percentile_format = '%12.{}f %2.12f %10d\n'.format(self.significant_figures) for iter_value in self.get_percentile_iterator(ticks_per_half_distance): value = iter_value.value_iterated_to / output_value_unit_scaling_ratio percentile = iter_value.percentile_level_iterated_to / 100 total_count = iter_value.total_count_to_this_value if iter_value.percentile_level_iterated_to != 100: other = 1 / (1 - iter_value.percentile_level_iterated_to / 100) out_file.write(percentile_format.encode() % (value, percentile, total_count, other)) else: out_file.write(last_line_percentile_format.encode() % (value, percentile, total_count)) mean = self.get_mean_value() / output_value_unit_scaling_ratio stddev = self.get_stddev() out_file.write('#[Mean = %12.{0}f, StdDeviation = %12.{0}f]\n'. format(self.significant_figures).encode() % (mean, stddev)) max = self.get_max_value() / output_value_unit_scaling_ratio total = self.get_total_count() out_file.write('#[Max = %12.{0}f, TotalCount = %12.{0}f]\n'.format( self.significant_figures).encode() % (max, total)) out_file.write(b'#[Buckets = %12d, SubBuckets = %12d]\n' % ( self.bucket_count, self.sub_bucket_count))
HdrHistogram/HdrHistogram_py
hdrh/log.py
HistogramLogWriter.output_interval_histogram
python
def output_interval_histogram(self, histogram, start_time_stamp_sec=0, end_time_stamp_sec=0, max_value_unit_ratio=1000000.0): '''Output an interval histogram, with the given timestamp and a configurable maxValueUnitRatio. (note that the specified timestamp will be used, and the timestamp in the actual histogram will be ignored). The max value reported with the interval line will be scaled by the given max_value_unit_ratio. The histogram start and end timestamps are assumed to be in msec units. Logging will be in seconds, realtive by a base time The default base time is 0. By covention, histogram start/end time are generally stamped with absolute times in msec since the epoch. For logging with absolute time stamps, the base time would remain zero. For logging with relative time stamps (time since a start point), Params: histogram The interval histogram to log. start_time_stamp_sec The start timestamp to log with the interval histogram, in seconds. default: using the start/end timestamp indicated in the histogram end_time_stamp_sec The end timestamp to log with the interval histogram, in seconds. default: using the start/end timestamp indicated in the histogram max_value_unit_ratio The ratio by which to divide the histogram's max value when reporting on it. default: 1,000,000 (which is the msec : nsec ratio ''' if not start_time_stamp_sec: start_time_stamp_sec = \ (histogram.get_start_time_stamp() - self.base_time) / 1000.0 if not end_time_stamp_sec: end_time_stamp_sec = (histogram.get_end_time_stamp() - self.base_time) / 1000.0 cpayload = histogram.encode() self.log.write("%f,%f,%f,%s\n" % (start_time_stamp_sec, end_time_stamp_sec - start_time_stamp_sec, histogram.get_max_value() // max_value_unit_ratio, cpayload.decode('utf-8')))
Output an interval histogram, with the given timestamp and a configurable maxValueUnitRatio. (note that the specified timestamp will be used, and the timestamp in the actual histogram will be ignored). The max value reported with the interval line will be scaled by the given max_value_unit_ratio. The histogram start and end timestamps are assumed to be in msec units. Logging will be in seconds, realtive by a base time The default base time is 0. By covention, histogram start/end time are generally stamped with absolute times in msec since the epoch. For logging with absolute time stamps, the base time would remain zero. For logging with relative time stamps (time since a start point), Params: histogram The interval histogram to log. start_time_stamp_sec The start timestamp to log with the interval histogram, in seconds. default: using the start/end timestamp indicated in the histogram end_time_stamp_sec The end timestamp to log with the interval histogram, in seconds. default: using the start/end timestamp indicated in the histogram max_value_unit_ratio The ratio by which to divide the histogram's max value when reporting on it. default: 1,000,000 (which is the msec : nsec ratio
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/log.py#L51-L92
[ "def get_max_value(self):\n if self.max_value == 0:\n return 0\n return self.get_highest_equivalent_value(self.max_value)\n", "def encode(self):\n '''Encode this histogram\n Return:\n a string containing the base64 encoded compressed histogram (V1 format)\n '''\n return self.encoder.encode()\n", "def get_start_time_stamp(self):\n return self.start_time_stamp_msec\n", "def get_end_time_stamp(self):\n return self.end_time_stamp_msec\n" ]
class HistogramLogWriter(): HISTOGRAM_LOG_FORMAT_VERSION = "1.2" def __init__(self, output_file): '''Constructs a new HistogramLogWriter that will write into the specified file. Params: output_file the File to write to ''' self.log = output_file self.base_time = 0 def output_start_time(self, start_time_msec): '''Log a start time in the log. Params: start_time_msec time (in milliseconds) since the absolute start time (the epoch) ''' self.log.write("#[StartTime: %f (seconds since epoch), %s]\n" % (float(start_time_msec) / 1000.0, datetime.fromtimestamp(start_time_msec).iso_format(' '))) def output_base_time(self, base_time_msec): '''Log a base time in the log. Params: base_time_msec time (in milliseconds) since the absolute start time (the epoch) ''' self.log.write("#[BaseTime: %f (seconds since epoch)]\n" % (float(base_time_msec) / 1000.0)) def output_comment(self, comment): '''Log a comment to the log. Comments will be preceded with with the '#' character. Params: comment the comment string. ''' self.log.write("#%s\n" % (comment)) def output_legend(self): '''Output a legend line to the log. ''' self.log.write("\"StartTimestamp\",\"Interval_Length\"," "\"Interval_Max\",\"Interval_Compressed_Histogram\"\n") def output_log_format_version(self): '''Output a log format version to the log. ''' self.output_comment("[Histogram log format version " + HistogramLogWriter.HISTOGRAM_LOG_FORMAT_VERSION + "]") def close(self): self.log.close()
HdrHistogram/HdrHistogram_py
hdrh/log.py
HistogramLogWriter.output_start_time
python
def output_start_time(self, start_time_msec): '''Log a start time in the log. Params: start_time_msec time (in milliseconds) since the absolute start time (the epoch) ''' self.log.write("#[StartTime: %f (seconds since epoch), %s]\n" % (float(start_time_msec) / 1000.0, datetime.fromtimestamp(start_time_msec).iso_format(' ')))
Log a start time in the log. Params: start_time_msec time (in milliseconds) since the absolute start time (the epoch)
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/log.py#L94-L101
null
class HistogramLogWriter(): HISTOGRAM_LOG_FORMAT_VERSION = "1.2" def __init__(self, output_file): '''Constructs a new HistogramLogWriter that will write into the specified file. Params: output_file the File to write to ''' self.log = output_file self.base_time = 0 def output_interval_histogram(self, histogram, start_time_stamp_sec=0, end_time_stamp_sec=0, max_value_unit_ratio=1000000.0): '''Output an interval histogram, with the given timestamp and a configurable maxValueUnitRatio. (note that the specified timestamp will be used, and the timestamp in the actual histogram will be ignored). The max value reported with the interval line will be scaled by the given max_value_unit_ratio. The histogram start and end timestamps are assumed to be in msec units. Logging will be in seconds, realtive by a base time The default base time is 0. By covention, histogram start/end time are generally stamped with absolute times in msec since the epoch. For logging with absolute time stamps, the base time would remain zero. For logging with relative time stamps (time since a start point), Params: histogram The interval histogram to log. start_time_stamp_sec The start timestamp to log with the interval histogram, in seconds. default: using the start/end timestamp indicated in the histogram end_time_stamp_sec The end timestamp to log with the interval histogram, in seconds. default: using the start/end timestamp indicated in the histogram max_value_unit_ratio The ratio by which to divide the histogram's max value when reporting on it. default: 1,000,000 (which is the msec : nsec ratio ''' if not start_time_stamp_sec: start_time_stamp_sec = \ (histogram.get_start_time_stamp() - self.base_time) / 1000.0 if not end_time_stamp_sec: end_time_stamp_sec = (histogram.get_end_time_stamp() - self.base_time) / 1000.0 cpayload = histogram.encode() self.log.write("%f,%f,%f,%s\n" % (start_time_stamp_sec, end_time_stamp_sec - start_time_stamp_sec, histogram.get_max_value() // max_value_unit_ratio, cpayload.decode('utf-8'))) def output_base_time(self, base_time_msec): '''Log a base time in the log. Params: base_time_msec time (in milliseconds) since the absolute start time (the epoch) ''' self.log.write("#[BaseTime: %f (seconds since epoch)]\n" % (float(base_time_msec) / 1000.0)) def output_comment(self, comment): '''Log a comment to the log. Comments will be preceded with with the '#' character. Params: comment the comment string. ''' self.log.write("#%s\n" % (comment)) def output_legend(self): '''Output a legend line to the log. ''' self.log.write("\"StartTimestamp\",\"Interval_Length\"," "\"Interval_Max\",\"Interval_Compressed_Histogram\"\n") def output_log_format_version(self): '''Output a log format version to the log. ''' self.output_comment("[Histogram log format version " + HistogramLogWriter.HISTOGRAM_LOG_FORMAT_VERSION + "]") def close(self): self.log.close()
HdrHistogram/HdrHistogram_py
hdrh/log.py
HistogramLogReader._decode_next_interval_histogram
python
def _decode_next_interval_histogram(self, dest_histogram, range_start_time_sec=0.0, range_end_time_sec=sys.maxsize, absolute=False): '''Read the next interval histogram from the log, if interval falls within an absolute or relative time range. Timestamps are assumed to appear in order in the log file, and as such this method will return a null upon encountering a timestamp larger than range_end_time_sec. Relative time range: the range is assumed to be in seconds relative to the actual timestamp value found in each interval line in the log Absolute time range: Absolute timestamps are calculated by adding the timestamp found with the recorded interval to the [latest, optional] start time found in the log. The start time is indicated in the log with a "#[StartTime: " followed by the start time in seconds. Params: dest_histogram if None, created a new histogram, else adds the new interval histogram to it range_start_time_sec The absolute or relative start of the expected time range, in seconds. range_start_time_sec The absolute or relative end of the expected time range, in seconds. absolute Defines if the passed range is absolute or relative Return: Returns an histogram object if an interval line was found with an associated start timestamp value that falls between start_time_sec and end_time_sec, or null if no such interval line is found. Upon encountering any unexpected format errors in reading the next interval from the file, this method will return None. The histogram returned will have it's timestamp set to the absolute timestamp calculated from adding the interval's indicated timestamp value to the latest [optional] start time found in the log. Exceptions: ValueError if there is a syntax error in one of the float fields ''' while 1: line = self.input_file.readline() if not line: return None if line[0] == '#': match_res = re_start_time.match(line) if match_res: self.start_time_sec = float(match_res.group(1)) self.observed_start_time = True continue match_res = re_base_time.match(line) if match_res: self.base_time_sec = float(match_res.group(1)) self.observed_base_time = True continue match_res = re_histogram_interval.match(line) if not match_res: # probably a legend line that starts with "\"StartTimestamp" continue # Decode: startTimestamp, intervalLength, maxTime, histogramPayload # Timestamp is expected to be in seconds log_time_stamp_in_sec = float(match_res.group(1)) interval_length_sec = float(match_res.group(2)) cpayload = match_res.group(4) if not self.observed_start_time: # No explicit start time noted. Use 1st observed time: self.start_time_sec = log_time_stamp_in_sec self.observed_start_time = True if not self.observed_base_time: # No explicit base time noted. # Deduce from 1st observed time (compared to start time): if log_time_stamp_in_sec < self.start_time_sec - (365 * 24 * 3600.0): # Criteria Note: if log timestamp is more than a year in # the past (compared to StartTime), # we assume that timestamps in the log are not absolute self.base_time_sec = self.start_time_sec else: # Timestamps are absolute self.base_time_sec = 0.0 self.observed_base_time = True absolute_start_time_stamp_sec = \ log_time_stamp_in_sec + self.base_time_sec offset_start_time_stamp_sec = \ absolute_start_time_stamp_sec - self.start_time_sec # Timestamp length is expect to be in seconds absolute_end_time_stamp_sec = \ absolute_start_time_stamp_sec + interval_length_sec if absolute: start_time_stamp_to_check_range_on = absolute_start_time_stamp_sec else: start_time_stamp_to_check_range_on = offset_start_time_stamp_sec if start_time_stamp_to_check_range_on < range_start_time_sec: continue if start_time_stamp_to_check_range_on > range_end_time_sec: return None if dest_histogram: # add the interval histogram to the destination histogram histogram = dest_histogram histogram.decode_and_add(cpayload) else: histogram = HdrHistogram.decode(cpayload) histogram.set_start_time_stamp(absolute_start_time_stamp_sec * 1000.0) histogram.set_end_time_stamp(absolute_end_time_stamp_sec * 1000.0) return histogram
Read the next interval histogram from the log, if interval falls within an absolute or relative time range. Timestamps are assumed to appear in order in the log file, and as such this method will return a null upon encountering a timestamp larger than range_end_time_sec. Relative time range: the range is assumed to be in seconds relative to the actual timestamp value found in each interval line in the log Absolute time range: Absolute timestamps are calculated by adding the timestamp found with the recorded interval to the [latest, optional] start time found in the log. The start time is indicated in the log with a "#[StartTime: " followed by the start time in seconds. Params: dest_histogram if None, created a new histogram, else adds the new interval histogram to it range_start_time_sec The absolute or relative start of the expected time range, in seconds. range_start_time_sec The absolute or relative end of the expected time range, in seconds. absolute Defines if the passed range is absolute or relative Return: Returns an histogram object if an interval line was found with an associated start timestamp value that falls between start_time_sec and end_time_sec, or null if no such interval line is found. Upon encountering any unexpected format errors in reading the next interval from the file, this method will return None. The histogram returned will have it's timestamp set to the absolute timestamp calculated from adding the interval's indicated timestamp value to the latest [optional] start time found in the log. Exceptions: ValueError if there is a syntax error in one of the float fields
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/log.py#L173-L289
[ "def decode(encoded_histogram, b64_wrap=True):\n '''Decode an encoded histogram and return a new histogram instance that\n has been initialized with the decoded content\n Return:\n a new histogram instance representing the decoded content\n Exception:\n TypeError in case of base64 decode error\n HdrCookieException:\n the main header has an invalid cookie\n the compressed payload header has an invalid cookie\n HdrLengthException:\n the decompressed size is too small for the HdrPayload structure\n or is not aligned or is too large for the passed payload class\n zlib.error:\n in case of zlib decompression error\n '''\n hdr_payload = HdrHistogramEncoder.decode(encoded_histogram, b64_wrap)\n payload = hdr_payload.payload\n histogram = HdrHistogram(payload.lowest_trackable_value,\n payload.highest_trackable_value,\n payload.significant_figures,\n hdr_payload=hdr_payload)\n return histogram\n" ]
class HistogramLogReader(): def __init__(self, input_file_name, reference_histogram): '''Constructs a new HistogramLogReader that produces intervals read from the specified file name. Params: input_file_name The name of the file to read from reference_histogram a histogram instance used as a reference to create new instances for all subsequent decoded interval histograms ''' self.start_time_sec = 0.0 self.observed_start_time = False self.base_time_sec = 0.0 self.observed_base_time = False self.input_file = open(input_file_name, "r") self.reference_histogram = reference_histogram def get_start_time_sec(self): '''get the latest start time found in the file so far (or 0.0), per the log file format explained above. Assuming the "#[StartTime:" comment line precedes the actual intervals recorded in the file, getStartTimeSec() can be safely used after each interval is read to determine's the offset of that interval's timestamp from the epoch. Return: latest Start Time found in the file (or 0.0 if non found) ''' return self.start_time_sec def get_next_interval_histogram(self, range_start_time_sec=0.0, range_end_time_sec=sys.maxsize, absolute=False): '''Read the next interval histogram from the log, if interval falls within an absolute or relative time range. Timestamps are assumed to appear in order in the log file, and as such this method will return a null upon encountering a timestamp larger than range_end_time_sec. Relative time range: the range is assumed to be in seconds relative to the actual timestamp value found in each interval line in the log Absolute time range: Absolute timestamps are calculated by adding the timestamp found with the recorded interval to the [latest, optional] start time found in the log. The start time is indicated in the log with a "#[StartTime: " followed by the start time in seconds. Params: range_start_time_sec The absolute or relative start of the expected time range, in seconds. range_start_time_sec The absolute or relative end of the expected time range, in seconds. absolute Defines if the passed range is absolute or relative Return: Returns an histogram object if an interval line was found with an associated start timestamp value that falls between start_time_sec and end_time_sec, or null if no such interval line is found. Upon encountering any unexpected format errors in reading the next interval from the file, this method will return None. The histogram returned will have it's timestamp set to the absolute timestamp calculated from adding the interval's indicated timestamp value to the latest [optional] start time found in the log. Exceptions: ValueError if there is a syntax error in one of the float fields ''' return self._decode_next_interval_histogram(None, range_start_time_sec, range_end_time_sec, absolute) def add_next_interval_histogram(self, dest_histogram=None, range_start_time_sec=0.0, range_end_time_sec=sys.maxsize, absolute=False): '''Read the next interval histogram from the log, if interval falls within an absolute or relative time range, and add it to the destination histogram (or to the reference histogram if dest_histogram is None) Timestamps are assumed to appear in order in the log file, and as such this method will return a null upon encountering a timestamp larger than range_end_time_sec. Relative time range: the range is assumed to be in seconds relative to the actual timestamp value found in each interval line in the log Absolute time range: Absolute timestamps are calculated by adding the timestamp found with the recorded interval to the [latest, optional] start time found in the log. The start time is indicated in the log with a "#[StartTime: " followed by the start time in seconds. Params: dest_histogram where to add the next interval histogram, if None the interal histogram will be added to the reference histogram passed in the constructor range_start_time_sec The absolute or relative start of the expected time range, in seconds. range_start_time_sec The absolute or relative end of the expected time range, in seconds. absolute Defines if the passed range is absolute or relative Return: Returns the destination histogram if an interval line was found with an associated start timestamp value that falls between start_time_sec and end_time_sec, or None if no such interval line is found. Upon encountering any unexpected format errors in reading the next interval from the file, this method will return None. The histogram returned will have it's timestamp set to the absolute timestamp calculated from adding the interval's indicated timestamp value to the latest [optional] start time found in the log. Exceptions: ValueError if there is a syntax error in one of the float fields ''' if not dest_histogram: dest_histogram = self.reference_histogram return self._decode_next_interval_histogram(dest_histogram, range_start_time_sec, range_end_time_sec, absolute) def close(self): self.input_file.close()
HdrHistogram/HdrHistogram_py
hdrh/log.py
HistogramLogReader.get_next_interval_histogram
python
def get_next_interval_histogram(self, range_start_time_sec=0.0, range_end_time_sec=sys.maxsize, absolute=False): '''Read the next interval histogram from the log, if interval falls within an absolute or relative time range. Timestamps are assumed to appear in order in the log file, and as such this method will return a null upon encountering a timestamp larger than range_end_time_sec. Relative time range: the range is assumed to be in seconds relative to the actual timestamp value found in each interval line in the log Absolute time range: Absolute timestamps are calculated by adding the timestamp found with the recorded interval to the [latest, optional] start time found in the log. The start time is indicated in the log with a "#[StartTime: " followed by the start time in seconds. Params: range_start_time_sec The absolute or relative start of the expected time range, in seconds. range_start_time_sec The absolute or relative end of the expected time range, in seconds. absolute Defines if the passed range is absolute or relative Return: Returns an histogram object if an interval line was found with an associated start timestamp value that falls between start_time_sec and end_time_sec, or null if no such interval line is found. Upon encountering any unexpected format errors in reading the next interval from the file, this method will return None. The histogram returned will have it's timestamp set to the absolute timestamp calculated from adding the interval's indicated timestamp value to the latest [optional] start time found in the log. Exceptions: ValueError if there is a syntax error in one of the float fields ''' return self._decode_next_interval_histogram(None, range_start_time_sec, range_end_time_sec, absolute)
Read the next interval histogram from the log, if interval falls within an absolute or relative time range. Timestamps are assumed to appear in order in the log file, and as such this method will return a null upon encountering a timestamp larger than range_end_time_sec. Relative time range: the range is assumed to be in seconds relative to the actual timestamp value found in each interval line in the log Absolute time range: Absolute timestamps are calculated by adding the timestamp found with the recorded interval to the [latest, optional] start time found in the log. The start time is indicated in the log with a "#[StartTime: " followed by the start time in seconds. Params: range_start_time_sec The absolute or relative start of the expected time range, in seconds. range_start_time_sec The absolute or relative end of the expected time range, in seconds. absolute Defines if the passed range is absolute or relative Return: Returns an histogram object if an interval line was found with an associated start timestamp value that falls between start_time_sec and end_time_sec, or null if no such interval line is found. Upon encountering any unexpected format errors in reading the next interval from the file, this method will return None. The histogram returned will have it's timestamp set to the absolute timestamp calculated from adding the interval's indicated timestamp value to the latest [optional] start time found in the log. Exceptions: ValueError if there is a syntax error in one of the float fields
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/log.py#L291-L337
[ "def _decode_next_interval_histogram(self,\n dest_histogram,\n range_start_time_sec=0.0,\n range_end_time_sec=sys.maxsize,\n absolute=False):\n '''Read the next interval histogram from the log, if interval falls\n within an absolute or relative time range.\n\n Timestamps are assumed to appear in order in the log file, and as such\n this method will return a null upon encountering a timestamp larger than\n range_end_time_sec.\n\n Relative time range:\n the range is assumed to be in seconds relative to\n the actual timestamp value found in each interval line in the log\n Absolute time range:\n Absolute timestamps are calculated by adding the timestamp found\n with the recorded interval to the [latest, optional] start time\n found in the log. The start time is indicated in the log with\n a \"#[StartTime: \" followed by the start time in seconds.\n\n Params:\n dest_histogram if None, created a new histogram, else adds\n the new interval histogram to it\n range_start_time_sec The absolute or relative start of the expected\n time range, in seconds.\n range_start_time_sec The absolute or relative end of the expected\n time range, in seconds.\n absolute Defines if the passed range is absolute or relative\n\n Return:\n Returns an histogram object if an interval line was found with an\n associated start timestamp value that falls between start_time_sec and\n end_time_sec,\n or null if no such interval line is found.\n Upon encountering any unexpected format errors in reading the next\n interval from the file, this method will return None.\n\n The histogram returned will have it's timestamp set to the absolute\n timestamp calculated from adding the interval's indicated timestamp\n value to the latest [optional] start time found in the log.\n\n Exceptions:\n ValueError if there is a syntax error in one of the float fields\n '''\n while 1:\n line = self.input_file.readline()\n if not line:\n return None\n if line[0] == '#':\n match_res = re_start_time.match(line)\n if match_res:\n self.start_time_sec = float(match_res.group(1))\n self.observed_start_time = True\n continue\n match_res = re_base_time.match(line)\n if match_res:\n self.base_time_sec = float(match_res.group(1))\n self.observed_base_time = True\n continue\n\n match_res = re_histogram_interval.match(line)\n if not match_res:\n # probably a legend line that starts with \"\\\"StartTimestamp\"\n continue\n # Decode: startTimestamp, intervalLength, maxTime, histogramPayload\n # Timestamp is expected to be in seconds\n log_time_stamp_in_sec = float(match_res.group(1))\n interval_length_sec = float(match_res.group(2))\n cpayload = match_res.group(4)\n\n if not self.observed_start_time:\n # No explicit start time noted. Use 1st observed time:\n self.start_time_sec = log_time_stamp_in_sec\n self.observed_start_time = True\n\n if not self.observed_base_time:\n # No explicit base time noted.\n # Deduce from 1st observed time (compared to start time):\n if log_time_stamp_in_sec < self.start_time_sec - (365 * 24 * 3600.0):\n # Criteria Note: if log timestamp is more than a year in\n # the past (compared to StartTime),\n # we assume that timestamps in the log are not absolute\n self.base_time_sec = self.start_time_sec\n else:\n # Timestamps are absolute\n self.base_time_sec = 0.0\n self.observed_base_time = True\n\n absolute_start_time_stamp_sec = \\\n log_time_stamp_in_sec + self.base_time_sec\n offset_start_time_stamp_sec = \\\n absolute_start_time_stamp_sec - self.start_time_sec\n\n # Timestamp length is expect to be in seconds\n absolute_end_time_stamp_sec = \\\n absolute_start_time_stamp_sec + interval_length_sec\n\n if absolute:\n start_time_stamp_to_check_range_on = absolute_start_time_stamp_sec\n else:\n start_time_stamp_to_check_range_on = offset_start_time_stamp_sec\n\n if start_time_stamp_to_check_range_on < range_start_time_sec:\n continue\n\n if start_time_stamp_to_check_range_on > range_end_time_sec:\n return None\n if dest_histogram:\n # add the interval histogram to the destination histogram\n histogram = dest_histogram\n histogram.decode_and_add(cpayload)\n else:\n histogram = HdrHistogram.decode(cpayload)\n histogram.set_start_time_stamp(absolute_start_time_stamp_sec * 1000.0)\n histogram.set_end_time_stamp(absolute_end_time_stamp_sec * 1000.0)\n return histogram\n" ]
class HistogramLogReader(): def __init__(self, input_file_name, reference_histogram): '''Constructs a new HistogramLogReader that produces intervals read from the specified file name. Params: input_file_name The name of the file to read from reference_histogram a histogram instance used as a reference to create new instances for all subsequent decoded interval histograms ''' self.start_time_sec = 0.0 self.observed_start_time = False self.base_time_sec = 0.0 self.observed_base_time = False self.input_file = open(input_file_name, "r") self.reference_histogram = reference_histogram def get_start_time_sec(self): '''get the latest start time found in the file so far (or 0.0), per the log file format explained above. Assuming the "#[StartTime:" comment line precedes the actual intervals recorded in the file, getStartTimeSec() can be safely used after each interval is read to determine's the offset of that interval's timestamp from the epoch. Return: latest Start Time found in the file (or 0.0 if non found) ''' return self.start_time_sec def _decode_next_interval_histogram(self, dest_histogram, range_start_time_sec=0.0, range_end_time_sec=sys.maxsize, absolute=False): '''Read the next interval histogram from the log, if interval falls within an absolute or relative time range. Timestamps are assumed to appear in order in the log file, and as such this method will return a null upon encountering a timestamp larger than range_end_time_sec. Relative time range: the range is assumed to be in seconds relative to the actual timestamp value found in each interval line in the log Absolute time range: Absolute timestamps are calculated by adding the timestamp found with the recorded interval to the [latest, optional] start time found in the log. The start time is indicated in the log with a "#[StartTime: " followed by the start time in seconds. Params: dest_histogram if None, created a new histogram, else adds the new interval histogram to it range_start_time_sec The absolute or relative start of the expected time range, in seconds. range_start_time_sec The absolute or relative end of the expected time range, in seconds. absolute Defines if the passed range is absolute or relative Return: Returns an histogram object if an interval line was found with an associated start timestamp value that falls between start_time_sec and end_time_sec, or null if no such interval line is found. Upon encountering any unexpected format errors in reading the next interval from the file, this method will return None. The histogram returned will have it's timestamp set to the absolute timestamp calculated from adding the interval's indicated timestamp value to the latest [optional] start time found in the log. Exceptions: ValueError if there is a syntax error in one of the float fields ''' while 1: line = self.input_file.readline() if not line: return None if line[0] == '#': match_res = re_start_time.match(line) if match_res: self.start_time_sec = float(match_res.group(1)) self.observed_start_time = True continue match_res = re_base_time.match(line) if match_res: self.base_time_sec = float(match_res.group(1)) self.observed_base_time = True continue match_res = re_histogram_interval.match(line) if not match_res: # probably a legend line that starts with "\"StartTimestamp" continue # Decode: startTimestamp, intervalLength, maxTime, histogramPayload # Timestamp is expected to be in seconds log_time_stamp_in_sec = float(match_res.group(1)) interval_length_sec = float(match_res.group(2)) cpayload = match_res.group(4) if not self.observed_start_time: # No explicit start time noted. Use 1st observed time: self.start_time_sec = log_time_stamp_in_sec self.observed_start_time = True if not self.observed_base_time: # No explicit base time noted. # Deduce from 1st observed time (compared to start time): if log_time_stamp_in_sec < self.start_time_sec - (365 * 24 * 3600.0): # Criteria Note: if log timestamp is more than a year in # the past (compared to StartTime), # we assume that timestamps in the log are not absolute self.base_time_sec = self.start_time_sec else: # Timestamps are absolute self.base_time_sec = 0.0 self.observed_base_time = True absolute_start_time_stamp_sec = \ log_time_stamp_in_sec + self.base_time_sec offset_start_time_stamp_sec = \ absolute_start_time_stamp_sec - self.start_time_sec # Timestamp length is expect to be in seconds absolute_end_time_stamp_sec = \ absolute_start_time_stamp_sec + interval_length_sec if absolute: start_time_stamp_to_check_range_on = absolute_start_time_stamp_sec else: start_time_stamp_to_check_range_on = offset_start_time_stamp_sec if start_time_stamp_to_check_range_on < range_start_time_sec: continue if start_time_stamp_to_check_range_on > range_end_time_sec: return None if dest_histogram: # add the interval histogram to the destination histogram histogram = dest_histogram histogram.decode_and_add(cpayload) else: histogram = HdrHistogram.decode(cpayload) histogram.set_start_time_stamp(absolute_start_time_stamp_sec * 1000.0) histogram.set_end_time_stamp(absolute_end_time_stamp_sec * 1000.0) return histogram def add_next_interval_histogram(self, dest_histogram=None, range_start_time_sec=0.0, range_end_time_sec=sys.maxsize, absolute=False): '''Read the next interval histogram from the log, if interval falls within an absolute or relative time range, and add it to the destination histogram (or to the reference histogram if dest_histogram is None) Timestamps are assumed to appear in order in the log file, and as such this method will return a null upon encountering a timestamp larger than range_end_time_sec. Relative time range: the range is assumed to be in seconds relative to the actual timestamp value found in each interval line in the log Absolute time range: Absolute timestamps are calculated by adding the timestamp found with the recorded interval to the [latest, optional] start time found in the log. The start time is indicated in the log with a "#[StartTime: " followed by the start time in seconds. Params: dest_histogram where to add the next interval histogram, if None the interal histogram will be added to the reference histogram passed in the constructor range_start_time_sec The absolute or relative start of the expected time range, in seconds. range_start_time_sec The absolute or relative end of the expected time range, in seconds. absolute Defines if the passed range is absolute or relative Return: Returns the destination histogram if an interval line was found with an associated start timestamp value that falls between start_time_sec and end_time_sec, or None if no such interval line is found. Upon encountering any unexpected format errors in reading the next interval from the file, this method will return None. The histogram returned will have it's timestamp set to the absolute timestamp calculated from adding the interval's indicated timestamp value to the latest [optional] start time found in the log. Exceptions: ValueError if there is a syntax error in one of the float fields ''' if not dest_histogram: dest_histogram = self.reference_histogram return self._decode_next_interval_histogram(dest_histogram, range_start_time_sec, range_end_time_sec, absolute) def close(self): self.input_file.close()
HdrHistogram/HdrHistogram_py
hdrh/log.py
HistogramLogReader.add_next_interval_histogram
python
def add_next_interval_histogram(self, dest_histogram=None, range_start_time_sec=0.0, range_end_time_sec=sys.maxsize, absolute=False): '''Read the next interval histogram from the log, if interval falls within an absolute or relative time range, and add it to the destination histogram (or to the reference histogram if dest_histogram is None) Timestamps are assumed to appear in order in the log file, and as such this method will return a null upon encountering a timestamp larger than range_end_time_sec. Relative time range: the range is assumed to be in seconds relative to the actual timestamp value found in each interval line in the log Absolute time range: Absolute timestamps are calculated by adding the timestamp found with the recorded interval to the [latest, optional] start time found in the log. The start time is indicated in the log with a "#[StartTime: " followed by the start time in seconds. Params: dest_histogram where to add the next interval histogram, if None the interal histogram will be added to the reference histogram passed in the constructor range_start_time_sec The absolute or relative start of the expected time range, in seconds. range_start_time_sec The absolute or relative end of the expected time range, in seconds. absolute Defines if the passed range is absolute or relative Return: Returns the destination histogram if an interval line was found with an associated start timestamp value that falls between start_time_sec and end_time_sec, or None if no such interval line is found. Upon encountering any unexpected format errors in reading the next interval from the file, this method will return None. The histogram returned will have it's timestamp set to the absolute timestamp calculated from adding the interval's indicated timestamp value to the latest [optional] start time found in the log. Exceptions: ValueError if there is a syntax error in one of the float fields ''' if not dest_histogram: dest_histogram = self.reference_histogram return self._decode_next_interval_histogram(dest_histogram, range_start_time_sec, range_end_time_sec, absolute)
Read the next interval histogram from the log, if interval falls within an absolute or relative time range, and add it to the destination histogram (or to the reference histogram if dest_histogram is None) Timestamps are assumed to appear in order in the log file, and as such this method will return a null upon encountering a timestamp larger than range_end_time_sec. Relative time range: the range is assumed to be in seconds relative to the actual timestamp value found in each interval line in the log Absolute time range: Absolute timestamps are calculated by adding the timestamp found with the recorded interval to the [latest, optional] start time found in the log. The start time is indicated in the log with a "#[StartTime: " followed by the start time in seconds. Params: dest_histogram where to add the next interval histogram, if None the interal histogram will be added to the reference histogram passed in the constructor range_start_time_sec The absolute or relative start of the expected time range, in seconds. range_start_time_sec The absolute or relative end of the expected time range, in seconds. absolute Defines if the passed range is absolute or relative Return: Returns the destination histogram if an interval line was found with an associated start timestamp value that falls between start_time_sec and end_time_sec, or None if no such interval line is found. Upon encountering any unexpected format errors in reading the next interval from the file, this method will return None. The histogram returned will have it's timestamp set to the absolute timestamp calculated from adding the interval's indicated timestamp value to the latest [optional] start time found in the log. Exceptions: ValueError if there is a syntax error in one of the float fields
train
https://github.com/HdrHistogram/HdrHistogram_py/blob/cb99981b0564a62e1aa02bd764efa6445923f8f7/hdrh/log.py#L339-L391
[ "def _decode_next_interval_histogram(self,\n dest_histogram,\n range_start_time_sec=0.0,\n range_end_time_sec=sys.maxsize,\n absolute=False):\n '''Read the next interval histogram from the log, if interval falls\n within an absolute or relative time range.\n\n Timestamps are assumed to appear in order in the log file, and as such\n this method will return a null upon encountering a timestamp larger than\n range_end_time_sec.\n\n Relative time range:\n the range is assumed to be in seconds relative to\n the actual timestamp value found in each interval line in the log\n Absolute time range:\n Absolute timestamps are calculated by adding the timestamp found\n with the recorded interval to the [latest, optional] start time\n found in the log. The start time is indicated in the log with\n a \"#[StartTime: \" followed by the start time in seconds.\n\n Params:\n dest_histogram if None, created a new histogram, else adds\n the new interval histogram to it\n range_start_time_sec The absolute or relative start of the expected\n time range, in seconds.\n range_start_time_sec The absolute or relative end of the expected\n time range, in seconds.\n absolute Defines if the passed range is absolute or relative\n\n Return:\n Returns an histogram object if an interval line was found with an\n associated start timestamp value that falls between start_time_sec and\n end_time_sec,\n or null if no such interval line is found.\n Upon encountering any unexpected format errors in reading the next\n interval from the file, this method will return None.\n\n The histogram returned will have it's timestamp set to the absolute\n timestamp calculated from adding the interval's indicated timestamp\n value to the latest [optional] start time found in the log.\n\n Exceptions:\n ValueError if there is a syntax error in one of the float fields\n '''\n while 1:\n line = self.input_file.readline()\n if not line:\n return None\n if line[0] == '#':\n match_res = re_start_time.match(line)\n if match_res:\n self.start_time_sec = float(match_res.group(1))\n self.observed_start_time = True\n continue\n match_res = re_base_time.match(line)\n if match_res:\n self.base_time_sec = float(match_res.group(1))\n self.observed_base_time = True\n continue\n\n match_res = re_histogram_interval.match(line)\n if not match_res:\n # probably a legend line that starts with \"\\\"StartTimestamp\"\n continue\n # Decode: startTimestamp, intervalLength, maxTime, histogramPayload\n # Timestamp is expected to be in seconds\n log_time_stamp_in_sec = float(match_res.group(1))\n interval_length_sec = float(match_res.group(2))\n cpayload = match_res.group(4)\n\n if not self.observed_start_time:\n # No explicit start time noted. Use 1st observed time:\n self.start_time_sec = log_time_stamp_in_sec\n self.observed_start_time = True\n\n if not self.observed_base_time:\n # No explicit base time noted.\n # Deduce from 1st observed time (compared to start time):\n if log_time_stamp_in_sec < self.start_time_sec - (365 * 24 * 3600.0):\n # Criteria Note: if log timestamp is more than a year in\n # the past (compared to StartTime),\n # we assume that timestamps in the log are not absolute\n self.base_time_sec = self.start_time_sec\n else:\n # Timestamps are absolute\n self.base_time_sec = 0.0\n self.observed_base_time = True\n\n absolute_start_time_stamp_sec = \\\n log_time_stamp_in_sec + self.base_time_sec\n offset_start_time_stamp_sec = \\\n absolute_start_time_stamp_sec - self.start_time_sec\n\n # Timestamp length is expect to be in seconds\n absolute_end_time_stamp_sec = \\\n absolute_start_time_stamp_sec + interval_length_sec\n\n if absolute:\n start_time_stamp_to_check_range_on = absolute_start_time_stamp_sec\n else:\n start_time_stamp_to_check_range_on = offset_start_time_stamp_sec\n\n if start_time_stamp_to_check_range_on < range_start_time_sec:\n continue\n\n if start_time_stamp_to_check_range_on > range_end_time_sec:\n return None\n if dest_histogram:\n # add the interval histogram to the destination histogram\n histogram = dest_histogram\n histogram.decode_and_add(cpayload)\n else:\n histogram = HdrHistogram.decode(cpayload)\n histogram.set_start_time_stamp(absolute_start_time_stamp_sec * 1000.0)\n histogram.set_end_time_stamp(absolute_end_time_stamp_sec * 1000.0)\n return histogram\n" ]
class HistogramLogReader(): def __init__(self, input_file_name, reference_histogram): '''Constructs a new HistogramLogReader that produces intervals read from the specified file name. Params: input_file_name The name of the file to read from reference_histogram a histogram instance used as a reference to create new instances for all subsequent decoded interval histograms ''' self.start_time_sec = 0.0 self.observed_start_time = False self.base_time_sec = 0.0 self.observed_base_time = False self.input_file = open(input_file_name, "r") self.reference_histogram = reference_histogram def get_start_time_sec(self): '''get the latest start time found in the file so far (or 0.0), per the log file format explained above. Assuming the "#[StartTime:" comment line precedes the actual intervals recorded in the file, getStartTimeSec() can be safely used after each interval is read to determine's the offset of that interval's timestamp from the epoch. Return: latest Start Time found in the file (or 0.0 if non found) ''' return self.start_time_sec def _decode_next_interval_histogram(self, dest_histogram, range_start_time_sec=0.0, range_end_time_sec=sys.maxsize, absolute=False): '''Read the next interval histogram from the log, if interval falls within an absolute or relative time range. Timestamps are assumed to appear in order in the log file, and as such this method will return a null upon encountering a timestamp larger than range_end_time_sec. Relative time range: the range is assumed to be in seconds relative to the actual timestamp value found in each interval line in the log Absolute time range: Absolute timestamps are calculated by adding the timestamp found with the recorded interval to the [latest, optional] start time found in the log. The start time is indicated in the log with a "#[StartTime: " followed by the start time in seconds. Params: dest_histogram if None, created a new histogram, else adds the new interval histogram to it range_start_time_sec The absolute or relative start of the expected time range, in seconds. range_start_time_sec The absolute or relative end of the expected time range, in seconds. absolute Defines if the passed range is absolute or relative Return: Returns an histogram object if an interval line was found with an associated start timestamp value that falls between start_time_sec and end_time_sec, or null if no such interval line is found. Upon encountering any unexpected format errors in reading the next interval from the file, this method will return None. The histogram returned will have it's timestamp set to the absolute timestamp calculated from adding the interval's indicated timestamp value to the latest [optional] start time found in the log. Exceptions: ValueError if there is a syntax error in one of the float fields ''' while 1: line = self.input_file.readline() if not line: return None if line[0] == '#': match_res = re_start_time.match(line) if match_res: self.start_time_sec = float(match_res.group(1)) self.observed_start_time = True continue match_res = re_base_time.match(line) if match_res: self.base_time_sec = float(match_res.group(1)) self.observed_base_time = True continue match_res = re_histogram_interval.match(line) if not match_res: # probably a legend line that starts with "\"StartTimestamp" continue # Decode: startTimestamp, intervalLength, maxTime, histogramPayload # Timestamp is expected to be in seconds log_time_stamp_in_sec = float(match_res.group(1)) interval_length_sec = float(match_res.group(2)) cpayload = match_res.group(4) if not self.observed_start_time: # No explicit start time noted. Use 1st observed time: self.start_time_sec = log_time_stamp_in_sec self.observed_start_time = True if not self.observed_base_time: # No explicit base time noted. # Deduce from 1st observed time (compared to start time): if log_time_stamp_in_sec < self.start_time_sec - (365 * 24 * 3600.0): # Criteria Note: if log timestamp is more than a year in # the past (compared to StartTime), # we assume that timestamps in the log are not absolute self.base_time_sec = self.start_time_sec else: # Timestamps are absolute self.base_time_sec = 0.0 self.observed_base_time = True absolute_start_time_stamp_sec = \ log_time_stamp_in_sec + self.base_time_sec offset_start_time_stamp_sec = \ absolute_start_time_stamp_sec - self.start_time_sec # Timestamp length is expect to be in seconds absolute_end_time_stamp_sec = \ absolute_start_time_stamp_sec + interval_length_sec if absolute: start_time_stamp_to_check_range_on = absolute_start_time_stamp_sec else: start_time_stamp_to_check_range_on = offset_start_time_stamp_sec if start_time_stamp_to_check_range_on < range_start_time_sec: continue if start_time_stamp_to_check_range_on > range_end_time_sec: return None if dest_histogram: # add the interval histogram to the destination histogram histogram = dest_histogram histogram.decode_and_add(cpayload) else: histogram = HdrHistogram.decode(cpayload) histogram.set_start_time_stamp(absolute_start_time_stamp_sec * 1000.0) histogram.set_end_time_stamp(absolute_end_time_stamp_sec * 1000.0) return histogram def get_next_interval_histogram(self, range_start_time_sec=0.0, range_end_time_sec=sys.maxsize, absolute=False): '''Read the next interval histogram from the log, if interval falls within an absolute or relative time range. Timestamps are assumed to appear in order in the log file, and as such this method will return a null upon encountering a timestamp larger than range_end_time_sec. Relative time range: the range is assumed to be in seconds relative to the actual timestamp value found in each interval line in the log Absolute time range: Absolute timestamps are calculated by adding the timestamp found with the recorded interval to the [latest, optional] start time found in the log. The start time is indicated in the log with a "#[StartTime: " followed by the start time in seconds. Params: range_start_time_sec The absolute or relative start of the expected time range, in seconds. range_start_time_sec The absolute or relative end of the expected time range, in seconds. absolute Defines if the passed range is absolute or relative Return: Returns an histogram object if an interval line was found with an associated start timestamp value that falls between start_time_sec and end_time_sec, or null if no such interval line is found. Upon encountering any unexpected format errors in reading the next interval from the file, this method will return None. The histogram returned will have it's timestamp set to the absolute timestamp calculated from adding the interval's indicated timestamp value to the latest [optional] start time found in the log. Exceptions: ValueError if there is a syntax error in one of the float fields ''' return self._decode_next_interval_histogram(None, range_start_time_sec, range_end_time_sec, absolute) def close(self): self.input_file.close()
dwkim78/upsilon
upsilon/utils/utils.py
sigma_clipping
python
def sigma_clipping(date, mag, err, threshold=3, iteration=1): # Check length. if (len(date) != len(mag)) \ or (len(date) != len(err)) \ or (len(mag) != len(err)): raise RuntimeError('The length of date, mag, and err must be same.') # By magnitudes for i in range(int(iteration)): mean = np.median(mag) std = np.std(mag) index = (mag >= mean - threshold*std) & (mag <= mean + threshold*std) date = date[index] mag = mag[index] err = err[index] return date, mag, err
Remove any fluctuated data points by magnitudes. Parameters ---------- date : array_like An array of dates. mag : array_like An array of magnitudes. err : array_like An array of magnitude errors. threshold : float, optional Threshold for sigma-clipping. iteration : int, optional The number of iteration. Returns ------- date : array_like Sigma-clipped dates. mag : array_like Sigma-clipped magnitudes. err : array_like Sigma-clipped magnitude errors.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/utils/utils.py#L4-L47
null
import numpy as np
dwkim78/upsilon
upsilon/extract_features/feature_set.py
get_feature_set_all
python
def get_feature_set_all(): features = get_feature_set() features.append('cusum') features.append('eta') features.append('n_points') features.append('period_SNR') features.append('period_log10FAP') features.append('period_uncertainty') features.append('weighted_mean') features.append('weighted_std') features.sort() return features
Return a list of entire features. A set of entire features regardless of being used to train a model or predict a class. Returns ------- feature_names : list A list of features' names.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/feature_set.py#L23-L49
[ "def get_feature_set():\n \"\"\"\n Return a list of features' names.\n\n Features' name that are used to train a model and predict a class.\n Sorted by the names.\n\n Returns\n -------\n feature_names : list\n A list of features' names.\n \"\"\"\n\n features = ['amplitude', 'hl_amp_ratio', 'kurtosis', 'period',\n 'phase_cusum', 'phase_eta', 'phi21', 'phi31', 'quartile31',\n 'r21', 'r31', 'shapiro_w', 'skewness', 'slope_per10',\n 'slope_per90', 'stetson_k']\n features.sort()\n\n return features\n" ]
def get_feature_set(): """ Return a list of features' names. Features' name that are used to train a model and predict a class. Sorted by the names. Returns ------- feature_names : list A list of features' names. """ features = ['amplitude', 'hl_amp_ratio', 'kurtosis', 'period', 'phase_cusum', 'phase_eta', 'phi21', 'phi31', 'quartile31', 'r21', 'r31', 'shapiro_w', 'skewness', 'slope_per10', 'slope_per90', 'stetson_k'] features.sort() return features if __name__ == '__main__': print(get_feature_set())
dwkim78/upsilon
upsilon/extract_features/extract_features.py
ExtractFeatures.shallow_run
python
def shallow_run(self): # Number of data points self.n_points = len(self.date) # Weight calculation. # All zero values. if not self.err.any(): self.err = np.ones(len(self.mag)) * np.std(self.mag) # Some zero values. elif not self.err.all(): np.putmask(self.err, self.err==0, np.median(self.err)) self.weight = 1. / self.err self.weighted_sum = np.sum(self.weight) # Simple statistics, mean, median and std. self.mean = np.mean(self.mag) self.median = np.median(self.mag) self.std = np.std(self.mag) # Weighted mean and std. self.weighted_mean = np.sum(self.mag * self.weight) / self.weighted_sum self.weighted_std = np.sqrt(np.sum((self.mag - self.weighted_mean) ** 2 \ * self.weight) / self.weighted_sum) # Skewness and kurtosis. self.skewness = ss.skew(self.mag) self.kurtosis = ss.kurtosis(self.mag) # Normalization-test. Shapiro-Wilk test. shapiro = ss.shapiro(self.mag) self.shapiro_w = shapiro[0] # self.shapiro_log10p = np.log10(shapiro[1]) # Percentile features. self.quartile31 = np.percentile(self.mag, 75) \ - np.percentile(self.mag, 25) # Stetson K. self.stetson_k = self.get_stetson_k(self.mag, self.median, self.err) # Ratio between higher and lower amplitude than average. self.hl_amp_ratio = self.half_mag_amplitude_ratio( self.mag, self.median, self.weight) # This second function's value is very similar with the above one. # self.hl_amp_ratio2 = self.half_mag_amplitude_ratio2( # self.mag, self.median) # Cusum self.cusum = self.get_cusum(self.mag) # Eta self.eta = self.get_eta(self.mag, self.weighted_std)
Derive not-period-based features.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/extract_features.py#L86-L139
null
class ExtractFeatures: """ Extract variability features of a light curve. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. err : array_like, optional An array of magnitude error. If None, std(mag) will be used. n_threads : int, optional The number of cores to use to derive periods. min_period : float, optional The minimum period to calculate. """ def __init__(self, date, mag, err=None, n_threads=4, min_period=0.03): # Set basic values. if not isinstance(date, np.ndarray): date = np.array(date) if not isinstance(mag, np.ndarray): mag = np.array(mag) self.date = date self.mag = mag if err is not None: if not isinstance(err, np.ndarray): err = np.array(err) self.err = err else: self.err = np.ones(len(self.mag)) * np.std(self.mag) # Check length. if (len(self.date) != len(self.mag)) or \ (len(self.date) != len(self.err)) or \ (len(self.mag) != len(self.err)): raise RuntimeError('The length of date, mag, and err must be same.') # if the number of data points is too small. min_n_data = 80 if len(self.date) < min_n_data: warnings.warn('The number of data points are less than %d.' % min_n_data) n_threads = int(n_threads) if n_threads > multiprocessing.cpu_count(): self.n_threads = multiprocessing.cpu_count() else: if n_threads <= 0: self.n_threads = 1 else: self.n_threads = n_threads min_period = float(min_period) if min_period <= 0: self.min_period = 0.03 else: self.min_period = min_period def run(self): """Run feature extraction modules.""" # shallow_run must be executed prior to deep_run # since shallow_run calculates several values needed for deep_run. self.shallow_run() self.deep_run() def deep_run(self): """Derive period-based features.""" # Lomb-Scargle period finding. self.get_period_LS(self.date, self.mag, self.n_threads, self.min_period) # Features based on a phase-folded light curve # such as Eta, slope-percentile, etc. # Should be called after the getPeriodLS() is called. # Created phased a folded light curve. # We use period * 2 to take eclipsing binaries into account. phase_folded_date = self.date % (self.period * 2.) sorted_index = np.argsort(phase_folded_date) folded_date = phase_folded_date[sorted_index] folded_mag = self.mag[sorted_index] # phase Eta self.phase_eta = self.get_eta(folded_mag, self.weighted_std) # Slope percentile. self.slope_per10, self.slope_per90 = \ self.slope_percentile(folded_date, folded_mag) # phase Cusum self.phase_cusum = self.get_cusum(folded_mag) def get_period_LS(self, date, mag, n_threads, min_period): """ Period finding using the Lomb-Scargle algorithm. Finding two periods. The second period is estimated after whitening the first period. Calculating various other features as well using derived periods. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. n_threads : int The number of threads to use. min_period : float The minimum period to calculate. """ # DO NOT CHANGE THESE PARAMETERS. oversampling = 3. hifac = int((max(date) - min(date)) / len(date) / min_period * 2.) # Minimum hifac if hifac < 100: hifac = 100 # Lomb-Scargle. fx, fy, nout, jmax, prob = pLS.fasper(date, mag, oversampling, hifac, n_threads) self.f = fx[jmax] self.period = 1. / self.f self.period_uncertainty = self.get_period_uncertainty(fx, fy, jmax) self.period_log10FAP = \ np.log10(pLS.getSignificance(fx, fy, nout, oversampling)[jmax]) # self.f_SNR1 = fy[jmax] / np.median(fy) self.period_SNR = (fy[jmax] - np.median(fy)) / np.std(fy) # Fit Fourier Series of order 3. order = 3 # Initial guess of Fourier coefficients. p0 = np.ones(order * 2 + 1) date_period = (date % self.period) / self.period p1, success = leastsq(self.residuals, p0, args=(date_period, mag, order)) # fitted_y = self.FourierSeries(p1, date_period, order) # print p1, self.mean, self.median # plt.plot(date_period, self.mag, 'b+') # plt.show() # Derive Fourier features for the first period. # Petersen, J. O., 1986, A&A self.amplitude = np.sqrt(p1[1] ** 2 + p1[2] ** 2) self.r21 = np.sqrt(p1[3] ** 2 + p1[4] ** 2) / self.amplitude self.r31 = np.sqrt(p1[5] ** 2 + p1[6] ** 2) / self.amplitude self.f_phase = np.arctan(-p1[1] / p1[2]) self.phi21 = np.arctan(-p1[3] / p1[4]) - 2. * self.f_phase self.phi31 = np.arctan(-p1[5] / p1[6]) - 3. * self.f_phase """ # Derive a second period. # Whitening a light curve. residual_mag = mag - fitted_y # Lomb-Scargle again to find the second period. omega_top, power_top = search_frequencies(date, residual_mag, err, #LS_kwargs={'generalized':True, 'subtract_mean':True}, n_eval=5000, n_retry=3, n_save=50) self.period2 = 2*np.pi/omega_top[np.where(power_top==np.max(power_top))][0] self.f2 = 1. / self.period2 self.f2_SNR = power_top[np.where(power_top==np.max(power_top))][0] \ * (len(self.date) - 1) / 2. # Fit Fourier Series again. p0 = [1.] * order * 2 date_period = (date % self.period) / self.period p2, success = leastsq(self.residuals, p0, args=(date_period, residual_mag, order)) fitted_y = self.FourierSeries(p2, date_period, order) #plt.plot(date%self.period2, residual_mag, 'b+') #plt.show() # Derive Fourier features for the first second. self.f2_amp = 2. * np.sqrt(p2[1]**2 + p2[2]**2) self.f2_R21 = np.sqrt(p2[3]**2 + p2[4]**2) / self.f2_amp self.f2_R31 = np.sqrt(p2[5]**2 + p2[6]**2) / self.f2_amp self.f2_R41 = np.sqrt(p2[7]**2 + p2[8]**2) / self.f2_amp self.f2_R51 = np.sqrt(p2[9]**2 + p2[10]**2) / self.f2_amp self.f2_phase = np.arctan(-p2[1] / p2[2]) self.f2_phi21 = np.arctan(-p2[3] / p2[4]) - 2. * self.f2_phase self.f2_phi31 = np.arctan(-p2[5] / p2[6]) - 3. * self.f2_phase self.f2_phi41 = np.arctan(-p2[7] / p2[8]) - 4. * self.f2_phase self.f2_phi51 = np.arctan(-p2[9] / p2[10]) - 5. * self.f2_phase # Calculate features using the first and second periods. self.f12_ratio = self.f2 / self.f1 self.f12_remain = self.f1 % self.f2 \ if self.f1 > self.f2 else self.f2 % self.f1 self.f12_amp = self.f2_amp / self.f1_amp self.f12_phase = self.f2_phase - self.f1_phase """ def get_period_uncertainty(self, fx, fy, jmax, fx_width=100): """ Get uncertainty of a period. The uncertainty is defined as the half width of the frequencies around the peak, that becomes lower than average + standard deviation of the power spectrum. Since we may not have fine resolution around the peak, we do not assume it is gaussian. So, no scaling factor of 2.355 (= 2 * sqrt(2 * ln2)) is applied. Parameters ---------- fx : array_like An array of frequencies. fy : array_like An array of amplitudes. jmax : int An index at the peak frequency. fx_width : int, optional Width of power spectrum to calculate uncertainty. Returns ------- p_uncertain : float Period uncertainty. """ # Get subset start_index = jmax - fx_width end_index = jmax + fx_width if start_index < 0: start_index = 0 if end_index > len(fx) - 1: end_index = len(fx) - 1 fx_subset = fx[start_index:end_index] fy_subset = fy[start_index:end_index] fy_mean = np.median(fy_subset) fy_std = np.std(fy_subset) # Find peak max_index = np.argmax(fy_subset) # Find list whose powers become lower than average + std. index = np.where(fy_subset <= fy_mean + fy_std)[0] # Find the edge at left and right. This is the full width. left_index = index[(index < max_index)] if len(left_index) == 0: left_index = 0 else: left_index = left_index[-1] right_index = index[(index > max_index)] if len(right_index) == 0: right_index = len(fy_subset) - 1 else: right_index = right_index[0] # We assume the half of the full width is the period uncertainty. half_width = (1. / fx_subset[left_index] - 1. / fx_subset[right_index]) / 2. period_uncertainty = half_width return period_uncertainty def residuals(self, pars, x, y, order): """ Residual of Fourier Series. Parameters ---------- pars : array_like Fourier series parameters. x : array_like An array of date. y : array_like An array of true values to fit. order : int An order of Fourier Series. """ return y - self.fourier_series(pars, x, order) def fourier_series(self, pars, x, order): """ Function to fit Fourier Series. Parameters ---------- x : array_like An array of date divided by period. It doesn't need to be sorted. pars : array_like Fourier series parameters. order : int An order of Fourier series. """ sum = pars[0] for i in range(order): sum += pars[i * 2 + 1] * np.sin(2 * np.pi * (i + 1) * x) \ + pars[i * 2 + 2] * np.cos(2 * np.pi * (i + 1) * x) return sum def get_stetson_k(self, mag, avg, err): """ Return Stetson K feature. Parameters ---------- mag : array_like An array of magnitude. avg : float An average value of magnitudes. err : array_like An array of magnitude errors. Returns ------- stetson_k : float Stetson K value. """ residual = (mag - avg) / err stetson_k = np.sum(np.fabs(residual)) \ / np.sqrt(np.sum(residual * residual)) / np.sqrt(len(mag)) return stetson_k def half_mag_amplitude_ratio(self, mag, avg, weight): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. weight : array_like An array of weight. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) lower_weight = weight[index] lower_weight_sum = np.sum(lower_weight) lower_mag = mag[index] lower_weighted_std = np.sum((lower_mag - avg) ** 2 * lower_weight) / \ lower_weight_sum # For higher (brighter) magnitude than average. index = np.where(mag <= avg) higher_weight = weight[index] higher_weight_sum = np.sum(higher_weight) higher_mag = mag[index] higher_weighted_std = np.sum((higher_mag - avg) ** 2 * higher_weight) / \ higher_weight_sum # Return ratio. return np.sqrt(lower_weighted_std / higher_weighted_std) def half_mag_amplitude_ratio2(self, mag, avg): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) fainter_mag = mag[index] lower_sum = np.sum((fainter_mag - avg) ** 2) / len(fainter_mag) # For higher (brighter) magnitude than average. index = np.where(mag <= avg) brighter_mag = mag[index] higher_sum = np.sum((avg - brighter_mag) ** 2) / len(brighter_mag) # Return ratio. return np.sqrt(lower_sum / higher_sum) def get_eta(self, mag, std): """ Return Eta feature. Parameters ---------- mag : array_like An array of magnitudes. std : array_like A standard deviation of magnitudes. Returns ------- eta : float The value of Eta index. """ diff = mag[1:] - mag[:len(mag) - 1] eta = np.sum(diff * diff) / (len(mag) - 1.) / std / std return eta def slope_percentile(self, date, mag): """ Return 10% and 90% percentile of slope. Parameters ---------- date : array_like An array of phase-folded date. Sorted. mag : array_like An array of phase-folded magnitudes. Sorted by date. Returns ------- per_10 : float 10% percentile values of slope. per_90 : float 90% percentile values of slope. """ date_diff = date[1:] - date[:len(date) - 1] mag_diff = mag[1:] - mag[:len(mag) - 1] # Remove zero mag_diff. index = np.where(mag_diff != 0.) date_diff = date_diff[index] mag_diff = mag_diff[index] # Derive slope. slope = date_diff / mag_diff percentile_10 = np.percentile(slope, 10.) percentile_90 = np.percentile(slope, 90.) return percentile_10, percentile_90 def get_cusum(self, mag): """ Return max - min of cumulative sum. Parameters ---------- mag : array_like An array of magnitudes. Returns ------- mm_cusum : float Max - min of cumulative sum. """ c = np.cumsum(mag - self.weighted_mean) / len(mag) / self.weighted_std return np.max(c) - np.min(c) def get_features2(self): """ Return all features with its names. Returns ------- names : list Feature names. values : list Feature values """ feature_names = [] feature_values = [] # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): # Omit input variables such as date, mag, err, etc. if not (name == 'date' or name == 'mag' or name == 'err' or name == 'n_threads' or name == 'min_period'): # Filter some other unnecessary features. if not (name == 'f' or name == 'f_phase' or name == 'period_log10FAP' or name == 'weight' or name == 'weighted_sum' or name == 'median' or name == 'mean' or name == 'std'): feature_names.append(name) # Sort by the names. # Sorting should be done to keep maintaining the same order of features. feature_names.sort() # Get feature values. for name in feature_names: feature_values.append(all_vars[name]) return feature_names, feature_values def get_features(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ ''' features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features ''' return self.get_features_all() def get_features_all(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list_all: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features
dwkim78/upsilon
upsilon/extract_features/extract_features.py
ExtractFeatures.deep_run
python
def deep_run(self): # Lomb-Scargle period finding. self.get_period_LS(self.date, self.mag, self.n_threads, self.min_period) # Features based on a phase-folded light curve # such as Eta, slope-percentile, etc. # Should be called after the getPeriodLS() is called. # Created phased a folded light curve. # We use period * 2 to take eclipsing binaries into account. phase_folded_date = self.date % (self.period * 2.) sorted_index = np.argsort(phase_folded_date) folded_date = phase_folded_date[sorted_index] folded_mag = self.mag[sorted_index] # phase Eta self.phase_eta = self.get_eta(folded_mag, self.weighted_std) # Slope percentile. self.slope_per10, self.slope_per90 = \ self.slope_percentile(folded_date, folded_mag) # phase Cusum self.phase_cusum = self.get_cusum(folded_mag)
Derive period-based features.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/extract_features.py#L141-L166
[ "def get_period_LS(self, date, mag, n_threads, min_period):\n \"\"\"\n Period finding using the Lomb-Scargle algorithm.\n\n Finding two periods. The second period is estimated after whitening\n the first period. Calculating various other features as well\n using derived periods.\n\n Parameters\n ----------\n date : array_like\n An array of observed date, in days.\n mag : array_like\n An array of observed magnitude.\n n_threads : int\n The number of threads to use.\n min_period : float\n The minimum period to calculate.\n \"\"\"\n\n # DO NOT CHANGE THESE PARAMETERS.\n oversampling = 3.\n hifac = int((max(date) - min(date)) / len(date) / min_period * 2.)\n\n # Minimum hifac\n if hifac < 100:\n hifac = 100\n\n # Lomb-Scargle.\n fx, fy, nout, jmax, prob = pLS.fasper(date, mag, oversampling, hifac,\n n_threads)\n\n self.f = fx[jmax]\n self.period = 1. / self.f\n self.period_uncertainty = self.get_period_uncertainty(fx, fy, jmax)\n self.period_log10FAP = \\\n np.log10(pLS.getSignificance(fx, fy, nout, oversampling)[jmax])\n # self.f_SNR1 = fy[jmax] / np.median(fy)\n self.period_SNR = (fy[jmax] - np.median(fy)) / np.std(fy)\n\n # Fit Fourier Series of order 3.\n order = 3\n # Initial guess of Fourier coefficients.\n p0 = np.ones(order * 2 + 1)\n date_period = (date % self.period) / self.period\n p1, success = leastsq(self.residuals, p0,\n args=(date_period, mag, order))\n # fitted_y = self.FourierSeries(p1, date_period, order)\n\n # print p1, self.mean, self.median\n # plt.plot(date_period, self.mag, 'b+')\n # plt.show()\n\n # Derive Fourier features for the first period.\n # Petersen, J. O., 1986, A&A\n self.amplitude = np.sqrt(p1[1] ** 2 + p1[2] ** 2)\n self.r21 = np.sqrt(p1[3] ** 2 + p1[4] ** 2) / self.amplitude\n self.r31 = np.sqrt(p1[5] ** 2 + p1[6] ** 2) / self.amplitude\n self.f_phase = np.arctan(-p1[1] / p1[2])\n self.phi21 = np.arctan(-p1[3] / p1[4]) - 2. * self.f_phase\n self.phi31 = np.arctan(-p1[5] / p1[6]) - 3. * self.f_phase\n\n \"\"\"\n # Derive a second period.\n # Whitening a light curve.\n residual_mag = mag - fitted_y\n\n # Lomb-Scargle again to find the second period.\n omega_top, power_top = search_frequencies(date, residual_mag, err,\n #LS_kwargs={'generalized':True, 'subtract_mean':True},\n n_eval=5000, n_retry=3, n_save=50)\n\n self.period2 = 2*np.pi/omega_top[np.where(power_top==np.max(power_top))][0]\n self.f2 = 1. / self.period2\n self.f2_SNR = power_top[np.where(power_top==np.max(power_top))][0] \\\n * (len(self.date) - 1) / 2.\n\n # Fit Fourier Series again.\n p0 = [1.] * order * 2\n date_period = (date % self.period) / self.period\n p2, success = leastsq(self.residuals, p0,\n args=(date_period, residual_mag, order))\n fitted_y = self.FourierSeries(p2, date_period, order)\n\n #plt.plot(date%self.period2, residual_mag, 'b+')\n #plt.show()\n\n # Derive Fourier features for the first second.\n self.f2_amp = 2. * np.sqrt(p2[1]**2 + p2[2]**2)\n self.f2_R21 = np.sqrt(p2[3]**2 + p2[4]**2) / self.f2_amp\n self.f2_R31 = np.sqrt(p2[5]**2 + p2[6]**2) / self.f2_amp\n self.f2_R41 = np.sqrt(p2[7]**2 + p2[8]**2) / self.f2_amp\n self.f2_R51 = np.sqrt(p2[9]**2 + p2[10]**2) / self.f2_amp\n self.f2_phase = np.arctan(-p2[1] / p2[2])\n self.f2_phi21 = np.arctan(-p2[3] / p2[4]) - 2. * self.f2_phase\n self.f2_phi31 = np.arctan(-p2[5] / p2[6]) - 3. * self.f2_phase\n self.f2_phi41 = np.arctan(-p2[7] / p2[8]) - 4. * self.f2_phase\n self.f2_phi51 = np.arctan(-p2[9] / p2[10]) - 5. * self.f2_phase\n\n # Calculate features using the first and second periods.\n self.f12_ratio = self.f2 / self.f1\n self.f12_remain = self.f1 % self.f2 \\\n if self.f1 > self.f2 else self.f2 % self.f1\n self.f12_amp = self.f2_amp / self.f1_amp\n self.f12_phase = self.f2_phase - self.f1_phase\n \"\"\"\n", "def get_eta(self, mag, std):\n \"\"\"\n Return Eta feature.\n\n Parameters\n ----------\n mag : array_like\n An array of magnitudes.\n std : array_like\n A standard deviation of magnitudes.\n\n Returns\n -------\n eta : float\n The value of Eta index.\n \"\"\"\n\n diff = mag[1:] - mag[:len(mag) - 1]\n eta = np.sum(diff * diff) / (len(mag) - 1.) / std / std\n\n return eta\n", "def slope_percentile(self, date, mag):\n \"\"\"\n Return 10% and 90% percentile of slope.\n\n Parameters\n ----------\n date : array_like\n An array of phase-folded date. Sorted.\n mag : array_like\n An array of phase-folded magnitudes. Sorted by date.\n\n Returns\n -------\n per_10 : float\n 10% percentile values of slope.\n per_90 : float\n 90% percentile values of slope.\n \"\"\"\n\n date_diff = date[1:] - date[:len(date) - 1]\n mag_diff = mag[1:] - mag[:len(mag) - 1]\n\n # Remove zero mag_diff.\n index = np.where(mag_diff != 0.)\n date_diff = date_diff[index]\n mag_diff = mag_diff[index]\n\n # Derive slope.\n slope = date_diff / mag_diff\n\n percentile_10 = np.percentile(slope, 10.)\n percentile_90 = np.percentile(slope, 90.)\n\n return percentile_10, percentile_90\n", "def get_cusum(self, mag):\n \"\"\"\n Return max - min of cumulative sum.\n\n Parameters\n ----------\n mag : array_like\n An array of magnitudes.\n\n Returns\n -------\n mm_cusum : float\n Max - min of cumulative sum.\n \"\"\"\n\n c = np.cumsum(mag - self.weighted_mean) / len(mag) / self.weighted_std\n\n return np.max(c) - np.min(c)\n" ]
class ExtractFeatures: """ Extract variability features of a light curve. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. err : array_like, optional An array of magnitude error. If None, std(mag) will be used. n_threads : int, optional The number of cores to use to derive periods. min_period : float, optional The minimum period to calculate. """ def __init__(self, date, mag, err=None, n_threads=4, min_period=0.03): # Set basic values. if not isinstance(date, np.ndarray): date = np.array(date) if not isinstance(mag, np.ndarray): mag = np.array(mag) self.date = date self.mag = mag if err is not None: if not isinstance(err, np.ndarray): err = np.array(err) self.err = err else: self.err = np.ones(len(self.mag)) * np.std(self.mag) # Check length. if (len(self.date) != len(self.mag)) or \ (len(self.date) != len(self.err)) or \ (len(self.mag) != len(self.err)): raise RuntimeError('The length of date, mag, and err must be same.') # if the number of data points is too small. min_n_data = 80 if len(self.date) < min_n_data: warnings.warn('The number of data points are less than %d.' % min_n_data) n_threads = int(n_threads) if n_threads > multiprocessing.cpu_count(): self.n_threads = multiprocessing.cpu_count() else: if n_threads <= 0: self.n_threads = 1 else: self.n_threads = n_threads min_period = float(min_period) if min_period <= 0: self.min_period = 0.03 else: self.min_period = min_period def run(self): """Run feature extraction modules.""" # shallow_run must be executed prior to deep_run # since shallow_run calculates several values needed for deep_run. self.shallow_run() self.deep_run() def shallow_run(self): """Derive not-period-based features.""" # Number of data points self.n_points = len(self.date) # Weight calculation. # All zero values. if not self.err.any(): self.err = np.ones(len(self.mag)) * np.std(self.mag) # Some zero values. elif not self.err.all(): np.putmask(self.err, self.err==0, np.median(self.err)) self.weight = 1. / self.err self.weighted_sum = np.sum(self.weight) # Simple statistics, mean, median and std. self.mean = np.mean(self.mag) self.median = np.median(self.mag) self.std = np.std(self.mag) # Weighted mean and std. self.weighted_mean = np.sum(self.mag * self.weight) / self.weighted_sum self.weighted_std = np.sqrt(np.sum((self.mag - self.weighted_mean) ** 2 \ * self.weight) / self.weighted_sum) # Skewness and kurtosis. self.skewness = ss.skew(self.mag) self.kurtosis = ss.kurtosis(self.mag) # Normalization-test. Shapiro-Wilk test. shapiro = ss.shapiro(self.mag) self.shapiro_w = shapiro[0] # self.shapiro_log10p = np.log10(shapiro[1]) # Percentile features. self.quartile31 = np.percentile(self.mag, 75) \ - np.percentile(self.mag, 25) # Stetson K. self.stetson_k = self.get_stetson_k(self.mag, self.median, self.err) # Ratio between higher and lower amplitude than average. self.hl_amp_ratio = self.half_mag_amplitude_ratio( self.mag, self.median, self.weight) # This second function's value is very similar with the above one. # self.hl_amp_ratio2 = self.half_mag_amplitude_ratio2( # self.mag, self.median) # Cusum self.cusum = self.get_cusum(self.mag) # Eta self.eta = self.get_eta(self.mag, self.weighted_std) def get_period_LS(self, date, mag, n_threads, min_period): """ Period finding using the Lomb-Scargle algorithm. Finding two periods. The second period is estimated after whitening the first period. Calculating various other features as well using derived periods. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. n_threads : int The number of threads to use. min_period : float The minimum period to calculate. """ # DO NOT CHANGE THESE PARAMETERS. oversampling = 3. hifac = int((max(date) - min(date)) / len(date) / min_period * 2.) # Minimum hifac if hifac < 100: hifac = 100 # Lomb-Scargle. fx, fy, nout, jmax, prob = pLS.fasper(date, mag, oversampling, hifac, n_threads) self.f = fx[jmax] self.period = 1. / self.f self.period_uncertainty = self.get_period_uncertainty(fx, fy, jmax) self.period_log10FAP = \ np.log10(pLS.getSignificance(fx, fy, nout, oversampling)[jmax]) # self.f_SNR1 = fy[jmax] / np.median(fy) self.period_SNR = (fy[jmax] - np.median(fy)) / np.std(fy) # Fit Fourier Series of order 3. order = 3 # Initial guess of Fourier coefficients. p0 = np.ones(order * 2 + 1) date_period = (date % self.period) / self.period p1, success = leastsq(self.residuals, p0, args=(date_period, mag, order)) # fitted_y = self.FourierSeries(p1, date_period, order) # print p1, self.mean, self.median # plt.plot(date_period, self.mag, 'b+') # plt.show() # Derive Fourier features for the first period. # Petersen, J. O., 1986, A&A self.amplitude = np.sqrt(p1[1] ** 2 + p1[2] ** 2) self.r21 = np.sqrt(p1[3] ** 2 + p1[4] ** 2) / self.amplitude self.r31 = np.sqrt(p1[5] ** 2 + p1[6] ** 2) / self.amplitude self.f_phase = np.arctan(-p1[1] / p1[2]) self.phi21 = np.arctan(-p1[3] / p1[4]) - 2. * self.f_phase self.phi31 = np.arctan(-p1[5] / p1[6]) - 3. * self.f_phase """ # Derive a second period. # Whitening a light curve. residual_mag = mag - fitted_y # Lomb-Scargle again to find the second period. omega_top, power_top = search_frequencies(date, residual_mag, err, #LS_kwargs={'generalized':True, 'subtract_mean':True}, n_eval=5000, n_retry=3, n_save=50) self.period2 = 2*np.pi/omega_top[np.where(power_top==np.max(power_top))][0] self.f2 = 1. / self.period2 self.f2_SNR = power_top[np.where(power_top==np.max(power_top))][0] \ * (len(self.date) - 1) / 2. # Fit Fourier Series again. p0 = [1.] * order * 2 date_period = (date % self.period) / self.period p2, success = leastsq(self.residuals, p0, args=(date_period, residual_mag, order)) fitted_y = self.FourierSeries(p2, date_period, order) #plt.plot(date%self.period2, residual_mag, 'b+') #plt.show() # Derive Fourier features for the first second. self.f2_amp = 2. * np.sqrt(p2[1]**2 + p2[2]**2) self.f2_R21 = np.sqrt(p2[3]**2 + p2[4]**2) / self.f2_amp self.f2_R31 = np.sqrt(p2[5]**2 + p2[6]**2) / self.f2_amp self.f2_R41 = np.sqrt(p2[7]**2 + p2[8]**2) / self.f2_amp self.f2_R51 = np.sqrt(p2[9]**2 + p2[10]**2) / self.f2_amp self.f2_phase = np.arctan(-p2[1] / p2[2]) self.f2_phi21 = np.arctan(-p2[3] / p2[4]) - 2. * self.f2_phase self.f2_phi31 = np.arctan(-p2[5] / p2[6]) - 3. * self.f2_phase self.f2_phi41 = np.arctan(-p2[7] / p2[8]) - 4. * self.f2_phase self.f2_phi51 = np.arctan(-p2[9] / p2[10]) - 5. * self.f2_phase # Calculate features using the first and second periods. self.f12_ratio = self.f2 / self.f1 self.f12_remain = self.f1 % self.f2 \ if self.f1 > self.f2 else self.f2 % self.f1 self.f12_amp = self.f2_amp / self.f1_amp self.f12_phase = self.f2_phase - self.f1_phase """ def get_period_uncertainty(self, fx, fy, jmax, fx_width=100): """ Get uncertainty of a period. The uncertainty is defined as the half width of the frequencies around the peak, that becomes lower than average + standard deviation of the power spectrum. Since we may not have fine resolution around the peak, we do not assume it is gaussian. So, no scaling factor of 2.355 (= 2 * sqrt(2 * ln2)) is applied. Parameters ---------- fx : array_like An array of frequencies. fy : array_like An array of amplitudes. jmax : int An index at the peak frequency. fx_width : int, optional Width of power spectrum to calculate uncertainty. Returns ------- p_uncertain : float Period uncertainty. """ # Get subset start_index = jmax - fx_width end_index = jmax + fx_width if start_index < 0: start_index = 0 if end_index > len(fx) - 1: end_index = len(fx) - 1 fx_subset = fx[start_index:end_index] fy_subset = fy[start_index:end_index] fy_mean = np.median(fy_subset) fy_std = np.std(fy_subset) # Find peak max_index = np.argmax(fy_subset) # Find list whose powers become lower than average + std. index = np.where(fy_subset <= fy_mean + fy_std)[0] # Find the edge at left and right. This is the full width. left_index = index[(index < max_index)] if len(left_index) == 0: left_index = 0 else: left_index = left_index[-1] right_index = index[(index > max_index)] if len(right_index) == 0: right_index = len(fy_subset) - 1 else: right_index = right_index[0] # We assume the half of the full width is the period uncertainty. half_width = (1. / fx_subset[left_index] - 1. / fx_subset[right_index]) / 2. period_uncertainty = half_width return period_uncertainty def residuals(self, pars, x, y, order): """ Residual of Fourier Series. Parameters ---------- pars : array_like Fourier series parameters. x : array_like An array of date. y : array_like An array of true values to fit. order : int An order of Fourier Series. """ return y - self.fourier_series(pars, x, order) def fourier_series(self, pars, x, order): """ Function to fit Fourier Series. Parameters ---------- x : array_like An array of date divided by period. It doesn't need to be sorted. pars : array_like Fourier series parameters. order : int An order of Fourier series. """ sum = pars[0] for i in range(order): sum += pars[i * 2 + 1] * np.sin(2 * np.pi * (i + 1) * x) \ + pars[i * 2 + 2] * np.cos(2 * np.pi * (i + 1) * x) return sum def get_stetson_k(self, mag, avg, err): """ Return Stetson K feature. Parameters ---------- mag : array_like An array of magnitude. avg : float An average value of magnitudes. err : array_like An array of magnitude errors. Returns ------- stetson_k : float Stetson K value. """ residual = (mag - avg) / err stetson_k = np.sum(np.fabs(residual)) \ / np.sqrt(np.sum(residual * residual)) / np.sqrt(len(mag)) return stetson_k def half_mag_amplitude_ratio(self, mag, avg, weight): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. weight : array_like An array of weight. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) lower_weight = weight[index] lower_weight_sum = np.sum(lower_weight) lower_mag = mag[index] lower_weighted_std = np.sum((lower_mag - avg) ** 2 * lower_weight) / \ lower_weight_sum # For higher (brighter) magnitude than average. index = np.where(mag <= avg) higher_weight = weight[index] higher_weight_sum = np.sum(higher_weight) higher_mag = mag[index] higher_weighted_std = np.sum((higher_mag - avg) ** 2 * higher_weight) / \ higher_weight_sum # Return ratio. return np.sqrt(lower_weighted_std / higher_weighted_std) def half_mag_amplitude_ratio2(self, mag, avg): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) fainter_mag = mag[index] lower_sum = np.sum((fainter_mag - avg) ** 2) / len(fainter_mag) # For higher (brighter) magnitude than average. index = np.where(mag <= avg) brighter_mag = mag[index] higher_sum = np.sum((avg - brighter_mag) ** 2) / len(brighter_mag) # Return ratio. return np.sqrt(lower_sum / higher_sum) def get_eta(self, mag, std): """ Return Eta feature. Parameters ---------- mag : array_like An array of magnitudes. std : array_like A standard deviation of magnitudes. Returns ------- eta : float The value of Eta index. """ diff = mag[1:] - mag[:len(mag) - 1] eta = np.sum(diff * diff) / (len(mag) - 1.) / std / std return eta def slope_percentile(self, date, mag): """ Return 10% and 90% percentile of slope. Parameters ---------- date : array_like An array of phase-folded date. Sorted. mag : array_like An array of phase-folded magnitudes. Sorted by date. Returns ------- per_10 : float 10% percentile values of slope. per_90 : float 90% percentile values of slope. """ date_diff = date[1:] - date[:len(date) - 1] mag_diff = mag[1:] - mag[:len(mag) - 1] # Remove zero mag_diff. index = np.where(mag_diff != 0.) date_diff = date_diff[index] mag_diff = mag_diff[index] # Derive slope. slope = date_diff / mag_diff percentile_10 = np.percentile(slope, 10.) percentile_90 = np.percentile(slope, 90.) return percentile_10, percentile_90 def get_cusum(self, mag): """ Return max - min of cumulative sum. Parameters ---------- mag : array_like An array of magnitudes. Returns ------- mm_cusum : float Max - min of cumulative sum. """ c = np.cumsum(mag - self.weighted_mean) / len(mag) / self.weighted_std return np.max(c) - np.min(c) def get_features2(self): """ Return all features with its names. Returns ------- names : list Feature names. values : list Feature values """ feature_names = [] feature_values = [] # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): # Omit input variables such as date, mag, err, etc. if not (name == 'date' or name == 'mag' or name == 'err' or name == 'n_threads' or name == 'min_period'): # Filter some other unnecessary features. if not (name == 'f' or name == 'f_phase' or name == 'period_log10FAP' or name == 'weight' or name == 'weighted_sum' or name == 'median' or name == 'mean' or name == 'std'): feature_names.append(name) # Sort by the names. # Sorting should be done to keep maintaining the same order of features. feature_names.sort() # Get feature values. for name in feature_names: feature_values.append(all_vars[name]) return feature_names, feature_values def get_features(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ ''' features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features ''' return self.get_features_all() def get_features_all(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list_all: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features
dwkim78/upsilon
upsilon/extract_features/extract_features.py
ExtractFeatures.get_period_LS
python
def get_period_LS(self, date, mag, n_threads, min_period): # DO NOT CHANGE THESE PARAMETERS. oversampling = 3. hifac = int((max(date) - min(date)) / len(date) / min_period * 2.) # Minimum hifac if hifac < 100: hifac = 100 # Lomb-Scargle. fx, fy, nout, jmax, prob = pLS.fasper(date, mag, oversampling, hifac, n_threads) self.f = fx[jmax] self.period = 1. / self.f self.period_uncertainty = self.get_period_uncertainty(fx, fy, jmax) self.period_log10FAP = \ np.log10(pLS.getSignificance(fx, fy, nout, oversampling)[jmax]) # self.f_SNR1 = fy[jmax] / np.median(fy) self.period_SNR = (fy[jmax] - np.median(fy)) / np.std(fy) # Fit Fourier Series of order 3. order = 3 # Initial guess of Fourier coefficients. p0 = np.ones(order * 2 + 1) date_period = (date % self.period) / self.period p1, success = leastsq(self.residuals, p0, args=(date_period, mag, order)) # fitted_y = self.FourierSeries(p1, date_period, order) # print p1, self.mean, self.median # plt.plot(date_period, self.mag, 'b+') # plt.show() # Derive Fourier features for the first period. # Petersen, J. O., 1986, A&A self.amplitude = np.sqrt(p1[1] ** 2 + p1[2] ** 2) self.r21 = np.sqrt(p1[3] ** 2 + p1[4] ** 2) / self.amplitude self.r31 = np.sqrt(p1[5] ** 2 + p1[6] ** 2) / self.amplitude self.f_phase = np.arctan(-p1[1] / p1[2]) self.phi21 = np.arctan(-p1[3] / p1[4]) - 2. * self.f_phase self.phi31 = np.arctan(-p1[5] / p1[6]) - 3. * self.f_phase """ # Derive a second period. # Whitening a light curve. residual_mag = mag - fitted_y # Lomb-Scargle again to find the second period. omega_top, power_top = search_frequencies(date, residual_mag, err, #LS_kwargs={'generalized':True, 'subtract_mean':True}, n_eval=5000, n_retry=3, n_save=50) self.period2 = 2*np.pi/omega_top[np.where(power_top==np.max(power_top))][0] self.f2 = 1. / self.period2 self.f2_SNR = power_top[np.where(power_top==np.max(power_top))][0] \ * (len(self.date) - 1) / 2. # Fit Fourier Series again. p0 = [1.] * order * 2 date_period = (date % self.period) / self.period p2, success = leastsq(self.residuals, p0, args=(date_period, residual_mag, order)) fitted_y = self.FourierSeries(p2, date_period, order) #plt.plot(date%self.period2, residual_mag, 'b+') #plt.show() # Derive Fourier features for the first second. self.f2_amp = 2. * np.sqrt(p2[1]**2 + p2[2]**2) self.f2_R21 = np.sqrt(p2[3]**2 + p2[4]**2) / self.f2_amp self.f2_R31 = np.sqrt(p2[5]**2 + p2[6]**2) / self.f2_amp self.f2_R41 = np.sqrt(p2[7]**2 + p2[8]**2) / self.f2_amp self.f2_R51 = np.sqrt(p2[9]**2 + p2[10]**2) / self.f2_amp self.f2_phase = np.arctan(-p2[1] / p2[2]) self.f2_phi21 = np.arctan(-p2[3] / p2[4]) - 2. * self.f2_phase self.f2_phi31 = np.arctan(-p2[5] / p2[6]) - 3. * self.f2_phase self.f2_phi41 = np.arctan(-p2[7] / p2[8]) - 4. * self.f2_phase self.f2_phi51 = np.arctan(-p2[9] / p2[10]) - 5. * self.f2_phase # Calculate features using the first and second periods. self.f12_ratio = self.f2 / self.f1 self.f12_remain = self.f1 % self.f2 \ if self.f1 > self.f2 else self.f2 % self.f1 self.f12_amp = self.f2_amp / self.f1_amp self.f12_phase = self.f2_phase - self.f1_phase """
Period finding using the Lomb-Scargle algorithm. Finding two periods. The second period is estimated after whitening the first period. Calculating various other features as well using derived periods. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. n_threads : int The number of threads to use. min_period : float The minimum period to calculate.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/extract_features.py#L168-L273
[ "def fasper(x, y, ofac, hifac, n_threads, MACC=4):\n \"\"\"\n Given abscissas x (which need not be equally spaced) and ordinates\n y, and given a desired oversampling factor ofac (a typical value\n being 4 or larger). this routine creates an array wk1 with a\n sequence of nout increasing frequencies (not angular frequencies)\n up to hifac times the \"average\" Nyquist frequency, and creates\n an array wk2 with the values of the Lomb normalized periodogram at\n those frequencies. The arrays x and y are not altered. This\n routine also returns jmax such that wk2(jmax) is the maximum\n element in wk2, and prob, an estimate of the significance of that\n maximum against the hypothesis of random noise. A small value of prob\n indicates that a significant periodic signal is present.\n\n Reference: \n Press, W. H. & Rybicki, G. B. 1989\n ApJ vol. 338, p. 277-280.\n Fast algorithm for spectral analysis of unevenly sampled data\n (1989ApJ...338..277P)\n\n Arguments:\n X : Abscissas array, (e.g. an array of times).\n Y : Ordinates array, (e.g. corresponding counts).\n Ofac : Oversampling factor.\n Hifac : Hifac * \"average\" Nyquist frequency = highest frequency\n for which values of the Lomb normalized periodogram will\n be calculated.\n n_threads : number of threads to use.\n\n Returns:\n Wk1 : An array of Lomb periodogram frequencies.\n Wk2 : An array of corresponding values of the Lomb periodogram.\n Nout : Wk1 & Wk2 dimensions (number of calculated frequencies)\n Jmax : The array index corresponding to the MAX( Wk2 ).\n Prob : False Alarm Probability of the largest Periodogram value\n MACC : Number of interpolation points per 1/4 cycle\n of highest frequency\n\n History:\n 02/23/2009, v1.0, MF\n Translation of IDL code (orig. Numerical recipies)\n \"\"\"\n #Check dimensions of input arrays\n n = long(len(x))\n if n != len(y):\n print('Incompatible arrays.')\n return\n\n #print x, y, hifac, ofac\n\n nout = int(0.5*ofac*hifac*n)\n nfreqt = long(ofac*hifac*n*MACC) #Size the FFT as next power\n nfreq = 64 # of 2 above nfreqt.\n\n while nfreq < nfreqt:\n nfreq = 2*nfreq\n\n ndim = long(2*nfreq)\n\n #Compute the mean, variance\n ave = y.mean()\n ##sample variance because the divisor is N-1\n var = ((y - y.mean())**2).sum()/(len(y) - 1) \n # and range of the data.\n xmin = x.min()\n xmax = x.max()\n xdif = xmax - xmin\n\n #extrapolate the data into the workspaces\n if is_pyfftw:\n wk1 = pyfftw.n_byte_align_empty(int(ndim), 16, 'complex') * 0.\n wk2 = pyfftw.n_byte_align_empty(int(ndim), 16, 'complex') * 0.\n else:\n wk1 = zeros(ndim, dtype='complex')\n wk2 = zeros(ndim, dtype='complex')\n\n fac = ndim/(xdif*ofac)\n fndim = ndim\n ck = ((x - xmin)*fac) % fndim\n ckk = (2.0*ck) % fndim\n\n for j in range(0, n):\n __spread__(y[j] - ave, wk1, ndim, ck[j], MACC)\n __spread__(1.0, wk2, ndim, ckk[j], MACC)\n\n #Take the Fast Fourier Transforms.\n if is_pyfftw:\n fft_wk1 = pyfftw.builders.ifft(wk1, planner_effort='FFTW_ESTIMATE',\n threads=n_threads)\n wk1 = fft_wk1() * len(wk1)\n fft_wk2 = pyfftw.builders.ifft(wk2, planner_effort='FFTW_ESTIMATE',\n threads=n_threads)\n wk2 = fft_wk2() * len(wk2)\n else:\n wk1 = ifft(wk1)*len(wk1)\n wk2 = ifft(wk2)*len(wk1)\n\n wk1 = wk1[1:nout + 1]\n wk2 = wk2[1:nout + 1]\n rwk1 = wk1.real\n iwk1 = wk1.imag\n rwk2 = wk2.real\n iwk2 = wk2.imag\n\n df = 1.0/(xdif*ofac)\n\n #Compute the Lomb value for each frequency\n hypo2 = 2.0*abs(wk2)\n hc2wt = rwk2/hypo2\n hs2wt = iwk2/hypo2\n\n cwt = sqrt(0.5 + hc2wt)\n swt = sign(hs2wt)*(sqrt(0.5 - hc2wt))\n den = 0.5*n + hc2wt*rwk2 + hs2wt*iwk2\n cterm = (cwt*rwk1 + swt*iwk1)**2./den\n sterm = (cwt*iwk1 - swt*rwk1)**2./(n - den)\n\n wk1 = df*(arange(nout, dtype='float') + 1.)\n wk2 = (cterm + sterm)/(2.0*var)\n pmax = wk2.max()\n jmax = wk2.argmax()\n\n #Significance estimation\n #expy = exp(-wk2) \n #effm = 2.0*(nout)/ofac \n #sig = effm*expy\n #ind = (sig > 0.01).nonzero()\n #sig[ind] = 1.0-(1.0-expy[ind])**effm\n\n #Estimate significance of largest peak value\n expy = exp(-pmax) \n effm = 2.0*(nout)/ofac\n prob = effm*expy\n\n if prob > 0.01: \n prob = 1.0 - (1.0 - expy)**effm\n\n return wk1, wk2, nout, jmax, prob\n", "def getSignificance(wk1, wk2, nout, ofac):\n \"\"\" \n Returns the peak false alarm probabilities\n Hence the lower is the probability and the more significant is the peak\n \"\"\"\n expy = exp(-wk2) \n effm = 2.0*(nout)/ofac \n sig = effm*expy\n ind = (sig > 0.01).nonzero()\n sig[ind] = 1.0 - (1.0 - expy[ind])**effm\n\n return sig\n", "def get_period_uncertainty(self, fx, fy, jmax, fx_width=100):\n \"\"\"\n Get uncertainty of a period.\n\n The uncertainty is defined as the half width of the frequencies\n around the peak, that becomes lower than average + standard deviation\n of the power spectrum.\n\n Since we may not have fine resolution around the peak,\n we do not assume it is gaussian. So, no scaling factor of\n 2.355 (= 2 * sqrt(2 * ln2)) is applied.\n\n Parameters\n ----------\n fx : array_like\n An array of frequencies.\n fy : array_like\n An array of amplitudes.\n jmax : int\n An index at the peak frequency.\n fx_width : int, optional\n Width of power spectrum to calculate uncertainty.\n\n Returns\n -------\n p_uncertain : float\n Period uncertainty.\n \"\"\"\n\n # Get subset\n start_index = jmax - fx_width\n end_index = jmax + fx_width\n if start_index < 0:\n start_index = 0\n if end_index > len(fx) - 1:\n end_index = len(fx) - 1\n\n fx_subset = fx[start_index:end_index]\n fy_subset = fy[start_index:end_index]\n fy_mean = np.median(fy_subset)\n fy_std = np.std(fy_subset)\n\n # Find peak\n max_index = np.argmax(fy_subset)\n\n # Find list whose powers become lower than average + std.\n index = np.where(fy_subset <= fy_mean + fy_std)[0]\n\n # Find the edge at left and right. This is the full width.\n left_index = index[(index < max_index)]\n if len(left_index) == 0:\n left_index = 0\n else:\n left_index = left_index[-1]\n right_index = index[(index > max_index)]\n if len(right_index) == 0:\n right_index = len(fy_subset) - 1\n else:\n right_index = right_index[0]\n\n # We assume the half of the full width is the period uncertainty.\n half_width = (1. / fx_subset[left_index]\n - 1. / fx_subset[right_index]) / 2.\n period_uncertainty = half_width\n\n return period_uncertainty\n" ]
class ExtractFeatures: """ Extract variability features of a light curve. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. err : array_like, optional An array of magnitude error. If None, std(mag) will be used. n_threads : int, optional The number of cores to use to derive periods. min_period : float, optional The minimum period to calculate. """ def __init__(self, date, mag, err=None, n_threads=4, min_period=0.03): # Set basic values. if not isinstance(date, np.ndarray): date = np.array(date) if not isinstance(mag, np.ndarray): mag = np.array(mag) self.date = date self.mag = mag if err is not None: if not isinstance(err, np.ndarray): err = np.array(err) self.err = err else: self.err = np.ones(len(self.mag)) * np.std(self.mag) # Check length. if (len(self.date) != len(self.mag)) or \ (len(self.date) != len(self.err)) or \ (len(self.mag) != len(self.err)): raise RuntimeError('The length of date, mag, and err must be same.') # if the number of data points is too small. min_n_data = 80 if len(self.date) < min_n_data: warnings.warn('The number of data points are less than %d.' % min_n_data) n_threads = int(n_threads) if n_threads > multiprocessing.cpu_count(): self.n_threads = multiprocessing.cpu_count() else: if n_threads <= 0: self.n_threads = 1 else: self.n_threads = n_threads min_period = float(min_period) if min_period <= 0: self.min_period = 0.03 else: self.min_period = min_period def run(self): """Run feature extraction modules.""" # shallow_run must be executed prior to deep_run # since shallow_run calculates several values needed for deep_run. self.shallow_run() self.deep_run() def shallow_run(self): """Derive not-period-based features.""" # Number of data points self.n_points = len(self.date) # Weight calculation. # All zero values. if not self.err.any(): self.err = np.ones(len(self.mag)) * np.std(self.mag) # Some zero values. elif not self.err.all(): np.putmask(self.err, self.err==0, np.median(self.err)) self.weight = 1. / self.err self.weighted_sum = np.sum(self.weight) # Simple statistics, mean, median and std. self.mean = np.mean(self.mag) self.median = np.median(self.mag) self.std = np.std(self.mag) # Weighted mean and std. self.weighted_mean = np.sum(self.mag * self.weight) / self.weighted_sum self.weighted_std = np.sqrt(np.sum((self.mag - self.weighted_mean) ** 2 \ * self.weight) / self.weighted_sum) # Skewness and kurtosis. self.skewness = ss.skew(self.mag) self.kurtosis = ss.kurtosis(self.mag) # Normalization-test. Shapiro-Wilk test. shapiro = ss.shapiro(self.mag) self.shapiro_w = shapiro[0] # self.shapiro_log10p = np.log10(shapiro[1]) # Percentile features. self.quartile31 = np.percentile(self.mag, 75) \ - np.percentile(self.mag, 25) # Stetson K. self.stetson_k = self.get_stetson_k(self.mag, self.median, self.err) # Ratio between higher and lower amplitude than average. self.hl_amp_ratio = self.half_mag_amplitude_ratio( self.mag, self.median, self.weight) # This second function's value is very similar with the above one. # self.hl_amp_ratio2 = self.half_mag_amplitude_ratio2( # self.mag, self.median) # Cusum self.cusum = self.get_cusum(self.mag) # Eta self.eta = self.get_eta(self.mag, self.weighted_std) def deep_run(self): """Derive period-based features.""" # Lomb-Scargle period finding. self.get_period_LS(self.date, self.mag, self.n_threads, self.min_period) # Features based on a phase-folded light curve # such as Eta, slope-percentile, etc. # Should be called after the getPeriodLS() is called. # Created phased a folded light curve. # We use period * 2 to take eclipsing binaries into account. phase_folded_date = self.date % (self.period * 2.) sorted_index = np.argsort(phase_folded_date) folded_date = phase_folded_date[sorted_index] folded_mag = self.mag[sorted_index] # phase Eta self.phase_eta = self.get_eta(folded_mag, self.weighted_std) # Slope percentile. self.slope_per10, self.slope_per90 = \ self.slope_percentile(folded_date, folded_mag) # phase Cusum self.phase_cusum = self.get_cusum(folded_mag) def get_period_uncertainty(self, fx, fy, jmax, fx_width=100): """ Get uncertainty of a period. The uncertainty is defined as the half width of the frequencies around the peak, that becomes lower than average + standard deviation of the power spectrum. Since we may not have fine resolution around the peak, we do not assume it is gaussian. So, no scaling factor of 2.355 (= 2 * sqrt(2 * ln2)) is applied. Parameters ---------- fx : array_like An array of frequencies. fy : array_like An array of amplitudes. jmax : int An index at the peak frequency. fx_width : int, optional Width of power spectrum to calculate uncertainty. Returns ------- p_uncertain : float Period uncertainty. """ # Get subset start_index = jmax - fx_width end_index = jmax + fx_width if start_index < 0: start_index = 0 if end_index > len(fx) - 1: end_index = len(fx) - 1 fx_subset = fx[start_index:end_index] fy_subset = fy[start_index:end_index] fy_mean = np.median(fy_subset) fy_std = np.std(fy_subset) # Find peak max_index = np.argmax(fy_subset) # Find list whose powers become lower than average + std. index = np.where(fy_subset <= fy_mean + fy_std)[0] # Find the edge at left and right. This is the full width. left_index = index[(index < max_index)] if len(left_index) == 0: left_index = 0 else: left_index = left_index[-1] right_index = index[(index > max_index)] if len(right_index) == 0: right_index = len(fy_subset) - 1 else: right_index = right_index[0] # We assume the half of the full width is the period uncertainty. half_width = (1. / fx_subset[left_index] - 1. / fx_subset[right_index]) / 2. period_uncertainty = half_width return period_uncertainty def residuals(self, pars, x, y, order): """ Residual of Fourier Series. Parameters ---------- pars : array_like Fourier series parameters. x : array_like An array of date. y : array_like An array of true values to fit. order : int An order of Fourier Series. """ return y - self.fourier_series(pars, x, order) def fourier_series(self, pars, x, order): """ Function to fit Fourier Series. Parameters ---------- x : array_like An array of date divided by period. It doesn't need to be sorted. pars : array_like Fourier series parameters. order : int An order of Fourier series. """ sum = pars[0] for i in range(order): sum += pars[i * 2 + 1] * np.sin(2 * np.pi * (i + 1) * x) \ + pars[i * 2 + 2] * np.cos(2 * np.pi * (i + 1) * x) return sum def get_stetson_k(self, mag, avg, err): """ Return Stetson K feature. Parameters ---------- mag : array_like An array of magnitude. avg : float An average value of magnitudes. err : array_like An array of magnitude errors. Returns ------- stetson_k : float Stetson K value. """ residual = (mag - avg) / err stetson_k = np.sum(np.fabs(residual)) \ / np.sqrt(np.sum(residual * residual)) / np.sqrt(len(mag)) return stetson_k def half_mag_amplitude_ratio(self, mag, avg, weight): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. weight : array_like An array of weight. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) lower_weight = weight[index] lower_weight_sum = np.sum(lower_weight) lower_mag = mag[index] lower_weighted_std = np.sum((lower_mag - avg) ** 2 * lower_weight) / \ lower_weight_sum # For higher (brighter) magnitude than average. index = np.where(mag <= avg) higher_weight = weight[index] higher_weight_sum = np.sum(higher_weight) higher_mag = mag[index] higher_weighted_std = np.sum((higher_mag - avg) ** 2 * higher_weight) / \ higher_weight_sum # Return ratio. return np.sqrt(lower_weighted_std / higher_weighted_std) def half_mag_amplitude_ratio2(self, mag, avg): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) fainter_mag = mag[index] lower_sum = np.sum((fainter_mag - avg) ** 2) / len(fainter_mag) # For higher (brighter) magnitude than average. index = np.where(mag <= avg) brighter_mag = mag[index] higher_sum = np.sum((avg - brighter_mag) ** 2) / len(brighter_mag) # Return ratio. return np.sqrt(lower_sum / higher_sum) def get_eta(self, mag, std): """ Return Eta feature. Parameters ---------- mag : array_like An array of magnitudes. std : array_like A standard deviation of magnitudes. Returns ------- eta : float The value of Eta index. """ diff = mag[1:] - mag[:len(mag) - 1] eta = np.sum(diff * diff) / (len(mag) - 1.) / std / std return eta def slope_percentile(self, date, mag): """ Return 10% and 90% percentile of slope. Parameters ---------- date : array_like An array of phase-folded date. Sorted. mag : array_like An array of phase-folded magnitudes. Sorted by date. Returns ------- per_10 : float 10% percentile values of slope. per_90 : float 90% percentile values of slope. """ date_diff = date[1:] - date[:len(date) - 1] mag_diff = mag[1:] - mag[:len(mag) - 1] # Remove zero mag_diff. index = np.where(mag_diff != 0.) date_diff = date_diff[index] mag_diff = mag_diff[index] # Derive slope. slope = date_diff / mag_diff percentile_10 = np.percentile(slope, 10.) percentile_90 = np.percentile(slope, 90.) return percentile_10, percentile_90 def get_cusum(self, mag): """ Return max - min of cumulative sum. Parameters ---------- mag : array_like An array of magnitudes. Returns ------- mm_cusum : float Max - min of cumulative sum. """ c = np.cumsum(mag - self.weighted_mean) / len(mag) / self.weighted_std return np.max(c) - np.min(c) def get_features2(self): """ Return all features with its names. Returns ------- names : list Feature names. values : list Feature values """ feature_names = [] feature_values = [] # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): # Omit input variables such as date, mag, err, etc. if not (name == 'date' or name == 'mag' or name == 'err' or name == 'n_threads' or name == 'min_period'): # Filter some other unnecessary features. if not (name == 'f' or name == 'f_phase' or name == 'period_log10FAP' or name == 'weight' or name == 'weighted_sum' or name == 'median' or name == 'mean' or name == 'std'): feature_names.append(name) # Sort by the names. # Sorting should be done to keep maintaining the same order of features. feature_names.sort() # Get feature values. for name in feature_names: feature_values.append(all_vars[name]) return feature_names, feature_values def get_features(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ ''' features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features ''' return self.get_features_all() def get_features_all(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list_all: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features
dwkim78/upsilon
upsilon/extract_features/extract_features.py
ExtractFeatures.get_period_uncertainty
python
def get_period_uncertainty(self, fx, fy, jmax, fx_width=100): # Get subset start_index = jmax - fx_width end_index = jmax + fx_width if start_index < 0: start_index = 0 if end_index > len(fx) - 1: end_index = len(fx) - 1 fx_subset = fx[start_index:end_index] fy_subset = fy[start_index:end_index] fy_mean = np.median(fy_subset) fy_std = np.std(fy_subset) # Find peak max_index = np.argmax(fy_subset) # Find list whose powers become lower than average + std. index = np.where(fy_subset <= fy_mean + fy_std)[0] # Find the edge at left and right. This is the full width. left_index = index[(index < max_index)] if len(left_index) == 0: left_index = 0 else: left_index = left_index[-1] right_index = index[(index > max_index)] if len(right_index) == 0: right_index = len(fy_subset) - 1 else: right_index = right_index[0] # We assume the half of the full width is the period uncertainty. half_width = (1. / fx_subset[left_index] - 1. / fx_subset[right_index]) / 2. period_uncertainty = half_width return period_uncertainty
Get uncertainty of a period. The uncertainty is defined as the half width of the frequencies around the peak, that becomes lower than average + standard deviation of the power spectrum. Since we may not have fine resolution around the peak, we do not assume it is gaussian. So, no scaling factor of 2.355 (= 2 * sqrt(2 * ln2)) is applied. Parameters ---------- fx : array_like An array of frequencies. fy : array_like An array of amplitudes. jmax : int An index at the peak frequency. fx_width : int, optional Width of power spectrum to calculate uncertainty. Returns ------- p_uncertain : float Period uncertainty.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/extract_features.py#L275-L340
null
class ExtractFeatures: """ Extract variability features of a light curve. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. err : array_like, optional An array of magnitude error. If None, std(mag) will be used. n_threads : int, optional The number of cores to use to derive periods. min_period : float, optional The minimum period to calculate. """ def __init__(self, date, mag, err=None, n_threads=4, min_period=0.03): # Set basic values. if not isinstance(date, np.ndarray): date = np.array(date) if not isinstance(mag, np.ndarray): mag = np.array(mag) self.date = date self.mag = mag if err is not None: if not isinstance(err, np.ndarray): err = np.array(err) self.err = err else: self.err = np.ones(len(self.mag)) * np.std(self.mag) # Check length. if (len(self.date) != len(self.mag)) or \ (len(self.date) != len(self.err)) or \ (len(self.mag) != len(self.err)): raise RuntimeError('The length of date, mag, and err must be same.') # if the number of data points is too small. min_n_data = 80 if len(self.date) < min_n_data: warnings.warn('The number of data points are less than %d.' % min_n_data) n_threads = int(n_threads) if n_threads > multiprocessing.cpu_count(): self.n_threads = multiprocessing.cpu_count() else: if n_threads <= 0: self.n_threads = 1 else: self.n_threads = n_threads min_period = float(min_period) if min_period <= 0: self.min_period = 0.03 else: self.min_period = min_period def run(self): """Run feature extraction modules.""" # shallow_run must be executed prior to deep_run # since shallow_run calculates several values needed for deep_run. self.shallow_run() self.deep_run() def shallow_run(self): """Derive not-period-based features.""" # Number of data points self.n_points = len(self.date) # Weight calculation. # All zero values. if not self.err.any(): self.err = np.ones(len(self.mag)) * np.std(self.mag) # Some zero values. elif not self.err.all(): np.putmask(self.err, self.err==0, np.median(self.err)) self.weight = 1. / self.err self.weighted_sum = np.sum(self.weight) # Simple statistics, mean, median and std. self.mean = np.mean(self.mag) self.median = np.median(self.mag) self.std = np.std(self.mag) # Weighted mean and std. self.weighted_mean = np.sum(self.mag * self.weight) / self.weighted_sum self.weighted_std = np.sqrt(np.sum((self.mag - self.weighted_mean) ** 2 \ * self.weight) / self.weighted_sum) # Skewness and kurtosis. self.skewness = ss.skew(self.mag) self.kurtosis = ss.kurtosis(self.mag) # Normalization-test. Shapiro-Wilk test. shapiro = ss.shapiro(self.mag) self.shapiro_w = shapiro[0] # self.shapiro_log10p = np.log10(shapiro[1]) # Percentile features. self.quartile31 = np.percentile(self.mag, 75) \ - np.percentile(self.mag, 25) # Stetson K. self.stetson_k = self.get_stetson_k(self.mag, self.median, self.err) # Ratio between higher and lower amplitude than average. self.hl_amp_ratio = self.half_mag_amplitude_ratio( self.mag, self.median, self.weight) # This second function's value is very similar with the above one. # self.hl_amp_ratio2 = self.half_mag_amplitude_ratio2( # self.mag, self.median) # Cusum self.cusum = self.get_cusum(self.mag) # Eta self.eta = self.get_eta(self.mag, self.weighted_std) def deep_run(self): """Derive period-based features.""" # Lomb-Scargle period finding. self.get_period_LS(self.date, self.mag, self.n_threads, self.min_period) # Features based on a phase-folded light curve # such as Eta, slope-percentile, etc. # Should be called after the getPeriodLS() is called. # Created phased a folded light curve. # We use period * 2 to take eclipsing binaries into account. phase_folded_date = self.date % (self.period * 2.) sorted_index = np.argsort(phase_folded_date) folded_date = phase_folded_date[sorted_index] folded_mag = self.mag[sorted_index] # phase Eta self.phase_eta = self.get_eta(folded_mag, self.weighted_std) # Slope percentile. self.slope_per10, self.slope_per90 = \ self.slope_percentile(folded_date, folded_mag) # phase Cusum self.phase_cusum = self.get_cusum(folded_mag) def get_period_LS(self, date, mag, n_threads, min_period): """ Period finding using the Lomb-Scargle algorithm. Finding two periods. The second period is estimated after whitening the first period. Calculating various other features as well using derived periods. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. n_threads : int The number of threads to use. min_period : float The minimum period to calculate. """ # DO NOT CHANGE THESE PARAMETERS. oversampling = 3. hifac = int((max(date) - min(date)) / len(date) / min_period * 2.) # Minimum hifac if hifac < 100: hifac = 100 # Lomb-Scargle. fx, fy, nout, jmax, prob = pLS.fasper(date, mag, oversampling, hifac, n_threads) self.f = fx[jmax] self.period = 1. / self.f self.period_uncertainty = self.get_period_uncertainty(fx, fy, jmax) self.period_log10FAP = \ np.log10(pLS.getSignificance(fx, fy, nout, oversampling)[jmax]) # self.f_SNR1 = fy[jmax] / np.median(fy) self.period_SNR = (fy[jmax] - np.median(fy)) / np.std(fy) # Fit Fourier Series of order 3. order = 3 # Initial guess of Fourier coefficients. p0 = np.ones(order * 2 + 1) date_period = (date % self.period) / self.period p1, success = leastsq(self.residuals, p0, args=(date_period, mag, order)) # fitted_y = self.FourierSeries(p1, date_period, order) # print p1, self.mean, self.median # plt.plot(date_period, self.mag, 'b+') # plt.show() # Derive Fourier features for the first period. # Petersen, J. O., 1986, A&A self.amplitude = np.sqrt(p1[1] ** 2 + p1[2] ** 2) self.r21 = np.sqrt(p1[3] ** 2 + p1[4] ** 2) / self.amplitude self.r31 = np.sqrt(p1[5] ** 2 + p1[6] ** 2) / self.amplitude self.f_phase = np.arctan(-p1[1] / p1[2]) self.phi21 = np.arctan(-p1[3] / p1[4]) - 2. * self.f_phase self.phi31 = np.arctan(-p1[5] / p1[6]) - 3. * self.f_phase """ # Derive a second period. # Whitening a light curve. residual_mag = mag - fitted_y # Lomb-Scargle again to find the second period. omega_top, power_top = search_frequencies(date, residual_mag, err, #LS_kwargs={'generalized':True, 'subtract_mean':True}, n_eval=5000, n_retry=3, n_save=50) self.period2 = 2*np.pi/omega_top[np.where(power_top==np.max(power_top))][0] self.f2 = 1. / self.period2 self.f2_SNR = power_top[np.where(power_top==np.max(power_top))][0] \ * (len(self.date) - 1) / 2. # Fit Fourier Series again. p0 = [1.] * order * 2 date_period = (date % self.period) / self.period p2, success = leastsq(self.residuals, p0, args=(date_period, residual_mag, order)) fitted_y = self.FourierSeries(p2, date_period, order) #plt.plot(date%self.period2, residual_mag, 'b+') #plt.show() # Derive Fourier features for the first second. self.f2_amp = 2. * np.sqrt(p2[1]**2 + p2[2]**2) self.f2_R21 = np.sqrt(p2[3]**2 + p2[4]**2) / self.f2_amp self.f2_R31 = np.sqrt(p2[5]**2 + p2[6]**2) / self.f2_amp self.f2_R41 = np.sqrt(p2[7]**2 + p2[8]**2) / self.f2_amp self.f2_R51 = np.sqrt(p2[9]**2 + p2[10]**2) / self.f2_amp self.f2_phase = np.arctan(-p2[1] / p2[2]) self.f2_phi21 = np.arctan(-p2[3] / p2[4]) - 2. * self.f2_phase self.f2_phi31 = np.arctan(-p2[5] / p2[6]) - 3. * self.f2_phase self.f2_phi41 = np.arctan(-p2[7] / p2[8]) - 4. * self.f2_phase self.f2_phi51 = np.arctan(-p2[9] / p2[10]) - 5. * self.f2_phase # Calculate features using the first and second periods. self.f12_ratio = self.f2 / self.f1 self.f12_remain = self.f1 % self.f2 \ if self.f1 > self.f2 else self.f2 % self.f1 self.f12_amp = self.f2_amp / self.f1_amp self.f12_phase = self.f2_phase - self.f1_phase """ def residuals(self, pars, x, y, order): """ Residual of Fourier Series. Parameters ---------- pars : array_like Fourier series parameters. x : array_like An array of date. y : array_like An array of true values to fit. order : int An order of Fourier Series. """ return y - self.fourier_series(pars, x, order) def fourier_series(self, pars, x, order): """ Function to fit Fourier Series. Parameters ---------- x : array_like An array of date divided by period. It doesn't need to be sorted. pars : array_like Fourier series parameters. order : int An order of Fourier series. """ sum = pars[0] for i in range(order): sum += pars[i * 2 + 1] * np.sin(2 * np.pi * (i + 1) * x) \ + pars[i * 2 + 2] * np.cos(2 * np.pi * (i + 1) * x) return sum def get_stetson_k(self, mag, avg, err): """ Return Stetson K feature. Parameters ---------- mag : array_like An array of magnitude. avg : float An average value of magnitudes. err : array_like An array of magnitude errors. Returns ------- stetson_k : float Stetson K value. """ residual = (mag - avg) / err stetson_k = np.sum(np.fabs(residual)) \ / np.sqrt(np.sum(residual * residual)) / np.sqrt(len(mag)) return stetson_k def half_mag_amplitude_ratio(self, mag, avg, weight): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. weight : array_like An array of weight. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) lower_weight = weight[index] lower_weight_sum = np.sum(lower_weight) lower_mag = mag[index] lower_weighted_std = np.sum((lower_mag - avg) ** 2 * lower_weight) / \ lower_weight_sum # For higher (brighter) magnitude than average. index = np.where(mag <= avg) higher_weight = weight[index] higher_weight_sum = np.sum(higher_weight) higher_mag = mag[index] higher_weighted_std = np.sum((higher_mag - avg) ** 2 * higher_weight) / \ higher_weight_sum # Return ratio. return np.sqrt(lower_weighted_std / higher_weighted_std) def half_mag_amplitude_ratio2(self, mag, avg): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) fainter_mag = mag[index] lower_sum = np.sum((fainter_mag - avg) ** 2) / len(fainter_mag) # For higher (brighter) magnitude than average. index = np.where(mag <= avg) brighter_mag = mag[index] higher_sum = np.sum((avg - brighter_mag) ** 2) / len(brighter_mag) # Return ratio. return np.sqrt(lower_sum / higher_sum) def get_eta(self, mag, std): """ Return Eta feature. Parameters ---------- mag : array_like An array of magnitudes. std : array_like A standard deviation of magnitudes. Returns ------- eta : float The value of Eta index. """ diff = mag[1:] - mag[:len(mag) - 1] eta = np.sum(diff * diff) / (len(mag) - 1.) / std / std return eta def slope_percentile(self, date, mag): """ Return 10% and 90% percentile of slope. Parameters ---------- date : array_like An array of phase-folded date. Sorted. mag : array_like An array of phase-folded magnitudes. Sorted by date. Returns ------- per_10 : float 10% percentile values of slope. per_90 : float 90% percentile values of slope. """ date_diff = date[1:] - date[:len(date) - 1] mag_diff = mag[1:] - mag[:len(mag) - 1] # Remove zero mag_diff. index = np.where(mag_diff != 0.) date_diff = date_diff[index] mag_diff = mag_diff[index] # Derive slope. slope = date_diff / mag_diff percentile_10 = np.percentile(slope, 10.) percentile_90 = np.percentile(slope, 90.) return percentile_10, percentile_90 def get_cusum(self, mag): """ Return max - min of cumulative sum. Parameters ---------- mag : array_like An array of magnitudes. Returns ------- mm_cusum : float Max - min of cumulative sum. """ c = np.cumsum(mag - self.weighted_mean) / len(mag) / self.weighted_std return np.max(c) - np.min(c) def get_features2(self): """ Return all features with its names. Returns ------- names : list Feature names. values : list Feature values """ feature_names = [] feature_values = [] # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): # Omit input variables such as date, mag, err, etc. if not (name == 'date' or name == 'mag' or name == 'err' or name == 'n_threads' or name == 'min_period'): # Filter some other unnecessary features. if not (name == 'f' or name == 'f_phase' or name == 'period_log10FAP' or name == 'weight' or name == 'weighted_sum' or name == 'median' or name == 'mean' or name == 'std'): feature_names.append(name) # Sort by the names. # Sorting should be done to keep maintaining the same order of features. feature_names.sort() # Get feature values. for name in feature_names: feature_values.append(all_vars[name]) return feature_names, feature_values def get_features(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ ''' features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features ''' return self.get_features_all() def get_features_all(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list_all: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features
dwkim78/upsilon
upsilon/extract_features/extract_features.py
ExtractFeatures.residuals
python
def residuals(self, pars, x, y, order): return y - self.fourier_series(pars, x, order)
Residual of Fourier Series. Parameters ---------- pars : array_like Fourier series parameters. x : array_like An array of date. y : array_like An array of true values to fit. order : int An order of Fourier Series.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/extract_features.py#L342-L358
[ "def fourier_series(self, pars, x, order):\n \"\"\"\n Function to fit Fourier Series.\n\n Parameters\n ----------\n x : array_like\n An array of date divided by period. It doesn't need to be sorted.\n pars : array_like\n Fourier series parameters.\n order : int\n An order of Fourier series.\n \"\"\"\n\n sum = pars[0]\n for i in range(order):\n sum += pars[i * 2 + 1] * np.sin(2 * np.pi * (i + 1) * x) \\\n + pars[i * 2 + 2] * np.cos(2 * np.pi * (i + 1) * x)\n\n return sum\n" ]
class ExtractFeatures: """ Extract variability features of a light curve. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. err : array_like, optional An array of magnitude error. If None, std(mag) will be used. n_threads : int, optional The number of cores to use to derive periods. min_period : float, optional The minimum period to calculate. """ def __init__(self, date, mag, err=None, n_threads=4, min_period=0.03): # Set basic values. if not isinstance(date, np.ndarray): date = np.array(date) if not isinstance(mag, np.ndarray): mag = np.array(mag) self.date = date self.mag = mag if err is not None: if not isinstance(err, np.ndarray): err = np.array(err) self.err = err else: self.err = np.ones(len(self.mag)) * np.std(self.mag) # Check length. if (len(self.date) != len(self.mag)) or \ (len(self.date) != len(self.err)) or \ (len(self.mag) != len(self.err)): raise RuntimeError('The length of date, mag, and err must be same.') # if the number of data points is too small. min_n_data = 80 if len(self.date) < min_n_data: warnings.warn('The number of data points are less than %d.' % min_n_data) n_threads = int(n_threads) if n_threads > multiprocessing.cpu_count(): self.n_threads = multiprocessing.cpu_count() else: if n_threads <= 0: self.n_threads = 1 else: self.n_threads = n_threads min_period = float(min_period) if min_period <= 0: self.min_period = 0.03 else: self.min_period = min_period def run(self): """Run feature extraction modules.""" # shallow_run must be executed prior to deep_run # since shallow_run calculates several values needed for deep_run. self.shallow_run() self.deep_run() def shallow_run(self): """Derive not-period-based features.""" # Number of data points self.n_points = len(self.date) # Weight calculation. # All zero values. if not self.err.any(): self.err = np.ones(len(self.mag)) * np.std(self.mag) # Some zero values. elif not self.err.all(): np.putmask(self.err, self.err==0, np.median(self.err)) self.weight = 1. / self.err self.weighted_sum = np.sum(self.weight) # Simple statistics, mean, median and std. self.mean = np.mean(self.mag) self.median = np.median(self.mag) self.std = np.std(self.mag) # Weighted mean and std. self.weighted_mean = np.sum(self.mag * self.weight) / self.weighted_sum self.weighted_std = np.sqrt(np.sum((self.mag - self.weighted_mean) ** 2 \ * self.weight) / self.weighted_sum) # Skewness and kurtosis. self.skewness = ss.skew(self.mag) self.kurtosis = ss.kurtosis(self.mag) # Normalization-test. Shapiro-Wilk test. shapiro = ss.shapiro(self.mag) self.shapiro_w = shapiro[0] # self.shapiro_log10p = np.log10(shapiro[1]) # Percentile features. self.quartile31 = np.percentile(self.mag, 75) \ - np.percentile(self.mag, 25) # Stetson K. self.stetson_k = self.get_stetson_k(self.mag, self.median, self.err) # Ratio between higher and lower amplitude than average. self.hl_amp_ratio = self.half_mag_amplitude_ratio( self.mag, self.median, self.weight) # This second function's value is very similar with the above one. # self.hl_amp_ratio2 = self.half_mag_amplitude_ratio2( # self.mag, self.median) # Cusum self.cusum = self.get_cusum(self.mag) # Eta self.eta = self.get_eta(self.mag, self.weighted_std) def deep_run(self): """Derive period-based features.""" # Lomb-Scargle period finding. self.get_period_LS(self.date, self.mag, self.n_threads, self.min_period) # Features based on a phase-folded light curve # such as Eta, slope-percentile, etc. # Should be called after the getPeriodLS() is called. # Created phased a folded light curve. # We use period * 2 to take eclipsing binaries into account. phase_folded_date = self.date % (self.period * 2.) sorted_index = np.argsort(phase_folded_date) folded_date = phase_folded_date[sorted_index] folded_mag = self.mag[sorted_index] # phase Eta self.phase_eta = self.get_eta(folded_mag, self.weighted_std) # Slope percentile. self.slope_per10, self.slope_per90 = \ self.slope_percentile(folded_date, folded_mag) # phase Cusum self.phase_cusum = self.get_cusum(folded_mag) def get_period_LS(self, date, mag, n_threads, min_period): """ Period finding using the Lomb-Scargle algorithm. Finding two periods. The second period is estimated after whitening the first period. Calculating various other features as well using derived periods. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. n_threads : int The number of threads to use. min_period : float The minimum period to calculate. """ # DO NOT CHANGE THESE PARAMETERS. oversampling = 3. hifac = int((max(date) - min(date)) / len(date) / min_period * 2.) # Minimum hifac if hifac < 100: hifac = 100 # Lomb-Scargle. fx, fy, nout, jmax, prob = pLS.fasper(date, mag, oversampling, hifac, n_threads) self.f = fx[jmax] self.period = 1. / self.f self.period_uncertainty = self.get_period_uncertainty(fx, fy, jmax) self.period_log10FAP = \ np.log10(pLS.getSignificance(fx, fy, nout, oversampling)[jmax]) # self.f_SNR1 = fy[jmax] / np.median(fy) self.period_SNR = (fy[jmax] - np.median(fy)) / np.std(fy) # Fit Fourier Series of order 3. order = 3 # Initial guess of Fourier coefficients. p0 = np.ones(order * 2 + 1) date_period = (date % self.period) / self.period p1, success = leastsq(self.residuals, p0, args=(date_period, mag, order)) # fitted_y = self.FourierSeries(p1, date_period, order) # print p1, self.mean, self.median # plt.plot(date_period, self.mag, 'b+') # plt.show() # Derive Fourier features for the first period. # Petersen, J. O., 1986, A&A self.amplitude = np.sqrt(p1[1] ** 2 + p1[2] ** 2) self.r21 = np.sqrt(p1[3] ** 2 + p1[4] ** 2) / self.amplitude self.r31 = np.sqrt(p1[5] ** 2 + p1[6] ** 2) / self.amplitude self.f_phase = np.arctan(-p1[1] / p1[2]) self.phi21 = np.arctan(-p1[3] / p1[4]) - 2. * self.f_phase self.phi31 = np.arctan(-p1[5] / p1[6]) - 3. * self.f_phase """ # Derive a second period. # Whitening a light curve. residual_mag = mag - fitted_y # Lomb-Scargle again to find the second period. omega_top, power_top = search_frequencies(date, residual_mag, err, #LS_kwargs={'generalized':True, 'subtract_mean':True}, n_eval=5000, n_retry=3, n_save=50) self.period2 = 2*np.pi/omega_top[np.where(power_top==np.max(power_top))][0] self.f2 = 1. / self.period2 self.f2_SNR = power_top[np.where(power_top==np.max(power_top))][0] \ * (len(self.date) - 1) / 2. # Fit Fourier Series again. p0 = [1.] * order * 2 date_period = (date % self.period) / self.period p2, success = leastsq(self.residuals, p0, args=(date_period, residual_mag, order)) fitted_y = self.FourierSeries(p2, date_period, order) #plt.plot(date%self.period2, residual_mag, 'b+') #plt.show() # Derive Fourier features for the first second. self.f2_amp = 2. * np.sqrt(p2[1]**2 + p2[2]**2) self.f2_R21 = np.sqrt(p2[3]**2 + p2[4]**2) / self.f2_amp self.f2_R31 = np.sqrt(p2[5]**2 + p2[6]**2) / self.f2_amp self.f2_R41 = np.sqrt(p2[7]**2 + p2[8]**2) / self.f2_amp self.f2_R51 = np.sqrt(p2[9]**2 + p2[10]**2) / self.f2_amp self.f2_phase = np.arctan(-p2[1] / p2[2]) self.f2_phi21 = np.arctan(-p2[3] / p2[4]) - 2. * self.f2_phase self.f2_phi31 = np.arctan(-p2[5] / p2[6]) - 3. * self.f2_phase self.f2_phi41 = np.arctan(-p2[7] / p2[8]) - 4. * self.f2_phase self.f2_phi51 = np.arctan(-p2[9] / p2[10]) - 5. * self.f2_phase # Calculate features using the first and second periods. self.f12_ratio = self.f2 / self.f1 self.f12_remain = self.f1 % self.f2 \ if self.f1 > self.f2 else self.f2 % self.f1 self.f12_amp = self.f2_amp / self.f1_amp self.f12_phase = self.f2_phase - self.f1_phase """ def get_period_uncertainty(self, fx, fy, jmax, fx_width=100): """ Get uncertainty of a period. The uncertainty is defined as the half width of the frequencies around the peak, that becomes lower than average + standard deviation of the power spectrum. Since we may not have fine resolution around the peak, we do not assume it is gaussian. So, no scaling factor of 2.355 (= 2 * sqrt(2 * ln2)) is applied. Parameters ---------- fx : array_like An array of frequencies. fy : array_like An array of amplitudes. jmax : int An index at the peak frequency. fx_width : int, optional Width of power spectrum to calculate uncertainty. Returns ------- p_uncertain : float Period uncertainty. """ # Get subset start_index = jmax - fx_width end_index = jmax + fx_width if start_index < 0: start_index = 0 if end_index > len(fx) - 1: end_index = len(fx) - 1 fx_subset = fx[start_index:end_index] fy_subset = fy[start_index:end_index] fy_mean = np.median(fy_subset) fy_std = np.std(fy_subset) # Find peak max_index = np.argmax(fy_subset) # Find list whose powers become lower than average + std. index = np.where(fy_subset <= fy_mean + fy_std)[0] # Find the edge at left and right. This is the full width. left_index = index[(index < max_index)] if len(left_index) == 0: left_index = 0 else: left_index = left_index[-1] right_index = index[(index > max_index)] if len(right_index) == 0: right_index = len(fy_subset) - 1 else: right_index = right_index[0] # We assume the half of the full width is the period uncertainty. half_width = (1. / fx_subset[left_index] - 1. / fx_subset[right_index]) / 2. period_uncertainty = half_width return period_uncertainty def fourier_series(self, pars, x, order): """ Function to fit Fourier Series. Parameters ---------- x : array_like An array of date divided by period. It doesn't need to be sorted. pars : array_like Fourier series parameters. order : int An order of Fourier series. """ sum = pars[0] for i in range(order): sum += pars[i * 2 + 1] * np.sin(2 * np.pi * (i + 1) * x) \ + pars[i * 2 + 2] * np.cos(2 * np.pi * (i + 1) * x) return sum def get_stetson_k(self, mag, avg, err): """ Return Stetson K feature. Parameters ---------- mag : array_like An array of magnitude. avg : float An average value of magnitudes. err : array_like An array of magnitude errors. Returns ------- stetson_k : float Stetson K value. """ residual = (mag - avg) / err stetson_k = np.sum(np.fabs(residual)) \ / np.sqrt(np.sum(residual * residual)) / np.sqrt(len(mag)) return stetson_k def half_mag_amplitude_ratio(self, mag, avg, weight): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. weight : array_like An array of weight. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) lower_weight = weight[index] lower_weight_sum = np.sum(lower_weight) lower_mag = mag[index] lower_weighted_std = np.sum((lower_mag - avg) ** 2 * lower_weight) / \ lower_weight_sum # For higher (brighter) magnitude than average. index = np.where(mag <= avg) higher_weight = weight[index] higher_weight_sum = np.sum(higher_weight) higher_mag = mag[index] higher_weighted_std = np.sum((higher_mag - avg) ** 2 * higher_weight) / \ higher_weight_sum # Return ratio. return np.sqrt(lower_weighted_std / higher_weighted_std) def half_mag_amplitude_ratio2(self, mag, avg): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) fainter_mag = mag[index] lower_sum = np.sum((fainter_mag - avg) ** 2) / len(fainter_mag) # For higher (brighter) magnitude than average. index = np.where(mag <= avg) brighter_mag = mag[index] higher_sum = np.sum((avg - brighter_mag) ** 2) / len(brighter_mag) # Return ratio. return np.sqrt(lower_sum / higher_sum) def get_eta(self, mag, std): """ Return Eta feature. Parameters ---------- mag : array_like An array of magnitudes. std : array_like A standard deviation of magnitudes. Returns ------- eta : float The value of Eta index. """ diff = mag[1:] - mag[:len(mag) - 1] eta = np.sum(diff * diff) / (len(mag) - 1.) / std / std return eta def slope_percentile(self, date, mag): """ Return 10% and 90% percentile of slope. Parameters ---------- date : array_like An array of phase-folded date. Sorted. mag : array_like An array of phase-folded magnitudes. Sorted by date. Returns ------- per_10 : float 10% percentile values of slope. per_90 : float 90% percentile values of slope. """ date_diff = date[1:] - date[:len(date) - 1] mag_diff = mag[1:] - mag[:len(mag) - 1] # Remove zero mag_diff. index = np.where(mag_diff != 0.) date_diff = date_diff[index] mag_diff = mag_diff[index] # Derive slope. slope = date_diff / mag_diff percentile_10 = np.percentile(slope, 10.) percentile_90 = np.percentile(slope, 90.) return percentile_10, percentile_90 def get_cusum(self, mag): """ Return max - min of cumulative sum. Parameters ---------- mag : array_like An array of magnitudes. Returns ------- mm_cusum : float Max - min of cumulative sum. """ c = np.cumsum(mag - self.weighted_mean) / len(mag) / self.weighted_std return np.max(c) - np.min(c) def get_features2(self): """ Return all features with its names. Returns ------- names : list Feature names. values : list Feature values """ feature_names = [] feature_values = [] # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): # Omit input variables such as date, mag, err, etc. if not (name == 'date' or name == 'mag' or name == 'err' or name == 'n_threads' or name == 'min_period'): # Filter some other unnecessary features. if not (name == 'f' or name == 'f_phase' or name == 'period_log10FAP' or name == 'weight' or name == 'weighted_sum' or name == 'median' or name == 'mean' or name == 'std'): feature_names.append(name) # Sort by the names. # Sorting should be done to keep maintaining the same order of features. feature_names.sort() # Get feature values. for name in feature_names: feature_values.append(all_vars[name]) return feature_names, feature_values def get_features(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ ''' features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features ''' return self.get_features_all() def get_features_all(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list_all: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features
dwkim78/upsilon
upsilon/extract_features/extract_features.py
ExtractFeatures.fourier_series
python
def fourier_series(self, pars, x, order): sum = pars[0] for i in range(order): sum += pars[i * 2 + 1] * np.sin(2 * np.pi * (i + 1) * x) \ + pars[i * 2 + 2] * np.cos(2 * np.pi * (i + 1) * x) return sum
Function to fit Fourier Series. Parameters ---------- x : array_like An array of date divided by period. It doesn't need to be sorted. pars : array_like Fourier series parameters. order : int An order of Fourier series.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/extract_features.py#L360-L379
null
class ExtractFeatures: """ Extract variability features of a light curve. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. err : array_like, optional An array of magnitude error. If None, std(mag) will be used. n_threads : int, optional The number of cores to use to derive periods. min_period : float, optional The minimum period to calculate. """ def __init__(self, date, mag, err=None, n_threads=4, min_period=0.03): # Set basic values. if not isinstance(date, np.ndarray): date = np.array(date) if not isinstance(mag, np.ndarray): mag = np.array(mag) self.date = date self.mag = mag if err is not None: if not isinstance(err, np.ndarray): err = np.array(err) self.err = err else: self.err = np.ones(len(self.mag)) * np.std(self.mag) # Check length. if (len(self.date) != len(self.mag)) or \ (len(self.date) != len(self.err)) or \ (len(self.mag) != len(self.err)): raise RuntimeError('The length of date, mag, and err must be same.') # if the number of data points is too small. min_n_data = 80 if len(self.date) < min_n_data: warnings.warn('The number of data points are less than %d.' % min_n_data) n_threads = int(n_threads) if n_threads > multiprocessing.cpu_count(): self.n_threads = multiprocessing.cpu_count() else: if n_threads <= 0: self.n_threads = 1 else: self.n_threads = n_threads min_period = float(min_period) if min_period <= 0: self.min_period = 0.03 else: self.min_period = min_period def run(self): """Run feature extraction modules.""" # shallow_run must be executed prior to deep_run # since shallow_run calculates several values needed for deep_run. self.shallow_run() self.deep_run() def shallow_run(self): """Derive not-period-based features.""" # Number of data points self.n_points = len(self.date) # Weight calculation. # All zero values. if not self.err.any(): self.err = np.ones(len(self.mag)) * np.std(self.mag) # Some zero values. elif not self.err.all(): np.putmask(self.err, self.err==0, np.median(self.err)) self.weight = 1. / self.err self.weighted_sum = np.sum(self.weight) # Simple statistics, mean, median and std. self.mean = np.mean(self.mag) self.median = np.median(self.mag) self.std = np.std(self.mag) # Weighted mean and std. self.weighted_mean = np.sum(self.mag * self.weight) / self.weighted_sum self.weighted_std = np.sqrt(np.sum((self.mag - self.weighted_mean) ** 2 \ * self.weight) / self.weighted_sum) # Skewness and kurtosis. self.skewness = ss.skew(self.mag) self.kurtosis = ss.kurtosis(self.mag) # Normalization-test. Shapiro-Wilk test. shapiro = ss.shapiro(self.mag) self.shapiro_w = shapiro[0] # self.shapiro_log10p = np.log10(shapiro[1]) # Percentile features. self.quartile31 = np.percentile(self.mag, 75) \ - np.percentile(self.mag, 25) # Stetson K. self.stetson_k = self.get_stetson_k(self.mag, self.median, self.err) # Ratio between higher and lower amplitude than average. self.hl_amp_ratio = self.half_mag_amplitude_ratio( self.mag, self.median, self.weight) # This second function's value is very similar with the above one. # self.hl_amp_ratio2 = self.half_mag_amplitude_ratio2( # self.mag, self.median) # Cusum self.cusum = self.get_cusum(self.mag) # Eta self.eta = self.get_eta(self.mag, self.weighted_std) def deep_run(self): """Derive period-based features.""" # Lomb-Scargle period finding. self.get_period_LS(self.date, self.mag, self.n_threads, self.min_period) # Features based on a phase-folded light curve # such as Eta, slope-percentile, etc. # Should be called after the getPeriodLS() is called. # Created phased a folded light curve. # We use period * 2 to take eclipsing binaries into account. phase_folded_date = self.date % (self.period * 2.) sorted_index = np.argsort(phase_folded_date) folded_date = phase_folded_date[sorted_index] folded_mag = self.mag[sorted_index] # phase Eta self.phase_eta = self.get_eta(folded_mag, self.weighted_std) # Slope percentile. self.slope_per10, self.slope_per90 = \ self.slope_percentile(folded_date, folded_mag) # phase Cusum self.phase_cusum = self.get_cusum(folded_mag) def get_period_LS(self, date, mag, n_threads, min_period): """ Period finding using the Lomb-Scargle algorithm. Finding two periods. The second period is estimated after whitening the first period. Calculating various other features as well using derived periods. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. n_threads : int The number of threads to use. min_period : float The minimum period to calculate. """ # DO NOT CHANGE THESE PARAMETERS. oversampling = 3. hifac = int((max(date) - min(date)) / len(date) / min_period * 2.) # Minimum hifac if hifac < 100: hifac = 100 # Lomb-Scargle. fx, fy, nout, jmax, prob = pLS.fasper(date, mag, oversampling, hifac, n_threads) self.f = fx[jmax] self.period = 1. / self.f self.period_uncertainty = self.get_period_uncertainty(fx, fy, jmax) self.period_log10FAP = \ np.log10(pLS.getSignificance(fx, fy, nout, oversampling)[jmax]) # self.f_SNR1 = fy[jmax] / np.median(fy) self.period_SNR = (fy[jmax] - np.median(fy)) / np.std(fy) # Fit Fourier Series of order 3. order = 3 # Initial guess of Fourier coefficients. p0 = np.ones(order * 2 + 1) date_period = (date % self.period) / self.period p1, success = leastsq(self.residuals, p0, args=(date_period, mag, order)) # fitted_y = self.FourierSeries(p1, date_period, order) # print p1, self.mean, self.median # plt.plot(date_period, self.mag, 'b+') # plt.show() # Derive Fourier features for the first period. # Petersen, J. O., 1986, A&A self.amplitude = np.sqrt(p1[1] ** 2 + p1[2] ** 2) self.r21 = np.sqrt(p1[3] ** 2 + p1[4] ** 2) / self.amplitude self.r31 = np.sqrt(p1[5] ** 2 + p1[6] ** 2) / self.amplitude self.f_phase = np.arctan(-p1[1] / p1[2]) self.phi21 = np.arctan(-p1[3] / p1[4]) - 2. * self.f_phase self.phi31 = np.arctan(-p1[5] / p1[6]) - 3. * self.f_phase """ # Derive a second period. # Whitening a light curve. residual_mag = mag - fitted_y # Lomb-Scargle again to find the second period. omega_top, power_top = search_frequencies(date, residual_mag, err, #LS_kwargs={'generalized':True, 'subtract_mean':True}, n_eval=5000, n_retry=3, n_save=50) self.period2 = 2*np.pi/omega_top[np.where(power_top==np.max(power_top))][0] self.f2 = 1. / self.period2 self.f2_SNR = power_top[np.where(power_top==np.max(power_top))][0] \ * (len(self.date) - 1) / 2. # Fit Fourier Series again. p0 = [1.] * order * 2 date_period = (date % self.period) / self.period p2, success = leastsq(self.residuals, p0, args=(date_period, residual_mag, order)) fitted_y = self.FourierSeries(p2, date_period, order) #plt.plot(date%self.period2, residual_mag, 'b+') #plt.show() # Derive Fourier features for the first second. self.f2_amp = 2. * np.sqrt(p2[1]**2 + p2[2]**2) self.f2_R21 = np.sqrt(p2[3]**2 + p2[4]**2) / self.f2_amp self.f2_R31 = np.sqrt(p2[5]**2 + p2[6]**2) / self.f2_amp self.f2_R41 = np.sqrt(p2[7]**2 + p2[8]**2) / self.f2_amp self.f2_R51 = np.sqrt(p2[9]**2 + p2[10]**2) / self.f2_amp self.f2_phase = np.arctan(-p2[1] / p2[2]) self.f2_phi21 = np.arctan(-p2[3] / p2[4]) - 2. * self.f2_phase self.f2_phi31 = np.arctan(-p2[5] / p2[6]) - 3. * self.f2_phase self.f2_phi41 = np.arctan(-p2[7] / p2[8]) - 4. * self.f2_phase self.f2_phi51 = np.arctan(-p2[9] / p2[10]) - 5. * self.f2_phase # Calculate features using the first and second periods. self.f12_ratio = self.f2 / self.f1 self.f12_remain = self.f1 % self.f2 \ if self.f1 > self.f2 else self.f2 % self.f1 self.f12_amp = self.f2_amp / self.f1_amp self.f12_phase = self.f2_phase - self.f1_phase """ def get_period_uncertainty(self, fx, fy, jmax, fx_width=100): """ Get uncertainty of a period. The uncertainty is defined as the half width of the frequencies around the peak, that becomes lower than average + standard deviation of the power spectrum. Since we may not have fine resolution around the peak, we do not assume it is gaussian. So, no scaling factor of 2.355 (= 2 * sqrt(2 * ln2)) is applied. Parameters ---------- fx : array_like An array of frequencies. fy : array_like An array of amplitudes. jmax : int An index at the peak frequency. fx_width : int, optional Width of power spectrum to calculate uncertainty. Returns ------- p_uncertain : float Period uncertainty. """ # Get subset start_index = jmax - fx_width end_index = jmax + fx_width if start_index < 0: start_index = 0 if end_index > len(fx) - 1: end_index = len(fx) - 1 fx_subset = fx[start_index:end_index] fy_subset = fy[start_index:end_index] fy_mean = np.median(fy_subset) fy_std = np.std(fy_subset) # Find peak max_index = np.argmax(fy_subset) # Find list whose powers become lower than average + std. index = np.where(fy_subset <= fy_mean + fy_std)[0] # Find the edge at left and right. This is the full width. left_index = index[(index < max_index)] if len(left_index) == 0: left_index = 0 else: left_index = left_index[-1] right_index = index[(index > max_index)] if len(right_index) == 0: right_index = len(fy_subset) - 1 else: right_index = right_index[0] # We assume the half of the full width is the period uncertainty. half_width = (1. / fx_subset[left_index] - 1. / fx_subset[right_index]) / 2. period_uncertainty = half_width return period_uncertainty def residuals(self, pars, x, y, order): """ Residual of Fourier Series. Parameters ---------- pars : array_like Fourier series parameters. x : array_like An array of date. y : array_like An array of true values to fit. order : int An order of Fourier Series. """ return y - self.fourier_series(pars, x, order) def get_stetson_k(self, mag, avg, err): """ Return Stetson K feature. Parameters ---------- mag : array_like An array of magnitude. avg : float An average value of magnitudes. err : array_like An array of magnitude errors. Returns ------- stetson_k : float Stetson K value. """ residual = (mag - avg) / err stetson_k = np.sum(np.fabs(residual)) \ / np.sqrt(np.sum(residual * residual)) / np.sqrt(len(mag)) return stetson_k def half_mag_amplitude_ratio(self, mag, avg, weight): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. weight : array_like An array of weight. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) lower_weight = weight[index] lower_weight_sum = np.sum(lower_weight) lower_mag = mag[index] lower_weighted_std = np.sum((lower_mag - avg) ** 2 * lower_weight) / \ lower_weight_sum # For higher (brighter) magnitude than average. index = np.where(mag <= avg) higher_weight = weight[index] higher_weight_sum = np.sum(higher_weight) higher_mag = mag[index] higher_weighted_std = np.sum((higher_mag - avg) ** 2 * higher_weight) / \ higher_weight_sum # Return ratio. return np.sqrt(lower_weighted_std / higher_weighted_std) def half_mag_amplitude_ratio2(self, mag, avg): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) fainter_mag = mag[index] lower_sum = np.sum((fainter_mag - avg) ** 2) / len(fainter_mag) # For higher (brighter) magnitude than average. index = np.where(mag <= avg) brighter_mag = mag[index] higher_sum = np.sum((avg - brighter_mag) ** 2) / len(brighter_mag) # Return ratio. return np.sqrt(lower_sum / higher_sum) def get_eta(self, mag, std): """ Return Eta feature. Parameters ---------- mag : array_like An array of magnitudes. std : array_like A standard deviation of magnitudes. Returns ------- eta : float The value of Eta index. """ diff = mag[1:] - mag[:len(mag) - 1] eta = np.sum(diff * diff) / (len(mag) - 1.) / std / std return eta def slope_percentile(self, date, mag): """ Return 10% and 90% percentile of slope. Parameters ---------- date : array_like An array of phase-folded date. Sorted. mag : array_like An array of phase-folded magnitudes. Sorted by date. Returns ------- per_10 : float 10% percentile values of slope. per_90 : float 90% percentile values of slope. """ date_diff = date[1:] - date[:len(date) - 1] mag_diff = mag[1:] - mag[:len(mag) - 1] # Remove zero mag_diff. index = np.where(mag_diff != 0.) date_diff = date_diff[index] mag_diff = mag_diff[index] # Derive slope. slope = date_diff / mag_diff percentile_10 = np.percentile(slope, 10.) percentile_90 = np.percentile(slope, 90.) return percentile_10, percentile_90 def get_cusum(self, mag): """ Return max - min of cumulative sum. Parameters ---------- mag : array_like An array of magnitudes. Returns ------- mm_cusum : float Max - min of cumulative sum. """ c = np.cumsum(mag - self.weighted_mean) / len(mag) / self.weighted_std return np.max(c) - np.min(c) def get_features2(self): """ Return all features with its names. Returns ------- names : list Feature names. values : list Feature values """ feature_names = [] feature_values = [] # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): # Omit input variables such as date, mag, err, etc. if not (name == 'date' or name == 'mag' or name == 'err' or name == 'n_threads' or name == 'min_period'): # Filter some other unnecessary features. if not (name == 'f' or name == 'f_phase' or name == 'period_log10FAP' or name == 'weight' or name == 'weighted_sum' or name == 'median' or name == 'mean' or name == 'std'): feature_names.append(name) # Sort by the names. # Sorting should be done to keep maintaining the same order of features. feature_names.sort() # Get feature values. for name in feature_names: feature_values.append(all_vars[name]) return feature_names, feature_values def get_features(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ ''' features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features ''' return self.get_features_all() def get_features_all(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list_all: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features
dwkim78/upsilon
upsilon/extract_features/extract_features.py
ExtractFeatures.get_stetson_k
python
def get_stetson_k(self, mag, avg, err): residual = (mag - avg) / err stetson_k = np.sum(np.fabs(residual)) \ / np.sqrt(np.sum(residual * residual)) / np.sqrt(len(mag)) return stetson_k
Return Stetson K feature. Parameters ---------- mag : array_like An array of magnitude. avg : float An average value of magnitudes. err : array_like An array of magnitude errors. Returns ------- stetson_k : float Stetson K value.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/extract_features.py#L381-L404
null
class ExtractFeatures: """ Extract variability features of a light curve. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. err : array_like, optional An array of magnitude error. If None, std(mag) will be used. n_threads : int, optional The number of cores to use to derive periods. min_period : float, optional The minimum period to calculate. """ def __init__(self, date, mag, err=None, n_threads=4, min_period=0.03): # Set basic values. if not isinstance(date, np.ndarray): date = np.array(date) if not isinstance(mag, np.ndarray): mag = np.array(mag) self.date = date self.mag = mag if err is not None: if not isinstance(err, np.ndarray): err = np.array(err) self.err = err else: self.err = np.ones(len(self.mag)) * np.std(self.mag) # Check length. if (len(self.date) != len(self.mag)) or \ (len(self.date) != len(self.err)) or \ (len(self.mag) != len(self.err)): raise RuntimeError('The length of date, mag, and err must be same.') # if the number of data points is too small. min_n_data = 80 if len(self.date) < min_n_data: warnings.warn('The number of data points are less than %d.' % min_n_data) n_threads = int(n_threads) if n_threads > multiprocessing.cpu_count(): self.n_threads = multiprocessing.cpu_count() else: if n_threads <= 0: self.n_threads = 1 else: self.n_threads = n_threads min_period = float(min_period) if min_period <= 0: self.min_period = 0.03 else: self.min_period = min_period def run(self): """Run feature extraction modules.""" # shallow_run must be executed prior to deep_run # since shallow_run calculates several values needed for deep_run. self.shallow_run() self.deep_run() def shallow_run(self): """Derive not-period-based features.""" # Number of data points self.n_points = len(self.date) # Weight calculation. # All zero values. if not self.err.any(): self.err = np.ones(len(self.mag)) * np.std(self.mag) # Some zero values. elif not self.err.all(): np.putmask(self.err, self.err==0, np.median(self.err)) self.weight = 1. / self.err self.weighted_sum = np.sum(self.weight) # Simple statistics, mean, median and std. self.mean = np.mean(self.mag) self.median = np.median(self.mag) self.std = np.std(self.mag) # Weighted mean and std. self.weighted_mean = np.sum(self.mag * self.weight) / self.weighted_sum self.weighted_std = np.sqrt(np.sum((self.mag - self.weighted_mean) ** 2 \ * self.weight) / self.weighted_sum) # Skewness and kurtosis. self.skewness = ss.skew(self.mag) self.kurtosis = ss.kurtosis(self.mag) # Normalization-test. Shapiro-Wilk test. shapiro = ss.shapiro(self.mag) self.shapiro_w = shapiro[0] # self.shapiro_log10p = np.log10(shapiro[1]) # Percentile features. self.quartile31 = np.percentile(self.mag, 75) \ - np.percentile(self.mag, 25) # Stetson K. self.stetson_k = self.get_stetson_k(self.mag, self.median, self.err) # Ratio between higher and lower amplitude than average. self.hl_amp_ratio = self.half_mag_amplitude_ratio( self.mag, self.median, self.weight) # This second function's value is very similar with the above one. # self.hl_amp_ratio2 = self.half_mag_amplitude_ratio2( # self.mag, self.median) # Cusum self.cusum = self.get_cusum(self.mag) # Eta self.eta = self.get_eta(self.mag, self.weighted_std) def deep_run(self): """Derive period-based features.""" # Lomb-Scargle period finding. self.get_period_LS(self.date, self.mag, self.n_threads, self.min_period) # Features based on a phase-folded light curve # such as Eta, slope-percentile, etc. # Should be called after the getPeriodLS() is called. # Created phased a folded light curve. # We use period * 2 to take eclipsing binaries into account. phase_folded_date = self.date % (self.period * 2.) sorted_index = np.argsort(phase_folded_date) folded_date = phase_folded_date[sorted_index] folded_mag = self.mag[sorted_index] # phase Eta self.phase_eta = self.get_eta(folded_mag, self.weighted_std) # Slope percentile. self.slope_per10, self.slope_per90 = \ self.slope_percentile(folded_date, folded_mag) # phase Cusum self.phase_cusum = self.get_cusum(folded_mag) def get_period_LS(self, date, mag, n_threads, min_period): """ Period finding using the Lomb-Scargle algorithm. Finding two periods. The second period is estimated after whitening the first period. Calculating various other features as well using derived periods. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. n_threads : int The number of threads to use. min_period : float The minimum period to calculate. """ # DO NOT CHANGE THESE PARAMETERS. oversampling = 3. hifac = int((max(date) - min(date)) / len(date) / min_period * 2.) # Minimum hifac if hifac < 100: hifac = 100 # Lomb-Scargle. fx, fy, nout, jmax, prob = pLS.fasper(date, mag, oversampling, hifac, n_threads) self.f = fx[jmax] self.period = 1. / self.f self.period_uncertainty = self.get_period_uncertainty(fx, fy, jmax) self.period_log10FAP = \ np.log10(pLS.getSignificance(fx, fy, nout, oversampling)[jmax]) # self.f_SNR1 = fy[jmax] / np.median(fy) self.period_SNR = (fy[jmax] - np.median(fy)) / np.std(fy) # Fit Fourier Series of order 3. order = 3 # Initial guess of Fourier coefficients. p0 = np.ones(order * 2 + 1) date_period = (date % self.period) / self.period p1, success = leastsq(self.residuals, p0, args=(date_period, mag, order)) # fitted_y = self.FourierSeries(p1, date_period, order) # print p1, self.mean, self.median # plt.plot(date_period, self.mag, 'b+') # plt.show() # Derive Fourier features for the first period. # Petersen, J. O., 1986, A&A self.amplitude = np.sqrt(p1[1] ** 2 + p1[2] ** 2) self.r21 = np.sqrt(p1[3] ** 2 + p1[4] ** 2) / self.amplitude self.r31 = np.sqrt(p1[5] ** 2 + p1[6] ** 2) / self.amplitude self.f_phase = np.arctan(-p1[1] / p1[2]) self.phi21 = np.arctan(-p1[3] / p1[4]) - 2. * self.f_phase self.phi31 = np.arctan(-p1[5] / p1[6]) - 3. * self.f_phase """ # Derive a second period. # Whitening a light curve. residual_mag = mag - fitted_y # Lomb-Scargle again to find the second period. omega_top, power_top = search_frequencies(date, residual_mag, err, #LS_kwargs={'generalized':True, 'subtract_mean':True}, n_eval=5000, n_retry=3, n_save=50) self.period2 = 2*np.pi/omega_top[np.where(power_top==np.max(power_top))][0] self.f2 = 1. / self.period2 self.f2_SNR = power_top[np.where(power_top==np.max(power_top))][0] \ * (len(self.date) - 1) / 2. # Fit Fourier Series again. p0 = [1.] * order * 2 date_period = (date % self.period) / self.period p2, success = leastsq(self.residuals, p0, args=(date_period, residual_mag, order)) fitted_y = self.FourierSeries(p2, date_period, order) #plt.plot(date%self.period2, residual_mag, 'b+') #plt.show() # Derive Fourier features for the first second. self.f2_amp = 2. * np.sqrt(p2[1]**2 + p2[2]**2) self.f2_R21 = np.sqrt(p2[3]**2 + p2[4]**2) / self.f2_amp self.f2_R31 = np.sqrt(p2[5]**2 + p2[6]**2) / self.f2_amp self.f2_R41 = np.sqrt(p2[7]**2 + p2[8]**2) / self.f2_amp self.f2_R51 = np.sqrt(p2[9]**2 + p2[10]**2) / self.f2_amp self.f2_phase = np.arctan(-p2[1] / p2[2]) self.f2_phi21 = np.arctan(-p2[3] / p2[4]) - 2. * self.f2_phase self.f2_phi31 = np.arctan(-p2[5] / p2[6]) - 3. * self.f2_phase self.f2_phi41 = np.arctan(-p2[7] / p2[8]) - 4. * self.f2_phase self.f2_phi51 = np.arctan(-p2[9] / p2[10]) - 5. * self.f2_phase # Calculate features using the first and second periods. self.f12_ratio = self.f2 / self.f1 self.f12_remain = self.f1 % self.f2 \ if self.f1 > self.f2 else self.f2 % self.f1 self.f12_amp = self.f2_amp / self.f1_amp self.f12_phase = self.f2_phase - self.f1_phase """ def get_period_uncertainty(self, fx, fy, jmax, fx_width=100): """ Get uncertainty of a period. The uncertainty is defined as the half width of the frequencies around the peak, that becomes lower than average + standard deviation of the power spectrum. Since we may not have fine resolution around the peak, we do not assume it is gaussian. So, no scaling factor of 2.355 (= 2 * sqrt(2 * ln2)) is applied. Parameters ---------- fx : array_like An array of frequencies. fy : array_like An array of amplitudes. jmax : int An index at the peak frequency. fx_width : int, optional Width of power spectrum to calculate uncertainty. Returns ------- p_uncertain : float Period uncertainty. """ # Get subset start_index = jmax - fx_width end_index = jmax + fx_width if start_index < 0: start_index = 0 if end_index > len(fx) - 1: end_index = len(fx) - 1 fx_subset = fx[start_index:end_index] fy_subset = fy[start_index:end_index] fy_mean = np.median(fy_subset) fy_std = np.std(fy_subset) # Find peak max_index = np.argmax(fy_subset) # Find list whose powers become lower than average + std. index = np.where(fy_subset <= fy_mean + fy_std)[0] # Find the edge at left and right. This is the full width. left_index = index[(index < max_index)] if len(left_index) == 0: left_index = 0 else: left_index = left_index[-1] right_index = index[(index > max_index)] if len(right_index) == 0: right_index = len(fy_subset) - 1 else: right_index = right_index[0] # We assume the half of the full width is the period uncertainty. half_width = (1. / fx_subset[left_index] - 1. / fx_subset[right_index]) / 2. period_uncertainty = half_width return period_uncertainty def residuals(self, pars, x, y, order): """ Residual of Fourier Series. Parameters ---------- pars : array_like Fourier series parameters. x : array_like An array of date. y : array_like An array of true values to fit. order : int An order of Fourier Series. """ return y - self.fourier_series(pars, x, order) def fourier_series(self, pars, x, order): """ Function to fit Fourier Series. Parameters ---------- x : array_like An array of date divided by period. It doesn't need to be sorted. pars : array_like Fourier series parameters. order : int An order of Fourier series. """ sum = pars[0] for i in range(order): sum += pars[i * 2 + 1] * np.sin(2 * np.pi * (i + 1) * x) \ + pars[i * 2 + 2] * np.cos(2 * np.pi * (i + 1) * x) return sum def half_mag_amplitude_ratio(self, mag, avg, weight): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. weight : array_like An array of weight. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) lower_weight = weight[index] lower_weight_sum = np.sum(lower_weight) lower_mag = mag[index] lower_weighted_std = np.sum((lower_mag - avg) ** 2 * lower_weight) / \ lower_weight_sum # For higher (brighter) magnitude than average. index = np.where(mag <= avg) higher_weight = weight[index] higher_weight_sum = np.sum(higher_weight) higher_mag = mag[index] higher_weighted_std = np.sum((higher_mag - avg) ** 2 * higher_weight) / \ higher_weight_sum # Return ratio. return np.sqrt(lower_weighted_std / higher_weighted_std) def half_mag_amplitude_ratio2(self, mag, avg): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) fainter_mag = mag[index] lower_sum = np.sum((fainter_mag - avg) ** 2) / len(fainter_mag) # For higher (brighter) magnitude than average. index = np.where(mag <= avg) brighter_mag = mag[index] higher_sum = np.sum((avg - brighter_mag) ** 2) / len(brighter_mag) # Return ratio. return np.sqrt(lower_sum / higher_sum) def get_eta(self, mag, std): """ Return Eta feature. Parameters ---------- mag : array_like An array of magnitudes. std : array_like A standard deviation of magnitudes. Returns ------- eta : float The value of Eta index. """ diff = mag[1:] - mag[:len(mag) - 1] eta = np.sum(diff * diff) / (len(mag) - 1.) / std / std return eta def slope_percentile(self, date, mag): """ Return 10% and 90% percentile of slope. Parameters ---------- date : array_like An array of phase-folded date. Sorted. mag : array_like An array of phase-folded magnitudes. Sorted by date. Returns ------- per_10 : float 10% percentile values of slope. per_90 : float 90% percentile values of slope. """ date_diff = date[1:] - date[:len(date) - 1] mag_diff = mag[1:] - mag[:len(mag) - 1] # Remove zero mag_diff. index = np.where(mag_diff != 0.) date_diff = date_diff[index] mag_diff = mag_diff[index] # Derive slope. slope = date_diff / mag_diff percentile_10 = np.percentile(slope, 10.) percentile_90 = np.percentile(slope, 90.) return percentile_10, percentile_90 def get_cusum(self, mag): """ Return max - min of cumulative sum. Parameters ---------- mag : array_like An array of magnitudes. Returns ------- mm_cusum : float Max - min of cumulative sum. """ c = np.cumsum(mag - self.weighted_mean) / len(mag) / self.weighted_std return np.max(c) - np.min(c) def get_features2(self): """ Return all features with its names. Returns ------- names : list Feature names. values : list Feature values """ feature_names = [] feature_values = [] # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): # Omit input variables such as date, mag, err, etc. if not (name == 'date' or name == 'mag' or name == 'err' or name == 'n_threads' or name == 'min_period'): # Filter some other unnecessary features. if not (name == 'f' or name == 'f_phase' or name == 'period_log10FAP' or name == 'weight' or name == 'weighted_sum' or name == 'median' or name == 'mean' or name == 'std'): feature_names.append(name) # Sort by the names. # Sorting should be done to keep maintaining the same order of features. feature_names.sort() # Get feature values. for name in feature_names: feature_values.append(all_vars[name]) return feature_names, feature_values def get_features(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ ''' features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features ''' return self.get_features_all() def get_features_all(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list_all: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features
dwkim78/upsilon
upsilon/extract_features/extract_features.py
ExtractFeatures.half_mag_amplitude_ratio
python
def half_mag_amplitude_ratio(self, mag, avg, weight): # For lower (fainter) magnitude than average. index = np.where(mag > avg) lower_weight = weight[index] lower_weight_sum = np.sum(lower_weight) lower_mag = mag[index] lower_weighted_std = np.sum((lower_mag - avg) ** 2 * lower_weight) / \ lower_weight_sum # For higher (brighter) magnitude than average. index = np.where(mag <= avg) higher_weight = weight[index] higher_weight_sum = np.sum(higher_weight) higher_mag = mag[index] higher_weighted_std = np.sum((higher_mag - avg) ** 2 * higher_weight) / \ higher_weight_sum # Return ratio. return np.sqrt(lower_weighted_std / higher_weighted_std)
Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. weight : array_like An array of weight. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/extract_features.py#L406-L449
null
class ExtractFeatures: """ Extract variability features of a light curve. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. err : array_like, optional An array of magnitude error. If None, std(mag) will be used. n_threads : int, optional The number of cores to use to derive periods. min_period : float, optional The minimum period to calculate. """ def __init__(self, date, mag, err=None, n_threads=4, min_period=0.03): # Set basic values. if not isinstance(date, np.ndarray): date = np.array(date) if not isinstance(mag, np.ndarray): mag = np.array(mag) self.date = date self.mag = mag if err is not None: if not isinstance(err, np.ndarray): err = np.array(err) self.err = err else: self.err = np.ones(len(self.mag)) * np.std(self.mag) # Check length. if (len(self.date) != len(self.mag)) or \ (len(self.date) != len(self.err)) or \ (len(self.mag) != len(self.err)): raise RuntimeError('The length of date, mag, and err must be same.') # if the number of data points is too small. min_n_data = 80 if len(self.date) < min_n_data: warnings.warn('The number of data points are less than %d.' % min_n_data) n_threads = int(n_threads) if n_threads > multiprocessing.cpu_count(): self.n_threads = multiprocessing.cpu_count() else: if n_threads <= 0: self.n_threads = 1 else: self.n_threads = n_threads min_period = float(min_period) if min_period <= 0: self.min_period = 0.03 else: self.min_period = min_period def run(self): """Run feature extraction modules.""" # shallow_run must be executed prior to deep_run # since shallow_run calculates several values needed for deep_run. self.shallow_run() self.deep_run() def shallow_run(self): """Derive not-period-based features.""" # Number of data points self.n_points = len(self.date) # Weight calculation. # All zero values. if not self.err.any(): self.err = np.ones(len(self.mag)) * np.std(self.mag) # Some zero values. elif not self.err.all(): np.putmask(self.err, self.err==0, np.median(self.err)) self.weight = 1. / self.err self.weighted_sum = np.sum(self.weight) # Simple statistics, mean, median and std. self.mean = np.mean(self.mag) self.median = np.median(self.mag) self.std = np.std(self.mag) # Weighted mean and std. self.weighted_mean = np.sum(self.mag * self.weight) / self.weighted_sum self.weighted_std = np.sqrt(np.sum((self.mag - self.weighted_mean) ** 2 \ * self.weight) / self.weighted_sum) # Skewness and kurtosis. self.skewness = ss.skew(self.mag) self.kurtosis = ss.kurtosis(self.mag) # Normalization-test. Shapiro-Wilk test. shapiro = ss.shapiro(self.mag) self.shapiro_w = shapiro[0] # self.shapiro_log10p = np.log10(shapiro[1]) # Percentile features. self.quartile31 = np.percentile(self.mag, 75) \ - np.percentile(self.mag, 25) # Stetson K. self.stetson_k = self.get_stetson_k(self.mag, self.median, self.err) # Ratio between higher and lower amplitude than average. self.hl_amp_ratio = self.half_mag_amplitude_ratio( self.mag, self.median, self.weight) # This second function's value is very similar with the above one. # self.hl_amp_ratio2 = self.half_mag_amplitude_ratio2( # self.mag, self.median) # Cusum self.cusum = self.get_cusum(self.mag) # Eta self.eta = self.get_eta(self.mag, self.weighted_std) def deep_run(self): """Derive period-based features.""" # Lomb-Scargle period finding. self.get_period_LS(self.date, self.mag, self.n_threads, self.min_period) # Features based on a phase-folded light curve # such as Eta, slope-percentile, etc. # Should be called after the getPeriodLS() is called. # Created phased a folded light curve. # We use period * 2 to take eclipsing binaries into account. phase_folded_date = self.date % (self.period * 2.) sorted_index = np.argsort(phase_folded_date) folded_date = phase_folded_date[sorted_index] folded_mag = self.mag[sorted_index] # phase Eta self.phase_eta = self.get_eta(folded_mag, self.weighted_std) # Slope percentile. self.slope_per10, self.slope_per90 = \ self.slope_percentile(folded_date, folded_mag) # phase Cusum self.phase_cusum = self.get_cusum(folded_mag) def get_period_LS(self, date, mag, n_threads, min_period): """ Period finding using the Lomb-Scargle algorithm. Finding two periods. The second period is estimated after whitening the first period. Calculating various other features as well using derived periods. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. n_threads : int The number of threads to use. min_period : float The minimum period to calculate. """ # DO NOT CHANGE THESE PARAMETERS. oversampling = 3. hifac = int((max(date) - min(date)) / len(date) / min_period * 2.) # Minimum hifac if hifac < 100: hifac = 100 # Lomb-Scargle. fx, fy, nout, jmax, prob = pLS.fasper(date, mag, oversampling, hifac, n_threads) self.f = fx[jmax] self.period = 1. / self.f self.period_uncertainty = self.get_period_uncertainty(fx, fy, jmax) self.period_log10FAP = \ np.log10(pLS.getSignificance(fx, fy, nout, oversampling)[jmax]) # self.f_SNR1 = fy[jmax] / np.median(fy) self.period_SNR = (fy[jmax] - np.median(fy)) / np.std(fy) # Fit Fourier Series of order 3. order = 3 # Initial guess of Fourier coefficients. p0 = np.ones(order * 2 + 1) date_period = (date % self.period) / self.period p1, success = leastsq(self.residuals, p0, args=(date_period, mag, order)) # fitted_y = self.FourierSeries(p1, date_period, order) # print p1, self.mean, self.median # plt.plot(date_period, self.mag, 'b+') # plt.show() # Derive Fourier features for the first period. # Petersen, J. O., 1986, A&A self.amplitude = np.sqrt(p1[1] ** 2 + p1[2] ** 2) self.r21 = np.sqrt(p1[3] ** 2 + p1[4] ** 2) / self.amplitude self.r31 = np.sqrt(p1[5] ** 2 + p1[6] ** 2) / self.amplitude self.f_phase = np.arctan(-p1[1] / p1[2]) self.phi21 = np.arctan(-p1[3] / p1[4]) - 2. * self.f_phase self.phi31 = np.arctan(-p1[5] / p1[6]) - 3. * self.f_phase """ # Derive a second period. # Whitening a light curve. residual_mag = mag - fitted_y # Lomb-Scargle again to find the second period. omega_top, power_top = search_frequencies(date, residual_mag, err, #LS_kwargs={'generalized':True, 'subtract_mean':True}, n_eval=5000, n_retry=3, n_save=50) self.period2 = 2*np.pi/omega_top[np.where(power_top==np.max(power_top))][0] self.f2 = 1. / self.period2 self.f2_SNR = power_top[np.where(power_top==np.max(power_top))][0] \ * (len(self.date) - 1) / 2. # Fit Fourier Series again. p0 = [1.] * order * 2 date_period = (date % self.period) / self.period p2, success = leastsq(self.residuals, p0, args=(date_period, residual_mag, order)) fitted_y = self.FourierSeries(p2, date_period, order) #plt.plot(date%self.period2, residual_mag, 'b+') #plt.show() # Derive Fourier features for the first second. self.f2_amp = 2. * np.sqrt(p2[1]**2 + p2[2]**2) self.f2_R21 = np.sqrt(p2[3]**2 + p2[4]**2) / self.f2_amp self.f2_R31 = np.sqrt(p2[5]**2 + p2[6]**2) / self.f2_amp self.f2_R41 = np.sqrt(p2[7]**2 + p2[8]**2) / self.f2_amp self.f2_R51 = np.sqrt(p2[9]**2 + p2[10]**2) / self.f2_amp self.f2_phase = np.arctan(-p2[1] / p2[2]) self.f2_phi21 = np.arctan(-p2[3] / p2[4]) - 2. * self.f2_phase self.f2_phi31 = np.arctan(-p2[5] / p2[6]) - 3. * self.f2_phase self.f2_phi41 = np.arctan(-p2[7] / p2[8]) - 4. * self.f2_phase self.f2_phi51 = np.arctan(-p2[9] / p2[10]) - 5. * self.f2_phase # Calculate features using the first and second periods. self.f12_ratio = self.f2 / self.f1 self.f12_remain = self.f1 % self.f2 \ if self.f1 > self.f2 else self.f2 % self.f1 self.f12_amp = self.f2_amp / self.f1_amp self.f12_phase = self.f2_phase - self.f1_phase """ def get_period_uncertainty(self, fx, fy, jmax, fx_width=100): """ Get uncertainty of a period. The uncertainty is defined as the half width of the frequencies around the peak, that becomes lower than average + standard deviation of the power spectrum. Since we may not have fine resolution around the peak, we do not assume it is gaussian. So, no scaling factor of 2.355 (= 2 * sqrt(2 * ln2)) is applied. Parameters ---------- fx : array_like An array of frequencies. fy : array_like An array of amplitudes. jmax : int An index at the peak frequency. fx_width : int, optional Width of power spectrum to calculate uncertainty. Returns ------- p_uncertain : float Period uncertainty. """ # Get subset start_index = jmax - fx_width end_index = jmax + fx_width if start_index < 0: start_index = 0 if end_index > len(fx) - 1: end_index = len(fx) - 1 fx_subset = fx[start_index:end_index] fy_subset = fy[start_index:end_index] fy_mean = np.median(fy_subset) fy_std = np.std(fy_subset) # Find peak max_index = np.argmax(fy_subset) # Find list whose powers become lower than average + std. index = np.where(fy_subset <= fy_mean + fy_std)[0] # Find the edge at left and right. This is the full width. left_index = index[(index < max_index)] if len(left_index) == 0: left_index = 0 else: left_index = left_index[-1] right_index = index[(index > max_index)] if len(right_index) == 0: right_index = len(fy_subset) - 1 else: right_index = right_index[0] # We assume the half of the full width is the period uncertainty. half_width = (1. / fx_subset[left_index] - 1. / fx_subset[right_index]) / 2. period_uncertainty = half_width return period_uncertainty def residuals(self, pars, x, y, order): """ Residual of Fourier Series. Parameters ---------- pars : array_like Fourier series parameters. x : array_like An array of date. y : array_like An array of true values to fit. order : int An order of Fourier Series. """ return y - self.fourier_series(pars, x, order) def fourier_series(self, pars, x, order): """ Function to fit Fourier Series. Parameters ---------- x : array_like An array of date divided by period. It doesn't need to be sorted. pars : array_like Fourier series parameters. order : int An order of Fourier series. """ sum = pars[0] for i in range(order): sum += pars[i * 2 + 1] * np.sin(2 * np.pi * (i + 1) * x) \ + pars[i * 2 + 2] * np.cos(2 * np.pi * (i + 1) * x) return sum def get_stetson_k(self, mag, avg, err): """ Return Stetson K feature. Parameters ---------- mag : array_like An array of magnitude. avg : float An average value of magnitudes. err : array_like An array of magnitude errors. Returns ------- stetson_k : float Stetson K value. """ residual = (mag - avg) / err stetson_k = np.sum(np.fabs(residual)) \ / np.sqrt(np.sum(residual * residual)) / np.sqrt(len(mag)) return stetson_k def half_mag_amplitude_ratio2(self, mag, avg): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) fainter_mag = mag[index] lower_sum = np.sum((fainter_mag - avg) ** 2) / len(fainter_mag) # For higher (brighter) magnitude than average. index = np.where(mag <= avg) brighter_mag = mag[index] higher_sum = np.sum((avg - brighter_mag) ** 2) / len(brighter_mag) # Return ratio. return np.sqrt(lower_sum / higher_sum) def get_eta(self, mag, std): """ Return Eta feature. Parameters ---------- mag : array_like An array of magnitudes. std : array_like A standard deviation of magnitudes. Returns ------- eta : float The value of Eta index. """ diff = mag[1:] - mag[:len(mag) - 1] eta = np.sum(diff * diff) / (len(mag) - 1.) / std / std return eta def slope_percentile(self, date, mag): """ Return 10% and 90% percentile of slope. Parameters ---------- date : array_like An array of phase-folded date. Sorted. mag : array_like An array of phase-folded magnitudes. Sorted by date. Returns ------- per_10 : float 10% percentile values of slope. per_90 : float 90% percentile values of slope. """ date_diff = date[1:] - date[:len(date) - 1] mag_diff = mag[1:] - mag[:len(mag) - 1] # Remove zero mag_diff. index = np.where(mag_diff != 0.) date_diff = date_diff[index] mag_diff = mag_diff[index] # Derive slope. slope = date_diff / mag_diff percentile_10 = np.percentile(slope, 10.) percentile_90 = np.percentile(slope, 90.) return percentile_10, percentile_90 def get_cusum(self, mag): """ Return max - min of cumulative sum. Parameters ---------- mag : array_like An array of magnitudes. Returns ------- mm_cusum : float Max - min of cumulative sum. """ c = np.cumsum(mag - self.weighted_mean) / len(mag) / self.weighted_std return np.max(c) - np.min(c) def get_features2(self): """ Return all features with its names. Returns ------- names : list Feature names. values : list Feature values """ feature_names = [] feature_values = [] # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): # Omit input variables such as date, mag, err, etc. if not (name == 'date' or name == 'mag' or name == 'err' or name == 'n_threads' or name == 'min_period'): # Filter some other unnecessary features. if not (name == 'f' or name == 'f_phase' or name == 'period_log10FAP' or name == 'weight' or name == 'weighted_sum' or name == 'median' or name == 'mean' or name == 'std'): feature_names.append(name) # Sort by the names. # Sorting should be done to keep maintaining the same order of features. feature_names.sort() # Get feature values. for name in feature_names: feature_values.append(all_vars[name]) return feature_names, feature_values def get_features(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ ''' features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features ''' return self.get_features_all() def get_features_all(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list_all: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features
dwkim78/upsilon
upsilon/extract_features/extract_features.py
ExtractFeatures.half_mag_amplitude_ratio2
python
def half_mag_amplitude_ratio2(self, mag, avg): # For lower (fainter) magnitude than average. index = np.where(mag > avg) fainter_mag = mag[index] lower_sum = np.sum((fainter_mag - avg) ** 2) / len(fainter_mag) # For higher (brighter) magnitude than average. index = np.where(mag <= avg) brighter_mag = mag[index] higher_sum = np.sum((avg - brighter_mag) ** 2) / len(brighter_mag) # Return ratio. return np.sqrt(lower_sum / higher_sum)
Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/extract_features.py#L451-L486
null
class ExtractFeatures: """ Extract variability features of a light curve. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. err : array_like, optional An array of magnitude error. If None, std(mag) will be used. n_threads : int, optional The number of cores to use to derive periods. min_period : float, optional The minimum period to calculate. """ def __init__(self, date, mag, err=None, n_threads=4, min_period=0.03): # Set basic values. if not isinstance(date, np.ndarray): date = np.array(date) if not isinstance(mag, np.ndarray): mag = np.array(mag) self.date = date self.mag = mag if err is not None: if not isinstance(err, np.ndarray): err = np.array(err) self.err = err else: self.err = np.ones(len(self.mag)) * np.std(self.mag) # Check length. if (len(self.date) != len(self.mag)) or \ (len(self.date) != len(self.err)) or \ (len(self.mag) != len(self.err)): raise RuntimeError('The length of date, mag, and err must be same.') # if the number of data points is too small. min_n_data = 80 if len(self.date) < min_n_data: warnings.warn('The number of data points are less than %d.' % min_n_data) n_threads = int(n_threads) if n_threads > multiprocessing.cpu_count(): self.n_threads = multiprocessing.cpu_count() else: if n_threads <= 0: self.n_threads = 1 else: self.n_threads = n_threads min_period = float(min_period) if min_period <= 0: self.min_period = 0.03 else: self.min_period = min_period def run(self): """Run feature extraction modules.""" # shallow_run must be executed prior to deep_run # since shallow_run calculates several values needed for deep_run. self.shallow_run() self.deep_run() def shallow_run(self): """Derive not-period-based features.""" # Number of data points self.n_points = len(self.date) # Weight calculation. # All zero values. if not self.err.any(): self.err = np.ones(len(self.mag)) * np.std(self.mag) # Some zero values. elif not self.err.all(): np.putmask(self.err, self.err==0, np.median(self.err)) self.weight = 1. / self.err self.weighted_sum = np.sum(self.weight) # Simple statistics, mean, median and std. self.mean = np.mean(self.mag) self.median = np.median(self.mag) self.std = np.std(self.mag) # Weighted mean and std. self.weighted_mean = np.sum(self.mag * self.weight) / self.weighted_sum self.weighted_std = np.sqrt(np.sum((self.mag - self.weighted_mean) ** 2 \ * self.weight) / self.weighted_sum) # Skewness and kurtosis. self.skewness = ss.skew(self.mag) self.kurtosis = ss.kurtosis(self.mag) # Normalization-test. Shapiro-Wilk test. shapiro = ss.shapiro(self.mag) self.shapiro_w = shapiro[0] # self.shapiro_log10p = np.log10(shapiro[1]) # Percentile features. self.quartile31 = np.percentile(self.mag, 75) \ - np.percentile(self.mag, 25) # Stetson K. self.stetson_k = self.get_stetson_k(self.mag, self.median, self.err) # Ratio between higher and lower amplitude than average. self.hl_amp_ratio = self.half_mag_amplitude_ratio( self.mag, self.median, self.weight) # This second function's value is very similar with the above one. # self.hl_amp_ratio2 = self.half_mag_amplitude_ratio2( # self.mag, self.median) # Cusum self.cusum = self.get_cusum(self.mag) # Eta self.eta = self.get_eta(self.mag, self.weighted_std) def deep_run(self): """Derive period-based features.""" # Lomb-Scargle period finding. self.get_period_LS(self.date, self.mag, self.n_threads, self.min_period) # Features based on a phase-folded light curve # such as Eta, slope-percentile, etc. # Should be called after the getPeriodLS() is called. # Created phased a folded light curve. # We use period * 2 to take eclipsing binaries into account. phase_folded_date = self.date % (self.period * 2.) sorted_index = np.argsort(phase_folded_date) folded_date = phase_folded_date[sorted_index] folded_mag = self.mag[sorted_index] # phase Eta self.phase_eta = self.get_eta(folded_mag, self.weighted_std) # Slope percentile. self.slope_per10, self.slope_per90 = \ self.slope_percentile(folded_date, folded_mag) # phase Cusum self.phase_cusum = self.get_cusum(folded_mag) def get_period_LS(self, date, mag, n_threads, min_period): """ Period finding using the Lomb-Scargle algorithm. Finding two periods. The second period is estimated after whitening the first period. Calculating various other features as well using derived periods. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. n_threads : int The number of threads to use. min_period : float The minimum period to calculate. """ # DO NOT CHANGE THESE PARAMETERS. oversampling = 3. hifac = int((max(date) - min(date)) / len(date) / min_period * 2.) # Minimum hifac if hifac < 100: hifac = 100 # Lomb-Scargle. fx, fy, nout, jmax, prob = pLS.fasper(date, mag, oversampling, hifac, n_threads) self.f = fx[jmax] self.period = 1. / self.f self.period_uncertainty = self.get_period_uncertainty(fx, fy, jmax) self.period_log10FAP = \ np.log10(pLS.getSignificance(fx, fy, nout, oversampling)[jmax]) # self.f_SNR1 = fy[jmax] / np.median(fy) self.period_SNR = (fy[jmax] - np.median(fy)) / np.std(fy) # Fit Fourier Series of order 3. order = 3 # Initial guess of Fourier coefficients. p0 = np.ones(order * 2 + 1) date_period = (date % self.period) / self.period p1, success = leastsq(self.residuals, p0, args=(date_period, mag, order)) # fitted_y = self.FourierSeries(p1, date_period, order) # print p1, self.mean, self.median # plt.plot(date_period, self.mag, 'b+') # plt.show() # Derive Fourier features for the first period. # Petersen, J. O., 1986, A&A self.amplitude = np.sqrt(p1[1] ** 2 + p1[2] ** 2) self.r21 = np.sqrt(p1[3] ** 2 + p1[4] ** 2) / self.amplitude self.r31 = np.sqrt(p1[5] ** 2 + p1[6] ** 2) / self.amplitude self.f_phase = np.arctan(-p1[1] / p1[2]) self.phi21 = np.arctan(-p1[3] / p1[4]) - 2. * self.f_phase self.phi31 = np.arctan(-p1[5] / p1[6]) - 3. * self.f_phase """ # Derive a second period. # Whitening a light curve. residual_mag = mag - fitted_y # Lomb-Scargle again to find the second period. omega_top, power_top = search_frequencies(date, residual_mag, err, #LS_kwargs={'generalized':True, 'subtract_mean':True}, n_eval=5000, n_retry=3, n_save=50) self.period2 = 2*np.pi/omega_top[np.where(power_top==np.max(power_top))][0] self.f2 = 1. / self.period2 self.f2_SNR = power_top[np.where(power_top==np.max(power_top))][0] \ * (len(self.date) - 1) / 2. # Fit Fourier Series again. p0 = [1.] * order * 2 date_period = (date % self.period) / self.period p2, success = leastsq(self.residuals, p0, args=(date_period, residual_mag, order)) fitted_y = self.FourierSeries(p2, date_period, order) #plt.plot(date%self.period2, residual_mag, 'b+') #plt.show() # Derive Fourier features for the first second. self.f2_amp = 2. * np.sqrt(p2[1]**2 + p2[2]**2) self.f2_R21 = np.sqrt(p2[3]**2 + p2[4]**2) / self.f2_amp self.f2_R31 = np.sqrt(p2[5]**2 + p2[6]**2) / self.f2_amp self.f2_R41 = np.sqrt(p2[7]**2 + p2[8]**2) / self.f2_amp self.f2_R51 = np.sqrt(p2[9]**2 + p2[10]**2) / self.f2_amp self.f2_phase = np.arctan(-p2[1] / p2[2]) self.f2_phi21 = np.arctan(-p2[3] / p2[4]) - 2. * self.f2_phase self.f2_phi31 = np.arctan(-p2[5] / p2[6]) - 3. * self.f2_phase self.f2_phi41 = np.arctan(-p2[7] / p2[8]) - 4. * self.f2_phase self.f2_phi51 = np.arctan(-p2[9] / p2[10]) - 5. * self.f2_phase # Calculate features using the first and second periods. self.f12_ratio = self.f2 / self.f1 self.f12_remain = self.f1 % self.f2 \ if self.f1 > self.f2 else self.f2 % self.f1 self.f12_amp = self.f2_amp / self.f1_amp self.f12_phase = self.f2_phase - self.f1_phase """ def get_period_uncertainty(self, fx, fy, jmax, fx_width=100): """ Get uncertainty of a period. The uncertainty is defined as the half width of the frequencies around the peak, that becomes lower than average + standard deviation of the power spectrum. Since we may not have fine resolution around the peak, we do not assume it is gaussian. So, no scaling factor of 2.355 (= 2 * sqrt(2 * ln2)) is applied. Parameters ---------- fx : array_like An array of frequencies. fy : array_like An array of amplitudes. jmax : int An index at the peak frequency. fx_width : int, optional Width of power spectrum to calculate uncertainty. Returns ------- p_uncertain : float Period uncertainty. """ # Get subset start_index = jmax - fx_width end_index = jmax + fx_width if start_index < 0: start_index = 0 if end_index > len(fx) - 1: end_index = len(fx) - 1 fx_subset = fx[start_index:end_index] fy_subset = fy[start_index:end_index] fy_mean = np.median(fy_subset) fy_std = np.std(fy_subset) # Find peak max_index = np.argmax(fy_subset) # Find list whose powers become lower than average + std. index = np.where(fy_subset <= fy_mean + fy_std)[0] # Find the edge at left and right. This is the full width. left_index = index[(index < max_index)] if len(left_index) == 0: left_index = 0 else: left_index = left_index[-1] right_index = index[(index > max_index)] if len(right_index) == 0: right_index = len(fy_subset) - 1 else: right_index = right_index[0] # We assume the half of the full width is the period uncertainty. half_width = (1. / fx_subset[left_index] - 1. / fx_subset[right_index]) / 2. period_uncertainty = half_width return period_uncertainty def residuals(self, pars, x, y, order): """ Residual of Fourier Series. Parameters ---------- pars : array_like Fourier series parameters. x : array_like An array of date. y : array_like An array of true values to fit. order : int An order of Fourier Series. """ return y - self.fourier_series(pars, x, order) def fourier_series(self, pars, x, order): """ Function to fit Fourier Series. Parameters ---------- x : array_like An array of date divided by period. It doesn't need to be sorted. pars : array_like Fourier series parameters. order : int An order of Fourier series. """ sum = pars[0] for i in range(order): sum += pars[i * 2 + 1] * np.sin(2 * np.pi * (i + 1) * x) \ + pars[i * 2 + 2] * np.cos(2 * np.pi * (i + 1) * x) return sum def get_stetson_k(self, mag, avg, err): """ Return Stetson K feature. Parameters ---------- mag : array_like An array of magnitude. avg : float An average value of magnitudes. err : array_like An array of magnitude errors. Returns ------- stetson_k : float Stetson K value. """ residual = (mag - avg) / err stetson_k = np.sum(np.fabs(residual)) \ / np.sqrt(np.sum(residual * residual)) / np.sqrt(len(mag)) return stetson_k def half_mag_amplitude_ratio(self, mag, avg, weight): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. weight : array_like An array of weight. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) lower_weight = weight[index] lower_weight_sum = np.sum(lower_weight) lower_mag = mag[index] lower_weighted_std = np.sum((lower_mag - avg) ** 2 * lower_weight) / \ lower_weight_sum # For higher (brighter) magnitude than average. index = np.where(mag <= avg) higher_weight = weight[index] higher_weight_sum = np.sum(higher_weight) higher_mag = mag[index] higher_weighted_std = np.sum((higher_mag - avg) ** 2 * higher_weight) / \ higher_weight_sum # Return ratio. return np.sqrt(lower_weighted_std / higher_weighted_std) def get_eta(self, mag, std): """ Return Eta feature. Parameters ---------- mag : array_like An array of magnitudes. std : array_like A standard deviation of magnitudes. Returns ------- eta : float The value of Eta index. """ diff = mag[1:] - mag[:len(mag) - 1] eta = np.sum(diff * diff) / (len(mag) - 1.) / std / std return eta def slope_percentile(self, date, mag): """ Return 10% and 90% percentile of slope. Parameters ---------- date : array_like An array of phase-folded date. Sorted. mag : array_like An array of phase-folded magnitudes. Sorted by date. Returns ------- per_10 : float 10% percentile values of slope. per_90 : float 90% percentile values of slope. """ date_diff = date[1:] - date[:len(date) - 1] mag_diff = mag[1:] - mag[:len(mag) - 1] # Remove zero mag_diff. index = np.where(mag_diff != 0.) date_diff = date_diff[index] mag_diff = mag_diff[index] # Derive slope. slope = date_diff / mag_diff percentile_10 = np.percentile(slope, 10.) percentile_90 = np.percentile(slope, 90.) return percentile_10, percentile_90 def get_cusum(self, mag): """ Return max - min of cumulative sum. Parameters ---------- mag : array_like An array of magnitudes. Returns ------- mm_cusum : float Max - min of cumulative sum. """ c = np.cumsum(mag - self.weighted_mean) / len(mag) / self.weighted_std return np.max(c) - np.min(c) def get_features2(self): """ Return all features with its names. Returns ------- names : list Feature names. values : list Feature values """ feature_names = [] feature_values = [] # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): # Omit input variables such as date, mag, err, etc. if not (name == 'date' or name == 'mag' or name == 'err' or name == 'n_threads' or name == 'min_period'): # Filter some other unnecessary features. if not (name == 'f' or name == 'f_phase' or name == 'period_log10FAP' or name == 'weight' or name == 'weighted_sum' or name == 'median' or name == 'mean' or name == 'std'): feature_names.append(name) # Sort by the names. # Sorting should be done to keep maintaining the same order of features. feature_names.sort() # Get feature values. for name in feature_names: feature_values.append(all_vars[name]) return feature_names, feature_values def get_features(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ ''' features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features ''' return self.get_features_all() def get_features_all(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list_all: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features
dwkim78/upsilon
upsilon/extract_features/extract_features.py
ExtractFeatures.get_eta
python
def get_eta(self, mag, std): diff = mag[1:] - mag[:len(mag) - 1] eta = np.sum(diff * diff) / (len(mag) - 1.) / std / std return eta
Return Eta feature. Parameters ---------- mag : array_like An array of magnitudes. std : array_like A standard deviation of magnitudes. Returns ------- eta : float The value of Eta index.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/extract_features.py#L488-L508
null
class ExtractFeatures: """ Extract variability features of a light curve. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. err : array_like, optional An array of magnitude error. If None, std(mag) will be used. n_threads : int, optional The number of cores to use to derive periods. min_period : float, optional The minimum period to calculate. """ def __init__(self, date, mag, err=None, n_threads=4, min_period=0.03): # Set basic values. if not isinstance(date, np.ndarray): date = np.array(date) if not isinstance(mag, np.ndarray): mag = np.array(mag) self.date = date self.mag = mag if err is not None: if not isinstance(err, np.ndarray): err = np.array(err) self.err = err else: self.err = np.ones(len(self.mag)) * np.std(self.mag) # Check length. if (len(self.date) != len(self.mag)) or \ (len(self.date) != len(self.err)) or \ (len(self.mag) != len(self.err)): raise RuntimeError('The length of date, mag, and err must be same.') # if the number of data points is too small. min_n_data = 80 if len(self.date) < min_n_data: warnings.warn('The number of data points are less than %d.' % min_n_data) n_threads = int(n_threads) if n_threads > multiprocessing.cpu_count(): self.n_threads = multiprocessing.cpu_count() else: if n_threads <= 0: self.n_threads = 1 else: self.n_threads = n_threads min_period = float(min_period) if min_period <= 0: self.min_period = 0.03 else: self.min_period = min_period def run(self): """Run feature extraction modules.""" # shallow_run must be executed prior to deep_run # since shallow_run calculates several values needed for deep_run. self.shallow_run() self.deep_run() def shallow_run(self): """Derive not-period-based features.""" # Number of data points self.n_points = len(self.date) # Weight calculation. # All zero values. if not self.err.any(): self.err = np.ones(len(self.mag)) * np.std(self.mag) # Some zero values. elif not self.err.all(): np.putmask(self.err, self.err==0, np.median(self.err)) self.weight = 1. / self.err self.weighted_sum = np.sum(self.weight) # Simple statistics, mean, median and std. self.mean = np.mean(self.mag) self.median = np.median(self.mag) self.std = np.std(self.mag) # Weighted mean and std. self.weighted_mean = np.sum(self.mag * self.weight) / self.weighted_sum self.weighted_std = np.sqrt(np.sum((self.mag - self.weighted_mean) ** 2 \ * self.weight) / self.weighted_sum) # Skewness and kurtosis. self.skewness = ss.skew(self.mag) self.kurtosis = ss.kurtosis(self.mag) # Normalization-test. Shapiro-Wilk test. shapiro = ss.shapiro(self.mag) self.shapiro_w = shapiro[0] # self.shapiro_log10p = np.log10(shapiro[1]) # Percentile features. self.quartile31 = np.percentile(self.mag, 75) \ - np.percentile(self.mag, 25) # Stetson K. self.stetson_k = self.get_stetson_k(self.mag, self.median, self.err) # Ratio between higher and lower amplitude than average. self.hl_amp_ratio = self.half_mag_amplitude_ratio( self.mag, self.median, self.weight) # This second function's value is very similar with the above one. # self.hl_amp_ratio2 = self.half_mag_amplitude_ratio2( # self.mag, self.median) # Cusum self.cusum = self.get_cusum(self.mag) # Eta self.eta = self.get_eta(self.mag, self.weighted_std) def deep_run(self): """Derive period-based features.""" # Lomb-Scargle period finding. self.get_period_LS(self.date, self.mag, self.n_threads, self.min_period) # Features based on a phase-folded light curve # such as Eta, slope-percentile, etc. # Should be called after the getPeriodLS() is called. # Created phased a folded light curve. # We use period * 2 to take eclipsing binaries into account. phase_folded_date = self.date % (self.period * 2.) sorted_index = np.argsort(phase_folded_date) folded_date = phase_folded_date[sorted_index] folded_mag = self.mag[sorted_index] # phase Eta self.phase_eta = self.get_eta(folded_mag, self.weighted_std) # Slope percentile. self.slope_per10, self.slope_per90 = \ self.slope_percentile(folded_date, folded_mag) # phase Cusum self.phase_cusum = self.get_cusum(folded_mag) def get_period_LS(self, date, mag, n_threads, min_period): """ Period finding using the Lomb-Scargle algorithm. Finding two periods. The second period is estimated after whitening the first period. Calculating various other features as well using derived periods. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. n_threads : int The number of threads to use. min_period : float The minimum period to calculate. """ # DO NOT CHANGE THESE PARAMETERS. oversampling = 3. hifac = int((max(date) - min(date)) / len(date) / min_period * 2.) # Minimum hifac if hifac < 100: hifac = 100 # Lomb-Scargle. fx, fy, nout, jmax, prob = pLS.fasper(date, mag, oversampling, hifac, n_threads) self.f = fx[jmax] self.period = 1. / self.f self.period_uncertainty = self.get_period_uncertainty(fx, fy, jmax) self.period_log10FAP = \ np.log10(pLS.getSignificance(fx, fy, nout, oversampling)[jmax]) # self.f_SNR1 = fy[jmax] / np.median(fy) self.period_SNR = (fy[jmax] - np.median(fy)) / np.std(fy) # Fit Fourier Series of order 3. order = 3 # Initial guess of Fourier coefficients. p0 = np.ones(order * 2 + 1) date_period = (date % self.period) / self.period p1, success = leastsq(self.residuals, p0, args=(date_period, mag, order)) # fitted_y = self.FourierSeries(p1, date_period, order) # print p1, self.mean, self.median # plt.plot(date_period, self.mag, 'b+') # plt.show() # Derive Fourier features for the first period. # Petersen, J. O., 1986, A&A self.amplitude = np.sqrt(p1[1] ** 2 + p1[2] ** 2) self.r21 = np.sqrt(p1[3] ** 2 + p1[4] ** 2) / self.amplitude self.r31 = np.sqrt(p1[5] ** 2 + p1[6] ** 2) / self.amplitude self.f_phase = np.arctan(-p1[1] / p1[2]) self.phi21 = np.arctan(-p1[3] / p1[4]) - 2. * self.f_phase self.phi31 = np.arctan(-p1[5] / p1[6]) - 3. * self.f_phase """ # Derive a second period. # Whitening a light curve. residual_mag = mag - fitted_y # Lomb-Scargle again to find the second period. omega_top, power_top = search_frequencies(date, residual_mag, err, #LS_kwargs={'generalized':True, 'subtract_mean':True}, n_eval=5000, n_retry=3, n_save=50) self.period2 = 2*np.pi/omega_top[np.where(power_top==np.max(power_top))][0] self.f2 = 1. / self.period2 self.f2_SNR = power_top[np.where(power_top==np.max(power_top))][0] \ * (len(self.date) - 1) / 2. # Fit Fourier Series again. p0 = [1.] * order * 2 date_period = (date % self.period) / self.period p2, success = leastsq(self.residuals, p0, args=(date_period, residual_mag, order)) fitted_y = self.FourierSeries(p2, date_period, order) #plt.plot(date%self.period2, residual_mag, 'b+') #plt.show() # Derive Fourier features for the first second. self.f2_amp = 2. * np.sqrt(p2[1]**2 + p2[2]**2) self.f2_R21 = np.sqrt(p2[3]**2 + p2[4]**2) / self.f2_amp self.f2_R31 = np.sqrt(p2[5]**2 + p2[6]**2) / self.f2_amp self.f2_R41 = np.sqrt(p2[7]**2 + p2[8]**2) / self.f2_amp self.f2_R51 = np.sqrt(p2[9]**2 + p2[10]**2) / self.f2_amp self.f2_phase = np.arctan(-p2[1] / p2[2]) self.f2_phi21 = np.arctan(-p2[3] / p2[4]) - 2. * self.f2_phase self.f2_phi31 = np.arctan(-p2[5] / p2[6]) - 3. * self.f2_phase self.f2_phi41 = np.arctan(-p2[7] / p2[8]) - 4. * self.f2_phase self.f2_phi51 = np.arctan(-p2[9] / p2[10]) - 5. * self.f2_phase # Calculate features using the first and second periods. self.f12_ratio = self.f2 / self.f1 self.f12_remain = self.f1 % self.f2 \ if self.f1 > self.f2 else self.f2 % self.f1 self.f12_amp = self.f2_amp / self.f1_amp self.f12_phase = self.f2_phase - self.f1_phase """ def get_period_uncertainty(self, fx, fy, jmax, fx_width=100): """ Get uncertainty of a period. The uncertainty is defined as the half width of the frequencies around the peak, that becomes lower than average + standard deviation of the power spectrum. Since we may not have fine resolution around the peak, we do not assume it is gaussian. So, no scaling factor of 2.355 (= 2 * sqrt(2 * ln2)) is applied. Parameters ---------- fx : array_like An array of frequencies. fy : array_like An array of amplitudes. jmax : int An index at the peak frequency. fx_width : int, optional Width of power spectrum to calculate uncertainty. Returns ------- p_uncertain : float Period uncertainty. """ # Get subset start_index = jmax - fx_width end_index = jmax + fx_width if start_index < 0: start_index = 0 if end_index > len(fx) - 1: end_index = len(fx) - 1 fx_subset = fx[start_index:end_index] fy_subset = fy[start_index:end_index] fy_mean = np.median(fy_subset) fy_std = np.std(fy_subset) # Find peak max_index = np.argmax(fy_subset) # Find list whose powers become lower than average + std. index = np.where(fy_subset <= fy_mean + fy_std)[0] # Find the edge at left and right. This is the full width. left_index = index[(index < max_index)] if len(left_index) == 0: left_index = 0 else: left_index = left_index[-1] right_index = index[(index > max_index)] if len(right_index) == 0: right_index = len(fy_subset) - 1 else: right_index = right_index[0] # We assume the half of the full width is the period uncertainty. half_width = (1. / fx_subset[left_index] - 1. / fx_subset[right_index]) / 2. period_uncertainty = half_width return period_uncertainty def residuals(self, pars, x, y, order): """ Residual of Fourier Series. Parameters ---------- pars : array_like Fourier series parameters. x : array_like An array of date. y : array_like An array of true values to fit. order : int An order of Fourier Series. """ return y - self.fourier_series(pars, x, order) def fourier_series(self, pars, x, order): """ Function to fit Fourier Series. Parameters ---------- x : array_like An array of date divided by period. It doesn't need to be sorted. pars : array_like Fourier series parameters. order : int An order of Fourier series. """ sum = pars[0] for i in range(order): sum += pars[i * 2 + 1] * np.sin(2 * np.pi * (i + 1) * x) \ + pars[i * 2 + 2] * np.cos(2 * np.pi * (i + 1) * x) return sum def get_stetson_k(self, mag, avg, err): """ Return Stetson K feature. Parameters ---------- mag : array_like An array of magnitude. avg : float An average value of magnitudes. err : array_like An array of magnitude errors. Returns ------- stetson_k : float Stetson K value. """ residual = (mag - avg) / err stetson_k = np.sum(np.fabs(residual)) \ / np.sqrt(np.sum(residual * residual)) / np.sqrt(len(mag)) return stetson_k def half_mag_amplitude_ratio(self, mag, avg, weight): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. weight : array_like An array of weight. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) lower_weight = weight[index] lower_weight_sum = np.sum(lower_weight) lower_mag = mag[index] lower_weighted_std = np.sum((lower_mag - avg) ** 2 * lower_weight) / \ lower_weight_sum # For higher (brighter) magnitude than average. index = np.where(mag <= avg) higher_weight = weight[index] higher_weight_sum = np.sum(higher_weight) higher_mag = mag[index] higher_weighted_std = np.sum((higher_mag - avg) ** 2 * higher_weight) / \ higher_weight_sum # Return ratio. return np.sqrt(lower_weighted_std / higher_weighted_std) def half_mag_amplitude_ratio2(self, mag, avg): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) fainter_mag = mag[index] lower_sum = np.sum((fainter_mag - avg) ** 2) / len(fainter_mag) # For higher (brighter) magnitude than average. index = np.where(mag <= avg) brighter_mag = mag[index] higher_sum = np.sum((avg - brighter_mag) ** 2) / len(brighter_mag) # Return ratio. return np.sqrt(lower_sum / higher_sum) def slope_percentile(self, date, mag): """ Return 10% and 90% percentile of slope. Parameters ---------- date : array_like An array of phase-folded date. Sorted. mag : array_like An array of phase-folded magnitudes. Sorted by date. Returns ------- per_10 : float 10% percentile values of slope. per_90 : float 90% percentile values of slope. """ date_diff = date[1:] - date[:len(date) - 1] mag_diff = mag[1:] - mag[:len(mag) - 1] # Remove zero mag_diff. index = np.where(mag_diff != 0.) date_diff = date_diff[index] mag_diff = mag_diff[index] # Derive slope. slope = date_diff / mag_diff percentile_10 = np.percentile(slope, 10.) percentile_90 = np.percentile(slope, 90.) return percentile_10, percentile_90 def get_cusum(self, mag): """ Return max - min of cumulative sum. Parameters ---------- mag : array_like An array of magnitudes. Returns ------- mm_cusum : float Max - min of cumulative sum. """ c = np.cumsum(mag - self.weighted_mean) / len(mag) / self.weighted_std return np.max(c) - np.min(c) def get_features2(self): """ Return all features with its names. Returns ------- names : list Feature names. values : list Feature values """ feature_names = [] feature_values = [] # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): # Omit input variables such as date, mag, err, etc. if not (name == 'date' or name == 'mag' or name == 'err' or name == 'n_threads' or name == 'min_period'): # Filter some other unnecessary features. if not (name == 'f' or name == 'f_phase' or name == 'period_log10FAP' or name == 'weight' or name == 'weighted_sum' or name == 'median' or name == 'mean' or name == 'std'): feature_names.append(name) # Sort by the names. # Sorting should be done to keep maintaining the same order of features. feature_names.sort() # Get feature values. for name in feature_names: feature_values.append(all_vars[name]) return feature_names, feature_values def get_features(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ ''' features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features ''' return self.get_features_all() def get_features_all(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list_all: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features
dwkim78/upsilon
upsilon/extract_features/extract_features.py
ExtractFeatures.slope_percentile
python
def slope_percentile(self, date, mag): date_diff = date[1:] - date[:len(date) - 1] mag_diff = mag[1:] - mag[:len(mag) - 1] # Remove zero mag_diff. index = np.where(mag_diff != 0.) date_diff = date_diff[index] mag_diff = mag_diff[index] # Derive slope. slope = date_diff / mag_diff percentile_10 = np.percentile(slope, 10.) percentile_90 = np.percentile(slope, 90.) return percentile_10, percentile_90
Return 10% and 90% percentile of slope. Parameters ---------- date : array_like An array of phase-folded date. Sorted. mag : array_like An array of phase-folded magnitudes. Sorted by date. Returns ------- per_10 : float 10% percentile values of slope. per_90 : float 90% percentile values of slope.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/extract_features.py#L510-L543
null
class ExtractFeatures: """ Extract variability features of a light curve. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. err : array_like, optional An array of magnitude error. If None, std(mag) will be used. n_threads : int, optional The number of cores to use to derive periods. min_period : float, optional The minimum period to calculate. """ def __init__(self, date, mag, err=None, n_threads=4, min_period=0.03): # Set basic values. if not isinstance(date, np.ndarray): date = np.array(date) if not isinstance(mag, np.ndarray): mag = np.array(mag) self.date = date self.mag = mag if err is not None: if not isinstance(err, np.ndarray): err = np.array(err) self.err = err else: self.err = np.ones(len(self.mag)) * np.std(self.mag) # Check length. if (len(self.date) != len(self.mag)) or \ (len(self.date) != len(self.err)) or \ (len(self.mag) != len(self.err)): raise RuntimeError('The length of date, mag, and err must be same.') # if the number of data points is too small. min_n_data = 80 if len(self.date) < min_n_data: warnings.warn('The number of data points are less than %d.' % min_n_data) n_threads = int(n_threads) if n_threads > multiprocessing.cpu_count(): self.n_threads = multiprocessing.cpu_count() else: if n_threads <= 0: self.n_threads = 1 else: self.n_threads = n_threads min_period = float(min_period) if min_period <= 0: self.min_period = 0.03 else: self.min_period = min_period def run(self): """Run feature extraction modules.""" # shallow_run must be executed prior to deep_run # since shallow_run calculates several values needed for deep_run. self.shallow_run() self.deep_run() def shallow_run(self): """Derive not-period-based features.""" # Number of data points self.n_points = len(self.date) # Weight calculation. # All zero values. if not self.err.any(): self.err = np.ones(len(self.mag)) * np.std(self.mag) # Some zero values. elif not self.err.all(): np.putmask(self.err, self.err==0, np.median(self.err)) self.weight = 1. / self.err self.weighted_sum = np.sum(self.weight) # Simple statistics, mean, median and std. self.mean = np.mean(self.mag) self.median = np.median(self.mag) self.std = np.std(self.mag) # Weighted mean and std. self.weighted_mean = np.sum(self.mag * self.weight) / self.weighted_sum self.weighted_std = np.sqrt(np.sum((self.mag - self.weighted_mean) ** 2 \ * self.weight) / self.weighted_sum) # Skewness and kurtosis. self.skewness = ss.skew(self.mag) self.kurtosis = ss.kurtosis(self.mag) # Normalization-test. Shapiro-Wilk test. shapiro = ss.shapiro(self.mag) self.shapiro_w = shapiro[0] # self.shapiro_log10p = np.log10(shapiro[1]) # Percentile features. self.quartile31 = np.percentile(self.mag, 75) \ - np.percentile(self.mag, 25) # Stetson K. self.stetson_k = self.get_stetson_k(self.mag, self.median, self.err) # Ratio between higher and lower amplitude than average. self.hl_amp_ratio = self.half_mag_amplitude_ratio( self.mag, self.median, self.weight) # This second function's value is very similar with the above one. # self.hl_amp_ratio2 = self.half_mag_amplitude_ratio2( # self.mag, self.median) # Cusum self.cusum = self.get_cusum(self.mag) # Eta self.eta = self.get_eta(self.mag, self.weighted_std) def deep_run(self): """Derive period-based features.""" # Lomb-Scargle period finding. self.get_period_LS(self.date, self.mag, self.n_threads, self.min_period) # Features based on a phase-folded light curve # such as Eta, slope-percentile, etc. # Should be called after the getPeriodLS() is called. # Created phased a folded light curve. # We use period * 2 to take eclipsing binaries into account. phase_folded_date = self.date % (self.period * 2.) sorted_index = np.argsort(phase_folded_date) folded_date = phase_folded_date[sorted_index] folded_mag = self.mag[sorted_index] # phase Eta self.phase_eta = self.get_eta(folded_mag, self.weighted_std) # Slope percentile. self.slope_per10, self.slope_per90 = \ self.slope_percentile(folded_date, folded_mag) # phase Cusum self.phase_cusum = self.get_cusum(folded_mag) def get_period_LS(self, date, mag, n_threads, min_period): """ Period finding using the Lomb-Scargle algorithm. Finding two periods. The second period is estimated after whitening the first period. Calculating various other features as well using derived periods. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. n_threads : int The number of threads to use. min_period : float The minimum period to calculate. """ # DO NOT CHANGE THESE PARAMETERS. oversampling = 3. hifac = int((max(date) - min(date)) / len(date) / min_period * 2.) # Minimum hifac if hifac < 100: hifac = 100 # Lomb-Scargle. fx, fy, nout, jmax, prob = pLS.fasper(date, mag, oversampling, hifac, n_threads) self.f = fx[jmax] self.period = 1. / self.f self.period_uncertainty = self.get_period_uncertainty(fx, fy, jmax) self.period_log10FAP = \ np.log10(pLS.getSignificance(fx, fy, nout, oversampling)[jmax]) # self.f_SNR1 = fy[jmax] / np.median(fy) self.period_SNR = (fy[jmax] - np.median(fy)) / np.std(fy) # Fit Fourier Series of order 3. order = 3 # Initial guess of Fourier coefficients. p0 = np.ones(order * 2 + 1) date_period = (date % self.period) / self.period p1, success = leastsq(self.residuals, p0, args=(date_period, mag, order)) # fitted_y = self.FourierSeries(p1, date_period, order) # print p1, self.mean, self.median # plt.plot(date_period, self.mag, 'b+') # plt.show() # Derive Fourier features for the first period. # Petersen, J. O., 1986, A&A self.amplitude = np.sqrt(p1[1] ** 2 + p1[2] ** 2) self.r21 = np.sqrt(p1[3] ** 2 + p1[4] ** 2) / self.amplitude self.r31 = np.sqrt(p1[5] ** 2 + p1[6] ** 2) / self.amplitude self.f_phase = np.arctan(-p1[1] / p1[2]) self.phi21 = np.arctan(-p1[3] / p1[4]) - 2. * self.f_phase self.phi31 = np.arctan(-p1[5] / p1[6]) - 3. * self.f_phase """ # Derive a second period. # Whitening a light curve. residual_mag = mag - fitted_y # Lomb-Scargle again to find the second period. omega_top, power_top = search_frequencies(date, residual_mag, err, #LS_kwargs={'generalized':True, 'subtract_mean':True}, n_eval=5000, n_retry=3, n_save=50) self.period2 = 2*np.pi/omega_top[np.where(power_top==np.max(power_top))][0] self.f2 = 1. / self.period2 self.f2_SNR = power_top[np.where(power_top==np.max(power_top))][0] \ * (len(self.date) - 1) / 2. # Fit Fourier Series again. p0 = [1.] * order * 2 date_period = (date % self.period) / self.period p2, success = leastsq(self.residuals, p0, args=(date_period, residual_mag, order)) fitted_y = self.FourierSeries(p2, date_period, order) #plt.plot(date%self.period2, residual_mag, 'b+') #plt.show() # Derive Fourier features for the first second. self.f2_amp = 2. * np.sqrt(p2[1]**2 + p2[2]**2) self.f2_R21 = np.sqrt(p2[3]**2 + p2[4]**2) / self.f2_amp self.f2_R31 = np.sqrt(p2[5]**2 + p2[6]**2) / self.f2_amp self.f2_R41 = np.sqrt(p2[7]**2 + p2[8]**2) / self.f2_amp self.f2_R51 = np.sqrt(p2[9]**2 + p2[10]**2) / self.f2_amp self.f2_phase = np.arctan(-p2[1] / p2[2]) self.f2_phi21 = np.arctan(-p2[3] / p2[4]) - 2. * self.f2_phase self.f2_phi31 = np.arctan(-p2[5] / p2[6]) - 3. * self.f2_phase self.f2_phi41 = np.arctan(-p2[7] / p2[8]) - 4. * self.f2_phase self.f2_phi51 = np.arctan(-p2[9] / p2[10]) - 5. * self.f2_phase # Calculate features using the first and second periods. self.f12_ratio = self.f2 / self.f1 self.f12_remain = self.f1 % self.f2 \ if self.f1 > self.f2 else self.f2 % self.f1 self.f12_amp = self.f2_amp / self.f1_amp self.f12_phase = self.f2_phase - self.f1_phase """ def get_period_uncertainty(self, fx, fy, jmax, fx_width=100): """ Get uncertainty of a period. The uncertainty is defined as the half width of the frequencies around the peak, that becomes lower than average + standard deviation of the power spectrum. Since we may not have fine resolution around the peak, we do not assume it is gaussian. So, no scaling factor of 2.355 (= 2 * sqrt(2 * ln2)) is applied. Parameters ---------- fx : array_like An array of frequencies. fy : array_like An array of amplitudes. jmax : int An index at the peak frequency. fx_width : int, optional Width of power spectrum to calculate uncertainty. Returns ------- p_uncertain : float Period uncertainty. """ # Get subset start_index = jmax - fx_width end_index = jmax + fx_width if start_index < 0: start_index = 0 if end_index > len(fx) - 1: end_index = len(fx) - 1 fx_subset = fx[start_index:end_index] fy_subset = fy[start_index:end_index] fy_mean = np.median(fy_subset) fy_std = np.std(fy_subset) # Find peak max_index = np.argmax(fy_subset) # Find list whose powers become lower than average + std. index = np.where(fy_subset <= fy_mean + fy_std)[0] # Find the edge at left and right. This is the full width. left_index = index[(index < max_index)] if len(left_index) == 0: left_index = 0 else: left_index = left_index[-1] right_index = index[(index > max_index)] if len(right_index) == 0: right_index = len(fy_subset) - 1 else: right_index = right_index[0] # We assume the half of the full width is the period uncertainty. half_width = (1. / fx_subset[left_index] - 1. / fx_subset[right_index]) / 2. period_uncertainty = half_width return period_uncertainty def residuals(self, pars, x, y, order): """ Residual of Fourier Series. Parameters ---------- pars : array_like Fourier series parameters. x : array_like An array of date. y : array_like An array of true values to fit. order : int An order of Fourier Series. """ return y - self.fourier_series(pars, x, order) def fourier_series(self, pars, x, order): """ Function to fit Fourier Series. Parameters ---------- x : array_like An array of date divided by period. It doesn't need to be sorted. pars : array_like Fourier series parameters. order : int An order of Fourier series. """ sum = pars[0] for i in range(order): sum += pars[i * 2 + 1] * np.sin(2 * np.pi * (i + 1) * x) \ + pars[i * 2 + 2] * np.cos(2 * np.pi * (i + 1) * x) return sum def get_stetson_k(self, mag, avg, err): """ Return Stetson K feature. Parameters ---------- mag : array_like An array of magnitude. avg : float An average value of magnitudes. err : array_like An array of magnitude errors. Returns ------- stetson_k : float Stetson K value. """ residual = (mag - avg) / err stetson_k = np.sum(np.fabs(residual)) \ / np.sqrt(np.sum(residual * residual)) / np.sqrt(len(mag)) return stetson_k def half_mag_amplitude_ratio(self, mag, avg, weight): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. weight : array_like An array of weight. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) lower_weight = weight[index] lower_weight_sum = np.sum(lower_weight) lower_mag = mag[index] lower_weighted_std = np.sum((lower_mag - avg) ** 2 * lower_weight) / \ lower_weight_sum # For higher (brighter) magnitude than average. index = np.where(mag <= avg) higher_weight = weight[index] higher_weight_sum = np.sum(higher_weight) higher_mag = mag[index] higher_weighted_std = np.sum((higher_mag - avg) ** 2 * higher_weight) / \ higher_weight_sum # Return ratio. return np.sqrt(lower_weighted_std / higher_weighted_std) def half_mag_amplitude_ratio2(self, mag, avg): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) fainter_mag = mag[index] lower_sum = np.sum((fainter_mag - avg) ** 2) / len(fainter_mag) # For higher (brighter) magnitude than average. index = np.where(mag <= avg) brighter_mag = mag[index] higher_sum = np.sum((avg - brighter_mag) ** 2) / len(brighter_mag) # Return ratio. return np.sqrt(lower_sum / higher_sum) def get_eta(self, mag, std): """ Return Eta feature. Parameters ---------- mag : array_like An array of magnitudes. std : array_like A standard deviation of magnitudes. Returns ------- eta : float The value of Eta index. """ diff = mag[1:] - mag[:len(mag) - 1] eta = np.sum(diff * diff) / (len(mag) - 1.) / std / std return eta def get_cusum(self, mag): """ Return max - min of cumulative sum. Parameters ---------- mag : array_like An array of magnitudes. Returns ------- mm_cusum : float Max - min of cumulative sum. """ c = np.cumsum(mag - self.weighted_mean) / len(mag) / self.weighted_std return np.max(c) - np.min(c) def get_features2(self): """ Return all features with its names. Returns ------- names : list Feature names. values : list Feature values """ feature_names = [] feature_values = [] # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): # Omit input variables such as date, mag, err, etc. if not (name == 'date' or name == 'mag' or name == 'err' or name == 'n_threads' or name == 'min_period'): # Filter some other unnecessary features. if not (name == 'f' or name == 'f_phase' or name == 'period_log10FAP' or name == 'weight' or name == 'weighted_sum' or name == 'median' or name == 'mean' or name == 'std'): feature_names.append(name) # Sort by the names. # Sorting should be done to keep maintaining the same order of features. feature_names.sort() # Get feature values. for name in feature_names: feature_values.append(all_vars[name]) return feature_names, feature_values def get_features(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ ''' features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features ''' return self.get_features_all() def get_features_all(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list_all: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features
dwkim78/upsilon
upsilon/extract_features/extract_features.py
ExtractFeatures.get_cusum
python
def get_cusum(self, mag): c = np.cumsum(mag - self.weighted_mean) / len(mag) / self.weighted_std return np.max(c) - np.min(c)
Return max - min of cumulative sum. Parameters ---------- mag : array_like An array of magnitudes. Returns ------- mm_cusum : float Max - min of cumulative sum.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/extract_features.py#L545-L562
null
class ExtractFeatures: """ Extract variability features of a light curve. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. err : array_like, optional An array of magnitude error. If None, std(mag) will be used. n_threads : int, optional The number of cores to use to derive periods. min_period : float, optional The minimum period to calculate. """ def __init__(self, date, mag, err=None, n_threads=4, min_period=0.03): # Set basic values. if not isinstance(date, np.ndarray): date = np.array(date) if not isinstance(mag, np.ndarray): mag = np.array(mag) self.date = date self.mag = mag if err is not None: if not isinstance(err, np.ndarray): err = np.array(err) self.err = err else: self.err = np.ones(len(self.mag)) * np.std(self.mag) # Check length. if (len(self.date) != len(self.mag)) or \ (len(self.date) != len(self.err)) or \ (len(self.mag) != len(self.err)): raise RuntimeError('The length of date, mag, and err must be same.') # if the number of data points is too small. min_n_data = 80 if len(self.date) < min_n_data: warnings.warn('The number of data points are less than %d.' % min_n_data) n_threads = int(n_threads) if n_threads > multiprocessing.cpu_count(): self.n_threads = multiprocessing.cpu_count() else: if n_threads <= 0: self.n_threads = 1 else: self.n_threads = n_threads min_period = float(min_period) if min_period <= 0: self.min_period = 0.03 else: self.min_period = min_period def run(self): """Run feature extraction modules.""" # shallow_run must be executed prior to deep_run # since shallow_run calculates several values needed for deep_run. self.shallow_run() self.deep_run() def shallow_run(self): """Derive not-period-based features.""" # Number of data points self.n_points = len(self.date) # Weight calculation. # All zero values. if not self.err.any(): self.err = np.ones(len(self.mag)) * np.std(self.mag) # Some zero values. elif not self.err.all(): np.putmask(self.err, self.err==0, np.median(self.err)) self.weight = 1. / self.err self.weighted_sum = np.sum(self.weight) # Simple statistics, mean, median and std. self.mean = np.mean(self.mag) self.median = np.median(self.mag) self.std = np.std(self.mag) # Weighted mean and std. self.weighted_mean = np.sum(self.mag * self.weight) / self.weighted_sum self.weighted_std = np.sqrt(np.sum((self.mag - self.weighted_mean) ** 2 \ * self.weight) / self.weighted_sum) # Skewness and kurtosis. self.skewness = ss.skew(self.mag) self.kurtosis = ss.kurtosis(self.mag) # Normalization-test. Shapiro-Wilk test. shapiro = ss.shapiro(self.mag) self.shapiro_w = shapiro[0] # self.shapiro_log10p = np.log10(shapiro[1]) # Percentile features. self.quartile31 = np.percentile(self.mag, 75) \ - np.percentile(self.mag, 25) # Stetson K. self.stetson_k = self.get_stetson_k(self.mag, self.median, self.err) # Ratio between higher and lower amplitude than average. self.hl_amp_ratio = self.half_mag_amplitude_ratio( self.mag, self.median, self.weight) # This second function's value is very similar with the above one. # self.hl_amp_ratio2 = self.half_mag_amplitude_ratio2( # self.mag, self.median) # Cusum self.cusum = self.get_cusum(self.mag) # Eta self.eta = self.get_eta(self.mag, self.weighted_std) def deep_run(self): """Derive period-based features.""" # Lomb-Scargle period finding. self.get_period_LS(self.date, self.mag, self.n_threads, self.min_period) # Features based on a phase-folded light curve # such as Eta, slope-percentile, etc. # Should be called after the getPeriodLS() is called. # Created phased a folded light curve. # We use period * 2 to take eclipsing binaries into account. phase_folded_date = self.date % (self.period * 2.) sorted_index = np.argsort(phase_folded_date) folded_date = phase_folded_date[sorted_index] folded_mag = self.mag[sorted_index] # phase Eta self.phase_eta = self.get_eta(folded_mag, self.weighted_std) # Slope percentile. self.slope_per10, self.slope_per90 = \ self.slope_percentile(folded_date, folded_mag) # phase Cusum self.phase_cusum = self.get_cusum(folded_mag) def get_period_LS(self, date, mag, n_threads, min_period): """ Period finding using the Lomb-Scargle algorithm. Finding two periods. The second period is estimated after whitening the first period. Calculating various other features as well using derived periods. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. n_threads : int The number of threads to use. min_period : float The minimum period to calculate. """ # DO NOT CHANGE THESE PARAMETERS. oversampling = 3. hifac = int((max(date) - min(date)) / len(date) / min_period * 2.) # Minimum hifac if hifac < 100: hifac = 100 # Lomb-Scargle. fx, fy, nout, jmax, prob = pLS.fasper(date, mag, oversampling, hifac, n_threads) self.f = fx[jmax] self.period = 1. / self.f self.period_uncertainty = self.get_period_uncertainty(fx, fy, jmax) self.period_log10FAP = \ np.log10(pLS.getSignificance(fx, fy, nout, oversampling)[jmax]) # self.f_SNR1 = fy[jmax] / np.median(fy) self.period_SNR = (fy[jmax] - np.median(fy)) / np.std(fy) # Fit Fourier Series of order 3. order = 3 # Initial guess of Fourier coefficients. p0 = np.ones(order * 2 + 1) date_period = (date % self.period) / self.period p1, success = leastsq(self.residuals, p0, args=(date_period, mag, order)) # fitted_y = self.FourierSeries(p1, date_period, order) # print p1, self.mean, self.median # plt.plot(date_period, self.mag, 'b+') # plt.show() # Derive Fourier features for the first period. # Petersen, J. O., 1986, A&A self.amplitude = np.sqrt(p1[1] ** 2 + p1[2] ** 2) self.r21 = np.sqrt(p1[3] ** 2 + p1[4] ** 2) / self.amplitude self.r31 = np.sqrt(p1[5] ** 2 + p1[6] ** 2) / self.amplitude self.f_phase = np.arctan(-p1[1] / p1[2]) self.phi21 = np.arctan(-p1[3] / p1[4]) - 2. * self.f_phase self.phi31 = np.arctan(-p1[5] / p1[6]) - 3. * self.f_phase """ # Derive a second period. # Whitening a light curve. residual_mag = mag - fitted_y # Lomb-Scargle again to find the second period. omega_top, power_top = search_frequencies(date, residual_mag, err, #LS_kwargs={'generalized':True, 'subtract_mean':True}, n_eval=5000, n_retry=3, n_save=50) self.period2 = 2*np.pi/omega_top[np.where(power_top==np.max(power_top))][0] self.f2 = 1. / self.period2 self.f2_SNR = power_top[np.where(power_top==np.max(power_top))][0] \ * (len(self.date) - 1) / 2. # Fit Fourier Series again. p0 = [1.] * order * 2 date_period = (date % self.period) / self.period p2, success = leastsq(self.residuals, p0, args=(date_period, residual_mag, order)) fitted_y = self.FourierSeries(p2, date_period, order) #plt.plot(date%self.period2, residual_mag, 'b+') #plt.show() # Derive Fourier features for the first second. self.f2_amp = 2. * np.sqrt(p2[1]**2 + p2[2]**2) self.f2_R21 = np.sqrt(p2[3]**2 + p2[4]**2) / self.f2_amp self.f2_R31 = np.sqrt(p2[5]**2 + p2[6]**2) / self.f2_amp self.f2_R41 = np.sqrt(p2[7]**2 + p2[8]**2) / self.f2_amp self.f2_R51 = np.sqrt(p2[9]**2 + p2[10]**2) / self.f2_amp self.f2_phase = np.arctan(-p2[1] / p2[2]) self.f2_phi21 = np.arctan(-p2[3] / p2[4]) - 2. * self.f2_phase self.f2_phi31 = np.arctan(-p2[5] / p2[6]) - 3. * self.f2_phase self.f2_phi41 = np.arctan(-p2[7] / p2[8]) - 4. * self.f2_phase self.f2_phi51 = np.arctan(-p2[9] / p2[10]) - 5. * self.f2_phase # Calculate features using the first and second periods. self.f12_ratio = self.f2 / self.f1 self.f12_remain = self.f1 % self.f2 \ if self.f1 > self.f2 else self.f2 % self.f1 self.f12_amp = self.f2_amp / self.f1_amp self.f12_phase = self.f2_phase - self.f1_phase """ def get_period_uncertainty(self, fx, fy, jmax, fx_width=100): """ Get uncertainty of a period. The uncertainty is defined as the half width of the frequencies around the peak, that becomes lower than average + standard deviation of the power spectrum. Since we may not have fine resolution around the peak, we do not assume it is gaussian. So, no scaling factor of 2.355 (= 2 * sqrt(2 * ln2)) is applied. Parameters ---------- fx : array_like An array of frequencies. fy : array_like An array of amplitudes. jmax : int An index at the peak frequency. fx_width : int, optional Width of power spectrum to calculate uncertainty. Returns ------- p_uncertain : float Period uncertainty. """ # Get subset start_index = jmax - fx_width end_index = jmax + fx_width if start_index < 0: start_index = 0 if end_index > len(fx) - 1: end_index = len(fx) - 1 fx_subset = fx[start_index:end_index] fy_subset = fy[start_index:end_index] fy_mean = np.median(fy_subset) fy_std = np.std(fy_subset) # Find peak max_index = np.argmax(fy_subset) # Find list whose powers become lower than average + std. index = np.where(fy_subset <= fy_mean + fy_std)[0] # Find the edge at left and right. This is the full width. left_index = index[(index < max_index)] if len(left_index) == 0: left_index = 0 else: left_index = left_index[-1] right_index = index[(index > max_index)] if len(right_index) == 0: right_index = len(fy_subset) - 1 else: right_index = right_index[0] # We assume the half of the full width is the period uncertainty. half_width = (1. / fx_subset[left_index] - 1. / fx_subset[right_index]) / 2. period_uncertainty = half_width return period_uncertainty def residuals(self, pars, x, y, order): """ Residual of Fourier Series. Parameters ---------- pars : array_like Fourier series parameters. x : array_like An array of date. y : array_like An array of true values to fit. order : int An order of Fourier Series. """ return y - self.fourier_series(pars, x, order) def fourier_series(self, pars, x, order): """ Function to fit Fourier Series. Parameters ---------- x : array_like An array of date divided by period. It doesn't need to be sorted. pars : array_like Fourier series parameters. order : int An order of Fourier series. """ sum = pars[0] for i in range(order): sum += pars[i * 2 + 1] * np.sin(2 * np.pi * (i + 1) * x) \ + pars[i * 2 + 2] * np.cos(2 * np.pi * (i + 1) * x) return sum def get_stetson_k(self, mag, avg, err): """ Return Stetson K feature. Parameters ---------- mag : array_like An array of magnitude. avg : float An average value of magnitudes. err : array_like An array of magnitude errors. Returns ------- stetson_k : float Stetson K value. """ residual = (mag - avg) / err stetson_k = np.sum(np.fabs(residual)) \ / np.sqrt(np.sum(residual * residual)) / np.sqrt(len(mag)) return stetson_k def half_mag_amplitude_ratio(self, mag, avg, weight): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. weight : array_like An array of weight. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) lower_weight = weight[index] lower_weight_sum = np.sum(lower_weight) lower_mag = mag[index] lower_weighted_std = np.sum((lower_mag - avg) ** 2 * lower_weight) / \ lower_weight_sum # For higher (brighter) magnitude than average. index = np.where(mag <= avg) higher_weight = weight[index] higher_weight_sum = np.sum(higher_weight) higher_mag = mag[index] higher_weighted_std = np.sum((higher_mag - avg) ** 2 * higher_weight) / \ higher_weight_sum # Return ratio. return np.sqrt(lower_weighted_std / higher_weighted_std) def half_mag_amplitude_ratio2(self, mag, avg): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) fainter_mag = mag[index] lower_sum = np.sum((fainter_mag - avg) ** 2) / len(fainter_mag) # For higher (brighter) magnitude than average. index = np.where(mag <= avg) brighter_mag = mag[index] higher_sum = np.sum((avg - brighter_mag) ** 2) / len(brighter_mag) # Return ratio. return np.sqrt(lower_sum / higher_sum) def get_eta(self, mag, std): """ Return Eta feature. Parameters ---------- mag : array_like An array of magnitudes. std : array_like A standard deviation of magnitudes. Returns ------- eta : float The value of Eta index. """ diff = mag[1:] - mag[:len(mag) - 1] eta = np.sum(diff * diff) / (len(mag) - 1.) / std / std return eta def slope_percentile(self, date, mag): """ Return 10% and 90% percentile of slope. Parameters ---------- date : array_like An array of phase-folded date. Sorted. mag : array_like An array of phase-folded magnitudes. Sorted by date. Returns ------- per_10 : float 10% percentile values of slope. per_90 : float 90% percentile values of slope. """ date_diff = date[1:] - date[:len(date) - 1] mag_diff = mag[1:] - mag[:len(mag) - 1] # Remove zero mag_diff. index = np.where(mag_diff != 0.) date_diff = date_diff[index] mag_diff = mag_diff[index] # Derive slope. slope = date_diff / mag_diff percentile_10 = np.percentile(slope, 10.) percentile_90 = np.percentile(slope, 90.) return percentile_10, percentile_90 def get_features2(self): """ Return all features with its names. Returns ------- names : list Feature names. values : list Feature values """ feature_names = [] feature_values = [] # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): # Omit input variables such as date, mag, err, etc. if not (name == 'date' or name == 'mag' or name == 'err' or name == 'n_threads' or name == 'min_period'): # Filter some other unnecessary features. if not (name == 'f' or name == 'f_phase' or name == 'period_log10FAP' or name == 'weight' or name == 'weighted_sum' or name == 'median' or name == 'mean' or name == 'std'): feature_names.append(name) # Sort by the names. # Sorting should be done to keep maintaining the same order of features. feature_names.sort() # Get feature values. for name in feature_names: feature_values.append(all_vars[name]) return feature_names, feature_values def get_features(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ ''' features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features ''' return self.get_features_all() def get_features_all(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list_all: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features
dwkim78/upsilon
upsilon/extract_features/extract_features.py
ExtractFeatures.get_features2
python
def get_features2(self): feature_names = [] feature_values = [] # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): # Omit input variables such as date, mag, err, etc. if not (name == 'date' or name == 'mag' or name == 'err' or name == 'n_threads' or name == 'min_period'): # Filter some other unnecessary features. if not (name == 'f' or name == 'f_phase' or name == 'period_log10FAP' or name == 'weight' or name == 'weighted_sum' or name == 'median' or name == 'mean' or name == 'std'): feature_names.append(name) # Sort by the names. # Sorting should be done to keep maintaining the same order of features. feature_names.sort() # Get feature values. for name in feature_names: feature_values.append(all_vars[name]) return feature_names, feature_values
Return all features with its names. Returns ------- names : list Feature names. values : list Feature values
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/extract_features.py#L564-L600
null
class ExtractFeatures: """ Extract variability features of a light curve. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. err : array_like, optional An array of magnitude error. If None, std(mag) will be used. n_threads : int, optional The number of cores to use to derive periods. min_period : float, optional The minimum period to calculate. """ def __init__(self, date, mag, err=None, n_threads=4, min_period=0.03): # Set basic values. if not isinstance(date, np.ndarray): date = np.array(date) if not isinstance(mag, np.ndarray): mag = np.array(mag) self.date = date self.mag = mag if err is not None: if not isinstance(err, np.ndarray): err = np.array(err) self.err = err else: self.err = np.ones(len(self.mag)) * np.std(self.mag) # Check length. if (len(self.date) != len(self.mag)) or \ (len(self.date) != len(self.err)) or \ (len(self.mag) != len(self.err)): raise RuntimeError('The length of date, mag, and err must be same.') # if the number of data points is too small. min_n_data = 80 if len(self.date) < min_n_data: warnings.warn('The number of data points are less than %d.' % min_n_data) n_threads = int(n_threads) if n_threads > multiprocessing.cpu_count(): self.n_threads = multiprocessing.cpu_count() else: if n_threads <= 0: self.n_threads = 1 else: self.n_threads = n_threads min_period = float(min_period) if min_period <= 0: self.min_period = 0.03 else: self.min_period = min_period def run(self): """Run feature extraction modules.""" # shallow_run must be executed prior to deep_run # since shallow_run calculates several values needed for deep_run. self.shallow_run() self.deep_run() def shallow_run(self): """Derive not-period-based features.""" # Number of data points self.n_points = len(self.date) # Weight calculation. # All zero values. if not self.err.any(): self.err = np.ones(len(self.mag)) * np.std(self.mag) # Some zero values. elif not self.err.all(): np.putmask(self.err, self.err==0, np.median(self.err)) self.weight = 1. / self.err self.weighted_sum = np.sum(self.weight) # Simple statistics, mean, median and std. self.mean = np.mean(self.mag) self.median = np.median(self.mag) self.std = np.std(self.mag) # Weighted mean and std. self.weighted_mean = np.sum(self.mag * self.weight) / self.weighted_sum self.weighted_std = np.sqrt(np.sum((self.mag - self.weighted_mean) ** 2 \ * self.weight) / self.weighted_sum) # Skewness and kurtosis. self.skewness = ss.skew(self.mag) self.kurtosis = ss.kurtosis(self.mag) # Normalization-test. Shapiro-Wilk test. shapiro = ss.shapiro(self.mag) self.shapiro_w = shapiro[0] # self.shapiro_log10p = np.log10(shapiro[1]) # Percentile features. self.quartile31 = np.percentile(self.mag, 75) \ - np.percentile(self.mag, 25) # Stetson K. self.stetson_k = self.get_stetson_k(self.mag, self.median, self.err) # Ratio between higher and lower amplitude than average. self.hl_amp_ratio = self.half_mag_amplitude_ratio( self.mag, self.median, self.weight) # This second function's value is very similar with the above one. # self.hl_amp_ratio2 = self.half_mag_amplitude_ratio2( # self.mag, self.median) # Cusum self.cusum = self.get_cusum(self.mag) # Eta self.eta = self.get_eta(self.mag, self.weighted_std) def deep_run(self): """Derive period-based features.""" # Lomb-Scargle period finding. self.get_period_LS(self.date, self.mag, self.n_threads, self.min_period) # Features based on a phase-folded light curve # such as Eta, slope-percentile, etc. # Should be called after the getPeriodLS() is called. # Created phased a folded light curve. # We use period * 2 to take eclipsing binaries into account. phase_folded_date = self.date % (self.period * 2.) sorted_index = np.argsort(phase_folded_date) folded_date = phase_folded_date[sorted_index] folded_mag = self.mag[sorted_index] # phase Eta self.phase_eta = self.get_eta(folded_mag, self.weighted_std) # Slope percentile. self.slope_per10, self.slope_per90 = \ self.slope_percentile(folded_date, folded_mag) # phase Cusum self.phase_cusum = self.get_cusum(folded_mag) def get_period_LS(self, date, mag, n_threads, min_period): """ Period finding using the Lomb-Scargle algorithm. Finding two periods. The second period is estimated after whitening the first period. Calculating various other features as well using derived periods. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. n_threads : int The number of threads to use. min_period : float The minimum period to calculate. """ # DO NOT CHANGE THESE PARAMETERS. oversampling = 3. hifac = int((max(date) - min(date)) / len(date) / min_period * 2.) # Minimum hifac if hifac < 100: hifac = 100 # Lomb-Scargle. fx, fy, nout, jmax, prob = pLS.fasper(date, mag, oversampling, hifac, n_threads) self.f = fx[jmax] self.period = 1. / self.f self.period_uncertainty = self.get_period_uncertainty(fx, fy, jmax) self.period_log10FAP = \ np.log10(pLS.getSignificance(fx, fy, nout, oversampling)[jmax]) # self.f_SNR1 = fy[jmax] / np.median(fy) self.period_SNR = (fy[jmax] - np.median(fy)) / np.std(fy) # Fit Fourier Series of order 3. order = 3 # Initial guess of Fourier coefficients. p0 = np.ones(order * 2 + 1) date_period = (date % self.period) / self.period p1, success = leastsq(self.residuals, p0, args=(date_period, mag, order)) # fitted_y = self.FourierSeries(p1, date_period, order) # print p1, self.mean, self.median # plt.plot(date_period, self.mag, 'b+') # plt.show() # Derive Fourier features for the first period. # Petersen, J. O., 1986, A&A self.amplitude = np.sqrt(p1[1] ** 2 + p1[2] ** 2) self.r21 = np.sqrt(p1[3] ** 2 + p1[4] ** 2) / self.amplitude self.r31 = np.sqrt(p1[5] ** 2 + p1[6] ** 2) / self.amplitude self.f_phase = np.arctan(-p1[1] / p1[2]) self.phi21 = np.arctan(-p1[3] / p1[4]) - 2. * self.f_phase self.phi31 = np.arctan(-p1[5] / p1[6]) - 3. * self.f_phase """ # Derive a second period. # Whitening a light curve. residual_mag = mag - fitted_y # Lomb-Scargle again to find the second period. omega_top, power_top = search_frequencies(date, residual_mag, err, #LS_kwargs={'generalized':True, 'subtract_mean':True}, n_eval=5000, n_retry=3, n_save=50) self.period2 = 2*np.pi/omega_top[np.where(power_top==np.max(power_top))][0] self.f2 = 1. / self.period2 self.f2_SNR = power_top[np.where(power_top==np.max(power_top))][0] \ * (len(self.date) - 1) / 2. # Fit Fourier Series again. p0 = [1.] * order * 2 date_period = (date % self.period) / self.period p2, success = leastsq(self.residuals, p0, args=(date_period, residual_mag, order)) fitted_y = self.FourierSeries(p2, date_period, order) #plt.plot(date%self.period2, residual_mag, 'b+') #plt.show() # Derive Fourier features for the first second. self.f2_amp = 2. * np.sqrt(p2[1]**2 + p2[2]**2) self.f2_R21 = np.sqrt(p2[3]**2 + p2[4]**2) / self.f2_amp self.f2_R31 = np.sqrt(p2[5]**2 + p2[6]**2) / self.f2_amp self.f2_R41 = np.sqrt(p2[7]**2 + p2[8]**2) / self.f2_amp self.f2_R51 = np.sqrt(p2[9]**2 + p2[10]**2) / self.f2_amp self.f2_phase = np.arctan(-p2[1] / p2[2]) self.f2_phi21 = np.arctan(-p2[3] / p2[4]) - 2. * self.f2_phase self.f2_phi31 = np.arctan(-p2[5] / p2[6]) - 3. * self.f2_phase self.f2_phi41 = np.arctan(-p2[7] / p2[8]) - 4. * self.f2_phase self.f2_phi51 = np.arctan(-p2[9] / p2[10]) - 5. * self.f2_phase # Calculate features using the first and second periods. self.f12_ratio = self.f2 / self.f1 self.f12_remain = self.f1 % self.f2 \ if self.f1 > self.f2 else self.f2 % self.f1 self.f12_amp = self.f2_amp / self.f1_amp self.f12_phase = self.f2_phase - self.f1_phase """ def get_period_uncertainty(self, fx, fy, jmax, fx_width=100): """ Get uncertainty of a period. The uncertainty is defined as the half width of the frequencies around the peak, that becomes lower than average + standard deviation of the power spectrum. Since we may not have fine resolution around the peak, we do not assume it is gaussian. So, no scaling factor of 2.355 (= 2 * sqrt(2 * ln2)) is applied. Parameters ---------- fx : array_like An array of frequencies. fy : array_like An array of amplitudes. jmax : int An index at the peak frequency. fx_width : int, optional Width of power spectrum to calculate uncertainty. Returns ------- p_uncertain : float Period uncertainty. """ # Get subset start_index = jmax - fx_width end_index = jmax + fx_width if start_index < 0: start_index = 0 if end_index > len(fx) - 1: end_index = len(fx) - 1 fx_subset = fx[start_index:end_index] fy_subset = fy[start_index:end_index] fy_mean = np.median(fy_subset) fy_std = np.std(fy_subset) # Find peak max_index = np.argmax(fy_subset) # Find list whose powers become lower than average + std. index = np.where(fy_subset <= fy_mean + fy_std)[0] # Find the edge at left and right. This is the full width. left_index = index[(index < max_index)] if len(left_index) == 0: left_index = 0 else: left_index = left_index[-1] right_index = index[(index > max_index)] if len(right_index) == 0: right_index = len(fy_subset) - 1 else: right_index = right_index[0] # We assume the half of the full width is the period uncertainty. half_width = (1. / fx_subset[left_index] - 1. / fx_subset[right_index]) / 2. period_uncertainty = half_width return period_uncertainty def residuals(self, pars, x, y, order): """ Residual of Fourier Series. Parameters ---------- pars : array_like Fourier series parameters. x : array_like An array of date. y : array_like An array of true values to fit. order : int An order of Fourier Series. """ return y - self.fourier_series(pars, x, order) def fourier_series(self, pars, x, order): """ Function to fit Fourier Series. Parameters ---------- x : array_like An array of date divided by period. It doesn't need to be sorted. pars : array_like Fourier series parameters. order : int An order of Fourier series. """ sum = pars[0] for i in range(order): sum += pars[i * 2 + 1] * np.sin(2 * np.pi * (i + 1) * x) \ + pars[i * 2 + 2] * np.cos(2 * np.pi * (i + 1) * x) return sum def get_stetson_k(self, mag, avg, err): """ Return Stetson K feature. Parameters ---------- mag : array_like An array of magnitude. avg : float An average value of magnitudes. err : array_like An array of magnitude errors. Returns ------- stetson_k : float Stetson K value. """ residual = (mag - avg) / err stetson_k = np.sum(np.fabs(residual)) \ / np.sqrt(np.sum(residual * residual)) / np.sqrt(len(mag)) return stetson_k def half_mag_amplitude_ratio(self, mag, avg, weight): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. weight : array_like An array of weight. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) lower_weight = weight[index] lower_weight_sum = np.sum(lower_weight) lower_mag = mag[index] lower_weighted_std = np.sum((lower_mag - avg) ** 2 * lower_weight) / \ lower_weight_sum # For higher (brighter) magnitude than average. index = np.where(mag <= avg) higher_weight = weight[index] higher_weight_sum = np.sum(higher_weight) higher_mag = mag[index] higher_weighted_std = np.sum((higher_mag - avg) ** 2 * higher_weight) / \ higher_weight_sum # Return ratio. return np.sqrt(lower_weighted_std / higher_weighted_std) def half_mag_amplitude_ratio2(self, mag, avg): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) fainter_mag = mag[index] lower_sum = np.sum((fainter_mag - avg) ** 2) / len(fainter_mag) # For higher (brighter) magnitude than average. index = np.where(mag <= avg) brighter_mag = mag[index] higher_sum = np.sum((avg - brighter_mag) ** 2) / len(brighter_mag) # Return ratio. return np.sqrt(lower_sum / higher_sum) def get_eta(self, mag, std): """ Return Eta feature. Parameters ---------- mag : array_like An array of magnitudes. std : array_like A standard deviation of magnitudes. Returns ------- eta : float The value of Eta index. """ diff = mag[1:] - mag[:len(mag) - 1] eta = np.sum(diff * diff) / (len(mag) - 1.) / std / std return eta def slope_percentile(self, date, mag): """ Return 10% and 90% percentile of slope. Parameters ---------- date : array_like An array of phase-folded date. Sorted. mag : array_like An array of phase-folded magnitudes. Sorted by date. Returns ------- per_10 : float 10% percentile values of slope. per_90 : float 90% percentile values of slope. """ date_diff = date[1:] - date[:len(date) - 1] mag_diff = mag[1:] - mag[:len(mag) - 1] # Remove zero mag_diff. index = np.where(mag_diff != 0.) date_diff = date_diff[index] mag_diff = mag_diff[index] # Derive slope. slope = date_diff / mag_diff percentile_10 = np.percentile(slope, 10.) percentile_90 = np.percentile(slope, 90.) return percentile_10, percentile_90 def get_cusum(self, mag): """ Return max - min of cumulative sum. Parameters ---------- mag : array_like An array of magnitudes. Returns ------- mm_cusum : float Max - min of cumulative sum. """ c = np.cumsum(mag - self.weighted_mean) / len(mag) / self.weighted_std return np.max(c) - np.min(c) def get_features(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ ''' features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features ''' return self.get_features_all() def get_features_all(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list_all: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features
dwkim78/upsilon
upsilon/extract_features/extract_features.py
ExtractFeatures.get_features_all
python
def get_features_all(self): features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list_all: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features
Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/extract_features.py#L631-L654
null
class ExtractFeatures: """ Extract variability features of a light curve. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. err : array_like, optional An array of magnitude error. If None, std(mag) will be used. n_threads : int, optional The number of cores to use to derive periods. min_period : float, optional The minimum period to calculate. """ def __init__(self, date, mag, err=None, n_threads=4, min_period=0.03): # Set basic values. if not isinstance(date, np.ndarray): date = np.array(date) if not isinstance(mag, np.ndarray): mag = np.array(mag) self.date = date self.mag = mag if err is not None: if not isinstance(err, np.ndarray): err = np.array(err) self.err = err else: self.err = np.ones(len(self.mag)) * np.std(self.mag) # Check length. if (len(self.date) != len(self.mag)) or \ (len(self.date) != len(self.err)) or \ (len(self.mag) != len(self.err)): raise RuntimeError('The length of date, mag, and err must be same.') # if the number of data points is too small. min_n_data = 80 if len(self.date) < min_n_data: warnings.warn('The number of data points are less than %d.' % min_n_data) n_threads = int(n_threads) if n_threads > multiprocessing.cpu_count(): self.n_threads = multiprocessing.cpu_count() else: if n_threads <= 0: self.n_threads = 1 else: self.n_threads = n_threads min_period = float(min_period) if min_period <= 0: self.min_period = 0.03 else: self.min_period = min_period def run(self): """Run feature extraction modules.""" # shallow_run must be executed prior to deep_run # since shallow_run calculates several values needed for deep_run. self.shallow_run() self.deep_run() def shallow_run(self): """Derive not-period-based features.""" # Number of data points self.n_points = len(self.date) # Weight calculation. # All zero values. if not self.err.any(): self.err = np.ones(len(self.mag)) * np.std(self.mag) # Some zero values. elif not self.err.all(): np.putmask(self.err, self.err==0, np.median(self.err)) self.weight = 1. / self.err self.weighted_sum = np.sum(self.weight) # Simple statistics, mean, median and std. self.mean = np.mean(self.mag) self.median = np.median(self.mag) self.std = np.std(self.mag) # Weighted mean and std. self.weighted_mean = np.sum(self.mag * self.weight) / self.weighted_sum self.weighted_std = np.sqrt(np.sum((self.mag - self.weighted_mean) ** 2 \ * self.weight) / self.weighted_sum) # Skewness and kurtosis. self.skewness = ss.skew(self.mag) self.kurtosis = ss.kurtosis(self.mag) # Normalization-test. Shapiro-Wilk test. shapiro = ss.shapiro(self.mag) self.shapiro_w = shapiro[0] # self.shapiro_log10p = np.log10(shapiro[1]) # Percentile features. self.quartile31 = np.percentile(self.mag, 75) \ - np.percentile(self.mag, 25) # Stetson K. self.stetson_k = self.get_stetson_k(self.mag, self.median, self.err) # Ratio between higher and lower amplitude than average. self.hl_amp_ratio = self.half_mag_amplitude_ratio( self.mag, self.median, self.weight) # This second function's value is very similar with the above one. # self.hl_amp_ratio2 = self.half_mag_amplitude_ratio2( # self.mag, self.median) # Cusum self.cusum = self.get_cusum(self.mag) # Eta self.eta = self.get_eta(self.mag, self.weighted_std) def deep_run(self): """Derive period-based features.""" # Lomb-Scargle period finding. self.get_period_LS(self.date, self.mag, self.n_threads, self.min_period) # Features based on a phase-folded light curve # such as Eta, slope-percentile, etc. # Should be called after the getPeriodLS() is called. # Created phased a folded light curve. # We use period * 2 to take eclipsing binaries into account. phase_folded_date = self.date % (self.period * 2.) sorted_index = np.argsort(phase_folded_date) folded_date = phase_folded_date[sorted_index] folded_mag = self.mag[sorted_index] # phase Eta self.phase_eta = self.get_eta(folded_mag, self.weighted_std) # Slope percentile. self.slope_per10, self.slope_per90 = \ self.slope_percentile(folded_date, folded_mag) # phase Cusum self.phase_cusum = self.get_cusum(folded_mag) def get_period_LS(self, date, mag, n_threads, min_period): """ Period finding using the Lomb-Scargle algorithm. Finding two periods. The second period is estimated after whitening the first period. Calculating various other features as well using derived periods. Parameters ---------- date : array_like An array of observed date, in days. mag : array_like An array of observed magnitude. n_threads : int The number of threads to use. min_period : float The minimum period to calculate. """ # DO NOT CHANGE THESE PARAMETERS. oversampling = 3. hifac = int((max(date) - min(date)) / len(date) / min_period * 2.) # Minimum hifac if hifac < 100: hifac = 100 # Lomb-Scargle. fx, fy, nout, jmax, prob = pLS.fasper(date, mag, oversampling, hifac, n_threads) self.f = fx[jmax] self.period = 1. / self.f self.period_uncertainty = self.get_period_uncertainty(fx, fy, jmax) self.period_log10FAP = \ np.log10(pLS.getSignificance(fx, fy, nout, oversampling)[jmax]) # self.f_SNR1 = fy[jmax] / np.median(fy) self.period_SNR = (fy[jmax] - np.median(fy)) / np.std(fy) # Fit Fourier Series of order 3. order = 3 # Initial guess of Fourier coefficients. p0 = np.ones(order * 2 + 1) date_period = (date % self.period) / self.period p1, success = leastsq(self.residuals, p0, args=(date_period, mag, order)) # fitted_y = self.FourierSeries(p1, date_period, order) # print p1, self.mean, self.median # plt.plot(date_period, self.mag, 'b+') # plt.show() # Derive Fourier features for the first period. # Petersen, J. O., 1986, A&A self.amplitude = np.sqrt(p1[1] ** 2 + p1[2] ** 2) self.r21 = np.sqrt(p1[3] ** 2 + p1[4] ** 2) / self.amplitude self.r31 = np.sqrt(p1[5] ** 2 + p1[6] ** 2) / self.amplitude self.f_phase = np.arctan(-p1[1] / p1[2]) self.phi21 = np.arctan(-p1[3] / p1[4]) - 2. * self.f_phase self.phi31 = np.arctan(-p1[5] / p1[6]) - 3. * self.f_phase """ # Derive a second period. # Whitening a light curve. residual_mag = mag - fitted_y # Lomb-Scargle again to find the second period. omega_top, power_top = search_frequencies(date, residual_mag, err, #LS_kwargs={'generalized':True, 'subtract_mean':True}, n_eval=5000, n_retry=3, n_save=50) self.period2 = 2*np.pi/omega_top[np.where(power_top==np.max(power_top))][0] self.f2 = 1. / self.period2 self.f2_SNR = power_top[np.where(power_top==np.max(power_top))][0] \ * (len(self.date) - 1) / 2. # Fit Fourier Series again. p0 = [1.] * order * 2 date_period = (date % self.period) / self.period p2, success = leastsq(self.residuals, p0, args=(date_period, residual_mag, order)) fitted_y = self.FourierSeries(p2, date_period, order) #plt.plot(date%self.period2, residual_mag, 'b+') #plt.show() # Derive Fourier features for the first second. self.f2_amp = 2. * np.sqrt(p2[1]**2 + p2[2]**2) self.f2_R21 = np.sqrt(p2[3]**2 + p2[4]**2) / self.f2_amp self.f2_R31 = np.sqrt(p2[5]**2 + p2[6]**2) / self.f2_amp self.f2_R41 = np.sqrt(p2[7]**2 + p2[8]**2) / self.f2_amp self.f2_R51 = np.sqrt(p2[9]**2 + p2[10]**2) / self.f2_amp self.f2_phase = np.arctan(-p2[1] / p2[2]) self.f2_phi21 = np.arctan(-p2[3] / p2[4]) - 2. * self.f2_phase self.f2_phi31 = np.arctan(-p2[5] / p2[6]) - 3. * self.f2_phase self.f2_phi41 = np.arctan(-p2[7] / p2[8]) - 4. * self.f2_phase self.f2_phi51 = np.arctan(-p2[9] / p2[10]) - 5. * self.f2_phase # Calculate features using the first and second periods. self.f12_ratio = self.f2 / self.f1 self.f12_remain = self.f1 % self.f2 \ if self.f1 > self.f2 else self.f2 % self.f1 self.f12_amp = self.f2_amp / self.f1_amp self.f12_phase = self.f2_phase - self.f1_phase """ def get_period_uncertainty(self, fx, fy, jmax, fx_width=100): """ Get uncertainty of a period. The uncertainty is defined as the half width of the frequencies around the peak, that becomes lower than average + standard deviation of the power spectrum. Since we may not have fine resolution around the peak, we do not assume it is gaussian. So, no scaling factor of 2.355 (= 2 * sqrt(2 * ln2)) is applied. Parameters ---------- fx : array_like An array of frequencies. fy : array_like An array of amplitudes. jmax : int An index at the peak frequency. fx_width : int, optional Width of power spectrum to calculate uncertainty. Returns ------- p_uncertain : float Period uncertainty. """ # Get subset start_index = jmax - fx_width end_index = jmax + fx_width if start_index < 0: start_index = 0 if end_index > len(fx) - 1: end_index = len(fx) - 1 fx_subset = fx[start_index:end_index] fy_subset = fy[start_index:end_index] fy_mean = np.median(fy_subset) fy_std = np.std(fy_subset) # Find peak max_index = np.argmax(fy_subset) # Find list whose powers become lower than average + std. index = np.where(fy_subset <= fy_mean + fy_std)[0] # Find the edge at left and right. This is the full width. left_index = index[(index < max_index)] if len(left_index) == 0: left_index = 0 else: left_index = left_index[-1] right_index = index[(index > max_index)] if len(right_index) == 0: right_index = len(fy_subset) - 1 else: right_index = right_index[0] # We assume the half of the full width is the period uncertainty. half_width = (1. / fx_subset[left_index] - 1. / fx_subset[right_index]) / 2. period_uncertainty = half_width return period_uncertainty def residuals(self, pars, x, y, order): """ Residual of Fourier Series. Parameters ---------- pars : array_like Fourier series parameters. x : array_like An array of date. y : array_like An array of true values to fit. order : int An order of Fourier Series. """ return y - self.fourier_series(pars, x, order) def fourier_series(self, pars, x, order): """ Function to fit Fourier Series. Parameters ---------- x : array_like An array of date divided by period. It doesn't need to be sorted. pars : array_like Fourier series parameters. order : int An order of Fourier series. """ sum = pars[0] for i in range(order): sum += pars[i * 2 + 1] * np.sin(2 * np.pi * (i + 1) * x) \ + pars[i * 2 + 2] * np.cos(2 * np.pi * (i + 1) * x) return sum def get_stetson_k(self, mag, avg, err): """ Return Stetson K feature. Parameters ---------- mag : array_like An array of magnitude. avg : float An average value of magnitudes. err : array_like An array of magnitude errors. Returns ------- stetson_k : float Stetson K value. """ residual = (mag - avg) / err stetson_k = np.sum(np.fabs(residual)) \ / np.sqrt(np.sum(residual * residual)) / np.sqrt(len(mag)) return stetson_k def half_mag_amplitude_ratio(self, mag, avg, weight): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. weight : array_like An array of weight. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) lower_weight = weight[index] lower_weight_sum = np.sum(lower_weight) lower_mag = mag[index] lower_weighted_std = np.sum((lower_mag - avg) ** 2 * lower_weight) / \ lower_weight_sum # For higher (brighter) magnitude than average. index = np.where(mag <= avg) higher_weight = weight[index] higher_weight_sum = np.sum(higher_weight) higher_mag = mag[index] higher_weighted_std = np.sum((higher_mag - avg) ** 2 * higher_weight) / \ higher_weight_sum # Return ratio. return np.sqrt(lower_weighted_std / higher_weighted_std) def half_mag_amplitude_ratio2(self, mag, avg): """ Return ratio of amplitude of higher and lower magnitudes. A ratio of amplitude of higher and lower magnitudes than average, considering weights. This ratio, by definition, should be higher for EB than for others. Parameters ---------- mag : array_like An array of magnitudes. avg : float An average value of magnitudes. Returns ------- hl_ratio : float Ratio of amplitude of higher and lower magnitudes than average. """ # For lower (fainter) magnitude than average. index = np.where(mag > avg) fainter_mag = mag[index] lower_sum = np.sum((fainter_mag - avg) ** 2) / len(fainter_mag) # For higher (brighter) magnitude than average. index = np.where(mag <= avg) brighter_mag = mag[index] higher_sum = np.sum((avg - brighter_mag) ** 2) / len(brighter_mag) # Return ratio. return np.sqrt(lower_sum / higher_sum) def get_eta(self, mag, std): """ Return Eta feature. Parameters ---------- mag : array_like An array of magnitudes. std : array_like A standard deviation of magnitudes. Returns ------- eta : float The value of Eta index. """ diff = mag[1:] - mag[:len(mag) - 1] eta = np.sum(diff * diff) / (len(mag) - 1.) / std / std return eta def slope_percentile(self, date, mag): """ Return 10% and 90% percentile of slope. Parameters ---------- date : array_like An array of phase-folded date. Sorted. mag : array_like An array of phase-folded magnitudes. Sorted by date. Returns ------- per_10 : float 10% percentile values of slope. per_90 : float 90% percentile values of slope. """ date_diff = date[1:] - date[:len(date) - 1] mag_diff = mag[1:] - mag[:len(mag) - 1] # Remove zero mag_diff. index = np.where(mag_diff != 0.) date_diff = date_diff[index] mag_diff = mag_diff[index] # Derive slope. slope = date_diff / mag_diff percentile_10 = np.percentile(slope, 10.) percentile_90 = np.percentile(slope, 90.) return percentile_10, percentile_90 def get_cusum(self, mag): """ Return max - min of cumulative sum. Parameters ---------- mag : array_like An array of magnitudes. Returns ------- mm_cusum : float Max - min of cumulative sum. """ c = np.cumsum(mag - self.weighted_mean) / len(mag) / self.weighted_std return np.max(c) - np.min(c) def get_features2(self): """ Return all features with its names. Returns ------- names : list Feature names. values : list Feature values """ feature_names = [] feature_values = [] # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): # Omit input variables such as date, mag, err, etc. if not (name == 'date' or name == 'mag' or name == 'err' or name == 'n_threads' or name == 'min_period'): # Filter some other unnecessary features. if not (name == 'f' or name == 'f_phase' or name == 'period_log10FAP' or name == 'weight' or name == 'weighted_sum' or name == 'median' or name == 'mean' or name == 'std'): feature_names.append(name) # Sort by the names. # Sorting should be done to keep maintaining the same order of features. feature_names.sort() # Get feature values. for name in feature_names: feature_values.append(all_vars[name]) return feature_names, feature_values def get_features(self): """ Return all features with its names. Regardless of being used for train and prediction. Sorted by the names. Returns ------- all_features : OrderedDict Features dictionary. """ ''' features = {} # Get all the names of features. all_vars = vars(self) for name in all_vars.keys(): if name in feature_names_list: features[name] = all_vars[name] # Sort by the keys (i.e. feature names). features = OrderedDict(sorted(features.items(), key=lambda t: t[0])) return features ''' return self.get_features_all()
dwkim78/upsilon
upsilon/extract_features/is_period_alias.py
is_period_alias
python
def is_period_alias(period): # Based on the period vs periodSN plot of EROS-2 dataset (Kim+ 2014). # Period alias occurs mostly at ~1 and ~30. # Check each 1, 2, 3, 4, 5 factors. for i in range(1, 6): # One-day and one-month alias if (.99 / float(i)) < period < (1.004 / float(i)): return True if (1.03 / float(i)) < period < (1.04 / float(i)): return True if (29.2 / float(i)) < period < (29.9 / float(i)): return True # From candidates from the two fields 01, 08. # All of them are close to one day (or sidereal) alias. if (0.96465 / float(i)) < period < (0.96485 / float(i)): return True if (0.96725 / float(i)) < period < (0.96745 / float(i)): return True if (0.98190 / float(i)) < period < (0.98230 / float(i)): return True if (1.01034 / float(i)) < period < (1.01076 / float(i)): return True if (1.01568 / float(i)) < period < (1.01604 / float(i)): return True if (1.01718 / float(i)) < period < (1.01742 / float(i)): return True # From the all candidates from the entire LMC fields. # Some of these could be overlapped with the above cuts. if (0.50776 / float(i)) < period < (0.50861 / float(i)): return True if (0.96434 / float(i)) < period < (0.9652 / float(i)): return True if (0.96688 / float(i)) < period < (0.96731 / float(i)): return True if (1.0722 / float(i)) < period < (1.0729 / float(i)): return True if (27.1 / float(i)) < period < (27.5 / float(i)): return True # Not in the range of any alias. return False
Check if a given period is possibly an alias. Parameters ---------- period : float A period to test if it is a possible alias or not. Returns ------- is_alias : boolean True if the given period is in a range of period alias.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/is_period_alias.py#L1-L57
null
dwkim78/upsilon
upsilon/datasets/base.py
load_EROS_lc
python
def load_EROS_lc(filename='lm0010n22323.time'): module_path = dirname(__file__) file_path = join(module_path, 'lightcurves', filename) data = np.loadtxt(file_path) date = data[:, 0] mag = data[:, 1] err = data[:, 2] return date, mag, err
Read an EROS light curve and return its data. Parameters ---------- filename : str, optional A light-curve filename. Returns ------- dates : numpy.ndarray An array of dates. magnitudes : numpy.ndarray An array of magnitudes. errors : numpy.ndarray An array of magnitudes errors.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/datasets/base.py#L12-L39
null
""" Base IO code for all datasets """ import sys import numpy as np from os.path import dirname from os.path import join def load_rf_model(): """ Return the UPSILoN random forests classifier. The classifier is trained using OGLE and EROS periodic variables (Kim et al. 2015). Returns ------- clf : sklearn.ensemble.RandomForestClassifier The UPSILoN random forests classifier. """ import gzip try: import cPickle as pickle except: import pickle module_path = dirname(__file__) file_path = join(module_path, 'models/rf.model.sub.github.gz') # For Python 3. if sys.version_info.major >= 3: clf = pickle.load(gzip.open(file_path, 'rb'), encoding='latin1') # For Python 2. else: clf = pickle.load(gzip.open(file_path, 'rb')) return clf
dwkim78/upsilon
upsilon/datasets/base.py
load_rf_model
python
def load_rf_model(): import gzip try: import cPickle as pickle except: import pickle module_path = dirname(__file__) file_path = join(module_path, 'models/rf.model.sub.github.gz') # For Python 3. if sys.version_info.major >= 3: clf = pickle.load(gzip.open(file_path, 'rb'), encoding='latin1') # For Python 2. else: clf = pickle.load(gzip.open(file_path, 'rb')) return clf
Return the UPSILoN random forests classifier. The classifier is trained using OGLE and EROS periodic variables (Kim et al. 2015). Returns ------- clf : sklearn.ensemble.RandomForestClassifier The UPSILoN random forests classifier.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/datasets/base.py#L42-L71
null
""" Base IO code for all datasets """ import sys import numpy as np from os.path import dirname from os.path import join def load_EROS_lc(filename='lm0010n22323.time'): """ Read an EROS light curve and return its data. Parameters ---------- filename : str, optional A light-curve filename. Returns ------- dates : numpy.ndarray An array of dates. magnitudes : numpy.ndarray An array of magnitudes. errors : numpy.ndarray An array of magnitudes errors. """ module_path = dirname(__file__) file_path = join(module_path, 'lightcurves', filename) data = np.loadtxt(file_path) date = data[:, 0] mag = data[:, 1] err = data[:, 2] return date, mag, err
dwkim78/upsilon
upsilon/extract_features/period_LS_pyfftw.py
fasper
python
def fasper(x, y, ofac, hifac, n_threads, MACC=4): #Check dimensions of input arrays n = long(len(x)) if n != len(y): print('Incompatible arrays.') return #print x, y, hifac, ofac nout = int(0.5*ofac*hifac*n) nfreqt = long(ofac*hifac*n*MACC) #Size the FFT as next power nfreq = 64 # of 2 above nfreqt. while nfreq < nfreqt: nfreq = 2*nfreq ndim = long(2*nfreq) #Compute the mean, variance ave = y.mean() ##sample variance because the divisor is N-1 var = ((y - y.mean())**2).sum()/(len(y) - 1) # and range of the data. xmin = x.min() xmax = x.max() xdif = xmax - xmin #extrapolate the data into the workspaces if is_pyfftw: wk1 = pyfftw.n_byte_align_empty(int(ndim), 16, 'complex') * 0. wk2 = pyfftw.n_byte_align_empty(int(ndim), 16, 'complex') * 0. else: wk1 = zeros(ndim, dtype='complex') wk2 = zeros(ndim, dtype='complex') fac = ndim/(xdif*ofac) fndim = ndim ck = ((x - xmin)*fac) % fndim ckk = (2.0*ck) % fndim for j in range(0, n): __spread__(y[j] - ave, wk1, ndim, ck[j], MACC) __spread__(1.0, wk2, ndim, ckk[j], MACC) #Take the Fast Fourier Transforms. if is_pyfftw: fft_wk1 = pyfftw.builders.ifft(wk1, planner_effort='FFTW_ESTIMATE', threads=n_threads) wk1 = fft_wk1() * len(wk1) fft_wk2 = pyfftw.builders.ifft(wk2, planner_effort='FFTW_ESTIMATE', threads=n_threads) wk2 = fft_wk2() * len(wk2) else: wk1 = ifft(wk1)*len(wk1) wk2 = ifft(wk2)*len(wk1) wk1 = wk1[1:nout + 1] wk2 = wk2[1:nout + 1] rwk1 = wk1.real iwk1 = wk1.imag rwk2 = wk2.real iwk2 = wk2.imag df = 1.0/(xdif*ofac) #Compute the Lomb value for each frequency hypo2 = 2.0*abs(wk2) hc2wt = rwk2/hypo2 hs2wt = iwk2/hypo2 cwt = sqrt(0.5 + hc2wt) swt = sign(hs2wt)*(sqrt(0.5 - hc2wt)) den = 0.5*n + hc2wt*rwk2 + hs2wt*iwk2 cterm = (cwt*rwk1 + swt*iwk1)**2./den sterm = (cwt*iwk1 - swt*rwk1)**2./(n - den) wk1 = df*(arange(nout, dtype='float') + 1.) wk2 = (cterm + sterm)/(2.0*var) pmax = wk2.max() jmax = wk2.argmax() #Significance estimation #expy = exp(-wk2) #effm = 2.0*(nout)/ofac #sig = effm*expy #ind = (sig > 0.01).nonzero() #sig[ind] = 1.0-(1.0-expy[ind])**effm #Estimate significance of largest peak value expy = exp(-pmax) effm = 2.0*(nout)/ofac prob = effm*expy if prob > 0.01: prob = 1.0 - (1.0 - expy)**effm return wk1, wk2, nout, jmax, prob
Given abscissas x (which need not be equally spaced) and ordinates y, and given a desired oversampling factor ofac (a typical value being 4 or larger). this routine creates an array wk1 with a sequence of nout increasing frequencies (not angular frequencies) up to hifac times the "average" Nyquist frequency, and creates an array wk2 with the values of the Lomb normalized periodogram at those frequencies. The arrays x and y are not altered. This routine also returns jmax such that wk2(jmax) is the maximum element in wk2, and prob, an estimate of the significance of that maximum against the hypothesis of random noise. A small value of prob indicates that a significant periodic signal is present. Reference: Press, W. H. & Rybicki, G. B. 1989 ApJ vol. 338, p. 277-280. Fast algorithm for spectral analysis of unevenly sampled data (1989ApJ...338..277P) Arguments: X : Abscissas array, (e.g. an array of times). Y : Ordinates array, (e.g. corresponding counts). Ofac : Oversampling factor. Hifac : Hifac * "average" Nyquist frequency = highest frequency for which values of the Lomb normalized periodogram will be calculated. n_threads : number of threads to use. Returns: Wk1 : An array of Lomb periodogram frequencies. Wk2 : An array of corresponding values of the Lomb periodogram. Nout : Wk1 & Wk2 dimensions (number of calculated frequencies) Jmax : The array index corresponding to the MAX( Wk2 ). Prob : False Alarm Probability of the largest Periodogram value MACC : Number of interpolation points per 1/4 cycle of highest frequency History: 02/23/2009, v1.0, MF Translation of IDL code (orig. Numerical recipies)
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/extract_features/period_LS_pyfftw.py#L95-L232
null
""" Fast algorithm for spectral analysis of unevenly sampled data ------------------------------------------------------------------------------- Modified by Dae-Won Kim. The code is originally from: http://www.astropython.org/snippet/2010/9/Fast-Lomb-Scargle-algorithm 28/04/2015 Performance improvement is pursued using the pyfftw. Generally, it becomes ~40% faster than the original code (tested on Mac OS X 10.9+). pyfftw also utilizes multiple cores. 04/07/2014 Performance improvement is pursued using Cython, but there is almost no improvement. This is because the most time-consuming part of the routine is numpy.ifft(), which is already highly optimized. Consequently, I decide to stick to the original code. 29/06/2014 Performance is compared with AstroML.time_series.search_frequencies. Given that Delta Scuti stars have very short period, I need a very fine and wide range of frequency grid to calculate. In such case, I observed this routine works better, in terms of both speed and accuracy, mainly because the AstroML function needs 'initial guess' and also 'limit_fractions' well defined. ------------------------------------------------------------------------------- The Lomb-Scargle method performs spectral analysis on unevenly sampled data and is known to be a powerful way to find, and test the significance of, weak periodic signals. The method has previously been thought to be 'slow', requiring of order 10(2)N(2) operations to analyze N data points. We show that Fast Fourier Transforms (FFTs) can be used in a novel way to make the computation of order 10(2)N log N. Despite its use of the FFT, the algorithm is in no way equivalent to conventional FFT periodogram analysis. Keywords: DATA SAMPLING, FAST FOURIER TRANSFORMATIONS, SPECTRUM ANALYSIS, SIGNAL PROCESSING Example: > import numpy > import lomb > x = numpy.arange(10) > y = numpy.sin(x) > fx,fy, nout, jmax, prob = lomb.fasper(x,y, 6., 6.) Reference: Press, W. H. & Rybicki, G. B. 1989 ApJ vol. 338, p. 277-280. Fast algorithm for spectral analysis of unevenly sampled data bib code: 1989ApJ...338..277P """ from numpy import * try: import pyfftw is_pyfftw = True except: from numpy.fft import * is_pyfftw = False def __spread__(y, yy, n, x, m): """ Given an array yy(0:n-1), extrapolate (spread) a value y into m actual array elements that best approximate the "fictional" (i.e., possible noninteger) array element number x. The weights used are coefficients of the Lagrange interpolating polynomial """ nfac = [0, 1, 1, 2, 6, 24, 120, 720, 5040, 40320, 362880] if m > 10. : print('factorial table too small in spread') return ix = long(x) if x == float(ix): yy[ix] = yy[ix]+y else: ilo = long(x - 0.5*float(m) + 1.0) ilo = min(max(ilo, 1), n - m + 1) ihi = ilo + m - 1 nden = nfac[m] fac = x - ilo for j in range(ilo + 1, ihi + 1): fac = fac*(x - j) yy[ihi] = yy[ihi] + y*fac/(nden*(x - ihi)) for j in range(ihi - 1, ilo - 1, -1): nden = (nden/(j + 1 - ilo))*(j - ihi) yy[j] = yy[j] + y*fac/(nden*(x - j)) def fasper(x, y, ofac, hifac, n_threads, MACC=4): """ Given abscissas x (which need not be equally spaced) and ordinates y, and given a desired oversampling factor ofac (a typical value being 4 or larger). this routine creates an array wk1 with a sequence of nout increasing frequencies (not angular frequencies) up to hifac times the "average" Nyquist frequency, and creates an array wk2 with the values of the Lomb normalized periodogram at those frequencies. The arrays x and y are not altered. This routine also returns jmax such that wk2(jmax) is the maximum element in wk2, and prob, an estimate of the significance of that maximum against the hypothesis of random noise. A small value of prob indicates that a significant periodic signal is present. Reference: Press, W. H. & Rybicki, G. B. 1989 ApJ vol. 338, p. 277-280. Fast algorithm for spectral analysis of unevenly sampled data (1989ApJ...338..277P) Arguments: X : Abscissas array, (e.g. an array of times). Y : Ordinates array, (e.g. corresponding counts). Ofac : Oversampling factor. Hifac : Hifac * "average" Nyquist frequency = highest frequency for which values of the Lomb normalized periodogram will be calculated. n_threads : number of threads to use. Returns: Wk1 : An array of Lomb periodogram frequencies. Wk2 : An array of corresponding values of the Lomb periodogram. Nout : Wk1 & Wk2 dimensions (number of calculated frequencies) Jmax : The array index corresponding to the MAX( Wk2 ). Prob : False Alarm Probability of the largest Periodogram value MACC : Number of interpolation points per 1/4 cycle of highest frequency History: 02/23/2009, v1.0, MF Translation of IDL code (orig. Numerical recipies) """ #Check dimensions of input arrays n = long(len(x)) if n != len(y): print('Incompatible arrays.') return #print x, y, hifac, ofac nout = int(0.5*ofac*hifac*n) nfreqt = long(ofac*hifac*n*MACC) #Size the FFT as next power nfreq = 64 # of 2 above nfreqt. while nfreq < nfreqt: nfreq = 2*nfreq ndim = long(2*nfreq) #Compute the mean, variance ave = y.mean() ##sample variance because the divisor is N-1 var = ((y - y.mean())**2).sum()/(len(y) - 1) # and range of the data. xmin = x.min() xmax = x.max() xdif = xmax - xmin #extrapolate the data into the workspaces if is_pyfftw: wk1 = pyfftw.n_byte_align_empty(int(ndim), 16, 'complex') * 0. wk2 = pyfftw.n_byte_align_empty(int(ndim), 16, 'complex') * 0. else: wk1 = zeros(ndim, dtype='complex') wk2 = zeros(ndim, dtype='complex') fac = ndim/(xdif*ofac) fndim = ndim ck = ((x - xmin)*fac) % fndim ckk = (2.0*ck) % fndim for j in range(0, n): __spread__(y[j] - ave, wk1, ndim, ck[j], MACC) __spread__(1.0, wk2, ndim, ckk[j], MACC) #Take the Fast Fourier Transforms. if is_pyfftw: fft_wk1 = pyfftw.builders.ifft(wk1, planner_effort='FFTW_ESTIMATE', threads=n_threads) wk1 = fft_wk1() * len(wk1) fft_wk2 = pyfftw.builders.ifft(wk2, planner_effort='FFTW_ESTIMATE', threads=n_threads) wk2 = fft_wk2() * len(wk2) else: wk1 = ifft(wk1)*len(wk1) wk2 = ifft(wk2)*len(wk1) wk1 = wk1[1:nout + 1] wk2 = wk2[1:nout + 1] rwk1 = wk1.real iwk1 = wk1.imag rwk2 = wk2.real iwk2 = wk2.imag df = 1.0/(xdif*ofac) #Compute the Lomb value for each frequency hypo2 = 2.0*abs(wk2) hc2wt = rwk2/hypo2 hs2wt = iwk2/hypo2 cwt = sqrt(0.5 + hc2wt) swt = sign(hs2wt)*(sqrt(0.5 - hc2wt)) den = 0.5*n + hc2wt*rwk2 + hs2wt*iwk2 cterm = (cwt*rwk1 + swt*iwk1)**2./den sterm = (cwt*iwk1 - swt*rwk1)**2./(n - den) wk1 = df*(arange(nout, dtype='float') + 1.) wk2 = (cterm + sterm)/(2.0*var) pmax = wk2.max() jmax = wk2.argmax() #Significance estimation #expy = exp(-wk2) #effm = 2.0*(nout)/ofac #sig = effm*expy #ind = (sig > 0.01).nonzero() #sig[ind] = 1.0-(1.0-expy[ind])**effm #Estimate significance of largest peak value expy = exp(-pmax) effm = 2.0*(nout)/ofac prob = effm*expy if prob > 0.01: prob = 1.0 - (1.0 - expy)**effm return wk1, wk2, nout, jmax, prob def getSignificance(wk1, wk2, nout, ofac): """ Returns the peak false alarm probabilities Hence the lower is the probability and the more significant is the peak """ expy = exp(-wk2) effm = 2.0*(nout)/ofac sig = effm*expy ind = (sig > 0.01).nonzero() sig[ind] = 1.0 - (1.0 - expy[ind])**effm return sig
dwkim78/upsilon
upsilon/predict/predict.py
predict
python
def predict(rf_model, features): import numpy as np from upsilon.extract_features.feature_set import get_feature_set feature_set = get_feature_set() # Grab only necessary features. cols = [feature for feature in features if feature in feature_set] cols = sorted(cols) filtered_features = [] for i in range(len(cols)): filtered_features.append(features[cols[i]]) filtered_features = np.array(filtered_features).reshape(1, -1) # Classify. classes = rf_model.classes_ # Note that we're classifying a single source, so [0] need tobe added. probabilities = rf_model.predict_proba(filtered_features)[0] # Classification flag. flag = 0 if features['period_SNR'] < 20. or is_period_alias(features['period']): flag = 1 # Return class, probability, and flag. max_index = np.where(probabilities == np.max(probabilities)) return classes[max_index][0], probabilities[max_index][0], flag
Return label and probability estimated. Parameters ---------- rf_model : sklearn.ensemble.RandomForestClassifier The UPSILoN random forests model. features : array_like A list of features estimated by UPSILoN. Returns ------- label : str A predicted label (i.e. class). probability : float Class probability. flag : int Classification flag.
train
https://github.com/dwkim78/upsilon/blob/5f381453f26582ef56e62fb8fed7317ce67861af/upsilon/predict/predict.py#L4-L50
[ "def is_period_alias(period):\n \"\"\"\n Check if a given period is possibly an alias.\n\n Parameters\n ----------\n period : float\n A period to test if it is a possible alias or not.\n\n Returns\n -------\n is_alias : boolean\n True if the given period is in a range of period alias.\n \"\"\"\n\n # Based on the period vs periodSN plot of EROS-2 dataset (Kim+ 2014).\n # Period alias occurs mostly at ~1 and ~30.\n # Check each 1, 2, 3, 4, 5 factors.\n for i in range(1, 6):\n # One-day and one-month alias\n if (.99 / float(i)) < period < (1.004 / float(i)):\n return True\n if (1.03 / float(i)) < period < (1.04 / float(i)):\n return True\n if (29.2 / float(i)) < period < (29.9 / float(i)):\n return True\n\n # From candidates from the two fields 01, 08.\n # All of them are close to one day (or sidereal) alias.\n if (0.96465 / float(i)) < period < (0.96485 / float(i)):\n return True\n if (0.96725 / float(i)) < period < (0.96745 / float(i)):\n return True\n if (0.98190 / float(i)) < period < (0.98230 / float(i)):\n return True\n if (1.01034 / float(i)) < period < (1.01076 / float(i)):\n return True\n if (1.01568 / float(i)) < period < (1.01604 / float(i)):\n return True\n if (1.01718 / float(i)) < period < (1.01742 / float(i)):\n return True\n\n # From the all candidates from the entire LMC fields.\n # Some of these could be overlapped with the above cuts.\n if (0.50776 / float(i)) < period < (0.50861 / float(i)):\n return True\n if (0.96434 / float(i)) < period < (0.9652 / float(i)):\n return True\n if (0.96688 / float(i)) < period < (0.96731 / float(i)):\n return True\n if (1.0722 / float(i)) < period < (1.0729 / float(i)):\n return True\n if (27.1 / float(i)) < period < (27.5 / float(i)):\n return True\n\n # Not in the range of any alias.\n return False\n", "def get_feature_set():\n \"\"\"\n Return a list of features' names.\n\n Features' name that are used to train a model and predict a class.\n Sorted by the names.\n\n Returns\n -------\n feature_names : list\n A list of features' names.\n \"\"\"\n\n features = ['amplitude', 'hl_amp_ratio', 'kurtosis', 'period',\n 'phase_cusum', 'phase_eta', 'phi21', 'phi31', 'quartile31',\n 'r21', 'r31', 'shapiro_w', 'skewness', 'slope_per10',\n 'slope_per90', 'stetson_k']\n features.sort()\n\n return features\n" ]
from upsilon.extract_features.is_period_alias import is_period_alias
unbit/sftpclone
sftpclone/sftpclone.py
configure_logging
python
def configure_logging(level=logging.DEBUG): if level == logging.DEBUG: # For debugging purposes, log from everyone! logging.basicConfig( level=logging.DEBUG, format='%(asctime)s - %(levelname)s - %(message)s' ) return logging logger = logging.getLogger(__name__) logger.setLevel(level) formatter = logging.Formatter('%(asctime)s - %(levelname)s - %(message)s') ch = logging.StreamHandler() ch.setLevel(level) ch.setFormatter(formatter) logger.addHandler(ch) return logger
Configure the module logging engine.
train
https://github.com/unbit/sftpclone/blob/1cc89478e680fc4e0d12b1a15b5bafd0390d05da/sftpclone/sftpclone.py#L33-L50
null
#!/usr/bin/env python # coding=utf-8 # Python 2.7 backward compatibility from __future__ import print_function from __future__ import unicode_literals from __future__ import absolute_import import paramiko import paramiko.py3compat import os import os.path import sys import errno from stat import S_ISDIR, S_ISLNK, S_ISREG, S_IMODE, S_IFMT import argparse import logging from getpass import getuser, getpass import glob import socket """SFTPClone: sync local and remote directories.""" logger = None try: # Not available in Python 2.x FileNotFoundError except NameError: FileNotFoundError = IOError def path_join(*args): """ Wrapper around `os.path.join`. Makes sure to join paths of the same type (bytes). """ args = (paramiko.py3compat.u(arg) for arg in args) return os.path.join(*args) def parse_username_password_hostname(remote_url): """ Parse a command line string and return username, password, remote hostname and remote path. :param remote_url: A command line string. :return: A tuple, containing username, password, remote hostname and remote path. """ assert remote_url assert ':' in remote_url if '@' in remote_url: username, hostname = remote_url.rsplit('@', 1) else: username, hostname = None, remote_url hostname, remote_path = hostname.split(':', 1) password = None if username and ':' in username: username, password = username.split(':', 1) assert hostname assert remote_path return username, password, hostname, remote_path def get_ssh_agent_keys(logger): """ Ask the SSH agent for a list of keys, and return it. :return: A reference to the SSH agent and a list of keys. """ agent, agent_keys = None, None try: agent = paramiko.agent.Agent() _agent_keys = agent.get_keys() if not _agent_keys: agent.close() logger.error( "SSH agent didn't provide any valid key. Trying to continue..." ) else: agent_keys = tuple(k for k in _agent_keys) except paramiko.SSHException: if agent: agent.close() agent = None logger.error("SSH agent speaks a non-compatible protocol. Ignoring it.") finally: return agent, agent_keys class SFTPClone(object): """The SFTPClone class.""" def __init__(self, local_path, remote_url, identity_files=None, port=None, fix_symlinks=False, ssh_config_path=None, ssh_agent=False, exclude_file=None, known_hosts_path=None, delete=True, allow_unknown=False, create_remote_directory=False, ): """Init the needed parameters and the SFTPClient.""" self.local_path = os.path.realpath(os.path.expanduser(local_path)) self.logger = logger or configure_logging() self.create_remote_directory = create_remote_directory if not os.path.exists(self.local_path): self.logger.error("Local path MUST exist. Exiting.") sys.exit(1) if exclude_file: with open(exclude_file) as f: # As in rsync's exclude from, ignore lines with leading ; and # # and treat each path as relative (thus by removing the leading # /) exclude_list = [ line.rstrip().lstrip("/") for line in f if not line.startswith((";", "#")) ] # actually, is a set of excluded files self.exclude_list = { g for pattern in exclude_list for g in glob.glob(path_join(self.local_path, pattern)) } else: self.exclude_list = set() username, password, hostname, self.remote_path = parse_username_password_hostname(remote_url) identity_files = identity_files or [] proxy_command = None if ssh_config_path: try: with open(os.path.expanduser(ssh_config_path)) as c_file: ssh_config = paramiko.SSHConfig() ssh_config.parse(c_file) c = ssh_config.lookup(hostname) hostname = c.get("hostname", hostname) username = c.get("user", username) port = int(c.get("port", port)) identity_files = c.get("identityfile", identity_files) proxy_command = c.get("proxycommand") except Exception as e: # it could be safe to continue anyway, # because parameters could have been manually specified self.logger.error( "Error while parsing ssh_config file: %s. Trying to continue anyway...", e ) # Set default values if not username: username = getuser() # defaults to current user port = port or 22 allow_unknown = allow_unknown or False self.chown = False self.fix_symlinks = fix_symlinks or False self.delete = delete if delete is not None else True if ssh_agent: agent, agent_keys = get_ssh_agent_keys(self.logger) else: agent, agent_keys = None, None if not identity_files and not password and not agent_keys: self.logger.error( "You need to specify either a password, an identity or to enable the ssh-agent support." ) sys.exit(1) # only root can change file owner if username == 'root': self.chown = True sock = (hostname, port) if proxy_command is not None: sock = paramiko.proxy.ProxyCommand(proxy_command) try: transport = paramiko.Transport(sock) except socket.gaierror: self.logger.error( "Hostname not known. Are you sure you inserted it correctly?") sys.exit(1) try: ssh_host = hostname if port == 22 else "[{}]:{}".format(hostname, port) known_hosts = None """ Before starting the transport session, we have to configure it. Specifically, we need to configure the preferred PK algorithm. If the system already knows a public key of a specific kind for a remote host, we have to peek its type as the preferred one. """ if known_hosts_path: known_hosts = paramiko.HostKeys() known_hosts_path = os.path.realpath( os.path.expanduser(known_hosts_path)) try: known_hosts.load(known_hosts_path) except IOError: self.logger.error( "Error while loading known hosts file at {}. Exiting...".format( known_hosts_path) ) sys.exit(1) known_keys = known_hosts.lookup(ssh_host) if known_keys is not None: # one or more keys are already known # set their type as preferred transport.get_security_options().key_types = \ tuple(known_keys.keys()) transport.start_client() if not known_hosts: self.logger.warning("Security warning: skipping known hosts check...") else: pubk = transport.get_remote_server_key() if ssh_host in known_hosts.keys(): if not known_hosts.check(ssh_host, pubk): self.logger.error( "Security warning: " "remote key fingerprint {} for hostname " "{} didn't match the one in known_hosts {}. " "Exiting...".format( pubk.get_base64(), ssh_host, known_hosts.lookup(hostname), ) ) sys.exit(1) elif not allow_unknown: prompt = ("The authenticity of host '{}' can't be established.\n" "{} key is {}.\n" "Are you sure you want to continue connecting? [y/n] ").format( ssh_host, pubk.get_name(), pubk.get_base64()) try: # Renamed to `input` in Python 3.x response = raw_input(prompt) except NameError: response = input(prompt) # Note: we do not modify the user's known_hosts file if not (response == "y" or response == "yes"): self.logger.error( "Host authentication failed." ) sys.exit(1) def perform_key_auth(pkey): try: transport.auth_publickey( username=username, key=pkey ) return True except paramiko.SSHException: self.logger.warning( "Authentication with identity {}... failed".format(pkey.get_base64()[:10]) ) return False if password: # Password auth, if specified. transport.auth_password( username=username, password=password ) elif agent_keys: # SSH agent keys have higher priority for pkey in agent_keys: if perform_key_auth(pkey): break # Authentication worked. else: # None of the keys worked. raise paramiko.SSHException elif identity_files: # Then follow identity file (specified from CL or ssh_config) # Try identity files one by one, until one works for key_path in identity_files: key_path = os.path.expanduser(key_path) try: key = paramiko.RSAKey.from_private_key_file(key_path) except paramiko.PasswordRequiredException: pk_password = getpass( "It seems that your identity from '{}' is encrypted. " "Please enter your password: ".format(key_path) ) try: key = paramiko.RSAKey.from_private_key_file(key_path, pk_password) except paramiko.SSHException: self.logger.error( "Incorrect passphrase. Cannot decode private key from '{}'.".format(key_path) ) continue except IOError or paramiko.SSHException: self.logger.error( "Something went wrong while opening '{}'. Skipping it.".format(key_path) ) continue if perform_key_auth(key): break # Authentication worked. else: # None of the keys worked. raise paramiko.SSHException else: # No authentication method specified, we shouldn't arrive here. assert False except paramiko.SSHException: self.logger.error( "None of the provided authentication methods worked. Exiting." ) transport.close() sys.exit(1) finally: if agent: agent.close() self.sftp = paramiko.SFTPClient.from_transport(transport) if self.remote_path.startswith("~"): # nasty hack to let getcwd work without changing dir! self.sftp.chdir('.') self.remote_path = self.remote_path.replace( "~", self.sftp.getcwd()) # home is the initial sftp dir @staticmethod def _file_need_upload(l_st, r_st): return True if \ l_st.st_size != r_st.st_size or int(l_st.st_mtime) != r_st.st_mtime \ else False @staticmethod def _must_be_deleted(local_path, r_st): """Return True if the remote correspondent of local_path has to be deleted. i.e. if it doesn't exists locally or if it has a different type from the remote one.""" # if the file doesn't exists if not os.path.lexists(local_path): return True # or if the file type is different l_st = os.lstat(local_path) if S_IFMT(r_st.st_mode) != S_IFMT(l_st.st_mode): return True return False def _match_modes(self, remote_path, l_st): """Match mod, utime and uid/gid with locals one.""" self.sftp.chmod(remote_path, S_IMODE(l_st.st_mode)) self.sftp.utime(remote_path, (l_st.st_atime, l_st.st_mtime)) if self.chown: self.sftp.chown(remote_path, l_st.st_uid, l_st.st_gid) def file_upload(self, local_path, remote_path, l_st): """Upload local_path to remote_path and set permission and mtime.""" self.sftp.put(local_path, remote_path) self._match_modes(remote_path, l_st) def remote_delete(self, remote_path, r_st): """Remove the remote directory node.""" # If it's a directory, then delete content and directory if S_ISDIR(r_st.st_mode): for item in self.sftp.listdir_attr(remote_path): full_path = path_join(remote_path, item.filename) self.remote_delete(full_path, item) self.sftp.rmdir(remote_path) # Or simply delete files else: try: self.sftp.remove(remote_path) except FileNotFoundError as e: self.logger.error( "error while removing {}. trace: {}".format(remote_path, e) ) def check_for_deletion(self, relative_path=None): """Traverse the entire remote_path tree. Find files/directories that need to be deleted, not being present in the local folder. """ if not relative_path: relative_path = str() # root of shared directory tree remote_path = path_join(self.remote_path, relative_path) local_path = path_join(self.local_path, relative_path) for remote_st in self.sftp.listdir_attr(remote_path): r_lstat = self.sftp.lstat(path_join(remote_path, remote_st.filename)) inner_remote_path = path_join(remote_path, remote_st.filename) inner_local_path = path_join(local_path, remote_st.filename) # check if remote_st is a symlink # otherwise could delete file outside shared directory if S_ISLNK(r_lstat.st_mode): if self._must_be_deleted(inner_local_path, r_lstat): self.remote_delete(inner_remote_path, r_lstat) continue if self._must_be_deleted(inner_local_path, remote_st): self.remote_delete(inner_remote_path, remote_st) elif S_ISDIR(remote_st.st_mode): self.check_for_deletion( path_join(relative_path, remote_st.filename) ) def create_update_symlink(self, link_destination, remote_path): """Create a new link pointing to link_destination in remote_path position.""" try: # if there's anything, delete it self.sftp.remove(remote_path) except IOError: # that's fine, nothing exists there! pass finally: # and recreate the link try: self.sftp.symlink(link_destination, remote_path) except OSError as e: # Sometimes, if links are "too" different, symlink fails. # Sadly, nothing we can do about it. self.logger.error("error while symlinking {} to {}: {}".format( remote_path, link_destination, e)) def node_check_for_upload_create(self, relative_path, f): """Check if the given directory tree node has to be uploaded/created on the remote folder.""" if not relative_path: # we're at the root of the shared directory tree relative_path = str() # the (absolute) local address of f. local_path = path_join(self.local_path, relative_path, f) try: l_st = os.lstat(local_path) except OSError as e: """A little background here. Sometimes, in big clusters configurations (mail, etc.), files could disappear or be moved, suddenly. There's nothing to do about it, system should be stopped before doing backups. Anyway, we log it, and skip it. """ self.logger.error("error while checking {}: {}".format(relative_path, e)) return if local_path in self.exclude_list: self.logger.info("Skipping excluded file %s.", local_path) return # the (absolute) remote address of f. remote_path = path_join(self.remote_path, relative_path, f) # First case: f is a directory if S_ISDIR(l_st.st_mode): # we check if the folder exists on the remote side # it has to be a folder, otherwise it would have already been # deleted try: self.sftp.stat(remote_path) except IOError: # it doesn't exist yet on remote side self.sftp.mkdir(remote_path) self._match_modes(remote_path, l_st) # now, we should traverse f too (recursion magic!) self.check_for_upload_create(path_join(relative_path, f)) # Second case: f is a symbolic link elif S_ISLNK(l_st.st_mode): # read the local link local_link = os.readlink(local_path) absolute_local_link = os.path.realpath(local_link) # is it absolute? is_absolute = local_link.startswith("/") # and does it point inside the shared directory? # add trailing slash (security) trailing_local_path = path_join(self.local_path, '') relpath = os.path.commonprefix( [absolute_local_link, trailing_local_path] ) == trailing_local_path if relpath: relative_link = absolute_local_link[len(trailing_local_path):] else: relative_link = None """ # Refactor them all, be efficient! # Case A: absolute link pointing outside shared directory # (we can only update the remote part) if is_absolute and not relpath: self.create_update_symlink(local_link, remote_path) # Case B: absolute link pointing inside shared directory # (we can leave it as it is or fix the prefix to match the one of the remote server) elif is_absolute and relpath: if self.fix_symlinks: self.create_update_symlink( join( self.remote_path, relative_link, ), remote_path ) else: self.create_update_symlink(local_link, remote_path) # Case C: relative link pointing outside shared directory # (all we can do is try to make the link anyway) elif not is_absolute and not relpath: self.create_update_symlink(local_link, remote_path) # Case D: relative link pointing inside shared directory # (we preserve the relativity and link it!) elif not is_absolute and relpath: self.create_update_symlink(local_link, remote_path) """ if is_absolute and relpath: if self.fix_symlinks: self.create_update_symlink( path_join( self.remote_path, relative_link, ), remote_path ) else: self.create_update_symlink(local_link, remote_path) # Third case: regular file elif S_ISREG(l_st.st_mode): try: r_st = self.sftp.lstat(remote_path) if self._file_need_upload(l_st, r_st): self.file_upload(local_path, remote_path, l_st) except IOError as e: if e.errno == errno.ENOENT: self.file_upload(local_path, remote_path, l_st) # Anything else. else: self.logger.warning("Skipping unsupported file %s.", local_path) def check_for_upload_create(self, relative_path=None): """Traverse the relative_path tree and check for files that need to be uploaded/created. Relativity here refers to the shared directory tree.""" for f in os.listdir( path_join( self.local_path, relative_path) if relative_path else self.local_path ): self.node_check_for_upload_create(relative_path, f) def run(self): """Run the sync. Confront the local and the remote directories and perform the needed changes.""" # Check if remote path is present try: self.sftp.stat(self.remote_path) except FileNotFoundError as e: if self.create_remote_directory: self.sftp.mkdir(self.remote_path) self.logger.info( "Created missing remote dir: '" + self.remote_path + "'") else: self.logger.error( "Remote folder does not exists. " "Add '-r' to create it if missing.") sys.exit(1) try: if self.delete: # First check for items to be removed self.check_for_deletion() # Now scan local for items to upload/create self.check_for_upload_create() except FileNotFoundError: # If this happens, probably the remote folder doesn't exist. self.logger.error( "Error while opening remote folder. Are you sure it does exist?") sys.exit(1) def create_parser(): """Create the CLI argument parser.""" parser = argparse.ArgumentParser( description='Sync a local and a remote folder through SFTP.' ) parser.add_argument( "path", type=str, metavar="local-path", help="the path of the local folder", ) parser.add_argument( "remote", type=str, metavar="user[:password]@hostname:remote-path", help="the ssh-url ([user[:password]@]hostname:remote-path) of the remote folder. " "The hostname can be specified as a ssh_config's hostname too. " "Every missing information will be gathered from there", ) parser.add_argument( "-k", "--key", metavar="identity-path", action="append", help="private key identity path (defaults to ~/.ssh/id_rsa)" ) parser.add_argument( "-l", "--logging", choices=['CRITICAL', 'ERROR', 'WARNING', 'INFO', 'DEBUG', 'NOTSET'], default='ERROR', help="set logging level" ) parser.add_argument( "-p", "--port", default=22, type=int, help="SSH remote port (defaults to 22)" ) parser.add_argument( "-f", "--fix-symlinks", action="store_true", help="fix symbolic links on remote side" ) parser.add_argument( "-a", "--ssh-agent", action="store_true", help="enable ssh-agent support" ) parser.add_argument( "-c", "--ssh-config", metavar="ssh_config path", default="~/.ssh/config", type=str, help="path to the ssh-configuration file (default to ~/.ssh/config)" ) parser.add_argument( "-n", "--known-hosts", metavar="known_hosts path", default="~/.ssh/known_hosts", type=str, help="path to the openSSH known_hosts file" ) parser.add_argument( "-d", "--disable-known-hosts", action="store_true", help="disable known_hosts fingerprint checking (security warning!)" ) parser.add_argument( "-e", "--exclude-from", metavar="exclude-from-file-path", type=str, help="exclude files matching pattern in exclude-from-file-path" ) parser.add_argument( "-t", "--do-not-delete", action="store_true", help="do not delete remote files missing from local folder" ) parser.add_argument( "-o", "--allow-unknown", action="store_true", help="allow connection to unknown hosts" ) parser.add_argument( "-r", "--create-remote-directory", action="store_true", help="Create remote base directory if missing on remote" ) return parser def main(args=None): """The main.""" parser = create_parser() args = vars(parser.parse_args(args)) log_mapping = { 'CRITICAL': logging.CRITICAL, 'ERROR': logging.ERROR, 'WARNING': logging.WARNING, 'INFO': logging.INFO, 'DEBUG': logging.DEBUG, 'NOTSET': logging.NOTSET, } log_level = log_mapping[args['logging']] del(args['logging']) global logger logger = configure_logging(log_level) args_mapping = { "path": "local_path", "remote": "remote_url", "ssh_config": "ssh_config_path", "exclude_from": "exclude_file", "known_hosts": "known_hosts_path", "do_not_delete": "delete", "key": "identity_files", } kwargs = { # convert the argument names to class constructor parameters args_mapping[k]: v for k, v in args.items() if v and k in args_mapping } kwargs.update({ k: v for k, v in args.items() if v and k not in args_mapping }) # Special case: disable known_hosts check if args['disable_known_hosts']: kwargs['known_hosts_path'] = None del(kwargs['disable_known_hosts']) # Toggle `do_not_delete` flag if "delete" in kwargs: kwargs["delete"] = not kwargs["delete"] # Manually set the default identity file. kwargs["identity_files"] = kwargs.get("identity_files", None) or ["~/.ssh/id_rsa"] sync = SFTPClone( **kwargs ) sync.run() if __name__ == '__main__': main()
unbit/sftpclone
sftpclone/sftpclone.py
path_join
python
def path_join(*args): args = (paramiko.py3compat.u(arg) for arg in args) return os.path.join(*args)
Wrapper around `os.path.join`. Makes sure to join paths of the same type (bytes).
train
https://github.com/unbit/sftpclone/blob/1cc89478e680fc4e0d12b1a15b5bafd0390d05da/sftpclone/sftpclone.py#L53-L59
null
#!/usr/bin/env python # coding=utf-8 # Python 2.7 backward compatibility from __future__ import print_function from __future__ import unicode_literals from __future__ import absolute_import import paramiko import paramiko.py3compat import os import os.path import sys import errno from stat import S_ISDIR, S_ISLNK, S_ISREG, S_IMODE, S_IFMT import argparse import logging from getpass import getuser, getpass import glob import socket """SFTPClone: sync local and remote directories.""" logger = None try: # Not available in Python 2.x FileNotFoundError except NameError: FileNotFoundError = IOError def configure_logging(level=logging.DEBUG): """Configure the module logging engine.""" if level == logging.DEBUG: # For debugging purposes, log from everyone! logging.basicConfig( level=logging.DEBUG, format='%(asctime)s - %(levelname)s - %(message)s' ) return logging logger = logging.getLogger(__name__) logger.setLevel(level) formatter = logging.Formatter('%(asctime)s - %(levelname)s - %(message)s') ch = logging.StreamHandler() ch.setLevel(level) ch.setFormatter(formatter) logger.addHandler(ch) return logger def parse_username_password_hostname(remote_url): """ Parse a command line string and return username, password, remote hostname and remote path. :param remote_url: A command line string. :return: A tuple, containing username, password, remote hostname and remote path. """ assert remote_url assert ':' in remote_url if '@' in remote_url: username, hostname = remote_url.rsplit('@', 1) else: username, hostname = None, remote_url hostname, remote_path = hostname.split(':', 1) password = None if username and ':' in username: username, password = username.split(':', 1) assert hostname assert remote_path return username, password, hostname, remote_path def get_ssh_agent_keys(logger): """ Ask the SSH agent for a list of keys, and return it. :return: A reference to the SSH agent and a list of keys. """ agent, agent_keys = None, None try: agent = paramiko.agent.Agent() _agent_keys = agent.get_keys() if not _agent_keys: agent.close() logger.error( "SSH agent didn't provide any valid key. Trying to continue..." ) else: agent_keys = tuple(k for k in _agent_keys) except paramiko.SSHException: if agent: agent.close() agent = None logger.error("SSH agent speaks a non-compatible protocol. Ignoring it.") finally: return agent, agent_keys class SFTPClone(object): """The SFTPClone class.""" def __init__(self, local_path, remote_url, identity_files=None, port=None, fix_symlinks=False, ssh_config_path=None, ssh_agent=False, exclude_file=None, known_hosts_path=None, delete=True, allow_unknown=False, create_remote_directory=False, ): """Init the needed parameters and the SFTPClient.""" self.local_path = os.path.realpath(os.path.expanduser(local_path)) self.logger = logger or configure_logging() self.create_remote_directory = create_remote_directory if not os.path.exists(self.local_path): self.logger.error("Local path MUST exist. Exiting.") sys.exit(1) if exclude_file: with open(exclude_file) as f: # As in rsync's exclude from, ignore lines with leading ; and # # and treat each path as relative (thus by removing the leading # /) exclude_list = [ line.rstrip().lstrip("/") for line in f if not line.startswith((";", "#")) ] # actually, is a set of excluded files self.exclude_list = { g for pattern in exclude_list for g in glob.glob(path_join(self.local_path, pattern)) } else: self.exclude_list = set() username, password, hostname, self.remote_path = parse_username_password_hostname(remote_url) identity_files = identity_files or [] proxy_command = None if ssh_config_path: try: with open(os.path.expanduser(ssh_config_path)) as c_file: ssh_config = paramiko.SSHConfig() ssh_config.parse(c_file) c = ssh_config.lookup(hostname) hostname = c.get("hostname", hostname) username = c.get("user", username) port = int(c.get("port", port)) identity_files = c.get("identityfile", identity_files) proxy_command = c.get("proxycommand") except Exception as e: # it could be safe to continue anyway, # because parameters could have been manually specified self.logger.error( "Error while parsing ssh_config file: %s. Trying to continue anyway...", e ) # Set default values if not username: username = getuser() # defaults to current user port = port or 22 allow_unknown = allow_unknown or False self.chown = False self.fix_symlinks = fix_symlinks or False self.delete = delete if delete is not None else True if ssh_agent: agent, agent_keys = get_ssh_agent_keys(self.logger) else: agent, agent_keys = None, None if not identity_files and not password and not agent_keys: self.logger.error( "You need to specify either a password, an identity or to enable the ssh-agent support." ) sys.exit(1) # only root can change file owner if username == 'root': self.chown = True sock = (hostname, port) if proxy_command is not None: sock = paramiko.proxy.ProxyCommand(proxy_command) try: transport = paramiko.Transport(sock) except socket.gaierror: self.logger.error( "Hostname not known. Are you sure you inserted it correctly?") sys.exit(1) try: ssh_host = hostname if port == 22 else "[{}]:{}".format(hostname, port) known_hosts = None """ Before starting the transport session, we have to configure it. Specifically, we need to configure the preferred PK algorithm. If the system already knows a public key of a specific kind for a remote host, we have to peek its type as the preferred one. """ if known_hosts_path: known_hosts = paramiko.HostKeys() known_hosts_path = os.path.realpath( os.path.expanduser(known_hosts_path)) try: known_hosts.load(known_hosts_path) except IOError: self.logger.error( "Error while loading known hosts file at {}. Exiting...".format( known_hosts_path) ) sys.exit(1) known_keys = known_hosts.lookup(ssh_host) if known_keys is not None: # one or more keys are already known # set their type as preferred transport.get_security_options().key_types = \ tuple(known_keys.keys()) transport.start_client() if not known_hosts: self.logger.warning("Security warning: skipping known hosts check...") else: pubk = transport.get_remote_server_key() if ssh_host in known_hosts.keys(): if not known_hosts.check(ssh_host, pubk): self.logger.error( "Security warning: " "remote key fingerprint {} for hostname " "{} didn't match the one in known_hosts {}. " "Exiting...".format( pubk.get_base64(), ssh_host, known_hosts.lookup(hostname), ) ) sys.exit(1) elif not allow_unknown: prompt = ("The authenticity of host '{}' can't be established.\n" "{} key is {}.\n" "Are you sure you want to continue connecting? [y/n] ").format( ssh_host, pubk.get_name(), pubk.get_base64()) try: # Renamed to `input` in Python 3.x response = raw_input(prompt) except NameError: response = input(prompt) # Note: we do not modify the user's known_hosts file if not (response == "y" or response == "yes"): self.logger.error( "Host authentication failed." ) sys.exit(1) def perform_key_auth(pkey): try: transport.auth_publickey( username=username, key=pkey ) return True except paramiko.SSHException: self.logger.warning( "Authentication with identity {}... failed".format(pkey.get_base64()[:10]) ) return False if password: # Password auth, if specified. transport.auth_password( username=username, password=password ) elif agent_keys: # SSH agent keys have higher priority for pkey in agent_keys: if perform_key_auth(pkey): break # Authentication worked. else: # None of the keys worked. raise paramiko.SSHException elif identity_files: # Then follow identity file (specified from CL or ssh_config) # Try identity files one by one, until one works for key_path in identity_files: key_path = os.path.expanduser(key_path) try: key = paramiko.RSAKey.from_private_key_file(key_path) except paramiko.PasswordRequiredException: pk_password = getpass( "It seems that your identity from '{}' is encrypted. " "Please enter your password: ".format(key_path) ) try: key = paramiko.RSAKey.from_private_key_file(key_path, pk_password) except paramiko.SSHException: self.logger.error( "Incorrect passphrase. Cannot decode private key from '{}'.".format(key_path) ) continue except IOError or paramiko.SSHException: self.logger.error( "Something went wrong while opening '{}'. Skipping it.".format(key_path) ) continue if perform_key_auth(key): break # Authentication worked. else: # None of the keys worked. raise paramiko.SSHException else: # No authentication method specified, we shouldn't arrive here. assert False except paramiko.SSHException: self.logger.error( "None of the provided authentication methods worked. Exiting." ) transport.close() sys.exit(1) finally: if agent: agent.close() self.sftp = paramiko.SFTPClient.from_transport(transport) if self.remote_path.startswith("~"): # nasty hack to let getcwd work without changing dir! self.sftp.chdir('.') self.remote_path = self.remote_path.replace( "~", self.sftp.getcwd()) # home is the initial sftp dir @staticmethod def _file_need_upload(l_st, r_st): return True if \ l_st.st_size != r_st.st_size or int(l_st.st_mtime) != r_st.st_mtime \ else False @staticmethod def _must_be_deleted(local_path, r_st): """Return True if the remote correspondent of local_path has to be deleted. i.e. if it doesn't exists locally or if it has a different type from the remote one.""" # if the file doesn't exists if not os.path.lexists(local_path): return True # or if the file type is different l_st = os.lstat(local_path) if S_IFMT(r_st.st_mode) != S_IFMT(l_st.st_mode): return True return False def _match_modes(self, remote_path, l_st): """Match mod, utime and uid/gid with locals one.""" self.sftp.chmod(remote_path, S_IMODE(l_st.st_mode)) self.sftp.utime(remote_path, (l_st.st_atime, l_st.st_mtime)) if self.chown: self.sftp.chown(remote_path, l_st.st_uid, l_st.st_gid) def file_upload(self, local_path, remote_path, l_st): """Upload local_path to remote_path and set permission and mtime.""" self.sftp.put(local_path, remote_path) self._match_modes(remote_path, l_st) def remote_delete(self, remote_path, r_st): """Remove the remote directory node.""" # If it's a directory, then delete content and directory if S_ISDIR(r_st.st_mode): for item in self.sftp.listdir_attr(remote_path): full_path = path_join(remote_path, item.filename) self.remote_delete(full_path, item) self.sftp.rmdir(remote_path) # Or simply delete files else: try: self.sftp.remove(remote_path) except FileNotFoundError as e: self.logger.error( "error while removing {}. trace: {}".format(remote_path, e) ) def check_for_deletion(self, relative_path=None): """Traverse the entire remote_path tree. Find files/directories that need to be deleted, not being present in the local folder. """ if not relative_path: relative_path = str() # root of shared directory tree remote_path = path_join(self.remote_path, relative_path) local_path = path_join(self.local_path, relative_path) for remote_st in self.sftp.listdir_attr(remote_path): r_lstat = self.sftp.lstat(path_join(remote_path, remote_st.filename)) inner_remote_path = path_join(remote_path, remote_st.filename) inner_local_path = path_join(local_path, remote_st.filename) # check if remote_st is a symlink # otherwise could delete file outside shared directory if S_ISLNK(r_lstat.st_mode): if self._must_be_deleted(inner_local_path, r_lstat): self.remote_delete(inner_remote_path, r_lstat) continue if self._must_be_deleted(inner_local_path, remote_st): self.remote_delete(inner_remote_path, remote_st) elif S_ISDIR(remote_st.st_mode): self.check_for_deletion( path_join(relative_path, remote_st.filename) ) def create_update_symlink(self, link_destination, remote_path): """Create a new link pointing to link_destination in remote_path position.""" try: # if there's anything, delete it self.sftp.remove(remote_path) except IOError: # that's fine, nothing exists there! pass finally: # and recreate the link try: self.sftp.symlink(link_destination, remote_path) except OSError as e: # Sometimes, if links are "too" different, symlink fails. # Sadly, nothing we can do about it. self.logger.error("error while symlinking {} to {}: {}".format( remote_path, link_destination, e)) def node_check_for_upload_create(self, relative_path, f): """Check if the given directory tree node has to be uploaded/created on the remote folder.""" if not relative_path: # we're at the root of the shared directory tree relative_path = str() # the (absolute) local address of f. local_path = path_join(self.local_path, relative_path, f) try: l_st = os.lstat(local_path) except OSError as e: """A little background here. Sometimes, in big clusters configurations (mail, etc.), files could disappear or be moved, suddenly. There's nothing to do about it, system should be stopped before doing backups. Anyway, we log it, and skip it. """ self.logger.error("error while checking {}: {}".format(relative_path, e)) return if local_path in self.exclude_list: self.logger.info("Skipping excluded file %s.", local_path) return # the (absolute) remote address of f. remote_path = path_join(self.remote_path, relative_path, f) # First case: f is a directory if S_ISDIR(l_st.st_mode): # we check if the folder exists on the remote side # it has to be a folder, otherwise it would have already been # deleted try: self.sftp.stat(remote_path) except IOError: # it doesn't exist yet on remote side self.sftp.mkdir(remote_path) self._match_modes(remote_path, l_st) # now, we should traverse f too (recursion magic!) self.check_for_upload_create(path_join(relative_path, f)) # Second case: f is a symbolic link elif S_ISLNK(l_st.st_mode): # read the local link local_link = os.readlink(local_path) absolute_local_link = os.path.realpath(local_link) # is it absolute? is_absolute = local_link.startswith("/") # and does it point inside the shared directory? # add trailing slash (security) trailing_local_path = path_join(self.local_path, '') relpath = os.path.commonprefix( [absolute_local_link, trailing_local_path] ) == trailing_local_path if relpath: relative_link = absolute_local_link[len(trailing_local_path):] else: relative_link = None """ # Refactor them all, be efficient! # Case A: absolute link pointing outside shared directory # (we can only update the remote part) if is_absolute and not relpath: self.create_update_symlink(local_link, remote_path) # Case B: absolute link pointing inside shared directory # (we can leave it as it is or fix the prefix to match the one of the remote server) elif is_absolute and relpath: if self.fix_symlinks: self.create_update_symlink( join( self.remote_path, relative_link, ), remote_path ) else: self.create_update_symlink(local_link, remote_path) # Case C: relative link pointing outside shared directory # (all we can do is try to make the link anyway) elif not is_absolute and not relpath: self.create_update_symlink(local_link, remote_path) # Case D: relative link pointing inside shared directory # (we preserve the relativity and link it!) elif not is_absolute and relpath: self.create_update_symlink(local_link, remote_path) """ if is_absolute and relpath: if self.fix_symlinks: self.create_update_symlink( path_join( self.remote_path, relative_link, ), remote_path ) else: self.create_update_symlink(local_link, remote_path) # Third case: regular file elif S_ISREG(l_st.st_mode): try: r_st = self.sftp.lstat(remote_path) if self._file_need_upload(l_st, r_st): self.file_upload(local_path, remote_path, l_st) except IOError as e: if e.errno == errno.ENOENT: self.file_upload(local_path, remote_path, l_st) # Anything else. else: self.logger.warning("Skipping unsupported file %s.", local_path) def check_for_upload_create(self, relative_path=None): """Traverse the relative_path tree and check for files that need to be uploaded/created. Relativity here refers to the shared directory tree.""" for f in os.listdir( path_join( self.local_path, relative_path) if relative_path else self.local_path ): self.node_check_for_upload_create(relative_path, f) def run(self): """Run the sync. Confront the local and the remote directories and perform the needed changes.""" # Check if remote path is present try: self.sftp.stat(self.remote_path) except FileNotFoundError as e: if self.create_remote_directory: self.sftp.mkdir(self.remote_path) self.logger.info( "Created missing remote dir: '" + self.remote_path + "'") else: self.logger.error( "Remote folder does not exists. " "Add '-r' to create it if missing.") sys.exit(1) try: if self.delete: # First check for items to be removed self.check_for_deletion() # Now scan local for items to upload/create self.check_for_upload_create() except FileNotFoundError: # If this happens, probably the remote folder doesn't exist. self.logger.error( "Error while opening remote folder. Are you sure it does exist?") sys.exit(1) def create_parser(): """Create the CLI argument parser.""" parser = argparse.ArgumentParser( description='Sync a local and a remote folder through SFTP.' ) parser.add_argument( "path", type=str, metavar="local-path", help="the path of the local folder", ) parser.add_argument( "remote", type=str, metavar="user[:password]@hostname:remote-path", help="the ssh-url ([user[:password]@]hostname:remote-path) of the remote folder. " "The hostname can be specified as a ssh_config's hostname too. " "Every missing information will be gathered from there", ) parser.add_argument( "-k", "--key", metavar="identity-path", action="append", help="private key identity path (defaults to ~/.ssh/id_rsa)" ) parser.add_argument( "-l", "--logging", choices=['CRITICAL', 'ERROR', 'WARNING', 'INFO', 'DEBUG', 'NOTSET'], default='ERROR', help="set logging level" ) parser.add_argument( "-p", "--port", default=22, type=int, help="SSH remote port (defaults to 22)" ) parser.add_argument( "-f", "--fix-symlinks", action="store_true", help="fix symbolic links on remote side" ) parser.add_argument( "-a", "--ssh-agent", action="store_true", help="enable ssh-agent support" ) parser.add_argument( "-c", "--ssh-config", metavar="ssh_config path", default="~/.ssh/config", type=str, help="path to the ssh-configuration file (default to ~/.ssh/config)" ) parser.add_argument( "-n", "--known-hosts", metavar="known_hosts path", default="~/.ssh/known_hosts", type=str, help="path to the openSSH known_hosts file" ) parser.add_argument( "-d", "--disable-known-hosts", action="store_true", help="disable known_hosts fingerprint checking (security warning!)" ) parser.add_argument( "-e", "--exclude-from", metavar="exclude-from-file-path", type=str, help="exclude files matching pattern in exclude-from-file-path" ) parser.add_argument( "-t", "--do-not-delete", action="store_true", help="do not delete remote files missing from local folder" ) parser.add_argument( "-o", "--allow-unknown", action="store_true", help="allow connection to unknown hosts" ) parser.add_argument( "-r", "--create-remote-directory", action="store_true", help="Create remote base directory if missing on remote" ) return parser def main(args=None): """The main.""" parser = create_parser() args = vars(parser.parse_args(args)) log_mapping = { 'CRITICAL': logging.CRITICAL, 'ERROR': logging.ERROR, 'WARNING': logging.WARNING, 'INFO': logging.INFO, 'DEBUG': logging.DEBUG, 'NOTSET': logging.NOTSET, } log_level = log_mapping[args['logging']] del(args['logging']) global logger logger = configure_logging(log_level) args_mapping = { "path": "local_path", "remote": "remote_url", "ssh_config": "ssh_config_path", "exclude_from": "exclude_file", "known_hosts": "known_hosts_path", "do_not_delete": "delete", "key": "identity_files", } kwargs = { # convert the argument names to class constructor parameters args_mapping[k]: v for k, v in args.items() if v and k in args_mapping } kwargs.update({ k: v for k, v in args.items() if v and k not in args_mapping }) # Special case: disable known_hosts check if args['disable_known_hosts']: kwargs['known_hosts_path'] = None del(kwargs['disable_known_hosts']) # Toggle `do_not_delete` flag if "delete" in kwargs: kwargs["delete"] = not kwargs["delete"] # Manually set the default identity file. kwargs["identity_files"] = kwargs.get("identity_files", None) or ["~/.ssh/id_rsa"] sync = SFTPClone( **kwargs ) sync.run() if __name__ == '__main__': main()
unbit/sftpclone
sftpclone/sftpclone.py
parse_username_password_hostname
python
def parse_username_password_hostname(remote_url): assert remote_url assert ':' in remote_url if '@' in remote_url: username, hostname = remote_url.rsplit('@', 1) else: username, hostname = None, remote_url hostname, remote_path = hostname.split(':', 1) password = None if username and ':' in username: username, password = username.split(':', 1) assert hostname assert remote_path return username, password, hostname, remote_path
Parse a command line string and return username, password, remote hostname and remote path. :param remote_url: A command line string. :return: A tuple, containing username, password, remote hostname and remote path.
train
https://github.com/unbit/sftpclone/blob/1cc89478e680fc4e0d12b1a15b5bafd0390d05da/sftpclone/sftpclone.py#L62-L85
null
#!/usr/bin/env python # coding=utf-8 # Python 2.7 backward compatibility from __future__ import print_function from __future__ import unicode_literals from __future__ import absolute_import import paramiko import paramiko.py3compat import os import os.path import sys import errno from stat import S_ISDIR, S_ISLNK, S_ISREG, S_IMODE, S_IFMT import argparse import logging from getpass import getuser, getpass import glob import socket """SFTPClone: sync local and remote directories.""" logger = None try: # Not available in Python 2.x FileNotFoundError except NameError: FileNotFoundError = IOError def configure_logging(level=logging.DEBUG): """Configure the module logging engine.""" if level == logging.DEBUG: # For debugging purposes, log from everyone! logging.basicConfig( level=logging.DEBUG, format='%(asctime)s - %(levelname)s - %(message)s' ) return logging logger = logging.getLogger(__name__) logger.setLevel(level) formatter = logging.Formatter('%(asctime)s - %(levelname)s - %(message)s') ch = logging.StreamHandler() ch.setLevel(level) ch.setFormatter(formatter) logger.addHandler(ch) return logger def path_join(*args): """ Wrapper around `os.path.join`. Makes sure to join paths of the same type (bytes). """ args = (paramiko.py3compat.u(arg) for arg in args) return os.path.join(*args) def get_ssh_agent_keys(logger): """ Ask the SSH agent for a list of keys, and return it. :return: A reference to the SSH agent and a list of keys. """ agent, agent_keys = None, None try: agent = paramiko.agent.Agent() _agent_keys = agent.get_keys() if not _agent_keys: agent.close() logger.error( "SSH agent didn't provide any valid key. Trying to continue..." ) else: agent_keys = tuple(k for k in _agent_keys) except paramiko.SSHException: if agent: agent.close() agent = None logger.error("SSH agent speaks a non-compatible protocol. Ignoring it.") finally: return agent, agent_keys class SFTPClone(object): """The SFTPClone class.""" def __init__(self, local_path, remote_url, identity_files=None, port=None, fix_symlinks=False, ssh_config_path=None, ssh_agent=False, exclude_file=None, known_hosts_path=None, delete=True, allow_unknown=False, create_remote_directory=False, ): """Init the needed parameters and the SFTPClient.""" self.local_path = os.path.realpath(os.path.expanduser(local_path)) self.logger = logger or configure_logging() self.create_remote_directory = create_remote_directory if not os.path.exists(self.local_path): self.logger.error("Local path MUST exist. Exiting.") sys.exit(1) if exclude_file: with open(exclude_file) as f: # As in rsync's exclude from, ignore lines with leading ; and # # and treat each path as relative (thus by removing the leading # /) exclude_list = [ line.rstrip().lstrip("/") for line in f if not line.startswith((";", "#")) ] # actually, is a set of excluded files self.exclude_list = { g for pattern in exclude_list for g in glob.glob(path_join(self.local_path, pattern)) } else: self.exclude_list = set() username, password, hostname, self.remote_path = parse_username_password_hostname(remote_url) identity_files = identity_files or [] proxy_command = None if ssh_config_path: try: with open(os.path.expanduser(ssh_config_path)) as c_file: ssh_config = paramiko.SSHConfig() ssh_config.parse(c_file) c = ssh_config.lookup(hostname) hostname = c.get("hostname", hostname) username = c.get("user", username) port = int(c.get("port", port)) identity_files = c.get("identityfile", identity_files) proxy_command = c.get("proxycommand") except Exception as e: # it could be safe to continue anyway, # because parameters could have been manually specified self.logger.error( "Error while parsing ssh_config file: %s. Trying to continue anyway...", e ) # Set default values if not username: username = getuser() # defaults to current user port = port or 22 allow_unknown = allow_unknown or False self.chown = False self.fix_symlinks = fix_symlinks or False self.delete = delete if delete is not None else True if ssh_agent: agent, agent_keys = get_ssh_agent_keys(self.logger) else: agent, agent_keys = None, None if not identity_files and not password and not agent_keys: self.logger.error( "You need to specify either a password, an identity or to enable the ssh-agent support." ) sys.exit(1) # only root can change file owner if username == 'root': self.chown = True sock = (hostname, port) if proxy_command is not None: sock = paramiko.proxy.ProxyCommand(proxy_command) try: transport = paramiko.Transport(sock) except socket.gaierror: self.logger.error( "Hostname not known. Are you sure you inserted it correctly?") sys.exit(1) try: ssh_host = hostname if port == 22 else "[{}]:{}".format(hostname, port) known_hosts = None """ Before starting the transport session, we have to configure it. Specifically, we need to configure the preferred PK algorithm. If the system already knows a public key of a specific kind for a remote host, we have to peek its type as the preferred one. """ if known_hosts_path: known_hosts = paramiko.HostKeys() known_hosts_path = os.path.realpath( os.path.expanduser(known_hosts_path)) try: known_hosts.load(known_hosts_path) except IOError: self.logger.error( "Error while loading known hosts file at {}. Exiting...".format( known_hosts_path) ) sys.exit(1) known_keys = known_hosts.lookup(ssh_host) if known_keys is not None: # one or more keys are already known # set their type as preferred transport.get_security_options().key_types = \ tuple(known_keys.keys()) transport.start_client() if not known_hosts: self.logger.warning("Security warning: skipping known hosts check...") else: pubk = transport.get_remote_server_key() if ssh_host in known_hosts.keys(): if not known_hosts.check(ssh_host, pubk): self.logger.error( "Security warning: " "remote key fingerprint {} for hostname " "{} didn't match the one in known_hosts {}. " "Exiting...".format( pubk.get_base64(), ssh_host, known_hosts.lookup(hostname), ) ) sys.exit(1) elif not allow_unknown: prompt = ("The authenticity of host '{}' can't be established.\n" "{} key is {}.\n" "Are you sure you want to continue connecting? [y/n] ").format( ssh_host, pubk.get_name(), pubk.get_base64()) try: # Renamed to `input` in Python 3.x response = raw_input(prompt) except NameError: response = input(prompt) # Note: we do not modify the user's known_hosts file if not (response == "y" or response == "yes"): self.logger.error( "Host authentication failed." ) sys.exit(1) def perform_key_auth(pkey): try: transport.auth_publickey( username=username, key=pkey ) return True except paramiko.SSHException: self.logger.warning( "Authentication with identity {}... failed".format(pkey.get_base64()[:10]) ) return False if password: # Password auth, if specified. transport.auth_password( username=username, password=password ) elif agent_keys: # SSH agent keys have higher priority for pkey in agent_keys: if perform_key_auth(pkey): break # Authentication worked. else: # None of the keys worked. raise paramiko.SSHException elif identity_files: # Then follow identity file (specified from CL or ssh_config) # Try identity files one by one, until one works for key_path in identity_files: key_path = os.path.expanduser(key_path) try: key = paramiko.RSAKey.from_private_key_file(key_path) except paramiko.PasswordRequiredException: pk_password = getpass( "It seems that your identity from '{}' is encrypted. " "Please enter your password: ".format(key_path) ) try: key = paramiko.RSAKey.from_private_key_file(key_path, pk_password) except paramiko.SSHException: self.logger.error( "Incorrect passphrase. Cannot decode private key from '{}'.".format(key_path) ) continue except IOError or paramiko.SSHException: self.logger.error( "Something went wrong while opening '{}'. Skipping it.".format(key_path) ) continue if perform_key_auth(key): break # Authentication worked. else: # None of the keys worked. raise paramiko.SSHException else: # No authentication method specified, we shouldn't arrive here. assert False except paramiko.SSHException: self.logger.error( "None of the provided authentication methods worked. Exiting." ) transport.close() sys.exit(1) finally: if agent: agent.close() self.sftp = paramiko.SFTPClient.from_transport(transport) if self.remote_path.startswith("~"): # nasty hack to let getcwd work without changing dir! self.sftp.chdir('.') self.remote_path = self.remote_path.replace( "~", self.sftp.getcwd()) # home is the initial sftp dir @staticmethod def _file_need_upload(l_st, r_st): return True if \ l_st.st_size != r_st.st_size or int(l_st.st_mtime) != r_st.st_mtime \ else False @staticmethod def _must_be_deleted(local_path, r_st): """Return True if the remote correspondent of local_path has to be deleted. i.e. if it doesn't exists locally or if it has a different type from the remote one.""" # if the file doesn't exists if not os.path.lexists(local_path): return True # or if the file type is different l_st = os.lstat(local_path) if S_IFMT(r_st.st_mode) != S_IFMT(l_st.st_mode): return True return False def _match_modes(self, remote_path, l_st): """Match mod, utime and uid/gid with locals one.""" self.sftp.chmod(remote_path, S_IMODE(l_st.st_mode)) self.sftp.utime(remote_path, (l_st.st_atime, l_st.st_mtime)) if self.chown: self.sftp.chown(remote_path, l_st.st_uid, l_st.st_gid) def file_upload(self, local_path, remote_path, l_st): """Upload local_path to remote_path and set permission and mtime.""" self.sftp.put(local_path, remote_path) self._match_modes(remote_path, l_st) def remote_delete(self, remote_path, r_st): """Remove the remote directory node.""" # If it's a directory, then delete content and directory if S_ISDIR(r_st.st_mode): for item in self.sftp.listdir_attr(remote_path): full_path = path_join(remote_path, item.filename) self.remote_delete(full_path, item) self.sftp.rmdir(remote_path) # Or simply delete files else: try: self.sftp.remove(remote_path) except FileNotFoundError as e: self.logger.error( "error while removing {}. trace: {}".format(remote_path, e) ) def check_for_deletion(self, relative_path=None): """Traverse the entire remote_path tree. Find files/directories that need to be deleted, not being present in the local folder. """ if not relative_path: relative_path = str() # root of shared directory tree remote_path = path_join(self.remote_path, relative_path) local_path = path_join(self.local_path, relative_path) for remote_st in self.sftp.listdir_attr(remote_path): r_lstat = self.sftp.lstat(path_join(remote_path, remote_st.filename)) inner_remote_path = path_join(remote_path, remote_st.filename) inner_local_path = path_join(local_path, remote_st.filename) # check if remote_st is a symlink # otherwise could delete file outside shared directory if S_ISLNK(r_lstat.st_mode): if self._must_be_deleted(inner_local_path, r_lstat): self.remote_delete(inner_remote_path, r_lstat) continue if self._must_be_deleted(inner_local_path, remote_st): self.remote_delete(inner_remote_path, remote_st) elif S_ISDIR(remote_st.st_mode): self.check_for_deletion( path_join(relative_path, remote_st.filename) ) def create_update_symlink(self, link_destination, remote_path): """Create a new link pointing to link_destination in remote_path position.""" try: # if there's anything, delete it self.sftp.remove(remote_path) except IOError: # that's fine, nothing exists there! pass finally: # and recreate the link try: self.sftp.symlink(link_destination, remote_path) except OSError as e: # Sometimes, if links are "too" different, symlink fails. # Sadly, nothing we can do about it. self.logger.error("error while symlinking {} to {}: {}".format( remote_path, link_destination, e)) def node_check_for_upload_create(self, relative_path, f): """Check if the given directory tree node has to be uploaded/created on the remote folder.""" if not relative_path: # we're at the root of the shared directory tree relative_path = str() # the (absolute) local address of f. local_path = path_join(self.local_path, relative_path, f) try: l_st = os.lstat(local_path) except OSError as e: """A little background here. Sometimes, in big clusters configurations (mail, etc.), files could disappear or be moved, suddenly. There's nothing to do about it, system should be stopped before doing backups. Anyway, we log it, and skip it. """ self.logger.error("error while checking {}: {}".format(relative_path, e)) return if local_path in self.exclude_list: self.logger.info("Skipping excluded file %s.", local_path) return # the (absolute) remote address of f. remote_path = path_join(self.remote_path, relative_path, f) # First case: f is a directory if S_ISDIR(l_st.st_mode): # we check if the folder exists on the remote side # it has to be a folder, otherwise it would have already been # deleted try: self.sftp.stat(remote_path) except IOError: # it doesn't exist yet on remote side self.sftp.mkdir(remote_path) self._match_modes(remote_path, l_st) # now, we should traverse f too (recursion magic!) self.check_for_upload_create(path_join(relative_path, f)) # Second case: f is a symbolic link elif S_ISLNK(l_st.st_mode): # read the local link local_link = os.readlink(local_path) absolute_local_link = os.path.realpath(local_link) # is it absolute? is_absolute = local_link.startswith("/") # and does it point inside the shared directory? # add trailing slash (security) trailing_local_path = path_join(self.local_path, '') relpath = os.path.commonprefix( [absolute_local_link, trailing_local_path] ) == trailing_local_path if relpath: relative_link = absolute_local_link[len(trailing_local_path):] else: relative_link = None """ # Refactor them all, be efficient! # Case A: absolute link pointing outside shared directory # (we can only update the remote part) if is_absolute and not relpath: self.create_update_symlink(local_link, remote_path) # Case B: absolute link pointing inside shared directory # (we can leave it as it is or fix the prefix to match the one of the remote server) elif is_absolute and relpath: if self.fix_symlinks: self.create_update_symlink( join( self.remote_path, relative_link, ), remote_path ) else: self.create_update_symlink(local_link, remote_path) # Case C: relative link pointing outside shared directory # (all we can do is try to make the link anyway) elif not is_absolute and not relpath: self.create_update_symlink(local_link, remote_path) # Case D: relative link pointing inside shared directory # (we preserve the relativity and link it!) elif not is_absolute and relpath: self.create_update_symlink(local_link, remote_path) """ if is_absolute and relpath: if self.fix_symlinks: self.create_update_symlink( path_join( self.remote_path, relative_link, ), remote_path ) else: self.create_update_symlink(local_link, remote_path) # Third case: regular file elif S_ISREG(l_st.st_mode): try: r_st = self.sftp.lstat(remote_path) if self._file_need_upload(l_st, r_st): self.file_upload(local_path, remote_path, l_st) except IOError as e: if e.errno == errno.ENOENT: self.file_upload(local_path, remote_path, l_st) # Anything else. else: self.logger.warning("Skipping unsupported file %s.", local_path) def check_for_upload_create(self, relative_path=None): """Traverse the relative_path tree and check for files that need to be uploaded/created. Relativity here refers to the shared directory tree.""" for f in os.listdir( path_join( self.local_path, relative_path) if relative_path else self.local_path ): self.node_check_for_upload_create(relative_path, f) def run(self): """Run the sync. Confront the local and the remote directories and perform the needed changes.""" # Check if remote path is present try: self.sftp.stat(self.remote_path) except FileNotFoundError as e: if self.create_remote_directory: self.sftp.mkdir(self.remote_path) self.logger.info( "Created missing remote dir: '" + self.remote_path + "'") else: self.logger.error( "Remote folder does not exists. " "Add '-r' to create it if missing.") sys.exit(1) try: if self.delete: # First check for items to be removed self.check_for_deletion() # Now scan local for items to upload/create self.check_for_upload_create() except FileNotFoundError: # If this happens, probably the remote folder doesn't exist. self.logger.error( "Error while opening remote folder. Are you sure it does exist?") sys.exit(1) def create_parser(): """Create the CLI argument parser.""" parser = argparse.ArgumentParser( description='Sync a local and a remote folder through SFTP.' ) parser.add_argument( "path", type=str, metavar="local-path", help="the path of the local folder", ) parser.add_argument( "remote", type=str, metavar="user[:password]@hostname:remote-path", help="the ssh-url ([user[:password]@]hostname:remote-path) of the remote folder. " "The hostname can be specified as a ssh_config's hostname too. " "Every missing information will be gathered from there", ) parser.add_argument( "-k", "--key", metavar="identity-path", action="append", help="private key identity path (defaults to ~/.ssh/id_rsa)" ) parser.add_argument( "-l", "--logging", choices=['CRITICAL', 'ERROR', 'WARNING', 'INFO', 'DEBUG', 'NOTSET'], default='ERROR', help="set logging level" ) parser.add_argument( "-p", "--port", default=22, type=int, help="SSH remote port (defaults to 22)" ) parser.add_argument( "-f", "--fix-symlinks", action="store_true", help="fix symbolic links on remote side" ) parser.add_argument( "-a", "--ssh-agent", action="store_true", help="enable ssh-agent support" ) parser.add_argument( "-c", "--ssh-config", metavar="ssh_config path", default="~/.ssh/config", type=str, help="path to the ssh-configuration file (default to ~/.ssh/config)" ) parser.add_argument( "-n", "--known-hosts", metavar="known_hosts path", default="~/.ssh/known_hosts", type=str, help="path to the openSSH known_hosts file" ) parser.add_argument( "-d", "--disable-known-hosts", action="store_true", help="disable known_hosts fingerprint checking (security warning!)" ) parser.add_argument( "-e", "--exclude-from", metavar="exclude-from-file-path", type=str, help="exclude files matching pattern in exclude-from-file-path" ) parser.add_argument( "-t", "--do-not-delete", action="store_true", help="do not delete remote files missing from local folder" ) parser.add_argument( "-o", "--allow-unknown", action="store_true", help="allow connection to unknown hosts" ) parser.add_argument( "-r", "--create-remote-directory", action="store_true", help="Create remote base directory if missing on remote" ) return parser def main(args=None): """The main.""" parser = create_parser() args = vars(parser.parse_args(args)) log_mapping = { 'CRITICAL': logging.CRITICAL, 'ERROR': logging.ERROR, 'WARNING': logging.WARNING, 'INFO': logging.INFO, 'DEBUG': logging.DEBUG, 'NOTSET': logging.NOTSET, } log_level = log_mapping[args['logging']] del(args['logging']) global logger logger = configure_logging(log_level) args_mapping = { "path": "local_path", "remote": "remote_url", "ssh_config": "ssh_config_path", "exclude_from": "exclude_file", "known_hosts": "known_hosts_path", "do_not_delete": "delete", "key": "identity_files", } kwargs = { # convert the argument names to class constructor parameters args_mapping[k]: v for k, v in args.items() if v and k in args_mapping } kwargs.update({ k: v for k, v in args.items() if v and k not in args_mapping }) # Special case: disable known_hosts check if args['disable_known_hosts']: kwargs['known_hosts_path'] = None del(kwargs['disable_known_hosts']) # Toggle `do_not_delete` flag if "delete" in kwargs: kwargs["delete"] = not kwargs["delete"] # Manually set the default identity file. kwargs["identity_files"] = kwargs.get("identity_files", None) or ["~/.ssh/id_rsa"] sync = SFTPClone( **kwargs ) sync.run() if __name__ == '__main__': main()
unbit/sftpclone
sftpclone/sftpclone.py
get_ssh_agent_keys
python
def get_ssh_agent_keys(logger): agent, agent_keys = None, None try: agent = paramiko.agent.Agent() _agent_keys = agent.get_keys() if not _agent_keys: agent.close() logger.error( "SSH agent didn't provide any valid key. Trying to continue..." ) else: agent_keys = tuple(k for k in _agent_keys) except paramiko.SSHException: if agent: agent.close() agent = None logger.error("SSH agent speaks a non-compatible protocol. Ignoring it.") finally: return agent, agent_keys
Ask the SSH agent for a list of keys, and return it. :return: A reference to the SSH agent and a list of keys.
train
https://github.com/unbit/sftpclone/blob/1cc89478e680fc4e0d12b1a15b5bafd0390d05da/sftpclone/sftpclone.py#L88-L113
null
#!/usr/bin/env python # coding=utf-8 # Python 2.7 backward compatibility from __future__ import print_function from __future__ import unicode_literals from __future__ import absolute_import import paramiko import paramiko.py3compat import os import os.path import sys import errno from stat import S_ISDIR, S_ISLNK, S_ISREG, S_IMODE, S_IFMT import argparse import logging from getpass import getuser, getpass import glob import socket """SFTPClone: sync local and remote directories.""" logger = None try: # Not available in Python 2.x FileNotFoundError except NameError: FileNotFoundError = IOError def configure_logging(level=logging.DEBUG): """Configure the module logging engine.""" if level == logging.DEBUG: # For debugging purposes, log from everyone! logging.basicConfig( level=logging.DEBUG, format='%(asctime)s - %(levelname)s - %(message)s' ) return logging logger = logging.getLogger(__name__) logger.setLevel(level) formatter = logging.Formatter('%(asctime)s - %(levelname)s - %(message)s') ch = logging.StreamHandler() ch.setLevel(level) ch.setFormatter(formatter) logger.addHandler(ch) return logger def path_join(*args): """ Wrapper around `os.path.join`. Makes sure to join paths of the same type (bytes). """ args = (paramiko.py3compat.u(arg) for arg in args) return os.path.join(*args) def parse_username_password_hostname(remote_url): """ Parse a command line string and return username, password, remote hostname and remote path. :param remote_url: A command line string. :return: A tuple, containing username, password, remote hostname and remote path. """ assert remote_url assert ':' in remote_url if '@' in remote_url: username, hostname = remote_url.rsplit('@', 1) else: username, hostname = None, remote_url hostname, remote_path = hostname.split(':', 1) password = None if username and ':' in username: username, password = username.split(':', 1) assert hostname assert remote_path return username, password, hostname, remote_path class SFTPClone(object): """The SFTPClone class.""" def __init__(self, local_path, remote_url, identity_files=None, port=None, fix_symlinks=False, ssh_config_path=None, ssh_agent=False, exclude_file=None, known_hosts_path=None, delete=True, allow_unknown=False, create_remote_directory=False, ): """Init the needed parameters and the SFTPClient.""" self.local_path = os.path.realpath(os.path.expanduser(local_path)) self.logger = logger or configure_logging() self.create_remote_directory = create_remote_directory if not os.path.exists(self.local_path): self.logger.error("Local path MUST exist. Exiting.") sys.exit(1) if exclude_file: with open(exclude_file) as f: # As in rsync's exclude from, ignore lines with leading ; and # # and treat each path as relative (thus by removing the leading # /) exclude_list = [ line.rstrip().lstrip("/") for line in f if not line.startswith((";", "#")) ] # actually, is a set of excluded files self.exclude_list = { g for pattern in exclude_list for g in glob.glob(path_join(self.local_path, pattern)) } else: self.exclude_list = set() username, password, hostname, self.remote_path = parse_username_password_hostname(remote_url) identity_files = identity_files or [] proxy_command = None if ssh_config_path: try: with open(os.path.expanduser(ssh_config_path)) as c_file: ssh_config = paramiko.SSHConfig() ssh_config.parse(c_file) c = ssh_config.lookup(hostname) hostname = c.get("hostname", hostname) username = c.get("user", username) port = int(c.get("port", port)) identity_files = c.get("identityfile", identity_files) proxy_command = c.get("proxycommand") except Exception as e: # it could be safe to continue anyway, # because parameters could have been manually specified self.logger.error( "Error while parsing ssh_config file: %s. Trying to continue anyway...", e ) # Set default values if not username: username = getuser() # defaults to current user port = port or 22 allow_unknown = allow_unknown or False self.chown = False self.fix_symlinks = fix_symlinks or False self.delete = delete if delete is not None else True if ssh_agent: agent, agent_keys = get_ssh_agent_keys(self.logger) else: agent, agent_keys = None, None if not identity_files and not password and not agent_keys: self.logger.error( "You need to specify either a password, an identity or to enable the ssh-agent support." ) sys.exit(1) # only root can change file owner if username == 'root': self.chown = True sock = (hostname, port) if proxy_command is not None: sock = paramiko.proxy.ProxyCommand(proxy_command) try: transport = paramiko.Transport(sock) except socket.gaierror: self.logger.error( "Hostname not known. Are you sure you inserted it correctly?") sys.exit(1) try: ssh_host = hostname if port == 22 else "[{}]:{}".format(hostname, port) known_hosts = None """ Before starting the transport session, we have to configure it. Specifically, we need to configure the preferred PK algorithm. If the system already knows a public key of a specific kind for a remote host, we have to peek its type as the preferred one. """ if known_hosts_path: known_hosts = paramiko.HostKeys() known_hosts_path = os.path.realpath( os.path.expanduser(known_hosts_path)) try: known_hosts.load(known_hosts_path) except IOError: self.logger.error( "Error while loading known hosts file at {}. Exiting...".format( known_hosts_path) ) sys.exit(1) known_keys = known_hosts.lookup(ssh_host) if known_keys is not None: # one or more keys are already known # set their type as preferred transport.get_security_options().key_types = \ tuple(known_keys.keys()) transport.start_client() if not known_hosts: self.logger.warning("Security warning: skipping known hosts check...") else: pubk = transport.get_remote_server_key() if ssh_host in known_hosts.keys(): if not known_hosts.check(ssh_host, pubk): self.logger.error( "Security warning: " "remote key fingerprint {} for hostname " "{} didn't match the one in known_hosts {}. " "Exiting...".format( pubk.get_base64(), ssh_host, known_hosts.lookup(hostname), ) ) sys.exit(1) elif not allow_unknown: prompt = ("The authenticity of host '{}' can't be established.\n" "{} key is {}.\n" "Are you sure you want to continue connecting? [y/n] ").format( ssh_host, pubk.get_name(), pubk.get_base64()) try: # Renamed to `input` in Python 3.x response = raw_input(prompt) except NameError: response = input(prompt) # Note: we do not modify the user's known_hosts file if not (response == "y" or response == "yes"): self.logger.error( "Host authentication failed." ) sys.exit(1) def perform_key_auth(pkey): try: transport.auth_publickey( username=username, key=pkey ) return True except paramiko.SSHException: self.logger.warning( "Authentication with identity {}... failed".format(pkey.get_base64()[:10]) ) return False if password: # Password auth, if specified. transport.auth_password( username=username, password=password ) elif agent_keys: # SSH agent keys have higher priority for pkey in agent_keys: if perform_key_auth(pkey): break # Authentication worked. else: # None of the keys worked. raise paramiko.SSHException elif identity_files: # Then follow identity file (specified from CL or ssh_config) # Try identity files one by one, until one works for key_path in identity_files: key_path = os.path.expanduser(key_path) try: key = paramiko.RSAKey.from_private_key_file(key_path) except paramiko.PasswordRequiredException: pk_password = getpass( "It seems that your identity from '{}' is encrypted. " "Please enter your password: ".format(key_path) ) try: key = paramiko.RSAKey.from_private_key_file(key_path, pk_password) except paramiko.SSHException: self.logger.error( "Incorrect passphrase. Cannot decode private key from '{}'.".format(key_path) ) continue except IOError or paramiko.SSHException: self.logger.error( "Something went wrong while opening '{}'. Skipping it.".format(key_path) ) continue if perform_key_auth(key): break # Authentication worked. else: # None of the keys worked. raise paramiko.SSHException else: # No authentication method specified, we shouldn't arrive here. assert False except paramiko.SSHException: self.logger.error( "None of the provided authentication methods worked. Exiting." ) transport.close() sys.exit(1) finally: if agent: agent.close() self.sftp = paramiko.SFTPClient.from_transport(transport) if self.remote_path.startswith("~"): # nasty hack to let getcwd work without changing dir! self.sftp.chdir('.') self.remote_path = self.remote_path.replace( "~", self.sftp.getcwd()) # home is the initial sftp dir @staticmethod def _file_need_upload(l_st, r_st): return True if \ l_st.st_size != r_st.st_size or int(l_st.st_mtime) != r_st.st_mtime \ else False @staticmethod def _must_be_deleted(local_path, r_st): """Return True if the remote correspondent of local_path has to be deleted. i.e. if it doesn't exists locally or if it has a different type from the remote one.""" # if the file doesn't exists if not os.path.lexists(local_path): return True # or if the file type is different l_st = os.lstat(local_path) if S_IFMT(r_st.st_mode) != S_IFMT(l_st.st_mode): return True return False def _match_modes(self, remote_path, l_st): """Match mod, utime and uid/gid with locals one.""" self.sftp.chmod(remote_path, S_IMODE(l_st.st_mode)) self.sftp.utime(remote_path, (l_st.st_atime, l_st.st_mtime)) if self.chown: self.sftp.chown(remote_path, l_st.st_uid, l_st.st_gid) def file_upload(self, local_path, remote_path, l_st): """Upload local_path to remote_path and set permission and mtime.""" self.sftp.put(local_path, remote_path) self._match_modes(remote_path, l_st) def remote_delete(self, remote_path, r_st): """Remove the remote directory node.""" # If it's a directory, then delete content and directory if S_ISDIR(r_st.st_mode): for item in self.sftp.listdir_attr(remote_path): full_path = path_join(remote_path, item.filename) self.remote_delete(full_path, item) self.sftp.rmdir(remote_path) # Or simply delete files else: try: self.sftp.remove(remote_path) except FileNotFoundError as e: self.logger.error( "error while removing {}. trace: {}".format(remote_path, e) ) def check_for_deletion(self, relative_path=None): """Traverse the entire remote_path tree. Find files/directories that need to be deleted, not being present in the local folder. """ if not relative_path: relative_path = str() # root of shared directory tree remote_path = path_join(self.remote_path, relative_path) local_path = path_join(self.local_path, relative_path) for remote_st in self.sftp.listdir_attr(remote_path): r_lstat = self.sftp.lstat(path_join(remote_path, remote_st.filename)) inner_remote_path = path_join(remote_path, remote_st.filename) inner_local_path = path_join(local_path, remote_st.filename) # check if remote_st is a symlink # otherwise could delete file outside shared directory if S_ISLNK(r_lstat.st_mode): if self._must_be_deleted(inner_local_path, r_lstat): self.remote_delete(inner_remote_path, r_lstat) continue if self._must_be_deleted(inner_local_path, remote_st): self.remote_delete(inner_remote_path, remote_st) elif S_ISDIR(remote_st.st_mode): self.check_for_deletion( path_join(relative_path, remote_st.filename) ) def create_update_symlink(self, link_destination, remote_path): """Create a new link pointing to link_destination in remote_path position.""" try: # if there's anything, delete it self.sftp.remove(remote_path) except IOError: # that's fine, nothing exists there! pass finally: # and recreate the link try: self.sftp.symlink(link_destination, remote_path) except OSError as e: # Sometimes, if links are "too" different, symlink fails. # Sadly, nothing we can do about it. self.logger.error("error while symlinking {} to {}: {}".format( remote_path, link_destination, e)) def node_check_for_upload_create(self, relative_path, f): """Check if the given directory tree node has to be uploaded/created on the remote folder.""" if not relative_path: # we're at the root of the shared directory tree relative_path = str() # the (absolute) local address of f. local_path = path_join(self.local_path, relative_path, f) try: l_st = os.lstat(local_path) except OSError as e: """A little background here. Sometimes, in big clusters configurations (mail, etc.), files could disappear or be moved, suddenly. There's nothing to do about it, system should be stopped before doing backups. Anyway, we log it, and skip it. """ self.logger.error("error while checking {}: {}".format(relative_path, e)) return if local_path in self.exclude_list: self.logger.info("Skipping excluded file %s.", local_path) return # the (absolute) remote address of f. remote_path = path_join(self.remote_path, relative_path, f) # First case: f is a directory if S_ISDIR(l_st.st_mode): # we check if the folder exists on the remote side # it has to be a folder, otherwise it would have already been # deleted try: self.sftp.stat(remote_path) except IOError: # it doesn't exist yet on remote side self.sftp.mkdir(remote_path) self._match_modes(remote_path, l_st) # now, we should traverse f too (recursion magic!) self.check_for_upload_create(path_join(relative_path, f)) # Second case: f is a symbolic link elif S_ISLNK(l_st.st_mode): # read the local link local_link = os.readlink(local_path) absolute_local_link = os.path.realpath(local_link) # is it absolute? is_absolute = local_link.startswith("/") # and does it point inside the shared directory? # add trailing slash (security) trailing_local_path = path_join(self.local_path, '') relpath = os.path.commonprefix( [absolute_local_link, trailing_local_path] ) == trailing_local_path if relpath: relative_link = absolute_local_link[len(trailing_local_path):] else: relative_link = None """ # Refactor them all, be efficient! # Case A: absolute link pointing outside shared directory # (we can only update the remote part) if is_absolute and not relpath: self.create_update_symlink(local_link, remote_path) # Case B: absolute link pointing inside shared directory # (we can leave it as it is or fix the prefix to match the one of the remote server) elif is_absolute and relpath: if self.fix_symlinks: self.create_update_symlink( join( self.remote_path, relative_link, ), remote_path ) else: self.create_update_symlink(local_link, remote_path) # Case C: relative link pointing outside shared directory # (all we can do is try to make the link anyway) elif not is_absolute and not relpath: self.create_update_symlink(local_link, remote_path) # Case D: relative link pointing inside shared directory # (we preserve the relativity and link it!) elif not is_absolute and relpath: self.create_update_symlink(local_link, remote_path) """ if is_absolute and relpath: if self.fix_symlinks: self.create_update_symlink( path_join( self.remote_path, relative_link, ), remote_path ) else: self.create_update_symlink(local_link, remote_path) # Third case: regular file elif S_ISREG(l_st.st_mode): try: r_st = self.sftp.lstat(remote_path) if self._file_need_upload(l_st, r_st): self.file_upload(local_path, remote_path, l_st) except IOError as e: if e.errno == errno.ENOENT: self.file_upload(local_path, remote_path, l_st) # Anything else. else: self.logger.warning("Skipping unsupported file %s.", local_path) def check_for_upload_create(self, relative_path=None): """Traverse the relative_path tree and check for files that need to be uploaded/created. Relativity here refers to the shared directory tree.""" for f in os.listdir( path_join( self.local_path, relative_path) if relative_path else self.local_path ): self.node_check_for_upload_create(relative_path, f) def run(self): """Run the sync. Confront the local and the remote directories and perform the needed changes.""" # Check if remote path is present try: self.sftp.stat(self.remote_path) except FileNotFoundError as e: if self.create_remote_directory: self.sftp.mkdir(self.remote_path) self.logger.info( "Created missing remote dir: '" + self.remote_path + "'") else: self.logger.error( "Remote folder does not exists. " "Add '-r' to create it if missing.") sys.exit(1) try: if self.delete: # First check for items to be removed self.check_for_deletion() # Now scan local for items to upload/create self.check_for_upload_create() except FileNotFoundError: # If this happens, probably the remote folder doesn't exist. self.logger.error( "Error while opening remote folder. Are you sure it does exist?") sys.exit(1) def create_parser(): """Create the CLI argument parser.""" parser = argparse.ArgumentParser( description='Sync a local and a remote folder through SFTP.' ) parser.add_argument( "path", type=str, metavar="local-path", help="the path of the local folder", ) parser.add_argument( "remote", type=str, metavar="user[:password]@hostname:remote-path", help="the ssh-url ([user[:password]@]hostname:remote-path) of the remote folder. " "The hostname can be specified as a ssh_config's hostname too. " "Every missing information will be gathered from there", ) parser.add_argument( "-k", "--key", metavar="identity-path", action="append", help="private key identity path (defaults to ~/.ssh/id_rsa)" ) parser.add_argument( "-l", "--logging", choices=['CRITICAL', 'ERROR', 'WARNING', 'INFO', 'DEBUG', 'NOTSET'], default='ERROR', help="set logging level" ) parser.add_argument( "-p", "--port", default=22, type=int, help="SSH remote port (defaults to 22)" ) parser.add_argument( "-f", "--fix-symlinks", action="store_true", help="fix symbolic links on remote side" ) parser.add_argument( "-a", "--ssh-agent", action="store_true", help="enable ssh-agent support" ) parser.add_argument( "-c", "--ssh-config", metavar="ssh_config path", default="~/.ssh/config", type=str, help="path to the ssh-configuration file (default to ~/.ssh/config)" ) parser.add_argument( "-n", "--known-hosts", metavar="known_hosts path", default="~/.ssh/known_hosts", type=str, help="path to the openSSH known_hosts file" ) parser.add_argument( "-d", "--disable-known-hosts", action="store_true", help="disable known_hosts fingerprint checking (security warning!)" ) parser.add_argument( "-e", "--exclude-from", metavar="exclude-from-file-path", type=str, help="exclude files matching pattern in exclude-from-file-path" ) parser.add_argument( "-t", "--do-not-delete", action="store_true", help="do not delete remote files missing from local folder" ) parser.add_argument( "-o", "--allow-unknown", action="store_true", help="allow connection to unknown hosts" ) parser.add_argument( "-r", "--create-remote-directory", action="store_true", help="Create remote base directory if missing on remote" ) return parser def main(args=None): """The main.""" parser = create_parser() args = vars(parser.parse_args(args)) log_mapping = { 'CRITICAL': logging.CRITICAL, 'ERROR': logging.ERROR, 'WARNING': logging.WARNING, 'INFO': logging.INFO, 'DEBUG': logging.DEBUG, 'NOTSET': logging.NOTSET, } log_level = log_mapping[args['logging']] del(args['logging']) global logger logger = configure_logging(log_level) args_mapping = { "path": "local_path", "remote": "remote_url", "ssh_config": "ssh_config_path", "exclude_from": "exclude_file", "known_hosts": "known_hosts_path", "do_not_delete": "delete", "key": "identity_files", } kwargs = { # convert the argument names to class constructor parameters args_mapping[k]: v for k, v in args.items() if v and k in args_mapping } kwargs.update({ k: v for k, v in args.items() if v and k not in args_mapping }) # Special case: disable known_hosts check if args['disable_known_hosts']: kwargs['known_hosts_path'] = None del(kwargs['disable_known_hosts']) # Toggle `do_not_delete` flag if "delete" in kwargs: kwargs["delete"] = not kwargs["delete"] # Manually set the default identity file. kwargs["identity_files"] = kwargs.get("identity_files", None) or ["~/.ssh/id_rsa"] sync = SFTPClone( **kwargs ) sync.run() if __name__ == '__main__': main()