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
jedie/PyHardLinkBackup
PyHardLinkBackup/phlb/filesystem_walk.py
PathLibFilter.iter
python
def iter(self, dir_entries): filter = self.filter for entry in dir_entries: path = filter(Path2(entry.path)) if path != False: yield path
:param dir_entries: list of os.DirEntry() instances
train
https://github.com/jedie/PyHardLinkBackup/blob/be28666834d2d9e3d8aac1b661cb2d5bd4056c29/PyHardLinkBackup/phlb/filesystem_walk.py#L74-L82
null
class PathLibFilter: def __init__(self, filter): """ :param filter: callable to filter in self.iter() """ assert callable(filter) self.filter = filter
jedie/PyHardLinkBackup
PyHardLinkBackup/phlb/add.py
add_dir_entry
python
def add_dir_entry(backup_run, dir_entry_path, process_bar, result): # print(dir_entry_path.pformat()) # print(dir_entry_path.stat.st_nlink) backup_entry = Path2(dir_entry_path.path) filesize = dir_entry_path.stat.st_size hash_filepath = Path2("%s%s%s" % (backup_entry.path, os.extsep, phlb_config.hash_name)) if hash_filepath.is_file(): with hash_filepath.open("r") as hash_file: hash_hexdigest = hash_file.read().strip() if filesize > 0: process_bar.update(filesize) else: with hash_filepath.open("w") as hash_file: callback = HashCallback(process_bar) with backup_entry.open("rb") as f: hash = calculate_hash(f, callback) hash_hexdigest = hash.hexdigest() hash_file.write(hash_hexdigest) old_backup_entry = deduplicate(backup_entry, hash_hexdigest) if old_backup_entry is None: result.add_new_file(filesize) else: result.add_stined_file(filesize) BackupEntry.objects.create( backup_run, dir_entry_path.path_instance, hash_hexdigest=hash_hexdigest # Path2() instance )
:param backup_run: :param dir_entry_path: filesystem_walk.DirEntryPath() instance :param process_bar:
train
https://github.com/jedie/PyHardLinkBackup/blob/be28666834d2d9e3d8aac1b661cb2d5bd4056c29/PyHardLinkBackup/phlb/add.py#L86-L120
[ "def deduplicate(backup_entry, hash_hexdigest):\n abs_dst_root = Path2(phlb_config.backup_path)\n\n try:\n backup_entry.relative_to(abs_dst_root)\n except ValueError as err:\n raise ValueError(\"Backup entry not in backup root path: %s\" % err)\n\n assert backup_entry.is_file(), \"Is not a...
""" Python HardLink Backup ~~~~~~~~~~~~~~~~~~~~~~ :copyleft: 2016 by Jens Diemer :license: GNU GPL v3 or above, see LICENSE for more details. """ import datetime import hashlib import logging import os # time.clock() on windows and time.time() on linux from click._compat import strip_ansi try: # https://github.com/tqdm/tqdm from tqdm import tqdm except ImportError as err: raise ImportError("Please install 'tqdm': %s" % err) # os.environ["DJANGO_SETTINGS_MODULE"] = "PyHardLinkBackup.django_project.settings" import django from pathlib_revised import Path2 # https://github.com/jedie/pathlib revised/ from PyHardLinkBackup.phlb import BACKUP_RUN_CONFIG_FILENAME from PyHardLinkBackup.phlb.deduplicate import deduplicate from PyHardLinkBackup.phlb.phlb_main import scan_dir_tree from PyHardLinkBackup.phlb.traceback_plus import exc_plus from PyHardLinkBackup.phlb.filesystem_walk import scandir_limited from PyHardLinkBackup.phlb.config import phlb_config from PyHardLinkBackup.phlb.human import human_filesize, to_percent from PyHardLinkBackup.backup_app.models import BackupEntry, BackupRun log = logging.getLogger("phlb.%s" % __name__) def calculate_hash(f, callback): # TODO: merge code! hash = hashlib.new(phlb_config.hash_name) f.seek(0) while True: data = f.read(phlb_config.chunk_size) if not data: break hash.update(data) callback(data) return hash class HashCallback: def __init__(self, process_bar): self.process_bar = process_bar def __call__(self, data): self.process_bar.update(len(data)) class DeduplicateResult: def __init__(self): self.total_stined_file_count = 0 self.total_stined_bytes = 0 self.total_new_file_count = 0 self.total_new_bytes = 0 self.total_fast_backup = 0 def add_new_file(self, size): self.total_new_file_count += 1 self.total_new_bytes += size def add_stined_file(self, size): self.total_stined_file_count += 1 self.total_stined_bytes += size def get_total_size(self): return self.total_new_bytes + self.total_stined_bytes def add_dir_entries(backup_run, filtered_dir_entries, result): total_size = sum([entry.stat.st_size for entry in filtered_dir_entries]) print("total size:", human_filesize(total_size)) path_iterator = sorted( filtered_dir_entries, key=lambda x: x.stat.st_mtime, # sort by last modify time reverse=True, # sort from newest to oldes ) # FIXME: The process bar will stuck if many small/null byte files are processed # Maybe: Change from bytes to file count and use a second bar if a big file # hash will be calculated. with tqdm(total=total_size, unit="B", unit_scale=True) as process_bar: for dir_entry in path_iterator: try: add_dir_entry(backup_run, dir_entry, process_bar, result) except Exception as err: # A unexpected error occurred. # Print and add traceback to summary log.error("Can't backup %s: %s" % (dir_entry, err)) for line in exc_plus(): log.error(strip_ansi(line)) def add_backup_entries(backup_run, result): backup_path = backup_run.path_part() filtered_dir_entries = scan_dir_tree( backup_path, extra_skip_patterns=( "*.%s" % phlb_config.hash_name, # skip all existing hash files BACKUP_RUN_CONFIG_FILENAME, # skip phlb_config.ini ), ) add_dir_entries(backup_run, filtered_dir_entries, result) def add_backup_run(backup_run_path): print("*** add backup run: %s" % backup_run_path.path) backup_name = backup_run_path.parent.stem date_part = backup_run_path.stem try: backup_datetime = datetime.datetime.strptime(date_part, phlb_config.sub_dir_formatter) except ValueError as err: print("\nERROR parsing datetime from given path: %s" % err) print(" * Is the given path right?") print() return backup_run = BackupRun.objects.create(name=backup_name, backup_datetime=backup_datetime, completed=False) result = DeduplicateResult() add_backup_entries(backup_run, result) print("*** backup run %s - %s added:" % (backup_name, date_part)) total_size = result.get_total_size() print( " * new content saved: %i files (%s %.1f%%)" % ( result.total_new_file_count, human_filesize(result.total_new_bytes), to_percent(result.total_new_bytes, total_size), ) ) print( " * stint space via hardlinks: %i files (%s %.1f%%)" % ( result.total_stined_file_count, human_filesize(result.total_stined_bytes), to_percent(result.total_stined_bytes, total_size), ) ) def add_backup_name(backup_name_path): backup_runs = scandir_limited(backup_name_path.path, limit=1) for dir_entry in backup_runs: backup_run_path = Path2(dir_entry.path) print(" * %s" % backup_run_path.stem) try: backup_run = BackupRun.objects.get_from_config_file(backup_run_path) except (FileNotFoundError, BackupRun.DoesNotExist) as err: print("Error: %s" % err) # no phlb_config.ini add_backup_run(backup_run_path) else: print("\tBackup exists:", backup_run) def add_all_backups(): abs_dst_root = Path2(phlb_config.backup_path) backup_names = scandir_limited(abs_dst_root.path, limit=1) for dir_entry in backup_names: backup_name_path = Path2(dir_entry.path) print("_" * 79) print("'%s' (path: %s)" % (backup_name_path.stem, backup_name_path.path)) add_backup_name(backup_name_path) def add_backups(): """ Scan all existing backup and add missing ones to database. """ django.setup() add_all_backups()
jedie/PyHardLinkBackup
PyHardLinkBackup/backup_app/migrations/0004_BackupRun_ini_file_20160203_1415.py
forwards_func
python
def forwards_func(apps, schema_editor): print("\n") create_count = 0 BackupRun = apps.get_model("backup_app", "BackupRun") # historical version of BackupRun backup_runs = BackupRun.objects.all() for backup_run in backup_runs: # Use the origin BackupRun model to get access to write_config() temp = OriginBackupRun(name=backup_run.name, backup_datetime=backup_run.backup_datetime) try: temp.write_config() except OSError as err: print("ERROR creating config file: %s" % err) else: create_count += 1 # print("%r created." % config_path.path) print("%i config files created.\n" % create_count)
manage migrate backup_app 0004_BackupRun_ini_file_20160203_1415
train
https://github.com/jedie/PyHardLinkBackup/blob/be28666834d2d9e3d8aac1b661cb2d5bd4056c29/PyHardLinkBackup/backup_app/migrations/0004_BackupRun_ini_file_20160203_1415.py#L9-L27
null
# coding: utf-8 from django.db import migrations, models from PyHardLinkBackup.backup_app.models import BackupRun as OriginBackupRun def reverse_func(apps, schema_editor): """ manage migrate backup_app 0003_auto_20160127_2002 """ print("\n") remove_count = 0 BackupRun = apps.get_model("backup_app", "BackupRun") backup_runs = BackupRun.objects.all() for backup_run in backup_runs: # Use the origin BackupRun model to get access to get_config_path() temp = OriginBackupRun(name=backup_run.name, backup_datetime=backup_run.backup_datetime) config_path = temp.get_config_path() try: config_path.unlink() except OSError as err: print("ERROR removing config file: %s" % err) else: remove_count += 1 # print("%r removed." % config_path.path) print("%i config files removed.\n" % remove_count) class Migration(migrations.Migration): dependencies = [("backup_app", "0003_auto_20160127_2002")] operations = [migrations.RunPython(forwards_func, reverse_func)]
jedie/PyHardLinkBackup
PyHardLinkBackup/backup_app/migrations/0004_BackupRun_ini_file_20160203_1415.py
reverse_func
python
def reverse_func(apps, schema_editor): print("\n") remove_count = 0 BackupRun = apps.get_model("backup_app", "BackupRun") backup_runs = BackupRun.objects.all() for backup_run in backup_runs: # Use the origin BackupRun model to get access to get_config_path() temp = OriginBackupRun(name=backup_run.name, backup_datetime=backup_run.backup_datetime) config_path = temp.get_config_path() try: config_path.unlink() except OSError as err: print("ERROR removing config file: %s" % err) else: remove_count += 1 # print("%r removed." % config_path.path) print("%i config files removed.\n" % remove_count)
manage migrate backup_app 0003_auto_20160127_2002
train
https://github.com/jedie/PyHardLinkBackup/blob/be28666834d2d9e3d8aac1b661cb2d5bd4056c29/PyHardLinkBackup/backup_app/migrations/0004_BackupRun_ini_file_20160203_1415.py#L30-L50
null
# coding: utf-8 from django.db import migrations, models from PyHardLinkBackup.backup_app.models import BackupRun as OriginBackupRun def forwards_func(apps, schema_editor): """ manage migrate backup_app 0004_BackupRun_ini_file_20160203_1415 """ print("\n") create_count = 0 BackupRun = apps.get_model("backup_app", "BackupRun") # historical version of BackupRun backup_runs = BackupRun.objects.all() for backup_run in backup_runs: # Use the origin BackupRun model to get access to write_config() temp = OriginBackupRun(name=backup_run.name, backup_datetime=backup_run.backup_datetime) try: temp.write_config() except OSError as err: print("ERROR creating config file: %s" % err) else: create_count += 1 # print("%r created." % config_path.path) print("%i config files created.\n" % create_count) class Migration(migrations.Migration): dependencies = [("backup_app", "0003_auto_20160127_2002")] operations = [migrations.RunPython(forwards_func, reverse_func)]
uyar/pygenstub
pygenstub.py
get_fields
python
def get_fields(node, fields_tag="field_list"): fields_nodes = [c for c in node.children if c.tagname == fields_tag] if len(fields_nodes) == 0: return {} assert len(fields_nodes) == 1, "multiple nodes with tag " + fields_tag fields_node = fields_nodes[0] fields = [ {f.tagname: f.rawsource.strip() for f in n.children} for n in fields_node.children if n.tagname == "field" ] return {f["field_name"]: f["field_body"] for f in fields}
Get the field names and their values from a node. :sig: (Document, str) -> Dict[str, str] :param node: Node to get the fields from. :param fields_tag: Tag of child node that contains the fields. :return: Names and values of fields.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L81-L99
null
# Copyright (C) 2016-2019 H. Turgut Uyar <uyar@tekir.org> # # pygenstub is free software: you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # pygenstub 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 General Public License for more details. # # You should have received a copy of the GNU General Public License # along with pygenstub. If not, see <http://www.gnu.org/licenses/>. """pygenstub is a utility for generating stub files from docstrings in source files. It takes a source file as input and creates a stub file with the same base name and the ``.pyi`` extension. For more information, please refer to the documentation: https://pygenstub.tekir.org/ """ from __future__ import absolute_import, division, print_function, unicode_literals import ast import inspect import logging import os import re import sys import textwrap from argparse import ArgumentParser from bisect import bisect from collections import OrderedDict from importlib import import_module from io import StringIO from pkgutil import get_loader, walk_packages from docutils.core import publish_doctree __version__ = "1.4.0" # sig: str PY3 = sys.version_info >= (3, 0) if not PY3: import __builtin__ as builtins from pathlib2 import Path else: import builtins from pathlib import Path # sigalias: Document = docutils.nodes.document BUILTIN_TYPES = {k for k, t in builtins.__dict__.items() if isinstance(t, type)} BUILTIN_TYPES.add("None") SIG_FIELD = "sig" # sig: str SIG_COMMENT = "# sig:" # sig: str SIG_ALIAS = "# sigalias:" # sig: str DECORATORS = {"property", "staticmethod", "classmethod"} # sig: Set[str] LINE_LENGTH_LIMIT = 79 INDENT = 4 * " " EDIT_WARNING = "THIS FILE IS AUTOMATICALLY GENERATED, DO NOT EDIT MANUALLY." _RE_QUALIFIED_TYPES = re.compile(r"\w+(?:\.\w+)*") _RE_COMMENT_IN_STRING = re.compile(r"""['"]\s*%(text)s\s*.*['"]""" % {"text": SIG_COMMENT}) _logger = logging.getLogger(__name__) def extract_signature(docstring): """Extract the signature from a docstring. :sig: (str) -> Optional[str] :param docstring: Docstring to extract the signature from. :return: Extracted signature, or ``None`` if there's no signature. """ root = publish_doctree(docstring, settings_overrides={"report_level": 5}) fields = get_fields(root) return fields.get(SIG_FIELD) def get_signature(node): """Get the signature of a function or a class. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef]) -> Optional[str] :param node: Node to get the signature from. :return: Value of signature field in node docstring, or ``None`` if there's no signature. """ docstring = ast.get_docstring(node) if docstring is None: return None return extract_signature(docstring) def split_parameter_types(parameters): """Split a parameter types declaration into individual types. The input is the left hand side of a signature (the part before the arrow), excluding the parentheses. :sig: (str) -> List[str] :param parameters: Comma separated parameter types. :return: Parameter types. """ if parameters == "": return [] # only consider the top level commas, ignore the ones in [] commas = [] bracket_depth = 0 for i, char in enumerate(parameters): if (char == ",") and (bracket_depth == 0): commas.append(i) elif char == "[": bracket_depth += 1 elif char == "]": bracket_depth -= 1 types = [] last_i = 0 for i in commas: types.append(parameters[last_i:i].strip()) last_i = i + 1 else: types.append(parameters[last_i:].strip()) return types def parse_signature(signature): """Parse a signature into its input and return parameter types. This will also collect the types that are required by any of the input and return types. :sig: (str) -> Tuple[List[str], str, Set[str]] :param signature: Signature to parse. :return: Input parameter types, return type, and all required types. """ if " -> " not in signature: # signature comment: no parameters, treat variable type as return type param_types, return_type = None, signature.strip() else: lhs, return_type = [s.strip() for s in signature.split(" -> ")] csv = lhs[1:-1].strip() # remove the parentheses around the parameter type list param_types = split_parameter_types(csv) requires = set(_RE_QUALIFIED_TYPES.findall(signature)) return param_types, return_type, requires class StubNode: """A node in a stub tree.""" def __init__(self): """Initialize this stub node. :sig: () -> None """ self.variables = [] # sig: List[VariableNode] self.variable_names = set() # sig: Set[str] self.children = [] # sig: List[Union[FunctionNode, ClassNode]] self.parent = None # sig: Optional[StubNode] def add_variable(self, node): """Add a variable node to this node. :sig: (VariableNode) -> None :param node: Variable node to add. """ if node.name not in self.variable_names: self.variables.append(node) self.variable_names.add(node.name) node.parent = self def add_child(self, node): """Add a function/method or class node to this node. :sig: (Union[FunctionNode, ClassNode]) -> None :param node: Function or class node to add. """ self.children.append(node) node.parent = self def get_code(self): """Get the stub code for this node. The stub code for a node consists of the type annotations of its variables, followed by the prototypes of its functions/methods and classes. :sig: () -> List[str] :return: Lines of stub code for this node. """ stub = [] for child in self.variables: stub.extend(child.get_code()) if ( (len(self.variables) > 0) and (len(self.children) > 0) and (not isinstance(self, ClassNode)) ): stub.append("") for child in self.children: stub.extend(child.get_code()) return stub class VariableNode(StubNode): """A node representing an assignment in a stub tree.""" def __init__(self, name, type_): """Initialize this variable node. :sig: (str, str) -> None :param name: Name of variable that is being assigned to. :param type_: Type of variable. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.type_ = type_ # sig: str def get_code(self): """Get the type annotation for this variable. :sig: () -> List[str] :return: Lines of stub code for this variable. """ return ["%(n)s = ... # type: %(t)s" % {"n": self.name, "t": self.type_}] class FunctionNode(StubNode): """A node representing a function in a stub tree.""" def __init__(self, name, parameters, rtype, decorators=None): """Initialize this function node. The parameters have to given as a list of triples where each item specifies the name of the parameter, its type, and whether it has a default value or not. :sig: (str, Sequence[Tuple[str, str, bool]], str, Optional[Sequence[str]]) -> None :param name: Name of function. :param parameters: List of parameter triples (name, type, has_default). :param rtype: Type of return value. :param decorators: Decorators of function. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.parameters = parameters # sig: Sequence[Tuple[str, str, bool]] self.rtype = rtype # sig: str self.decorators = decorators if decorators is not None else [] # sig: Sequence[str] self._async = False # sig: bool def get_code(self): """Get the stub code for this function. :sig: () -> List[str] :return: Lines of stub code for this function. """ stub = [] for deco in self.decorators: if (deco in DECORATORS) or deco.endswith(".setter"): stub.append("@" + deco) parameters = [] for name, type_, has_default in self.parameters: decl = "%(n)s%(t)s%(d)s" % { "n": name, "t": ": " + type_ if type_ else "", "d": " = ..." if has_default else "", } parameters.append(decl) slots = { "a": "async " if self._async else "", "n": self.name, "p": ", ".join(parameters), "r": self.rtype, } prototype = "%(a)sdef %(n)s(%(p)s) -> %(r)s: ..." % slots if len(prototype) <= LINE_LENGTH_LIMIT: stub.append(prototype) elif len(INDENT + slots["p"]) <= LINE_LENGTH_LIMIT: stub.append("%(a)sdef %(n)s(" % slots) stub.append(INDENT + slots["p"]) stub.append(") -> %(r)s: ..." % slots) else: stub.append("%(a)sdef %(n)s(" % slots) for param in parameters: stub.append(INDENT + param + ",") stub.append(") -> %(r)s: ..." % slots) return stub class ClassNode(StubNode): """A node representing a class in a stub tree.""" def __init__(self, name, bases, signature=None): """Initialize this class node. :sig: (str, Sequence[str], Optional[str]) -> None :param name: Name of class. :param bases: Base classes of class. :param signature: Signature of class, to be used in __init__ method. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.bases = bases # sig: Sequence[str] self.signature = signature # sig: Optional[str] def get_code(self): """Get the stub code for this class. :sig: () -> List[str] :return: Lines of stub code for this class. """ stub = [] bases = ("(" + ", ".join(self.bases) + ")") if len(self.bases) > 0 else "" slots = {"n": self.name, "b": bases} if (len(self.children) == 0) and (len(self.variables) == 0): stub.append("class %(n)s%(b)s: ..." % slots) else: stub.append("class %(n)s%(b)s:" % slots) super_code = super().get_code() if PY3 else StubNode.get_code(self) for line in super_code: stub.append(INDENT + line) return stub def get_aliases(lines): """Get the type aliases in the source. :sig: (Sequence[str]) -> Dict[str, str] :param lines: Lines of the source code. :return: Aliases and their their definitions. """ aliases = {} for line in lines: line = line.strip() if len(line) > 0 and line.startswith(SIG_ALIAS): _, content = line.split(SIG_ALIAS) alias, signature = [t.strip() for t in content.split("=")] aliases[alias] = signature return aliases class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue() def get_stub(source, generic=False): """Get the stub code for a source code. :sig: (str, bool) -> str :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. :return: Generated stub code. """ generator = StubGenerator(source, generic=generic) stub = generator.generate_stub() return stub def get_mod_paths(mod_name, out_dir): """Get source and stub paths for a module.""" paths = [] try: mod = get_loader(mod_name) source = Path(mod.path) if source.name.endswith(".py"): source_rel = Path(*mod_name.split(".")) if source.name == "__init__.py": source_rel = source_rel.joinpath("__init__.py") destination = Path(out_dir, source_rel.with_suffix(".pyi")) paths.append((source, destination)) except Exception as e: _logger.debug(e) _logger.warning("cannot handle module, skipping: %s", mod_name) return paths def get_pkg_paths(pkg_name, out_dir): """Recursively get all source and stub paths for a package.""" paths = [] try: pkg = import_module(pkg_name) if not hasattr(pkg, "__path__"): return get_mod_paths(pkg_name, out_dir) for mod_info in walk_packages(pkg.__path__, pkg.__name__ + "."): mod_paths = get_mod_paths(mod_info.name, out_dir) paths.extend(mod_paths) except Exception as e: _logger.debug(e) _logger.warning("cannot handle package, skipping: %s", pkg_name) return paths ############################################################ # SPHINX ############################################################ def process_docstring(app, what, name, obj, options, lines): """Modify the docstring before generating documentation. This will insert type declarations for parameters and return type into the docstring, and remove the signature field so that it will be excluded from the generated document. """ aliases = getattr(app, "_sigaliases", None) if aliases is None: if what == "module": aliases = get_aliases(inspect.getsource(obj).splitlines()) app._sigaliases = aliases sig_marker = ":" + SIG_FIELD + ":" is_class = what in ("class", "exception") signature = extract_signature("\n".join(lines)) if signature is None: if not is_class: return init_method = getattr(obj, "__init__") init_doc = init_method.__doc__ init_lines = init_doc.splitlines()[1:] if len(init_lines) > 1: init_doc = textwrap.dedent("\n".join(init_lines[1:])) init_lines = init_doc.splitlines() if sig_marker not in init_doc: return sig_started = False for line in init_lines: if line.lstrip().startswith(sig_marker): sig_started = True if sig_started: lines.append(line) signature = extract_signature("\n".join(lines)) if is_class: obj = init_method param_types, rtype, _ = parse_signature(signature) param_names = [p for p in inspect.signature(obj).parameters] if is_class and (param_names[0] == "self"): del param_names[0] # if something goes wrong, don't insert parameter types if len(param_names) == len(param_types): for name, type_ in zip(param_names, param_types): find = ":param %(name)s:" % {"name": name} alias = aliases.get(type_) if alias is not None: type_ = "*%(type)s* :sup:`%(alias)s`" % {"type": type_, "alias": alias} for i, line in enumerate(lines): if line.startswith(find): lines.insert(i, ":type %(name)s: %(type)s" % {"name": name, "type": type_}) break if not is_class: for i, line in enumerate(lines): if line.startswith((":return:", ":returns:")): lines.insert(i, ":rtype: " + rtype) break # remove the signature field sig_start = 0 while sig_start < len(lines): if lines[sig_start].startswith(sig_marker): break sig_start += 1 sig_end = sig_start + 1 while sig_end < len(lines): if (not lines[sig_end]) or (lines[sig_end][0] != " "): break sig_end += 1 for i in reversed(range(sig_start, sig_end)): del lines[i] def setup(app): """Register to Sphinx.""" app.connect("autodoc-process-docstring", process_docstring) return {"version": __version__} ############################################################ # MAIN ############################################################ def main(argv=None): """Start the command line interface.""" parser = ArgumentParser(prog="pygenstub") parser.add_argument("--version", action="version", version="%(prog)s " + __version__) parser.add_argument("files", nargs="*", help="generate stubs for given files") parser.add_argument( "-m", "--module", action="append", metavar="MODULE", dest="modules", default=[], help="generate stubs for given modules", ) parser.add_argument( "-o", "--output", metavar="PATH", dest="out_dir", help="change the output directory" ) parser.add_argument( "--generic", action="store_true", default=False, help="generate generic stubs" ) parser.add_argument("--debug", action="store_true", help="enable debug messages") argv = argv if argv is not None else sys.argv arguments = parser.parse_args(argv[1:]) # set debug mode if arguments.debug: logging.basicConfig(level=logging.DEBUG) _logger.debug("running in debug mode") out_dir = arguments.out_dir if arguments.out_dir is not None else "" if (out_dir == "") and (len(arguments.modules) > 0): print("Output directory must be given when generating stubs for modules.") sys.exit(1) modules = [] for path in arguments.files: paths = Path(path).glob("**/*.py") if Path(path).is_dir() else [Path(path)] for source in paths: if str(source).startswith(os.path.pardir): source = source.absolute().resolve() if (out_dir != "") and source.is_absolute(): source = source.relative_to(source.root) destination = Path(out_dir, source.with_suffix(".pyi")) modules.append((source, destination)) for mod_name in arguments.modules: modules.extend(get_pkg_paths(mod_name, out_dir)) for source, destination in modules: _logger.info("generating stub for %s to path %s", source, destination) with source.open() as f: code = f.read() try: stub = get_stub(code, generic=arguments.generic) except Exception as e: print(source, "-", e, file=sys.stderr) continue if stub != "": if not destination.parent.exists(): destination.parent.mkdir(parents=True) with destination.open("w") as f: f.write("# " + EDIT_WARNING + "\n\n" + stub) if __name__ == "__main__": main()
uyar/pygenstub
pygenstub.py
extract_signature
python
def extract_signature(docstring): root = publish_doctree(docstring, settings_overrides={"report_level": 5}) fields = get_fields(root) return fields.get(SIG_FIELD)
Extract the signature from a docstring. :sig: (str) -> Optional[str] :param docstring: Docstring to extract the signature from. :return: Extracted signature, or ``None`` if there's no signature.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L102-L111
[ "def get_fields(node, fields_tag=\"field_list\"):\n \"\"\"Get the field names and their values from a node.\n\n :sig: (Document, str) -> Dict[str, str]\n :param node: Node to get the fields from.\n :param fields_tag: Tag of child node that contains the fields.\n :return: Names and values of fields.\n...
# Copyright (C) 2016-2019 H. Turgut Uyar <uyar@tekir.org> # # pygenstub is free software: you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # pygenstub 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 General Public License for more details. # # You should have received a copy of the GNU General Public License # along with pygenstub. If not, see <http://www.gnu.org/licenses/>. """pygenstub is a utility for generating stub files from docstrings in source files. It takes a source file as input and creates a stub file with the same base name and the ``.pyi`` extension. For more information, please refer to the documentation: https://pygenstub.tekir.org/ """ from __future__ import absolute_import, division, print_function, unicode_literals import ast import inspect import logging import os import re import sys import textwrap from argparse import ArgumentParser from bisect import bisect from collections import OrderedDict from importlib import import_module from io import StringIO from pkgutil import get_loader, walk_packages from docutils.core import publish_doctree __version__ = "1.4.0" # sig: str PY3 = sys.version_info >= (3, 0) if not PY3: import __builtin__ as builtins from pathlib2 import Path else: import builtins from pathlib import Path # sigalias: Document = docutils.nodes.document BUILTIN_TYPES = {k for k, t in builtins.__dict__.items() if isinstance(t, type)} BUILTIN_TYPES.add("None") SIG_FIELD = "sig" # sig: str SIG_COMMENT = "# sig:" # sig: str SIG_ALIAS = "# sigalias:" # sig: str DECORATORS = {"property", "staticmethod", "classmethod"} # sig: Set[str] LINE_LENGTH_LIMIT = 79 INDENT = 4 * " " EDIT_WARNING = "THIS FILE IS AUTOMATICALLY GENERATED, DO NOT EDIT MANUALLY." _RE_QUALIFIED_TYPES = re.compile(r"\w+(?:\.\w+)*") _RE_COMMENT_IN_STRING = re.compile(r"""['"]\s*%(text)s\s*.*['"]""" % {"text": SIG_COMMENT}) _logger = logging.getLogger(__name__) def get_fields(node, fields_tag="field_list"): """Get the field names and their values from a node. :sig: (Document, str) -> Dict[str, str] :param node: Node to get the fields from. :param fields_tag: Tag of child node that contains the fields. :return: Names and values of fields. """ fields_nodes = [c for c in node.children if c.tagname == fields_tag] if len(fields_nodes) == 0: return {} assert len(fields_nodes) == 1, "multiple nodes with tag " + fields_tag fields_node = fields_nodes[0] fields = [ {f.tagname: f.rawsource.strip() for f in n.children} for n in fields_node.children if n.tagname == "field" ] return {f["field_name"]: f["field_body"] for f in fields} def get_signature(node): """Get the signature of a function or a class. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef]) -> Optional[str] :param node: Node to get the signature from. :return: Value of signature field in node docstring, or ``None`` if there's no signature. """ docstring = ast.get_docstring(node) if docstring is None: return None return extract_signature(docstring) def split_parameter_types(parameters): """Split a parameter types declaration into individual types. The input is the left hand side of a signature (the part before the arrow), excluding the parentheses. :sig: (str) -> List[str] :param parameters: Comma separated parameter types. :return: Parameter types. """ if parameters == "": return [] # only consider the top level commas, ignore the ones in [] commas = [] bracket_depth = 0 for i, char in enumerate(parameters): if (char == ",") and (bracket_depth == 0): commas.append(i) elif char == "[": bracket_depth += 1 elif char == "]": bracket_depth -= 1 types = [] last_i = 0 for i in commas: types.append(parameters[last_i:i].strip()) last_i = i + 1 else: types.append(parameters[last_i:].strip()) return types def parse_signature(signature): """Parse a signature into its input and return parameter types. This will also collect the types that are required by any of the input and return types. :sig: (str) -> Tuple[List[str], str, Set[str]] :param signature: Signature to parse. :return: Input parameter types, return type, and all required types. """ if " -> " not in signature: # signature comment: no parameters, treat variable type as return type param_types, return_type = None, signature.strip() else: lhs, return_type = [s.strip() for s in signature.split(" -> ")] csv = lhs[1:-1].strip() # remove the parentheses around the parameter type list param_types = split_parameter_types(csv) requires = set(_RE_QUALIFIED_TYPES.findall(signature)) return param_types, return_type, requires class StubNode: """A node in a stub tree.""" def __init__(self): """Initialize this stub node. :sig: () -> None """ self.variables = [] # sig: List[VariableNode] self.variable_names = set() # sig: Set[str] self.children = [] # sig: List[Union[FunctionNode, ClassNode]] self.parent = None # sig: Optional[StubNode] def add_variable(self, node): """Add a variable node to this node. :sig: (VariableNode) -> None :param node: Variable node to add. """ if node.name not in self.variable_names: self.variables.append(node) self.variable_names.add(node.name) node.parent = self def add_child(self, node): """Add a function/method or class node to this node. :sig: (Union[FunctionNode, ClassNode]) -> None :param node: Function or class node to add. """ self.children.append(node) node.parent = self def get_code(self): """Get the stub code for this node. The stub code for a node consists of the type annotations of its variables, followed by the prototypes of its functions/methods and classes. :sig: () -> List[str] :return: Lines of stub code for this node. """ stub = [] for child in self.variables: stub.extend(child.get_code()) if ( (len(self.variables) > 0) and (len(self.children) > 0) and (not isinstance(self, ClassNode)) ): stub.append("") for child in self.children: stub.extend(child.get_code()) return stub class VariableNode(StubNode): """A node representing an assignment in a stub tree.""" def __init__(self, name, type_): """Initialize this variable node. :sig: (str, str) -> None :param name: Name of variable that is being assigned to. :param type_: Type of variable. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.type_ = type_ # sig: str def get_code(self): """Get the type annotation for this variable. :sig: () -> List[str] :return: Lines of stub code for this variable. """ return ["%(n)s = ... # type: %(t)s" % {"n": self.name, "t": self.type_}] class FunctionNode(StubNode): """A node representing a function in a stub tree.""" def __init__(self, name, parameters, rtype, decorators=None): """Initialize this function node. The parameters have to given as a list of triples where each item specifies the name of the parameter, its type, and whether it has a default value or not. :sig: (str, Sequence[Tuple[str, str, bool]], str, Optional[Sequence[str]]) -> None :param name: Name of function. :param parameters: List of parameter triples (name, type, has_default). :param rtype: Type of return value. :param decorators: Decorators of function. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.parameters = parameters # sig: Sequence[Tuple[str, str, bool]] self.rtype = rtype # sig: str self.decorators = decorators if decorators is not None else [] # sig: Sequence[str] self._async = False # sig: bool def get_code(self): """Get the stub code for this function. :sig: () -> List[str] :return: Lines of stub code for this function. """ stub = [] for deco in self.decorators: if (deco in DECORATORS) or deco.endswith(".setter"): stub.append("@" + deco) parameters = [] for name, type_, has_default in self.parameters: decl = "%(n)s%(t)s%(d)s" % { "n": name, "t": ": " + type_ if type_ else "", "d": " = ..." if has_default else "", } parameters.append(decl) slots = { "a": "async " if self._async else "", "n": self.name, "p": ", ".join(parameters), "r": self.rtype, } prototype = "%(a)sdef %(n)s(%(p)s) -> %(r)s: ..." % slots if len(prototype) <= LINE_LENGTH_LIMIT: stub.append(prototype) elif len(INDENT + slots["p"]) <= LINE_LENGTH_LIMIT: stub.append("%(a)sdef %(n)s(" % slots) stub.append(INDENT + slots["p"]) stub.append(") -> %(r)s: ..." % slots) else: stub.append("%(a)sdef %(n)s(" % slots) for param in parameters: stub.append(INDENT + param + ",") stub.append(") -> %(r)s: ..." % slots) return stub class ClassNode(StubNode): """A node representing a class in a stub tree.""" def __init__(self, name, bases, signature=None): """Initialize this class node. :sig: (str, Sequence[str], Optional[str]) -> None :param name: Name of class. :param bases: Base classes of class. :param signature: Signature of class, to be used in __init__ method. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.bases = bases # sig: Sequence[str] self.signature = signature # sig: Optional[str] def get_code(self): """Get the stub code for this class. :sig: () -> List[str] :return: Lines of stub code for this class. """ stub = [] bases = ("(" + ", ".join(self.bases) + ")") if len(self.bases) > 0 else "" slots = {"n": self.name, "b": bases} if (len(self.children) == 0) and (len(self.variables) == 0): stub.append("class %(n)s%(b)s: ..." % slots) else: stub.append("class %(n)s%(b)s:" % slots) super_code = super().get_code() if PY3 else StubNode.get_code(self) for line in super_code: stub.append(INDENT + line) return stub def get_aliases(lines): """Get the type aliases in the source. :sig: (Sequence[str]) -> Dict[str, str] :param lines: Lines of the source code. :return: Aliases and their their definitions. """ aliases = {} for line in lines: line = line.strip() if len(line) > 0 and line.startswith(SIG_ALIAS): _, content = line.split(SIG_ALIAS) alias, signature = [t.strip() for t in content.split("=")] aliases[alias] = signature return aliases class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue() def get_stub(source, generic=False): """Get the stub code for a source code. :sig: (str, bool) -> str :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. :return: Generated stub code. """ generator = StubGenerator(source, generic=generic) stub = generator.generate_stub() return stub def get_mod_paths(mod_name, out_dir): """Get source and stub paths for a module.""" paths = [] try: mod = get_loader(mod_name) source = Path(mod.path) if source.name.endswith(".py"): source_rel = Path(*mod_name.split(".")) if source.name == "__init__.py": source_rel = source_rel.joinpath("__init__.py") destination = Path(out_dir, source_rel.with_suffix(".pyi")) paths.append((source, destination)) except Exception as e: _logger.debug(e) _logger.warning("cannot handle module, skipping: %s", mod_name) return paths def get_pkg_paths(pkg_name, out_dir): """Recursively get all source and stub paths for a package.""" paths = [] try: pkg = import_module(pkg_name) if not hasattr(pkg, "__path__"): return get_mod_paths(pkg_name, out_dir) for mod_info in walk_packages(pkg.__path__, pkg.__name__ + "."): mod_paths = get_mod_paths(mod_info.name, out_dir) paths.extend(mod_paths) except Exception as e: _logger.debug(e) _logger.warning("cannot handle package, skipping: %s", pkg_name) return paths ############################################################ # SPHINX ############################################################ def process_docstring(app, what, name, obj, options, lines): """Modify the docstring before generating documentation. This will insert type declarations for parameters and return type into the docstring, and remove the signature field so that it will be excluded from the generated document. """ aliases = getattr(app, "_sigaliases", None) if aliases is None: if what == "module": aliases = get_aliases(inspect.getsource(obj).splitlines()) app._sigaliases = aliases sig_marker = ":" + SIG_FIELD + ":" is_class = what in ("class", "exception") signature = extract_signature("\n".join(lines)) if signature is None: if not is_class: return init_method = getattr(obj, "__init__") init_doc = init_method.__doc__ init_lines = init_doc.splitlines()[1:] if len(init_lines) > 1: init_doc = textwrap.dedent("\n".join(init_lines[1:])) init_lines = init_doc.splitlines() if sig_marker not in init_doc: return sig_started = False for line in init_lines: if line.lstrip().startswith(sig_marker): sig_started = True if sig_started: lines.append(line) signature = extract_signature("\n".join(lines)) if is_class: obj = init_method param_types, rtype, _ = parse_signature(signature) param_names = [p for p in inspect.signature(obj).parameters] if is_class and (param_names[0] == "self"): del param_names[0] # if something goes wrong, don't insert parameter types if len(param_names) == len(param_types): for name, type_ in zip(param_names, param_types): find = ":param %(name)s:" % {"name": name} alias = aliases.get(type_) if alias is not None: type_ = "*%(type)s* :sup:`%(alias)s`" % {"type": type_, "alias": alias} for i, line in enumerate(lines): if line.startswith(find): lines.insert(i, ":type %(name)s: %(type)s" % {"name": name, "type": type_}) break if not is_class: for i, line in enumerate(lines): if line.startswith((":return:", ":returns:")): lines.insert(i, ":rtype: " + rtype) break # remove the signature field sig_start = 0 while sig_start < len(lines): if lines[sig_start].startswith(sig_marker): break sig_start += 1 sig_end = sig_start + 1 while sig_end < len(lines): if (not lines[sig_end]) or (lines[sig_end][0] != " "): break sig_end += 1 for i in reversed(range(sig_start, sig_end)): del lines[i] def setup(app): """Register to Sphinx.""" app.connect("autodoc-process-docstring", process_docstring) return {"version": __version__} ############################################################ # MAIN ############################################################ def main(argv=None): """Start the command line interface.""" parser = ArgumentParser(prog="pygenstub") parser.add_argument("--version", action="version", version="%(prog)s " + __version__) parser.add_argument("files", nargs="*", help="generate stubs for given files") parser.add_argument( "-m", "--module", action="append", metavar="MODULE", dest="modules", default=[], help="generate stubs for given modules", ) parser.add_argument( "-o", "--output", metavar="PATH", dest="out_dir", help="change the output directory" ) parser.add_argument( "--generic", action="store_true", default=False, help="generate generic stubs" ) parser.add_argument("--debug", action="store_true", help="enable debug messages") argv = argv if argv is not None else sys.argv arguments = parser.parse_args(argv[1:]) # set debug mode if arguments.debug: logging.basicConfig(level=logging.DEBUG) _logger.debug("running in debug mode") out_dir = arguments.out_dir if arguments.out_dir is not None else "" if (out_dir == "") and (len(arguments.modules) > 0): print("Output directory must be given when generating stubs for modules.") sys.exit(1) modules = [] for path in arguments.files: paths = Path(path).glob("**/*.py") if Path(path).is_dir() else [Path(path)] for source in paths: if str(source).startswith(os.path.pardir): source = source.absolute().resolve() if (out_dir != "") and source.is_absolute(): source = source.relative_to(source.root) destination = Path(out_dir, source.with_suffix(".pyi")) modules.append((source, destination)) for mod_name in arguments.modules: modules.extend(get_pkg_paths(mod_name, out_dir)) for source, destination in modules: _logger.info("generating stub for %s to path %s", source, destination) with source.open() as f: code = f.read() try: stub = get_stub(code, generic=arguments.generic) except Exception as e: print(source, "-", e, file=sys.stderr) continue if stub != "": if not destination.parent.exists(): destination.parent.mkdir(parents=True) with destination.open("w") as f: f.write("# " + EDIT_WARNING + "\n\n" + stub) if __name__ == "__main__": main()
uyar/pygenstub
pygenstub.py
split_parameter_types
python
def split_parameter_types(parameters): if parameters == "": return [] # only consider the top level commas, ignore the ones in [] commas = [] bracket_depth = 0 for i, char in enumerate(parameters): if (char == ",") and (bracket_depth == 0): commas.append(i) elif char == "[": bracket_depth += 1 elif char == "]": bracket_depth -= 1 types = [] last_i = 0 for i in commas: types.append(parameters[last_i:i].strip()) last_i = i + 1 else: types.append(parameters[last_i:].strip()) return types
Split a parameter types declaration into individual types. The input is the left hand side of a signature (the part before the arrow), excluding the parentheses. :sig: (str) -> List[str] :param parameters: Comma separated parameter types. :return: Parameter types.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L127-L158
null
# Copyright (C) 2016-2019 H. Turgut Uyar <uyar@tekir.org> # # pygenstub is free software: you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # pygenstub 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 General Public License for more details. # # You should have received a copy of the GNU General Public License # along with pygenstub. If not, see <http://www.gnu.org/licenses/>. """pygenstub is a utility for generating stub files from docstrings in source files. It takes a source file as input and creates a stub file with the same base name and the ``.pyi`` extension. For more information, please refer to the documentation: https://pygenstub.tekir.org/ """ from __future__ import absolute_import, division, print_function, unicode_literals import ast import inspect import logging import os import re import sys import textwrap from argparse import ArgumentParser from bisect import bisect from collections import OrderedDict from importlib import import_module from io import StringIO from pkgutil import get_loader, walk_packages from docutils.core import publish_doctree __version__ = "1.4.0" # sig: str PY3 = sys.version_info >= (3, 0) if not PY3: import __builtin__ as builtins from pathlib2 import Path else: import builtins from pathlib import Path # sigalias: Document = docutils.nodes.document BUILTIN_TYPES = {k for k, t in builtins.__dict__.items() if isinstance(t, type)} BUILTIN_TYPES.add("None") SIG_FIELD = "sig" # sig: str SIG_COMMENT = "# sig:" # sig: str SIG_ALIAS = "# sigalias:" # sig: str DECORATORS = {"property", "staticmethod", "classmethod"} # sig: Set[str] LINE_LENGTH_LIMIT = 79 INDENT = 4 * " " EDIT_WARNING = "THIS FILE IS AUTOMATICALLY GENERATED, DO NOT EDIT MANUALLY." _RE_QUALIFIED_TYPES = re.compile(r"\w+(?:\.\w+)*") _RE_COMMENT_IN_STRING = re.compile(r"""['"]\s*%(text)s\s*.*['"]""" % {"text": SIG_COMMENT}) _logger = logging.getLogger(__name__) def get_fields(node, fields_tag="field_list"): """Get the field names and their values from a node. :sig: (Document, str) -> Dict[str, str] :param node: Node to get the fields from. :param fields_tag: Tag of child node that contains the fields. :return: Names and values of fields. """ fields_nodes = [c for c in node.children if c.tagname == fields_tag] if len(fields_nodes) == 0: return {} assert len(fields_nodes) == 1, "multiple nodes with tag " + fields_tag fields_node = fields_nodes[0] fields = [ {f.tagname: f.rawsource.strip() for f in n.children} for n in fields_node.children if n.tagname == "field" ] return {f["field_name"]: f["field_body"] for f in fields} def extract_signature(docstring): """Extract the signature from a docstring. :sig: (str) -> Optional[str] :param docstring: Docstring to extract the signature from. :return: Extracted signature, or ``None`` if there's no signature. """ root = publish_doctree(docstring, settings_overrides={"report_level": 5}) fields = get_fields(root) return fields.get(SIG_FIELD) def get_signature(node): """Get the signature of a function or a class. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef]) -> Optional[str] :param node: Node to get the signature from. :return: Value of signature field in node docstring, or ``None`` if there's no signature. """ docstring = ast.get_docstring(node) if docstring is None: return None return extract_signature(docstring) def parse_signature(signature): """Parse a signature into its input and return parameter types. This will also collect the types that are required by any of the input and return types. :sig: (str) -> Tuple[List[str], str, Set[str]] :param signature: Signature to parse. :return: Input parameter types, return type, and all required types. """ if " -> " not in signature: # signature comment: no parameters, treat variable type as return type param_types, return_type = None, signature.strip() else: lhs, return_type = [s.strip() for s in signature.split(" -> ")] csv = lhs[1:-1].strip() # remove the parentheses around the parameter type list param_types = split_parameter_types(csv) requires = set(_RE_QUALIFIED_TYPES.findall(signature)) return param_types, return_type, requires class StubNode: """A node in a stub tree.""" def __init__(self): """Initialize this stub node. :sig: () -> None """ self.variables = [] # sig: List[VariableNode] self.variable_names = set() # sig: Set[str] self.children = [] # sig: List[Union[FunctionNode, ClassNode]] self.parent = None # sig: Optional[StubNode] def add_variable(self, node): """Add a variable node to this node. :sig: (VariableNode) -> None :param node: Variable node to add. """ if node.name not in self.variable_names: self.variables.append(node) self.variable_names.add(node.name) node.parent = self def add_child(self, node): """Add a function/method or class node to this node. :sig: (Union[FunctionNode, ClassNode]) -> None :param node: Function or class node to add. """ self.children.append(node) node.parent = self def get_code(self): """Get the stub code for this node. The stub code for a node consists of the type annotations of its variables, followed by the prototypes of its functions/methods and classes. :sig: () -> List[str] :return: Lines of stub code for this node. """ stub = [] for child in self.variables: stub.extend(child.get_code()) if ( (len(self.variables) > 0) and (len(self.children) > 0) and (not isinstance(self, ClassNode)) ): stub.append("") for child in self.children: stub.extend(child.get_code()) return stub class VariableNode(StubNode): """A node representing an assignment in a stub tree.""" def __init__(self, name, type_): """Initialize this variable node. :sig: (str, str) -> None :param name: Name of variable that is being assigned to. :param type_: Type of variable. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.type_ = type_ # sig: str def get_code(self): """Get the type annotation for this variable. :sig: () -> List[str] :return: Lines of stub code for this variable. """ return ["%(n)s = ... # type: %(t)s" % {"n": self.name, "t": self.type_}] class FunctionNode(StubNode): """A node representing a function in a stub tree.""" def __init__(self, name, parameters, rtype, decorators=None): """Initialize this function node. The parameters have to given as a list of triples where each item specifies the name of the parameter, its type, and whether it has a default value or not. :sig: (str, Sequence[Tuple[str, str, bool]], str, Optional[Sequence[str]]) -> None :param name: Name of function. :param parameters: List of parameter triples (name, type, has_default). :param rtype: Type of return value. :param decorators: Decorators of function. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.parameters = parameters # sig: Sequence[Tuple[str, str, bool]] self.rtype = rtype # sig: str self.decorators = decorators if decorators is not None else [] # sig: Sequence[str] self._async = False # sig: bool def get_code(self): """Get the stub code for this function. :sig: () -> List[str] :return: Lines of stub code for this function. """ stub = [] for deco in self.decorators: if (deco in DECORATORS) or deco.endswith(".setter"): stub.append("@" + deco) parameters = [] for name, type_, has_default in self.parameters: decl = "%(n)s%(t)s%(d)s" % { "n": name, "t": ": " + type_ if type_ else "", "d": " = ..." if has_default else "", } parameters.append(decl) slots = { "a": "async " if self._async else "", "n": self.name, "p": ", ".join(parameters), "r": self.rtype, } prototype = "%(a)sdef %(n)s(%(p)s) -> %(r)s: ..." % slots if len(prototype) <= LINE_LENGTH_LIMIT: stub.append(prototype) elif len(INDENT + slots["p"]) <= LINE_LENGTH_LIMIT: stub.append("%(a)sdef %(n)s(" % slots) stub.append(INDENT + slots["p"]) stub.append(") -> %(r)s: ..." % slots) else: stub.append("%(a)sdef %(n)s(" % slots) for param in parameters: stub.append(INDENT + param + ",") stub.append(") -> %(r)s: ..." % slots) return stub class ClassNode(StubNode): """A node representing a class in a stub tree.""" def __init__(self, name, bases, signature=None): """Initialize this class node. :sig: (str, Sequence[str], Optional[str]) -> None :param name: Name of class. :param bases: Base classes of class. :param signature: Signature of class, to be used in __init__ method. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.bases = bases # sig: Sequence[str] self.signature = signature # sig: Optional[str] def get_code(self): """Get the stub code for this class. :sig: () -> List[str] :return: Lines of stub code for this class. """ stub = [] bases = ("(" + ", ".join(self.bases) + ")") if len(self.bases) > 0 else "" slots = {"n": self.name, "b": bases} if (len(self.children) == 0) and (len(self.variables) == 0): stub.append("class %(n)s%(b)s: ..." % slots) else: stub.append("class %(n)s%(b)s:" % slots) super_code = super().get_code() if PY3 else StubNode.get_code(self) for line in super_code: stub.append(INDENT + line) return stub def get_aliases(lines): """Get the type aliases in the source. :sig: (Sequence[str]) -> Dict[str, str] :param lines: Lines of the source code. :return: Aliases and their their definitions. """ aliases = {} for line in lines: line = line.strip() if len(line) > 0 and line.startswith(SIG_ALIAS): _, content = line.split(SIG_ALIAS) alias, signature = [t.strip() for t in content.split("=")] aliases[alias] = signature return aliases class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue() def get_stub(source, generic=False): """Get the stub code for a source code. :sig: (str, bool) -> str :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. :return: Generated stub code. """ generator = StubGenerator(source, generic=generic) stub = generator.generate_stub() return stub def get_mod_paths(mod_name, out_dir): """Get source and stub paths for a module.""" paths = [] try: mod = get_loader(mod_name) source = Path(mod.path) if source.name.endswith(".py"): source_rel = Path(*mod_name.split(".")) if source.name == "__init__.py": source_rel = source_rel.joinpath("__init__.py") destination = Path(out_dir, source_rel.with_suffix(".pyi")) paths.append((source, destination)) except Exception as e: _logger.debug(e) _logger.warning("cannot handle module, skipping: %s", mod_name) return paths def get_pkg_paths(pkg_name, out_dir): """Recursively get all source and stub paths for a package.""" paths = [] try: pkg = import_module(pkg_name) if not hasattr(pkg, "__path__"): return get_mod_paths(pkg_name, out_dir) for mod_info in walk_packages(pkg.__path__, pkg.__name__ + "."): mod_paths = get_mod_paths(mod_info.name, out_dir) paths.extend(mod_paths) except Exception as e: _logger.debug(e) _logger.warning("cannot handle package, skipping: %s", pkg_name) return paths ############################################################ # SPHINX ############################################################ def process_docstring(app, what, name, obj, options, lines): """Modify the docstring before generating documentation. This will insert type declarations for parameters and return type into the docstring, and remove the signature field so that it will be excluded from the generated document. """ aliases = getattr(app, "_sigaliases", None) if aliases is None: if what == "module": aliases = get_aliases(inspect.getsource(obj).splitlines()) app._sigaliases = aliases sig_marker = ":" + SIG_FIELD + ":" is_class = what in ("class", "exception") signature = extract_signature("\n".join(lines)) if signature is None: if not is_class: return init_method = getattr(obj, "__init__") init_doc = init_method.__doc__ init_lines = init_doc.splitlines()[1:] if len(init_lines) > 1: init_doc = textwrap.dedent("\n".join(init_lines[1:])) init_lines = init_doc.splitlines() if sig_marker not in init_doc: return sig_started = False for line in init_lines: if line.lstrip().startswith(sig_marker): sig_started = True if sig_started: lines.append(line) signature = extract_signature("\n".join(lines)) if is_class: obj = init_method param_types, rtype, _ = parse_signature(signature) param_names = [p for p in inspect.signature(obj).parameters] if is_class and (param_names[0] == "self"): del param_names[0] # if something goes wrong, don't insert parameter types if len(param_names) == len(param_types): for name, type_ in zip(param_names, param_types): find = ":param %(name)s:" % {"name": name} alias = aliases.get(type_) if alias is not None: type_ = "*%(type)s* :sup:`%(alias)s`" % {"type": type_, "alias": alias} for i, line in enumerate(lines): if line.startswith(find): lines.insert(i, ":type %(name)s: %(type)s" % {"name": name, "type": type_}) break if not is_class: for i, line in enumerate(lines): if line.startswith((":return:", ":returns:")): lines.insert(i, ":rtype: " + rtype) break # remove the signature field sig_start = 0 while sig_start < len(lines): if lines[sig_start].startswith(sig_marker): break sig_start += 1 sig_end = sig_start + 1 while sig_end < len(lines): if (not lines[sig_end]) or (lines[sig_end][0] != " "): break sig_end += 1 for i in reversed(range(sig_start, sig_end)): del lines[i] def setup(app): """Register to Sphinx.""" app.connect("autodoc-process-docstring", process_docstring) return {"version": __version__} ############################################################ # MAIN ############################################################ def main(argv=None): """Start the command line interface.""" parser = ArgumentParser(prog="pygenstub") parser.add_argument("--version", action="version", version="%(prog)s " + __version__) parser.add_argument("files", nargs="*", help="generate stubs for given files") parser.add_argument( "-m", "--module", action="append", metavar="MODULE", dest="modules", default=[], help="generate stubs for given modules", ) parser.add_argument( "-o", "--output", metavar="PATH", dest="out_dir", help="change the output directory" ) parser.add_argument( "--generic", action="store_true", default=False, help="generate generic stubs" ) parser.add_argument("--debug", action="store_true", help="enable debug messages") argv = argv if argv is not None else sys.argv arguments = parser.parse_args(argv[1:]) # set debug mode if arguments.debug: logging.basicConfig(level=logging.DEBUG) _logger.debug("running in debug mode") out_dir = arguments.out_dir if arguments.out_dir is not None else "" if (out_dir == "") and (len(arguments.modules) > 0): print("Output directory must be given when generating stubs for modules.") sys.exit(1) modules = [] for path in arguments.files: paths = Path(path).glob("**/*.py") if Path(path).is_dir() else [Path(path)] for source in paths: if str(source).startswith(os.path.pardir): source = source.absolute().resolve() if (out_dir != "") and source.is_absolute(): source = source.relative_to(source.root) destination = Path(out_dir, source.with_suffix(".pyi")) modules.append((source, destination)) for mod_name in arguments.modules: modules.extend(get_pkg_paths(mod_name, out_dir)) for source, destination in modules: _logger.info("generating stub for %s to path %s", source, destination) with source.open() as f: code = f.read() try: stub = get_stub(code, generic=arguments.generic) except Exception as e: print(source, "-", e, file=sys.stderr) continue if stub != "": if not destination.parent.exists(): destination.parent.mkdir(parents=True) with destination.open("w") as f: f.write("# " + EDIT_WARNING + "\n\n" + stub) if __name__ == "__main__": main()
uyar/pygenstub
pygenstub.py
parse_signature
python
def parse_signature(signature): if " -> " not in signature: # signature comment: no parameters, treat variable type as return type param_types, return_type = None, signature.strip() else: lhs, return_type = [s.strip() for s in signature.split(" -> ")] csv = lhs[1:-1].strip() # remove the parentheses around the parameter type list param_types = split_parameter_types(csv) requires = set(_RE_QUALIFIED_TYPES.findall(signature)) return param_types, return_type, requires
Parse a signature into its input and return parameter types. This will also collect the types that are required by any of the input and return types. :sig: (str) -> Tuple[List[str], str, Set[str]] :param signature: Signature to parse. :return: Input parameter types, return type, and all required types.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L161-L179
[ "def split_parameter_types(parameters):\n \"\"\"Split a parameter types declaration into individual types.\n\n The input is the left hand side of a signature (the part before the arrow),\n excluding the parentheses.\n\n :sig: (str) -> List[str]\n :param parameters: Comma separated parameter types.\n ...
# Copyright (C) 2016-2019 H. Turgut Uyar <uyar@tekir.org> # # pygenstub is free software: you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # pygenstub 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 General Public License for more details. # # You should have received a copy of the GNU General Public License # along with pygenstub. If not, see <http://www.gnu.org/licenses/>. """pygenstub is a utility for generating stub files from docstrings in source files. It takes a source file as input and creates a stub file with the same base name and the ``.pyi`` extension. For more information, please refer to the documentation: https://pygenstub.tekir.org/ """ from __future__ import absolute_import, division, print_function, unicode_literals import ast import inspect import logging import os import re import sys import textwrap from argparse import ArgumentParser from bisect import bisect from collections import OrderedDict from importlib import import_module from io import StringIO from pkgutil import get_loader, walk_packages from docutils.core import publish_doctree __version__ = "1.4.0" # sig: str PY3 = sys.version_info >= (3, 0) if not PY3: import __builtin__ as builtins from pathlib2 import Path else: import builtins from pathlib import Path # sigalias: Document = docutils.nodes.document BUILTIN_TYPES = {k for k, t in builtins.__dict__.items() if isinstance(t, type)} BUILTIN_TYPES.add("None") SIG_FIELD = "sig" # sig: str SIG_COMMENT = "# sig:" # sig: str SIG_ALIAS = "# sigalias:" # sig: str DECORATORS = {"property", "staticmethod", "classmethod"} # sig: Set[str] LINE_LENGTH_LIMIT = 79 INDENT = 4 * " " EDIT_WARNING = "THIS FILE IS AUTOMATICALLY GENERATED, DO NOT EDIT MANUALLY." _RE_QUALIFIED_TYPES = re.compile(r"\w+(?:\.\w+)*") _RE_COMMENT_IN_STRING = re.compile(r"""['"]\s*%(text)s\s*.*['"]""" % {"text": SIG_COMMENT}) _logger = logging.getLogger(__name__) def get_fields(node, fields_tag="field_list"): """Get the field names and their values from a node. :sig: (Document, str) -> Dict[str, str] :param node: Node to get the fields from. :param fields_tag: Tag of child node that contains the fields. :return: Names and values of fields. """ fields_nodes = [c for c in node.children if c.tagname == fields_tag] if len(fields_nodes) == 0: return {} assert len(fields_nodes) == 1, "multiple nodes with tag " + fields_tag fields_node = fields_nodes[0] fields = [ {f.tagname: f.rawsource.strip() for f in n.children} for n in fields_node.children if n.tagname == "field" ] return {f["field_name"]: f["field_body"] for f in fields} def extract_signature(docstring): """Extract the signature from a docstring. :sig: (str) -> Optional[str] :param docstring: Docstring to extract the signature from. :return: Extracted signature, or ``None`` if there's no signature. """ root = publish_doctree(docstring, settings_overrides={"report_level": 5}) fields = get_fields(root) return fields.get(SIG_FIELD) def get_signature(node): """Get the signature of a function or a class. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef]) -> Optional[str] :param node: Node to get the signature from. :return: Value of signature field in node docstring, or ``None`` if there's no signature. """ docstring = ast.get_docstring(node) if docstring is None: return None return extract_signature(docstring) def split_parameter_types(parameters): """Split a parameter types declaration into individual types. The input is the left hand side of a signature (the part before the arrow), excluding the parentheses. :sig: (str) -> List[str] :param parameters: Comma separated parameter types. :return: Parameter types. """ if parameters == "": return [] # only consider the top level commas, ignore the ones in [] commas = [] bracket_depth = 0 for i, char in enumerate(parameters): if (char == ",") and (bracket_depth == 0): commas.append(i) elif char == "[": bracket_depth += 1 elif char == "]": bracket_depth -= 1 types = [] last_i = 0 for i in commas: types.append(parameters[last_i:i].strip()) last_i = i + 1 else: types.append(parameters[last_i:].strip()) return types class StubNode: """A node in a stub tree.""" def __init__(self): """Initialize this stub node. :sig: () -> None """ self.variables = [] # sig: List[VariableNode] self.variable_names = set() # sig: Set[str] self.children = [] # sig: List[Union[FunctionNode, ClassNode]] self.parent = None # sig: Optional[StubNode] def add_variable(self, node): """Add a variable node to this node. :sig: (VariableNode) -> None :param node: Variable node to add. """ if node.name not in self.variable_names: self.variables.append(node) self.variable_names.add(node.name) node.parent = self def add_child(self, node): """Add a function/method or class node to this node. :sig: (Union[FunctionNode, ClassNode]) -> None :param node: Function or class node to add. """ self.children.append(node) node.parent = self def get_code(self): """Get the stub code for this node. The stub code for a node consists of the type annotations of its variables, followed by the prototypes of its functions/methods and classes. :sig: () -> List[str] :return: Lines of stub code for this node. """ stub = [] for child in self.variables: stub.extend(child.get_code()) if ( (len(self.variables) > 0) and (len(self.children) > 0) and (not isinstance(self, ClassNode)) ): stub.append("") for child in self.children: stub.extend(child.get_code()) return stub class VariableNode(StubNode): """A node representing an assignment in a stub tree.""" def __init__(self, name, type_): """Initialize this variable node. :sig: (str, str) -> None :param name: Name of variable that is being assigned to. :param type_: Type of variable. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.type_ = type_ # sig: str def get_code(self): """Get the type annotation for this variable. :sig: () -> List[str] :return: Lines of stub code for this variable. """ return ["%(n)s = ... # type: %(t)s" % {"n": self.name, "t": self.type_}] class FunctionNode(StubNode): """A node representing a function in a stub tree.""" def __init__(self, name, parameters, rtype, decorators=None): """Initialize this function node. The parameters have to given as a list of triples where each item specifies the name of the parameter, its type, and whether it has a default value or not. :sig: (str, Sequence[Tuple[str, str, bool]], str, Optional[Sequence[str]]) -> None :param name: Name of function. :param parameters: List of parameter triples (name, type, has_default). :param rtype: Type of return value. :param decorators: Decorators of function. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.parameters = parameters # sig: Sequence[Tuple[str, str, bool]] self.rtype = rtype # sig: str self.decorators = decorators if decorators is not None else [] # sig: Sequence[str] self._async = False # sig: bool def get_code(self): """Get the stub code for this function. :sig: () -> List[str] :return: Lines of stub code for this function. """ stub = [] for deco in self.decorators: if (deco in DECORATORS) or deco.endswith(".setter"): stub.append("@" + deco) parameters = [] for name, type_, has_default in self.parameters: decl = "%(n)s%(t)s%(d)s" % { "n": name, "t": ": " + type_ if type_ else "", "d": " = ..." if has_default else "", } parameters.append(decl) slots = { "a": "async " if self._async else "", "n": self.name, "p": ", ".join(parameters), "r": self.rtype, } prototype = "%(a)sdef %(n)s(%(p)s) -> %(r)s: ..." % slots if len(prototype) <= LINE_LENGTH_LIMIT: stub.append(prototype) elif len(INDENT + slots["p"]) <= LINE_LENGTH_LIMIT: stub.append("%(a)sdef %(n)s(" % slots) stub.append(INDENT + slots["p"]) stub.append(") -> %(r)s: ..." % slots) else: stub.append("%(a)sdef %(n)s(" % slots) for param in parameters: stub.append(INDENT + param + ",") stub.append(") -> %(r)s: ..." % slots) return stub class ClassNode(StubNode): """A node representing a class in a stub tree.""" def __init__(self, name, bases, signature=None): """Initialize this class node. :sig: (str, Sequence[str], Optional[str]) -> None :param name: Name of class. :param bases: Base classes of class. :param signature: Signature of class, to be used in __init__ method. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.bases = bases # sig: Sequence[str] self.signature = signature # sig: Optional[str] def get_code(self): """Get the stub code for this class. :sig: () -> List[str] :return: Lines of stub code for this class. """ stub = [] bases = ("(" + ", ".join(self.bases) + ")") if len(self.bases) > 0 else "" slots = {"n": self.name, "b": bases} if (len(self.children) == 0) and (len(self.variables) == 0): stub.append("class %(n)s%(b)s: ..." % slots) else: stub.append("class %(n)s%(b)s:" % slots) super_code = super().get_code() if PY3 else StubNode.get_code(self) for line in super_code: stub.append(INDENT + line) return stub def get_aliases(lines): """Get the type aliases in the source. :sig: (Sequence[str]) -> Dict[str, str] :param lines: Lines of the source code. :return: Aliases and their their definitions. """ aliases = {} for line in lines: line = line.strip() if len(line) > 0 and line.startswith(SIG_ALIAS): _, content = line.split(SIG_ALIAS) alias, signature = [t.strip() for t in content.split("=")] aliases[alias] = signature return aliases class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue() def get_stub(source, generic=False): """Get the stub code for a source code. :sig: (str, bool) -> str :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. :return: Generated stub code. """ generator = StubGenerator(source, generic=generic) stub = generator.generate_stub() return stub def get_mod_paths(mod_name, out_dir): """Get source and stub paths for a module.""" paths = [] try: mod = get_loader(mod_name) source = Path(mod.path) if source.name.endswith(".py"): source_rel = Path(*mod_name.split(".")) if source.name == "__init__.py": source_rel = source_rel.joinpath("__init__.py") destination = Path(out_dir, source_rel.with_suffix(".pyi")) paths.append((source, destination)) except Exception as e: _logger.debug(e) _logger.warning("cannot handle module, skipping: %s", mod_name) return paths def get_pkg_paths(pkg_name, out_dir): """Recursively get all source and stub paths for a package.""" paths = [] try: pkg = import_module(pkg_name) if not hasattr(pkg, "__path__"): return get_mod_paths(pkg_name, out_dir) for mod_info in walk_packages(pkg.__path__, pkg.__name__ + "."): mod_paths = get_mod_paths(mod_info.name, out_dir) paths.extend(mod_paths) except Exception as e: _logger.debug(e) _logger.warning("cannot handle package, skipping: %s", pkg_name) return paths ############################################################ # SPHINX ############################################################ def process_docstring(app, what, name, obj, options, lines): """Modify the docstring before generating documentation. This will insert type declarations for parameters and return type into the docstring, and remove the signature field so that it will be excluded from the generated document. """ aliases = getattr(app, "_sigaliases", None) if aliases is None: if what == "module": aliases = get_aliases(inspect.getsource(obj).splitlines()) app._sigaliases = aliases sig_marker = ":" + SIG_FIELD + ":" is_class = what in ("class", "exception") signature = extract_signature("\n".join(lines)) if signature is None: if not is_class: return init_method = getattr(obj, "__init__") init_doc = init_method.__doc__ init_lines = init_doc.splitlines()[1:] if len(init_lines) > 1: init_doc = textwrap.dedent("\n".join(init_lines[1:])) init_lines = init_doc.splitlines() if sig_marker not in init_doc: return sig_started = False for line in init_lines: if line.lstrip().startswith(sig_marker): sig_started = True if sig_started: lines.append(line) signature = extract_signature("\n".join(lines)) if is_class: obj = init_method param_types, rtype, _ = parse_signature(signature) param_names = [p for p in inspect.signature(obj).parameters] if is_class and (param_names[0] == "self"): del param_names[0] # if something goes wrong, don't insert parameter types if len(param_names) == len(param_types): for name, type_ in zip(param_names, param_types): find = ":param %(name)s:" % {"name": name} alias = aliases.get(type_) if alias is not None: type_ = "*%(type)s* :sup:`%(alias)s`" % {"type": type_, "alias": alias} for i, line in enumerate(lines): if line.startswith(find): lines.insert(i, ":type %(name)s: %(type)s" % {"name": name, "type": type_}) break if not is_class: for i, line in enumerate(lines): if line.startswith((":return:", ":returns:")): lines.insert(i, ":rtype: " + rtype) break # remove the signature field sig_start = 0 while sig_start < len(lines): if lines[sig_start].startswith(sig_marker): break sig_start += 1 sig_end = sig_start + 1 while sig_end < len(lines): if (not lines[sig_end]) or (lines[sig_end][0] != " "): break sig_end += 1 for i in reversed(range(sig_start, sig_end)): del lines[i] def setup(app): """Register to Sphinx.""" app.connect("autodoc-process-docstring", process_docstring) return {"version": __version__} ############################################################ # MAIN ############################################################ def main(argv=None): """Start the command line interface.""" parser = ArgumentParser(prog="pygenstub") parser.add_argument("--version", action="version", version="%(prog)s " + __version__) parser.add_argument("files", nargs="*", help="generate stubs for given files") parser.add_argument( "-m", "--module", action="append", metavar="MODULE", dest="modules", default=[], help="generate stubs for given modules", ) parser.add_argument( "-o", "--output", metavar="PATH", dest="out_dir", help="change the output directory" ) parser.add_argument( "--generic", action="store_true", default=False, help="generate generic stubs" ) parser.add_argument("--debug", action="store_true", help="enable debug messages") argv = argv if argv is not None else sys.argv arguments = parser.parse_args(argv[1:]) # set debug mode if arguments.debug: logging.basicConfig(level=logging.DEBUG) _logger.debug("running in debug mode") out_dir = arguments.out_dir if arguments.out_dir is not None else "" if (out_dir == "") and (len(arguments.modules) > 0): print("Output directory must be given when generating stubs for modules.") sys.exit(1) modules = [] for path in arguments.files: paths = Path(path).glob("**/*.py") if Path(path).is_dir() else [Path(path)] for source in paths: if str(source).startswith(os.path.pardir): source = source.absolute().resolve() if (out_dir != "") and source.is_absolute(): source = source.relative_to(source.root) destination = Path(out_dir, source.with_suffix(".pyi")) modules.append((source, destination)) for mod_name in arguments.modules: modules.extend(get_pkg_paths(mod_name, out_dir)) for source, destination in modules: _logger.info("generating stub for %s to path %s", source, destination) with source.open() as f: code = f.read() try: stub = get_stub(code, generic=arguments.generic) except Exception as e: print(source, "-", e, file=sys.stderr) continue if stub != "": if not destination.parent.exists(): destination.parent.mkdir(parents=True) with destination.open("w") as f: f.write("# " + EDIT_WARNING + "\n\n" + stub) if __name__ == "__main__": main()
uyar/pygenstub
pygenstub.py
get_aliases
python
def get_aliases(lines): aliases = {} for line in lines: line = line.strip() if len(line) > 0 and line.startswith(SIG_ALIAS): _, content = line.split(SIG_ALIAS) alias, signature = [t.strip() for t in content.split("=")] aliases[alias] = signature return aliases
Get the type aliases in the source. :sig: (Sequence[str]) -> Dict[str, str] :param lines: Lines of the source code. :return: Aliases and their their definitions.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L372-L386
null
# Copyright (C) 2016-2019 H. Turgut Uyar <uyar@tekir.org> # # pygenstub is free software: you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # pygenstub 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 General Public License for more details. # # You should have received a copy of the GNU General Public License # along with pygenstub. If not, see <http://www.gnu.org/licenses/>. """pygenstub is a utility for generating stub files from docstrings in source files. It takes a source file as input and creates a stub file with the same base name and the ``.pyi`` extension. For more information, please refer to the documentation: https://pygenstub.tekir.org/ """ from __future__ import absolute_import, division, print_function, unicode_literals import ast import inspect import logging import os import re import sys import textwrap from argparse import ArgumentParser from bisect import bisect from collections import OrderedDict from importlib import import_module from io import StringIO from pkgutil import get_loader, walk_packages from docutils.core import publish_doctree __version__ = "1.4.0" # sig: str PY3 = sys.version_info >= (3, 0) if not PY3: import __builtin__ as builtins from pathlib2 import Path else: import builtins from pathlib import Path # sigalias: Document = docutils.nodes.document BUILTIN_TYPES = {k for k, t in builtins.__dict__.items() if isinstance(t, type)} BUILTIN_TYPES.add("None") SIG_FIELD = "sig" # sig: str SIG_COMMENT = "# sig:" # sig: str SIG_ALIAS = "# sigalias:" # sig: str DECORATORS = {"property", "staticmethod", "classmethod"} # sig: Set[str] LINE_LENGTH_LIMIT = 79 INDENT = 4 * " " EDIT_WARNING = "THIS FILE IS AUTOMATICALLY GENERATED, DO NOT EDIT MANUALLY." _RE_QUALIFIED_TYPES = re.compile(r"\w+(?:\.\w+)*") _RE_COMMENT_IN_STRING = re.compile(r"""['"]\s*%(text)s\s*.*['"]""" % {"text": SIG_COMMENT}) _logger = logging.getLogger(__name__) def get_fields(node, fields_tag="field_list"): """Get the field names and their values from a node. :sig: (Document, str) -> Dict[str, str] :param node: Node to get the fields from. :param fields_tag: Tag of child node that contains the fields. :return: Names and values of fields. """ fields_nodes = [c for c in node.children if c.tagname == fields_tag] if len(fields_nodes) == 0: return {} assert len(fields_nodes) == 1, "multiple nodes with tag " + fields_tag fields_node = fields_nodes[0] fields = [ {f.tagname: f.rawsource.strip() for f in n.children} for n in fields_node.children if n.tagname == "field" ] return {f["field_name"]: f["field_body"] for f in fields} def extract_signature(docstring): """Extract the signature from a docstring. :sig: (str) -> Optional[str] :param docstring: Docstring to extract the signature from. :return: Extracted signature, or ``None`` if there's no signature. """ root = publish_doctree(docstring, settings_overrides={"report_level": 5}) fields = get_fields(root) return fields.get(SIG_FIELD) def get_signature(node): """Get the signature of a function or a class. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef]) -> Optional[str] :param node: Node to get the signature from. :return: Value of signature field in node docstring, or ``None`` if there's no signature. """ docstring = ast.get_docstring(node) if docstring is None: return None return extract_signature(docstring) def split_parameter_types(parameters): """Split a parameter types declaration into individual types. The input is the left hand side of a signature (the part before the arrow), excluding the parentheses. :sig: (str) -> List[str] :param parameters: Comma separated parameter types. :return: Parameter types. """ if parameters == "": return [] # only consider the top level commas, ignore the ones in [] commas = [] bracket_depth = 0 for i, char in enumerate(parameters): if (char == ",") and (bracket_depth == 0): commas.append(i) elif char == "[": bracket_depth += 1 elif char == "]": bracket_depth -= 1 types = [] last_i = 0 for i in commas: types.append(parameters[last_i:i].strip()) last_i = i + 1 else: types.append(parameters[last_i:].strip()) return types def parse_signature(signature): """Parse a signature into its input and return parameter types. This will also collect the types that are required by any of the input and return types. :sig: (str) -> Tuple[List[str], str, Set[str]] :param signature: Signature to parse. :return: Input parameter types, return type, and all required types. """ if " -> " not in signature: # signature comment: no parameters, treat variable type as return type param_types, return_type = None, signature.strip() else: lhs, return_type = [s.strip() for s in signature.split(" -> ")] csv = lhs[1:-1].strip() # remove the parentheses around the parameter type list param_types = split_parameter_types(csv) requires = set(_RE_QUALIFIED_TYPES.findall(signature)) return param_types, return_type, requires class StubNode: """A node in a stub tree.""" def __init__(self): """Initialize this stub node. :sig: () -> None """ self.variables = [] # sig: List[VariableNode] self.variable_names = set() # sig: Set[str] self.children = [] # sig: List[Union[FunctionNode, ClassNode]] self.parent = None # sig: Optional[StubNode] def add_variable(self, node): """Add a variable node to this node. :sig: (VariableNode) -> None :param node: Variable node to add. """ if node.name not in self.variable_names: self.variables.append(node) self.variable_names.add(node.name) node.parent = self def add_child(self, node): """Add a function/method or class node to this node. :sig: (Union[FunctionNode, ClassNode]) -> None :param node: Function or class node to add. """ self.children.append(node) node.parent = self def get_code(self): """Get the stub code for this node. The stub code for a node consists of the type annotations of its variables, followed by the prototypes of its functions/methods and classes. :sig: () -> List[str] :return: Lines of stub code for this node. """ stub = [] for child in self.variables: stub.extend(child.get_code()) if ( (len(self.variables) > 0) and (len(self.children) > 0) and (not isinstance(self, ClassNode)) ): stub.append("") for child in self.children: stub.extend(child.get_code()) return stub class VariableNode(StubNode): """A node representing an assignment in a stub tree.""" def __init__(self, name, type_): """Initialize this variable node. :sig: (str, str) -> None :param name: Name of variable that is being assigned to. :param type_: Type of variable. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.type_ = type_ # sig: str def get_code(self): """Get the type annotation for this variable. :sig: () -> List[str] :return: Lines of stub code for this variable. """ return ["%(n)s = ... # type: %(t)s" % {"n": self.name, "t": self.type_}] class FunctionNode(StubNode): """A node representing a function in a stub tree.""" def __init__(self, name, parameters, rtype, decorators=None): """Initialize this function node. The parameters have to given as a list of triples where each item specifies the name of the parameter, its type, and whether it has a default value or not. :sig: (str, Sequence[Tuple[str, str, bool]], str, Optional[Sequence[str]]) -> None :param name: Name of function. :param parameters: List of parameter triples (name, type, has_default). :param rtype: Type of return value. :param decorators: Decorators of function. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.parameters = parameters # sig: Sequence[Tuple[str, str, bool]] self.rtype = rtype # sig: str self.decorators = decorators if decorators is not None else [] # sig: Sequence[str] self._async = False # sig: bool def get_code(self): """Get the stub code for this function. :sig: () -> List[str] :return: Lines of stub code for this function. """ stub = [] for deco in self.decorators: if (deco in DECORATORS) or deco.endswith(".setter"): stub.append("@" + deco) parameters = [] for name, type_, has_default in self.parameters: decl = "%(n)s%(t)s%(d)s" % { "n": name, "t": ": " + type_ if type_ else "", "d": " = ..." if has_default else "", } parameters.append(decl) slots = { "a": "async " if self._async else "", "n": self.name, "p": ", ".join(parameters), "r": self.rtype, } prototype = "%(a)sdef %(n)s(%(p)s) -> %(r)s: ..." % slots if len(prototype) <= LINE_LENGTH_LIMIT: stub.append(prototype) elif len(INDENT + slots["p"]) <= LINE_LENGTH_LIMIT: stub.append("%(a)sdef %(n)s(" % slots) stub.append(INDENT + slots["p"]) stub.append(") -> %(r)s: ..." % slots) else: stub.append("%(a)sdef %(n)s(" % slots) for param in parameters: stub.append(INDENT + param + ",") stub.append(") -> %(r)s: ..." % slots) return stub class ClassNode(StubNode): """A node representing a class in a stub tree.""" def __init__(self, name, bases, signature=None): """Initialize this class node. :sig: (str, Sequence[str], Optional[str]) -> None :param name: Name of class. :param bases: Base classes of class. :param signature: Signature of class, to be used in __init__ method. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.bases = bases # sig: Sequence[str] self.signature = signature # sig: Optional[str] def get_code(self): """Get the stub code for this class. :sig: () -> List[str] :return: Lines of stub code for this class. """ stub = [] bases = ("(" + ", ".join(self.bases) + ")") if len(self.bases) > 0 else "" slots = {"n": self.name, "b": bases} if (len(self.children) == 0) and (len(self.variables) == 0): stub.append("class %(n)s%(b)s: ..." % slots) else: stub.append("class %(n)s%(b)s:" % slots) super_code = super().get_code() if PY3 else StubNode.get_code(self) for line in super_code: stub.append(INDENT + line) return stub class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue() def get_stub(source, generic=False): """Get the stub code for a source code. :sig: (str, bool) -> str :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. :return: Generated stub code. """ generator = StubGenerator(source, generic=generic) stub = generator.generate_stub() return stub def get_mod_paths(mod_name, out_dir): """Get source and stub paths for a module.""" paths = [] try: mod = get_loader(mod_name) source = Path(mod.path) if source.name.endswith(".py"): source_rel = Path(*mod_name.split(".")) if source.name == "__init__.py": source_rel = source_rel.joinpath("__init__.py") destination = Path(out_dir, source_rel.with_suffix(".pyi")) paths.append((source, destination)) except Exception as e: _logger.debug(e) _logger.warning("cannot handle module, skipping: %s", mod_name) return paths def get_pkg_paths(pkg_name, out_dir): """Recursively get all source and stub paths for a package.""" paths = [] try: pkg = import_module(pkg_name) if not hasattr(pkg, "__path__"): return get_mod_paths(pkg_name, out_dir) for mod_info in walk_packages(pkg.__path__, pkg.__name__ + "."): mod_paths = get_mod_paths(mod_info.name, out_dir) paths.extend(mod_paths) except Exception as e: _logger.debug(e) _logger.warning("cannot handle package, skipping: %s", pkg_name) return paths ############################################################ # SPHINX ############################################################ def process_docstring(app, what, name, obj, options, lines): """Modify the docstring before generating documentation. This will insert type declarations for parameters and return type into the docstring, and remove the signature field so that it will be excluded from the generated document. """ aliases = getattr(app, "_sigaliases", None) if aliases is None: if what == "module": aliases = get_aliases(inspect.getsource(obj).splitlines()) app._sigaliases = aliases sig_marker = ":" + SIG_FIELD + ":" is_class = what in ("class", "exception") signature = extract_signature("\n".join(lines)) if signature is None: if not is_class: return init_method = getattr(obj, "__init__") init_doc = init_method.__doc__ init_lines = init_doc.splitlines()[1:] if len(init_lines) > 1: init_doc = textwrap.dedent("\n".join(init_lines[1:])) init_lines = init_doc.splitlines() if sig_marker not in init_doc: return sig_started = False for line in init_lines: if line.lstrip().startswith(sig_marker): sig_started = True if sig_started: lines.append(line) signature = extract_signature("\n".join(lines)) if is_class: obj = init_method param_types, rtype, _ = parse_signature(signature) param_names = [p for p in inspect.signature(obj).parameters] if is_class and (param_names[0] == "self"): del param_names[0] # if something goes wrong, don't insert parameter types if len(param_names) == len(param_types): for name, type_ in zip(param_names, param_types): find = ":param %(name)s:" % {"name": name} alias = aliases.get(type_) if alias is not None: type_ = "*%(type)s* :sup:`%(alias)s`" % {"type": type_, "alias": alias} for i, line in enumerate(lines): if line.startswith(find): lines.insert(i, ":type %(name)s: %(type)s" % {"name": name, "type": type_}) break if not is_class: for i, line in enumerate(lines): if line.startswith((":return:", ":returns:")): lines.insert(i, ":rtype: " + rtype) break # remove the signature field sig_start = 0 while sig_start < len(lines): if lines[sig_start].startswith(sig_marker): break sig_start += 1 sig_end = sig_start + 1 while sig_end < len(lines): if (not lines[sig_end]) or (lines[sig_end][0] != " "): break sig_end += 1 for i in reversed(range(sig_start, sig_end)): del lines[i] def setup(app): """Register to Sphinx.""" app.connect("autodoc-process-docstring", process_docstring) return {"version": __version__} ############################################################ # MAIN ############################################################ def main(argv=None): """Start the command line interface.""" parser = ArgumentParser(prog="pygenstub") parser.add_argument("--version", action="version", version="%(prog)s " + __version__) parser.add_argument("files", nargs="*", help="generate stubs for given files") parser.add_argument( "-m", "--module", action="append", metavar="MODULE", dest="modules", default=[], help="generate stubs for given modules", ) parser.add_argument( "-o", "--output", metavar="PATH", dest="out_dir", help="change the output directory" ) parser.add_argument( "--generic", action="store_true", default=False, help="generate generic stubs" ) parser.add_argument("--debug", action="store_true", help="enable debug messages") argv = argv if argv is not None else sys.argv arguments = parser.parse_args(argv[1:]) # set debug mode if arguments.debug: logging.basicConfig(level=logging.DEBUG) _logger.debug("running in debug mode") out_dir = arguments.out_dir if arguments.out_dir is not None else "" if (out_dir == "") and (len(arguments.modules) > 0): print("Output directory must be given when generating stubs for modules.") sys.exit(1) modules = [] for path in arguments.files: paths = Path(path).glob("**/*.py") if Path(path).is_dir() else [Path(path)] for source in paths: if str(source).startswith(os.path.pardir): source = source.absolute().resolve() if (out_dir != "") and source.is_absolute(): source = source.relative_to(source.root) destination = Path(out_dir, source.with_suffix(".pyi")) modules.append((source, destination)) for mod_name in arguments.modules: modules.extend(get_pkg_paths(mod_name, out_dir)) for source, destination in modules: _logger.info("generating stub for %s to path %s", source, destination) with source.open() as f: code = f.read() try: stub = get_stub(code, generic=arguments.generic) except Exception as e: print(source, "-", e, file=sys.stderr) continue if stub != "": if not destination.parent.exists(): destination.parent.mkdir(parents=True) with destination.open("w") as f: f.write("# " + EDIT_WARNING + "\n\n" + stub) if __name__ == "__main__": main()
uyar/pygenstub
pygenstub.py
get_stub
python
def get_stub(source, generic=False): generator = StubGenerator(source, generic=generic) stub = generator.generate_stub() return stub
Get the stub code for a source code. :sig: (str, bool) -> str :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. :return: Generated stub code.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L745-L755
[ "def generate_stub(self):\n \"\"\"Generate the stub code for this source.\n\n :sig: () -> str\n :return: Generated stub code.\n \"\"\"\n needed_types = self.required_types - BUILTIN_TYPES\n\n needed_types -= self.defined_types\n _logger.debug(\"defined types: %s\", self.defined_types)\n\n mo...
# Copyright (C) 2016-2019 H. Turgut Uyar <uyar@tekir.org> # # pygenstub is free software: you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # pygenstub 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 General Public License for more details. # # You should have received a copy of the GNU General Public License # along with pygenstub. If not, see <http://www.gnu.org/licenses/>. """pygenstub is a utility for generating stub files from docstrings in source files. It takes a source file as input and creates a stub file with the same base name and the ``.pyi`` extension. For more information, please refer to the documentation: https://pygenstub.tekir.org/ """ from __future__ import absolute_import, division, print_function, unicode_literals import ast import inspect import logging import os import re import sys import textwrap from argparse import ArgumentParser from bisect import bisect from collections import OrderedDict from importlib import import_module from io import StringIO from pkgutil import get_loader, walk_packages from docutils.core import publish_doctree __version__ = "1.4.0" # sig: str PY3 = sys.version_info >= (3, 0) if not PY3: import __builtin__ as builtins from pathlib2 import Path else: import builtins from pathlib import Path # sigalias: Document = docutils.nodes.document BUILTIN_TYPES = {k for k, t in builtins.__dict__.items() if isinstance(t, type)} BUILTIN_TYPES.add("None") SIG_FIELD = "sig" # sig: str SIG_COMMENT = "# sig:" # sig: str SIG_ALIAS = "# sigalias:" # sig: str DECORATORS = {"property", "staticmethod", "classmethod"} # sig: Set[str] LINE_LENGTH_LIMIT = 79 INDENT = 4 * " " EDIT_WARNING = "THIS FILE IS AUTOMATICALLY GENERATED, DO NOT EDIT MANUALLY." _RE_QUALIFIED_TYPES = re.compile(r"\w+(?:\.\w+)*") _RE_COMMENT_IN_STRING = re.compile(r"""['"]\s*%(text)s\s*.*['"]""" % {"text": SIG_COMMENT}) _logger = logging.getLogger(__name__) def get_fields(node, fields_tag="field_list"): """Get the field names and their values from a node. :sig: (Document, str) -> Dict[str, str] :param node: Node to get the fields from. :param fields_tag: Tag of child node that contains the fields. :return: Names and values of fields. """ fields_nodes = [c for c in node.children if c.tagname == fields_tag] if len(fields_nodes) == 0: return {} assert len(fields_nodes) == 1, "multiple nodes with tag " + fields_tag fields_node = fields_nodes[0] fields = [ {f.tagname: f.rawsource.strip() for f in n.children} for n in fields_node.children if n.tagname == "field" ] return {f["field_name"]: f["field_body"] for f in fields} def extract_signature(docstring): """Extract the signature from a docstring. :sig: (str) -> Optional[str] :param docstring: Docstring to extract the signature from. :return: Extracted signature, or ``None`` if there's no signature. """ root = publish_doctree(docstring, settings_overrides={"report_level": 5}) fields = get_fields(root) return fields.get(SIG_FIELD) def get_signature(node): """Get the signature of a function or a class. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef]) -> Optional[str] :param node: Node to get the signature from. :return: Value of signature field in node docstring, or ``None`` if there's no signature. """ docstring = ast.get_docstring(node) if docstring is None: return None return extract_signature(docstring) def split_parameter_types(parameters): """Split a parameter types declaration into individual types. The input is the left hand side of a signature (the part before the arrow), excluding the parentheses. :sig: (str) -> List[str] :param parameters: Comma separated parameter types. :return: Parameter types. """ if parameters == "": return [] # only consider the top level commas, ignore the ones in [] commas = [] bracket_depth = 0 for i, char in enumerate(parameters): if (char == ",") and (bracket_depth == 0): commas.append(i) elif char == "[": bracket_depth += 1 elif char == "]": bracket_depth -= 1 types = [] last_i = 0 for i in commas: types.append(parameters[last_i:i].strip()) last_i = i + 1 else: types.append(parameters[last_i:].strip()) return types def parse_signature(signature): """Parse a signature into its input and return parameter types. This will also collect the types that are required by any of the input and return types. :sig: (str) -> Tuple[List[str], str, Set[str]] :param signature: Signature to parse. :return: Input parameter types, return type, and all required types. """ if " -> " not in signature: # signature comment: no parameters, treat variable type as return type param_types, return_type = None, signature.strip() else: lhs, return_type = [s.strip() for s in signature.split(" -> ")] csv = lhs[1:-1].strip() # remove the parentheses around the parameter type list param_types = split_parameter_types(csv) requires = set(_RE_QUALIFIED_TYPES.findall(signature)) return param_types, return_type, requires class StubNode: """A node in a stub tree.""" def __init__(self): """Initialize this stub node. :sig: () -> None """ self.variables = [] # sig: List[VariableNode] self.variable_names = set() # sig: Set[str] self.children = [] # sig: List[Union[FunctionNode, ClassNode]] self.parent = None # sig: Optional[StubNode] def add_variable(self, node): """Add a variable node to this node. :sig: (VariableNode) -> None :param node: Variable node to add. """ if node.name not in self.variable_names: self.variables.append(node) self.variable_names.add(node.name) node.parent = self def add_child(self, node): """Add a function/method or class node to this node. :sig: (Union[FunctionNode, ClassNode]) -> None :param node: Function or class node to add. """ self.children.append(node) node.parent = self def get_code(self): """Get the stub code for this node. The stub code for a node consists of the type annotations of its variables, followed by the prototypes of its functions/methods and classes. :sig: () -> List[str] :return: Lines of stub code for this node. """ stub = [] for child in self.variables: stub.extend(child.get_code()) if ( (len(self.variables) > 0) and (len(self.children) > 0) and (not isinstance(self, ClassNode)) ): stub.append("") for child in self.children: stub.extend(child.get_code()) return stub class VariableNode(StubNode): """A node representing an assignment in a stub tree.""" def __init__(self, name, type_): """Initialize this variable node. :sig: (str, str) -> None :param name: Name of variable that is being assigned to. :param type_: Type of variable. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.type_ = type_ # sig: str def get_code(self): """Get the type annotation for this variable. :sig: () -> List[str] :return: Lines of stub code for this variable. """ return ["%(n)s = ... # type: %(t)s" % {"n": self.name, "t": self.type_}] class FunctionNode(StubNode): """A node representing a function in a stub tree.""" def __init__(self, name, parameters, rtype, decorators=None): """Initialize this function node. The parameters have to given as a list of triples where each item specifies the name of the parameter, its type, and whether it has a default value or not. :sig: (str, Sequence[Tuple[str, str, bool]], str, Optional[Sequence[str]]) -> None :param name: Name of function. :param parameters: List of parameter triples (name, type, has_default). :param rtype: Type of return value. :param decorators: Decorators of function. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.parameters = parameters # sig: Sequence[Tuple[str, str, bool]] self.rtype = rtype # sig: str self.decorators = decorators if decorators is not None else [] # sig: Sequence[str] self._async = False # sig: bool def get_code(self): """Get the stub code for this function. :sig: () -> List[str] :return: Lines of stub code for this function. """ stub = [] for deco in self.decorators: if (deco in DECORATORS) or deco.endswith(".setter"): stub.append("@" + deco) parameters = [] for name, type_, has_default in self.parameters: decl = "%(n)s%(t)s%(d)s" % { "n": name, "t": ": " + type_ if type_ else "", "d": " = ..." if has_default else "", } parameters.append(decl) slots = { "a": "async " if self._async else "", "n": self.name, "p": ", ".join(parameters), "r": self.rtype, } prototype = "%(a)sdef %(n)s(%(p)s) -> %(r)s: ..." % slots if len(prototype) <= LINE_LENGTH_LIMIT: stub.append(prototype) elif len(INDENT + slots["p"]) <= LINE_LENGTH_LIMIT: stub.append("%(a)sdef %(n)s(" % slots) stub.append(INDENT + slots["p"]) stub.append(") -> %(r)s: ..." % slots) else: stub.append("%(a)sdef %(n)s(" % slots) for param in parameters: stub.append(INDENT + param + ",") stub.append(") -> %(r)s: ..." % slots) return stub class ClassNode(StubNode): """A node representing a class in a stub tree.""" def __init__(self, name, bases, signature=None): """Initialize this class node. :sig: (str, Sequence[str], Optional[str]) -> None :param name: Name of class. :param bases: Base classes of class. :param signature: Signature of class, to be used in __init__ method. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.bases = bases # sig: Sequence[str] self.signature = signature # sig: Optional[str] def get_code(self): """Get the stub code for this class. :sig: () -> List[str] :return: Lines of stub code for this class. """ stub = [] bases = ("(" + ", ".join(self.bases) + ")") if len(self.bases) > 0 else "" slots = {"n": self.name, "b": bases} if (len(self.children) == 0) and (len(self.variables) == 0): stub.append("class %(n)s%(b)s: ..." % slots) else: stub.append("class %(n)s%(b)s:" % slots) super_code = super().get_code() if PY3 else StubNode.get_code(self) for line in super_code: stub.append(INDENT + line) return stub def get_aliases(lines): """Get the type aliases in the source. :sig: (Sequence[str]) -> Dict[str, str] :param lines: Lines of the source code. :return: Aliases and their their definitions. """ aliases = {} for line in lines: line = line.strip() if len(line) > 0 and line.startswith(SIG_ALIAS): _, content = line.split(SIG_ALIAS) alias, signature = [t.strip() for t in content.split("=")] aliases[alias] = signature return aliases class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue() def get_mod_paths(mod_name, out_dir): """Get source and stub paths for a module.""" paths = [] try: mod = get_loader(mod_name) source = Path(mod.path) if source.name.endswith(".py"): source_rel = Path(*mod_name.split(".")) if source.name == "__init__.py": source_rel = source_rel.joinpath("__init__.py") destination = Path(out_dir, source_rel.with_suffix(".pyi")) paths.append((source, destination)) except Exception as e: _logger.debug(e) _logger.warning("cannot handle module, skipping: %s", mod_name) return paths def get_pkg_paths(pkg_name, out_dir): """Recursively get all source and stub paths for a package.""" paths = [] try: pkg = import_module(pkg_name) if not hasattr(pkg, "__path__"): return get_mod_paths(pkg_name, out_dir) for mod_info in walk_packages(pkg.__path__, pkg.__name__ + "."): mod_paths = get_mod_paths(mod_info.name, out_dir) paths.extend(mod_paths) except Exception as e: _logger.debug(e) _logger.warning("cannot handle package, skipping: %s", pkg_name) return paths ############################################################ # SPHINX ############################################################ def process_docstring(app, what, name, obj, options, lines): """Modify the docstring before generating documentation. This will insert type declarations for parameters and return type into the docstring, and remove the signature field so that it will be excluded from the generated document. """ aliases = getattr(app, "_sigaliases", None) if aliases is None: if what == "module": aliases = get_aliases(inspect.getsource(obj).splitlines()) app._sigaliases = aliases sig_marker = ":" + SIG_FIELD + ":" is_class = what in ("class", "exception") signature = extract_signature("\n".join(lines)) if signature is None: if not is_class: return init_method = getattr(obj, "__init__") init_doc = init_method.__doc__ init_lines = init_doc.splitlines()[1:] if len(init_lines) > 1: init_doc = textwrap.dedent("\n".join(init_lines[1:])) init_lines = init_doc.splitlines() if sig_marker not in init_doc: return sig_started = False for line in init_lines: if line.lstrip().startswith(sig_marker): sig_started = True if sig_started: lines.append(line) signature = extract_signature("\n".join(lines)) if is_class: obj = init_method param_types, rtype, _ = parse_signature(signature) param_names = [p for p in inspect.signature(obj).parameters] if is_class and (param_names[0] == "self"): del param_names[0] # if something goes wrong, don't insert parameter types if len(param_names) == len(param_types): for name, type_ in zip(param_names, param_types): find = ":param %(name)s:" % {"name": name} alias = aliases.get(type_) if alias is not None: type_ = "*%(type)s* :sup:`%(alias)s`" % {"type": type_, "alias": alias} for i, line in enumerate(lines): if line.startswith(find): lines.insert(i, ":type %(name)s: %(type)s" % {"name": name, "type": type_}) break if not is_class: for i, line in enumerate(lines): if line.startswith((":return:", ":returns:")): lines.insert(i, ":rtype: " + rtype) break # remove the signature field sig_start = 0 while sig_start < len(lines): if lines[sig_start].startswith(sig_marker): break sig_start += 1 sig_end = sig_start + 1 while sig_end < len(lines): if (not lines[sig_end]) or (lines[sig_end][0] != " "): break sig_end += 1 for i in reversed(range(sig_start, sig_end)): del lines[i] def setup(app): """Register to Sphinx.""" app.connect("autodoc-process-docstring", process_docstring) return {"version": __version__} ############################################################ # MAIN ############################################################ def main(argv=None): """Start the command line interface.""" parser = ArgumentParser(prog="pygenstub") parser.add_argument("--version", action="version", version="%(prog)s " + __version__) parser.add_argument("files", nargs="*", help="generate stubs for given files") parser.add_argument( "-m", "--module", action="append", metavar="MODULE", dest="modules", default=[], help="generate stubs for given modules", ) parser.add_argument( "-o", "--output", metavar="PATH", dest="out_dir", help="change the output directory" ) parser.add_argument( "--generic", action="store_true", default=False, help="generate generic stubs" ) parser.add_argument("--debug", action="store_true", help="enable debug messages") argv = argv if argv is not None else sys.argv arguments = parser.parse_args(argv[1:]) # set debug mode if arguments.debug: logging.basicConfig(level=logging.DEBUG) _logger.debug("running in debug mode") out_dir = arguments.out_dir if arguments.out_dir is not None else "" if (out_dir == "") and (len(arguments.modules) > 0): print("Output directory must be given when generating stubs for modules.") sys.exit(1) modules = [] for path in arguments.files: paths = Path(path).glob("**/*.py") if Path(path).is_dir() else [Path(path)] for source in paths: if str(source).startswith(os.path.pardir): source = source.absolute().resolve() if (out_dir != "") and source.is_absolute(): source = source.relative_to(source.root) destination = Path(out_dir, source.with_suffix(".pyi")) modules.append((source, destination)) for mod_name in arguments.modules: modules.extend(get_pkg_paths(mod_name, out_dir)) for source, destination in modules: _logger.info("generating stub for %s to path %s", source, destination) with source.open() as f: code = f.read() try: stub = get_stub(code, generic=arguments.generic) except Exception as e: print(source, "-", e, file=sys.stderr) continue if stub != "": if not destination.parent.exists(): destination.parent.mkdir(parents=True) with destination.open("w") as f: f.write("# " + EDIT_WARNING + "\n\n" + stub) if __name__ == "__main__": main()
uyar/pygenstub
pygenstub.py
get_mod_paths
python
def get_mod_paths(mod_name, out_dir): paths = [] try: mod = get_loader(mod_name) source = Path(mod.path) if source.name.endswith(".py"): source_rel = Path(*mod_name.split(".")) if source.name == "__init__.py": source_rel = source_rel.joinpath("__init__.py") destination = Path(out_dir, source_rel.with_suffix(".pyi")) paths.append((source, destination)) except Exception as e: _logger.debug(e) _logger.warning("cannot handle module, skipping: %s", mod_name) return paths
Get source and stub paths for a module.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L758-L773
null
# Copyright (C) 2016-2019 H. Turgut Uyar <uyar@tekir.org> # # pygenstub is free software: you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # pygenstub 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 General Public License for more details. # # You should have received a copy of the GNU General Public License # along with pygenstub. If not, see <http://www.gnu.org/licenses/>. """pygenstub is a utility for generating stub files from docstrings in source files. It takes a source file as input and creates a stub file with the same base name and the ``.pyi`` extension. For more information, please refer to the documentation: https://pygenstub.tekir.org/ """ from __future__ import absolute_import, division, print_function, unicode_literals import ast import inspect import logging import os import re import sys import textwrap from argparse import ArgumentParser from bisect import bisect from collections import OrderedDict from importlib import import_module from io import StringIO from pkgutil import get_loader, walk_packages from docutils.core import publish_doctree __version__ = "1.4.0" # sig: str PY3 = sys.version_info >= (3, 0) if not PY3: import __builtin__ as builtins from pathlib2 import Path else: import builtins from pathlib import Path # sigalias: Document = docutils.nodes.document BUILTIN_TYPES = {k for k, t in builtins.__dict__.items() if isinstance(t, type)} BUILTIN_TYPES.add("None") SIG_FIELD = "sig" # sig: str SIG_COMMENT = "# sig:" # sig: str SIG_ALIAS = "# sigalias:" # sig: str DECORATORS = {"property", "staticmethod", "classmethod"} # sig: Set[str] LINE_LENGTH_LIMIT = 79 INDENT = 4 * " " EDIT_WARNING = "THIS FILE IS AUTOMATICALLY GENERATED, DO NOT EDIT MANUALLY." _RE_QUALIFIED_TYPES = re.compile(r"\w+(?:\.\w+)*") _RE_COMMENT_IN_STRING = re.compile(r"""['"]\s*%(text)s\s*.*['"]""" % {"text": SIG_COMMENT}) _logger = logging.getLogger(__name__) def get_fields(node, fields_tag="field_list"): """Get the field names and their values from a node. :sig: (Document, str) -> Dict[str, str] :param node: Node to get the fields from. :param fields_tag: Tag of child node that contains the fields. :return: Names and values of fields. """ fields_nodes = [c for c in node.children if c.tagname == fields_tag] if len(fields_nodes) == 0: return {} assert len(fields_nodes) == 1, "multiple nodes with tag " + fields_tag fields_node = fields_nodes[0] fields = [ {f.tagname: f.rawsource.strip() for f in n.children} for n in fields_node.children if n.tagname == "field" ] return {f["field_name"]: f["field_body"] for f in fields} def extract_signature(docstring): """Extract the signature from a docstring. :sig: (str) -> Optional[str] :param docstring: Docstring to extract the signature from. :return: Extracted signature, or ``None`` if there's no signature. """ root = publish_doctree(docstring, settings_overrides={"report_level": 5}) fields = get_fields(root) return fields.get(SIG_FIELD) def get_signature(node): """Get the signature of a function or a class. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef]) -> Optional[str] :param node: Node to get the signature from. :return: Value of signature field in node docstring, or ``None`` if there's no signature. """ docstring = ast.get_docstring(node) if docstring is None: return None return extract_signature(docstring) def split_parameter_types(parameters): """Split a parameter types declaration into individual types. The input is the left hand side of a signature (the part before the arrow), excluding the parentheses. :sig: (str) -> List[str] :param parameters: Comma separated parameter types. :return: Parameter types. """ if parameters == "": return [] # only consider the top level commas, ignore the ones in [] commas = [] bracket_depth = 0 for i, char in enumerate(parameters): if (char == ",") and (bracket_depth == 0): commas.append(i) elif char == "[": bracket_depth += 1 elif char == "]": bracket_depth -= 1 types = [] last_i = 0 for i in commas: types.append(parameters[last_i:i].strip()) last_i = i + 1 else: types.append(parameters[last_i:].strip()) return types def parse_signature(signature): """Parse a signature into its input and return parameter types. This will also collect the types that are required by any of the input and return types. :sig: (str) -> Tuple[List[str], str, Set[str]] :param signature: Signature to parse. :return: Input parameter types, return type, and all required types. """ if " -> " not in signature: # signature comment: no parameters, treat variable type as return type param_types, return_type = None, signature.strip() else: lhs, return_type = [s.strip() for s in signature.split(" -> ")] csv = lhs[1:-1].strip() # remove the parentheses around the parameter type list param_types = split_parameter_types(csv) requires = set(_RE_QUALIFIED_TYPES.findall(signature)) return param_types, return_type, requires class StubNode: """A node in a stub tree.""" def __init__(self): """Initialize this stub node. :sig: () -> None """ self.variables = [] # sig: List[VariableNode] self.variable_names = set() # sig: Set[str] self.children = [] # sig: List[Union[FunctionNode, ClassNode]] self.parent = None # sig: Optional[StubNode] def add_variable(self, node): """Add a variable node to this node. :sig: (VariableNode) -> None :param node: Variable node to add. """ if node.name not in self.variable_names: self.variables.append(node) self.variable_names.add(node.name) node.parent = self def add_child(self, node): """Add a function/method or class node to this node. :sig: (Union[FunctionNode, ClassNode]) -> None :param node: Function or class node to add. """ self.children.append(node) node.parent = self def get_code(self): """Get the stub code for this node. The stub code for a node consists of the type annotations of its variables, followed by the prototypes of its functions/methods and classes. :sig: () -> List[str] :return: Lines of stub code for this node. """ stub = [] for child in self.variables: stub.extend(child.get_code()) if ( (len(self.variables) > 0) and (len(self.children) > 0) and (not isinstance(self, ClassNode)) ): stub.append("") for child in self.children: stub.extend(child.get_code()) return stub class VariableNode(StubNode): """A node representing an assignment in a stub tree.""" def __init__(self, name, type_): """Initialize this variable node. :sig: (str, str) -> None :param name: Name of variable that is being assigned to. :param type_: Type of variable. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.type_ = type_ # sig: str def get_code(self): """Get the type annotation for this variable. :sig: () -> List[str] :return: Lines of stub code for this variable. """ return ["%(n)s = ... # type: %(t)s" % {"n": self.name, "t": self.type_}] class FunctionNode(StubNode): """A node representing a function in a stub tree.""" def __init__(self, name, parameters, rtype, decorators=None): """Initialize this function node. The parameters have to given as a list of triples where each item specifies the name of the parameter, its type, and whether it has a default value or not. :sig: (str, Sequence[Tuple[str, str, bool]], str, Optional[Sequence[str]]) -> None :param name: Name of function. :param parameters: List of parameter triples (name, type, has_default). :param rtype: Type of return value. :param decorators: Decorators of function. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.parameters = parameters # sig: Sequence[Tuple[str, str, bool]] self.rtype = rtype # sig: str self.decorators = decorators if decorators is not None else [] # sig: Sequence[str] self._async = False # sig: bool def get_code(self): """Get the stub code for this function. :sig: () -> List[str] :return: Lines of stub code for this function. """ stub = [] for deco in self.decorators: if (deco in DECORATORS) or deco.endswith(".setter"): stub.append("@" + deco) parameters = [] for name, type_, has_default in self.parameters: decl = "%(n)s%(t)s%(d)s" % { "n": name, "t": ": " + type_ if type_ else "", "d": " = ..." if has_default else "", } parameters.append(decl) slots = { "a": "async " if self._async else "", "n": self.name, "p": ", ".join(parameters), "r": self.rtype, } prototype = "%(a)sdef %(n)s(%(p)s) -> %(r)s: ..." % slots if len(prototype) <= LINE_LENGTH_LIMIT: stub.append(prototype) elif len(INDENT + slots["p"]) <= LINE_LENGTH_LIMIT: stub.append("%(a)sdef %(n)s(" % slots) stub.append(INDENT + slots["p"]) stub.append(") -> %(r)s: ..." % slots) else: stub.append("%(a)sdef %(n)s(" % slots) for param in parameters: stub.append(INDENT + param + ",") stub.append(") -> %(r)s: ..." % slots) return stub class ClassNode(StubNode): """A node representing a class in a stub tree.""" def __init__(self, name, bases, signature=None): """Initialize this class node. :sig: (str, Sequence[str], Optional[str]) -> None :param name: Name of class. :param bases: Base classes of class. :param signature: Signature of class, to be used in __init__ method. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.bases = bases # sig: Sequence[str] self.signature = signature # sig: Optional[str] def get_code(self): """Get the stub code for this class. :sig: () -> List[str] :return: Lines of stub code for this class. """ stub = [] bases = ("(" + ", ".join(self.bases) + ")") if len(self.bases) > 0 else "" slots = {"n": self.name, "b": bases} if (len(self.children) == 0) and (len(self.variables) == 0): stub.append("class %(n)s%(b)s: ..." % slots) else: stub.append("class %(n)s%(b)s:" % slots) super_code = super().get_code() if PY3 else StubNode.get_code(self) for line in super_code: stub.append(INDENT + line) return stub def get_aliases(lines): """Get the type aliases in the source. :sig: (Sequence[str]) -> Dict[str, str] :param lines: Lines of the source code. :return: Aliases and their their definitions. """ aliases = {} for line in lines: line = line.strip() if len(line) > 0 and line.startswith(SIG_ALIAS): _, content = line.split(SIG_ALIAS) alias, signature = [t.strip() for t in content.split("=")] aliases[alias] = signature return aliases class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue() def get_stub(source, generic=False): """Get the stub code for a source code. :sig: (str, bool) -> str :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. :return: Generated stub code. """ generator = StubGenerator(source, generic=generic) stub = generator.generate_stub() return stub def get_pkg_paths(pkg_name, out_dir): """Recursively get all source and stub paths for a package.""" paths = [] try: pkg = import_module(pkg_name) if not hasattr(pkg, "__path__"): return get_mod_paths(pkg_name, out_dir) for mod_info in walk_packages(pkg.__path__, pkg.__name__ + "."): mod_paths = get_mod_paths(mod_info.name, out_dir) paths.extend(mod_paths) except Exception as e: _logger.debug(e) _logger.warning("cannot handle package, skipping: %s", pkg_name) return paths ############################################################ # SPHINX ############################################################ def process_docstring(app, what, name, obj, options, lines): """Modify the docstring before generating documentation. This will insert type declarations for parameters and return type into the docstring, and remove the signature field so that it will be excluded from the generated document. """ aliases = getattr(app, "_sigaliases", None) if aliases is None: if what == "module": aliases = get_aliases(inspect.getsource(obj).splitlines()) app._sigaliases = aliases sig_marker = ":" + SIG_FIELD + ":" is_class = what in ("class", "exception") signature = extract_signature("\n".join(lines)) if signature is None: if not is_class: return init_method = getattr(obj, "__init__") init_doc = init_method.__doc__ init_lines = init_doc.splitlines()[1:] if len(init_lines) > 1: init_doc = textwrap.dedent("\n".join(init_lines[1:])) init_lines = init_doc.splitlines() if sig_marker not in init_doc: return sig_started = False for line in init_lines: if line.lstrip().startswith(sig_marker): sig_started = True if sig_started: lines.append(line) signature = extract_signature("\n".join(lines)) if is_class: obj = init_method param_types, rtype, _ = parse_signature(signature) param_names = [p for p in inspect.signature(obj).parameters] if is_class and (param_names[0] == "self"): del param_names[0] # if something goes wrong, don't insert parameter types if len(param_names) == len(param_types): for name, type_ in zip(param_names, param_types): find = ":param %(name)s:" % {"name": name} alias = aliases.get(type_) if alias is not None: type_ = "*%(type)s* :sup:`%(alias)s`" % {"type": type_, "alias": alias} for i, line in enumerate(lines): if line.startswith(find): lines.insert(i, ":type %(name)s: %(type)s" % {"name": name, "type": type_}) break if not is_class: for i, line in enumerate(lines): if line.startswith((":return:", ":returns:")): lines.insert(i, ":rtype: " + rtype) break # remove the signature field sig_start = 0 while sig_start < len(lines): if lines[sig_start].startswith(sig_marker): break sig_start += 1 sig_end = sig_start + 1 while sig_end < len(lines): if (not lines[sig_end]) or (lines[sig_end][0] != " "): break sig_end += 1 for i in reversed(range(sig_start, sig_end)): del lines[i] def setup(app): """Register to Sphinx.""" app.connect("autodoc-process-docstring", process_docstring) return {"version": __version__} ############################################################ # MAIN ############################################################ def main(argv=None): """Start the command line interface.""" parser = ArgumentParser(prog="pygenstub") parser.add_argument("--version", action="version", version="%(prog)s " + __version__) parser.add_argument("files", nargs="*", help="generate stubs for given files") parser.add_argument( "-m", "--module", action="append", metavar="MODULE", dest="modules", default=[], help="generate stubs for given modules", ) parser.add_argument( "-o", "--output", metavar="PATH", dest="out_dir", help="change the output directory" ) parser.add_argument( "--generic", action="store_true", default=False, help="generate generic stubs" ) parser.add_argument("--debug", action="store_true", help="enable debug messages") argv = argv if argv is not None else sys.argv arguments = parser.parse_args(argv[1:]) # set debug mode if arguments.debug: logging.basicConfig(level=logging.DEBUG) _logger.debug("running in debug mode") out_dir = arguments.out_dir if arguments.out_dir is not None else "" if (out_dir == "") and (len(arguments.modules) > 0): print("Output directory must be given when generating stubs for modules.") sys.exit(1) modules = [] for path in arguments.files: paths = Path(path).glob("**/*.py") if Path(path).is_dir() else [Path(path)] for source in paths: if str(source).startswith(os.path.pardir): source = source.absolute().resolve() if (out_dir != "") and source.is_absolute(): source = source.relative_to(source.root) destination = Path(out_dir, source.with_suffix(".pyi")) modules.append((source, destination)) for mod_name in arguments.modules: modules.extend(get_pkg_paths(mod_name, out_dir)) for source, destination in modules: _logger.info("generating stub for %s to path %s", source, destination) with source.open() as f: code = f.read() try: stub = get_stub(code, generic=arguments.generic) except Exception as e: print(source, "-", e, file=sys.stderr) continue if stub != "": if not destination.parent.exists(): destination.parent.mkdir(parents=True) with destination.open("w") as f: f.write("# " + EDIT_WARNING + "\n\n" + stub) if __name__ == "__main__": main()
uyar/pygenstub
pygenstub.py
get_pkg_paths
python
def get_pkg_paths(pkg_name, out_dir): paths = [] try: pkg = import_module(pkg_name) if not hasattr(pkg, "__path__"): return get_mod_paths(pkg_name, out_dir) for mod_info in walk_packages(pkg.__path__, pkg.__name__ + "."): mod_paths = get_mod_paths(mod_info.name, out_dir) paths.extend(mod_paths) except Exception as e: _logger.debug(e) _logger.warning("cannot handle package, skipping: %s", pkg_name) return paths
Recursively get all source and stub paths for a package.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L776-L789
[ "def get_mod_paths(mod_name, out_dir):\n \"\"\"Get source and stub paths for a module.\"\"\"\n paths = []\n try:\n mod = get_loader(mod_name)\n source = Path(mod.path)\n if source.name.endswith(\".py\"):\n source_rel = Path(*mod_name.split(\".\"))\n if source.name...
# Copyright (C) 2016-2019 H. Turgut Uyar <uyar@tekir.org> # # pygenstub is free software: you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # pygenstub 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 General Public License for more details. # # You should have received a copy of the GNU General Public License # along with pygenstub. If not, see <http://www.gnu.org/licenses/>. """pygenstub is a utility for generating stub files from docstrings in source files. It takes a source file as input and creates a stub file with the same base name and the ``.pyi`` extension. For more information, please refer to the documentation: https://pygenstub.tekir.org/ """ from __future__ import absolute_import, division, print_function, unicode_literals import ast import inspect import logging import os import re import sys import textwrap from argparse import ArgumentParser from bisect import bisect from collections import OrderedDict from importlib import import_module from io import StringIO from pkgutil import get_loader, walk_packages from docutils.core import publish_doctree __version__ = "1.4.0" # sig: str PY3 = sys.version_info >= (3, 0) if not PY3: import __builtin__ as builtins from pathlib2 import Path else: import builtins from pathlib import Path # sigalias: Document = docutils.nodes.document BUILTIN_TYPES = {k for k, t in builtins.__dict__.items() if isinstance(t, type)} BUILTIN_TYPES.add("None") SIG_FIELD = "sig" # sig: str SIG_COMMENT = "# sig:" # sig: str SIG_ALIAS = "# sigalias:" # sig: str DECORATORS = {"property", "staticmethod", "classmethod"} # sig: Set[str] LINE_LENGTH_LIMIT = 79 INDENT = 4 * " " EDIT_WARNING = "THIS FILE IS AUTOMATICALLY GENERATED, DO NOT EDIT MANUALLY." _RE_QUALIFIED_TYPES = re.compile(r"\w+(?:\.\w+)*") _RE_COMMENT_IN_STRING = re.compile(r"""['"]\s*%(text)s\s*.*['"]""" % {"text": SIG_COMMENT}) _logger = logging.getLogger(__name__) def get_fields(node, fields_tag="field_list"): """Get the field names and their values from a node. :sig: (Document, str) -> Dict[str, str] :param node: Node to get the fields from. :param fields_tag: Tag of child node that contains the fields. :return: Names and values of fields. """ fields_nodes = [c for c in node.children if c.tagname == fields_tag] if len(fields_nodes) == 0: return {} assert len(fields_nodes) == 1, "multiple nodes with tag " + fields_tag fields_node = fields_nodes[0] fields = [ {f.tagname: f.rawsource.strip() for f in n.children} for n in fields_node.children if n.tagname == "field" ] return {f["field_name"]: f["field_body"] for f in fields} def extract_signature(docstring): """Extract the signature from a docstring. :sig: (str) -> Optional[str] :param docstring: Docstring to extract the signature from. :return: Extracted signature, or ``None`` if there's no signature. """ root = publish_doctree(docstring, settings_overrides={"report_level": 5}) fields = get_fields(root) return fields.get(SIG_FIELD) def get_signature(node): """Get the signature of a function or a class. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef]) -> Optional[str] :param node: Node to get the signature from. :return: Value of signature field in node docstring, or ``None`` if there's no signature. """ docstring = ast.get_docstring(node) if docstring is None: return None return extract_signature(docstring) def split_parameter_types(parameters): """Split a parameter types declaration into individual types. The input is the left hand side of a signature (the part before the arrow), excluding the parentheses. :sig: (str) -> List[str] :param parameters: Comma separated parameter types. :return: Parameter types. """ if parameters == "": return [] # only consider the top level commas, ignore the ones in [] commas = [] bracket_depth = 0 for i, char in enumerate(parameters): if (char == ",") and (bracket_depth == 0): commas.append(i) elif char == "[": bracket_depth += 1 elif char == "]": bracket_depth -= 1 types = [] last_i = 0 for i in commas: types.append(parameters[last_i:i].strip()) last_i = i + 1 else: types.append(parameters[last_i:].strip()) return types def parse_signature(signature): """Parse a signature into its input and return parameter types. This will also collect the types that are required by any of the input and return types. :sig: (str) -> Tuple[List[str], str, Set[str]] :param signature: Signature to parse. :return: Input parameter types, return type, and all required types. """ if " -> " not in signature: # signature comment: no parameters, treat variable type as return type param_types, return_type = None, signature.strip() else: lhs, return_type = [s.strip() for s in signature.split(" -> ")] csv = lhs[1:-1].strip() # remove the parentheses around the parameter type list param_types = split_parameter_types(csv) requires = set(_RE_QUALIFIED_TYPES.findall(signature)) return param_types, return_type, requires class StubNode: """A node in a stub tree.""" def __init__(self): """Initialize this stub node. :sig: () -> None """ self.variables = [] # sig: List[VariableNode] self.variable_names = set() # sig: Set[str] self.children = [] # sig: List[Union[FunctionNode, ClassNode]] self.parent = None # sig: Optional[StubNode] def add_variable(self, node): """Add a variable node to this node. :sig: (VariableNode) -> None :param node: Variable node to add. """ if node.name not in self.variable_names: self.variables.append(node) self.variable_names.add(node.name) node.parent = self def add_child(self, node): """Add a function/method or class node to this node. :sig: (Union[FunctionNode, ClassNode]) -> None :param node: Function or class node to add. """ self.children.append(node) node.parent = self def get_code(self): """Get the stub code for this node. The stub code for a node consists of the type annotations of its variables, followed by the prototypes of its functions/methods and classes. :sig: () -> List[str] :return: Lines of stub code for this node. """ stub = [] for child in self.variables: stub.extend(child.get_code()) if ( (len(self.variables) > 0) and (len(self.children) > 0) and (not isinstance(self, ClassNode)) ): stub.append("") for child in self.children: stub.extend(child.get_code()) return stub class VariableNode(StubNode): """A node representing an assignment in a stub tree.""" def __init__(self, name, type_): """Initialize this variable node. :sig: (str, str) -> None :param name: Name of variable that is being assigned to. :param type_: Type of variable. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.type_ = type_ # sig: str def get_code(self): """Get the type annotation for this variable. :sig: () -> List[str] :return: Lines of stub code for this variable. """ return ["%(n)s = ... # type: %(t)s" % {"n": self.name, "t": self.type_}] class FunctionNode(StubNode): """A node representing a function in a stub tree.""" def __init__(self, name, parameters, rtype, decorators=None): """Initialize this function node. The parameters have to given as a list of triples where each item specifies the name of the parameter, its type, and whether it has a default value or not. :sig: (str, Sequence[Tuple[str, str, bool]], str, Optional[Sequence[str]]) -> None :param name: Name of function. :param parameters: List of parameter triples (name, type, has_default). :param rtype: Type of return value. :param decorators: Decorators of function. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.parameters = parameters # sig: Sequence[Tuple[str, str, bool]] self.rtype = rtype # sig: str self.decorators = decorators if decorators is not None else [] # sig: Sequence[str] self._async = False # sig: bool def get_code(self): """Get the stub code for this function. :sig: () -> List[str] :return: Lines of stub code for this function. """ stub = [] for deco in self.decorators: if (deco in DECORATORS) or deco.endswith(".setter"): stub.append("@" + deco) parameters = [] for name, type_, has_default in self.parameters: decl = "%(n)s%(t)s%(d)s" % { "n": name, "t": ": " + type_ if type_ else "", "d": " = ..." if has_default else "", } parameters.append(decl) slots = { "a": "async " if self._async else "", "n": self.name, "p": ", ".join(parameters), "r": self.rtype, } prototype = "%(a)sdef %(n)s(%(p)s) -> %(r)s: ..." % slots if len(prototype) <= LINE_LENGTH_LIMIT: stub.append(prototype) elif len(INDENT + slots["p"]) <= LINE_LENGTH_LIMIT: stub.append("%(a)sdef %(n)s(" % slots) stub.append(INDENT + slots["p"]) stub.append(") -> %(r)s: ..." % slots) else: stub.append("%(a)sdef %(n)s(" % slots) for param in parameters: stub.append(INDENT + param + ",") stub.append(") -> %(r)s: ..." % slots) return stub class ClassNode(StubNode): """A node representing a class in a stub tree.""" def __init__(self, name, bases, signature=None): """Initialize this class node. :sig: (str, Sequence[str], Optional[str]) -> None :param name: Name of class. :param bases: Base classes of class. :param signature: Signature of class, to be used in __init__ method. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.bases = bases # sig: Sequence[str] self.signature = signature # sig: Optional[str] def get_code(self): """Get the stub code for this class. :sig: () -> List[str] :return: Lines of stub code for this class. """ stub = [] bases = ("(" + ", ".join(self.bases) + ")") if len(self.bases) > 0 else "" slots = {"n": self.name, "b": bases} if (len(self.children) == 0) and (len(self.variables) == 0): stub.append("class %(n)s%(b)s: ..." % slots) else: stub.append("class %(n)s%(b)s:" % slots) super_code = super().get_code() if PY3 else StubNode.get_code(self) for line in super_code: stub.append(INDENT + line) return stub def get_aliases(lines): """Get the type aliases in the source. :sig: (Sequence[str]) -> Dict[str, str] :param lines: Lines of the source code. :return: Aliases and their their definitions. """ aliases = {} for line in lines: line = line.strip() if len(line) > 0 and line.startswith(SIG_ALIAS): _, content = line.split(SIG_ALIAS) alias, signature = [t.strip() for t in content.split("=")] aliases[alias] = signature return aliases class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue() def get_stub(source, generic=False): """Get the stub code for a source code. :sig: (str, bool) -> str :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. :return: Generated stub code. """ generator = StubGenerator(source, generic=generic) stub = generator.generate_stub() return stub def get_mod_paths(mod_name, out_dir): """Get source and stub paths for a module.""" paths = [] try: mod = get_loader(mod_name) source = Path(mod.path) if source.name.endswith(".py"): source_rel = Path(*mod_name.split(".")) if source.name == "__init__.py": source_rel = source_rel.joinpath("__init__.py") destination = Path(out_dir, source_rel.with_suffix(".pyi")) paths.append((source, destination)) except Exception as e: _logger.debug(e) _logger.warning("cannot handle module, skipping: %s", mod_name) return paths ############################################################ # SPHINX ############################################################ def process_docstring(app, what, name, obj, options, lines): """Modify the docstring before generating documentation. This will insert type declarations for parameters and return type into the docstring, and remove the signature field so that it will be excluded from the generated document. """ aliases = getattr(app, "_sigaliases", None) if aliases is None: if what == "module": aliases = get_aliases(inspect.getsource(obj).splitlines()) app._sigaliases = aliases sig_marker = ":" + SIG_FIELD + ":" is_class = what in ("class", "exception") signature = extract_signature("\n".join(lines)) if signature is None: if not is_class: return init_method = getattr(obj, "__init__") init_doc = init_method.__doc__ init_lines = init_doc.splitlines()[1:] if len(init_lines) > 1: init_doc = textwrap.dedent("\n".join(init_lines[1:])) init_lines = init_doc.splitlines() if sig_marker not in init_doc: return sig_started = False for line in init_lines: if line.lstrip().startswith(sig_marker): sig_started = True if sig_started: lines.append(line) signature = extract_signature("\n".join(lines)) if is_class: obj = init_method param_types, rtype, _ = parse_signature(signature) param_names = [p for p in inspect.signature(obj).parameters] if is_class and (param_names[0] == "self"): del param_names[0] # if something goes wrong, don't insert parameter types if len(param_names) == len(param_types): for name, type_ in zip(param_names, param_types): find = ":param %(name)s:" % {"name": name} alias = aliases.get(type_) if alias is not None: type_ = "*%(type)s* :sup:`%(alias)s`" % {"type": type_, "alias": alias} for i, line in enumerate(lines): if line.startswith(find): lines.insert(i, ":type %(name)s: %(type)s" % {"name": name, "type": type_}) break if not is_class: for i, line in enumerate(lines): if line.startswith((":return:", ":returns:")): lines.insert(i, ":rtype: " + rtype) break # remove the signature field sig_start = 0 while sig_start < len(lines): if lines[sig_start].startswith(sig_marker): break sig_start += 1 sig_end = sig_start + 1 while sig_end < len(lines): if (not lines[sig_end]) or (lines[sig_end][0] != " "): break sig_end += 1 for i in reversed(range(sig_start, sig_end)): del lines[i] def setup(app): """Register to Sphinx.""" app.connect("autodoc-process-docstring", process_docstring) return {"version": __version__} ############################################################ # MAIN ############################################################ def main(argv=None): """Start the command line interface.""" parser = ArgumentParser(prog="pygenstub") parser.add_argument("--version", action="version", version="%(prog)s " + __version__) parser.add_argument("files", nargs="*", help="generate stubs for given files") parser.add_argument( "-m", "--module", action="append", metavar="MODULE", dest="modules", default=[], help="generate stubs for given modules", ) parser.add_argument( "-o", "--output", metavar="PATH", dest="out_dir", help="change the output directory" ) parser.add_argument( "--generic", action="store_true", default=False, help="generate generic stubs" ) parser.add_argument("--debug", action="store_true", help="enable debug messages") argv = argv if argv is not None else sys.argv arguments = parser.parse_args(argv[1:]) # set debug mode if arguments.debug: logging.basicConfig(level=logging.DEBUG) _logger.debug("running in debug mode") out_dir = arguments.out_dir if arguments.out_dir is not None else "" if (out_dir == "") and (len(arguments.modules) > 0): print("Output directory must be given when generating stubs for modules.") sys.exit(1) modules = [] for path in arguments.files: paths = Path(path).glob("**/*.py") if Path(path).is_dir() else [Path(path)] for source in paths: if str(source).startswith(os.path.pardir): source = source.absolute().resolve() if (out_dir != "") and source.is_absolute(): source = source.relative_to(source.root) destination = Path(out_dir, source.with_suffix(".pyi")) modules.append((source, destination)) for mod_name in arguments.modules: modules.extend(get_pkg_paths(mod_name, out_dir)) for source, destination in modules: _logger.info("generating stub for %s to path %s", source, destination) with source.open() as f: code = f.read() try: stub = get_stub(code, generic=arguments.generic) except Exception as e: print(source, "-", e, file=sys.stderr) continue if stub != "": if not destination.parent.exists(): destination.parent.mkdir(parents=True) with destination.open("w") as f: f.write("# " + EDIT_WARNING + "\n\n" + stub) if __name__ == "__main__": main()
uyar/pygenstub
pygenstub.py
process_docstring
python
def process_docstring(app, what, name, obj, options, lines): aliases = getattr(app, "_sigaliases", None) if aliases is None: if what == "module": aliases = get_aliases(inspect.getsource(obj).splitlines()) app._sigaliases = aliases sig_marker = ":" + SIG_FIELD + ":" is_class = what in ("class", "exception") signature = extract_signature("\n".join(lines)) if signature is None: if not is_class: return init_method = getattr(obj, "__init__") init_doc = init_method.__doc__ init_lines = init_doc.splitlines()[1:] if len(init_lines) > 1: init_doc = textwrap.dedent("\n".join(init_lines[1:])) init_lines = init_doc.splitlines() if sig_marker not in init_doc: return sig_started = False for line in init_lines: if line.lstrip().startswith(sig_marker): sig_started = True if sig_started: lines.append(line) signature = extract_signature("\n".join(lines)) if is_class: obj = init_method param_types, rtype, _ = parse_signature(signature) param_names = [p for p in inspect.signature(obj).parameters] if is_class and (param_names[0] == "self"): del param_names[0] # if something goes wrong, don't insert parameter types if len(param_names) == len(param_types): for name, type_ in zip(param_names, param_types): find = ":param %(name)s:" % {"name": name} alias = aliases.get(type_) if alias is not None: type_ = "*%(type)s* :sup:`%(alias)s`" % {"type": type_, "alias": alias} for i, line in enumerate(lines): if line.startswith(find): lines.insert(i, ":type %(name)s: %(type)s" % {"name": name, "type": type_}) break if not is_class: for i, line in enumerate(lines): if line.startswith((":return:", ":returns:")): lines.insert(i, ":rtype: " + rtype) break # remove the signature field sig_start = 0 while sig_start < len(lines): if lines[sig_start].startswith(sig_marker): break sig_start += 1 sig_end = sig_start + 1 while sig_end < len(lines): if (not lines[sig_end]) or (lines[sig_end][0] != " "): break sig_end += 1 for i in reversed(range(sig_start, sig_end)): del lines[i]
Modify the docstring before generating documentation. This will insert type declarations for parameters and return type into the docstring, and remove the signature field so that it will be excluded from the generated document.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L797-L874
[ "def extract_signature(docstring):\n \"\"\"Extract the signature from a docstring.\n\n :sig: (str) -> Optional[str]\n :param docstring: Docstring to extract the signature from.\n :return: Extracted signature, or ``None`` if there's no signature.\n \"\"\"\n root = publish_doctree(docstring, setting...
# Copyright (C) 2016-2019 H. Turgut Uyar <uyar@tekir.org> # # pygenstub is free software: you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # pygenstub 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 General Public License for more details. # # You should have received a copy of the GNU General Public License # along with pygenstub. If not, see <http://www.gnu.org/licenses/>. """pygenstub is a utility for generating stub files from docstrings in source files. It takes a source file as input and creates a stub file with the same base name and the ``.pyi`` extension. For more information, please refer to the documentation: https://pygenstub.tekir.org/ """ from __future__ import absolute_import, division, print_function, unicode_literals import ast import inspect import logging import os import re import sys import textwrap from argparse import ArgumentParser from bisect import bisect from collections import OrderedDict from importlib import import_module from io import StringIO from pkgutil import get_loader, walk_packages from docutils.core import publish_doctree __version__ = "1.4.0" # sig: str PY3 = sys.version_info >= (3, 0) if not PY3: import __builtin__ as builtins from pathlib2 import Path else: import builtins from pathlib import Path # sigalias: Document = docutils.nodes.document BUILTIN_TYPES = {k for k, t in builtins.__dict__.items() if isinstance(t, type)} BUILTIN_TYPES.add("None") SIG_FIELD = "sig" # sig: str SIG_COMMENT = "# sig:" # sig: str SIG_ALIAS = "# sigalias:" # sig: str DECORATORS = {"property", "staticmethod", "classmethod"} # sig: Set[str] LINE_LENGTH_LIMIT = 79 INDENT = 4 * " " EDIT_WARNING = "THIS FILE IS AUTOMATICALLY GENERATED, DO NOT EDIT MANUALLY." _RE_QUALIFIED_TYPES = re.compile(r"\w+(?:\.\w+)*") _RE_COMMENT_IN_STRING = re.compile(r"""['"]\s*%(text)s\s*.*['"]""" % {"text": SIG_COMMENT}) _logger = logging.getLogger(__name__) def get_fields(node, fields_tag="field_list"): """Get the field names and their values from a node. :sig: (Document, str) -> Dict[str, str] :param node: Node to get the fields from. :param fields_tag: Tag of child node that contains the fields. :return: Names and values of fields. """ fields_nodes = [c for c in node.children if c.tagname == fields_tag] if len(fields_nodes) == 0: return {} assert len(fields_nodes) == 1, "multiple nodes with tag " + fields_tag fields_node = fields_nodes[0] fields = [ {f.tagname: f.rawsource.strip() for f in n.children} for n in fields_node.children if n.tagname == "field" ] return {f["field_name"]: f["field_body"] for f in fields} def extract_signature(docstring): """Extract the signature from a docstring. :sig: (str) -> Optional[str] :param docstring: Docstring to extract the signature from. :return: Extracted signature, or ``None`` if there's no signature. """ root = publish_doctree(docstring, settings_overrides={"report_level": 5}) fields = get_fields(root) return fields.get(SIG_FIELD) def get_signature(node): """Get the signature of a function or a class. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef]) -> Optional[str] :param node: Node to get the signature from. :return: Value of signature field in node docstring, or ``None`` if there's no signature. """ docstring = ast.get_docstring(node) if docstring is None: return None return extract_signature(docstring) def split_parameter_types(parameters): """Split a parameter types declaration into individual types. The input is the left hand side of a signature (the part before the arrow), excluding the parentheses. :sig: (str) -> List[str] :param parameters: Comma separated parameter types. :return: Parameter types. """ if parameters == "": return [] # only consider the top level commas, ignore the ones in [] commas = [] bracket_depth = 0 for i, char in enumerate(parameters): if (char == ",") and (bracket_depth == 0): commas.append(i) elif char == "[": bracket_depth += 1 elif char == "]": bracket_depth -= 1 types = [] last_i = 0 for i in commas: types.append(parameters[last_i:i].strip()) last_i = i + 1 else: types.append(parameters[last_i:].strip()) return types def parse_signature(signature): """Parse a signature into its input and return parameter types. This will also collect the types that are required by any of the input and return types. :sig: (str) -> Tuple[List[str], str, Set[str]] :param signature: Signature to parse. :return: Input parameter types, return type, and all required types. """ if " -> " not in signature: # signature comment: no parameters, treat variable type as return type param_types, return_type = None, signature.strip() else: lhs, return_type = [s.strip() for s in signature.split(" -> ")] csv = lhs[1:-1].strip() # remove the parentheses around the parameter type list param_types = split_parameter_types(csv) requires = set(_RE_QUALIFIED_TYPES.findall(signature)) return param_types, return_type, requires class StubNode: """A node in a stub tree.""" def __init__(self): """Initialize this stub node. :sig: () -> None """ self.variables = [] # sig: List[VariableNode] self.variable_names = set() # sig: Set[str] self.children = [] # sig: List[Union[FunctionNode, ClassNode]] self.parent = None # sig: Optional[StubNode] def add_variable(self, node): """Add a variable node to this node. :sig: (VariableNode) -> None :param node: Variable node to add. """ if node.name not in self.variable_names: self.variables.append(node) self.variable_names.add(node.name) node.parent = self def add_child(self, node): """Add a function/method or class node to this node. :sig: (Union[FunctionNode, ClassNode]) -> None :param node: Function or class node to add. """ self.children.append(node) node.parent = self def get_code(self): """Get the stub code for this node. The stub code for a node consists of the type annotations of its variables, followed by the prototypes of its functions/methods and classes. :sig: () -> List[str] :return: Lines of stub code for this node. """ stub = [] for child in self.variables: stub.extend(child.get_code()) if ( (len(self.variables) > 0) and (len(self.children) > 0) and (not isinstance(self, ClassNode)) ): stub.append("") for child in self.children: stub.extend(child.get_code()) return stub class VariableNode(StubNode): """A node representing an assignment in a stub tree.""" def __init__(self, name, type_): """Initialize this variable node. :sig: (str, str) -> None :param name: Name of variable that is being assigned to. :param type_: Type of variable. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.type_ = type_ # sig: str def get_code(self): """Get the type annotation for this variable. :sig: () -> List[str] :return: Lines of stub code for this variable. """ return ["%(n)s = ... # type: %(t)s" % {"n": self.name, "t": self.type_}] class FunctionNode(StubNode): """A node representing a function in a stub tree.""" def __init__(self, name, parameters, rtype, decorators=None): """Initialize this function node. The parameters have to given as a list of triples where each item specifies the name of the parameter, its type, and whether it has a default value or not. :sig: (str, Sequence[Tuple[str, str, bool]], str, Optional[Sequence[str]]) -> None :param name: Name of function. :param parameters: List of parameter triples (name, type, has_default). :param rtype: Type of return value. :param decorators: Decorators of function. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.parameters = parameters # sig: Sequence[Tuple[str, str, bool]] self.rtype = rtype # sig: str self.decorators = decorators if decorators is not None else [] # sig: Sequence[str] self._async = False # sig: bool def get_code(self): """Get the stub code for this function. :sig: () -> List[str] :return: Lines of stub code for this function. """ stub = [] for deco in self.decorators: if (deco in DECORATORS) or deco.endswith(".setter"): stub.append("@" + deco) parameters = [] for name, type_, has_default in self.parameters: decl = "%(n)s%(t)s%(d)s" % { "n": name, "t": ": " + type_ if type_ else "", "d": " = ..." if has_default else "", } parameters.append(decl) slots = { "a": "async " if self._async else "", "n": self.name, "p": ", ".join(parameters), "r": self.rtype, } prototype = "%(a)sdef %(n)s(%(p)s) -> %(r)s: ..." % slots if len(prototype) <= LINE_LENGTH_LIMIT: stub.append(prototype) elif len(INDENT + slots["p"]) <= LINE_LENGTH_LIMIT: stub.append("%(a)sdef %(n)s(" % slots) stub.append(INDENT + slots["p"]) stub.append(") -> %(r)s: ..." % slots) else: stub.append("%(a)sdef %(n)s(" % slots) for param in parameters: stub.append(INDENT + param + ",") stub.append(") -> %(r)s: ..." % slots) return stub class ClassNode(StubNode): """A node representing a class in a stub tree.""" def __init__(self, name, bases, signature=None): """Initialize this class node. :sig: (str, Sequence[str], Optional[str]) -> None :param name: Name of class. :param bases: Base classes of class. :param signature: Signature of class, to be used in __init__ method. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.bases = bases # sig: Sequence[str] self.signature = signature # sig: Optional[str] def get_code(self): """Get the stub code for this class. :sig: () -> List[str] :return: Lines of stub code for this class. """ stub = [] bases = ("(" + ", ".join(self.bases) + ")") if len(self.bases) > 0 else "" slots = {"n": self.name, "b": bases} if (len(self.children) == 0) and (len(self.variables) == 0): stub.append("class %(n)s%(b)s: ..." % slots) else: stub.append("class %(n)s%(b)s:" % slots) super_code = super().get_code() if PY3 else StubNode.get_code(self) for line in super_code: stub.append(INDENT + line) return stub def get_aliases(lines): """Get the type aliases in the source. :sig: (Sequence[str]) -> Dict[str, str] :param lines: Lines of the source code. :return: Aliases and their their definitions. """ aliases = {} for line in lines: line = line.strip() if len(line) > 0 and line.startswith(SIG_ALIAS): _, content = line.split(SIG_ALIAS) alias, signature = [t.strip() for t in content.split("=")] aliases[alias] = signature return aliases class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue() def get_stub(source, generic=False): """Get the stub code for a source code. :sig: (str, bool) -> str :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. :return: Generated stub code. """ generator = StubGenerator(source, generic=generic) stub = generator.generate_stub() return stub def get_mod_paths(mod_name, out_dir): """Get source and stub paths for a module.""" paths = [] try: mod = get_loader(mod_name) source = Path(mod.path) if source.name.endswith(".py"): source_rel = Path(*mod_name.split(".")) if source.name == "__init__.py": source_rel = source_rel.joinpath("__init__.py") destination = Path(out_dir, source_rel.with_suffix(".pyi")) paths.append((source, destination)) except Exception as e: _logger.debug(e) _logger.warning("cannot handle module, skipping: %s", mod_name) return paths def get_pkg_paths(pkg_name, out_dir): """Recursively get all source and stub paths for a package.""" paths = [] try: pkg = import_module(pkg_name) if not hasattr(pkg, "__path__"): return get_mod_paths(pkg_name, out_dir) for mod_info in walk_packages(pkg.__path__, pkg.__name__ + "."): mod_paths = get_mod_paths(mod_info.name, out_dir) paths.extend(mod_paths) except Exception as e: _logger.debug(e) _logger.warning("cannot handle package, skipping: %s", pkg_name) return paths ############################################################ # SPHINX ############################################################ def setup(app): """Register to Sphinx.""" app.connect("autodoc-process-docstring", process_docstring) return {"version": __version__} ############################################################ # MAIN ############################################################ def main(argv=None): """Start the command line interface.""" parser = ArgumentParser(prog="pygenstub") parser.add_argument("--version", action="version", version="%(prog)s " + __version__) parser.add_argument("files", nargs="*", help="generate stubs for given files") parser.add_argument( "-m", "--module", action="append", metavar="MODULE", dest="modules", default=[], help="generate stubs for given modules", ) parser.add_argument( "-o", "--output", metavar="PATH", dest="out_dir", help="change the output directory" ) parser.add_argument( "--generic", action="store_true", default=False, help="generate generic stubs" ) parser.add_argument("--debug", action="store_true", help="enable debug messages") argv = argv if argv is not None else sys.argv arguments = parser.parse_args(argv[1:]) # set debug mode if arguments.debug: logging.basicConfig(level=logging.DEBUG) _logger.debug("running in debug mode") out_dir = arguments.out_dir if arguments.out_dir is not None else "" if (out_dir == "") and (len(arguments.modules) > 0): print("Output directory must be given when generating stubs for modules.") sys.exit(1) modules = [] for path in arguments.files: paths = Path(path).glob("**/*.py") if Path(path).is_dir() else [Path(path)] for source in paths: if str(source).startswith(os.path.pardir): source = source.absolute().resolve() if (out_dir != "") and source.is_absolute(): source = source.relative_to(source.root) destination = Path(out_dir, source.with_suffix(".pyi")) modules.append((source, destination)) for mod_name in arguments.modules: modules.extend(get_pkg_paths(mod_name, out_dir)) for source, destination in modules: _logger.info("generating stub for %s to path %s", source, destination) with source.open() as f: code = f.read() try: stub = get_stub(code, generic=arguments.generic) except Exception as e: print(source, "-", e, file=sys.stderr) continue if stub != "": if not destination.parent.exists(): destination.parent.mkdir(parents=True) with destination.open("w") as f: f.write("# " + EDIT_WARNING + "\n\n" + stub) if __name__ == "__main__": main()
uyar/pygenstub
pygenstub.py
main
python
def main(argv=None): parser = ArgumentParser(prog="pygenstub") parser.add_argument("--version", action="version", version="%(prog)s " + __version__) parser.add_argument("files", nargs="*", help="generate stubs for given files") parser.add_argument( "-m", "--module", action="append", metavar="MODULE", dest="modules", default=[], help="generate stubs for given modules", ) parser.add_argument( "-o", "--output", metavar="PATH", dest="out_dir", help="change the output directory" ) parser.add_argument( "--generic", action="store_true", default=False, help="generate generic stubs" ) parser.add_argument("--debug", action="store_true", help="enable debug messages") argv = argv if argv is not None else sys.argv arguments = parser.parse_args(argv[1:]) # set debug mode if arguments.debug: logging.basicConfig(level=logging.DEBUG) _logger.debug("running in debug mode") out_dir = arguments.out_dir if arguments.out_dir is not None else "" if (out_dir == "") and (len(arguments.modules) > 0): print("Output directory must be given when generating stubs for modules.") sys.exit(1) modules = [] for path in arguments.files: paths = Path(path).glob("**/*.py") if Path(path).is_dir() else [Path(path)] for source in paths: if str(source).startswith(os.path.pardir): source = source.absolute().resolve() if (out_dir != "") and source.is_absolute(): source = source.relative_to(source.root) destination = Path(out_dir, source.with_suffix(".pyi")) modules.append((source, destination)) for mod_name in arguments.modules: modules.extend(get_pkg_paths(mod_name, out_dir)) for source, destination in modules: _logger.info("generating stub for %s to path %s", source, destination) with source.open() as f: code = f.read() try: stub = get_stub(code, generic=arguments.generic) except Exception as e: print(source, "-", e, file=sys.stderr) continue if stub != "": if not destination.parent.exists(): destination.parent.mkdir(parents=True) with destination.open("w") as f: f.write("# " + EDIT_WARNING + "\n\n" + stub)
Start the command line interface.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L888-L951
[ "def get_stub(source, generic=False):\n \"\"\"Get the stub code for a source code.\n\n :sig: (str, bool) -> str\n :param source: Source code to generate the stub for.\n :param generic: Whether to produce generic stubs.\n :return: Generated stub code.\n \"\"\"\n generator = StubGenerator(source,...
# Copyright (C) 2016-2019 H. Turgut Uyar <uyar@tekir.org> # # pygenstub is free software: you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # pygenstub 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 General Public License for more details. # # You should have received a copy of the GNU General Public License # along with pygenstub. If not, see <http://www.gnu.org/licenses/>. """pygenstub is a utility for generating stub files from docstrings in source files. It takes a source file as input and creates a stub file with the same base name and the ``.pyi`` extension. For more information, please refer to the documentation: https://pygenstub.tekir.org/ """ from __future__ import absolute_import, division, print_function, unicode_literals import ast import inspect import logging import os import re import sys import textwrap from argparse import ArgumentParser from bisect import bisect from collections import OrderedDict from importlib import import_module from io import StringIO from pkgutil import get_loader, walk_packages from docutils.core import publish_doctree __version__ = "1.4.0" # sig: str PY3 = sys.version_info >= (3, 0) if not PY3: import __builtin__ as builtins from pathlib2 import Path else: import builtins from pathlib import Path # sigalias: Document = docutils.nodes.document BUILTIN_TYPES = {k for k, t in builtins.__dict__.items() if isinstance(t, type)} BUILTIN_TYPES.add("None") SIG_FIELD = "sig" # sig: str SIG_COMMENT = "# sig:" # sig: str SIG_ALIAS = "# sigalias:" # sig: str DECORATORS = {"property", "staticmethod", "classmethod"} # sig: Set[str] LINE_LENGTH_LIMIT = 79 INDENT = 4 * " " EDIT_WARNING = "THIS FILE IS AUTOMATICALLY GENERATED, DO NOT EDIT MANUALLY." _RE_QUALIFIED_TYPES = re.compile(r"\w+(?:\.\w+)*") _RE_COMMENT_IN_STRING = re.compile(r"""['"]\s*%(text)s\s*.*['"]""" % {"text": SIG_COMMENT}) _logger = logging.getLogger(__name__) def get_fields(node, fields_tag="field_list"): """Get the field names and their values from a node. :sig: (Document, str) -> Dict[str, str] :param node: Node to get the fields from. :param fields_tag: Tag of child node that contains the fields. :return: Names and values of fields. """ fields_nodes = [c for c in node.children if c.tagname == fields_tag] if len(fields_nodes) == 0: return {} assert len(fields_nodes) == 1, "multiple nodes with tag " + fields_tag fields_node = fields_nodes[0] fields = [ {f.tagname: f.rawsource.strip() for f in n.children} for n in fields_node.children if n.tagname == "field" ] return {f["field_name"]: f["field_body"] for f in fields} def extract_signature(docstring): """Extract the signature from a docstring. :sig: (str) -> Optional[str] :param docstring: Docstring to extract the signature from. :return: Extracted signature, or ``None`` if there's no signature. """ root = publish_doctree(docstring, settings_overrides={"report_level": 5}) fields = get_fields(root) return fields.get(SIG_FIELD) def get_signature(node): """Get the signature of a function or a class. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef]) -> Optional[str] :param node: Node to get the signature from. :return: Value of signature field in node docstring, or ``None`` if there's no signature. """ docstring = ast.get_docstring(node) if docstring is None: return None return extract_signature(docstring) def split_parameter_types(parameters): """Split a parameter types declaration into individual types. The input is the left hand side of a signature (the part before the arrow), excluding the parentheses. :sig: (str) -> List[str] :param parameters: Comma separated parameter types. :return: Parameter types. """ if parameters == "": return [] # only consider the top level commas, ignore the ones in [] commas = [] bracket_depth = 0 for i, char in enumerate(parameters): if (char == ",") and (bracket_depth == 0): commas.append(i) elif char == "[": bracket_depth += 1 elif char == "]": bracket_depth -= 1 types = [] last_i = 0 for i in commas: types.append(parameters[last_i:i].strip()) last_i = i + 1 else: types.append(parameters[last_i:].strip()) return types def parse_signature(signature): """Parse a signature into its input and return parameter types. This will also collect the types that are required by any of the input and return types. :sig: (str) -> Tuple[List[str], str, Set[str]] :param signature: Signature to parse. :return: Input parameter types, return type, and all required types. """ if " -> " not in signature: # signature comment: no parameters, treat variable type as return type param_types, return_type = None, signature.strip() else: lhs, return_type = [s.strip() for s in signature.split(" -> ")] csv = lhs[1:-1].strip() # remove the parentheses around the parameter type list param_types = split_parameter_types(csv) requires = set(_RE_QUALIFIED_TYPES.findall(signature)) return param_types, return_type, requires class StubNode: """A node in a stub tree.""" def __init__(self): """Initialize this stub node. :sig: () -> None """ self.variables = [] # sig: List[VariableNode] self.variable_names = set() # sig: Set[str] self.children = [] # sig: List[Union[FunctionNode, ClassNode]] self.parent = None # sig: Optional[StubNode] def add_variable(self, node): """Add a variable node to this node. :sig: (VariableNode) -> None :param node: Variable node to add. """ if node.name not in self.variable_names: self.variables.append(node) self.variable_names.add(node.name) node.parent = self def add_child(self, node): """Add a function/method or class node to this node. :sig: (Union[FunctionNode, ClassNode]) -> None :param node: Function or class node to add. """ self.children.append(node) node.parent = self def get_code(self): """Get the stub code for this node. The stub code for a node consists of the type annotations of its variables, followed by the prototypes of its functions/methods and classes. :sig: () -> List[str] :return: Lines of stub code for this node. """ stub = [] for child in self.variables: stub.extend(child.get_code()) if ( (len(self.variables) > 0) and (len(self.children) > 0) and (not isinstance(self, ClassNode)) ): stub.append("") for child in self.children: stub.extend(child.get_code()) return stub class VariableNode(StubNode): """A node representing an assignment in a stub tree.""" def __init__(self, name, type_): """Initialize this variable node. :sig: (str, str) -> None :param name: Name of variable that is being assigned to. :param type_: Type of variable. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.type_ = type_ # sig: str def get_code(self): """Get the type annotation for this variable. :sig: () -> List[str] :return: Lines of stub code for this variable. """ return ["%(n)s = ... # type: %(t)s" % {"n": self.name, "t": self.type_}] class FunctionNode(StubNode): """A node representing a function in a stub tree.""" def __init__(self, name, parameters, rtype, decorators=None): """Initialize this function node. The parameters have to given as a list of triples where each item specifies the name of the parameter, its type, and whether it has a default value or not. :sig: (str, Sequence[Tuple[str, str, bool]], str, Optional[Sequence[str]]) -> None :param name: Name of function. :param parameters: List of parameter triples (name, type, has_default). :param rtype: Type of return value. :param decorators: Decorators of function. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.parameters = parameters # sig: Sequence[Tuple[str, str, bool]] self.rtype = rtype # sig: str self.decorators = decorators if decorators is not None else [] # sig: Sequence[str] self._async = False # sig: bool def get_code(self): """Get the stub code for this function. :sig: () -> List[str] :return: Lines of stub code for this function. """ stub = [] for deco in self.decorators: if (deco in DECORATORS) or deco.endswith(".setter"): stub.append("@" + deco) parameters = [] for name, type_, has_default in self.parameters: decl = "%(n)s%(t)s%(d)s" % { "n": name, "t": ": " + type_ if type_ else "", "d": " = ..." if has_default else "", } parameters.append(decl) slots = { "a": "async " if self._async else "", "n": self.name, "p": ", ".join(parameters), "r": self.rtype, } prototype = "%(a)sdef %(n)s(%(p)s) -> %(r)s: ..." % slots if len(prototype) <= LINE_LENGTH_LIMIT: stub.append(prototype) elif len(INDENT + slots["p"]) <= LINE_LENGTH_LIMIT: stub.append("%(a)sdef %(n)s(" % slots) stub.append(INDENT + slots["p"]) stub.append(") -> %(r)s: ..." % slots) else: stub.append("%(a)sdef %(n)s(" % slots) for param in parameters: stub.append(INDENT + param + ",") stub.append(") -> %(r)s: ..." % slots) return stub class ClassNode(StubNode): """A node representing a class in a stub tree.""" def __init__(self, name, bases, signature=None): """Initialize this class node. :sig: (str, Sequence[str], Optional[str]) -> None :param name: Name of class. :param bases: Base classes of class. :param signature: Signature of class, to be used in __init__ method. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.bases = bases # sig: Sequence[str] self.signature = signature # sig: Optional[str] def get_code(self): """Get the stub code for this class. :sig: () -> List[str] :return: Lines of stub code for this class. """ stub = [] bases = ("(" + ", ".join(self.bases) + ")") if len(self.bases) > 0 else "" slots = {"n": self.name, "b": bases} if (len(self.children) == 0) and (len(self.variables) == 0): stub.append("class %(n)s%(b)s: ..." % slots) else: stub.append("class %(n)s%(b)s:" % slots) super_code = super().get_code() if PY3 else StubNode.get_code(self) for line in super_code: stub.append(INDENT + line) return stub def get_aliases(lines): """Get the type aliases in the source. :sig: (Sequence[str]) -> Dict[str, str] :param lines: Lines of the source code. :return: Aliases and their their definitions. """ aliases = {} for line in lines: line = line.strip() if len(line) > 0 and line.startswith(SIG_ALIAS): _, content = line.split(SIG_ALIAS) alias, signature = [t.strip() for t in content.split("=")] aliases[alias] = signature return aliases class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue() def get_stub(source, generic=False): """Get the stub code for a source code. :sig: (str, bool) -> str :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. :return: Generated stub code. """ generator = StubGenerator(source, generic=generic) stub = generator.generate_stub() return stub def get_mod_paths(mod_name, out_dir): """Get source and stub paths for a module.""" paths = [] try: mod = get_loader(mod_name) source = Path(mod.path) if source.name.endswith(".py"): source_rel = Path(*mod_name.split(".")) if source.name == "__init__.py": source_rel = source_rel.joinpath("__init__.py") destination = Path(out_dir, source_rel.with_suffix(".pyi")) paths.append((source, destination)) except Exception as e: _logger.debug(e) _logger.warning("cannot handle module, skipping: %s", mod_name) return paths def get_pkg_paths(pkg_name, out_dir): """Recursively get all source and stub paths for a package.""" paths = [] try: pkg = import_module(pkg_name) if not hasattr(pkg, "__path__"): return get_mod_paths(pkg_name, out_dir) for mod_info in walk_packages(pkg.__path__, pkg.__name__ + "."): mod_paths = get_mod_paths(mod_info.name, out_dir) paths.extend(mod_paths) except Exception as e: _logger.debug(e) _logger.warning("cannot handle package, skipping: %s", pkg_name) return paths ############################################################ # SPHINX ############################################################ def process_docstring(app, what, name, obj, options, lines): """Modify the docstring before generating documentation. This will insert type declarations for parameters and return type into the docstring, and remove the signature field so that it will be excluded from the generated document. """ aliases = getattr(app, "_sigaliases", None) if aliases is None: if what == "module": aliases = get_aliases(inspect.getsource(obj).splitlines()) app._sigaliases = aliases sig_marker = ":" + SIG_FIELD + ":" is_class = what in ("class", "exception") signature = extract_signature("\n".join(lines)) if signature is None: if not is_class: return init_method = getattr(obj, "__init__") init_doc = init_method.__doc__ init_lines = init_doc.splitlines()[1:] if len(init_lines) > 1: init_doc = textwrap.dedent("\n".join(init_lines[1:])) init_lines = init_doc.splitlines() if sig_marker not in init_doc: return sig_started = False for line in init_lines: if line.lstrip().startswith(sig_marker): sig_started = True if sig_started: lines.append(line) signature = extract_signature("\n".join(lines)) if is_class: obj = init_method param_types, rtype, _ = parse_signature(signature) param_names = [p for p in inspect.signature(obj).parameters] if is_class and (param_names[0] == "self"): del param_names[0] # if something goes wrong, don't insert parameter types if len(param_names) == len(param_types): for name, type_ in zip(param_names, param_types): find = ":param %(name)s:" % {"name": name} alias = aliases.get(type_) if alias is not None: type_ = "*%(type)s* :sup:`%(alias)s`" % {"type": type_, "alias": alias} for i, line in enumerate(lines): if line.startswith(find): lines.insert(i, ":type %(name)s: %(type)s" % {"name": name, "type": type_}) break if not is_class: for i, line in enumerate(lines): if line.startswith((":return:", ":returns:")): lines.insert(i, ":rtype: " + rtype) break # remove the signature field sig_start = 0 while sig_start < len(lines): if lines[sig_start].startswith(sig_marker): break sig_start += 1 sig_end = sig_start + 1 while sig_end < len(lines): if (not lines[sig_end]) or (lines[sig_end][0] != " "): break sig_end += 1 for i in reversed(range(sig_start, sig_end)): del lines[i] def setup(app): """Register to Sphinx.""" app.connect("autodoc-process-docstring", process_docstring) return {"version": __version__} ############################################################ # MAIN ############################################################ if __name__ == "__main__": main()
uyar/pygenstub
pygenstub.py
StubNode.add_variable
python
def add_variable(self, node): if node.name not in self.variable_names: self.variables.append(node) self.variable_names.add(node.name) node.parent = self
Add a variable node to this node. :sig: (VariableNode) -> None :param node: Variable node to add.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L195-L204
null
class StubNode: """A node in a stub tree.""" def __init__(self): """Initialize this stub node. :sig: () -> None """ self.variables = [] # sig: List[VariableNode] self.variable_names = set() # sig: Set[str] self.children = [] # sig: List[Union[FunctionNode, ClassNode]] self.parent = None # sig: Optional[StubNode] def add_child(self, node): """Add a function/method or class node to this node. :sig: (Union[FunctionNode, ClassNode]) -> None :param node: Function or class node to add. """ self.children.append(node) node.parent = self def get_code(self): """Get the stub code for this node. The stub code for a node consists of the type annotations of its variables, followed by the prototypes of its functions/methods and classes. :sig: () -> List[str] :return: Lines of stub code for this node. """ stub = [] for child in self.variables: stub.extend(child.get_code()) if ( (len(self.variables) > 0) and (len(self.children) > 0) and (not isinstance(self, ClassNode)) ): stub.append("") for child in self.children: stub.extend(child.get_code()) return stub
uyar/pygenstub
pygenstub.py
StubNode.get_code
python
def get_code(self): stub = [] for child in self.variables: stub.extend(child.get_code()) if ( (len(self.variables) > 0) and (len(self.children) > 0) and (not isinstance(self, ClassNode)) ): stub.append("") for child in self.children: stub.extend(child.get_code()) return stub
Get the stub code for this node. The stub code for a node consists of the type annotations of its variables, followed by the prototypes of its functions/methods and classes. :sig: () -> List[str] :return: Lines of stub code for this node.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L215-L235
null
class StubNode: """A node in a stub tree.""" def __init__(self): """Initialize this stub node. :sig: () -> None """ self.variables = [] # sig: List[VariableNode] self.variable_names = set() # sig: Set[str] self.children = [] # sig: List[Union[FunctionNode, ClassNode]] self.parent = None # sig: Optional[StubNode] def add_variable(self, node): """Add a variable node to this node. :sig: (VariableNode) -> None :param node: Variable node to add. """ if node.name not in self.variable_names: self.variables.append(node) self.variable_names.add(node.name) node.parent = self def add_child(self, node): """Add a function/method or class node to this node. :sig: (Union[FunctionNode, ClassNode]) -> None :param node: Function or class node to add. """ self.children.append(node) node.parent = self
uyar/pygenstub
pygenstub.py
FunctionNode.get_code
python
def get_code(self): stub = [] for deco in self.decorators: if (deco in DECORATORS) or deco.endswith(".setter"): stub.append("@" + deco) parameters = [] for name, type_, has_default in self.parameters: decl = "%(n)s%(t)s%(d)s" % { "n": name, "t": ": " + type_ if type_ else "", "d": " = ..." if has_default else "", } parameters.append(decl) slots = { "a": "async " if self._async else "", "n": self.name, "p": ", ".join(parameters), "r": self.rtype, } prototype = "%(a)sdef %(n)s(%(p)s) -> %(r)s: ..." % slots if len(prototype) <= LINE_LENGTH_LIMIT: stub.append(prototype) elif len(INDENT + slots["p"]) <= LINE_LENGTH_LIMIT: stub.append("%(a)sdef %(n)s(" % slots) stub.append(INDENT + slots["p"]) stub.append(") -> %(r)s: ..." % slots) else: stub.append("%(a)sdef %(n)s(" % slots) for param in parameters: stub.append(INDENT + param + ",") stub.append(") -> %(r)s: ..." % slots) return stub
Get the stub code for this function. :sig: () -> List[str] :return: Lines of stub code for this function.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L290-L331
null
class FunctionNode(StubNode): """A node representing a function in a stub tree.""" def __init__(self, name, parameters, rtype, decorators=None): """Initialize this function node. The parameters have to given as a list of triples where each item specifies the name of the parameter, its type, and whether it has a default value or not. :sig: (str, Sequence[Tuple[str, str, bool]], str, Optional[Sequence[str]]) -> None :param name: Name of function. :param parameters: List of parameter triples (name, type, has_default). :param rtype: Type of return value. :param decorators: Decorators of function. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.parameters = parameters # sig: Sequence[Tuple[str, str, bool]] self.rtype = rtype # sig: str self.decorators = decorators if decorators is not None else [] # sig: Sequence[str] self._async = False # sig: bool
uyar/pygenstub
pygenstub.py
ClassNode.get_code
python
def get_code(self): stub = [] bases = ("(" + ", ".join(self.bases) + ")") if len(self.bases) > 0 else "" slots = {"n": self.name, "b": bases} if (len(self.children) == 0) and (len(self.variables) == 0): stub.append("class %(n)s%(b)s: ..." % slots) else: stub.append("class %(n)s%(b)s:" % slots) super_code = super().get_code() if PY3 else StubNode.get_code(self) for line in super_code: stub.append(INDENT + line) return stub
Get the stub code for this class. :sig: () -> List[str] :return: Lines of stub code for this class.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L353-L369
[ "def get_code(self):\n \"\"\"Get the stub code for this node.\n\n The stub code for a node consists of the type annotations of its variables,\n followed by the prototypes of its functions/methods and classes.\n\n :sig: () -> List[str]\n :return: Lines of stub code for this node.\n \"\"\"\n stub...
class ClassNode(StubNode): """A node representing a class in a stub tree.""" def __init__(self, name, bases, signature=None): """Initialize this class node. :sig: (str, Sequence[str], Optional[str]) -> None :param name: Name of class. :param bases: Base classes of class. :param signature: Signature of class, to be used in __init__ method. """ if not PY3: StubNode.__init__(self) else: super().__init__() self.name = name # sig: str self.bases = bases # sig: Sequence[str] self.signature = signature # sig: Optional[str]
uyar/pygenstub
pygenstub.py
StubGenerator.collect_aliases
python
def collect_aliases(self): self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias}
Collect the type aliases in the source. :sig: () -> None
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L417-L426
[ "def parse_signature(signature):\n \"\"\"Parse a signature into its input and return parameter types.\n\n This will also collect the types that are required by any of the input\n and return types.\n\n :sig: (str) -> Tuple[List[str], str, Set[str]]\n :param signature: Signature to parse.\n :return:...
class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue()
uyar/pygenstub
pygenstub.py
StubGenerator.visit_Import
python
def visit_Import(self, node): line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name
Visit an import node.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L428-L436
null
class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue()
uyar/pygenstub
pygenstub.py
StubGenerator.visit_ImportFrom
python
def visit_ImportFrom(self, node): line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name
Visit an from-import node.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L438-L446
null
class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue()
uyar/pygenstub
pygenstub.py
StubGenerator.visit_Assign
python
def visit_Assign(self, node): line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node)
Visit an assignment node.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L448-L480
[ "def parse_signature(signature):\n \"\"\"Parse a signature into its input and return parameter types.\n\n This will also collect the types that are required by any of the input\n and return types.\n\n :sig: (str) -> Tuple[List[str], str, Set[str]]\n :param signature: Signature to parse.\n :return:...
class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue()
uyar/pygenstub
pygenstub.py
StubGenerator.get_function_node
python
def get_function_node(self, node): decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node
Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L482-L575
[ "def get_signature(node):\n \"\"\"Get the signature of a function or a class.\n\n :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef]) -> Optional[str]\n :param node: Node to get the signature from.\n :return: Value of signature field in node docstring, or ``None`` if there's no signature....
class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue()
uyar/pygenstub
pygenstub.py
StubGenerator.visit_FunctionDef
python
def visit_FunctionDef(self, node): node = self.get_function_node(node) if node is not None: node._async = False
Visit a function node.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L577-L581
[ "def get_function_node(self, node):\n \"\"\"Process a function node.\n\n :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode\n :param node: Node to process.\n :return: Generated function node in stub tree.\n \"\"\"\n decorators = []\n for d in node.decorator_list:\n if h...
class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue()
uyar/pygenstub
pygenstub.py
StubGenerator.visit_AsyncFunctionDef
python
def visit_AsyncFunctionDef(self, node): node = self.get_function_node(node) if node is not None: node._async = True
Visit an async function node.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L583-L587
[ "def get_function_node(self, node):\n \"\"\"Process a function node.\n\n :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode\n :param node: Node to process.\n :return: Generated function node in stub tree.\n \"\"\"\n decorators = []\n for d in node.decorator_list:\n if h...
class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue()
uyar/pygenstub
pygenstub.py
StubGenerator.visit_ClassDef
python
def visit_ClassDef(self, node): self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1]
Visit a class node.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L589-L612
[ "def get_signature(node):\n \"\"\"Get the signature of a function or a class.\n\n :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef]) -> Optional[str]\n :param node: Node to get the signature from.\n :return: Value of signature field in node docstring, or ``None`` if there's no signature....
class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue()
uyar/pygenstub
pygenstub.py
StubGenerator.generate_import_from
python
def generate_import_from(module_, names): regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line
Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L615-L639
null
class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_stub(self): """Generate the stub code for this source. :sig: () -> str :return: Generated stub code. """ needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue()
uyar/pygenstub
pygenstub.py
StubGenerator.generate_stub
python
def generate_stub(self): needed_types = self.required_types - BUILTIN_TYPES needed_types -= self.defined_types _logger.debug("defined types: %s", self.defined_types) module_vars = {v.name for v in self.root.variables} _logger.debug("module variables: %s", module_vars) qualified_types = {n for n in needed_types if "." in n} qualified_namespaces = {".".join(n.split(".")[:-1]) for n in qualified_types} needed_namespaces = qualified_namespaces - module_vars needed_types -= qualified_types _logger.debug("needed namespaces: %s", needed_namespaces) imported_names = {n.split("::")[0] for n in self.imported_names} imported_types = imported_names & (needed_types | needed_namespaces) needed_types -= imported_types needed_namespaces -= imported_names _logger.debug("used imported types: %s", imported_types) try: typing_mod = __import__("typing") typing_types = {n for n in needed_types if hasattr(typing_mod, n)} needed_types -= typing_types _logger.debug("types from typing module: %s", typing_types) except ImportError: typing_types = set() _logger.warn("typing module not installed") if len(needed_types) > 0: raise ValueError("Unknown types: " + ", ".join(needed_types)) out = StringIO() started = False if len(typing_types) > 0: line = self.generate_import_from("typing", typing_types) out.write(line + "\n") started = True if len(imported_types) > 0: if started: out.write("\n") # preserve the import order in the source file for name in self.imported_names: if name.split("::")[0] in imported_types: line = self.generate_import_from(self.imported_names[name], {name}) out.write(line + "\n") started = True if len(needed_namespaces) > 0: if started: out.write("\n") as_names = {n.split("::")[0]: n for n in self.imported_namespaces if "::" in n} for module_ in sorted(needed_namespaces): if module_ in as_names: a, n = as_names[module_].split("::") out.write("import " + n + " as " + a + "\n") else: out.write("import " + module_ + "\n") started = True if len(self.aliases) > 0: if started: out.write("\n") for alias, signature in self.aliases.items(): out.write("%s = %s\n" % (alias, signature)) started = True if started: out.write("\n") stub_lines = self.root.get_code() n_lines = len(stub_lines) for line_no in range(n_lines): prev_line = stub_lines[line_no - 1] if line_no > 0 else None line = stub_lines[line_no] next_line = stub_lines[line_no + 1] if line_no < (n_lines - 1) else None if ( line.startswith("class ") and (prev_line is not None) and ( (not prev_line.startswith("class ")) or (next_line and next_line.startswith(" ")) ) ): out.write("\n") if ( line.startswith("def ") and (prev_line is not None) and (prev_line.startswith((" ", "class "))) ): out.write("\n") out.write(line + "\n") line_no += 1 return out.getvalue()
Generate the stub code for this source. :sig: () -> str :return: Generated stub code.
train
https://github.com/uyar/pygenstub/blob/a6b18a823382d3c6be29c411fb33c58b6090d22c/pygenstub.py#L641-L742
[ "def generate_import_from(module_, names):\n \"\"\"Generate an import line.\n\n :sig: (str, Set[str]) -> str\n :param module_: Name of module to import the names from.\n :param names: Names to import.\n :return: Import line in stub code.\n \"\"\"\n regular_names = [n for n in names if \"::\" no...
class StubGenerator(ast.NodeVisitor): """A transformer that generates stub declarations from a source code.""" def __init__(self, source, generic=False): """Initialize this stub generator. :sig: (str, bool) -> None :param source: Source code to generate the stub for. :param generic: Whether to produce generic stubs. """ self.root = StubNode() # sig: StubNode self.generic = generic # sig: bool self.imported_namespaces = OrderedDict() # sig: OrderedDict[str, str] self.imported_names = OrderedDict() # sig: OrderedDict[str, str] self.defined_types = set() # sig: Set[str] self.required_types = set() # sig: Set[str] self.aliases = OrderedDict() # sig: OrderedDict[str, str] self._parents = [self.root] # sig: List[StubNode] self._code_lines = source.splitlines() # sig: List[str] self.collect_aliases() ast_tree = ast.parse(source) self.visit(ast_tree) def collect_aliases(self): """Collect the type aliases in the source. :sig: () -> None """ self.aliases = get_aliases(self._code_lines) for alias, signature in self.aliases.items(): _, _, requires = parse_signature(signature) self.required_types |= requires self.defined_types |= {alias} def visit_Import(self, node): """Visit an import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_namespaces[imported_name] = module_name def visit_ImportFrom(self, node): """Visit an from-import node.""" line = self._code_lines[node.lineno - 1] module_name = line.split("from")[1].split("import")[0].strip() for name in node.names: imported_name = name.name if name.asname: imported_name = name.asname + "::" + imported_name self.imported_names[imported_name] = module_name def visit_Assign(self, node): """Visit an assignment node.""" line = self._code_lines[node.lineno - 1] if SIG_COMMENT in line: line = _RE_COMMENT_IN_STRING.sub("", line) if (SIG_COMMENT not in line) and (not self.generic): return if SIG_COMMENT in line: _, signature = line.split(SIG_COMMENT) _, return_type, requires = parse_signature(signature) self.required_types |= requires parent = self._parents[-1] for var in node.targets: if isinstance(var, ast.Name): name, p = var.id, parent elif ( isinstance(var, ast.Attribute) and isinstance(var.value, ast.Name) and (var.value.id == "self") ): name, p = var.attr, parent.parent else: name, p = None, None if name is not None: if self.generic: return_type = "Any" self.required_types.add(return_type) stub_node = VariableNode(name, return_type) p.add_variable(stub_node) def get_function_node(self, node): """Process a function node. :sig: (Union[ast.FunctionDef, ast.AsyncFunctionDef]) -> FunctionNode :param node: Node to process. :return: Generated function node in stub tree. """ decorators = [] for d in node.decorator_list: if hasattr(d, "id"): decorators.append(d.id) elif hasattr(d, "func"): decorators.append(d.func.id) elif hasattr(d, "value"): decorators.append(d.value.id + "." + d.attr) signature = get_signature(node) if signature is None: parent = self._parents[-1] if isinstance(parent, ClassNode) and (node.name == "__init__"): signature = parent.signature if (signature is None) and (not self.generic): return None param_names = [arg.arg if PY3 else arg.id for arg in node.args.args] n_args = len(param_names) if signature is None: param_types, rtype, requires = ["Any"] * n_args, "Any", {"Any"} else: _logger.debug("parsing signature for %s", node.name) param_types, rtype, requires = parse_signature(signature) # TODO: only in classes if ((n_args > 0) and (param_names[0] == "self")) or ( (n_args > 0) and (param_names[0] == "cls") and ("classmethod" in decorators) ): if signature is None: param_types[0] = "" else: param_types.insert(0, "") _logger.debug("parameter types: %s", param_types) _logger.debug("return type: %s", rtype) _logger.debug("required types: %s", requires) self.required_types |= requires if node.args.vararg is not None: param_names.append("*" + (node.args.vararg.arg if PY3 else node.args.vararg)) param_types.append("") if node.args.kwarg is not None: param_names.append("**" + (node.args.kwarg.arg if PY3 else node.args.kwarg)) param_types.append("") kwonly_args = getattr(node.args, "kwonlyargs", []) if len(kwonly_args) > 0: param_names.extend([arg.arg for arg in kwonly_args]) if signature is None: param_types.extend(["Any"] * len(kwonly_args)) if len(param_types) != len(param_names): raise ValueError("Parameter names and types don't match: " + node.name) param_locs = [(a.lineno, a.col_offset) for a in (node.args.args + kwonly_args)] param_defaults = { bisect(param_locs, (d.lineno, d.col_offset)) - 1 for d in node.args.defaults } kwonly_defaults = getattr(node.args, "kw_defaults", []) for i, d in enumerate(kwonly_defaults): if d is not None: param_defaults.add(n_args + i) params = [ (name, type_, i in param_defaults) for i, (name, type_) in enumerate(zip(param_names, param_types)) ] if len(kwonly_args) > 0: params.insert(n_args, ("*", "", False)) stub_node = FunctionNode( node.name, parameters=params, rtype=rtype, decorators=decorators ) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] return stub_node def visit_FunctionDef(self, node): """Visit a function node.""" node = self.get_function_node(node) if node is not None: node._async = False def visit_AsyncFunctionDef(self, node): """Visit an async function node.""" node = self.get_function_node(node) if node is not None: node._async = True def visit_ClassDef(self, node): """Visit a class node.""" self.defined_types.add(node.name) bases = [] for n in node.bases: base_parts = [] while True: if not isinstance(n, ast.Attribute): base_parts.append(n.id) break else: base_parts.append(n.attr) n = n.value bases.append(".".join(base_parts[::-1])) self.required_types |= set(bases) signature = get_signature(node) stub_node = ClassNode(node.name, bases=bases, signature=signature) self._parents[-1].add_child(stub_node) self._parents.append(stub_node) self.generic_visit(node) del self._parents[-1] @staticmethod def generate_import_from(module_, names): """Generate an import line. :sig: (str, Set[str]) -> str :param module_: Name of module to import the names from. :param names: Names to import. :return: Import line in stub code. """ regular_names = [n for n in names if "::" not in n] as_names = [n for n in names if "::" in n] line = "" if len(regular_names) > 0: slots = {"m": module_, "n": ", ".join(sorted(regular_names))} line = "from %(m)s import %(n)s" % slots if len(line) > LINE_LENGTH_LIMIT: slots["n"] = INDENT + (",\n" + INDENT).join(sorted(regular_names)) + "," line = "from %(m)s import (\n%(n)s\n)" % slots if len(as_names) > 0: line += "\n" for as_name in as_names: a, n = as_name.split("::") line += "from %(m)s import %(n)s as %(a)s" % {"m": module_, "n": n, "a": a} return line
rfverbruggen/rachiopy
rachiopy/__init__.py
Rachio._request
python
def _request(self, path, method, body=None): url = '/'.join([_SERVER, path]) (resp, content) = _HTTP.request(url, method, headers=self._headers, body=body) content_type = resp.get('content-type') if content_type and content_type.startswith('application/json'): content = json.loads(content.decode('UTF-8')) return (resp, content)
Make a request from the API.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/__init__.py#L32-L42
null
class Rachio(object): """Represent the Rachio API.""" def __init__(self, authtoken): """Rachio class initializer.""" self._headers = {'Content-Type': 'application/json', 'Authorization': 'Bearer %s' % authtoken} self.person = Person(self) self.device = Device(self) self.zone = Zone(self) self.schedulerule = Schedulerule(self) self.flexschedulerule = FlexSchedulerule(self) self.notification = Notification(self) def get(self, path): """Make a GET request from the API.""" return self._request(path, 'GET') def delete(self, path): """Make a DELETE request from the API.""" return self._request(path, 'DELETE') def put(self, path, payload): """Make a PUT request from the API.""" body = json.dumps(payload) return self._request(path, 'PUT', body) def post(self, path, payload): """Make a POST request from the API.""" body = json.dumps(payload) return self._request(path, 'POST', body)
rfverbruggen/rachiopy
rachiopy/__init__.py
Rachio.put
python
def put(self, path, payload): body = json.dumps(payload) return self._request(path, 'PUT', body)
Make a PUT request from the API.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/__init__.py#L52-L55
[ "def _request(self, path, method, body=None):\n \"\"\"Make a request from the API.\"\"\"\n url = '/'.join([_SERVER, path])\n (resp, content) = _HTTP.request(url, method,\n headers=self._headers, body=body)\n\n content_type = resp.get('content-type')\n if content_typ...
class Rachio(object): """Represent the Rachio API.""" def __init__(self, authtoken): """Rachio class initializer.""" self._headers = {'Content-Type': 'application/json', 'Authorization': 'Bearer %s' % authtoken} self.person = Person(self) self.device = Device(self) self.zone = Zone(self) self.schedulerule = Schedulerule(self) self.flexschedulerule = FlexSchedulerule(self) self.notification = Notification(self) def _request(self, path, method, body=None): """Make a request from the API.""" url = '/'.join([_SERVER, path]) (resp, content) = _HTTP.request(url, method, headers=self._headers, body=body) content_type = resp.get('content-type') if content_type and content_type.startswith('application/json'): content = json.loads(content.decode('UTF-8')) return (resp, content) def get(self, path): """Make a GET request from the API.""" return self._request(path, 'GET') def delete(self, path): """Make a DELETE request from the API.""" return self._request(path, 'DELETE') def post(self, path, payload): """Make a POST request from the API.""" body = json.dumps(payload) return self._request(path, 'POST', body)
rfverbruggen/rachiopy
rachiopy/__init__.py
Rachio.post
python
def post(self, path, payload): body = json.dumps(payload) return self._request(path, 'POST', body)
Make a POST request from the API.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/__init__.py#L57-L60
[ "def _request(self, path, method, body=None):\n \"\"\"Make a request from the API.\"\"\"\n url = '/'.join([_SERVER, path])\n (resp, content) = _HTTP.request(url, method,\n headers=self._headers, body=body)\n\n content_type = resp.get('content-type')\n if content_typ...
class Rachio(object): """Represent the Rachio API.""" def __init__(self, authtoken): """Rachio class initializer.""" self._headers = {'Content-Type': 'application/json', 'Authorization': 'Bearer %s' % authtoken} self.person = Person(self) self.device = Device(self) self.zone = Zone(self) self.schedulerule = Schedulerule(self) self.flexschedulerule = FlexSchedulerule(self) self.notification = Notification(self) def _request(self, path, method, body=None): """Make a request from the API.""" url = '/'.join([_SERVER, path]) (resp, content) = _HTTP.request(url, method, headers=self._headers, body=body) content_type = resp.get('content-type') if content_type and content_type.startswith('application/json'): content = json.loads(content.decode('UTF-8')) return (resp, content) def get(self, path): """Make a GET request from the API.""" return self._request(path, 'GET') def delete(self, path): """Make a DELETE request from the API.""" return self._request(path, 'DELETE') def put(self, path, payload): """Make a PUT request from the API.""" body = json.dumps(payload) return self._request(path, 'PUT', body)
rfverbruggen/rachiopy
rachiopy/schedulerule.py
Schedulerule.skip
python
def skip(self, sched_rule_id): path = 'schedulerule/skip' payload = {'id': sched_rule_id} return self.rachio.put(path, payload)
Skip a schedule rule (watering time).
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/schedulerule.py#L11-L15
null
class Schedulerule(object): """Schedulerule class with methods for /schedulerule/ API calls.""" def __init__(self, rachio): """Schedulerule class initializer.""" self.rachio = rachio def start(self, sched_rule_id): """Start a schedule rule (watering time).""" path = 'schedulerule/start' payload = {'id': sched_rule_id} return self.rachio.put(path, payload) def seasonalAdjustment(self, sched_rule_id, adjustment): """Seasonal adjustment for a schedule rule (watering time). This adjustment amount will be applied to the overall run time of the selected schedule while overriding any current adjustments. """ path = 'schedulerule/seasonal_adjustment' payload = {'id': sched_rule_id, 'adjustment': adjustment} return self.rachio.put(path, payload) def get(self, sched_rule_id): """Retrieve the information for a scheduleRule entity.""" path = '/'.join(['schedulerule', sched_rule_id]) return self.rachio.get(path)
rfverbruggen/rachiopy
rachiopy/schedulerule.py
Schedulerule.start
python
def start(self, sched_rule_id): path = 'schedulerule/start' payload = {'id': sched_rule_id} return self.rachio.put(path, payload)
Start a schedule rule (watering time).
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/schedulerule.py#L17-L21
null
class Schedulerule(object): """Schedulerule class with methods for /schedulerule/ API calls.""" def __init__(self, rachio): """Schedulerule class initializer.""" self.rachio = rachio def skip(self, sched_rule_id): """Skip a schedule rule (watering time).""" path = 'schedulerule/skip' payload = {'id': sched_rule_id} return self.rachio.put(path, payload) def seasonalAdjustment(self, sched_rule_id, adjustment): """Seasonal adjustment for a schedule rule (watering time). This adjustment amount will be applied to the overall run time of the selected schedule while overriding any current adjustments. """ path = 'schedulerule/seasonal_adjustment' payload = {'id': sched_rule_id, 'adjustment': adjustment} return self.rachio.put(path, payload) def get(self, sched_rule_id): """Retrieve the information for a scheduleRule entity.""" path = '/'.join(['schedulerule', sched_rule_id]) return self.rachio.get(path)
rfverbruggen/rachiopy
rachiopy/schedulerule.py
Schedulerule.seasonalAdjustment
python
def seasonalAdjustment(self, sched_rule_id, adjustment): path = 'schedulerule/seasonal_adjustment' payload = {'id': sched_rule_id, 'adjustment': adjustment} return self.rachio.put(path, payload)
Seasonal adjustment for a schedule rule (watering time). This adjustment amount will be applied to the overall run time of the selected schedule while overriding any current adjustments.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/schedulerule.py#L23-L31
null
class Schedulerule(object): """Schedulerule class with methods for /schedulerule/ API calls.""" def __init__(self, rachio): """Schedulerule class initializer.""" self.rachio = rachio def skip(self, sched_rule_id): """Skip a schedule rule (watering time).""" path = 'schedulerule/skip' payload = {'id': sched_rule_id} return self.rachio.put(path, payload) def start(self, sched_rule_id): """Start a schedule rule (watering time).""" path = 'schedulerule/start' payload = {'id': sched_rule_id} return self.rachio.put(path, payload) def get(self, sched_rule_id): """Retrieve the information for a scheduleRule entity.""" path = '/'.join(['schedulerule', sched_rule_id]) return self.rachio.get(path)
rfverbruggen/rachiopy
rachiopy/schedulerule.py
Schedulerule.get
python
def get(self, sched_rule_id): path = '/'.join(['schedulerule', sched_rule_id]) return self.rachio.get(path)
Retrieve the information for a scheduleRule entity.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/schedulerule.py#L33-L36
null
class Schedulerule(object): """Schedulerule class with methods for /schedulerule/ API calls.""" def __init__(self, rachio): """Schedulerule class initializer.""" self.rachio = rachio def skip(self, sched_rule_id): """Skip a schedule rule (watering time).""" path = 'schedulerule/skip' payload = {'id': sched_rule_id} return self.rachio.put(path, payload) def start(self, sched_rule_id): """Start a schedule rule (watering time).""" path = 'schedulerule/start' payload = {'id': sched_rule_id} return self.rachio.put(path, payload) def seasonalAdjustment(self, sched_rule_id, adjustment): """Seasonal adjustment for a schedule rule (watering time). This adjustment amount will be applied to the overall run time of the selected schedule while overriding any current adjustments. """ path = 'schedulerule/seasonal_adjustment' payload = {'id': sched_rule_id, 'adjustment': adjustment} return self.rachio.put(path, payload)
rfverbruggen/rachiopy
rachiopy/zone.py
Zone.start
python
def start(self, zone_id, duration): path = 'zone/start' payload = {'id': zone_id, 'duration': duration} return self.rachio.put(path, payload)
Start a zone.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/zone.py#L11-L15
null
class Zone(object): """Zone class with methods for /zone/ API calls.""" def __init__(self, rachio): """Zone class initializer.""" self.rachio = rachio def startMultiple(self, zones): """Start multiple zones.""" path = 'zone/start_multiple' payload = {'zones': zones} return self.rachio.put(path, payload) def schedule(self): """Create an empty zone schedule.""" return ZoneSchedule(self) def get(self, zone_id): """Retrieve the information for a zone entity.""" path = '/'.join(['zone', zone_id]) return self.rachio.get(path)
rfverbruggen/rachiopy
rachiopy/zone.py
Zone.startMultiple
python
def startMultiple(self, zones): path = 'zone/start_multiple' payload = {'zones': zones} return self.rachio.put(path, payload)
Start multiple zones.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/zone.py#L17-L21
null
class Zone(object): """Zone class with methods for /zone/ API calls.""" def __init__(self, rachio): """Zone class initializer.""" self.rachio = rachio def start(self, zone_id, duration): """Start a zone.""" path = 'zone/start' payload = {'id': zone_id, 'duration': duration} return self.rachio.put(path, payload) def schedule(self): """Create an empty zone schedule.""" return ZoneSchedule(self) def get(self, zone_id): """Retrieve the information for a zone entity.""" path = '/'.join(['zone', zone_id]) return self.rachio.get(path)
rfverbruggen/rachiopy
rachiopy/zone.py
Zone.get
python
def get(self, zone_id): path = '/'.join(['zone', zone_id]) return self.rachio.get(path)
Retrieve the information for a zone entity.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/zone.py#L27-L30
null
class Zone(object): """Zone class with methods for /zone/ API calls.""" def __init__(self, rachio): """Zone class initializer.""" self.rachio = rachio def start(self, zone_id, duration): """Start a zone.""" path = 'zone/start' payload = {'id': zone_id, 'duration': duration} return self.rachio.put(path, payload) def startMultiple(self, zones): """Start multiple zones.""" path = 'zone/start_multiple' payload = {'zones': zones} return self.rachio.put(path, payload) def schedule(self): """Create an empty zone schedule.""" return ZoneSchedule(self)
rfverbruggen/rachiopy
rachiopy/zone.py
ZoneSchedule.start
python
def start(self): zones = [{"id": data[0], "duration": data[1], "sortOrder": count} for (count, data) in enumerate(self._zones, 1)] self._api.startMultiple(zones)
Start the schedule.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/zone.py#L45-L49
null
class ZoneSchedule(object): """Help with starting multiple zones.""" def __init__(self, zone_api): """Zoneschedule class initializer.""" self._api = zone_api self._zones = [] def enqueue(self, zone_id, duration): """Add a zone and duration to the schedule.""" self._zones.append((zone_id, duration)) def __enter__(self): """Allow a schedule to be created in a with block.""" return self def __exit__(self, exc_type, exc_val, exc_tb): """Allow the schedule to be executed by leaving with block.""" self.start()
rfverbruggen/rachiopy
rachiopy/flexschedulerule.py
FlexSchedulerule.get
python
def get(self, flex_sched_rule_id): path = '/'.join(['flexschedulerule', flex_sched_rule_id]) return self.rachio.get(path)
Retrieve the information for a flexscheduleRule entity.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/flexschedulerule.py#L11-L14
null
class FlexSchedulerule(object): """FlexSchedulerule class with methods for /flexschedulerule/ calls.""" def __init__(self, rachio): """Flexschedulerule class initializer.""" self.rachio = rachio
rfverbruggen/rachiopy
rachiopy/device.py
Device.get
python
def get(self, dev_id): path = '/'.join(['device', dev_id]) return self.rachio.get(path)
Retrieve the information for a device entity.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/device.py#L11-L14
null
class Device(object): """Device class with /device/ API calls.""" def __init__(self, rachio): """Device class initializer.""" self.rachio = rachio def getCurrentSchedule(self, dev_id): """Retrieve current schedule running, if any.""" path = '/'.join(['device', dev_id, 'current_schedule']) return self.rachio.get(path) def getEvent(self, dev_id, starttime, endtime): """Retrieve events for a device entity.""" path = 'device/%s/event?startTime=%s&endTime=%s' % \ (dev_id, starttime, endtime) return self.rachio.get(path) def getScheduleItem(self, dev_id): """Retrieve the next two weeks of schedule items for a device.""" path = '/'.join(['device', dev_id, 'scheduleitem']) return self.rachio.get(path) def getForecast(self, dev_id, units): """Retrieve current and predicted forecast.""" assert units in ['US', 'METRIC'], 'units must be either US or METRIC' path = 'device/%s/forecast?units=%s' % (dev_id, units) return self.rachio.get(path) def stopWater(self, dev_id): """Stop all watering on device.""" path = 'device/stop_water' payload = {'id': dev_id} return self.rachio.put(path, payload) def rainDelay(self, dev_id, duration): """Rain delay device.""" path = 'device/rain_delay' payload = {'id': dev_id, 'duration': duration} return self.rachio.put(path, payload) def on(self, dev_id): """Turn ON all features of the device. schedules, weather intelligence, water budget, etc. """ path = 'device/on' payload = {'id': dev_id} return self.rachio.put(path, payload) def off(self, dev_id): """Turn OFF all features of the device. schedules, weather intelligence, water budget, etc. """ path = 'device/off' payload = {'id': dev_id} return self.rachio.put(path, payload)
rfverbruggen/rachiopy
rachiopy/device.py
Device.getEvent
python
def getEvent(self, dev_id, starttime, endtime): path = 'device/%s/event?startTime=%s&endTime=%s' % \ (dev_id, starttime, endtime) return self.rachio.get(path)
Retrieve events for a device entity.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/device.py#L21-L25
null
class Device(object): """Device class with /device/ API calls.""" def __init__(self, rachio): """Device class initializer.""" self.rachio = rachio def get(self, dev_id): """Retrieve the information for a device entity.""" path = '/'.join(['device', dev_id]) return self.rachio.get(path) def getCurrentSchedule(self, dev_id): """Retrieve current schedule running, if any.""" path = '/'.join(['device', dev_id, 'current_schedule']) return self.rachio.get(path) def getScheduleItem(self, dev_id): """Retrieve the next two weeks of schedule items for a device.""" path = '/'.join(['device', dev_id, 'scheduleitem']) return self.rachio.get(path) def getForecast(self, dev_id, units): """Retrieve current and predicted forecast.""" assert units in ['US', 'METRIC'], 'units must be either US or METRIC' path = 'device/%s/forecast?units=%s' % (dev_id, units) return self.rachio.get(path) def stopWater(self, dev_id): """Stop all watering on device.""" path = 'device/stop_water' payload = {'id': dev_id} return self.rachio.put(path, payload) def rainDelay(self, dev_id, duration): """Rain delay device.""" path = 'device/rain_delay' payload = {'id': dev_id, 'duration': duration} return self.rachio.put(path, payload) def on(self, dev_id): """Turn ON all features of the device. schedules, weather intelligence, water budget, etc. """ path = 'device/on' payload = {'id': dev_id} return self.rachio.put(path, payload) def off(self, dev_id): """Turn OFF all features of the device. schedules, weather intelligence, water budget, etc. """ path = 'device/off' payload = {'id': dev_id} return self.rachio.put(path, payload)
rfverbruggen/rachiopy
rachiopy/device.py
Device.getForecast
python
def getForecast(self, dev_id, units): assert units in ['US', 'METRIC'], 'units must be either US or METRIC' path = 'device/%s/forecast?units=%s' % (dev_id, units) return self.rachio.get(path)
Retrieve current and predicted forecast.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/device.py#L32-L36
null
class Device(object): """Device class with /device/ API calls.""" def __init__(self, rachio): """Device class initializer.""" self.rachio = rachio def get(self, dev_id): """Retrieve the information for a device entity.""" path = '/'.join(['device', dev_id]) return self.rachio.get(path) def getCurrentSchedule(self, dev_id): """Retrieve current schedule running, if any.""" path = '/'.join(['device', dev_id, 'current_schedule']) return self.rachio.get(path) def getEvent(self, dev_id, starttime, endtime): """Retrieve events for a device entity.""" path = 'device/%s/event?startTime=%s&endTime=%s' % \ (dev_id, starttime, endtime) return self.rachio.get(path) def getScheduleItem(self, dev_id): """Retrieve the next two weeks of schedule items for a device.""" path = '/'.join(['device', dev_id, 'scheduleitem']) return self.rachio.get(path) def stopWater(self, dev_id): """Stop all watering on device.""" path = 'device/stop_water' payload = {'id': dev_id} return self.rachio.put(path, payload) def rainDelay(self, dev_id, duration): """Rain delay device.""" path = 'device/rain_delay' payload = {'id': dev_id, 'duration': duration} return self.rachio.put(path, payload) def on(self, dev_id): """Turn ON all features of the device. schedules, weather intelligence, water budget, etc. """ path = 'device/on' payload = {'id': dev_id} return self.rachio.put(path, payload) def off(self, dev_id): """Turn OFF all features of the device. schedules, weather intelligence, water budget, etc. """ path = 'device/off' payload = {'id': dev_id} return self.rachio.put(path, payload)
rfverbruggen/rachiopy
rachiopy/device.py
Device.stopWater
python
def stopWater(self, dev_id): path = 'device/stop_water' payload = {'id': dev_id} return self.rachio.put(path, payload)
Stop all watering on device.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/device.py#L38-L42
null
class Device(object): """Device class with /device/ API calls.""" def __init__(self, rachio): """Device class initializer.""" self.rachio = rachio def get(self, dev_id): """Retrieve the information for a device entity.""" path = '/'.join(['device', dev_id]) return self.rachio.get(path) def getCurrentSchedule(self, dev_id): """Retrieve current schedule running, if any.""" path = '/'.join(['device', dev_id, 'current_schedule']) return self.rachio.get(path) def getEvent(self, dev_id, starttime, endtime): """Retrieve events for a device entity.""" path = 'device/%s/event?startTime=%s&endTime=%s' % \ (dev_id, starttime, endtime) return self.rachio.get(path) def getScheduleItem(self, dev_id): """Retrieve the next two weeks of schedule items for a device.""" path = '/'.join(['device', dev_id, 'scheduleitem']) return self.rachio.get(path) def getForecast(self, dev_id, units): """Retrieve current and predicted forecast.""" assert units in ['US', 'METRIC'], 'units must be either US or METRIC' path = 'device/%s/forecast?units=%s' % (dev_id, units) return self.rachio.get(path) def rainDelay(self, dev_id, duration): """Rain delay device.""" path = 'device/rain_delay' payload = {'id': dev_id, 'duration': duration} return self.rachio.put(path, payload) def on(self, dev_id): """Turn ON all features of the device. schedules, weather intelligence, water budget, etc. """ path = 'device/on' payload = {'id': dev_id} return self.rachio.put(path, payload) def off(self, dev_id): """Turn OFF all features of the device. schedules, weather intelligence, water budget, etc. """ path = 'device/off' payload = {'id': dev_id} return self.rachio.put(path, payload)
rfverbruggen/rachiopy
rachiopy/device.py
Device.rainDelay
python
def rainDelay(self, dev_id, duration): path = 'device/rain_delay' payload = {'id': dev_id, 'duration': duration} return self.rachio.put(path, payload)
Rain delay device.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/device.py#L44-L48
null
class Device(object): """Device class with /device/ API calls.""" def __init__(self, rachio): """Device class initializer.""" self.rachio = rachio def get(self, dev_id): """Retrieve the information for a device entity.""" path = '/'.join(['device', dev_id]) return self.rachio.get(path) def getCurrentSchedule(self, dev_id): """Retrieve current schedule running, if any.""" path = '/'.join(['device', dev_id, 'current_schedule']) return self.rachio.get(path) def getEvent(self, dev_id, starttime, endtime): """Retrieve events for a device entity.""" path = 'device/%s/event?startTime=%s&endTime=%s' % \ (dev_id, starttime, endtime) return self.rachio.get(path) def getScheduleItem(self, dev_id): """Retrieve the next two weeks of schedule items for a device.""" path = '/'.join(['device', dev_id, 'scheduleitem']) return self.rachio.get(path) def getForecast(self, dev_id, units): """Retrieve current and predicted forecast.""" assert units in ['US', 'METRIC'], 'units must be either US or METRIC' path = 'device/%s/forecast?units=%s' % (dev_id, units) return self.rachio.get(path) def stopWater(self, dev_id): """Stop all watering on device.""" path = 'device/stop_water' payload = {'id': dev_id} return self.rachio.put(path, payload) def on(self, dev_id): """Turn ON all features of the device. schedules, weather intelligence, water budget, etc. """ path = 'device/on' payload = {'id': dev_id} return self.rachio.put(path, payload) def off(self, dev_id): """Turn OFF all features of the device. schedules, weather intelligence, water budget, etc. """ path = 'device/off' payload = {'id': dev_id} return self.rachio.put(path, payload)
rfverbruggen/rachiopy
rachiopy/device.py
Device.on
python
def on(self, dev_id): path = 'device/on' payload = {'id': dev_id} return self.rachio.put(path, payload)
Turn ON all features of the device. schedules, weather intelligence, water budget, etc.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/device.py#L50-L57
null
class Device(object): """Device class with /device/ API calls.""" def __init__(self, rachio): """Device class initializer.""" self.rachio = rachio def get(self, dev_id): """Retrieve the information for a device entity.""" path = '/'.join(['device', dev_id]) return self.rachio.get(path) def getCurrentSchedule(self, dev_id): """Retrieve current schedule running, if any.""" path = '/'.join(['device', dev_id, 'current_schedule']) return self.rachio.get(path) def getEvent(self, dev_id, starttime, endtime): """Retrieve events for a device entity.""" path = 'device/%s/event?startTime=%s&endTime=%s' % \ (dev_id, starttime, endtime) return self.rachio.get(path) def getScheduleItem(self, dev_id): """Retrieve the next two weeks of schedule items for a device.""" path = '/'.join(['device', dev_id, 'scheduleitem']) return self.rachio.get(path) def getForecast(self, dev_id, units): """Retrieve current and predicted forecast.""" assert units in ['US', 'METRIC'], 'units must be either US or METRIC' path = 'device/%s/forecast?units=%s' % (dev_id, units) return self.rachio.get(path) def stopWater(self, dev_id): """Stop all watering on device.""" path = 'device/stop_water' payload = {'id': dev_id} return self.rachio.put(path, payload) def rainDelay(self, dev_id, duration): """Rain delay device.""" path = 'device/rain_delay' payload = {'id': dev_id, 'duration': duration} return self.rachio.put(path, payload) def off(self, dev_id): """Turn OFF all features of the device. schedules, weather intelligence, water budget, etc. """ path = 'device/off' payload = {'id': dev_id} return self.rachio.put(path, payload)
rfverbruggen/rachiopy
rachiopy/device.py
Device.off
python
def off(self, dev_id): path = 'device/off' payload = {'id': dev_id} return self.rachio.put(path, payload)
Turn OFF all features of the device. schedules, weather intelligence, water budget, etc.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/device.py#L59-L66
null
class Device(object): """Device class with /device/ API calls.""" def __init__(self, rachio): """Device class initializer.""" self.rachio = rachio def get(self, dev_id): """Retrieve the information for a device entity.""" path = '/'.join(['device', dev_id]) return self.rachio.get(path) def getCurrentSchedule(self, dev_id): """Retrieve current schedule running, if any.""" path = '/'.join(['device', dev_id, 'current_schedule']) return self.rachio.get(path) def getEvent(self, dev_id, starttime, endtime): """Retrieve events for a device entity.""" path = 'device/%s/event?startTime=%s&endTime=%s' % \ (dev_id, starttime, endtime) return self.rachio.get(path) def getScheduleItem(self, dev_id): """Retrieve the next two weeks of schedule items for a device.""" path = '/'.join(['device', dev_id, 'scheduleitem']) return self.rachio.get(path) def getForecast(self, dev_id, units): """Retrieve current and predicted forecast.""" assert units in ['US', 'METRIC'], 'units must be either US or METRIC' path = 'device/%s/forecast?units=%s' % (dev_id, units) return self.rachio.get(path) def stopWater(self, dev_id): """Stop all watering on device.""" path = 'device/stop_water' payload = {'id': dev_id} return self.rachio.put(path, payload) def rainDelay(self, dev_id, duration): """Rain delay device.""" path = 'device/rain_delay' payload = {'id': dev_id, 'duration': duration} return self.rachio.put(path, payload) def on(self, dev_id): """Turn ON all features of the device. schedules, weather intelligence, water budget, etc. """ path = 'device/on' payload = {'id': dev_id} return self.rachio.put(path, payload)
rfverbruggen/rachiopy
rachiopy/notification.py
Notification.postWebhook
python
def postWebhook(self, dev_id, external_id, url, event_types): path = 'notification/webhook' payload = {'device': {'id': dev_id}, 'externalId': external_id, 'url': url, 'eventTypes': event_types} return self.rachio.post(path, payload)
Add a webhook to a device. externalId can be used as opaque data that is tied to your company, and passed back in each webhook event response.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/notification.py#L24-L34
null
class Notification(object): """Notification class with methods for /notification/ API calls.""" def __init__(self, rachio): """Notification class initializer.""" self.rachio = rachio def getWebhookEventType(self): """Retrieve the list of events types. Event types that are available to any webhook for subscription. """ path = 'notification/webhook_event_type' return self.rachio.get(path) def getDeviceWebhook(self, dev_id): """Retrieve all webhooks for a device.""" path = '/'.join(['notification', dev_id, 'webhook']) return self.rachio.get(path) def putWebhook(self, hook_id, external_id, url, event_types): """Update a webhook.""" path = 'notification/webhook' payload = {'id': hook_id, 'externalId': external_id, 'url': url, 'eventTypes': event_types} return self.rachio.put(path, payload) def deleteWebhook(self, hook_id): """Remove a webhook.""" path = '/'.join(['notification', 'webhook', hook_id]) return self.rachio.delete(path) def get(self, hook_id): """Get a webhook.""" path = '/'.join(['notification', 'webhook', hook_id]) return self.rachio.get(path)
rfverbruggen/rachiopy
rachiopy/notification.py
Notification.putWebhook
python
def putWebhook(self, hook_id, external_id, url, event_types): path = 'notification/webhook' payload = {'id': hook_id, 'externalId': external_id, 'url': url, 'eventTypes': event_types} return self.rachio.put(path, payload)
Update a webhook.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/notification.py#L36-L41
null
class Notification(object): """Notification class with methods for /notification/ API calls.""" def __init__(self, rachio): """Notification class initializer.""" self.rachio = rachio def getWebhookEventType(self): """Retrieve the list of events types. Event types that are available to any webhook for subscription. """ path = 'notification/webhook_event_type' return self.rachio.get(path) def getDeviceWebhook(self, dev_id): """Retrieve all webhooks for a device.""" path = '/'.join(['notification', dev_id, 'webhook']) return self.rachio.get(path) def postWebhook(self, dev_id, external_id, url, event_types): """Add a webhook to a device. externalId can be used as opaque data that is tied to your company, and passed back in each webhook event response. """ path = 'notification/webhook' payload = {'device': {'id': dev_id}, 'externalId': external_id, 'url': url, 'eventTypes': event_types} return self.rachio.post(path, payload) def deleteWebhook(self, hook_id): """Remove a webhook.""" path = '/'.join(['notification', 'webhook', hook_id]) return self.rachio.delete(path) def get(self, hook_id): """Get a webhook.""" path = '/'.join(['notification', 'webhook', hook_id]) return self.rachio.get(path)
rfverbruggen/rachiopy
rachiopy/notification.py
Notification.deleteWebhook
python
def deleteWebhook(self, hook_id): path = '/'.join(['notification', 'webhook', hook_id]) return self.rachio.delete(path)
Remove a webhook.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/notification.py#L43-L46
null
class Notification(object): """Notification class with methods for /notification/ API calls.""" def __init__(self, rachio): """Notification class initializer.""" self.rachio = rachio def getWebhookEventType(self): """Retrieve the list of events types. Event types that are available to any webhook for subscription. """ path = 'notification/webhook_event_type' return self.rachio.get(path) def getDeviceWebhook(self, dev_id): """Retrieve all webhooks for a device.""" path = '/'.join(['notification', dev_id, 'webhook']) return self.rachio.get(path) def postWebhook(self, dev_id, external_id, url, event_types): """Add a webhook to a device. externalId can be used as opaque data that is tied to your company, and passed back in each webhook event response. """ path = 'notification/webhook' payload = {'device': {'id': dev_id}, 'externalId': external_id, 'url': url, 'eventTypes': event_types} return self.rachio.post(path, payload) def putWebhook(self, hook_id, external_id, url, event_types): """Update a webhook.""" path = 'notification/webhook' payload = {'id': hook_id, 'externalId': external_id, 'url': url, 'eventTypes': event_types} return self.rachio.put(path, payload) def get(self, hook_id): """Get a webhook.""" path = '/'.join(['notification', 'webhook', hook_id]) return self.rachio.get(path)
rfverbruggen/rachiopy
rachiopy/notification.py
Notification.get
python
def get(self, hook_id): path = '/'.join(['notification', 'webhook', hook_id]) return self.rachio.get(path)
Get a webhook.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/notification.py#L48-L51
null
class Notification(object): """Notification class with methods for /notification/ API calls.""" def __init__(self, rachio): """Notification class initializer.""" self.rachio = rachio def getWebhookEventType(self): """Retrieve the list of events types. Event types that are available to any webhook for subscription. """ path = 'notification/webhook_event_type' return self.rachio.get(path) def getDeviceWebhook(self, dev_id): """Retrieve all webhooks for a device.""" path = '/'.join(['notification', dev_id, 'webhook']) return self.rachio.get(path) def postWebhook(self, dev_id, external_id, url, event_types): """Add a webhook to a device. externalId can be used as opaque data that is tied to your company, and passed back in each webhook event response. """ path = 'notification/webhook' payload = {'device': {'id': dev_id}, 'externalId': external_id, 'url': url, 'eventTypes': event_types} return self.rachio.post(path, payload) def putWebhook(self, hook_id, external_id, url, event_types): """Update a webhook.""" path = 'notification/webhook' payload = {'id': hook_id, 'externalId': external_id, 'url': url, 'eventTypes': event_types} return self.rachio.put(path, payload) def deleteWebhook(self, hook_id): """Remove a webhook.""" path = '/'.join(['notification', 'webhook', hook_id]) return self.rachio.delete(path)
rfverbruggen/rachiopy
rachiopy/person.py
Person.get
python
def get(self, user_id): path = '/'.join(['person', user_id]) return self.rachio.get(path)
Retrieve the information for a person entity.
train
https://github.com/rfverbruggen/rachiopy/blob/c91abc9984f0f453e60fa905285c1b640c3390ae/rachiopy/person.py#L16-L19
null
class Person(object): """Person class with methods for /person/ API calls.""" def __init__(self, rachio): """Person class initializer.""" self.rachio = rachio def getInfo(self): """Retrieve the id for the person entity currently logged in.""" path = 'person/info' return self.rachio.get(path)
tapilab/brandelion
brandelion/cli/analyze.py
parse_json
python
def parse_json(json_file, include_date=False): if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e)
Yield screen_name, text tuples from a json file.
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L50-L71
null
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def extract_tweets(json_file): """ Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user. """ for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets) def preprocess(s): """ >>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t' """ # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower() def vectorize(json_file, vec, dofit=True): """ Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account. """ ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): """ Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids. """ result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result def iter_follower_file(fname): """ Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ... """ with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:]) # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def jaccard_merge(brands, exemplars): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores def compute_log_degrees(brands, exemplars): """ For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower. """ counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) def proportion_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores # COSINE SIMILARITY def _cosine(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b))) def cosine(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the cosine similarity betwee a brand's followers and the exemplars. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def cosine_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores def adamic(brands, exemplars): """ Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.""" print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores def compute_rarity_scores(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores def rarity(brands, exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def compute_rarity_scores_log(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores def rarity_log(brands, exemplars): """ Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/analyze.py
extract_tweets
python
def extract_tweets(json_file): for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets)
Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user.
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L74-L79
[ "def parse_json(json_file, include_date=False):\n \"\"\" Yield screen_name, text tuples from a json file. \"\"\"\n if json_file[-2:] == 'gz':\n fh = gzip.open(json_file, 'rt')\n else:\n fh = io.open(json_file, mode='rt', encoding='utf8')\n for line in fh:\n try:\n jj = js...
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def parse_json(json_file, include_date=False): """ Yield screen_name, text tuples from a json file. """ if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e) def preprocess(s): """ >>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t' """ # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower() def vectorize(json_file, vec, dofit=True): """ Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account. """ ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): """ Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids. """ result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result def iter_follower_file(fname): """ Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ... """ with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:]) # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def jaccard_merge(brands, exemplars): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores def compute_log_degrees(brands, exemplars): """ For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower. """ counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) def proportion_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores # COSINE SIMILARITY def _cosine(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b))) def cosine(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the cosine similarity betwee a brand's followers and the exemplars. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def cosine_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores def adamic(brands, exemplars): """ Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.""" print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores def compute_rarity_scores(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores def rarity(brands, exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def compute_rarity_scores_log(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores def rarity_log(brands, exemplars): """ Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/analyze.py
preprocess
python
def preprocess(s): # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower()
>>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t'
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L82-L96
null
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def parse_json(json_file, include_date=False): """ Yield screen_name, text tuples from a json file. """ if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e) def extract_tweets(json_file): """ Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user. """ for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets) def vectorize(json_file, vec, dofit=True): """ Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account. """ ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): """ Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids. """ result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result def iter_follower_file(fname): """ Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ... """ with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:]) # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def jaccard_merge(brands, exemplars): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores def compute_log_degrees(brands, exemplars): """ For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower. """ counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) def proportion_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores # COSINE SIMILARITY def _cosine(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b))) def cosine(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the cosine similarity betwee a brand's followers and the exemplars. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def cosine_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores def adamic(brands, exemplars): """ Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.""" print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores def compute_rarity_scores(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores def rarity(brands, exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def compute_rarity_scores_log(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores def rarity_log(brands, exemplars): """ Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/analyze.py
vectorize
python
def vectorize(json_file, vec, dofit=True): ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X
Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account.
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L99-L109
[ "def extract_tweets(json_file):\n \"\"\" Yield screen_name, string tuples, where the string is the\n concatenation of all tweets of this user. \"\"\"\n for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]):\n tweets = [t[1] for t in tweet_iter]\n yield screen_name, ' '...
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def parse_json(json_file, include_date=False): """ Yield screen_name, text tuples from a json file. """ if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e) def extract_tweets(json_file): """ Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user. """ for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets) def preprocess(s): """ >>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t' """ # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower() def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): """ Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids. """ result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result def iter_follower_file(fname): """ Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ... """ with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:]) # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def jaccard_merge(brands, exemplars): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores def compute_log_degrees(brands, exemplars): """ For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower. """ counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) def proportion_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores # COSINE SIMILARITY def _cosine(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b))) def cosine(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the cosine similarity betwee a brand's followers and the exemplars. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def cosine_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores def adamic(brands, exemplars): """ Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.""" print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores def compute_rarity_scores(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores def rarity(brands, exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def compute_rarity_scores_log(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores def rarity_log(brands, exemplars): """ Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/analyze.py
read_follower_file
python
def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result
Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids.
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L171-L184
null
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def parse_json(json_file, include_date=False): """ Yield screen_name, text tuples from a json file. """ if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e) def extract_tweets(json_file): """ Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user. """ for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets) def preprocess(s): """ >>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t' """ # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower() def vectorize(json_file, vec, dofit=True): """ Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account. """ ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def iter_follower_file(fname): """ Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ... """ with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:]) # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def jaccard_merge(brands, exemplars): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores def compute_log_degrees(brands, exemplars): """ For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower. """ counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) def proportion_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores # COSINE SIMILARITY def _cosine(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b))) def cosine(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the cosine similarity betwee a brand's followers and the exemplars. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def cosine_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores def adamic(brands, exemplars): """ Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.""" print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores def compute_rarity_scores(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores def rarity(brands, exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def compute_rarity_scores_log(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores def rarity_log(brands, exemplars): """ Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/analyze.py
iter_follower_file
python
def iter_follower_file(fname): with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:])
Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ...
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L187-L196
null
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def parse_json(json_file, include_date=False): """ Yield screen_name, text tuples from a json file. """ if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e) def extract_tweets(json_file): """ Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user. """ for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets) def preprocess(s): """ >>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t' """ # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower() def vectorize(json_file, vec, dofit=True): """ Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account. """ ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): """ Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids. """ result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def jaccard_merge(brands, exemplars): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores def compute_log_degrees(brands, exemplars): """ For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower. """ counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) def proportion_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores # COSINE SIMILARITY def _cosine(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b))) def cosine(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the cosine similarity betwee a brand's followers and the exemplars. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def cosine_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores def adamic(brands, exemplars): """ Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.""" print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores def compute_rarity_scores(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores def rarity(brands, exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def compute_rarity_scores_log(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores def rarity_log(brands, exemplars): """ Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/analyze.py
jaccard_merge
python
def jaccard_merge(brands, exemplars): scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores
Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L235-L246
[ "def _jaccard(a, b):\n \"\"\" Return the Jaccard similarity between two sets a and b. \"\"\"\n return 1. * len(a & b) / len(a | b)\n" ]
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def parse_json(json_file, include_date=False): """ Yield screen_name, text tuples from a json file. """ if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e) def extract_tweets(json_file): """ Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user. """ for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets) def preprocess(s): """ >>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t' """ # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower() def vectorize(json_file, vec, dofit=True): """ Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account. """ ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): """ Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids. """ result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result def iter_follower_file(fname): """ Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ... """ with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:]) # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def compute_log_degrees(brands, exemplars): """ For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower. """ counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) def proportion_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores # COSINE SIMILARITY def _cosine(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b))) def cosine(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the cosine similarity betwee a brand's followers and the exemplars. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def cosine_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores def adamic(brands, exemplars): """ Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.""" print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores def compute_rarity_scores(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores def rarity(brands, exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def compute_rarity_scores_log(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores def rarity_log(brands, exemplars): """ Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/analyze.py
compute_log_degrees
python
def compute_log_degrees(brands, exemplars): counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts
For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower.
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L249-L259
null
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def parse_json(json_file, include_date=False): """ Yield screen_name, text tuples from a json file. """ if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e) def extract_tweets(json_file): """ Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user. """ for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets) def preprocess(s): """ >>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t' """ # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower() def vectorize(json_file, vec, dofit=True): """ Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account. """ ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): """ Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids. """ result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result def iter_follower_file(fname): """ Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ... """ with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:]) # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def jaccard_merge(brands, exemplars): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) def proportion_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores # COSINE SIMILARITY def _cosine(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b))) def cosine(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the cosine similarity betwee a brand's followers and the exemplars. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def cosine_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores def adamic(brands, exemplars): """ Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.""" print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores def compute_rarity_scores(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores def rarity(brands, exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def compute_rarity_scores_log(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores def rarity_log(brands, exemplars): """ Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/analyze.py
proportion_merge
python
def proportion_merge(brands, exemplars): scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores
Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L298-L308
[ "def _proportion(a, b):\n \"\"\" Return the len(a & b) / len(a) \"\"\"\n return 1. * len(a & b) / len(a)\n" ]
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def parse_json(json_file, include_date=False): """ Yield screen_name, text tuples from a json file. """ if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e) def extract_tweets(json_file): """ Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user. """ for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets) def preprocess(s): """ >>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t' """ # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower() def vectorize(json_file, vec, dofit=True): """ Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account. """ ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): """ Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids. """ result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result def iter_follower_file(fname): """ Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ... """ with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:]) # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def jaccard_merge(brands, exemplars): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores def compute_log_degrees(brands, exemplars): """ For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower. """ counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) # COSINE SIMILARITY def _cosine(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b))) def cosine(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the cosine similarity betwee a brand's followers and the exemplars. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def cosine_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores def adamic(brands, exemplars): """ Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.""" print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores def compute_rarity_scores(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores def rarity(brands, exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def compute_rarity_scores_log(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores def rarity_log(brands, exemplars): """ Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/analyze.py
_cosine
python
def _cosine(a, b): return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b)))
Return the len(a & b) / len(a)
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L314-L316
null
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def parse_json(json_file, include_date=False): """ Yield screen_name, text tuples from a json file. """ if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e) def extract_tweets(json_file): """ Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user. """ for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets) def preprocess(s): """ >>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t' """ # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower() def vectorize(json_file, vec, dofit=True): """ Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account. """ ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): """ Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids. """ result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result def iter_follower_file(fname): """ Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ... """ with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:]) # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def jaccard_merge(brands, exemplars): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores def compute_log_degrees(brands, exemplars): """ For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower. """ counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) def proportion_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores # COSINE SIMILARITY def cosine(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the cosine similarity betwee a brand's followers and the exemplars. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def cosine_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores def adamic(brands, exemplars): """ Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.""" print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores def compute_rarity_scores(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores def rarity(brands, exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def compute_rarity_scores_log(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores def rarity_log(brands, exemplars): """ Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/analyze.py
cosine
python
def cosine(brands, exemplars, weighted_avg=False, sqrt=False): scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores
Return the cosine similarity betwee a brand's followers and the exemplars.
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L319-L332
null
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def parse_json(json_file, include_date=False): """ Yield screen_name, text tuples from a json file. """ if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e) def extract_tweets(json_file): """ Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user. """ for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets) def preprocess(s): """ >>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t' """ # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower() def vectorize(json_file, vec, dofit=True): """ Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account. """ ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): """ Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids. """ result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result def iter_follower_file(fname): """ Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ... """ with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:]) # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def jaccard_merge(brands, exemplars): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores def compute_log_degrees(brands, exemplars): """ For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower. """ counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) def proportion_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores # COSINE SIMILARITY def _cosine(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b))) def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def cosine_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores def adamic(brands, exemplars): """ Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.""" print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores def compute_rarity_scores(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores def rarity(brands, exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def compute_rarity_scores_log(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores def rarity_log(brands, exemplars): """ Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/analyze.py
cosine_merge
python
def cosine_merge(brands, exemplars): scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores
Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L347-L357
[ "def _cosine(a, b):\n \"\"\" Return the len(a & b) / len(a) \"\"\"\n return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b)))\n" ]
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def parse_json(json_file, include_date=False): """ Yield screen_name, text tuples from a json file. """ if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e) def extract_tweets(json_file): """ Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user. """ for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets) def preprocess(s): """ >>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t' """ # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower() def vectorize(json_file, vec, dofit=True): """ Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account. """ ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): """ Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids. """ result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result def iter_follower_file(fname): """ Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ... """ with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:]) # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def jaccard_merge(brands, exemplars): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores def compute_log_degrees(brands, exemplars): """ For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower. """ counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) def proportion_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores # COSINE SIMILARITY def _cosine(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b))) def cosine(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the cosine similarity betwee a brand's followers and the exemplars. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def adamic(brands, exemplars): """ Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.""" print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores def compute_rarity_scores(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores def rarity(brands, exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def compute_rarity_scores_log(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores def rarity_log(brands, exemplars): """ Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/analyze.py
adamic
python
def adamic(brands, exemplars): print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores
Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L360-L376
null
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def parse_json(json_file, include_date=False): """ Yield screen_name, text tuples from a json file. """ if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e) def extract_tweets(json_file): """ Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user. """ for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets) def preprocess(s): """ >>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t' """ # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower() def vectorize(json_file, vec, dofit=True): """ Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account. """ ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): """ Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids. """ result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result def iter_follower_file(fname): """ Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ... """ with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:]) # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def jaccard_merge(brands, exemplars): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores def compute_log_degrees(brands, exemplars): """ For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower. """ counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) def proportion_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores # COSINE SIMILARITY def _cosine(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b))) def cosine(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the cosine similarity betwee a brand's followers and the exemplars. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def cosine_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores def compute_rarity_scores(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores def rarity(brands, exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def compute_rarity_scores_log(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores def rarity_log(brands, exemplars): """ Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/analyze.py
compute_rarity_scores
python
def compute_rarity_scores(exemplars): scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores
Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)]
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L379-L390
null
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def parse_json(json_file, include_date=False): """ Yield screen_name, text tuples from a json file. """ if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e) def extract_tweets(json_file): """ Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user. """ for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets) def preprocess(s): """ >>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t' """ # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower() def vectorize(json_file, vec, dofit=True): """ Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account. """ ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): """ Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids. """ result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result def iter_follower_file(fname): """ Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ... """ with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:]) # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def jaccard_merge(brands, exemplars): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores def compute_log_degrees(brands, exemplars): """ For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower. """ counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) def proportion_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores # COSINE SIMILARITY def _cosine(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b))) def cosine(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the cosine similarity betwee a brand's followers and the exemplars. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def cosine_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores def adamic(brands, exemplars): """ Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.""" print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores def rarity(brands, exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def compute_rarity_scores_log(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores def rarity_log(brands, exemplars): """ Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/analyze.py
rarity
python
def rarity(brands, exemplars): rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores
Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L393-L400
[ "def compute_rarity_scores(exemplars):\n \"\"\" Compute a score for each follower that is sum_i (1/n_i), where n_i is\n the degree of the ith exemplar they follow.\n >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items()\n [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)]\n \"\"\"\n ...
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def parse_json(json_file, include_date=False): """ Yield screen_name, text tuples from a json file. """ if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e) def extract_tweets(json_file): """ Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user. """ for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets) def preprocess(s): """ >>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t' """ # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower() def vectorize(json_file, vec, dofit=True): """ Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account. """ ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): """ Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids. """ result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result def iter_follower_file(fname): """ Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ... """ with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:]) # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def jaccard_merge(brands, exemplars): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores def compute_log_degrees(brands, exemplars): """ For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower. """ counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) def proportion_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores # COSINE SIMILARITY def _cosine(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b))) def cosine(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the cosine similarity betwee a brand's followers and the exemplars. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def cosine_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores def adamic(brands, exemplars): """ Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.""" print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores def compute_rarity_scores(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores def compute_rarity_scores_log(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores def rarity_log(brands, exemplars): """ Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/analyze.py
compute_rarity_scores_log
python
def compute_rarity_scores_log(exemplars): scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores
Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)]
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L403-L414
null
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def parse_json(json_file, include_date=False): """ Yield screen_name, text tuples from a json file. """ if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e) def extract_tweets(json_file): """ Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user. """ for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets) def preprocess(s): """ >>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t' """ # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower() def vectorize(json_file, vec, dofit=True): """ Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account. """ ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): """ Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids. """ result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result def iter_follower_file(fname): """ Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ... """ with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:]) # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def jaccard_merge(brands, exemplars): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores def compute_log_degrees(brands, exemplars): """ For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower. """ counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) def proportion_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores # COSINE SIMILARITY def _cosine(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b))) def cosine(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the cosine similarity betwee a brand's followers and the exemplars. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def cosine_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores def adamic(brands, exemplars): """ Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.""" print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores def compute_rarity_scores(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores def rarity(brands, exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def rarity_log(brands, exemplars): """ Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/analyze.py
rarity_log
python
def rarity_log(brands, exemplars): rarity = compute_rarity_scores_log(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores
Compute a score for each follower that is sum_i (1/log(n_i)), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/analyze.py#L417-L424
[ "def compute_rarity_scores_log(exemplars):\n \"\"\" Compute a score for each follower that is sum_i (1/n_i), where n_i is\n the degree of the ith exemplar they follow.\n >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items()\n [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)]\n \"\"\"\n...
#!/usr/bin/env python # -*- coding: utf-8 -*- """Analyze social and linguistic brand data. usage: brandelion analyze --text --brand-tweets <file> --exemplar-tweets <file> --sample-tweets <file> --output <file> [--text-method <string>] brandelion analyze --network --brand-followers <file> --exemplar-followers <file> --output <file> [--network-method <string> --min-followers <n> --max-followers <n> --sample-exemplars <p> --seed <s>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --brand-tweets <file> File containing tweets from brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --exemplar-tweets <file> File containing tweets from exemplar accounts. --sample-tweets <file> File containing tweets from representative sample of Twitter. --text-method <string> Method to do text analysis [default: chi2] --network-method <string> Method to do text analysis [default: jaccard] -o, --output <file> File to store results -t, --text Analyze text of tweets. -n, --network Analyze followers. --min-followers <n> Ignore exemplars that don't have at least n followers [default: 0] --max-followers <n> Ignore exemplars that have more than least n followers [default: 1e10] --sample-exemplars <p> Sample p percent of the exemplars, uniformly at random. [default: 100] --seed <s> Seed for random sampling. [default: 12345] """ from collections import Counter, defaultdict from docopt import docopt import io from itertools import groupby import gzip import json import math import numpy as np import os import re import random from scipy.sparse import vstack import sys from sklearn.feature_extraction.text import CountVectorizer from sklearn.feature_selection import chi2 as skchi2 from sklearn import linear_model from . import report ### TEXT ANALYSIS ### def parse_json(json_file, include_date=False): """ Yield screen_name, text tuples from a json file. """ if json_file[-2:] == 'gz': fh = gzip.open(json_file, 'rt') else: fh = io.open(json_file, mode='rt', encoding='utf8') for line in fh: try: jj = json.loads(line) if type(jj) is not list: jj = [jj] for j in jj: if include_date: yield (j['user']['screen_name'].lower(), j['text'], j['created_at']) else: if 'full_text' in j: # get untruncated text if available. yield (j['user']['screen_name'].lower(), j['full_text']) else: yield (j['user']['screen_name'].lower(), j['text']) except Exception as e: sys.stderr.write('skipping json error: %s\n' % e) def extract_tweets(json_file): """ Yield screen_name, string tuples, where the string is the concatenation of all tweets of this user. """ for screen_name, tweet_iter in groupby(parse_json(json_file), lambda x: x[0]): tweets = [t[1] for t in tweet_iter] yield screen_name, ' '.join(tweets) def preprocess(s): """ >>> preprocess('#hi there http://www.foo.com @you isn"t RT &lt;&gt;') 'hashtaghi hashtaghi there isn"t' """ # s = re.sub('@\S+', 'thisisamention', s) # map all mentions to thisisamention s = re.sub(r'@\S+', ' ', s) # map all mentions to thisisamention # s = re.sub('http\S+', 'http', s) # keep only http from urls s = re.sub(r'http\S+', ' ', s) # keep only http from urls s = re.sub(r'#(\S+)', r'hashtag\1 hashtag\1', s) # #foo -> hashtagfoo hashtagfoo (for retaining hashtags even using bigrams) # s = re.sub(r'[0-9]+', '9', s) # 1234 -> 9 s = re.sub(r'\bRT\b', ' ', s, re.IGNORECASE) s = re.sub(r'&[a-z]+;', ' ', s, re.IGNORECASE) s = re.sub(r'\s+', ' ', s).strip() return s.lower() def vectorize(json_file, vec, dofit=True): """ Return a matrix where each row corresponds to a Twitter account, and each column corresponds to the number of times a term is used by that account. """ ## CountVectorizer, efficiently. screen_names = [x[0] for x in extract_tweets(json_file)] if dofit: X = vec.fit_transform(x[1] for x in extract_tweets(json_file)) else: X = vec.transform(x[1] for x in extract_tweets(json_file)) return screen_names, X def chi2(exemplars, samples, n=300): y = np.array(([1.] * exemplars.shape[0]) + ([.0] * samples.shape[0])) X = vstack((exemplars, samples)).tocsr() clf = linear_model.LogisticRegression(penalty='l2') clf.fit(X, y) coef = clf.coef_[0] chis, pvals = skchi2(X, y) top_indices = chis.argsort()[::-1] top_indices = [i for i in top_indices if coef[i] > 0] for idx in range(len(coef)): coef[idx] = 0. for idx in top_indices[:n]: coef[idx] = chis[idx] return coef def do_score(vec, coef): return np.sum(coef[vec.nonzero()[1]]) / np.sum(coef) def write_top_words(fname, vocab, scores): outf = io.open(fname, 'w', encoding='utf8') for i in np.argsort(scores)[::-1]: if scores[i] > 0: outf.write('%s %g\n' % (vocab[i], scores[i])) outf.close() def analyze_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, outfile, analyze_fn): analyze = getattr(sys.modules[__name__], analyze_fn) vec = CountVectorizer(min_df=3, preprocessor=preprocess, ngram_range=(2, 2), binary=True) _, exemplar_vectors = vectorize(exemplar_tweets_file, vec, dofit=True) print('read tweets for %d exemplar accounts' % exemplar_vectors.shape[0]) brands, brand_vectors = vectorize(brand_tweets_file, vec, dofit=False) print('read tweets for %d brand accounts' % brand_vectors.shape[0]) _, sample_vectors = vectorize(sample_tweets_file, vec, dofit=False) print('read tweets for %d sample accounts' % sample_vectors.shape[0]) scores = analyze(exemplar_vectors, sample_vectors) vocab = vec.get_feature_names() write_top_words(outfile + '.topwords', vocab, scores) print('top 10 ngrams:\n', '\n'.join(['%s=%.4g' % (vocab[i], scores[i]) for i in np.argsort(scores)[::-1][:10]])) outf = open(outfile, 'wt') for bi, brand_vec in enumerate(brand_vectors): outf.write('%s %g\n' % (brands[bi], do_score(brand_vec, scores))) outf.flush() ### FOLLOWER ANALYSIS ### def get_twitter_handles(fname): handles = set() with open(fname, 'rt') as f: for line in f: handles.add(line[:90].split()[0].lower()) return handles def read_follower_file(fname, min_followers=0, max_followers=1e10, blacklist=set()): """ Read a file of follower information and return a dictionary mapping screen_name to a set of follower ids. """ result = {} with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: if parts[1].lower() not in blacklist: followers = set(int(x) for x in parts[2:]) if len(followers) > min_followers and len(followers) <= max_followers: result[parts[1].lower()] = followers else: print('skipping exemplar', parts[1].lower()) return result def iter_follower_file(fname): """ Iterator from a file of follower information and return a tuple of screen_name, follower ids. File format is: <iso timestamp> <screen_name> <follower_id1> <follower_ids2> ... """ with open(fname, 'rt') as f: for line in f: parts = line.split() if len(parts) > 3: yield parts[1].lower(), set(int(x) for x in parts[2:]) # JACCARD def _jaccard(a, b): """ Return the Jaccard similarity between two sets a and b. """ return 1. * len(a & b) / len(a | b) def jaccard(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_jaccard(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_jaccard(followers, others) for others in exemplars.values()) / len(exemplars) # limit to exemplars with less than 40k followers: scores[brand] = 1. * sum(_jaccard(brands[brand], others) for others in exemplars.itervalues() if len(others) < 40000) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def jaccard_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, True, False) def jaccard_sqrt_no_weighted_avg(brands, exemplars): return jaccard(brands, exemplars, False, True) def jaccard_sqrt(brands, exemplars): return jaccard(brands, exemplars, weighted_avg=True, sqrt=True) def jaccard_merge(brands, exemplars): """ Return the average Jaccard similarity between a brand's followers and the followers of each exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _jaccard(followers, exemplar_followers) return scores def compute_log_degrees(brands, exemplars): """ For each follower, let Z be the total number of brands they follow. Return a dictionary of 1. / log(Z), for each follower. """ counts = Counter() for followers in brands.values(): # + exemplars.values(): # Include exemplars in these counts? No, don't want to penalize people who follow many exemplars. counts.update(followers) counts.update(counts.keys()) # Add 1 to each count. for k in counts: counts[k] = 1. / math.log(counts[k]) return counts # PROPORTION def _proportion(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / len(a) def proportion(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the proportion of a brand's followers who also follow an exemplar. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_proportion(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_proportion(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def proportion_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=False) def proportion_sqrt_no_weighted_avg(brands, exemplars): return proportion(brands, exemplars, weighted_avg=False, sqrt=True) def proportion_sqrt(brands, exemplars): return proportion(brands, exemplars, weighted_avg=True, sqrt=True) def proportion_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _proportion(followers, exemplar_followers) return scores # COSINE SIMILARITY def _cosine(a, b): """ Return the len(a & b) / len(a) """ return 1. * len(a & b) / (math.sqrt(len(a)) * math.sqrt(len(b))) def cosine(brands, exemplars, weighted_avg=False, sqrt=False): """ Return the cosine similarity betwee a brand's followers and the exemplars. """ scores = {} for brand, followers in brands: if weighted_avg: scores[brand] = np.average([_cosine(followers, others) for others in exemplars.values()], weights=[1. / len(others) for others in exemplars.values()]) else: scores[brand] = 1. * sum(_cosine(followers, others) for others in exemplars.values()) / len(exemplars) if sqrt: scores = dict([(b, math.sqrt(s)) for b, s in scores.items()]) return scores def cosine_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=False) def cosine_sqrt_no_weighted_avg(brands, exemplars): return cosine(brands, exemplars, weighted_avg=False, sqrt=True) def cosine_sqrt(brands, exemplars): return cosine(brands, exemplars, weighted_avg=True, sqrt=True) def cosine_merge(brands, exemplars): """ Return the proportion of a brand's followers who also follower an exemplar. We merge all exemplar followers into one big pseudo-account.""" scores = {} exemplar_followers = set() for followers in exemplars.values(): exemplar_followers |= followers for brand, followers in brands: scores[brand] = _cosine(followers, exemplar_followers) return scores def adamic(brands, exemplars): """ Return the average Adamic/Adar similarity between a brand's followers and the followers of each exemplar. We approximate the number of followed accounts per user by only considering those in our brand set.""" print('adamic deprecated...requires loading all brands in memory.') return degrees = compute_log_degrees(brands, exemplars) scores = {} exemplar_sums = dict([(exemplar, sum(degrees[z] for z in exemplars[exemplar])) for exemplar in exemplars]) for brand in sorted(brands): brand_sum = sum(degrees[z] for z in brands[brand]) total = 0. for exemplar in exemplars: total += sum(degrees[z] for z in brands[brand] & exemplars[exemplar]) / (brand_sum + exemplar_sums[exemplar]) scores[brand] = total / len(exemplars) return scores def compute_rarity_scores(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / len(followers) for f in followers: scores[f] += score return scores def rarity(brands, exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. The score for a brand is then the average of their follower scores.""" rarity = compute_rarity_scores(exemplars) scores = {} for brand, followers in brands: scores[brand] = sum(rarity[f] for f in followers) / len(followers) return scores def compute_rarity_scores_log(exemplars): """ Compute a score for each follower that is sum_i (1/n_i), where n_i is the degree of the ith exemplar they follow. >>> compute_rarity_scores({'e1':{1,2,3,4}, 'e2':{4,5}}).items() [(1, 0.25), (2, 0.25), (3, 0.25), (4, 0.75), (5, 0.5)] """ scores = defaultdict(lambda: 0.) for followers in exemplars.values(): score = 1. / math.log(len(followers)) for f in followers: scores[f] += score return scores def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def analyze_followers(brand_follower_file, exemplar_follower_file, outfile, analyze_fn, min_followers, max_followers, sample_exemplars): brands = iter_follower_file(brand_follower_file) exemplars = read_follower_file(exemplar_follower_file, min_followers=min_followers, max_followers=max_followers, blacklist=get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) if sample_exemplars < 100: # sample a subset of exemplars. exemplars = dict([(k, exemplars[k]) for k in random.sample(exemplars.keys(), int(len(exemplars) * sample_exemplars / 100.))]) print('sampled %d exemplars' % (len(exemplars))) analyze = getattr(sys.modules[__name__], analyze_fn) scores = analyze(brands, exemplars) mkdirs(outfile) outf = open(outfile, 'wt') for brand in sorted(scores): outf.write('%s %g\n' % (brand, scores[brand])) outf.flush() outf.close() print('results written to', outfile) def main(): args = docopt(__doc__) print(args) if '--seed' in args: random.seed(args['--seed']) if args['--network']: analyze_followers(args['--brand-followers'], args['--exemplar-followers'], args['--output'], args['--network-method'], int(args['--min-followers']), int(float(args['--max-followers'])), float(args['--sample-exemplars'])) if args['--text']: analyze_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--output'], args['--text-method']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/collect.py
iter_lines
python
def iter_lines(filename): with open(filename, 'rt') as idfile: for line in idfile: screen_name = line.strip() if len(screen_name) > 0: yield screen_name.split()[0]
Iterate over screen names in a file, one per line.
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/collect.py#L47-L53
null
#!/usr/bin/env python # -*- coding: utf-8 -*- """Collect Twitter data for brands. usage: brandelion collect --tweets --input <file> --output <file> --max=<N> brandelion collect --followers --input <file> --output <file> --max=<N> [--loop] brandelion collect --exemplars --query <string> --output <file> Options -h, --help -i, --input <file> File containing list of Twitter accounts, one per line. -l, --loop If true, keep looping to collect more data continuously. -o, --output <file> File to store results -t, --tweets Fetch tweets. -f, --followers Fetch followers -e, --exemplars Fetch exemplars from Twitter lists -q, --query <string> A single string used to search for Twitter lists. -m, --max=<N> Maximum number of followers or tweets to collect per account [default: 50000]. """ from collections import Counter import datetime from docopt import docopt import gzip import io import json import re import requests import sys import time import traceback import requests import twutil import json import time ##import config from init.py: from .. import config from googleapiclient.discovery import build from googleapiclient.errors import HttpError as GoogleHttpError def fetch_followers(account_file, outfile, limit, do_loop): """ Fetch up to limit followers for each Twitter account in account_file. Write results to outfile file in format: screen_name user_id follower_id_1 follower_id_2 ...""" print('Fetching followers for accounts in %s' % account_file) niters = 1 while True: outf = gzip.open(outfile, 'wt') for screen_name in iter_lines(account_file): timestamp = datetime.datetime.now().isoformat() print('collecting followers for', screen_name) followers = twutil.collect.followers_for_screen_name(screen_name, limit) if len(followers) > 0: outf.write('%s %s %s\n' % (timestamp, screen_name, ' '.join(followers))) outf.flush() else: print('unknown user', screen_name) outf.close() if not do_loop: return else: if niters == 1: outfile = '%s.%d' % (outfile, niters) else: outfile = outfile[:outfile.rindex('.')] + '.%d' % niters niters += 1 def fetch_tweets(account_file, outfile, limit): """ Fetch up to limit tweets for each account in account_file and write to outfile. """ print('fetching tweets for accounts in', account_file) outf = io.open(outfile, 'wt') for screen_name in iter_lines(account_file): print('\nFetching tweets for %s' % screen_name) for tweet in twutil.collect.tweets_for_user(screen_name, limit): tweet['user']['screen_name'] = screen_name outf.write('%s\n' % json.dumps(tweet, ensure_ascii=False)) outf.flush() #DEPRECATED # def fetch_lists(keyword, max_results=20): # """ # Fetch the urls of up to max_results Twitter lists that match the provided keyword. # >>> len(fetch_lists('politics', max_results=4)) # 4 # """ # res_per_page = 8 # start = 0 # results = [] # while len(results) < max_results: # url = 'http://ajax.googleapis.com/ajax/services/search/web?v=1.0&q=site:twitter.com+inurl:lists+%s&rsz=%d&start=%d' % (keyword, # res_per_page, # start) # js = json.loads(requests.get(url).text) # if not js['responseData']: # print('something went wrong in google search:\n', js) # return results[:max_results] # else: # for r in js['responseData']['results']: # results.append(r['url']) # start += res_per_page # time.sleep(.4) # return results[:max_results] #NEW FETCH lists def google_search(search_term, api_key, cse_id, **kwargs): final_urls = [] try: service = build("customsearch", "v1", developerKey=api_key) res = service.cse().list(q=search_term, cx=cse_id, **kwargs).execute() for item in res['items']: final_urls.append(item['formattedUrl']) return final_urls except GoogleHttpError: print("something wrong with Google HTTP Errpr") return final_urls def fetch_lists(keyword,max_results=20): """ Fetch the urls of up to max_results Twitter lists that match the provided keyword. >>> len(fetch_lists('politics', max_results=4)) 4 """ #CONFIG FILE READ api_key=config.get('GOOGLE_CSE_KEYS','API_KEY') cse_id=config.get('GOOGLE_CSE_KEYS','CSE_ID') results = [] start_c = 1 search_term = "inurl:lists + "+keyword while len(results)<max_results: temp_res = google_search(search_term,api_key,cse_id,num=10,start=start_c) if len(temp_res) == 0: print("Google API Error, returning retrieved results") return results results.extend(temp_res) start_c += 10 return results[:max_results] def fetch_list_members(list_url): """ Get all members of the list specified by the given url. E.g., https://twitter.com/lore77/lists/libri-cultura-education """ match = re.match(r'.+twitter\.com\/(.+)\/lists\/(.+)', list_url) if not match: print('cannot parse list url %s' % list_url) return [] screen_name, slug = match.groups() print('collecting list %s/%s' % (screen_name, slug)) return twutil.collect.list_members(slug, screen_name) def fetch_exemplars(keyword, outfile, n=50): """ Fetch top lists matching this keyword, then return Twitter screen names along with the number of different lists on which each appers.. """ list_urls = fetch_lists(keyword, n) print('found %d lists for %s' % (len(list_urls), keyword)) counts = Counter() for list_url in list_urls: counts.update(fetch_list_members(list_url)) # Write to file. outf = io.open(outfile, 'wt') for handle in sorted(counts): outf.write('%s\t%d\n' % (handle, counts[handle])) outf.close() print('saved exemplars to', outfile) def main(): args = docopt(__doc__) if args['--followers']: fetch_followers(args['--input'], args['--output'], int(args['--max']), args['--loop']) elif args['--tweets']: fetch_tweets(args['--input'], args['--output'], int(args['--max'])) else: fetch_exemplars(args['--query'], args['--output']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/collect.py
fetch_followers
python
def fetch_followers(account_file, outfile, limit, do_loop): print('Fetching followers for accounts in %s' % account_file) niters = 1 while True: outf = gzip.open(outfile, 'wt') for screen_name in iter_lines(account_file): timestamp = datetime.datetime.now().isoformat() print('collecting followers for', screen_name) followers = twutil.collect.followers_for_screen_name(screen_name, limit) if len(followers) > 0: outf.write('%s %s %s\n' % (timestamp, screen_name, ' '.join(followers))) outf.flush() else: print('unknown user', screen_name) outf.close() if not do_loop: return else: if niters == 1: outfile = '%s.%d' % (outfile, niters) else: outfile = outfile[:outfile.rindex('.')] + '.%d' % niters niters += 1
Fetch up to limit followers for each Twitter account in account_file. Write results to outfile file in format: screen_name user_id follower_id_1 follower_id_2 ...
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/collect.py#L56-L82
[ "def iter_lines(filename):\n \"\"\" Iterate over screen names in a file, one per line.\"\"\"\n with open(filename, 'rt') as idfile:\n for line in idfile:\n screen_name = line.strip()\n if len(screen_name) > 0:\n yield screen_name.split()[0]\n" ]
#!/usr/bin/env python # -*- coding: utf-8 -*- """Collect Twitter data for brands. usage: brandelion collect --tweets --input <file> --output <file> --max=<N> brandelion collect --followers --input <file> --output <file> --max=<N> [--loop] brandelion collect --exemplars --query <string> --output <file> Options -h, --help -i, --input <file> File containing list of Twitter accounts, one per line. -l, --loop If true, keep looping to collect more data continuously. -o, --output <file> File to store results -t, --tweets Fetch tweets. -f, --followers Fetch followers -e, --exemplars Fetch exemplars from Twitter lists -q, --query <string> A single string used to search for Twitter lists. -m, --max=<N> Maximum number of followers or tweets to collect per account [default: 50000]. """ from collections import Counter import datetime from docopt import docopt import gzip import io import json import re import requests import sys import time import traceback import requests import twutil import json import time ##import config from init.py: from .. import config from googleapiclient.discovery import build from googleapiclient.errors import HttpError as GoogleHttpError def iter_lines(filename): """ Iterate over screen names in a file, one per line.""" with open(filename, 'rt') as idfile: for line in idfile: screen_name = line.strip() if len(screen_name) > 0: yield screen_name.split()[0] def fetch_tweets(account_file, outfile, limit): """ Fetch up to limit tweets for each account in account_file and write to outfile. """ print('fetching tweets for accounts in', account_file) outf = io.open(outfile, 'wt') for screen_name in iter_lines(account_file): print('\nFetching tweets for %s' % screen_name) for tweet in twutil.collect.tweets_for_user(screen_name, limit): tweet['user']['screen_name'] = screen_name outf.write('%s\n' % json.dumps(tweet, ensure_ascii=False)) outf.flush() #DEPRECATED # def fetch_lists(keyword, max_results=20): # """ # Fetch the urls of up to max_results Twitter lists that match the provided keyword. # >>> len(fetch_lists('politics', max_results=4)) # 4 # """ # res_per_page = 8 # start = 0 # results = [] # while len(results) < max_results: # url = 'http://ajax.googleapis.com/ajax/services/search/web?v=1.0&q=site:twitter.com+inurl:lists+%s&rsz=%d&start=%d' % (keyword, # res_per_page, # start) # js = json.loads(requests.get(url).text) # if not js['responseData']: # print('something went wrong in google search:\n', js) # return results[:max_results] # else: # for r in js['responseData']['results']: # results.append(r['url']) # start += res_per_page # time.sleep(.4) # return results[:max_results] #NEW FETCH lists def google_search(search_term, api_key, cse_id, **kwargs): final_urls = [] try: service = build("customsearch", "v1", developerKey=api_key) res = service.cse().list(q=search_term, cx=cse_id, **kwargs).execute() for item in res['items']: final_urls.append(item['formattedUrl']) return final_urls except GoogleHttpError: print("something wrong with Google HTTP Errpr") return final_urls def fetch_lists(keyword,max_results=20): """ Fetch the urls of up to max_results Twitter lists that match the provided keyword. >>> len(fetch_lists('politics', max_results=4)) 4 """ #CONFIG FILE READ api_key=config.get('GOOGLE_CSE_KEYS','API_KEY') cse_id=config.get('GOOGLE_CSE_KEYS','CSE_ID') results = [] start_c = 1 search_term = "inurl:lists + "+keyword while len(results)<max_results: temp_res = google_search(search_term,api_key,cse_id,num=10,start=start_c) if len(temp_res) == 0: print("Google API Error, returning retrieved results") return results results.extend(temp_res) start_c += 10 return results[:max_results] def fetch_list_members(list_url): """ Get all members of the list specified by the given url. E.g., https://twitter.com/lore77/lists/libri-cultura-education """ match = re.match(r'.+twitter\.com\/(.+)\/lists\/(.+)', list_url) if not match: print('cannot parse list url %s' % list_url) return [] screen_name, slug = match.groups() print('collecting list %s/%s' % (screen_name, slug)) return twutil.collect.list_members(slug, screen_name) def fetch_exemplars(keyword, outfile, n=50): """ Fetch top lists matching this keyword, then return Twitter screen names along with the number of different lists on which each appers.. """ list_urls = fetch_lists(keyword, n) print('found %d lists for %s' % (len(list_urls), keyword)) counts = Counter() for list_url in list_urls: counts.update(fetch_list_members(list_url)) # Write to file. outf = io.open(outfile, 'wt') for handle in sorted(counts): outf.write('%s\t%d\n' % (handle, counts[handle])) outf.close() print('saved exemplars to', outfile) def main(): args = docopt(__doc__) if args['--followers']: fetch_followers(args['--input'], args['--output'], int(args['--max']), args['--loop']) elif args['--tweets']: fetch_tweets(args['--input'], args['--output'], int(args['--max'])) else: fetch_exemplars(args['--query'], args['--output']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/collect.py
fetch_tweets
python
def fetch_tweets(account_file, outfile, limit): print('fetching tweets for accounts in', account_file) outf = io.open(outfile, 'wt') for screen_name in iter_lines(account_file): print('\nFetching tweets for %s' % screen_name) for tweet in twutil.collect.tweets_for_user(screen_name, limit): tweet['user']['screen_name'] = screen_name outf.write('%s\n' % json.dumps(tweet, ensure_ascii=False)) outf.flush()
Fetch up to limit tweets for each account in account_file and write to outfile.
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/collect.py#L85-L95
[ "def iter_lines(filename):\n \"\"\" Iterate over screen names in a file, one per line.\"\"\"\n with open(filename, 'rt') as idfile:\n for line in idfile:\n screen_name = line.strip()\n if len(screen_name) > 0:\n yield screen_name.split()[0]\n" ]
#!/usr/bin/env python # -*- coding: utf-8 -*- """Collect Twitter data for brands. usage: brandelion collect --tweets --input <file> --output <file> --max=<N> brandelion collect --followers --input <file> --output <file> --max=<N> [--loop] brandelion collect --exemplars --query <string> --output <file> Options -h, --help -i, --input <file> File containing list of Twitter accounts, one per line. -l, --loop If true, keep looping to collect more data continuously. -o, --output <file> File to store results -t, --tweets Fetch tweets. -f, --followers Fetch followers -e, --exemplars Fetch exemplars from Twitter lists -q, --query <string> A single string used to search for Twitter lists. -m, --max=<N> Maximum number of followers or tweets to collect per account [default: 50000]. """ from collections import Counter import datetime from docopt import docopt import gzip import io import json import re import requests import sys import time import traceback import requests import twutil import json import time ##import config from init.py: from .. import config from googleapiclient.discovery import build from googleapiclient.errors import HttpError as GoogleHttpError def iter_lines(filename): """ Iterate over screen names in a file, one per line.""" with open(filename, 'rt') as idfile: for line in idfile: screen_name = line.strip() if len(screen_name) > 0: yield screen_name.split()[0] def fetch_followers(account_file, outfile, limit, do_loop): """ Fetch up to limit followers for each Twitter account in account_file. Write results to outfile file in format: screen_name user_id follower_id_1 follower_id_2 ...""" print('Fetching followers for accounts in %s' % account_file) niters = 1 while True: outf = gzip.open(outfile, 'wt') for screen_name in iter_lines(account_file): timestamp = datetime.datetime.now().isoformat() print('collecting followers for', screen_name) followers = twutil.collect.followers_for_screen_name(screen_name, limit) if len(followers) > 0: outf.write('%s %s %s\n' % (timestamp, screen_name, ' '.join(followers))) outf.flush() else: print('unknown user', screen_name) outf.close() if not do_loop: return else: if niters == 1: outfile = '%s.%d' % (outfile, niters) else: outfile = outfile[:outfile.rindex('.')] + '.%d' % niters niters += 1 #DEPRECATED # def fetch_lists(keyword, max_results=20): # """ # Fetch the urls of up to max_results Twitter lists that match the provided keyword. # >>> len(fetch_lists('politics', max_results=4)) # 4 # """ # res_per_page = 8 # start = 0 # results = [] # while len(results) < max_results: # url = 'http://ajax.googleapis.com/ajax/services/search/web?v=1.0&q=site:twitter.com+inurl:lists+%s&rsz=%d&start=%d' % (keyword, # res_per_page, # start) # js = json.loads(requests.get(url).text) # if not js['responseData']: # print('something went wrong in google search:\n', js) # return results[:max_results] # else: # for r in js['responseData']['results']: # results.append(r['url']) # start += res_per_page # time.sleep(.4) # return results[:max_results] #NEW FETCH lists def google_search(search_term, api_key, cse_id, **kwargs): final_urls = [] try: service = build("customsearch", "v1", developerKey=api_key) res = service.cse().list(q=search_term, cx=cse_id, **kwargs).execute() for item in res['items']: final_urls.append(item['formattedUrl']) return final_urls except GoogleHttpError: print("something wrong with Google HTTP Errpr") return final_urls def fetch_lists(keyword,max_results=20): """ Fetch the urls of up to max_results Twitter lists that match the provided keyword. >>> len(fetch_lists('politics', max_results=4)) 4 """ #CONFIG FILE READ api_key=config.get('GOOGLE_CSE_KEYS','API_KEY') cse_id=config.get('GOOGLE_CSE_KEYS','CSE_ID') results = [] start_c = 1 search_term = "inurl:lists + "+keyword while len(results)<max_results: temp_res = google_search(search_term,api_key,cse_id,num=10,start=start_c) if len(temp_res) == 0: print("Google API Error, returning retrieved results") return results results.extend(temp_res) start_c += 10 return results[:max_results] def fetch_list_members(list_url): """ Get all members of the list specified by the given url. E.g., https://twitter.com/lore77/lists/libri-cultura-education """ match = re.match(r'.+twitter\.com\/(.+)\/lists\/(.+)', list_url) if not match: print('cannot parse list url %s' % list_url) return [] screen_name, slug = match.groups() print('collecting list %s/%s' % (screen_name, slug)) return twutil.collect.list_members(slug, screen_name) def fetch_exemplars(keyword, outfile, n=50): """ Fetch top lists matching this keyword, then return Twitter screen names along with the number of different lists on which each appers.. """ list_urls = fetch_lists(keyword, n) print('found %d lists for %s' % (len(list_urls), keyword)) counts = Counter() for list_url in list_urls: counts.update(fetch_list_members(list_url)) # Write to file. outf = io.open(outfile, 'wt') for handle in sorted(counts): outf.write('%s\t%d\n' % (handle, counts[handle])) outf.close() print('saved exemplars to', outfile) def main(): args = docopt(__doc__) if args['--followers']: fetch_followers(args['--input'], args['--output'], int(args['--max']), args['--loop']) elif args['--tweets']: fetch_tweets(args['--input'], args['--output'], int(args['--max'])) else: fetch_exemplars(args['--query'], args['--output']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/collect.py
fetch_lists
python
def fetch_lists(keyword,max_results=20): #CONFIG FILE READ api_key=config.get('GOOGLE_CSE_KEYS','API_KEY') cse_id=config.get('GOOGLE_CSE_KEYS','CSE_ID') results = [] start_c = 1 search_term = "inurl:lists + "+keyword while len(results)<max_results: temp_res = google_search(search_term,api_key,cse_id,num=10,start=start_c) if len(temp_res) == 0: print("Google API Error, returning retrieved results") return results results.extend(temp_res) start_c += 10 return results[:max_results]
Fetch the urls of up to max_results Twitter lists that match the provided keyword. >>> len(fetch_lists('politics', max_results=4)) 4
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/collect.py#L138-L158
[ "def google_search(search_term, api_key, cse_id, **kwargs):\n final_urls = []\n try:\n service = build(\"customsearch\", \"v1\", developerKey=api_key)\n res = service.cse().list(q=search_term, cx=cse_id, **kwargs).execute()\n for item in res['items']:\n final_urls.append(item['...
#!/usr/bin/env python # -*- coding: utf-8 -*- """Collect Twitter data for brands. usage: brandelion collect --tweets --input <file> --output <file> --max=<N> brandelion collect --followers --input <file> --output <file> --max=<N> [--loop] brandelion collect --exemplars --query <string> --output <file> Options -h, --help -i, --input <file> File containing list of Twitter accounts, one per line. -l, --loop If true, keep looping to collect more data continuously. -o, --output <file> File to store results -t, --tweets Fetch tweets. -f, --followers Fetch followers -e, --exemplars Fetch exemplars from Twitter lists -q, --query <string> A single string used to search for Twitter lists. -m, --max=<N> Maximum number of followers or tweets to collect per account [default: 50000]. """ from collections import Counter import datetime from docopt import docopt import gzip import io import json import re import requests import sys import time import traceback import requests import twutil import json import time ##import config from init.py: from .. import config from googleapiclient.discovery import build from googleapiclient.errors import HttpError as GoogleHttpError def iter_lines(filename): """ Iterate over screen names in a file, one per line.""" with open(filename, 'rt') as idfile: for line in idfile: screen_name = line.strip() if len(screen_name) > 0: yield screen_name.split()[0] def fetch_followers(account_file, outfile, limit, do_loop): """ Fetch up to limit followers for each Twitter account in account_file. Write results to outfile file in format: screen_name user_id follower_id_1 follower_id_2 ...""" print('Fetching followers for accounts in %s' % account_file) niters = 1 while True: outf = gzip.open(outfile, 'wt') for screen_name in iter_lines(account_file): timestamp = datetime.datetime.now().isoformat() print('collecting followers for', screen_name) followers = twutil.collect.followers_for_screen_name(screen_name, limit) if len(followers) > 0: outf.write('%s %s %s\n' % (timestamp, screen_name, ' '.join(followers))) outf.flush() else: print('unknown user', screen_name) outf.close() if not do_loop: return else: if niters == 1: outfile = '%s.%d' % (outfile, niters) else: outfile = outfile[:outfile.rindex('.')] + '.%d' % niters niters += 1 def fetch_tweets(account_file, outfile, limit): """ Fetch up to limit tweets for each account in account_file and write to outfile. """ print('fetching tweets for accounts in', account_file) outf = io.open(outfile, 'wt') for screen_name in iter_lines(account_file): print('\nFetching tweets for %s' % screen_name) for tweet in twutil.collect.tweets_for_user(screen_name, limit): tweet['user']['screen_name'] = screen_name outf.write('%s\n' % json.dumps(tweet, ensure_ascii=False)) outf.flush() #DEPRECATED # def fetch_lists(keyword, max_results=20): # """ # Fetch the urls of up to max_results Twitter lists that match the provided keyword. # >>> len(fetch_lists('politics', max_results=4)) # 4 # """ # res_per_page = 8 # start = 0 # results = [] # while len(results) < max_results: # url = 'http://ajax.googleapis.com/ajax/services/search/web?v=1.0&q=site:twitter.com+inurl:lists+%s&rsz=%d&start=%d' % (keyword, # res_per_page, # start) # js = json.loads(requests.get(url).text) # if not js['responseData']: # print('something went wrong in google search:\n', js) # return results[:max_results] # else: # for r in js['responseData']['results']: # results.append(r['url']) # start += res_per_page # time.sleep(.4) # return results[:max_results] #NEW FETCH lists def google_search(search_term, api_key, cse_id, **kwargs): final_urls = [] try: service = build("customsearch", "v1", developerKey=api_key) res = service.cse().list(q=search_term, cx=cse_id, **kwargs).execute() for item in res['items']: final_urls.append(item['formattedUrl']) return final_urls except GoogleHttpError: print("something wrong with Google HTTP Errpr") return final_urls def fetch_list_members(list_url): """ Get all members of the list specified by the given url. E.g., https://twitter.com/lore77/lists/libri-cultura-education """ match = re.match(r'.+twitter\.com\/(.+)\/lists\/(.+)', list_url) if not match: print('cannot parse list url %s' % list_url) return [] screen_name, slug = match.groups() print('collecting list %s/%s' % (screen_name, slug)) return twutil.collect.list_members(slug, screen_name) def fetch_exemplars(keyword, outfile, n=50): """ Fetch top lists matching this keyword, then return Twitter screen names along with the number of different lists on which each appers.. """ list_urls = fetch_lists(keyword, n) print('found %d lists for %s' % (len(list_urls), keyword)) counts = Counter() for list_url in list_urls: counts.update(fetch_list_members(list_url)) # Write to file. outf = io.open(outfile, 'wt') for handle in sorted(counts): outf.write('%s\t%d\n' % (handle, counts[handle])) outf.close() print('saved exemplars to', outfile) def main(): args = docopt(__doc__) if args['--followers']: fetch_followers(args['--input'], args['--output'], int(args['--max']), args['--loop']) elif args['--tweets']: fetch_tweets(args['--input'], args['--output'], int(args['--max'])) else: fetch_exemplars(args['--query'], args['--output']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/collect.py
fetch_list_members
python
def fetch_list_members(list_url): match = re.match(r'.+twitter\.com\/(.+)\/lists\/(.+)', list_url) if not match: print('cannot parse list url %s' % list_url) return [] screen_name, slug = match.groups() print('collecting list %s/%s' % (screen_name, slug)) return twutil.collect.list_members(slug, screen_name)
Get all members of the list specified by the given url. E.g., https://twitter.com/lore77/lists/libri-cultura-education
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/collect.py#L160-L168
null
#!/usr/bin/env python # -*- coding: utf-8 -*- """Collect Twitter data for brands. usage: brandelion collect --tweets --input <file> --output <file> --max=<N> brandelion collect --followers --input <file> --output <file> --max=<N> [--loop] brandelion collect --exemplars --query <string> --output <file> Options -h, --help -i, --input <file> File containing list of Twitter accounts, one per line. -l, --loop If true, keep looping to collect more data continuously. -o, --output <file> File to store results -t, --tweets Fetch tweets. -f, --followers Fetch followers -e, --exemplars Fetch exemplars from Twitter lists -q, --query <string> A single string used to search for Twitter lists. -m, --max=<N> Maximum number of followers or tweets to collect per account [default: 50000]. """ from collections import Counter import datetime from docopt import docopt import gzip import io import json import re import requests import sys import time import traceback import requests import twutil import json import time ##import config from init.py: from .. import config from googleapiclient.discovery import build from googleapiclient.errors import HttpError as GoogleHttpError def iter_lines(filename): """ Iterate over screen names in a file, one per line.""" with open(filename, 'rt') as idfile: for line in idfile: screen_name = line.strip() if len(screen_name) > 0: yield screen_name.split()[0] def fetch_followers(account_file, outfile, limit, do_loop): """ Fetch up to limit followers for each Twitter account in account_file. Write results to outfile file in format: screen_name user_id follower_id_1 follower_id_2 ...""" print('Fetching followers for accounts in %s' % account_file) niters = 1 while True: outf = gzip.open(outfile, 'wt') for screen_name in iter_lines(account_file): timestamp = datetime.datetime.now().isoformat() print('collecting followers for', screen_name) followers = twutil.collect.followers_for_screen_name(screen_name, limit) if len(followers) > 0: outf.write('%s %s %s\n' % (timestamp, screen_name, ' '.join(followers))) outf.flush() else: print('unknown user', screen_name) outf.close() if not do_loop: return else: if niters == 1: outfile = '%s.%d' % (outfile, niters) else: outfile = outfile[:outfile.rindex('.')] + '.%d' % niters niters += 1 def fetch_tweets(account_file, outfile, limit): """ Fetch up to limit tweets for each account in account_file and write to outfile. """ print('fetching tweets for accounts in', account_file) outf = io.open(outfile, 'wt') for screen_name in iter_lines(account_file): print('\nFetching tweets for %s' % screen_name) for tweet in twutil.collect.tweets_for_user(screen_name, limit): tweet['user']['screen_name'] = screen_name outf.write('%s\n' % json.dumps(tweet, ensure_ascii=False)) outf.flush() #DEPRECATED # def fetch_lists(keyword, max_results=20): # """ # Fetch the urls of up to max_results Twitter lists that match the provided keyword. # >>> len(fetch_lists('politics', max_results=4)) # 4 # """ # res_per_page = 8 # start = 0 # results = [] # while len(results) < max_results: # url = 'http://ajax.googleapis.com/ajax/services/search/web?v=1.0&q=site:twitter.com+inurl:lists+%s&rsz=%d&start=%d' % (keyword, # res_per_page, # start) # js = json.loads(requests.get(url).text) # if not js['responseData']: # print('something went wrong in google search:\n', js) # return results[:max_results] # else: # for r in js['responseData']['results']: # results.append(r['url']) # start += res_per_page # time.sleep(.4) # return results[:max_results] #NEW FETCH lists def google_search(search_term, api_key, cse_id, **kwargs): final_urls = [] try: service = build("customsearch", "v1", developerKey=api_key) res = service.cse().list(q=search_term, cx=cse_id, **kwargs).execute() for item in res['items']: final_urls.append(item['formattedUrl']) return final_urls except GoogleHttpError: print("something wrong with Google HTTP Errpr") return final_urls def fetch_lists(keyword,max_results=20): """ Fetch the urls of up to max_results Twitter lists that match the provided keyword. >>> len(fetch_lists('politics', max_results=4)) 4 """ #CONFIG FILE READ api_key=config.get('GOOGLE_CSE_KEYS','API_KEY') cse_id=config.get('GOOGLE_CSE_KEYS','CSE_ID') results = [] start_c = 1 search_term = "inurl:lists + "+keyword while len(results)<max_results: temp_res = google_search(search_term,api_key,cse_id,num=10,start=start_c) if len(temp_res) == 0: print("Google API Error, returning retrieved results") return results results.extend(temp_res) start_c += 10 return results[:max_results] def fetch_exemplars(keyword, outfile, n=50): """ Fetch top lists matching this keyword, then return Twitter screen names along with the number of different lists on which each appers.. """ list_urls = fetch_lists(keyword, n) print('found %d lists for %s' % (len(list_urls), keyword)) counts = Counter() for list_url in list_urls: counts.update(fetch_list_members(list_url)) # Write to file. outf = io.open(outfile, 'wt') for handle in sorted(counts): outf.write('%s\t%d\n' % (handle, counts[handle])) outf.close() print('saved exemplars to', outfile) def main(): args = docopt(__doc__) if args['--followers']: fetch_followers(args['--input'], args['--output'], int(args['--max']), args['--loop']) elif args['--tweets']: fetch_tweets(args['--input'], args['--output'], int(args['--max'])) else: fetch_exemplars(args['--query'], args['--output']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/collect.py
fetch_exemplars
python
def fetch_exemplars(keyword, outfile, n=50): list_urls = fetch_lists(keyword, n) print('found %d lists for %s' % (len(list_urls), keyword)) counts = Counter() for list_url in list_urls: counts.update(fetch_list_members(list_url)) # Write to file. outf = io.open(outfile, 'wt') for handle in sorted(counts): outf.write('%s\t%d\n' % (handle, counts[handle])) outf.close() print('saved exemplars to', outfile)
Fetch top lists matching this keyword, then return Twitter screen names along with the number of different lists on which each appers..
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/collect.py#L171-L184
[ "def fetch_lists(keyword,max_results=20):\n \"\"\"\n Fetch the urls of up to max_results Twitter lists that match the provided keyword.\n >>> len(fetch_lists('politics', max_results=4))\n 4\n \"\"\"\n #CONFIG FILE READ\n api_key=config.get('GOOGLE_CSE_KEYS','API_KEY')\n cse_id=config.get('GO...
#!/usr/bin/env python # -*- coding: utf-8 -*- """Collect Twitter data for brands. usage: brandelion collect --tweets --input <file> --output <file> --max=<N> brandelion collect --followers --input <file> --output <file> --max=<N> [--loop] brandelion collect --exemplars --query <string> --output <file> Options -h, --help -i, --input <file> File containing list of Twitter accounts, one per line. -l, --loop If true, keep looping to collect more data continuously. -o, --output <file> File to store results -t, --tweets Fetch tweets. -f, --followers Fetch followers -e, --exemplars Fetch exemplars from Twitter lists -q, --query <string> A single string used to search for Twitter lists. -m, --max=<N> Maximum number of followers or tweets to collect per account [default: 50000]. """ from collections import Counter import datetime from docopt import docopt import gzip import io import json import re import requests import sys import time import traceback import requests import twutil import json import time ##import config from init.py: from .. import config from googleapiclient.discovery import build from googleapiclient.errors import HttpError as GoogleHttpError def iter_lines(filename): """ Iterate over screen names in a file, one per line.""" with open(filename, 'rt') as idfile: for line in idfile: screen_name = line.strip() if len(screen_name) > 0: yield screen_name.split()[0] def fetch_followers(account_file, outfile, limit, do_loop): """ Fetch up to limit followers for each Twitter account in account_file. Write results to outfile file in format: screen_name user_id follower_id_1 follower_id_2 ...""" print('Fetching followers for accounts in %s' % account_file) niters = 1 while True: outf = gzip.open(outfile, 'wt') for screen_name in iter_lines(account_file): timestamp = datetime.datetime.now().isoformat() print('collecting followers for', screen_name) followers = twutil.collect.followers_for_screen_name(screen_name, limit) if len(followers) > 0: outf.write('%s %s %s\n' % (timestamp, screen_name, ' '.join(followers))) outf.flush() else: print('unknown user', screen_name) outf.close() if not do_loop: return else: if niters == 1: outfile = '%s.%d' % (outfile, niters) else: outfile = outfile[:outfile.rindex('.')] + '.%d' % niters niters += 1 def fetch_tweets(account_file, outfile, limit): """ Fetch up to limit tweets for each account in account_file and write to outfile. """ print('fetching tweets for accounts in', account_file) outf = io.open(outfile, 'wt') for screen_name in iter_lines(account_file): print('\nFetching tweets for %s' % screen_name) for tweet in twutil.collect.tweets_for_user(screen_name, limit): tweet['user']['screen_name'] = screen_name outf.write('%s\n' % json.dumps(tweet, ensure_ascii=False)) outf.flush() #DEPRECATED # def fetch_lists(keyword, max_results=20): # """ # Fetch the urls of up to max_results Twitter lists that match the provided keyword. # >>> len(fetch_lists('politics', max_results=4)) # 4 # """ # res_per_page = 8 # start = 0 # results = [] # while len(results) < max_results: # url = 'http://ajax.googleapis.com/ajax/services/search/web?v=1.0&q=site:twitter.com+inurl:lists+%s&rsz=%d&start=%d' % (keyword, # res_per_page, # start) # js = json.loads(requests.get(url).text) # if not js['responseData']: # print('something went wrong in google search:\n', js) # return results[:max_results] # else: # for r in js['responseData']['results']: # results.append(r['url']) # start += res_per_page # time.sleep(.4) # return results[:max_results] #NEW FETCH lists def google_search(search_term, api_key, cse_id, **kwargs): final_urls = [] try: service = build("customsearch", "v1", developerKey=api_key) res = service.cse().list(q=search_term, cx=cse_id, **kwargs).execute() for item in res['items']: final_urls.append(item['formattedUrl']) return final_urls except GoogleHttpError: print("something wrong with Google HTTP Errpr") return final_urls def fetch_lists(keyword,max_results=20): """ Fetch the urls of up to max_results Twitter lists that match the provided keyword. >>> len(fetch_lists('politics', max_results=4)) 4 """ #CONFIG FILE READ api_key=config.get('GOOGLE_CSE_KEYS','API_KEY') cse_id=config.get('GOOGLE_CSE_KEYS','CSE_ID') results = [] start_c = 1 search_term = "inurl:lists + "+keyword while len(results)<max_results: temp_res = google_search(search_term,api_key,cse_id,num=10,start=start_c) if len(temp_res) == 0: print("Google API Error, returning retrieved results") return results results.extend(temp_res) start_c += 10 return results[:max_results] def fetch_list_members(list_url): """ Get all members of the list specified by the given url. E.g., https://twitter.com/lore77/lists/libri-cultura-education """ match = re.match(r'.+twitter\.com\/(.+)\/lists\/(.+)', list_url) if not match: print('cannot parse list url %s' % list_url) return [] screen_name, slug = match.groups() print('collecting list %s/%s' % (screen_name, slug)) return twutil.collect.list_members(slug, screen_name) def main(): args = docopt(__doc__) if args['--followers']: fetch_followers(args['--input'], args['--output'], int(args['--max']), args['--loop']) elif args['--tweets']: fetch_tweets(args['--input'], args['--output'], int(args['--max'])) else: fetch_exemplars(args['--query'], args['--output']) if __name__ == '__main__': main()
tapilab/brandelion
brandelion/cli/diagnose.py
correlation_by_exemplar
python
def correlation_by_exemplar(brands, exemplars, validation_scores, analyze_fn_str, outf): analyze_fn = getattr(analyze, analyze_fn_str) keys = sorted(k for k in validation_scores.keys() if k in set(x[0] for x in brands)) truth = [validation_scores[k] for k in keys] result = {} outf.write('exemplar\tcorr\tn_followers\n') outf.flush() for exemplar in exemplars: single_exemplar = {exemplar: exemplars[exemplar]} social_scores = analyze_fn(brands, single_exemplar) predicted = [social_scores[k] for k in keys] outf.write('%s\t%g\t%d\n' % (exemplar, scistat.pearsonr(predicted, truth)[0], len(exemplars[exemplar]))) outf.flush() result[exemplar] = scistat.pearsonr(predicted, truth)[0] outf.close() return result
Report the overall correlation with the validation scores using each exemplar in isolation.
train
https://github.com/tapilab/brandelion/blob/40a5a5333cf704182c8666d1fbbbdadc7ff88546/brandelion/cli/diagnose.py#L50-L66
null
#!/usr/bin/env python # -*- coding: utf-8 -*- """Run diagnostics on a dataset. usage: brandelion diagnose --network --brand-followers <file> --exemplar-followers <file> --validation <file> --output <file> [--network-method <string>] Options -h, --help --brand-followers <file> File containing follower data for brand accounts. --exemplar-followers <file> File containing follower data for exemplar accounts. --network-method <string> Method to do text analysis [default: jaccard] -n, --network Analyze followers. -o, --output <file> File to store results -t, --text Analyze text. -v, --validation <file> File containing third-party scores for each brand by Twitter name, (e.g., surveys), for comparison. """ from docopt import docopt import os import scipy.stats as scistat import random from . import analyze, report random.seed(123) def read_scores(fname): scores = {} for line in open(fname): parts = line.strip().lower().split() if len(parts) > 1: scores[parts[0]] = float(parts[1]) return scores def validate(scores, validation): keys = sorted(validation.keys()) predicted = [scores[k] for k in keys] truth = [validation[k] for k in keys] print('Pearson:', scistat.pearsonr(predicted, truth)) def diagnose_text(brand_tweets_file, exemplar_tweets_file, sample_tweets_file, validation_file): pass def mkdirs(filename): report.mkdirs(os.path.dirname(filename)) def diagnose_followers(brand_follower_file, exemplar_follower_file, validation_file, analyze_fn, output_file): mkdirs(output_file) outf = open(output_file, 'wt') brands = analyze.read_follower_file(brand_follower_file).items() exemplars = analyze.read_follower_file(exemplar_follower_file, blacklist=analyze.get_twitter_handles(brand_follower_file)) print('read follower data for %d exemplars' % (len(exemplars))) scores = report.read_scores(validation_file) return correlation_by_exemplar(brands, exemplars, scores, analyze_fn, outf) def main(): args = docopt(__doc__) print(args) if '--network' in args: diagnose_followers(args['--brand-followers'], args['--exemplar-followers'], args['--validation'], args['--network-method'], args['--output']) if '--text' in args: diagnose_text(args['--brand-tweets'], args['--exemplar-tweets'], args['--sample-tweets'], args['--validation'], args['--text-method']) if __name__ == '__main__': main()
pyQode/pyqode.cobol
pyqode/cobol/widgets/code_edit.py
CobolCodeEdit._do_home_key
python
def _do_home_key(self, event=None, select=False): # get nb char to first significative char min_column = self.indenter_mode.min_column text = api.TextHelper(self).current_line_text()[min_column:] indent = len(text) - len(text.lstrip()) delta = (self.textCursor().positionInBlock() - indent - min_column) cursor = self.textCursor() move = QtGui.QTextCursor.MoveAnchor if select: move = QtGui.QTextCursor.KeepAnchor if delta > 0: cursor.movePosition(QtGui.QTextCursor.Left, move, delta) else: cursor.movePosition(QtGui.QTextCursor.StartOfBlock, move) cursor.movePosition(QtGui.QTextCursor.Right, cursor.MoveAnchor, min_column) self.setTextCursor(cursor) if event: event.accept()
Performs home key action
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/widgets/code_edit.py#L206-L224
null
class CobolCodeEdit(api.CodeEdit): """ CodeEdit specialized for cobol source code editing. """ class CobolFileManager(FileManager): def _get_icon(self): return QtGui.QIcon(icons.ICON_MIMETYPE) mimetypes = ['text/x-cobol'] extensions = [".COB", ".CBL", ".PCO", ".CPY", ".SCB", ".SQB"] @classmethod def all_extensions(cls): return cls.extensions + [ext.lower() for ext in cls.extensions] @property def free_format(self): return self._free_format @free_format.setter def free_format(self, free_fmt): if free_fmt != self._free_format: self._free_format = free_fmt self.margins.enabled = not free_fmt self.syntax_highlighter.rehighlight() self._update_backend_format() @property def lower_case_keywords(self): return self._lower_case_keywords @lower_case_keywords.setter def lower_case_keywords(self, value): self._lower_case_keywords = value self._update_backend_proposed_kw_case() @property def comment_indicator(self): return self._comment_indicator @comment_indicator.setter def comment_indicator(self, value): self._comment_indicator = value def __init__(self, parent=None, color_scheme='qt', free_format=False): super().__init__(parent) self.file = self.CobolFileManager(self) self._lower_case_keywords = False self._free_format = None self._comment_indicator = '*> ' self.word_separators.remove('-') self._start_server() self._setup_panels() self._setup_modes(color_scheme) self.encoding_panel = self.panels.append( panels.EncodingPanel(), api.Panel.Position.TOP ) self.read_only_panel = self.panels.append( panels.ReadOnlyPanel(), api.Panel.Position.TOP ) self.free_format = free_format def _start_server(self): if hasattr(sys, "frozen"): cwd = os.path.dirname(sys.executable) base = 'cobol-backend' srv = base + '.exe' if sys.platform == 'win32' else base srv = os.path.join(cwd, srv) self.backend.start(srv) else: self.backend.start(server.__file__) def close(self, *args, **kwargs): self.cursor_history = None self.extended_selection_mode = None self.case_converter = None self.auto_complete = None self.outline_mode = None self.add_separator() self.goto_def_mode = None self.code_completion_mode = None self.file_watcher = None self.auto_indent_mode = None self.caret_line_mode = None self.zoom_mode = None self.indenter_mode = None self.auto_indent_mode = None self.left_margin = None self.right_margin = None self.comments_mode = None self.offset_calculator = None self.occurences_highlighter_mode = None self.backspace_mode = None self.search_panel = None self.folding_panel = None self.line_nbr_panel = None self.checker_panel = None self.global_checker_panel = None self.encoding_panel = None self.read_only_panel = None self.margins = None super().close(*args, **kwargs) def _setup_modes(self, color_scheme): self.cursor_history = self.modes.append(modes.CursorHistoryMode()) self.extended_selection_mode = self.modes.append( modes.ExtendedSelectionMode() ) self.case_converter = self.modes.append( modes.CaseConverterMode() ) self.auto_complete = self.modes.append( modes.AutoCompleteMode()) self.outline_mode = self.modes.append( modes.OutlineMode(get_outline)) self.add_separator() self.goto_def_mode = self.modes.append( cobmodes.GoToDefinitionMode() ) self.code_completion_mode = self.modes.append( modes.CodeCompletionMode() ) self.code_completion_mode.trigger_symbols[:] = [] self.file_watcher = self.modes.append( modes.FileWatcherMode() ) self.auto_indent_mode = self.modes.append( cobmodes.CobolAutoIndentMode() ) self.caret_line_mode = self.modes.append( modes.CaretLineHighlighterMode() ) self.zoom_mode = self.modes.append( modes.ZoomMode() ) self.indenter_mode = self.modes.append( cobmodes.IndenterMode() ) self.auto_indent_mode = self.modes.append( modes.AutoIndentMode() ) self.modes.append(cobmodes.CobolSyntaxHighlighter( self.document(), color_scheme=api.ColorScheme(color_scheme))) self.syntax_highlighter.fold_detector = CobolFoldDetector() self.margins = self.modes.append(cobmodes.MarginsMode()) self.comments_mode = self.modes.append(cobmodes.CommentsMode()) self.offset_calculator = self.modes.append( cobmodes.OffsetCalculatorMode()) self.occurences_highlighter_mode = self.modes.append( modes.OccurrencesHighlighterMode() ) self.occurences_highlighter_mode.case_sensitive = False self.backspace_mode = self.modes.append( cobmodes.SmartBackSpaceMode() ) def _setup_panels(self): self.search_panel = self.panels.append( panels.SearchAndReplacePanel(), api.Panel.Position.BOTTOM ) self.folding_panel = self.panels.append( panels.FoldingPanel(), api.Panel.Position.LEFT ) self.line_nbr_panel = self.panels.append( panels.LineNumberPanel(), api.Panel.Position.LEFT ) self.checker_panel = self.panels.append( panels.CheckerPanel(), api.Panel.Position.LEFT ) self.global_checker_panel = self.panels.append( panels.GlobalCheckerPanel(), api.Panel.Position.RIGHT) def _update_backend_format(self): from pyqode.cobol.backend.workers import set_free_format try: self.backend.send_request(set_free_format, self.free_format) except NotRunning: QtCore.QTimer.singleShot(100, self._update_backend_format) def _update_backend_proposed_kw_case(self): from pyqode.cobol.backend.workers import set_lower_case_keywords try: self.backend.send_request(set_lower_case_keywords, self.lower_case_keywords) except NotRunning: QtCore.QTimer.singleShot( 100, self._update_backend_proposed_kw_case)
pyQode/pyqode.cobol
pyqode/cobol/modes/goto.py
GoToDefinitionMode.on_state_changed
python
def on_state_changed(self, state): super(GoToDefinitionMode, self).on_state_changed(state) if state: self.editor.mouse_moved.connect(self._on_mouse_moved) self.editor.mouse_released.connect(self._on_mouse_released) self.editor.add_action(self.action_goto, sub_menu='COBOL') self.editor.mouse_double_clicked.connect( self._timer.cancel_requests) else: self.editor.mouse_moved.disconnect(self._on_mouse_moved) self.editor.mouse_released.disconnect(self._on_mouse_released) self.editor.remove_action(self.action_goto, sub_menu='Python') self.editor.mouse_double_clicked.disconnect( self._timer.cancel_requests)
Connects/disconnects slots to/from signals when the mode state changed.
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/modes/goto.py#L29-L46
null
class GoToDefinitionMode(Mode, QObject): """ Go to the definition of the symbol under the word cursor. """ #: Signal emitted when a word is clicked. The parameter is a #: QTextCursor with the clicked word set as the selected text. word_clicked = Signal(QTextCursor) def __init__(self): QObject.__init__(self) Mode.__init__(self) self._previous_cursor_start = -1 self._previous_cursor_end = -1 self._definition = None self._deco = None self._pending = False self.action_goto = QAction(_("Go to assignments"), self) self.action_goto.setShortcut('F7') self.action_goto.triggered.connect(self.request_goto) self.word_clicked.connect(self.request_goto) self._timer = DelayJobRunner(delay=200) def _select_word_under_mouse_cursor(self): """ Selects the word under the mouse cursor. """ cursor = TextHelper(self.editor).word_under_mouse_cursor() if (self._previous_cursor_start != cursor.selectionStart() and self._previous_cursor_end != cursor.selectionEnd()): self._remove_decoration() self._add_decoration(cursor) self._previous_cursor_start = cursor.selectionStart() self._previous_cursor_end = cursor.selectionEnd() def _on_mouse_moved(self, event): """ mouse moved callback """ if event.modifiers() & Qt.ControlModifier: self._select_word_under_mouse_cursor() else: self._remove_decoration() self.editor.set_mouse_cursor(Qt.IBeamCursor) self._previous_cursor_start = -1 self._previous_cursor_end = -1 def _on_mouse_released(self, event): """ mouse pressed callback """ if event.button() == 1 and self._deco: cursor = TextHelper(self.editor).word_under_mouse_cursor() if cursor and cursor.selectedText(): self._timer.request_job(self.word_clicked.emit, cursor) def find_definition(self, symbol, definition): if symbol.lower() == definition.name.lower().replace(" section", "").replace(" division", ""): return definition for ch in definition.children: d = self.find_definition(symbol, ch) if d is not None: return d return None def select_word(self, cursor): symbol = cursor.selectedText() analyser = self.editor.outline_mode for definition in analyser.definitions: node = self.find_definition(symbol, definition) if node is not None: break else: node = None self._definition = None if node and node.line != cursor.block().blockNumber(): self._definition = node if self._deco is None: if cursor.selectedText(): self._deco = TextDecoration(cursor) self._deco.set_foreground(Qt.blue) self._deco.set_as_underlined() self.editor.decorations.append(self._deco) return True return False def _add_decoration(self, cursor): """ Adds a decoration for the word under ``cursor``. """ if self.select_word(cursor): self.editor.set_mouse_cursor(Qt.PointingHandCursor) else: self.editor.set_mouse_cursor(Qt.IBeamCursor) def _remove_decoration(self): """ Removes the word under cursor's decoration """ if self._deco is not None: self.editor.decorations.remove(self._deco) self._deco = None def request_goto(self, tc=None): """ Request a go to assignment. :param tc: Text cursor which contains the text that we must look for its assignment. Can be None to go to the text that is under the text cursor. :type tc: QtGui.QTextCursor """ if not tc: tc = TextHelper(self.editor).word_under_cursor( select_whole_word=True) if not self._definition or isinstance(self.sender(), QAction): self.select_word(tc) if self._definition is not None: QTimer.singleShot(100, self._goto_def) def _goto_def(self): if self._definition: line = self._definition.line col = self._definition.column TextHelper(self.editor).goto_line(line, move=True, column=col)
pyQode/pyqode.cobol
pyqode/cobol/modes/goto.py
GoToDefinitionMode._on_mouse_moved
python
def _on_mouse_moved(self, event): if event.modifiers() & Qt.ControlModifier: self._select_word_under_mouse_cursor() else: self._remove_decoration() self.editor.set_mouse_cursor(Qt.IBeamCursor) self._previous_cursor_start = -1 self._previous_cursor_end = -1
mouse moved callback
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/modes/goto.py#L58-L66
null
class GoToDefinitionMode(Mode, QObject): """ Go to the definition of the symbol under the word cursor. """ #: Signal emitted when a word is clicked. The parameter is a #: QTextCursor with the clicked word set as the selected text. word_clicked = Signal(QTextCursor) def __init__(self): QObject.__init__(self) Mode.__init__(self) self._previous_cursor_start = -1 self._previous_cursor_end = -1 self._definition = None self._deco = None self._pending = False self.action_goto = QAction(_("Go to assignments"), self) self.action_goto.setShortcut('F7') self.action_goto.triggered.connect(self.request_goto) self.word_clicked.connect(self.request_goto) self._timer = DelayJobRunner(delay=200) def on_state_changed(self, state): """ Connects/disconnects slots to/from signals when the mode state changed. """ super(GoToDefinitionMode, self).on_state_changed(state) if state: self.editor.mouse_moved.connect(self._on_mouse_moved) self.editor.mouse_released.connect(self._on_mouse_released) self.editor.add_action(self.action_goto, sub_menu='COBOL') self.editor.mouse_double_clicked.connect( self._timer.cancel_requests) else: self.editor.mouse_moved.disconnect(self._on_mouse_moved) self.editor.mouse_released.disconnect(self._on_mouse_released) self.editor.remove_action(self.action_goto, sub_menu='Python') self.editor.mouse_double_clicked.disconnect( self._timer.cancel_requests) def _select_word_under_mouse_cursor(self): """ Selects the word under the mouse cursor. """ cursor = TextHelper(self.editor).word_under_mouse_cursor() if (self._previous_cursor_start != cursor.selectionStart() and self._previous_cursor_end != cursor.selectionEnd()): self._remove_decoration() self._add_decoration(cursor) self._previous_cursor_start = cursor.selectionStart() self._previous_cursor_end = cursor.selectionEnd() def _on_mouse_released(self, event): """ mouse pressed callback """ if event.button() == 1 and self._deco: cursor = TextHelper(self.editor).word_under_mouse_cursor() if cursor and cursor.selectedText(): self._timer.request_job(self.word_clicked.emit, cursor) def find_definition(self, symbol, definition): if symbol.lower() == definition.name.lower().replace(" section", "").replace(" division", ""): return definition for ch in definition.children: d = self.find_definition(symbol, ch) if d is not None: return d return None def select_word(self, cursor): symbol = cursor.selectedText() analyser = self.editor.outline_mode for definition in analyser.definitions: node = self.find_definition(symbol, definition) if node is not None: break else: node = None self._definition = None if node and node.line != cursor.block().blockNumber(): self._definition = node if self._deco is None: if cursor.selectedText(): self._deco = TextDecoration(cursor) self._deco.set_foreground(Qt.blue) self._deco.set_as_underlined() self.editor.decorations.append(self._deco) return True return False def _add_decoration(self, cursor): """ Adds a decoration for the word under ``cursor``. """ if self.select_word(cursor): self.editor.set_mouse_cursor(Qt.PointingHandCursor) else: self.editor.set_mouse_cursor(Qt.IBeamCursor) def _remove_decoration(self): """ Removes the word under cursor's decoration """ if self._deco is not None: self.editor.decorations.remove(self._deco) self._deco = None def request_goto(self, tc=None): """ Request a go to assignment. :param tc: Text cursor which contains the text that we must look for its assignment. Can be None to go to the text that is under the text cursor. :type tc: QtGui.QTextCursor """ if not tc: tc = TextHelper(self.editor).word_under_cursor( select_whole_word=True) if not self._definition or isinstance(self.sender(), QAction): self.select_word(tc) if self._definition is not None: QTimer.singleShot(100, self._goto_def) def _goto_def(self): if self._definition: line = self._definition.line col = self._definition.column TextHelper(self.editor).goto_line(line, move=True, column=col)
pyQode/pyqode.cobol
pyqode/cobol/modes/goto.py
GoToDefinitionMode._add_decoration
python
def _add_decoration(self, cursor): if self.select_word(cursor): self.editor.set_mouse_cursor(Qt.PointingHandCursor) else: self.editor.set_mouse_cursor(Qt.IBeamCursor)
Adds a decoration for the word under ``cursor``.
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/modes/goto.py#L105-L112
null
class GoToDefinitionMode(Mode, QObject): """ Go to the definition of the symbol under the word cursor. """ #: Signal emitted when a word is clicked. The parameter is a #: QTextCursor with the clicked word set as the selected text. word_clicked = Signal(QTextCursor) def __init__(self): QObject.__init__(self) Mode.__init__(self) self._previous_cursor_start = -1 self._previous_cursor_end = -1 self._definition = None self._deco = None self._pending = False self.action_goto = QAction(_("Go to assignments"), self) self.action_goto.setShortcut('F7') self.action_goto.triggered.connect(self.request_goto) self.word_clicked.connect(self.request_goto) self._timer = DelayJobRunner(delay=200) def on_state_changed(self, state): """ Connects/disconnects slots to/from signals when the mode state changed. """ super(GoToDefinitionMode, self).on_state_changed(state) if state: self.editor.mouse_moved.connect(self._on_mouse_moved) self.editor.mouse_released.connect(self._on_mouse_released) self.editor.add_action(self.action_goto, sub_menu='COBOL') self.editor.mouse_double_clicked.connect( self._timer.cancel_requests) else: self.editor.mouse_moved.disconnect(self._on_mouse_moved) self.editor.mouse_released.disconnect(self._on_mouse_released) self.editor.remove_action(self.action_goto, sub_menu='Python') self.editor.mouse_double_clicked.disconnect( self._timer.cancel_requests) def _select_word_under_mouse_cursor(self): """ Selects the word under the mouse cursor. """ cursor = TextHelper(self.editor).word_under_mouse_cursor() if (self._previous_cursor_start != cursor.selectionStart() and self._previous_cursor_end != cursor.selectionEnd()): self._remove_decoration() self._add_decoration(cursor) self._previous_cursor_start = cursor.selectionStart() self._previous_cursor_end = cursor.selectionEnd() def _on_mouse_moved(self, event): """ mouse moved callback """ if event.modifiers() & Qt.ControlModifier: self._select_word_under_mouse_cursor() else: self._remove_decoration() self.editor.set_mouse_cursor(Qt.IBeamCursor) self._previous_cursor_start = -1 self._previous_cursor_end = -1 def _on_mouse_released(self, event): """ mouse pressed callback """ if event.button() == 1 and self._deco: cursor = TextHelper(self.editor).word_under_mouse_cursor() if cursor and cursor.selectedText(): self._timer.request_job(self.word_clicked.emit, cursor) def find_definition(self, symbol, definition): if symbol.lower() == definition.name.lower().replace(" section", "").replace(" division", ""): return definition for ch in definition.children: d = self.find_definition(symbol, ch) if d is not None: return d return None def select_word(self, cursor): symbol = cursor.selectedText() analyser = self.editor.outline_mode for definition in analyser.definitions: node = self.find_definition(symbol, definition) if node is not None: break else: node = None self._definition = None if node and node.line != cursor.block().blockNumber(): self._definition = node if self._deco is None: if cursor.selectedText(): self._deco = TextDecoration(cursor) self._deco.set_foreground(Qt.blue) self._deco.set_as_underlined() self.editor.decorations.append(self._deco) return True return False def _remove_decoration(self): """ Removes the word under cursor's decoration """ if self._deco is not None: self.editor.decorations.remove(self._deco) self._deco = None def request_goto(self, tc=None): """ Request a go to assignment. :param tc: Text cursor which contains the text that we must look for its assignment. Can be None to go to the text that is under the text cursor. :type tc: QtGui.QTextCursor """ if not tc: tc = TextHelper(self.editor).word_under_cursor( select_whole_word=True) if not self._definition or isinstance(self.sender(), QAction): self.select_word(tc) if self._definition is not None: QTimer.singleShot(100, self._goto_def) def _goto_def(self): if self._definition: line = self._definition.line col = self._definition.column TextHelper(self.editor).goto_line(line, move=True, column=col)
pyQode/pyqode.cobol
pyqode/cobol/modes/goto.py
GoToDefinitionMode.request_goto
python
def request_goto(self, tc=None): if not tc: tc = TextHelper(self.editor).word_under_cursor( select_whole_word=True) if not self._definition or isinstance(self.sender(), QAction): self.select_word(tc) if self._definition is not None: QTimer.singleShot(100, self._goto_def)
Request a go to assignment. :param tc: Text cursor which contains the text that we must look for its assignment. Can be None to go to the text that is under the text cursor. :type tc: QtGui.QTextCursor
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/modes/goto.py#L122-L137
null
class GoToDefinitionMode(Mode, QObject): """ Go to the definition of the symbol under the word cursor. """ #: Signal emitted when a word is clicked. The parameter is a #: QTextCursor with the clicked word set as the selected text. word_clicked = Signal(QTextCursor) def __init__(self): QObject.__init__(self) Mode.__init__(self) self._previous_cursor_start = -1 self._previous_cursor_end = -1 self._definition = None self._deco = None self._pending = False self.action_goto = QAction(_("Go to assignments"), self) self.action_goto.setShortcut('F7') self.action_goto.triggered.connect(self.request_goto) self.word_clicked.connect(self.request_goto) self._timer = DelayJobRunner(delay=200) def on_state_changed(self, state): """ Connects/disconnects slots to/from signals when the mode state changed. """ super(GoToDefinitionMode, self).on_state_changed(state) if state: self.editor.mouse_moved.connect(self._on_mouse_moved) self.editor.mouse_released.connect(self._on_mouse_released) self.editor.add_action(self.action_goto, sub_menu='COBOL') self.editor.mouse_double_clicked.connect( self._timer.cancel_requests) else: self.editor.mouse_moved.disconnect(self._on_mouse_moved) self.editor.mouse_released.disconnect(self._on_mouse_released) self.editor.remove_action(self.action_goto, sub_menu='Python') self.editor.mouse_double_clicked.disconnect( self._timer.cancel_requests) def _select_word_under_mouse_cursor(self): """ Selects the word under the mouse cursor. """ cursor = TextHelper(self.editor).word_under_mouse_cursor() if (self._previous_cursor_start != cursor.selectionStart() and self._previous_cursor_end != cursor.selectionEnd()): self._remove_decoration() self._add_decoration(cursor) self._previous_cursor_start = cursor.selectionStart() self._previous_cursor_end = cursor.selectionEnd() def _on_mouse_moved(self, event): """ mouse moved callback """ if event.modifiers() & Qt.ControlModifier: self._select_word_under_mouse_cursor() else: self._remove_decoration() self.editor.set_mouse_cursor(Qt.IBeamCursor) self._previous_cursor_start = -1 self._previous_cursor_end = -1 def _on_mouse_released(self, event): """ mouse pressed callback """ if event.button() == 1 and self._deco: cursor = TextHelper(self.editor).word_under_mouse_cursor() if cursor and cursor.selectedText(): self._timer.request_job(self.word_clicked.emit, cursor) def find_definition(self, symbol, definition): if symbol.lower() == definition.name.lower().replace(" section", "").replace(" division", ""): return definition for ch in definition.children: d = self.find_definition(symbol, ch) if d is not None: return d return None def select_word(self, cursor): symbol = cursor.selectedText() analyser = self.editor.outline_mode for definition in analyser.definitions: node = self.find_definition(symbol, definition) if node is not None: break else: node = None self._definition = None if node and node.line != cursor.block().blockNumber(): self._definition = node if self._deco is None: if cursor.selectedText(): self._deco = TextDecoration(cursor) self._deco.set_foreground(Qt.blue) self._deco.set_as_underlined() self.editor.decorations.append(self._deco) return True return False def _add_decoration(self, cursor): """ Adds a decoration for the word under ``cursor``. """ if self.select_word(cursor): self.editor.set_mouse_cursor(Qt.PointingHandCursor) else: self.editor.set_mouse_cursor(Qt.IBeamCursor) def _remove_decoration(self): """ Removes the word under cursor's decoration """ if self._deco is not None: self.editor.decorations.remove(self._deco) self._deco = None def _goto_def(self): if self._definition: line = self._definition.line col = self._definition.column TextHelper(self.editor).goto_line(line, move=True, column=col)
pyQode/pyqode.cobol
pyqode/cobol/api/folding.py
CobolFoldDetector.normalize_text
python
def normalize_text(self, text): if not self.editor.free_format: text = ' ' * 6 + text[6:] return text.upper()
Normalize text, when fixed format is ON, replace the first 6 chars by a space.
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/api/folding.py#L43-L49
null
class CobolFoldDetector(FoldDetector): def __init__(self): super().__init__() self.proc_division = None self._proc_div_txt = "" self.data_division = None self._data_div_txt = "" self.variables = set() self.divisions = [] def is_valid(self, block): return block is not None and block.isValid() def stripped_texts(self, block, prev_block): ctext = block.text().rstrip().upper() if self.is_valid(prev_block): ptext = prev_block.text().rstrip().upper() else: ptext = '' return ctext, ptext def is_in_data_division(self, block): for div_block, div_type in reversed(self.divisions): if div_block.blockNumber() < block.blockNumber(): return div_type == 'data' return False def is_in_proc_division(self, block): for div_block, div_type in reversed(self.divisions): if div_block.blockNumber() < block.blockNumber(): return div_type == 'procedure' return False def get_indent(self, normalized_text): indent = len(normalized_text) - len(normalized_text.lstrip()) return indent + indent % 2 def detect_fold_level(self, prev_block, block): ctext, ptext = self.stripped_texts(block, prev_block) if not self.editor.free_format: ctext = self.normalize_text(ctext) ptext = self.normalize_text(ptext) if regex.DIVISION.indexIn(ctext) != -1 and not ctext.lstrip().startswith('*'): return OFFSET_DIVISION elif regex.SECTION.indexIn(ctext) != -1 and not ctext.lstrip().startswith('*'): return OFFSET_SECTION else: # anywhere else, folding is mostly based on the indentation level indent = self.get_indent(ctext) pindent = self.get_indent(ptext) if ctext.strip().upper().startswith('END-') and self.is_valid(prev_block) and pindent > indent: # find previous block with the same indent, use it's fold level + 1 to include # the end-branch statement in the fold scope pblock = prev_block while self.is_valid(pblock) and (pindent != indent or len(ptext.strip()) == 0): pblock = pblock.previous() ptext = self.normalize_text(pblock.text()) pindent = self.get_indent(ptext) lvl = TextBlockHelper.get_fold_lvl(pblock.next()) else: lvl = OFFSET_OTHER + indent # if not self.editor.free_format and (ctext.lstrip().startswith('-') or ctext.lstrip().startswith('*')): if not self.editor.free_format and (ctext.lstrip().startswith('-')): # use previous fold level lvl = TextBlockHelper.get_fold_lvl(prev_block) if not self.editor.free_format and ctext.strip().startswith('*'): if regex.DIVISION.indexIn(ptext) != -1 and not ptext.lstrip().startswith('*'): lvl = OFFSET_SECTION elif regex.SECTION.indexIn(ptext) != -1 and not ptext.lstrip().startswith('*'): return OFFSET_SECTION + 2 else: lvl = TextBlockHelper.get_fold_lvl(prev_block) return lvl
pyQode/pyqode.cobol
pyqode/cobol/api/pic.py
_clean_code
python
def _clean_code(code): lines = [] # cleanup lines, the parser is very sensitive to extra spaces,... for l in code.splitlines(): # remove last . if l.endswith('.'): l = l[:-1] # the parser doe not like VALUE xxx. if "VALUE" in l: l = l[:l.find("VALUE")] # the parser does not like extra spaces between "PIC X(xxx)" and "." indent = len(l) - len(l.lstrip()) tokens = l.split(" ") while "" in tokens: tokens.remove("") if tokens and not tokens[-1].endswith("."): tokens[-1] += "." lines.append(" " * indent + " ".join(tokens)) return lines
Cleans the received code (the parser does not like extra spaces not a VALUE statement). Returns the cleaned code as a list of lines. :param code: The COBOL code to clean :return The list of code lines (cleaned)
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/api/pic.py#L17-L44
null
from .parsers.pic import process_cobol class PicFieldInfo(object): """ This structure holds the information about a PIC field. """ offset = 0 name = "" level = 0 pic = "" occurs = None redefines = None indexed_by = None def get_field_infos(code, free_format): """ Gets the list of pic fields information from line |start| to line |end|. :param code: code to parse :returns: the list of pic fields info found in the specified text. """ offset = 0 field_infos = [] lines = _clean_code(code) previous_offset = 0 for row in process_cobol(lines, free_format): fi = PicFieldInfo() fi.name = row["name"] fi.level = row["level"] fi.pic = row["pic"] fi.occurs = row["occurs"] fi.redefines = row["redefines"] fi.indexed_by = row["indexed_by"] # find item that was redefined and use its offset if fi.redefines: for fib in field_infos: if fib.name == fi.redefines: offset = fib.offset # level 1 should have their offset set to 1 if fi.level == 1: offset = 1 # level 78 have no offset if fi.level == 78: offset = 0 # level 77 have offset always to 1 if fi.level == 77: offset = 1 # set item offset fi.offset = offset # special case: level 88 have the same level as its parent if fi.level == 88: fi.offset = previous_offset else: previous_offset = offset field_infos.append(fi) # compute offset of next PIC field. if row['pic']: offset += row['pic_info']['length'] return field_infos
pyQode/pyqode.cobol
pyqode/cobol/api/pic.py
get_field_infos
python
def get_field_infos(code, free_format): offset = 0 field_infos = [] lines = _clean_code(code) previous_offset = 0 for row in process_cobol(lines, free_format): fi = PicFieldInfo() fi.name = row["name"] fi.level = row["level"] fi.pic = row["pic"] fi.occurs = row["occurs"] fi.redefines = row["redefines"] fi.indexed_by = row["indexed_by"] # find item that was redefined and use its offset if fi.redefines: for fib in field_infos: if fib.name == fi.redefines: offset = fib.offset # level 1 should have their offset set to 1 if fi.level == 1: offset = 1 # level 78 have no offset if fi.level == 78: offset = 0 # level 77 have offset always to 1 if fi.level == 77: offset = 1 # set item offset fi.offset = offset # special case: level 88 have the same level as its parent if fi.level == 88: fi.offset = previous_offset else: previous_offset = offset field_infos.append(fi) # compute offset of next PIC field. if row['pic']: offset += row['pic_info']['length'] return field_infos
Gets the list of pic fields information from line |start| to line |end|. :param code: code to parse :returns: the list of pic fields info found in the specified text.
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/api/pic.py#L47-L103
[ "def process_cobol(lines, free_format):\n return clean_names(denormalize_cobol(parse_cobol(clean_cobol(lines, free_format))),\n ensure_unique_names=False, strip_prefix=False,\n make_database_safe=False)\n", "def _clean_code(code):\n \"\"\"\n Cleans the received...
from .parsers.pic import process_cobol class PicFieldInfo(object): """ This structure holds the information about a PIC field. """ offset = 0 name = "" level = 0 pic = "" occurs = None redefines = None indexed_by = None def _clean_code(code): """ Cleans the received code (the parser does not like extra spaces not a VALUE statement). Returns the cleaned code as a list of lines. :param code: The COBOL code to clean :return The list of code lines (cleaned) """ lines = [] # cleanup lines, the parser is very sensitive to extra spaces,... for l in code.splitlines(): # remove last . if l.endswith('.'): l = l[:-1] # the parser doe not like VALUE xxx. if "VALUE" in l: l = l[:l.find("VALUE")] # the parser does not like extra spaces between "PIC X(xxx)" and "." indent = len(l) - len(l.lstrip()) tokens = l.split(" ") while "" in tokens: tokens.remove("") if tokens and not tokens[-1].endswith("."): tokens[-1] += "." lines.append(" " * indent + " ".join(tokens)) return lines
pyQode/pyqode.cobol
pyqode/cobol/modes/indenter.py
IndenterMode.unindent_selection
python
def unindent_selection(self, cursor): doc = self.editor.document() tab_len = self.editor.tab_length nb_lines = len(cursor.selection().toPlainText().splitlines()) if nb_lines == 0: nb_lines = 1 block = doc.findBlock(cursor.selectionStart()) assert isinstance(block, QtGui.QTextBlock) i = 0 _logger().debug('unindent selection: %d lines', nb_lines) while i < nb_lines: txt = block.text()[self.min_column:] _logger().debug('line to unindent: %s', txt) _logger().debug('self.editor.use_spaces_instead_of_tabs: %r', self.editor.use_spaces_instead_of_tabs) if self.editor.use_spaces_instead_of_tabs: indentation = len(txt) - len(txt.lstrip()) else: indentation = len(txt) - len(txt.replace('\t', '')) _logger().debug('unindent line %d: %d spaces (min indent=%d)', i, indentation, self.min_column) if indentation > 0: c = QtGui.QTextCursor(block) c.movePosition(c.StartOfLine, cursor.MoveAnchor) c.movePosition(c.Right, cursor.MoveAnchor, indentation + self.min_column) max_spaces = indentation % tab_len if max_spaces == 0: max_spaces = tab_len spaces = self.count_deletable_spaces(c, max_spaces) for _ in range(spaces): c.deletePreviousChar() block = block.next() i += 1 return cursor
Un-indents selected text :param cursor: QTextCursor
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/modes/indenter.py#L70-L107
[ "def _logger():\n return logging.getLogger(__name__)\n" ]
class IndenterMode(Mode): """ Implements classic indentation/tabulation (Tab/Shift+Tab) It inserts/removes tabulations (a series of spaces defined by the tabLength settings) at the cursor position if there is no selection, otherwise it fully indents/un-indents selected lines. To trigger an indentation/un-indentation programatically, you must emit :attr:`pyqode.core.api.CodeEdit.indent_requested` or :attr:`pyqode.core.api.CodeEdit.unindent_requested`. """ @property def min_column(self): return 0 if self.editor.free_format else 7 def __init__(self): super(IndenterMode, self).__init__() def on_state_changed(self, state): if state: self.editor.indent_requested.connect(self.indent) self.editor.unindent_requested.connect(self.unindent) else: self.editor.indent_requested.disconnect(self.indent) self.editor.unindent_requested.disconnect(self.unindent) def indent_selection(self, cursor): """ Indent selected text :param cursor: QTextCursor """ doc = self.editor.document() tab_len = self.editor.tab_length cursor.beginEditBlock() nb_lines = len(cursor.selection().toPlainText().splitlines()) if (cursor.atBlockStart() and cursor.position() == cursor.selectionEnd()): nb_lines += 1 block = doc.findBlock(cursor.selectionStart()) i = 0 # indent every lines while i < nb_lines: nb_space_to_add = tab_len cursor = QtGui.QTextCursor(block) cursor.movePosition(cursor.StartOfLine, cursor.MoveAnchor) cursor.movePosition(cursor.Right, cursor.MoveAnchor, self.min_column) if self.editor.use_spaces_instead_of_tabs: for _ in range(nb_space_to_add): cursor.insertText(" ") else: cursor.insertText('\t') block = block.next() i += 1 cursor.endEditBlock() def indent(self): """ Indents text at cursor position. """ cursor = self.editor.textCursor() assert isinstance(cursor, QtGui.QTextCursor) if cursor.hasSelection(): self.indent_selection(cursor) else: # simply insert indentation at the cursor position tab_len = self.editor.tab_length if cursor.positionInBlock() < self.min_column and not cursor.atBlockEnd(): cursor.movePosition(cursor.Right, cursor.MoveAnchor, self.min_column) cursor.beginEditBlock() if self.editor.use_spaces_instead_of_tabs: nb_space_to_add = tab_len - (cursor.positionInBlock() - self.min_column) % tab_len cursor.insertText(nb_space_to_add * " ") else: cursor.insertText('\t') cursor.endEditBlock() self.editor.setTextCursor(cursor) def count_deletable_spaces(self, cursor, max_spaces): # count the number of spaces deletable, stop at tab len max_spaces = abs(max_spaces) if max_spaces > self.editor.tab_length: max_spaces = self.editor.tab_length spaces = 0 trav_cursor = QtGui.QTextCursor(cursor) while spaces < max_spaces or trav_cursor.atBlockStart(): pos = trav_cursor.position() trav_cursor.movePosition(cursor.Left, cursor.KeepAnchor) char = trav_cursor.selectedText() if char == " ": spaces += 1 else: break trav_cursor.setPosition(pos - 1) return spaces def unindent(self): """ Un-indents text at cursor position. """ _logger().debug('unindent') cursor = self.editor.textCursor() _logger().debug('cursor has selection %r', cursor.hasSelection()) if cursor.hasSelection(): cursor.beginEditBlock() self.unindent_selection(cursor) cursor.endEditBlock() self.editor.setTextCursor(cursor) else: tab_len = self.editor.tab_length indentation = cursor.positionInBlock() indentation -= self.min_column if indentation == 0: return max_spaces = indentation % tab_len if max_spaces == 0: max_spaces = tab_len spaces = self.count_deletable_spaces(cursor, max_spaces) _logger().info('deleting %d space before cursor' % spaces) cursor.beginEditBlock() for _ in range(spaces): cursor.deletePreviousChar() cursor.endEditBlock() self.editor.setTextCursor(cursor) _logger().debug(cursor.block().text())
pyQode/pyqode.cobol
pyqode/cobol/modes/indenter.py
IndenterMode.indent
python
def indent(self): cursor = self.editor.textCursor() assert isinstance(cursor, QtGui.QTextCursor) if cursor.hasSelection(): self.indent_selection(cursor) else: # simply insert indentation at the cursor position tab_len = self.editor.tab_length if cursor.positionInBlock() < self.min_column and not cursor.atBlockEnd(): cursor.movePosition(cursor.Right, cursor.MoveAnchor, self.min_column) cursor.beginEditBlock() if self.editor.use_spaces_instead_of_tabs: nb_space_to_add = tab_len - (cursor.positionInBlock() - self.min_column) % tab_len cursor.insertText(nb_space_to_add * " ") else: cursor.insertText('\t') cursor.endEditBlock() self.editor.setTextCursor(cursor)
Indents text at cursor position.
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/modes/indenter.py#L109-L129
null
class IndenterMode(Mode): """ Implements classic indentation/tabulation (Tab/Shift+Tab) It inserts/removes tabulations (a series of spaces defined by the tabLength settings) at the cursor position if there is no selection, otherwise it fully indents/un-indents selected lines. To trigger an indentation/un-indentation programatically, you must emit :attr:`pyqode.core.api.CodeEdit.indent_requested` or :attr:`pyqode.core.api.CodeEdit.unindent_requested`. """ @property def min_column(self): return 0 if self.editor.free_format else 7 def __init__(self): super(IndenterMode, self).__init__() def on_state_changed(self, state): if state: self.editor.indent_requested.connect(self.indent) self.editor.unindent_requested.connect(self.unindent) else: self.editor.indent_requested.disconnect(self.indent) self.editor.unindent_requested.disconnect(self.unindent) def indent_selection(self, cursor): """ Indent selected text :param cursor: QTextCursor """ doc = self.editor.document() tab_len = self.editor.tab_length cursor.beginEditBlock() nb_lines = len(cursor.selection().toPlainText().splitlines()) if (cursor.atBlockStart() and cursor.position() == cursor.selectionEnd()): nb_lines += 1 block = doc.findBlock(cursor.selectionStart()) i = 0 # indent every lines while i < nb_lines: nb_space_to_add = tab_len cursor = QtGui.QTextCursor(block) cursor.movePosition(cursor.StartOfLine, cursor.MoveAnchor) cursor.movePosition(cursor.Right, cursor.MoveAnchor, self.min_column) if self.editor.use_spaces_instead_of_tabs: for _ in range(nb_space_to_add): cursor.insertText(" ") else: cursor.insertText('\t') block = block.next() i += 1 cursor.endEditBlock() def unindent_selection(self, cursor): """ Un-indents selected text :param cursor: QTextCursor """ doc = self.editor.document() tab_len = self.editor.tab_length nb_lines = len(cursor.selection().toPlainText().splitlines()) if nb_lines == 0: nb_lines = 1 block = doc.findBlock(cursor.selectionStart()) assert isinstance(block, QtGui.QTextBlock) i = 0 _logger().debug('unindent selection: %d lines', nb_lines) while i < nb_lines: txt = block.text()[self.min_column:] _logger().debug('line to unindent: %s', txt) _logger().debug('self.editor.use_spaces_instead_of_tabs: %r', self.editor.use_spaces_instead_of_tabs) if self.editor.use_spaces_instead_of_tabs: indentation = len(txt) - len(txt.lstrip()) else: indentation = len(txt) - len(txt.replace('\t', '')) _logger().debug('unindent line %d: %d spaces (min indent=%d)', i, indentation, self.min_column) if indentation > 0: c = QtGui.QTextCursor(block) c.movePosition(c.StartOfLine, cursor.MoveAnchor) c.movePosition(c.Right, cursor.MoveAnchor, indentation + self.min_column) max_spaces = indentation % tab_len if max_spaces == 0: max_spaces = tab_len spaces = self.count_deletable_spaces(c, max_spaces) for _ in range(spaces): c.deletePreviousChar() block = block.next() i += 1 return cursor def count_deletable_spaces(self, cursor, max_spaces): # count the number of spaces deletable, stop at tab len max_spaces = abs(max_spaces) if max_spaces > self.editor.tab_length: max_spaces = self.editor.tab_length spaces = 0 trav_cursor = QtGui.QTextCursor(cursor) while spaces < max_spaces or trav_cursor.atBlockStart(): pos = trav_cursor.position() trav_cursor.movePosition(cursor.Left, cursor.KeepAnchor) char = trav_cursor.selectedText() if char == " ": spaces += 1 else: break trav_cursor.setPosition(pos - 1) return spaces def unindent(self): """ Un-indents text at cursor position. """ _logger().debug('unindent') cursor = self.editor.textCursor() _logger().debug('cursor has selection %r', cursor.hasSelection()) if cursor.hasSelection(): cursor.beginEditBlock() self.unindent_selection(cursor) cursor.endEditBlock() self.editor.setTextCursor(cursor) else: tab_len = self.editor.tab_length indentation = cursor.positionInBlock() indentation -= self.min_column if indentation == 0: return max_spaces = indentation % tab_len if max_spaces == 0: max_spaces = tab_len spaces = self.count_deletable_spaces(cursor, max_spaces) _logger().info('deleting %d space before cursor' % spaces) cursor.beginEditBlock() for _ in range(spaces): cursor.deletePreviousChar() cursor.endEditBlock() self.editor.setTextCursor(cursor) _logger().debug(cursor.block().text())
pyQode/pyqode.cobol
pyqode/cobol/modes/indenter.py
IndenterMode.unindent
python
def unindent(self): _logger().debug('unindent') cursor = self.editor.textCursor() _logger().debug('cursor has selection %r', cursor.hasSelection()) if cursor.hasSelection(): cursor.beginEditBlock() self.unindent_selection(cursor) cursor.endEditBlock() self.editor.setTextCursor(cursor) else: tab_len = self.editor.tab_length indentation = cursor.positionInBlock() indentation -= self.min_column if indentation == 0: return max_spaces = indentation % tab_len if max_spaces == 0: max_spaces = tab_len spaces = self.count_deletable_spaces(cursor, max_spaces) _logger().info('deleting %d space before cursor' % spaces) cursor.beginEditBlock() for _ in range(spaces): cursor.deletePreviousChar() cursor.endEditBlock() self.editor.setTextCursor(cursor) _logger().debug(cursor.block().text())
Un-indents text at cursor position.
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/modes/indenter.py#L149-L178
[ "def _logger():\n return logging.getLogger(__name__)\n" ]
class IndenterMode(Mode): """ Implements classic indentation/tabulation (Tab/Shift+Tab) It inserts/removes tabulations (a series of spaces defined by the tabLength settings) at the cursor position if there is no selection, otherwise it fully indents/un-indents selected lines. To trigger an indentation/un-indentation programatically, you must emit :attr:`pyqode.core.api.CodeEdit.indent_requested` or :attr:`pyqode.core.api.CodeEdit.unindent_requested`. """ @property def min_column(self): return 0 if self.editor.free_format else 7 def __init__(self): super(IndenterMode, self).__init__() def on_state_changed(self, state): if state: self.editor.indent_requested.connect(self.indent) self.editor.unindent_requested.connect(self.unindent) else: self.editor.indent_requested.disconnect(self.indent) self.editor.unindent_requested.disconnect(self.unindent) def indent_selection(self, cursor): """ Indent selected text :param cursor: QTextCursor """ doc = self.editor.document() tab_len = self.editor.tab_length cursor.beginEditBlock() nb_lines = len(cursor.selection().toPlainText().splitlines()) if (cursor.atBlockStart() and cursor.position() == cursor.selectionEnd()): nb_lines += 1 block = doc.findBlock(cursor.selectionStart()) i = 0 # indent every lines while i < nb_lines: nb_space_to_add = tab_len cursor = QtGui.QTextCursor(block) cursor.movePosition(cursor.StartOfLine, cursor.MoveAnchor) cursor.movePosition(cursor.Right, cursor.MoveAnchor, self.min_column) if self.editor.use_spaces_instead_of_tabs: for _ in range(nb_space_to_add): cursor.insertText(" ") else: cursor.insertText('\t') block = block.next() i += 1 cursor.endEditBlock() def unindent_selection(self, cursor): """ Un-indents selected text :param cursor: QTextCursor """ doc = self.editor.document() tab_len = self.editor.tab_length nb_lines = len(cursor.selection().toPlainText().splitlines()) if nb_lines == 0: nb_lines = 1 block = doc.findBlock(cursor.selectionStart()) assert isinstance(block, QtGui.QTextBlock) i = 0 _logger().debug('unindent selection: %d lines', nb_lines) while i < nb_lines: txt = block.text()[self.min_column:] _logger().debug('line to unindent: %s', txt) _logger().debug('self.editor.use_spaces_instead_of_tabs: %r', self.editor.use_spaces_instead_of_tabs) if self.editor.use_spaces_instead_of_tabs: indentation = len(txt) - len(txt.lstrip()) else: indentation = len(txt) - len(txt.replace('\t', '')) _logger().debug('unindent line %d: %d spaces (min indent=%d)', i, indentation, self.min_column) if indentation > 0: c = QtGui.QTextCursor(block) c.movePosition(c.StartOfLine, cursor.MoveAnchor) c.movePosition(c.Right, cursor.MoveAnchor, indentation + self.min_column) max_spaces = indentation % tab_len if max_spaces == 0: max_spaces = tab_len spaces = self.count_deletable_spaces(c, max_spaces) for _ in range(spaces): c.deletePreviousChar() block = block.next() i += 1 return cursor def indent(self): """ Indents text at cursor position. """ cursor = self.editor.textCursor() assert isinstance(cursor, QtGui.QTextCursor) if cursor.hasSelection(): self.indent_selection(cursor) else: # simply insert indentation at the cursor position tab_len = self.editor.tab_length if cursor.positionInBlock() < self.min_column and not cursor.atBlockEnd(): cursor.movePosition(cursor.Right, cursor.MoveAnchor, self.min_column) cursor.beginEditBlock() if self.editor.use_spaces_instead_of_tabs: nb_space_to_add = tab_len - (cursor.positionInBlock() - self.min_column) % tab_len cursor.insertText(nb_space_to_add * " ") else: cursor.insertText('\t') cursor.endEditBlock() self.editor.setTextCursor(cursor) def count_deletable_spaces(self, cursor, max_spaces): # count the number of spaces deletable, stop at tab len max_spaces = abs(max_spaces) if max_spaces > self.editor.tab_length: max_spaces = self.editor.tab_length spaces = 0 trav_cursor = QtGui.QTextCursor(cursor) while spaces < max_spaces or trav_cursor.atBlockStart(): pos = trav_cursor.position() trav_cursor.movePosition(cursor.Left, cursor.KeepAnchor) char = trav_cursor.selectedText() if char == " ": spaces += 1 else: break trav_cursor.setPosition(pos - 1) return spaces
pyQode/pyqode.cobol
pyqode/cobol/widgets/pic_offsets.py
PicOffsetsTable.set_editor
python
def set_editor(self, editor): if self._editor is not None: try: self._editor.offset_calculator.pic_infos_available.disconnect( self._update) except (AttributeError, RuntimeError, ReferenceError): # see https://github.com/OpenCobolIDE/OpenCobolIDE/issues/89 pass self._editor = weakref.proxy(editor) if editor else editor try: self._editor.offset_calculator.pic_infos_available.connect( self._update) except AttributeError: pass
Sets the associated editor, when the editor's offset calculator mode emit the signal pic_infos_available, the table is automatically refreshed. You can also refresh manually by calling :meth:`update_pic_infos`.
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/widgets/pic_offsets.py#L25-L45
null
class PicOffsetsTable(QtWidgets.QTableWidget): """ Displays the pic field offsets. """ #: signal emitted when the widget should be shown (i.e. when the pic info #: has been updated) show_requested = QtCore.Signal() def __init__(self, parent=None): super().__init__(parent) self._update([]) self._editor = None self.verticalHeader().setVisible(False) self.setColumnCount(4) self.setHorizontalHeaderLabels( ['Level', 'Name', 'Offset', 'PIC']) self.setEditTriggers(self.NoEditTriggers) self.setSelectionBehavior(self.SelectRows) self.setSelectionMode(self.SingleSelection) def update_pic_infos(self, infos): """ Update the pic filed informations shown in the table. """ self._update(infos) def _update(self, infos): self.clearContents() self.setRowCount(len(infos)) # process each info in a separate row for i, info in enumerate(infos): self.setItem( i, 0, QtWidgets.QTableWidgetItem("%s" % info.level)) self.setItem( i, 1, QtWidgets.QTableWidgetItem(info.name)) self.setItem( i, 2, QtWidgets.QTableWidgetItem("%s" % info.offset)) self.setItem( i, 3, QtWidgets.QTableWidgetItem(info.pic)) self.setSortingEnabled(False) self.show_requested.emit()
pyQode/pyqode.cobol
pyqode/cobol/api/parsers/names.py
cmp_name
python
def cmp_name(first_node, second_node): if len(first_node.children) == len(second_node.children): for first_child, second_child in zip(first_node.children, second_node.children): for key in first_child.__dict__.keys(): if key.startswith('_'): continue if first_child.__dict__[key] != second_child.__dict__[key]: return 1 ret_val = cmp_name(first_child, second_child) if ret_val != 0: return 1 else: return 1 return 0
Compare two name recursively. :param first_node: First node :param second_node: Second state :return: 0 if same name, 1 if names are differents.
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/api/parsers/names.py#L99-L122
[ "def cmp_name(first_node, second_node):\n \"\"\"\n Compare two name recursively.\n\n :param first_node: First node\n\n :param second_node: Second state\n\n :return: 0 if same name, 1 if names are differents.\n \"\"\"\n if len(first_node.children) == len(second_node.children):\n for first...
""" This parser parses the defined names in a cobol program and store them under the appropriate division/section. The code comes from OpenCobolIDE and has been left mostly intact. """ import logging import re from pyqode.cobol.api import icons from pyqode.cobol.api.keywords import ALL_KEYWORDS from pyqode.core.share import Definition from pyqode.cobol.api import regex def _logger(): return logging.getLogger(__name__) class Name(object): """ A Name is a node in the simplified abstract syntax tree. """ class Type: """ Enumerates the possible name types (div, section, paragraph,...) """ Root = -1 Division = 0 Section = 1 Variable = 2 Paragraph = 3 def __init__(self, node_type, line, column, name, description=None): self.node_type = node_type self.line = line self.column = column self.end_line = -1 self.name = name if description is None: description = name self.description = description.replace(".", "") self.children = [] def add_child(self, child): """ Add a child to the node :param child: The child node to add """ self.children.append(child) child.parent = self def find(self, name): """ Finds a possible child whose name match the name parameter. :param name: name of the child node to look up :type name: str :return: DocumentNode or None """ for c in self.children: if c.name == name: return c result = c.find(name) if result: return result def __repr__(self): type_names = { self.Type.Root: "Root", self.Type.Division: "Division", self.Type.Paragraph: "Paragraph", self.Type.Section: "Section", self.Type.Variable: "Variable" } return "%s(name=%s, line=%s, end_line=%s)" % ( type_names[self.node_type], self.name, self.line, self.end_line) def to_definition(self): """ Converts the name instance to a pyqode.core.share.Definition """ icon = { Name.Type.Root: icons.ICON_MIMETYPE, Name.Type.Division: icons.ICON_DIVISION, Name.Type.Section: icons.ICON_SECTION, Name.Type.Variable: icons.ICON_VAR, Name.Type.Paragraph: icons.ICON_FUNC }[self.node_type] d = Definition(self.name, self.line, self.column, icon, self.description) for ch in self.children: d.add_child(ch.to_definition()) return d def parse_division(l, c, line, root_node, last_section_node): """ Extracts a division node from a line :param l: The line number (starting from 0) :param c: The column number :param line: The line string (without indentation) :param root_node: The document root node. :return: tuple(last_div_node, last_section_node) """ name = line name = name.replace(".", "") # trim whitespaces/tabs between XXX and DIVISION tokens = [t for t in name.split(' ') if t] node = Name(Name.Type.Division, l, c, '%s %s' % (tokens[0], tokens[1])) root_node.add_child(node) last_div_node = node # do not take previous sections into account if last_section_node: last_section_node.end_line = l last_section_node = None return last_div_node, last_section_node def parse_section(l, c, last_div_node, last_vars, line): """ Extracts a section node from a line. :param l: The line number (starting from 0) :param last_div_node: The last div node found :param last_vars: The last vars dict :param line: The line string (without indentation) :return: last_section_node """ name = line name = name.replace(".", "") node = Name(Name.Type.Section, l, c, name) last_div_node.add_child(node) last_section_node = node # do not take previous var into account last_vars.clear() return last_section_node def parse_pic_field(l, c, last_section_node, last_vars, line): """ Parse a pic field line. Return A VariableNode or None in case of malformed code. :param l: The line number (starting from 0) :param c: The column number (starting from 0) :param last_section_node: The last section node found :param last_vars: The last vars dict :param line: The line string (without indentation) :return: The extracted variable node """ parent_node = None raw_tokens = line.split(" ") tokens = [] for t in raw_tokens: if not t.isspace() and t != "": tokens.append(t) try: if tokens[0].upper() == "FD": lvl = 1 else: lvl = int(tokens[0], 16) name = tokens[1] except ValueError: return None except IndexError: # line not complete return None name = name.replace(".", "") if name in ALL_KEYWORDS or name in ['-', '/']: return None m = re.findall(r'pic.*\.', line, re.IGNORECASE) if m: description = ' '.join([t for t in m[0].split(' ') if t]) else: description = line try: index = description.lower().index('value') except ValueError: description = description.replace('.', '') else: description = description[index:].replace('value', '')[:80] if lvl == int('78', 16): lvl = 1 if lvl == 1: parent_node = last_section_node last_vars.clear() else: # find parent level levels = sorted(last_vars.keys(), reverse=True) for lv in levels: if lv < lvl: parent_node = last_vars[lv] break if not parent_node: # malformed code return None # todo: enabled this with an option in pyqode 3.0 # if lvl == int('88', 16): # return None if not name or name.upper().strip() == 'PIC': name = 'FILLER' node = Name(Name.Type.Variable, l, c, name, description) parent_node.add_child(node) last_vars[lvl] = node # remove closed variables levels = sorted(last_vars.keys(), reverse=True) for l in levels: if l > lvl: last_vars.pop(l) return node def parse_paragraph(l, c, last_div_node, last_section_node, line): """ Extracts a paragraph node :param l: The line number (starting from 0) :param last_div_node: The last div node found :param last_section_node: The last section node found :param line: The line string (without indentation) :return: The extracted paragraph node """ if not line.endswith('.'): return None name = line.replace(".", "") if name.strip() == '': return None if name.upper() in ALL_KEYWORDS: return None parent_node = last_div_node if last_section_node is not None: parent_node = last_section_node node = Name(Name.Type.Paragraph, l, c, name) parent_node.add_child(node) return node def defined_names(code, free_format=False): """ Parses a cobol document and build a name tree. For convenience, it also returns the list of variables (PIC) and procedures (paragraphs). :param code: cobol code to parse. Default is None. :param free_format: True if the source code must be considered as coded in free format. :return: A tuple made up of the name tree root node, the list of variables and the list of paragraphs. :rtype: Name, list of Name, list of Name """ root_node = Name(Name.Type.Root, 0, 0, 'root') variables = [] paragraphs = [] lines = code.splitlines() last_div_node = None last_section_node = None last_vars = {} last_par = None for i, line in enumerate(lines): if not free_format: if len(line) >= 6: line = 6 * " " + line[6:] column = len(line) - len(line.lstrip()) if not line.isspace() and not line.strip().startswith("*"): line = line.strip() # DIVISIONS if regex.DIVISION.indexIn(line.upper()) != -1 and 'EXIT' not in line.upper(): # remember if last_div_node is not None: last_div_node.end_line = i last_div_node, last_section_node = parse_division( i, column, line, root_node, last_section_node) # SECTIONS elif regex.SECTION.indexIn(line.upper()) != -1 and 'EXIT' not in line.upper(): if last_section_node: last_section_node.end_line = i if last_div_node is None: name = 'PROCEDURE DIVISION' for to_check in ['WORKING-STORAGE', 'LOCAL-STORAGE', 'LINKAGE', 'REPORT ', 'SCREEN']: if to_check in line.upper(): name = 'DATA DIVISION' last_div_node = Name(Name.Type.Division, -1, -1, name, name) root_node.add_child(last_div_node) last_section_node = parse_section( i, column, last_div_node, last_vars, line) # VARIABLES # PARAGRAPHS elif (last_div_node is not None and "PROCEDURE DIVISION" in last_div_node.name.upper()): tokens = line.upper().split(" ") if len(tokens) == 1 and not tokens[0] in ALL_KEYWORDS: p = parse_paragraph( i, column, last_div_node, last_section_node, line) if p: paragraphs.append(p) if last_par: last_par.end_line = i last_par = p elif regex.VAR_PATTERN.indexIn(line.upper()) != -1 or line.upper().lstrip().startswith('FD'): if last_div_node is None: last_div_node = Name(Name.Type.Division, -1, -1, 'DATA DIVISION', '') root_node.add_child(last_div_node) if last_section_node is None: last_section_node = Name(Name.Type.Section, -1, -1, 'WORKING-STORAGE SECTION', '') last_div_node.add_child(last_section_node) v = parse_pic_field( i, column, last_section_node, last_vars, line) if v: variables.append(v) # close last div if last_par: last_par.end_line = len(lines) - 1 if last_div_node: last_div_node.end_line = len(lines) if root_node and last_div_node: root_node.end_line = last_div_node.end_line return root_node, variables, paragraphs if __name__ == '__main__': with open('/Users/Colin/test.cbl') as f: root, variables, paragraphs = defined_names(f.read()) print(root)
pyQode/pyqode.cobol
pyqode/cobol/api/parsers/names.py
parse_division
python
def parse_division(l, c, line, root_node, last_section_node): name = line name = name.replace(".", "") # trim whitespaces/tabs between XXX and DIVISION tokens = [t for t in name.split(' ') if t] node = Name(Name.Type.Division, l, c, '%s %s' % (tokens[0], tokens[1])) root_node.add_child(node) last_div_node = node # do not take previous sections into account if last_section_node: last_section_node.end_line = l last_section_node = None return last_div_node, last_section_node
Extracts a division node from a line :param l: The line number (starting from 0) :param c: The column number :param line: The line string (without indentation) :param root_node: The document root node. :return: tuple(last_div_node, last_section_node)
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/api/parsers/names.py#L125-L150
null
""" This parser parses the defined names in a cobol program and store them under the appropriate division/section. The code comes from OpenCobolIDE and has been left mostly intact. """ import logging import re from pyqode.cobol.api import icons from pyqode.cobol.api.keywords import ALL_KEYWORDS from pyqode.core.share import Definition from pyqode.cobol.api import regex def _logger(): return logging.getLogger(__name__) class Name(object): """ A Name is a node in the simplified abstract syntax tree. """ class Type: """ Enumerates the possible name types (div, section, paragraph,...) """ Root = -1 Division = 0 Section = 1 Variable = 2 Paragraph = 3 def __init__(self, node_type, line, column, name, description=None): self.node_type = node_type self.line = line self.column = column self.end_line = -1 self.name = name if description is None: description = name self.description = description.replace(".", "") self.children = [] def add_child(self, child): """ Add a child to the node :param child: The child node to add """ self.children.append(child) child.parent = self def find(self, name): """ Finds a possible child whose name match the name parameter. :param name: name of the child node to look up :type name: str :return: DocumentNode or None """ for c in self.children: if c.name == name: return c result = c.find(name) if result: return result def __repr__(self): type_names = { self.Type.Root: "Root", self.Type.Division: "Division", self.Type.Paragraph: "Paragraph", self.Type.Section: "Section", self.Type.Variable: "Variable" } return "%s(name=%s, line=%s, end_line=%s)" % ( type_names[self.node_type], self.name, self.line, self.end_line) def to_definition(self): """ Converts the name instance to a pyqode.core.share.Definition """ icon = { Name.Type.Root: icons.ICON_MIMETYPE, Name.Type.Division: icons.ICON_DIVISION, Name.Type.Section: icons.ICON_SECTION, Name.Type.Variable: icons.ICON_VAR, Name.Type.Paragraph: icons.ICON_FUNC }[self.node_type] d = Definition(self.name, self.line, self.column, icon, self.description) for ch in self.children: d.add_child(ch.to_definition()) return d def cmp_name(first_node, second_node): """ Compare two name recursively. :param first_node: First node :param second_node: Second state :return: 0 if same name, 1 if names are differents. """ if len(first_node.children) == len(second_node.children): for first_child, second_child in zip(first_node.children, second_node.children): for key in first_child.__dict__.keys(): if key.startswith('_'): continue if first_child.__dict__[key] != second_child.__dict__[key]: return 1 ret_val = cmp_name(first_child, second_child) if ret_val != 0: return 1 else: return 1 return 0 def parse_section(l, c, last_div_node, last_vars, line): """ Extracts a section node from a line. :param l: The line number (starting from 0) :param last_div_node: The last div node found :param last_vars: The last vars dict :param line: The line string (without indentation) :return: last_section_node """ name = line name = name.replace(".", "") node = Name(Name.Type.Section, l, c, name) last_div_node.add_child(node) last_section_node = node # do not take previous var into account last_vars.clear() return last_section_node def parse_pic_field(l, c, last_section_node, last_vars, line): """ Parse a pic field line. Return A VariableNode or None in case of malformed code. :param l: The line number (starting from 0) :param c: The column number (starting from 0) :param last_section_node: The last section node found :param last_vars: The last vars dict :param line: The line string (without indentation) :return: The extracted variable node """ parent_node = None raw_tokens = line.split(" ") tokens = [] for t in raw_tokens: if not t.isspace() and t != "": tokens.append(t) try: if tokens[0].upper() == "FD": lvl = 1 else: lvl = int(tokens[0], 16) name = tokens[1] except ValueError: return None except IndexError: # line not complete return None name = name.replace(".", "") if name in ALL_KEYWORDS or name in ['-', '/']: return None m = re.findall(r'pic.*\.', line, re.IGNORECASE) if m: description = ' '.join([t for t in m[0].split(' ') if t]) else: description = line try: index = description.lower().index('value') except ValueError: description = description.replace('.', '') else: description = description[index:].replace('value', '')[:80] if lvl == int('78', 16): lvl = 1 if lvl == 1: parent_node = last_section_node last_vars.clear() else: # find parent level levels = sorted(last_vars.keys(), reverse=True) for lv in levels: if lv < lvl: parent_node = last_vars[lv] break if not parent_node: # malformed code return None # todo: enabled this with an option in pyqode 3.0 # if lvl == int('88', 16): # return None if not name or name.upper().strip() == 'PIC': name = 'FILLER' node = Name(Name.Type.Variable, l, c, name, description) parent_node.add_child(node) last_vars[lvl] = node # remove closed variables levels = sorted(last_vars.keys(), reverse=True) for l in levels: if l > lvl: last_vars.pop(l) return node def parse_paragraph(l, c, last_div_node, last_section_node, line): """ Extracts a paragraph node :param l: The line number (starting from 0) :param last_div_node: The last div node found :param last_section_node: The last section node found :param line: The line string (without indentation) :return: The extracted paragraph node """ if not line.endswith('.'): return None name = line.replace(".", "") if name.strip() == '': return None if name.upper() in ALL_KEYWORDS: return None parent_node = last_div_node if last_section_node is not None: parent_node = last_section_node node = Name(Name.Type.Paragraph, l, c, name) parent_node.add_child(node) return node def defined_names(code, free_format=False): """ Parses a cobol document and build a name tree. For convenience, it also returns the list of variables (PIC) and procedures (paragraphs). :param code: cobol code to parse. Default is None. :param free_format: True if the source code must be considered as coded in free format. :return: A tuple made up of the name tree root node, the list of variables and the list of paragraphs. :rtype: Name, list of Name, list of Name """ root_node = Name(Name.Type.Root, 0, 0, 'root') variables = [] paragraphs = [] lines = code.splitlines() last_div_node = None last_section_node = None last_vars = {} last_par = None for i, line in enumerate(lines): if not free_format: if len(line) >= 6: line = 6 * " " + line[6:] column = len(line) - len(line.lstrip()) if not line.isspace() and not line.strip().startswith("*"): line = line.strip() # DIVISIONS if regex.DIVISION.indexIn(line.upper()) != -1 and 'EXIT' not in line.upper(): # remember if last_div_node is not None: last_div_node.end_line = i last_div_node, last_section_node = parse_division( i, column, line, root_node, last_section_node) # SECTIONS elif regex.SECTION.indexIn(line.upper()) != -1 and 'EXIT' not in line.upper(): if last_section_node: last_section_node.end_line = i if last_div_node is None: name = 'PROCEDURE DIVISION' for to_check in ['WORKING-STORAGE', 'LOCAL-STORAGE', 'LINKAGE', 'REPORT ', 'SCREEN']: if to_check in line.upper(): name = 'DATA DIVISION' last_div_node = Name(Name.Type.Division, -1, -1, name, name) root_node.add_child(last_div_node) last_section_node = parse_section( i, column, last_div_node, last_vars, line) # VARIABLES # PARAGRAPHS elif (last_div_node is not None and "PROCEDURE DIVISION" in last_div_node.name.upper()): tokens = line.upper().split(" ") if len(tokens) == 1 and not tokens[0] in ALL_KEYWORDS: p = parse_paragraph( i, column, last_div_node, last_section_node, line) if p: paragraphs.append(p) if last_par: last_par.end_line = i last_par = p elif regex.VAR_PATTERN.indexIn(line.upper()) != -1 or line.upper().lstrip().startswith('FD'): if last_div_node is None: last_div_node = Name(Name.Type.Division, -1, -1, 'DATA DIVISION', '') root_node.add_child(last_div_node) if last_section_node is None: last_section_node = Name(Name.Type.Section, -1, -1, 'WORKING-STORAGE SECTION', '') last_div_node.add_child(last_section_node) v = parse_pic_field( i, column, last_section_node, last_vars, line) if v: variables.append(v) # close last div if last_par: last_par.end_line = len(lines) - 1 if last_div_node: last_div_node.end_line = len(lines) if root_node and last_div_node: root_node.end_line = last_div_node.end_line return root_node, variables, paragraphs if __name__ == '__main__': with open('/Users/Colin/test.cbl') as f: root, variables, paragraphs = defined_names(f.read()) print(root)
pyQode/pyqode.cobol
pyqode/cobol/api/parsers/names.py
parse_section
python
def parse_section(l, c, last_div_node, last_vars, line): name = line name = name.replace(".", "") node = Name(Name.Type.Section, l, c, name) last_div_node.add_child(node) last_section_node = node # do not take previous var into account last_vars.clear() return last_section_node
Extracts a section node from a line. :param l: The line number (starting from 0) :param last_div_node: The last div node found :param last_vars: The last vars dict :param line: The line string (without indentation) :return: last_section_node
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/api/parsers/names.py#L153-L174
null
""" This parser parses the defined names in a cobol program and store them under the appropriate division/section. The code comes from OpenCobolIDE and has been left mostly intact. """ import logging import re from pyqode.cobol.api import icons from pyqode.cobol.api.keywords import ALL_KEYWORDS from pyqode.core.share import Definition from pyqode.cobol.api import regex def _logger(): return logging.getLogger(__name__) class Name(object): """ A Name is a node in the simplified abstract syntax tree. """ class Type: """ Enumerates the possible name types (div, section, paragraph,...) """ Root = -1 Division = 0 Section = 1 Variable = 2 Paragraph = 3 def __init__(self, node_type, line, column, name, description=None): self.node_type = node_type self.line = line self.column = column self.end_line = -1 self.name = name if description is None: description = name self.description = description.replace(".", "") self.children = [] def add_child(self, child): """ Add a child to the node :param child: The child node to add """ self.children.append(child) child.parent = self def find(self, name): """ Finds a possible child whose name match the name parameter. :param name: name of the child node to look up :type name: str :return: DocumentNode or None """ for c in self.children: if c.name == name: return c result = c.find(name) if result: return result def __repr__(self): type_names = { self.Type.Root: "Root", self.Type.Division: "Division", self.Type.Paragraph: "Paragraph", self.Type.Section: "Section", self.Type.Variable: "Variable" } return "%s(name=%s, line=%s, end_line=%s)" % ( type_names[self.node_type], self.name, self.line, self.end_line) def to_definition(self): """ Converts the name instance to a pyqode.core.share.Definition """ icon = { Name.Type.Root: icons.ICON_MIMETYPE, Name.Type.Division: icons.ICON_DIVISION, Name.Type.Section: icons.ICON_SECTION, Name.Type.Variable: icons.ICON_VAR, Name.Type.Paragraph: icons.ICON_FUNC }[self.node_type] d = Definition(self.name, self.line, self.column, icon, self.description) for ch in self.children: d.add_child(ch.to_definition()) return d def cmp_name(first_node, second_node): """ Compare two name recursively. :param first_node: First node :param second_node: Second state :return: 0 if same name, 1 if names are differents. """ if len(first_node.children) == len(second_node.children): for first_child, second_child in zip(first_node.children, second_node.children): for key in first_child.__dict__.keys(): if key.startswith('_'): continue if first_child.__dict__[key] != second_child.__dict__[key]: return 1 ret_val = cmp_name(first_child, second_child) if ret_val != 0: return 1 else: return 1 return 0 def parse_division(l, c, line, root_node, last_section_node): """ Extracts a division node from a line :param l: The line number (starting from 0) :param c: The column number :param line: The line string (without indentation) :param root_node: The document root node. :return: tuple(last_div_node, last_section_node) """ name = line name = name.replace(".", "") # trim whitespaces/tabs between XXX and DIVISION tokens = [t for t in name.split(' ') if t] node = Name(Name.Type.Division, l, c, '%s %s' % (tokens[0], tokens[1])) root_node.add_child(node) last_div_node = node # do not take previous sections into account if last_section_node: last_section_node.end_line = l last_section_node = None return last_div_node, last_section_node def parse_pic_field(l, c, last_section_node, last_vars, line): """ Parse a pic field line. Return A VariableNode or None in case of malformed code. :param l: The line number (starting from 0) :param c: The column number (starting from 0) :param last_section_node: The last section node found :param last_vars: The last vars dict :param line: The line string (without indentation) :return: The extracted variable node """ parent_node = None raw_tokens = line.split(" ") tokens = [] for t in raw_tokens: if not t.isspace() and t != "": tokens.append(t) try: if tokens[0].upper() == "FD": lvl = 1 else: lvl = int(tokens[0], 16) name = tokens[1] except ValueError: return None except IndexError: # line not complete return None name = name.replace(".", "") if name in ALL_KEYWORDS or name in ['-', '/']: return None m = re.findall(r'pic.*\.', line, re.IGNORECASE) if m: description = ' '.join([t for t in m[0].split(' ') if t]) else: description = line try: index = description.lower().index('value') except ValueError: description = description.replace('.', '') else: description = description[index:].replace('value', '')[:80] if lvl == int('78', 16): lvl = 1 if lvl == 1: parent_node = last_section_node last_vars.clear() else: # find parent level levels = sorted(last_vars.keys(), reverse=True) for lv in levels: if lv < lvl: parent_node = last_vars[lv] break if not parent_node: # malformed code return None # todo: enabled this with an option in pyqode 3.0 # if lvl == int('88', 16): # return None if not name or name.upper().strip() == 'PIC': name = 'FILLER' node = Name(Name.Type.Variable, l, c, name, description) parent_node.add_child(node) last_vars[lvl] = node # remove closed variables levels = sorted(last_vars.keys(), reverse=True) for l in levels: if l > lvl: last_vars.pop(l) return node def parse_paragraph(l, c, last_div_node, last_section_node, line): """ Extracts a paragraph node :param l: The line number (starting from 0) :param last_div_node: The last div node found :param last_section_node: The last section node found :param line: The line string (without indentation) :return: The extracted paragraph node """ if not line.endswith('.'): return None name = line.replace(".", "") if name.strip() == '': return None if name.upper() in ALL_KEYWORDS: return None parent_node = last_div_node if last_section_node is not None: parent_node = last_section_node node = Name(Name.Type.Paragraph, l, c, name) parent_node.add_child(node) return node def defined_names(code, free_format=False): """ Parses a cobol document and build a name tree. For convenience, it also returns the list of variables (PIC) and procedures (paragraphs). :param code: cobol code to parse. Default is None. :param free_format: True if the source code must be considered as coded in free format. :return: A tuple made up of the name tree root node, the list of variables and the list of paragraphs. :rtype: Name, list of Name, list of Name """ root_node = Name(Name.Type.Root, 0, 0, 'root') variables = [] paragraphs = [] lines = code.splitlines() last_div_node = None last_section_node = None last_vars = {} last_par = None for i, line in enumerate(lines): if not free_format: if len(line) >= 6: line = 6 * " " + line[6:] column = len(line) - len(line.lstrip()) if not line.isspace() and not line.strip().startswith("*"): line = line.strip() # DIVISIONS if regex.DIVISION.indexIn(line.upper()) != -1 and 'EXIT' not in line.upper(): # remember if last_div_node is not None: last_div_node.end_line = i last_div_node, last_section_node = parse_division( i, column, line, root_node, last_section_node) # SECTIONS elif regex.SECTION.indexIn(line.upper()) != -1 and 'EXIT' not in line.upper(): if last_section_node: last_section_node.end_line = i if last_div_node is None: name = 'PROCEDURE DIVISION' for to_check in ['WORKING-STORAGE', 'LOCAL-STORAGE', 'LINKAGE', 'REPORT ', 'SCREEN']: if to_check in line.upper(): name = 'DATA DIVISION' last_div_node = Name(Name.Type.Division, -1, -1, name, name) root_node.add_child(last_div_node) last_section_node = parse_section( i, column, last_div_node, last_vars, line) # VARIABLES # PARAGRAPHS elif (last_div_node is not None and "PROCEDURE DIVISION" in last_div_node.name.upper()): tokens = line.upper().split(" ") if len(tokens) == 1 and not tokens[0] in ALL_KEYWORDS: p = parse_paragraph( i, column, last_div_node, last_section_node, line) if p: paragraphs.append(p) if last_par: last_par.end_line = i last_par = p elif regex.VAR_PATTERN.indexIn(line.upper()) != -1 or line.upper().lstrip().startswith('FD'): if last_div_node is None: last_div_node = Name(Name.Type.Division, -1, -1, 'DATA DIVISION', '') root_node.add_child(last_div_node) if last_section_node is None: last_section_node = Name(Name.Type.Section, -1, -1, 'WORKING-STORAGE SECTION', '') last_div_node.add_child(last_section_node) v = parse_pic_field( i, column, last_section_node, last_vars, line) if v: variables.append(v) # close last div if last_par: last_par.end_line = len(lines) - 1 if last_div_node: last_div_node.end_line = len(lines) if root_node and last_div_node: root_node.end_line = last_div_node.end_line return root_node, variables, paragraphs if __name__ == '__main__': with open('/Users/Colin/test.cbl') as f: root, variables, paragraphs = defined_names(f.read()) print(root)
pyQode/pyqode.cobol
pyqode/cobol/api/parsers/names.py
parse_pic_field
python
def parse_pic_field(l, c, last_section_node, last_vars, line): parent_node = None raw_tokens = line.split(" ") tokens = [] for t in raw_tokens: if not t.isspace() and t != "": tokens.append(t) try: if tokens[0].upper() == "FD": lvl = 1 else: lvl = int(tokens[0], 16) name = tokens[1] except ValueError: return None except IndexError: # line not complete return None name = name.replace(".", "") if name in ALL_KEYWORDS or name in ['-', '/']: return None m = re.findall(r'pic.*\.', line, re.IGNORECASE) if m: description = ' '.join([t for t in m[0].split(' ') if t]) else: description = line try: index = description.lower().index('value') except ValueError: description = description.replace('.', '') else: description = description[index:].replace('value', '')[:80] if lvl == int('78', 16): lvl = 1 if lvl == 1: parent_node = last_section_node last_vars.clear() else: # find parent level levels = sorted(last_vars.keys(), reverse=True) for lv in levels: if lv < lvl: parent_node = last_vars[lv] break if not parent_node: # malformed code return None # todo: enabled this with an option in pyqode 3.0 # if lvl == int('88', 16): # return None if not name or name.upper().strip() == 'PIC': name = 'FILLER' node = Name(Name.Type.Variable, l, c, name, description) parent_node.add_child(node) last_vars[lvl] = node # remove closed variables levels = sorted(last_vars.keys(), reverse=True) for l in levels: if l > lvl: last_vars.pop(l) return node
Parse a pic field line. Return A VariableNode or None in case of malformed code. :param l: The line number (starting from 0) :param c: The column number (starting from 0) :param last_section_node: The last section node found :param last_vars: The last vars dict :param line: The line string (without indentation) :return: The extracted variable node
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/api/parsers/names.py#L177-L248
null
""" This parser parses the defined names in a cobol program and store them under the appropriate division/section. The code comes from OpenCobolIDE and has been left mostly intact. """ import logging import re from pyqode.cobol.api import icons from pyqode.cobol.api.keywords import ALL_KEYWORDS from pyqode.core.share import Definition from pyqode.cobol.api import regex def _logger(): return logging.getLogger(__name__) class Name(object): """ A Name is a node in the simplified abstract syntax tree. """ class Type: """ Enumerates the possible name types (div, section, paragraph,...) """ Root = -1 Division = 0 Section = 1 Variable = 2 Paragraph = 3 def __init__(self, node_type, line, column, name, description=None): self.node_type = node_type self.line = line self.column = column self.end_line = -1 self.name = name if description is None: description = name self.description = description.replace(".", "") self.children = [] def add_child(self, child): """ Add a child to the node :param child: The child node to add """ self.children.append(child) child.parent = self def find(self, name): """ Finds a possible child whose name match the name parameter. :param name: name of the child node to look up :type name: str :return: DocumentNode or None """ for c in self.children: if c.name == name: return c result = c.find(name) if result: return result def __repr__(self): type_names = { self.Type.Root: "Root", self.Type.Division: "Division", self.Type.Paragraph: "Paragraph", self.Type.Section: "Section", self.Type.Variable: "Variable" } return "%s(name=%s, line=%s, end_line=%s)" % ( type_names[self.node_type], self.name, self.line, self.end_line) def to_definition(self): """ Converts the name instance to a pyqode.core.share.Definition """ icon = { Name.Type.Root: icons.ICON_MIMETYPE, Name.Type.Division: icons.ICON_DIVISION, Name.Type.Section: icons.ICON_SECTION, Name.Type.Variable: icons.ICON_VAR, Name.Type.Paragraph: icons.ICON_FUNC }[self.node_type] d = Definition(self.name, self.line, self.column, icon, self.description) for ch in self.children: d.add_child(ch.to_definition()) return d def cmp_name(first_node, second_node): """ Compare two name recursively. :param first_node: First node :param second_node: Second state :return: 0 if same name, 1 if names are differents. """ if len(first_node.children) == len(second_node.children): for first_child, second_child in zip(first_node.children, second_node.children): for key in first_child.__dict__.keys(): if key.startswith('_'): continue if first_child.__dict__[key] != second_child.__dict__[key]: return 1 ret_val = cmp_name(first_child, second_child) if ret_val != 0: return 1 else: return 1 return 0 def parse_division(l, c, line, root_node, last_section_node): """ Extracts a division node from a line :param l: The line number (starting from 0) :param c: The column number :param line: The line string (without indentation) :param root_node: The document root node. :return: tuple(last_div_node, last_section_node) """ name = line name = name.replace(".", "") # trim whitespaces/tabs between XXX and DIVISION tokens = [t for t in name.split(' ') if t] node = Name(Name.Type.Division, l, c, '%s %s' % (tokens[0], tokens[1])) root_node.add_child(node) last_div_node = node # do not take previous sections into account if last_section_node: last_section_node.end_line = l last_section_node = None return last_div_node, last_section_node def parse_section(l, c, last_div_node, last_vars, line): """ Extracts a section node from a line. :param l: The line number (starting from 0) :param last_div_node: The last div node found :param last_vars: The last vars dict :param line: The line string (without indentation) :return: last_section_node """ name = line name = name.replace(".", "") node = Name(Name.Type.Section, l, c, name) last_div_node.add_child(node) last_section_node = node # do not take previous var into account last_vars.clear() return last_section_node def parse_paragraph(l, c, last_div_node, last_section_node, line): """ Extracts a paragraph node :param l: The line number (starting from 0) :param last_div_node: The last div node found :param last_section_node: The last section node found :param line: The line string (without indentation) :return: The extracted paragraph node """ if not line.endswith('.'): return None name = line.replace(".", "") if name.strip() == '': return None if name.upper() in ALL_KEYWORDS: return None parent_node = last_div_node if last_section_node is not None: parent_node = last_section_node node = Name(Name.Type.Paragraph, l, c, name) parent_node.add_child(node) return node def defined_names(code, free_format=False): """ Parses a cobol document and build a name tree. For convenience, it also returns the list of variables (PIC) and procedures (paragraphs). :param code: cobol code to parse. Default is None. :param free_format: True if the source code must be considered as coded in free format. :return: A tuple made up of the name tree root node, the list of variables and the list of paragraphs. :rtype: Name, list of Name, list of Name """ root_node = Name(Name.Type.Root, 0, 0, 'root') variables = [] paragraphs = [] lines = code.splitlines() last_div_node = None last_section_node = None last_vars = {} last_par = None for i, line in enumerate(lines): if not free_format: if len(line) >= 6: line = 6 * " " + line[6:] column = len(line) - len(line.lstrip()) if not line.isspace() and not line.strip().startswith("*"): line = line.strip() # DIVISIONS if regex.DIVISION.indexIn(line.upper()) != -1 and 'EXIT' not in line.upper(): # remember if last_div_node is not None: last_div_node.end_line = i last_div_node, last_section_node = parse_division( i, column, line, root_node, last_section_node) # SECTIONS elif regex.SECTION.indexIn(line.upper()) != -1 and 'EXIT' not in line.upper(): if last_section_node: last_section_node.end_line = i if last_div_node is None: name = 'PROCEDURE DIVISION' for to_check in ['WORKING-STORAGE', 'LOCAL-STORAGE', 'LINKAGE', 'REPORT ', 'SCREEN']: if to_check in line.upper(): name = 'DATA DIVISION' last_div_node = Name(Name.Type.Division, -1, -1, name, name) root_node.add_child(last_div_node) last_section_node = parse_section( i, column, last_div_node, last_vars, line) # VARIABLES # PARAGRAPHS elif (last_div_node is not None and "PROCEDURE DIVISION" in last_div_node.name.upper()): tokens = line.upper().split(" ") if len(tokens) == 1 and not tokens[0] in ALL_KEYWORDS: p = parse_paragraph( i, column, last_div_node, last_section_node, line) if p: paragraphs.append(p) if last_par: last_par.end_line = i last_par = p elif regex.VAR_PATTERN.indexIn(line.upper()) != -1 or line.upper().lstrip().startswith('FD'): if last_div_node is None: last_div_node = Name(Name.Type.Division, -1, -1, 'DATA DIVISION', '') root_node.add_child(last_div_node) if last_section_node is None: last_section_node = Name(Name.Type.Section, -1, -1, 'WORKING-STORAGE SECTION', '') last_div_node.add_child(last_section_node) v = parse_pic_field( i, column, last_section_node, last_vars, line) if v: variables.append(v) # close last div if last_par: last_par.end_line = len(lines) - 1 if last_div_node: last_div_node.end_line = len(lines) if root_node and last_div_node: root_node.end_line = last_div_node.end_line return root_node, variables, paragraphs if __name__ == '__main__': with open('/Users/Colin/test.cbl') as f: root, variables, paragraphs = defined_names(f.read()) print(root)
pyQode/pyqode.cobol
pyqode/cobol/api/parsers/names.py
parse_paragraph
python
def parse_paragraph(l, c, last_div_node, last_section_node, line): if not line.endswith('.'): return None name = line.replace(".", "") if name.strip() == '': return None if name.upper() in ALL_KEYWORDS: return None parent_node = last_div_node if last_section_node is not None: parent_node = last_section_node node = Name(Name.Type.Paragraph, l, c, name) parent_node.add_child(node) return node
Extracts a paragraph node :param l: The line number (starting from 0) :param last_div_node: The last div node found :param last_section_node: The last section node found :param line: The line string (without indentation) :return: The extracted paragraph node
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/api/parsers/names.py#L251-L273
null
""" This parser parses the defined names in a cobol program and store them under the appropriate division/section. The code comes from OpenCobolIDE and has been left mostly intact. """ import logging import re from pyqode.cobol.api import icons from pyqode.cobol.api.keywords import ALL_KEYWORDS from pyqode.core.share import Definition from pyqode.cobol.api import regex def _logger(): return logging.getLogger(__name__) class Name(object): """ A Name is a node in the simplified abstract syntax tree. """ class Type: """ Enumerates the possible name types (div, section, paragraph,...) """ Root = -1 Division = 0 Section = 1 Variable = 2 Paragraph = 3 def __init__(self, node_type, line, column, name, description=None): self.node_type = node_type self.line = line self.column = column self.end_line = -1 self.name = name if description is None: description = name self.description = description.replace(".", "") self.children = [] def add_child(self, child): """ Add a child to the node :param child: The child node to add """ self.children.append(child) child.parent = self def find(self, name): """ Finds a possible child whose name match the name parameter. :param name: name of the child node to look up :type name: str :return: DocumentNode or None """ for c in self.children: if c.name == name: return c result = c.find(name) if result: return result def __repr__(self): type_names = { self.Type.Root: "Root", self.Type.Division: "Division", self.Type.Paragraph: "Paragraph", self.Type.Section: "Section", self.Type.Variable: "Variable" } return "%s(name=%s, line=%s, end_line=%s)" % ( type_names[self.node_type], self.name, self.line, self.end_line) def to_definition(self): """ Converts the name instance to a pyqode.core.share.Definition """ icon = { Name.Type.Root: icons.ICON_MIMETYPE, Name.Type.Division: icons.ICON_DIVISION, Name.Type.Section: icons.ICON_SECTION, Name.Type.Variable: icons.ICON_VAR, Name.Type.Paragraph: icons.ICON_FUNC }[self.node_type] d = Definition(self.name, self.line, self.column, icon, self.description) for ch in self.children: d.add_child(ch.to_definition()) return d def cmp_name(first_node, second_node): """ Compare two name recursively. :param first_node: First node :param second_node: Second state :return: 0 if same name, 1 if names are differents. """ if len(first_node.children) == len(second_node.children): for first_child, second_child in zip(first_node.children, second_node.children): for key in first_child.__dict__.keys(): if key.startswith('_'): continue if first_child.__dict__[key] != second_child.__dict__[key]: return 1 ret_val = cmp_name(first_child, second_child) if ret_val != 0: return 1 else: return 1 return 0 def parse_division(l, c, line, root_node, last_section_node): """ Extracts a division node from a line :param l: The line number (starting from 0) :param c: The column number :param line: The line string (without indentation) :param root_node: The document root node. :return: tuple(last_div_node, last_section_node) """ name = line name = name.replace(".", "") # trim whitespaces/tabs between XXX and DIVISION tokens = [t for t in name.split(' ') if t] node = Name(Name.Type.Division, l, c, '%s %s' % (tokens[0], tokens[1])) root_node.add_child(node) last_div_node = node # do not take previous sections into account if last_section_node: last_section_node.end_line = l last_section_node = None return last_div_node, last_section_node def parse_section(l, c, last_div_node, last_vars, line): """ Extracts a section node from a line. :param l: The line number (starting from 0) :param last_div_node: The last div node found :param last_vars: The last vars dict :param line: The line string (without indentation) :return: last_section_node """ name = line name = name.replace(".", "") node = Name(Name.Type.Section, l, c, name) last_div_node.add_child(node) last_section_node = node # do not take previous var into account last_vars.clear() return last_section_node def parse_pic_field(l, c, last_section_node, last_vars, line): """ Parse a pic field line. Return A VariableNode or None in case of malformed code. :param l: The line number (starting from 0) :param c: The column number (starting from 0) :param last_section_node: The last section node found :param last_vars: The last vars dict :param line: The line string (without indentation) :return: The extracted variable node """ parent_node = None raw_tokens = line.split(" ") tokens = [] for t in raw_tokens: if not t.isspace() and t != "": tokens.append(t) try: if tokens[0].upper() == "FD": lvl = 1 else: lvl = int(tokens[0], 16) name = tokens[1] except ValueError: return None except IndexError: # line not complete return None name = name.replace(".", "") if name in ALL_KEYWORDS or name in ['-', '/']: return None m = re.findall(r'pic.*\.', line, re.IGNORECASE) if m: description = ' '.join([t for t in m[0].split(' ') if t]) else: description = line try: index = description.lower().index('value') except ValueError: description = description.replace('.', '') else: description = description[index:].replace('value', '')[:80] if lvl == int('78', 16): lvl = 1 if lvl == 1: parent_node = last_section_node last_vars.clear() else: # find parent level levels = sorted(last_vars.keys(), reverse=True) for lv in levels: if lv < lvl: parent_node = last_vars[lv] break if not parent_node: # malformed code return None # todo: enabled this with an option in pyqode 3.0 # if lvl == int('88', 16): # return None if not name or name.upper().strip() == 'PIC': name = 'FILLER' node = Name(Name.Type.Variable, l, c, name, description) parent_node.add_child(node) last_vars[lvl] = node # remove closed variables levels = sorted(last_vars.keys(), reverse=True) for l in levels: if l > lvl: last_vars.pop(l) return node def defined_names(code, free_format=False): """ Parses a cobol document and build a name tree. For convenience, it also returns the list of variables (PIC) and procedures (paragraphs). :param code: cobol code to parse. Default is None. :param free_format: True if the source code must be considered as coded in free format. :return: A tuple made up of the name tree root node, the list of variables and the list of paragraphs. :rtype: Name, list of Name, list of Name """ root_node = Name(Name.Type.Root, 0, 0, 'root') variables = [] paragraphs = [] lines = code.splitlines() last_div_node = None last_section_node = None last_vars = {} last_par = None for i, line in enumerate(lines): if not free_format: if len(line) >= 6: line = 6 * " " + line[6:] column = len(line) - len(line.lstrip()) if not line.isspace() and not line.strip().startswith("*"): line = line.strip() # DIVISIONS if regex.DIVISION.indexIn(line.upper()) != -1 and 'EXIT' not in line.upper(): # remember if last_div_node is not None: last_div_node.end_line = i last_div_node, last_section_node = parse_division( i, column, line, root_node, last_section_node) # SECTIONS elif regex.SECTION.indexIn(line.upper()) != -1 and 'EXIT' not in line.upper(): if last_section_node: last_section_node.end_line = i if last_div_node is None: name = 'PROCEDURE DIVISION' for to_check in ['WORKING-STORAGE', 'LOCAL-STORAGE', 'LINKAGE', 'REPORT ', 'SCREEN']: if to_check in line.upper(): name = 'DATA DIVISION' last_div_node = Name(Name.Type.Division, -1, -1, name, name) root_node.add_child(last_div_node) last_section_node = parse_section( i, column, last_div_node, last_vars, line) # VARIABLES # PARAGRAPHS elif (last_div_node is not None and "PROCEDURE DIVISION" in last_div_node.name.upper()): tokens = line.upper().split(" ") if len(tokens) == 1 and not tokens[0] in ALL_KEYWORDS: p = parse_paragraph( i, column, last_div_node, last_section_node, line) if p: paragraphs.append(p) if last_par: last_par.end_line = i last_par = p elif regex.VAR_PATTERN.indexIn(line.upper()) != -1 or line.upper().lstrip().startswith('FD'): if last_div_node is None: last_div_node = Name(Name.Type.Division, -1, -1, 'DATA DIVISION', '') root_node.add_child(last_div_node) if last_section_node is None: last_section_node = Name(Name.Type.Section, -1, -1, 'WORKING-STORAGE SECTION', '') last_div_node.add_child(last_section_node) v = parse_pic_field( i, column, last_section_node, last_vars, line) if v: variables.append(v) # close last div if last_par: last_par.end_line = len(lines) - 1 if last_div_node: last_div_node.end_line = len(lines) if root_node and last_div_node: root_node.end_line = last_div_node.end_line return root_node, variables, paragraphs if __name__ == '__main__': with open('/Users/Colin/test.cbl') as f: root, variables, paragraphs = defined_names(f.read()) print(root)
pyQode/pyqode.cobol
pyqode/cobol/api/parsers/names.py
defined_names
python
def defined_names(code, free_format=False): root_node = Name(Name.Type.Root, 0, 0, 'root') variables = [] paragraphs = [] lines = code.splitlines() last_div_node = None last_section_node = None last_vars = {} last_par = None for i, line in enumerate(lines): if not free_format: if len(line) >= 6: line = 6 * " " + line[6:] column = len(line) - len(line.lstrip()) if not line.isspace() and not line.strip().startswith("*"): line = line.strip() # DIVISIONS if regex.DIVISION.indexIn(line.upper()) != -1 and 'EXIT' not in line.upper(): # remember if last_div_node is not None: last_div_node.end_line = i last_div_node, last_section_node = parse_division( i, column, line, root_node, last_section_node) # SECTIONS elif regex.SECTION.indexIn(line.upper()) != -1 and 'EXIT' not in line.upper(): if last_section_node: last_section_node.end_line = i if last_div_node is None: name = 'PROCEDURE DIVISION' for to_check in ['WORKING-STORAGE', 'LOCAL-STORAGE', 'LINKAGE', 'REPORT ', 'SCREEN']: if to_check in line.upper(): name = 'DATA DIVISION' last_div_node = Name(Name.Type.Division, -1, -1, name, name) root_node.add_child(last_div_node) last_section_node = parse_section( i, column, last_div_node, last_vars, line) # VARIABLES # PARAGRAPHS elif (last_div_node is not None and "PROCEDURE DIVISION" in last_div_node.name.upper()): tokens = line.upper().split(" ") if len(tokens) == 1 and not tokens[0] in ALL_KEYWORDS: p = parse_paragraph( i, column, last_div_node, last_section_node, line) if p: paragraphs.append(p) if last_par: last_par.end_line = i last_par = p elif regex.VAR_PATTERN.indexIn(line.upper()) != -1 or line.upper().lstrip().startswith('FD'): if last_div_node is None: last_div_node = Name(Name.Type.Division, -1, -1, 'DATA DIVISION', '') root_node.add_child(last_div_node) if last_section_node is None: last_section_node = Name(Name.Type.Section, -1, -1, 'WORKING-STORAGE SECTION', '') last_div_node.add_child(last_section_node) v = parse_pic_field( i, column, last_section_node, last_vars, line) if v: variables.append(v) # close last div if last_par: last_par.end_line = len(lines) - 1 if last_div_node: last_div_node.end_line = len(lines) if root_node and last_div_node: root_node.end_line = last_div_node.end_line return root_node, variables, paragraphs
Parses a cobol document and build a name tree. For convenience, it also returns the list of variables (PIC) and procedures (paragraphs). :param code: cobol code to parse. Default is None. :param free_format: True if the source code must be considered as coded in free format. :return: A tuple made up of the name tree root node, the list of variables and the list of paragraphs. :rtype: Name, list of Name, list of Name
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/api/parsers/names.py#L276-L360
null
""" This parser parses the defined names in a cobol program and store them under the appropriate division/section. The code comes from OpenCobolIDE and has been left mostly intact. """ import logging import re from pyqode.cobol.api import icons from pyqode.cobol.api.keywords import ALL_KEYWORDS from pyqode.core.share import Definition from pyqode.cobol.api import regex def _logger(): return logging.getLogger(__name__) class Name(object): """ A Name is a node in the simplified abstract syntax tree. """ class Type: """ Enumerates the possible name types (div, section, paragraph,...) """ Root = -1 Division = 0 Section = 1 Variable = 2 Paragraph = 3 def __init__(self, node_type, line, column, name, description=None): self.node_type = node_type self.line = line self.column = column self.end_line = -1 self.name = name if description is None: description = name self.description = description.replace(".", "") self.children = [] def add_child(self, child): """ Add a child to the node :param child: The child node to add """ self.children.append(child) child.parent = self def find(self, name): """ Finds a possible child whose name match the name parameter. :param name: name of the child node to look up :type name: str :return: DocumentNode or None """ for c in self.children: if c.name == name: return c result = c.find(name) if result: return result def __repr__(self): type_names = { self.Type.Root: "Root", self.Type.Division: "Division", self.Type.Paragraph: "Paragraph", self.Type.Section: "Section", self.Type.Variable: "Variable" } return "%s(name=%s, line=%s, end_line=%s)" % ( type_names[self.node_type], self.name, self.line, self.end_line) def to_definition(self): """ Converts the name instance to a pyqode.core.share.Definition """ icon = { Name.Type.Root: icons.ICON_MIMETYPE, Name.Type.Division: icons.ICON_DIVISION, Name.Type.Section: icons.ICON_SECTION, Name.Type.Variable: icons.ICON_VAR, Name.Type.Paragraph: icons.ICON_FUNC }[self.node_type] d = Definition(self.name, self.line, self.column, icon, self.description) for ch in self.children: d.add_child(ch.to_definition()) return d def cmp_name(first_node, second_node): """ Compare two name recursively. :param first_node: First node :param second_node: Second state :return: 0 if same name, 1 if names are differents. """ if len(first_node.children) == len(second_node.children): for first_child, second_child in zip(first_node.children, second_node.children): for key in first_child.__dict__.keys(): if key.startswith('_'): continue if first_child.__dict__[key] != second_child.__dict__[key]: return 1 ret_val = cmp_name(first_child, second_child) if ret_val != 0: return 1 else: return 1 return 0 def parse_division(l, c, line, root_node, last_section_node): """ Extracts a division node from a line :param l: The line number (starting from 0) :param c: The column number :param line: The line string (without indentation) :param root_node: The document root node. :return: tuple(last_div_node, last_section_node) """ name = line name = name.replace(".", "") # trim whitespaces/tabs between XXX and DIVISION tokens = [t for t in name.split(' ') if t] node = Name(Name.Type.Division, l, c, '%s %s' % (tokens[0], tokens[1])) root_node.add_child(node) last_div_node = node # do not take previous sections into account if last_section_node: last_section_node.end_line = l last_section_node = None return last_div_node, last_section_node def parse_section(l, c, last_div_node, last_vars, line): """ Extracts a section node from a line. :param l: The line number (starting from 0) :param last_div_node: The last div node found :param last_vars: The last vars dict :param line: The line string (without indentation) :return: last_section_node """ name = line name = name.replace(".", "") node = Name(Name.Type.Section, l, c, name) last_div_node.add_child(node) last_section_node = node # do not take previous var into account last_vars.clear() return last_section_node def parse_pic_field(l, c, last_section_node, last_vars, line): """ Parse a pic field line. Return A VariableNode or None in case of malformed code. :param l: The line number (starting from 0) :param c: The column number (starting from 0) :param last_section_node: The last section node found :param last_vars: The last vars dict :param line: The line string (without indentation) :return: The extracted variable node """ parent_node = None raw_tokens = line.split(" ") tokens = [] for t in raw_tokens: if not t.isspace() and t != "": tokens.append(t) try: if tokens[0].upper() == "FD": lvl = 1 else: lvl = int(tokens[0], 16) name = tokens[1] except ValueError: return None except IndexError: # line not complete return None name = name.replace(".", "") if name in ALL_KEYWORDS or name in ['-', '/']: return None m = re.findall(r'pic.*\.', line, re.IGNORECASE) if m: description = ' '.join([t for t in m[0].split(' ') if t]) else: description = line try: index = description.lower().index('value') except ValueError: description = description.replace('.', '') else: description = description[index:].replace('value', '')[:80] if lvl == int('78', 16): lvl = 1 if lvl == 1: parent_node = last_section_node last_vars.clear() else: # find parent level levels = sorted(last_vars.keys(), reverse=True) for lv in levels: if lv < lvl: parent_node = last_vars[lv] break if not parent_node: # malformed code return None # todo: enabled this with an option in pyqode 3.0 # if lvl == int('88', 16): # return None if not name or name.upper().strip() == 'PIC': name = 'FILLER' node = Name(Name.Type.Variable, l, c, name, description) parent_node.add_child(node) last_vars[lvl] = node # remove closed variables levels = sorted(last_vars.keys(), reverse=True) for l in levels: if l > lvl: last_vars.pop(l) return node def parse_paragraph(l, c, last_div_node, last_section_node, line): """ Extracts a paragraph node :param l: The line number (starting from 0) :param last_div_node: The last div node found :param last_section_node: The last section node found :param line: The line string (without indentation) :return: The extracted paragraph node """ if not line.endswith('.'): return None name = line.replace(".", "") if name.strip() == '': return None if name.upper() in ALL_KEYWORDS: return None parent_node = last_div_node if last_section_node is not None: parent_node = last_section_node node = Name(Name.Type.Paragraph, l, c, name) parent_node.add_child(node) return node if __name__ == '__main__': with open('/Users/Colin/test.cbl') as f: root, variables, paragraphs = defined_names(f.read()) print(root)
pyQode/pyqode.cobol
pyqode/cobol/api/parsers/names.py
Name.find
python
def find(self, name): for c in self.children: if c.name == name: return c result = c.find(name) if result: return result
Finds a possible child whose name match the name parameter. :param name: name of the child node to look up :type name: str :return: DocumentNode or None
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/api/parsers/names.py#L55-L69
null
class Name(object): """ A Name is a node in the simplified abstract syntax tree. """ class Type: """ Enumerates the possible name types (div, section, paragraph,...) """ Root = -1 Division = 0 Section = 1 Variable = 2 Paragraph = 3 def __init__(self, node_type, line, column, name, description=None): self.node_type = node_type self.line = line self.column = column self.end_line = -1 self.name = name if description is None: description = name self.description = description.replace(".", "") self.children = [] def add_child(self, child): """ Add a child to the node :param child: The child node to add """ self.children.append(child) child.parent = self def __repr__(self): type_names = { self.Type.Root: "Root", self.Type.Division: "Division", self.Type.Paragraph: "Paragraph", self.Type.Section: "Section", self.Type.Variable: "Variable" } return "%s(name=%s, line=%s, end_line=%s)" % ( type_names[self.node_type], self.name, self.line, self.end_line) def to_definition(self): """ Converts the name instance to a pyqode.core.share.Definition """ icon = { Name.Type.Root: icons.ICON_MIMETYPE, Name.Type.Division: icons.ICON_DIVISION, Name.Type.Section: icons.ICON_SECTION, Name.Type.Variable: icons.ICON_VAR, Name.Type.Paragraph: icons.ICON_FUNC }[self.node_type] d = Definition(self.name, self.line, self.column, icon, self.description) for ch in self.children: d.add_child(ch.to_definition()) return d
pyQode/pyqode.cobol
pyqode/cobol/api/parsers/names.py
Name.to_definition
python
def to_definition(self): icon = { Name.Type.Root: icons.ICON_MIMETYPE, Name.Type.Division: icons.ICON_DIVISION, Name.Type.Section: icons.ICON_SECTION, Name.Type.Variable: icons.ICON_VAR, Name.Type.Paragraph: icons.ICON_FUNC }[self.node_type] d = Definition(self.name, self.line, self.column, icon, self.description) for ch in self.children: d.add_child(ch.to_definition()) return d
Converts the name instance to a pyqode.core.share.Definition
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/api/parsers/names.py#L82-L96
null
class Name(object): """ A Name is a node in the simplified abstract syntax tree. """ class Type: """ Enumerates the possible name types (div, section, paragraph,...) """ Root = -1 Division = 0 Section = 1 Variable = 2 Paragraph = 3 def __init__(self, node_type, line, column, name, description=None): self.node_type = node_type self.line = line self.column = column self.end_line = -1 self.name = name if description is None: description = name self.description = description.replace(".", "") self.children = [] def add_child(self, child): """ Add a child to the node :param child: The child node to add """ self.children.append(child) child.parent = self def find(self, name): """ Finds a possible child whose name match the name parameter. :param name: name of the child node to look up :type name: str :return: DocumentNode or None """ for c in self.children: if c.name == name: return c result = c.find(name) if result: return result def __repr__(self): type_names = { self.Type.Root: "Root", self.Type.Division: "Division", self.Type.Paragraph: "Paragraph", self.Type.Section: "Section", self.Type.Variable: "Variable" } return "%s(name=%s, line=%s, end_line=%s)" % ( type_names[self.node_type], self.name, self.line, self.end_line)
pyQode/pyqode.cobol
pyqode/cobol/api/parsers/pic.py
parse_cobol
python
def parse_cobol(lines): output = [] intify = ["level", "occurs"] # All in 1 line now, let's parse for row in lines: match = CobolPatterns.row_pattern.match(row.strip()) if not match: _logger().warning("Found unmatched row %s" % row.strip()) continue match = match.groupdict() for i in intify: match[i] = int(match[i]) if match[i] is not None else None if match['pic'] is not None: match['pic_info'] = parse_pic_string(match['pic']) output.append(match) return output
Parses the COBOL - converts the COBOL line into a dictionary containing the information - parses the pic information into type, length, precision - ~~handles redefines~~ -> our implementation does not do that anymore because we want to display item that was redefined.
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/api/parsers/pic.py#L114-L143
[ "def _logger():\n return logging.getLogger(__name__)\n", "def parse_pic_string(pic_str):\n # Expand repeating chars\n while True:\n match = CobolPatterns.pic_pattern_repeats.search(pic_str)\n\n if not match:\n break\n\n expanded_str = match.group(1) * int(match.group(2))\n...
""" This module contains the Python-Cobol pic parser written by Brian Peterson (https://github.com/bpeterso2000/pycobol ) and enhanced by Paulus Schoutsen (https://github.com/balloob/Python-COBOL/blob/master/cobol.py ), licensed under the GPL v3. We slightly modified this module to: - add python3 support - use the pyqode logger mechanism - fix a few PEP8 issues. """ import logging import re def _logger(): return logging.getLogger(__name__) class CobolPatterns: opt_pattern_format = "({})?" row_pattern_base = r'^(?P<level>\d{2})\s*(?P<name>\S+)' row_pattern_redefines = r"\s+REDEFINES\s(?P<redefines>\S+)" row_pattern_pic = r'\s+(COMP-\d\s+)?PIC\s+(?P<pic>\S+)' row_pattern_occurs = r'\s+OCCURS (?P<occurs>\d+) TIMES' row_pattern_indexed_by = r"\s+INDEXED BY\s(?P<indexed_by>\S+)" row_pattern_end = r'\.$' row_pattern = re.compile( row_pattern_base + opt_pattern_format.format(row_pattern_redefines) + opt_pattern_format.format(row_pattern_pic) + opt_pattern_format.format(row_pattern_occurs) + opt_pattern_format.format(row_pattern_indexed_by) + row_pattern_end, re.IGNORECASE ) pic_pattern_repeats = re.compile(r'(.)\((\d+)\)') pic_pattern_float = re.compile(r'S?[9Z]*[.V][9Z]+') pic_pattern_integer = re.compile(r'S?[9Z]+') # Parse the pic string def parse_pic_string(pic_str): # Expand repeating chars while True: match = CobolPatterns.pic_pattern_repeats.search(pic_str) if not match: break expanded_str = match.group(1) * int(match.group(2)) pic_str = CobolPatterns.pic_pattern_repeats.sub( expanded_str, pic_str, 1) # Match to types if CobolPatterns.pic_pattern_float.match(pic_str): data_type = 'Float' elif CobolPatterns.pic_pattern_integer.match(pic_str): data_type = 'Integer' else: data_type = 'Char' # Handle signed if pic_str[0] == "S": data_type = "Signed " + data_type # Handle precision decimal_pos = 0 if 'V' in pic_str: decimal_pos = len(pic_str[pic_str.index('V') + 1:]) pic_str = pic_str.replace('V', '') return { 'type': data_type, 'length': len(pic_str), 'precision': decimal_pos } # Cleans the COBOL by converting the cobol information to single lines def clean_cobol(lines, free_format): holder = [] output = [] for row in lines: if not free_format: row = row[6:72].rstrip() if row == "" or row[0] in ('*', '/'): continue holder.append(row if len(holder) == 0 else row.strip()) if row[-1] == ".": output.append(" ".join(holder)) holder = [] if len(holder) > 0: _logger().warning( "Probably invalid COBOL - found unfinished line: %s" % " ".join(holder)) return output # Helper function # Gets all the lines that have a higher level then the parent_level until # a line with equal or lower level then parent_level is encountered def get_subgroup(parent_level, lines): output = [] for row in lines: if row["level"] > parent_level: output.append(row) else: return output return output def denormalize_cobol(lines): return handle_occurs(lines, 1) # Helper function # Will go ahead and denormalize the COBOL # Beacuse the OCCURS are removed the INDEXED BY will also be removed def handle_occurs(lines, occurs, level_diff=0, name_postfix=""): output = [] for i in range(1, occurs + 1): skip_till = 0 new_name_postfix = (name_postfix if occurs == 1 else name_postfix + '-' + str(i)) for index, row in enumerate(lines): if index < skip_till: continue new_row = row.copy() new_row['level'] += level_diff # Not needed when flattened new_row['indexed_by'] = None if row['occurs'] is None: # First time occurs is just 1, we don't want to add _1 after # *every* field new_row['name'] = row['name'] + new_name_postfix # + "-" + str(i) if occurs > 1 else row['name'] + name_postfix output.append(new_row) else: if row["pic"] is not None: # If it has occurs and pic just repeat the same line # multiple times new_row['occurs'] = None for j in range(1, row["occurs"] + 1): row_to_add = new_row.copy() # First time occurs is just 1, we don't want to add # 1 after *every* field row_to_add["name"] = (row['name'] + new_name_postfix + '-' + str(j)) # + "-" + str(i) + "-" + str(j) if occurs > 1 else # row['name'] + name_postfix + "-" + str(j) output.append(row_to_add) else: # Get all the lines that have to occur occur_lines = get_subgroup(row['level'], lines[index + 1:]) # Calculate the new level difference that has to be applied new_level_diff = level_diff + row['level'] - occur_lines[ 0]['level'] output += handle_occurs(occur_lines, row['occurs'], new_level_diff, new_name_postfix) skip_till = index + len(occur_lines) + 1 return output def clean_names(lines, ensure_unique_names=False, strip_prefix=False, make_database_safe=False): """ Clean the names. Options to: - strip prefixes on names - enforce unique names - make database safe names by converting - to _ """ names = {} for row in lines: if strip_prefix: row['name'] = row['name'][row['name'].find('-') + 1:] if row['indexed_by'] is not None: row['indexed_by'] = row['indexed_by'][row['indexed_by'].find( '-') + 1:] if ensure_unique_names: i = 1 while (row['name'] if i == 1 else row['name'] + "-" + str(i)) in names: i += 1 names[row['name'] if i == 1 else row['name'] + "-" + str(i)] = 1 if i > 1: row['name'] = row['name'] + "-" + str(i) if make_database_safe: row['name'] = row['name'].replace("-", "_") return lines def process_cobol(lines, free_format): return clean_names(denormalize_cobol(parse_cobol(clean_cobol(lines, free_format))), ensure_unique_names=False, strip_prefix=False, make_database_safe=False)
pyQode/pyqode.cobol
pyqode/cobol/api/parsers/pic.py
clean_names
python
def clean_names(lines, ensure_unique_names=False, strip_prefix=False, make_database_safe=False): names = {} for row in lines: if strip_prefix: row['name'] = row['name'][row['name'].find('-') + 1:] if row['indexed_by'] is not None: row['indexed_by'] = row['indexed_by'][row['indexed_by'].find( '-') + 1:] if ensure_unique_names: i = 1 while (row['name'] if i == 1 else row['name'] + "-" + str(i)) in names: i += 1 names[row['name'] if i == 1 else row['name'] + "-" + str(i)] = 1 if i > 1: row['name'] = row['name'] + "-" + str(i) if make_database_safe: row['name'] = row['name'].replace("-", "_") return lines
Clean the names. Options to: - strip prefixes on names - enforce unique names - make database safe names by converting - to _
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/api/parsers/pic.py#L230-L263
null
""" This module contains the Python-Cobol pic parser written by Brian Peterson (https://github.com/bpeterso2000/pycobol ) and enhanced by Paulus Schoutsen (https://github.com/balloob/Python-COBOL/blob/master/cobol.py ), licensed under the GPL v3. We slightly modified this module to: - add python3 support - use the pyqode logger mechanism - fix a few PEP8 issues. """ import logging import re def _logger(): return logging.getLogger(__name__) class CobolPatterns: opt_pattern_format = "({})?" row_pattern_base = r'^(?P<level>\d{2})\s*(?P<name>\S+)' row_pattern_redefines = r"\s+REDEFINES\s(?P<redefines>\S+)" row_pattern_pic = r'\s+(COMP-\d\s+)?PIC\s+(?P<pic>\S+)' row_pattern_occurs = r'\s+OCCURS (?P<occurs>\d+) TIMES' row_pattern_indexed_by = r"\s+INDEXED BY\s(?P<indexed_by>\S+)" row_pattern_end = r'\.$' row_pattern = re.compile( row_pattern_base + opt_pattern_format.format(row_pattern_redefines) + opt_pattern_format.format(row_pattern_pic) + opt_pattern_format.format(row_pattern_occurs) + opt_pattern_format.format(row_pattern_indexed_by) + row_pattern_end, re.IGNORECASE ) pic_pattern_repeats = re.compile(r'(.)\((\d+)\)') pic_pattern_float = re.compile(r'S?[9Z]*[.V][9Z]+') pic_pattern_integer = re.compile(r'S?[9Z]+') # Parse the pic string def parse_pic_string(pic_str): # Expand repeating chars while True: match = CobolPatterns.pic_pattern_repeats.search(pic_str) if not match: break expanded_str = match.group(1) * int(match.group(2)) pic_str = CobolPatterns.pic_pattern_repeats.sub( expanded_str, pic_str, 1) # Match to types if CobolPatterns.pic_pattern_float.match(pic_str): data_type = 'Float' elif CobolPatterns.pic_pattern_integer.match(pic_str): data_type = 'Integer' else: data_type = 'Char' # Handle signed if pic_str[0] == "S": data_type = "Signed " + data_type # Handle precision decimal_pos = 0 if 'V' in pic_str: decimal_pos = len(pic_str[pic_str.index('V') + 1:]) pic_str = pic_str.replace('V', '') return { 'type': data_type, 'length': len(pic_str), 'precision': decimal_pos } # Cleans the COBOL by converting the cobol information to single lines def clean_cobol(lines, free_format): holder = [] output = [] for row in lines: if not free_format: row = row[6:72].rstrip() if row == "" or row[0] in ('*', '/'): continue holder.append(row if len(holder) == 0 else row.strip()) if row[-1] == ".": output.append(" ".join(holder)) holder = [] if len(holder) > 0: _logger().warning( "Probably invalid COBOL - found unfinished line: %s" % " ".join(holder)) return output def parse_cobol(lines): """ Parses the COBOL - converts the COBOL line into a dictionary containing the information - parses the pic information into type, length, precision - ~~handles redefines~~ -> our implementation does not do that anymore because we want to display item that was redefined. """ output = [] intify = ["level", "occurs"] # All in 1 line now, let's parse for row in lines: match = CobolPatterns.row_pattern.match(row.strip()) if not match: _logger().warning("Found unmatched row %s" % row.strip()) continue match = match.groupdict() for i in intify: match[i] = int(match[i]) if match[i] is not None else None if match['pic'] is not None: match['pic_info'] = parse_pic_string(match['pic']) output.append(match) return output # Helper function # Gets all the lines that have a higher level then the parent_level until # a line with equal or lower level then parent_level is encountered def get_subgroup(parent_level, lines): output = [] for row in lines: if row["level"] > parent_level: output.append(row) else: return output return output def denormalize_cobol(lines): return handle_occurs(lines, 1) # Helper function # Will go ahead and denormalize the COBOL # Beacuse the OCCURS are removed the INDEXED BY will also be removed def handle_occurs(lines, occurs, level_diff=0, name_postfix=""): output = [] for i in range(1, occurs + 1): skip_till = 0 new_name_postfix = (name_postfix if occurs == 1 else name_postfix + '-' + str(i)) for index, row in enumerate(lines): if index < skip_till: continue new_row = row.copy() new_row['level'] += level_diff # Not needed when flattened new_row['indexed_by'] = None if row['occurs'] is None: # First time occurs is just 1, we don't want to add _1 after # *every* field new_row['name'] = row['name'] + new_name_postfix # + "-" + str(i) if occurs > 1 else row['name'] + name_postfix output.append(new_row) else: if row["pic"] is not None: # If it has occurs and pic just repeat the same line # multiple times new_row['occurs'] = None for j in range(1, row["occurs"] + 1): row_to_add = new_row.copy() # First time occurs is just 1, we don't want to add # 1 after *every* field row_to_add["name"] = (row['name'] + new_name_postfix + '-' + str(j)) # + "-" + str(i) + "-" + str(j) if occurs > 1 else # row['name'] + name_postfix + "-" + str(j) output.append(row_to_add) else: # Get all the lines that have to occur occur_lines = get_subgroup(row['level'], lines[index + 1:]) # Calculate the new level difference that has to be applied new_level_diff = level_diff + row['level'] - occur_lines[ 0]['level'] output += handle_occurs(occur_lines, row['occurs'], new_level_diff, new_name_postfix) skip_till = index + len(occur_lines) + 1 return output def process_cobol(lines, free_format): return clean_names(denormalize_cobol(parse_cobol(clean_cobol(lines, free_format))), ensure_unique_names=False, strip_prefix=False, make_database_safe=False)
pyQode/pyqode.cobol
pyqode/cobol/modes/margins.py
MarginsMode.on_state_changed
python
def on_state_changed(self, state): if state: self.editor.painted.connect(self._paint_margins) self.editor.key_pressed.connect(self._on_key_pressed) self.editor.repaint() else: self.editor.painted.disconnect(self._paint_margins) self.editor.key_pressed.disconnect(self._on_key_pressed) self.editor.repaint()
Connects/Disconnects to the painted event of the editor. :param state: Enable state
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/modes/margins.py#L63-L76
null
class MarginsMode(Mode): """ Displays four margins at the specified columns. """ @property def colors(self): """ Gets/sets the colors of the 4 margins """ return self._colors @colors.setter def colors(self, value): self._colors = value self._pens = [QtGui.QPen(c) for c in value] TextHelper(self.editor).mark_whole_doc_dirty() self.editor.repaint() if self.editor: for clone in self.editor.clones: try: clone.modes.get(self.__class__).colors = value except KeyError: # this should never happen since we're working with clones pass @property def positions(self): """ Gets/sets the 4 position of the margins (tuple made up from 4 integers). The positions are 0 based (use 0 for column 1,...). """ return self._positions @positions.setter def positions(self, value): self._positions = value if self.editor: for clone in self.editor.clones: try: clone.modes.get(self.__class__).position = value except KeyError: # this should never happen since we're working with clones pass def __init__(self): super(MarginsMode, self).__init__() self._positions = [7, 11, 72, 79] self._colors = [QtGui.QColor('red') for _ in range(4)] self._pens = [QtGui.QPen(c) for c in self._colors] def _paint_margins(self, event): """ Paints the right margin after editor paint event. """ font = QtGui.QFont(self.editor.font_name, self.editor.font_size + self.editor.zoom_level) metrics = QtGui.QFontMetricsF(font) painter = QtGui.QPainter(self.editor.viewport()) for pos, pen in zip(self._positions, self._pens): if pos < 0: # margin not visible continue offset = self.editor.contentOffset().x() + \ self.editor.document().documentMargin() x_pos = round(metrics.width(' ') * pos) + offset painter.setPen(pen) painter.drawLine(x_pos, 0, x_pos, 2 ** 16) def _on_key_pressed(self, event): Qt = QtCore.Qt modifier_match = int(event.modifiers()) == \ Qt.ControlModifier + Qt.AltModifier if event.key() == Qt.Key_Left and modifier_match: self._go_to_previous_margin() event.accept() elif event.key() == Qt.Key_Right and modifier_match: self._go_to_next_margin() event.accept() def _go_to_next_margin(self): tc = self.editor.textCursor() max_column = len(tc.block().text()) current_column = TextHelper(self.editor).current_column_nbr() next_column = max_column for p in self._positions: if p < 0: continue if p > current_column: next_column = p if next_column > max_column: next_column = max_column break tc.movePosition(tc.Right, tc.MoveAnchor, next_column - current_column) self.editor.setTextCursor(tc) def _go_to_previous_margin(self): tc = self.editor.textCursor() min_column = 1 current_column = TextHelper(self.editor).current_column_nbr() next_column = min_column for p in reversed(self._positions): if p < 0: continue if p < current_column: next_column = p if next_column < min_column: next_column = min_column break tc.movePosition(tc.Left, tc.MoveAnchor, current_column - next_column) self.editor.setTextCursor(tc) def clone_settings(self, original): self.colors = original.colors self.positions = original.positions
pyQode/pyqode.cobol
pyqode/cobol/modes/margins.py
MarginsMode._paint_margins
python
def _paint_margins(self, event): font = QtGui.QFont(self.editor.font_name, self.editor.font_size + self.editor.zoom_level) metrics = QtGui.QFontMetricsF(font) painter = QtGui.QPainter(self.editor.viewport()) for pos, pen in zip(self._positions, self._pens): if pos < 0: # margin not visible continue offset = self.editor.contentOffset().x() + \ self.editor.document().documentMargin() x_pos = round(metrics.width(' ') * pos) + offset painter.setPen(pen) painter.drawLine(x_pos, 0, x_pos, 2 ** 16)
Paints the right margin after editor paint event.
train
https://github.com/pyQode/pyqode.cobol/blob/eedae4e320a4b2d0c44abb2c3061091321648fb7/pyqode/cobol/modes/margins.py#L78-L92
null
class MarginsMode(Mode): """ Displays four margins at the specified columns. """ @property def colors(self): """ Gets/sets the colors of the 4 margins """ return self._colors @colors.setter def colors(self, value): self._colors = value self._pens = [QtGui.QPen(c) for c in value] TextHelper(self.editor).mark_whole_doc_dirty() self.editor.repaint() if self.editor: for clone in self.editor.clones: try: clone.modes.get(self.__class__).colors = value except KeyError: # this should never happen since we're working with clones pass @property def positions(self): """ Gets/sets the 4 position of the margins (tuple made up from 4 integers). The positions are 0 based (use 0 for column 1,...). """ return self._positions @positions.setter def positions(self, value): self._positions = value if self.editor: for clone in self.editor.clones: try: clone.modes.get(self.__class__).position = value except KeyError: # this should never happen since we're working with clones pass def __init__(self): super(MarginsMode, self).__init__() self._positions = [7, 11, 72, 79] self._colors = [QtGui.QColor('red') for _ in range(4)] self._pens = [QtGui.QPen(c) for c in self._colors] def on_state_changed(self, state): """ Connects/Disconnects to the painted event of the editor. :param state: Enable state """ if state: self.editor.painted.connect(self._paint_margins) self.editor.key_pressed.connect(self._on_key_pressed) self.editor.repaint() else: self.editor.painted.disconnect(self._paint_margins) self.editor.key_pressed.disconnect(self._on_key_pressed) self.editor.repaint() def _on_key_pressed(self, event): Qt = QtCore.Qt modifier_match = int(event.modifiers()) == \ Qt.ControlModifier + Qt.AltModifier if event.key() == Qt.Key_Left and modifier_match: self._go_to_previous_margin() event.accept() elif event.key() == Qt.Key_Right and modifier_match: self._go_to_next_margin() event.accept() def _go_to_next_margin(self): tc = self.editor.textCursor() max_column = len(tc.block().text()) current_column = TextHelper(self.editor).current_column_nbr() next_column = max_column for p in self._positions: if p < 0: continue if p > current_column: next_column = p if next_column > max_column: next_column = max_column break tc.movePosition(tc.Right, tc.MoveAnchor, next_column - current_column) self.editor.setTextCursor(tc) def _go_to_previous_margin(self): tc = self.editor.textCursor() min_column = 1 current_column = TextHelper(self.editor).current_column_nbr() next_column = min_column for p in reversed(self._positions): if p < 0: continue if p < current_column: next_column = p if next_column < min_column: next_column = min_column break tc.movePosition(tc.Left, tc.MoveAnchor, current_column - next_column) self.editor.setTextCursor(tc) def clone_settings(self, original): self.colors = original.colors self.positions = original.positions