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
crm416/semantic
semantic/solver.py
MathService._calculate
python
def _calculate(numbers, symbols): if len(numbers) is 1: return numbers[0] precedence = [[pow], [mul, div], [add, sub]] # Find most important operation for op_group in precedence: for i, op in enumerate(symbols): if op in op_group: # Apply operation a = numbers[i] b = numbers[i + 1] result = MathService._applyBinary(a, b, op) new_numbers = numbers[:i] + [result] + numbers[i + 2:] new_symbols = symbols[:i] + symbols[i + 1:] return MathService._calculate(new_numbers, new_symbols)
Calculates a final value given a set of numbers and symbols.
train
https://github.com/crm416/semantic/blob/46deb8fefb3ea58aad2fedc8d0d62f3ee254b8fe/semantic/solver.py#L126-L144
[ "def _applyBinary(a, b, op):\n a = float(a)\n b = float(b)\n return op(a, b)\n", "def _calculate(numbers, symbols):\n \"\"\"Calculates a final value given a set of numbers and symbols.\"\"\"\n if len(numbers) is 1:\n return numbers[0]\n\n precedence = [[pow], [mul, div], [add, sub]]\n\n # Find most important operation\n for op_group in precedence:\n for i, op in enumerate(symbols):\n if op in op_group:\n # Apply operation\n a = numbers[i]\n b = numbers[i + 1]\n result = MathService._applyBinary(a, b, op)\n new_numbers = numbers[:i] + [result] + numbers[i + 2:]\n new_symbols = symbols[:i] + symbols[i + 1:]\n\n return MathService._calculate(new_numbers, new_symbols)\n" ]
class MathService(object): __constants__ = { 'e': e, 'E': e, 'EE': e, 'pi': pi, 'pie': pi } __unaryOperators__ = { 'log': log, 'sine': sin, 'sin': sin, 'cosine': cos, 'cos': cos, 'tan': tan, 'tangent': tan, 'arcsine': asin, 'arcsin': asin, 'asin': asin, 'arccosine': acos, 'arccos': acos, 'acos': acos, 'arctanget': atan, 'arctan': atan, 'atan': atan, 'sqrt': sqrt } __binaryOperators__ = { 'plus': add, 'add': add, 'sum': add, 'minus': sub, 'sub': sub, 'subtract': sub, 'less': sub, 'over': div, 'divide': div, 'times': mul, 'multiply': mul, 'to': pow } @staticmethod def _applyBinary(a, b, op): a = float(a) b = float(b) return op(a, b) @staticmethod def _applyUnary(a, op): a = float(a) return op(a) @staticmethod def _preprocess(inp): """Revise wording to match canonical and expected forms.""" inp = re.sub(r'(\b)a(\b)', r'\g<1>one\g<2>', inp) inp = re.sub(r'to the (.*) power', r'to \g<1>', inp) inp = re.sub(r'to the (.*?)(\b)', r'to \g<1>\g<2>', inp) inp = re.sub(r'log of', r'log', inp) inp = re.sub(r'(square )?root( of)?', r'sqrt', inp) inp = re.sub(r'squared', r'to two', inp) inp = re.sub(r'cubed', r'to three', inp) inp = re.sub(r'divided?( by)?', r'divide', inp) inp = re.sub(r'(\b)over(\b)', r'\g<1>divide\g<2>', inp) inp = re.sub(r'(\b)EE(\b)', r'\g<1>e\g<2>', inp) inp = re.sub(r'(\b)E(\b)', r'\g<1>e\g<2>', inp) inp = re.sub(r'(\b)pie(\b)', r'\g<1>pi\g<2>', inp) inp = re.sub(r'(\b)PI(\b)', r'\g<1>pi\g<2>', inp) def findImplicitMultiplications(inp): """Replace omitted 'times' references.""" def findConstantMultiplications(inp): split = inp.split(' ') revision = "" converter = NumberService() for i, w in enumerate(split): if i > 0 and w in MathService.__constants__: if converter.isValid(split[i - 1]): revision += " times" if not revision: revision = w else: revision += " " + w return revision def findUnaryMultiplications(inp): split = inp.split(' ') revision = "" for i, w in enumerate(split): if i > 0 and w in MathService.__unaryOperators__: last_op = split[i - 1] binary = last_op in MathService.__binaryOperators__ unary = last_op in MathService.__unaryOperators__ if last_op and not (binary or unary): revision += " times" if not revision: revision = w else: revision += " " + w return revision return findUnaryMultiplications(findConstantMultiplications(inp)) return findImplicitMultiplications(inp) @staticmethod def parseEquation(self, inp): """Solves the equation specified by the input string. Args: inp (str): An equation, specified in words, containing some combination of numbers, binary, and unary operations. Returns: The floating-point result of carrying out the computation. """ inp = MathService._preprocess(inp) split = inp.split(' ') # Recursive call on unary operators for i, w in enumerate(split): if w in self.__unaryOperators__: op = self.__unaryOperators__[w] # Split equation into halves eq1 = ' '.join(split[:i]) eq2 = ' '.join(split[i + 1:]) # Calculate second half result = MathService._applyUnary(self.parseEquation(eq2), op) return self.parseEquation(eq1 + " " + str(result)) def extractNumbersAndSymbols(inp): numbers = [] symbols = [] # Divide into values (numbers), operators (symbols) next_number = "" for w in inp.split(' '): if w in self.__binaryOperators__: symbols.append(self.__binaryOperators__[w]) if next_number: numbers.append(next_number) next_number = "" else: if next_number: next_number += " " next_number += w if next_number: numbers.append(next_number) # Cast numbers from words to integers def convert(n): if n in self.__constants__: return self.__constants__[n] converter = NumberService() return converter.parse(n) numbers = [convert(n) for n in numbers] return numbers, symbols numbers, symbols = extractNumbersAndSymbols(inp) return MathService._calculate(numbers, symbols)
crm416/semantic
semantic/solver.py
MathService.parseEquation
python
def parseEquation(self, inp): inp = MathService._preprocess(inp) split = inp.split(' ') # Recursive call on unary operators for i, w in enumerate(split): if w in self.__unaryOperators__: op = self.__unaryOperators__[w] # Split equation into halves eq1 = ' '.join(split[:i]) eq2 = ' '.join(split[i + 1:]) # Calculate second half result = MathService._applyUnary(self.parseEquation(eq2), op) return self.parseEquation(eq1 + " " + str(result)) def extractNumbersAndSymbols(inp): numbers = [] symbols = [] # Divide into values (numbers), operators (symbols) next_number = "" for w in inp.split(' '): if w in self.__binaryOperators__: symbols.append(self.__binaryOperators__[w]) if next_number: numbers.append(next_number) next_number = "" else: if next_number: next_number += " " next_number += w if next_number: numbers.append(next_number) # Cast numbers from words to integers def convert(n): if n in self.__constants__: return self.__constants__[n] converter = NumberService() return converter.parse(n) numbers = [convert(n) for n in numbers] return numbers, symbols numbers, symbols = extractNumbersAndSymbols(inp) return MathService._calculate(numbers, symbols)
Solves the equation specified by the input string. Args: inp (str): An equation, specified in words, containing some combination of numbers, binary, and unary operations. Returns: The floating-point result of carrying out the computation.
train
https://github.com/crm416/semantic/blob/46deb8fefb3ea58aad2fedc8d0d62f3ee254b8fe/semantic/solver.py#L146-L209
[ "def _applyUnary(a, op):\n a = float(a)\n return op(a)\n", "def _preprocess(inp):\n \"\"\"Revise wording to match canonical and expected forms.\"\"\"\n inp = re.sub(r'(\\b)a(\\b)', r'\\g<1>one\\g<2>', inp)\n inp = re.sub(r'to the (.*) power', r'to \\g<1>', inp)\n inp = re.sub(r'to the (.*?)(\\b)', r'to \\g<1>\\g<2>', inp)\n inp = re.sub(r'log of', r'log', inp)\n inp = re.sub(r'(square )?root( of)?', r'sqrt', inp)\n inp = re.sub(r'squared', r'to two', inp)\n inp = re.sub(r'cubed', r'to three', inp)\n inp = re.sub(r'divided?( by)?', r'divide', inp)\n inp = re.sub(r'(\\b)over(\\b)', r'\\g<1>divide\\g<2>', inp)\n inp = re.sub(r'(\\b)EE(\\b)', r'\\g<1>e\\g<2>', inp)\n inp = re.sub(r'(\\b)E(\\b)', r'\\g<1>e\\g<2>', inp)\n inp = re.sub(r'(\\b)pie(\\b)', r'\\g<1>pi\\g<2>', inp)\n inp = re.sub(r'(\\b)PI(\\b)', r'\\g<1>pi\\g<2>', inp)\n\n def findImplicitMultiplications(inp):\n \"\"\"Replace omitted 'times' references.\"\"\"\n\n def findConstantMultiplications(inp):\n split = inp.split(' ')\n revision = \"\"\n\n converter = NumberService()\n for i, w in enumerate(split):\n if i > 0 and w in MathService.__constants__:\n if converter.isValid(split[i - 1]):\n revision += \" times\"\n if not revision:\n revision = w\n else:\n revision += \" \" + w\n\n return revision\n\n def findUnaryMultiplications(inp):\n split = inp.split(' ')\n revision = \"\"\n\n for i, w in enumerate(split):\n if i > 0 and w in MathService.__unaryOperators__:\n last_op = split[i - 1]\n\n binary = last_op in MathService.__binaryOperators__\n unary = last_op in MathService.__unaryOperators__\n\n if last_op and not (binary or unary):\n revision += \" times\"\n if not revision:\n revision = w\n else:\n revision += \" \" + w\n\n return revision\n\n return findUnaryMultiplications(findConstantMultiplications(inp))\n\n return findImplicitMultiplications(inp)\n", "def _calculate(numbers, symbols):\n \"\"\"Calculates a final value given a set of numbers and symbols.\"\"\"\n if len(numbers) is 1:\n return numbers[0]\n\n precedence = [[pow], [mul, div], [add, sub]]\n\n # Find most important operation\n for op_group in precedence:\n for i, op in enumerate(symbols):\n if op in op_group:\n # Apply operation\n a = numbers[i]\n b = numbers[i + 1]\n result = MathService._applyBinary(a, b, op)\n new_numbers = numbers[:i] + [result] + numbers[i + 2:]\n new_symbols = symbols[:i] + symbols[i + 1:]\n\n return MathService._calculate(new_numbers, new_symbols)\n", "def parseEquation(self, inp):\n \"\"\"Solves the equation specified by the input string.\n\n Args:\n inp (str): An equation, specified in words, containing some\n combination of numbers, binary, and unary operations.\n\n Returns:\n The floating-point result of carrying out the computation.\n \"\"\"\n inp = MathService._preprocess(inp)\n split = inp.split(' ')\n\n # Recursive call on unary operators\n for i, w in enumerate(split):\n if w in self.__unaryOperators__:\n op = self.__unaryOperators__[w]\n\n # Split equation into halves\n eq1 = ' '.join(split[:i])\n eq2 = ' '.join(split[i + 1:])\n\n # Calculate second half\n result = MathService._applyUnary(self.parseEquation(eq2), op)\n\n return self.parseEquation(eq1 + \" \" + str(result))\n\n def extractNumbersAndSymbols(inp):\n numbers = []\n symbols = []\n\n # Divide into values (numbers), operators (symbols)\n next_number = \"\"\n for w in inp.split(' '):\n if w in self.__binaryOperators__:\n symbols.append(self.__binaryOperators__[w])\n\n if next_number:\n numbers.append(next_number)\n next_number = \"\"\n\n else:\n if next_number:\n next_number += \" \"\n next_number += w\n\n if next_number:\n numbers.append(next_number)\n\n # Cast numbers from words to integers\n def convert(n):\n if n in self.__constants__:\n return self.__constants__[n]\n\n converter = NumberService()\n return converter.parse(n)\n\n numbers = [convert(n) for n in numbers]\n\n return numbers, symbols\n\n numbers, symbols = extractNumbersAndSymbols(inp)\n\n return MathService._calculate(numbers, symbols)\n", "def extractNumbersAndSymbols(inp):\n numbers = []\n symbols = []\n\n # Divide into values (numbers), operators (symbols)\n next_number = \"\"\n for w in inp.split(' '):\n if w in self.__binaryOperators__:\n symbols.append(self.__binaryOperators__[w])\n\n if next_number:\n numbers.append(next_number)\n next_number = \"\"\n\n else:\n if next_number:\n next_number += \" \"\n next_number += w\n\n if next_number:\n numbers.append(next_number)\n\n # Cast numbers from words to integers\n def convert(n):\n if n in self.__constants__:\n return self.__constants__[n]\n\n converter = NumberService()\n return converter.parse(n)\n\n numbers = [convert(n) for n in numbers]\n\n return numbers, symbols\n" ]
class MathService(object): __constants__ = { 'e': e, 'E': e, 'EE': e, 'pi': pi, 'pie': pi } __unaryOperators__ = { 'log': log, 'sine': sin, 'sin': sin, 'cosine': cos, 'cos': cos, 'tan': tan, 'tangent': tan, 'arcsine': asin, 'arcsin': asin, 'asin': asin, 'arccosine': acos, 'arccos': acos, 'acos': acos, 'arctanget': atan, 'arctan': atan, 'atan': atan, 'sqrt': sqrt } __binaryOperators__ = { 'plus': add, 'add': add, 'sum': add, 'minus': sub, 'sub': sub, 'subtract': sub, 'less': sub, 'over': div, 'divide': div, 'times': mul, 'multiply': mul, 'to': pow } @staticmethod def _applyBinary(a, b, op): a = float(a) b = float(b) return op(a, b) @staticmethod def _applyUnary(a, op): a = float(a) return op(a) @staticmethod def _preprocess(inp): """Revise wording to match canonical and expected forms.""" inp = re.sub(r'(\b)a(\b)', r'\g<1>one\g<2>', inp) inp = re.sub(r'to the (.*) power', r'to \g<1>', inp) inp = re.sub(r'to the (.*?)(\b)', r'to \g<1>\g<2>', inp) inp = re.sub(r'log of', r'log', inp) inp = re.sub(r'(square )?root( of)?', r'sqrt', inp) inp = re.sub(r'squared', r'to two', inp) inp = re.sub(r'cubed', r'to three', inp) inp = re.sub(r'divided?( by)?', r'divide', inp) inp = re.sub(r'(\b)over(\b)', r'\g<1>divide\g<2>', inp) inp = re.sub(r'(\b)EE(\b)', r'\g<1>e\g<2>', inp) inp = re.sub(r'(\b)E(\b)', r'\g<1>e\g<2>', inp) inp = re.sub(r'(\b)pie(\b)', r'\g<1>pi\g<2>', inp) inp = re.sub(r'(\b)PI(\b)', r'\g<1>pi\g<2>', inp) def findImplicitMultiplications(inp): """Replace omitted 'times' references.""" def findConstantMultiplications(inp): split = inp.split(' ') revision = "" converter = NumberService() for i, w in enumerate(split): if i > 0 and w in MathService.__constants__: if converter.isValid(split[i - 1]): revision += " times" if not revision: revision = w else: revision += " " + w return revision def findUnaryMultiplications(inp): split = inp.split(' ') revision = "" for i, w in enumerate(split): if i > 0 and w in MathService.__unaryOperators__: last_op = split[i - 1] binary = last_op in MathService.__binaryOperators__ unary = last_op in MathService.__unaryOperators__ if last_op and not (binary or unary): revision += " times" if not revision: revision = w else: revision += " " + w return revision return findUnaryMultiplications(findConstantMultiplications(inp)) return findImplicitMultiplications(inp) @staticmethod def _calculate(numbers, symbols): """Calculates a final value given a set of numbers and symbols.""" if len(numbers) is 1: return numbers[0] precedence = [[pow], [mul, div], [add, sub]] # Find most important operation for op_group in precedence: for i, op in enumerate(symbols): if op in op_group: # Apply operation a = numbers[i] b = numbers[i + 1] result = MathService._applyBinary(a, b, op) new_numbers = numbers[:i] + [result] + numbers[i + 2:] new_symbols = symbols[:i] + symbols[i + 1:] return MathService._calculate(new_numbers, new_symbols)
crm416/semantic
semantic/numbers.py
NumberService.parse
python
def parse(self, words): def exact(words): """If already represented as float or int, convert.""" try: return float(words) except: return None guess = exact(words) if guess is not None: return guess split = words.split(' ') # Replace final ordinal/fraction with number if split[-1] in self.__fractions__: split[-1] = self.__fractions__[split[-1]] elif split[-1] in self.__ordinals__: split[-1] = self.__ordinals__[split[-1]] parsed_ordinals = ' '.join(split) return self.parseFloat(parsed_ordinals)
A general method for parsing word-representations of numbers. Supports floats and integers. Args: words (str): Description of an arbitrary number. Returns: A double representation of the words.
train
https://github.com/crm416/semantic/blob/46deb8fefb3ea58aad2fedc8d0d62f3ee254b8fe/semantic/numbers.py#L91-L122
[ "def parseFloat(self, words):\n \"\"\"Convert a floating-point number described in words to a double.\n\n Supports two kinds of descriptions: those with a 'point' (e.g.,\n \"one point two five\") and those with a fraction (e.g., \"one and\n a quarter\").\n\n Args:\n words (str): Description of the floating-point number.\n\n Returns:\n A double representation of the words.\n \"\"\"\n def pointFloat(words):\n m = re.search(r'(.*) point (.*)', words)\n if m:\n whole = m.group(1)\n frac = m.group(2)\n total = 0.0\n coeff = 0.10\n for digit in frac.split(' '):\n total += coeff * self.parse(digit)\n coeff /= 10.0\n\n return self.parseInt(whole) + total\n return None\n\n def fractionFloat(words):\n m = re.search(r'(.*) and (.*)', words)\n if m:\n whole = self.parseInt(m.group(1))\n frac = m.group(2)\n\n # Replace plurals\n frac = re.sub(r'(\\w+)s(\\b)', '\\g<1>\\g<2>', frac)\n\n # Convert 'a' to 'one' (e.g., 'a third' to 'one third')\n frac = re.sub(r'(\\b)a(\\b)', '\\g<1>one\\g<2>', frac)\n\n split = frac.split(' ')\n\n # Split fraction into num (regular integer), denom (ordinal)\n num = split[:1]\n denom = split[1:]\n\n while denom:\n try:\n # Test for valid num, denom\n num_value = self.parse(' '.join(num))\n denom_value = self.parse(' '.join(denom))\n return whole + float(num_value) / denom_value\n except:\n # Add another word to num\n num += denom[:1]\n denom = denom[1:]\n return None\n\n # Extract \"one point two five\"-type float\n result = pointFloat(words)\n if result:\n return result\n\n # Extract \"one and a quarter\"-type float\n result = fractionFloat(words)\n if result:\n return result\n\n # Parse as integer\n return self.parseInt(words)\n", "def exact(words):\n \"\"\"If already represented as float or int, convert.\"\"\"\n try:\n return float(words)\n except:\n return None\n" ]
class NumberService(object): __small__ = { 'zero': 0, 'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6, 'seven': 7, 'eight': 8, 'nine': 9, 'ten': 10, 'eleven': 11, 'twelve': 12, 'thirteen': 13, 'fourteen': 14, 'fifteen': 15, 'sixteen': 16, 'seventeen': 17, 'eighteen': 18, 'nineteen': 19, 'twenty': 20, 'thirty': 30, 'forty': 40, 'fifty': 50, 'sixty': 60, 'seventy': 70, 'eighty': 80, 'ninety': 90 } __magnitude__ = { 'thousand': 1000, 'million': 1000000, 'billion': 1000000000, 'trillion': 1000000000000, 'quadrillion': 1000000000000000, 'quintillion': 1000000000000000000, 'sextillion': 1000000000000000000000, 'septillion': 1000000000000000000000000, 'octillion': 1000000000000000000000000000, 'nonillion': 1000000000000000000000000000000, 'decillion': 1000000000000000000000000000000000, } __ordinals__ = { 'first': 'one', 'second': 'two', 'third': 'three', 'fourth': 'four', 'fifth': 'five', 'sixth': 'six', 'seventh': 'seventh', 'eighth': 'eight', 'ninth': 'nine', 'tenth': 'ten', 'eleventh': 'eleven', 'twelth': 'twelve', 'thirteenth': 'thirteen', 'fifteenth': 'fifteen', 'sixteenth': 'sixteen', 'seventeenth': 'seventeen', 'eighteenth': 'eighteen', 'nineteenth': 'nineteen', 'twentieth': 'twenty', 'thirtieth': 'thirty', 'fortieth': 'forty', 'fiftieth': 'fifty', 'sixtieth': 'sixty', 'seventieth': 'seventy', 'eightieth': 'eighty', 'ninetieth': 'ninety', 'hundredth': 'hundred' } __fractions__ = { 'quarter': 'four', 'half': 'two', 'halve': 'two' } class NumberException(Exception): def __init__(self, msg): Exception.__init__(self, msg) def parseFloat(self, words): """Convert a floating-point number described in words to a double. Supports two kinds of descriptions: those with a 'point' (e.g., "one point two five") and those with a fraction (e.g., "one and a quarter"). Args: words (str): Description of the floating-point number. Returns: A double representation of the words. """ def pointFloat(words): m = re.search(r'(.*) point (.*)', words) if m: whole = m.group(1) frac = m.group(2) total = 0.0 coeff = 0.10 for digit in frac.split(' '): total += coeff * self.parse(digit) coeff /= 10.0 return self.parseInt(whole) + total return None def fractionFloat(words): m = re.search(r'(.*) and (.*)', words) if m: whole = self.parseInt(m.group(1)) frac = m.group(2) # Replace plurals frac = re.sub(r'(\w+)s(\b)', '\g<1>\g<2>', frac) # Convert 'a' to 'one' (e.g., 'a third' to 'one third') frac = re.sub(r'(\b)a(\b)', '\g<1>one\g<2>', frac) split = frac.split(' ') # Split fraction into num (regular integer), denom (ordinal) num = split[:1] denom = split[1:] while denom: try: # Test for valid num, denom num_value = self.parse(' '.join(num)) denom_value = self.parse(' '.join(denom)) return whole + float(num_value) / denom_value except: # Add another word to num num += denom[:1] denom = denom[1:] return None # Extract "one point two five"-type float result = pointFloat(words) if result: return result # Extract "one and a quarter"-type float result = fractionFloat(words) if result: return result # Parse as integer return self.parseInt(words) def parseInt(self, words): """Parses words to the integer they describe. Args: words (str): Description of the integer. Returns: An integer representation of the words. """ # Remove 'and', case-sensitivity words = words.replace(" and ", " ").lower() # 'a' -> 'one' words = re.sub(r'(\b)a(\b)', '\g<1>one\g<2>', words) def textToNumber(s): """ Converts raw number string to an integer. Based on text2num.py by Greg Hewill. """ a = re.split(r"[\s-]+", s) n = 0 g = 0 for w in a: x = NumberService.__small__.get(w, None) if x is not None: g += x elif w == "hundred": g *= 100 else: x = NumberService.__magnitude__.get(w, None) if x is not None: n += g * x g = 0 else: raise NumberService.NumberException( "Unknown number: " + w) return n + g return textToNumber(words) def isValid(self, inp): try: self.parse(inp) return True except: return False @staticmethod def parseMagnitude(m): """Parses a number m into a human-ready string representation. For example, crops off floats if they're too accurate. Arguments: m (float): Floating-point number to be cleaned. Returns: Human-ready string description of the number. """ m = NumberService().parse(m) def toDecimalPrecision(n, k): return float("%.*f" % (k, round(n, k))) # Cast to two digits of precision digits = 2 magnitude = toDecimalPrecision(m, digits) # If value is really small, keep going while not magnitude: digits += 1 magnitude = toDecimalPrecision(m, digits) # If item is less than one, go one beyond 'necessary' number of digits if m < 1.0: magnitude = toDecimalPrecision(m, digits + 1) # Ignore decimal accuracy if irrelevant if int(magnitude) == magnitude: magnitude = int(magnitude) # Adjust for scientific notation magString = str(magnitude) magString = re.sub(r'(\d)e-(\d+)', '\g<1> times ten to the negative \g<2>', magString) magString = re.sub(r'(\d)e\+(\d+)', '\g<1> times ten to the \g<2>', magString) magString = re.sub(r'-(\d+)', 'negative \g<1>', magString) magString = re.sub(r'\b0(\d+)', '\g<1>', magString) return magString def longestNumber(self, inp): """Extracts the longest valid numerical description from a string. Not guaranteed to return a result even if some valid numerical description exists (i.e., method is not particularly advanced). Args: inp (str): An arbitrary string, hopefully containing a number. Returns: The number with the longest string description in input, or None if not found. """ split = inp.split(' ') # Assume just a single number numStart = None numEnd = None for i, w in enumerate(split): if self.isValid(w): if numStart is None: numStart = i numEnd = i else: # Check for ordinal, which would signify end w = re.sub(r'(\w+)s(\b)', '\g<1>\g<2>', w) if w in self.__ordinals__: if self.isValid(' '.join(split[numStart:i + 1])): numEnd = i break description = ' '.join(split[numStart:numEnd + 1]) return self.parse(description)
crm416/semantic
semantic/numbers.py
NumberService.parseFloat
python
def parseFloat(self, words): def pointFloat(words): m = re.search(r'(.*) point (.*)', words) if m: whole = m.group(1) frac = m.group(2) total = 0.0 coeff = 0.10 for digit in frac.split(' '): total += coeff * self.parse(digit) coeff /= 10.0 return self.parseInt(whole) + total return None def fractionFloat(words): m = re.search(r'(.*) and (.*)', words) if m: whole = self.parseInt(m.group(1)) frac = m.group(2) # Replace plurals frac = re.sub(r'(\w+)s(\b)', '\g<1>\g<2>', frac) # Convert 'a' to 'one' (e.g., 'a third' to 'one third') frac = re.sub(r'(\b)a(\b)', '\g<1>one\g<2>', frac) split = frac.split(' ') # Split fraction into num (regular integer), denom (ordinal) num = split[:1] denom = split[1:] while denom: try: # Test for valid num, denom num_value = self.parse(' '.join(num)) denom_value = self.parse(' '.join(denom)) return whole + float(num_value) / denom_value except: # Add another word to num num += denom[:1] denom = denom[1:] return None # Extract "one point two five"-type float result = pointFloat(words) if result: return result # Extract "one and a quarter"-type float result = fractionFloat(words) if result: return result # Parse as integer return self.parseInt(words)
Convert a floating-point number described in words to a double. Supports two kinds of descriptions: those with a 'point' (e.g., "one point two five") and those with a fraction (e.g., "one and a quarter"). Args: words (str): Description of the floating-point number. Returns: A double representation of the words.
train
https://github.com/crm416/semantic/blob/46deb8fefb3ea58aad2fedc8d0d62f3ee254b8fe/semantic/numbers.py#L124-L192
[ "def parseInt(self, words):\n \"\"\"Parses words to the integer they describe.\n\n Args:\n words (str): Description of the integer.\n\n Returns:\n An integer representation of the words.\n \"\"\"\n # Remove 'and', case-sensitivity\n words = words.replace(\" and \", \" \").lower()\n # 'a' -> 'one'\n words = re.sub(r'(\\b)a(\\b)', '\\g<1>one\\g<2>', words)\n\n def textToNumber(s):\n \"\"\"\n Converts raw number string to an integer.\n Based on text2num.py by Greg Hewill.\n \"\"\"\n a = re.split(r\"[\\s-]+\", s)\n n = 0\n g = 0\n for w in a:\n x = NumberService.__small__.get(w, None)\n if x is not None:\n g += x\n elif w == \"hundred\":\n g *= 100\n else:\n x = NumberService.__magnitude__.get(w, None)\n if x is not None:\n n += g * x\n g = 0\n else:\n raise NumberService.NumberException(\n \"Unknown number: \" + w)\n return n + g\n\n return textToNumber(words)\n", "def pointFloat(words):\n m = re.search(r'(.*) point (.*)', words)\n if m:\n whole = m.group(1)\n frac = m.group(2)\n total = 0.0\n coeff = 0.10\n for digit in frac.split(' '):\n total += coeff * self.parse(digit)\n coeff /= 10.0\n\n return self.parseInt(whole) + total\n return None\n", "def fractionFloat(words):\n m = re.search(r'(.*) and (.*)', words)\n if m:\n whole = self.parseInt(m.group(1))\n frac = m.group(2)\n\n # Replace plurals\n frac = re.sub(r'(\\w+)s(\\b)', '\\g<1>\\g<2>', frac)\n\n # Convert 'a' to 'one' (e.g., 'a third' to 'one third')\n frac = re.sub(r'(\\b)a(\\b)', '\\g<1>one\\g<2>', frac)\n\n split = frac.split(' ')\n\n # Split fraction into num (regular integer), denom (ordinal)\n num = split[:1]\n denom = split[1:]\n\n while denom:\n try:\n # Test for valid num, denom\n num_value = self.parse(' '.join(num))\n denom_value = self.parse(' '.join(denom))\n return whole + float(num_value) / denom_value\n except:\n # Add another word to num\n num += denom[:1]\n denom = denom[1:]\n return None\n" ]
class NumberService(object): __small__ = { 'zero': 0, 'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6, 'seven': 7, 'eight': 8, 'nine': 9, 'ten': 10, 'eleven': 11, 'twelve': 12, 'thirteen': 13, 'fourteen': 14, 'fifteen': 15, 'sixteen': 16, 'seventeen': 17, 'eighteen': 18, 'nineteen': 19, 'twenty': 20, 'thirty': 30, 'forty': 40, 'fifty': 50, 'sixty': 60, 'seventy': 70, 'eighty': 80, 'ninety': 90 } __magnitude__ = { 'thousand': 1000, 'million': 1000000, 'billion': 1000000000, 'trillion': 1000000000000, 'quadrillion': 1000000000000000, 'quintillion': 1000000000000000000, 'sextillion': 1000000000000000000000, 'septillion': 1000000000000000000000000, 'octillion': 1000000000000000000000000000, 'nonillion': 1000000000000000000000000000000, 'decillion': 1000000000000000000000000000000000, } __ordinals__ = { 'first': 'one', 'second': 'two', 'third': 'three', 'fourth': 'four', 'fifth': 'five', 'sixth': 'six', 'seventh': 'seventh', 'eighth': 'eight', 'ninth': 'nine', 'tenth': 'ten', 'eleventh': 'eleven', 'twelth': 'twelve', 'thirteenth': 'thirteen', 'fifteenth': 'fifteen', 'sixteenth': 'sixteen', 'seventeenth': 'seventeen', 'eighteenth': 'eighteen', 'nineteenth': 'nineteen', 'twentieth': 'twenty', 'thirtieth': 'thirty', 'fortieth': 'forty', 'fiftieth': 'fifty', 'sixtieth': 'sixty', 'seventieth': 'seventy', 'eightieth': 'eighty', 'ninetieth': 'ninety', 'hundredth': 'hundred' } __fractions__ = { 'quarter': 'four', 'half': 'two', 'halve': 'two' } class NumberException(Exception): def __init__(self, msg): Exception.__init__(self, msg) def parse(self, words): """A general method for parsing word-representations of numbers. Supports floats and integers. Args: words (str): Description of an arbitrary number. Returns: A double representation of the words. """ def exact(words): """If already represented as float or int, convert.""" try: return float(words) except: return None guess = exact(words) if guess is not None: return guess split = words.split(' ') # Replace final ordinal/fraction with number if split[-1] in self.__fractions__: split[-1] = self.__fractions__[split[-1]] elif split[-1] in self.__ordinals__: split[-1] = self.__ordinals__[split[-1]] parsed_ordinals = ' '.join(split) return self.parseFloat(parsed_ordinals) def parseInt(self, words): """Parses words to the integer they describe. Args: words (str): Description of the integer. Returns: An integer representation of the words. """ # Remove 'and', case-sensitivity words = words.replace(" and ", " ").lower() # 'a' -> 'one' words = re.sub(r'(\b)a(\b)', '\g<1>one\g<2>', words) def textToNumber(s): """ Converts raw number string to an integer. Based on text2num.py by Greg Hewill. """ a = re.split(r"[\s-]+", s) n = 0 g = 0 for w in a: x = NumberService.__small__.get(w, None) if x is not None: g += x elif w == "hundred": g *= 100 else: x = NumberService.__magnitude__.get(w, None) if x is not None: n += g * x g = 0 else: raise NumberService.NumberException( "Unknown number: " + w) return n + g return textToNumber(words) def isValid(self, inp): try: self.parse(inp) return True except: return False @staticmethod def parseMagnitude(m): """Parses a number m into a human-ready string representation. For example, crops off floats if they're too accurate. Arguments: m (float): Floating-point number to be cleaned. Returns: Human-ready string description of the number. """ m = NumberService().parse(m) def toDecimalPrecision(n, k): return float("%.*f" % (k, round(n, k))) # Cast to two digits of precision digits = 2 magnitude = toDecimalPrecision(m, digits) # If value is really small, keep going while not magnitude: digits += 1 magnitude = toDecimalPrecision(m, digits) # If item is less than one, go one beyond 'necessary' number of digits if m < 1.0: magnitude = toDecimalPrecision(m, digits + 1) # Ignore decimal accuracy if irrelevant if int(magnitude) == magnitude: magnitude = int(magnitude) # Adjust for scientific notation magString = str(magnitude) magString = re.sub(r'(\d)e-(\d+)', '\g<1> times ten to the negative \g<2>', magString) magString = re.sub(r'(\d)e\+(\d+)', '\g<1> times ten to the \g<2>', magString) magString = re.sub(r'-(\d+)', 'negative \g<1>', magString) magString = re.sub(r'\b0(\d+)', '\g<1>', magString) return magString def longestNumber(self, inp): """Extracts the longest valid numerical description from a string. Not guaranteed to return a result even if some valid numerical description exists (i.e., method is not particularly advanced). Args: inp (str): An arbitrary string, hopefully containing a number. Returns: The number with the longest string description in input, or None if not found. """ split = inp.split(' ') # Assume just a single number numStart = None numEnd = None for i, w in enumerate(split): if self.isValid(w): if numStart is None: numStart = i numEnd = i else: # Check for ordinal, which would signify end w = re.sub(r'(\w+)s(\b)', '\g<1>\g<2>', w) if w in self.__ordinals__: if self.isValid(' '.join(split[numStart:i + 1])): numEnd = i break description = ' '.join(split[numStart:numEnd + 1]) return self.parse(description)
crm416/semantic
semantic/numbers.py
NumberService.parseInt
python
def parseInt(self, words): # Remove 'and', case-sensitivity words = words.replace(" and ", " ").lower() # 'a' -> 'one' words = re.sub(r'(\b)a(\b)', '\g<1>one\g<2>', words) def textToNumber(s): """ Converts raw number string to an integer. Based on text2num.py by Greg Hewill. """ a = re.split(r"[\s-]+", s) n = 0 g = 0 for w in a: x = NumberService.__small__.get(w, None) if x is not None: g += x elif w == "hundred": g *= 100 else: x = NumberService.__magnitude__.get(w, None) if x is not None: n += g * x g = 0 else: raise NumberService.NumberException( "Unknown number: " + w) return n + g return textToNumber(words)
Parses words to the integer they describe. Args: words (str): Description of the integer. Returns: An integer representation of the words.
train
https://github.com/crm416/semantic/blob/46deb8fefb3ea58aad2fedc8d0d62f3ee254b8fe/semantic/numbers.py#L194-L232
[ "def textToNumber(s):\n \"\"\"\n Converts raw number string to an integer.\n Based on text2num.py by Greg Hewill.\n \"\"\"\n a = re.split(r\"[\\s-]+\", s)\n n = 0\n g = 0\n for w in a:\n x = NumberService.__small__.get(w, None)\n if x is not None:\n g += x\n elif w == \"hundred\":\n g *= 100\n else:\n x = NumberService.__magnitude__.get(w, None)\n if x is not None:\n n += g * x\n g = 0\n else:\n raise NumberService.NumberException(\n \"Unknown number: \" + w)\n return n + g\n" ]
class NumberService(object): __small__ = { 'zero': 0, 'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6, 'seven': 7, 'eight': 8, 'nine': 9, 'ten': 10, 'eleven': 11, 'twelve': 12, 'thirteen': 13, 'fourteen': 14, 'fifteen': 15, 'sixteen': 16, 'seventeen': 17, 'eighteen': 18, 'nineteen': 19, 'twenty': 20, 'thirty': 30, 'forty': 40, 'fifty': 50, 'sixty': 60, 'seventy': 70, 'eighty': 80, 'ninety': 90 } __magnitude__ = { 'thousand': 1000, 'million': 1000000, 'billion': 1000000000, 'trillion': 1000000000000, 'quadrillion': 1000000000000000, 'quintillion': 1000000000000000000, 'sextillion': 1000000000000000000000, 'septillion': 1000000000000000000000000, 'octillion': 1000000000000000000000000000, 'nonillion': 1000000000000000000000000000000, 'decillion': 1000000000000000000000000000000000, } __ordinals__ = { 'first': 'one', 'second': 'two', 'third': 'three', 'fourth': 'four', 'fifth': 'five', 'sixth': 'six', 'seventh': 'seventh', 'eighth': 'eight', 'ninth': 'nine', 'tenth': 'ten', 'eleventh': 'eleven', 'twelth': 'twelve', 'thirteenth': 'thirteen', 'fifteenth': 'fifteen', 'sixteenth': 'sixteen', 'seventeenth': 'seventeen', 'eighteenth': 'eighteen', 'nineteenth': 'nineteen', 'twentieth': 'twenty', 'thirtieth': 'thirty', 'fortieth': 'forty', 'fiftieth': 'fifty', 'sixtieth': 'sixty', 'seventieth': 'seventy', 'eightieth': 'eighty', 'ninetieth': 'ninety', 'hundredth': 'hundred' } __fractions__ = { 'quarter': 'four', 'half': 'two', 'halve': 'two' } class NumberException(Exception): def __init__(self, msg): Exception.__init__(self, msg) def parse(self, words): """A general method for parsing word-representations of numbers. Supports floats and integers. Args: words (str): Description of an arbitrary number. Returns: A double representation of the words. """ def exact(words): """If already represented as float or int, convert.""" try: return float(words) except: return None guess = exact(words) if guess is not None: return guess split = words.split(' ') # Replace final ordinal/fraction with number if split[-1] in self.__fractions__: split[-1] = self.__fractions__[split[-1]] elif split[-1] in self.__ordinals__: split[-1] = self.__ordinals__[split[-1]] parsed_ordinals = ' '.join(split) return self.parseFloat(parsed_ordinals) def parseFloat(self, words): """Convert a floating-point number described in words to a double. Supports two kinds of descriptions: those with a 'point' (e.g., "one point two five") and those with a fraction (e.g., "one and a quarter"). Args: words (str): Description of the floating-point number. Returns: A double representation of the words. """ def pointFloat(words): m = re.search(r'(.*) point (.*)', words) if m: whole = m.group(1) frac = m.group(2) total = 0.0 coeff = 0.10 for digit in frac.split(' '): total += coeff * self.parse(digit) coeff /= 10.0 return self.parseInt(whole) + total return None def fractionFloat(words): m = re.search(r'(.*) and (.*)', words) if m: whole = self.parseInt(m.group(1)) frac = m.group(2) # Replace plurals frac = re.sub(r'(\w+)s(\b)', '\g<1>\g<2>', frac) # Convert 'a' to 'one' (e.g., 'a third' to 'one third') frac = re.sub(r'(\b)a(\b)', '\g<1>one\g<2>', frac) split = frac.split(' ') # Split fraction into num (regular integer), denom (ordinal) num = split[:1] denom = split[1:] while denom: try: # Test for valid num, denom num_value = self.parse(' '.join(num)) denom_value = self.parse(' '.join(denom)) return whole + float(num_value) / denom_value except: # Add another word to num num += denom[:1] denom = denom[1:] return None # Extract "one point two five"-type float result = pointFloat(words) if result: return result # Extract "one and a quarter"-type float result = fractionFloat(words) if result: return result # Parse as integer return self.parseInt(words) def isValid(self, inp): try: self.parse(inp) return True except: return False @staticmethod def parseMagnitude(m): """Parses a number m into a human-ready string representation. For example, crops off floats if they're too accurate. Arguments: m (float): Floating-point number to be cleaned. Returns: Human-ready string description of the number. """ m = NumberService().parse(m) def toDecimalPrecision(n, k): return float("%.*f" % (k, round(n, k))) # Cast to two digits of precision digits = 2 magnitude = toDecimalPrecision(m, digits) # If value is really small, keep going while not magnitude: digits += 1 magnitude = toDecimalPrecision(m, digits) # If item is less than one, go one beyond 'necessary' number of digits if m < 1.0: magnitude = toDecimalPrecision(m, digits + 1) # Ignore decimal accuracy if irrelevant if int(magnitude) == magnitude: magnitude = int(magnitude) # Adjust for scientific notation magString = str(magnitude) magString = re.sub(r'(\d)e-(\d+)', '\g<1> times ten to the negative \g<2>', magString) magString = re.sub(r'(\d)e\+(\d+)', '\g<1> times ten to the \g<2>', magString) magString = re.sub(r'-(\d+)', 'negative \g<1>', magString) magString = re.sub(r'\b0(\d+)', '\g<1>', magString) return magString def longestNumber(self, inp): """Extracts the longest valid numerical description from a string. Not guaranteed to return a result even if some valid numerical description exists (i.e., method is not particularly advanced). Args: inp (str): An arbitrary string, hopefully containing a number. Returns: The number with the longest string description in input, or None if not found. """ split = inp.split(' ') # Assume just a single number numStart = None numEnd = None for i, w in enumerate(split): if self.isValid(w): if numStart is None: numStart = i numEnd = i else: # Check for ordinal, which would signify end w = re.sub(r'(\w+)s(\b)', '\g<1>\g<2>', w) if w in self.__ordinals__: if self.isValid(' '.join(split[numStart:i + 1])): numEnd = i break description = ' '.join(split[numStart:numEnd + 1]) return self.parse(description)
crm416/semantic
semantic/numbers.py
NumberService.parseMagnitude
python
def parseMagnitude(m): m = NumberService().parse(m) def toDecimalPrecision(n, k): return float("%.*f" % (k, round(n, k))) # Cast to two digits of precision digits = 2 magnitude = toDecimalPrecision(m, digits) # If value is really small, keep going while not magnitude: digits += 1 magnitude = toDecimalPrecision(m, digits) # If item is less than one, go one beyond 'necessary' number of digits if m < 1.0: magnitude = toDecimalPrecision(m, digits + 1) # Ignore decimal accuracy if irrelevant if int(magnitude) == magnitude: magnitude = int(magnitude) # Adjust for scientific notation magString = str(magnitude) magString = re.sub(r'(\d)e-(\d+)', '\g<1> times ten to the negative \g<2>', magString) magString = re.sub(r'(\d)e\+(\d+)', '\g<1> times ten to the \g<2>', magString) magString = re.sub(r'-(\d+)', 'negative \g<1>', magString) magString = re.sub(r'\b0(\d+)', '\g<1>', magString) return magString
Parses a number m into a human-ready string representation. For example, crops off floats if they're too accurate. Arguments: m (float): Floating-point number to be cleaned. Returns: Human-ready string description of the number.
train
https://github.com/crm416/semantic/blob/46deb8fefb3ea58aad2fedc8d0d62f3ee254b8fe/semantic/numbers.py#L242-L282
[ "def parse(self, words):\n \"\"\"A general method for parsing word-representations of numbers.\n Supports floats and integers.\n\n Args:\n words (str): Description of an arbitrary number.\n\n Returns:\n A double representation of the words.\n \"\"\"\n def exact(words):\n \"\"\"If already represented as float or int, convert.\"\"\"\n try:\n return float(words)\n except:\n return None\n\n guess = exact(words)\n if guess is not None:\n return guess\n\n split = words.split(' ')\n\n # Replace final ordinal/fraction with number\n if split[-1] in self.__fractions__:\n split[-1] = self.__fractions__[split[-1]]\n elif split[-1] in self.__ordinals__:\n split[-1] = self.__ordinals__[split[-1]]\n\n parsed_ordinals = ' '.join(split)\n\n return self.parseFloat(parsed_ordinals)\n", "def toDecimalPrecision(n, k):\n return float(\"%.*f\" % (k, round(n, k)))\n" ]
class NumberService(object): __small__ = { 'zero': 0, 'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6, 'seven': 7, 'eight': 8, 'nine': 9, 'ten': 10, 'eleven': 11, 'twelve': 12, 'thirteen': 13, 'fourteen': 14, 'fifteen': 15, 'sixteen': 16, 'seventeen': 17, 'eighteen': 18, 'nineteen': 19, 'twenty': 20, 'thirty': 30, 'forty': 40, 'fifty': 50, 'sixty': 60, 'seventy': 70, 'eighty': 80, 'ninety': 90 } __magnitude__ = { 'thousand': 1000, 'million': 1000000, 'billion': 1000000000, 'trillion': 1000000000000, 'quadrillion': 1000000000000000, 'quintillion': 1000000000000000000, 'sextillion': 1000000000000000000000, 'septillion': 1000000000000000000000000, 'octillion': 1000000000000000000000000000, 'nonillion': 1000000000000000000000000000000, 'decillion': 1000000000000000000000000000000000, } __ordinals__ = { 'first': 'one', 'second': 'two', 'third': 'three', 'fourth': 'four', 'fifth': 'five', 'sixth': 'six', 'seventh': 'seventh', 'eighth': 'eight', 'ninth': 'nine', 'tenth': 'ten', 'eleventh': 'eleven', 'twelth': 'twelve', 'thirteenth': 'thirteen', 'fifteenth': 'fifteen', 'sixteenth': 'sixteen', 'seventeenth': 'seventeen', 'eighteenth': 'eighteen', 'nineteenth': 'nineteen', 'twentieth': 'twenty', 'thirtieth': 'thirty', 'fortieth': 'forty', 'fiftieth': 'fifty', 'sixtieth': 'sixty', 'seventieth': 'seventy', 'eightieth': 'eighty', 'ninetieth': 'ninety', 'hundredth': 'hundred' } __fractions__ = { 'quarter': 'four', 'half': 'two', 'halve': 'two' } class NumberException(Exception): def __init__(self, msg): Exception.__init__(self, msg) def parse(self, words): """A general method for parsing word-representations of numbers. Supports floats and integers. Args: words (str): Description of an arbitrary number. Returns: A double representation of the words. """ def exact(words): """If already represented as float or int, convert.""" try: return float(words) except: return None guess = exact(words) if guess is not None: return guess split = words.split(' ') # Replace final ordinal/fraction with number if split[-1] in self.__fractions__: split[-1] = self.__fractions__[split[-1]] elif split[-1] in self.__ordinals__: split[-1] = self.__ordinals__[split[-1]] parsed_ordinals = ' '.join(split) return self.parseFloat(parsed_ordinals) def parseFloat(self, words): """Convert a floating-point number described in words to a double. Supports two kinds of descriptions: those with a 'point' (e.g., "one point two five") and those with a fraction (e.g., "one and a quarter"). Args: words (str): Description of the floating-point number. Returns: A double representation of the words. """ def pointFloat(words): m = re.search(r'(.*) point (.*)', words) if m: whole = m.group(1) frac = m.group(2) total = 0.0 coeff = 0.10 for digit in frac.split(' '): total += coeff * self.parse(digit) coeff /= 10.0 return self.parseInt(whole) + total return None def fractionFloat(words): m = re.search(r'(.*) and (.*)', words) if m: whole = self.parseInt(m.group(1)) frac = m.group(2) # Replace plurals frac = re.sub(r'(\w+)s(\b)', '\g<1>\g<2>', frac) # Convert 'a' to 'one' (e.g., 'a third' to 'one third') frac = re.sub(r'(\b)a(\b)', '\g<1>one\g<2>', frac) split = frac.split(' ') # Split fraction into num (regular integer), denom (ordinal) num = split[:1] denom = split[1:] while denom: try: # Test for valid num, denom num_value = self.parse(' '.join(num)) denom_value = self.parse(' '.join(denom)) return whole + float(num_value) / denom_value except: # Add another word to num num += denom[:1] denom = denom[1:] return None # Extract "one point two five"-type float result = pointFloat(words) if result: return result # Extract "one and a quarter"-type float result = fractionFloat(words) if result: return result # Parse as integer return self.parseInt(words) def parseInt(self, words): """Parses words to the integer they describe. Args: words (str): Description of the integer. Returns: An integer representation of the words. """ # Remove 'and', case-sensitivity words = words.replace(" and ", " ").lower() # 'a' -> 'one' words = re.sub(r'(\b)a(\b)', '\g<1>one\g<2>', words) def textToNumber(s): """ Converts raw number string to an integer. Based on text2num.py by Greg Hewill. """ a = re.split(r"[\s-]+", s) n = 0 g = 0 for w in a: x = NumberService.__small__.get(w, None) if x is not None: g += x elif w == "hundred": g *= 100 else: x = NumberService.__magnitude__.get(w, None) if x is not None: n += g * x g = 0 else: raise NumberService.NumberException( "Unknown number: " + w) return n + g return textToNumber(words) def isValid(self, inp): try: self.parse(inp) return True except: return False @staticmethod def longestNumber(self, inp): """Extracts the longest valid numerical description from a string. Not guaranteed to return a result even if some valid numerical description exists (i.e., method is not particularly advanced). Args: inp (str): An arbitrary string, hopefully containing a number. Returns: The number with the longest string description in input, or None if not found. """ split = inp.split(' ') # Assume just a single number numStart = None numEnd = None for i, w in enumerate(split): if self.isValid(w): if numStart is None: numStart = i numEnd = i else: # Check for ordinal, which would signify end w = re.sub(r'(\w+)s(\b)', '\g<1>\g<2>', w) if w in self.__ordinals__: if self.isValid(' '.join(split[numStart:i + 1])): numEnd = i break description = ' '.join(split[numStart:numEnd + 1]) return self.parse(description)
crm416/semantic
semantic/numbers.py
NumberService.longestNumber
python
def longestNumber(self, inp): split = inp.split(' ') # Assume just a single number numStart = None numEnd = None for i, w in enumerate(split): if self.isValid(w): if numStart is None: numStart = i numEnd = i else: # Check for ordinal, which would signify end w = re.sub(r'(\w+)s(\b)', '\g<1>\g<2>', w) if w in self.__ordinals__: if self.isValid(' '.join(split[numStart:i + 1])): numEnd = i break description = ' '.join(split[numStart:numEnd + 1]) return self.parse(description)
Extracts the longest valid numerical description from a string. Not guaranteed to return a result even if some valid numerical description exists (i.e., method is not particularly advanced). Args: inp (str): An arbitrary string, hopefully containing a number. Returns: The number with the longest string description in input, or None if not found.
train
https://github.com/crm416/semantic/blob/46deb8fefb3ea58aad2fedc8d0d62f3ee254b8fe/semantic/numbers.py#L284-L314
[ "def parse(self, words):\n \"\"\"A general method for parsing word-representations of numbers.\n Supports floats and integers.\n\n Args:\n words (str): Description of an arbitrary number.\n\n Returns:\n A double representation of the words.\n \"\"\"\n def exact(words):\n \"\"\"If already represented as float or int, convert.\"\"\"\n try:\n return float(words)\n except:\n return None\n\n guess = exact(words)\n if guess is not None:\n return guess\n\n split = words.split(' ')\n\n # Replace final ordinal/fraction with number\n if split[-1] in self.__fractions__:\n split[-1] = self.__fractions__[split[-1]]\n elif split[-1] in self.__ordinals__:\n split[-1] = self.__ordinals__[split[-1]]\n\n parsed_ordinals = ' '.join(split)\n\n return self.parseFloat(parsed_ordinals)\n", "def isValid(self, inp):\n try:\n self.parse(inp)\n return True\n except:\n return False\n" ]
class NumberService(object): __small__ = { 'zero': 0, 'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6, 'seven': 7, 'eight': 8, 'nine': 9, 'ten': 10, 'eleven': 11, 'twelve': 12, 'thirteen': 13, 'fourteen': 14, 'fifteen': 15, 'sixteen': 16, 'seventeen': 17, 'eighteen': 18, 'nineteen': 19, 'twenty': 20, 'thirty': 30, 'forty': 40, 'fifty': 50, 'sixty': 60, 'seventy': 70, 'eighty': 80, 'ninety': 90 } __magnitude__ = { 'thousand': 1000, 'million': 1000000, 'billion': 1000000000, 'trillion': 1000000000000, 'quadrillion': 1000000000000000, 'quintillion': 1000000000000000000, 'sextillion': 1000000000000000000000, 'septillion': 1000000000000000000000000, 'octillion': 1000000000000000000000000000, 'nonillion': 1000000000000000000000000000000, 'decillion': 1000000000000000000000000000000000, } __ordinals__ = { 'first': 'one', 'second': 'two', 'third': 'three', 'fourth': 'four', 'fifth': 'five', 'sixth': 'six', 'seventh': 'seventh', 'eighth': 'eight', 'ninth': 'nine', 'tenth': 'ten', 'eleventh': 'eleven', 'twelth': 'twelve', 'thirteenth': 'thirteen', 'fifteenth': 'fifteen', 'sixteenth': 'sixteen', 'seventeenth': 'seventeen', 'eighteenth': 'eighteen', 'nineteenth': 'nineteen', 'twentieth': 'twenty', 'thirtieth': 'thirty', 'fortieth': 'forty', 'fiftieth': 'fifty', 'sixtieth': 'sixty', 'seventieth': 'seventy', 'eightieth': 'eighty', 'ninetieth': 'ninety', 'hundredth': 'hundred' } __fractions__ = { 'quarter': 'four', 'half': 'two', 'halve': 'two' } class NumberException(Exception): def __init__(self, msg): Exception.__init__(self, msg) def parse(self, words): """A general method for parsing word-representations of numbers. Supports floats and integers. Args: words (str): Description of an arbitrary number. Returns: A double representation of the words. """ def exact(words): """If already represented as float or int, convert.""" try: return float(words) except: return None guess = exact(words) if guess is not None: return guess split = words.split(' ') # Replace final ordinal/fraction with number if split[-1] in self.__fractions__: split[-1] = self.__fractions__[split[-1]] elif split[-1] in self.__ordinals__: split[-1] = self.__ordinals__[split[-1]] parsed_ordinals = ' '.join(split) return self.parseFloat(parsed_ordinals) def parseFloat(self, words): """Convert a floating-point number described in words to a double. Supports two kinds of descriptions: those with a 'point' (e.g., "one point two five") and those with a fraction (e.g., "one and a quarter"). Args: words (str): Description of the floating-point number. Returns: A double representation of the words. """ def pointFloat(words): m = re.search(r'(.*) point (.*)', words) if m: whole = m.group(1) frac = m.group(2) total = 0.0 coeff = 0.10 for digit in frac.split(' '): total += coeff * self.parse(digit) coeff /= 10.0 return self.parseInt(whole) + total return None def fractionFloat(words): m = re.search(r'(.*) and (.*)', words) if m: whole = self.parseInt(m.group(1)) frac = m.group(2) # Replace plurals frac = re.sub(r'(\w+)s(\b)', '\g<1>\g<2>', frac) # Convert 'a' to 'one' (e.g., 'a third' to 'one third') frac = re.sub(r'(\b)a(\b)', '\g<1>one\g<2>', frac) split = frac.split(' ') # Split fraction into num (regular integer), denom (ordinal) num = split[:1] denom = split[1:] while denom: try: # Test for valid num, denom num_value = self.parse(' '.join(num)) denom_value = self.parse(' '.join(denom)) return whole + float(num_value) / denom_value except: # Add another word to num num += denom[:1] denom = denom[1:] return None # Extract "one point two five"-type float result = pointFloat(words) if result: return result # Extract "one and a quarter"-type float result = fractionFloat(words) if result: return result # Parse as integer return self.parseInt(words) def parseInt(self, words): """Parses words to the integer they describe. Args: words (str): Description of the integer. Returns: An integer representation of the words. """ # Remove 'and', case-sensitivity words = words.replace(" and ", " ").lower() # 'a' -> 'one' words = re.sub(r'(\b)a(\b)', '\g<1>one\g<2>', words) def textToNumber(s): """ Converts raw number string to an integer. Based on text2num.py by Greg Hewill. """ a = re.split(r"[\s-]+", s) n = 0 g = 0 for w in a: x = NumberService.__small__.get(w, None) if x is not None: g += x elif w == "hundred": g *= 100 else: x = NumberService.__magnitude__.get(w, None) if x is not None: n += g * x g = 0 else: raise NumberService.NumberException( "Unknown number: " + w) return n + g return textToNumber(words) def isValid(self, inp): try: self.parse(inp) return True except: return False @staticmethod def parseMagnitude(m): """Parses a number m into a human-ready string representation. For example, crops off floats if they're too accurate. Arguments: m (float): Floating-point number to be cleaned. Returns: Human-ready string description of the number. """ m = NumberService().parse(m) def toDecimalPrecision(n, k): return float("%.*f" % (k, round(n, k))) # Cast to two digits of precision digits = 2 magnitude = toDecimalPrecision(m, digits) # If value is really small, keep going while not magnitude: digits += 1 magnitude = toDecimalPrecision(m, digits) # If item is less than one, go one beyond 'necessary' number of digits if m < 1.0: magnitude = toDecimalPrecision(m, digits + 1) # Ignore decimal accuracy if irrelevant if int(magnitude) == magnitude: magnitude = int(magnitude) # Adjust for scientific notation magString = str(magnitude) magString = re.sub(r'(\d)e-(\d+)', '\g<1> times ten to the negative \g<2>', magString) magString = re.sub(r'(\d)e\+(\d+)', '\g<1> times ten to the \g<2>', magString) magString = re.sub(r'-(\d+)', 'negative \g<1>', magString) magString = re.sub(r'\b0(\d+)', '\g<1>', magString) return magString
crm416/semantic
semantic/dates.py
extractDates
python
def extractDates(inp, tz=None, now=None): service = DateService(tz=tz, now=now) return service.extractDates(inp)
Extract semantic date information from an input string. This is a convenience method which would only be used if you'd rather not initialize a DateService object. Args: inp (str): The input string to be parsed. tz: An optional Pytz timezone. All datetime objects returned will be relative to the supplied timezone, or timezone-less if none is supplied. now: The time to which all returned datetime objects should be relative. For example, if the text is "In 5 hours", the datetime returned will be now + datetime.timedelta(hours=5). Uses datetime.datetime.now() if none is supplied. Returns: A list of datetime objects extracted from input.
train
https://github.com/crm416/semantic/blob/46deb8fefb3ea58aad2fedc8d0d62f3ee254b8fe/semantic/dates.py#L476-L495
[ "def extractDates(self, inp):\n \"\"\"Extract semantic date information from an input string.\n In effect, runs both parseDay and parseTime on the input\n string and merges the results to produce a comprehensive\n datetime object.\n\n Args:\n inp (str): Input string to be parsed.\n\n Returns:\n A list of datetime objects containing the extracted dates from the\n input snippet, or an empty list if not found.\n \"\"\"\n def merge(param):\n day, time = param\n if not (day or time):\n return None\n\n if not day:\n return time\n if not time:\n return day\n\n return datetime.datetime(\n day.year, day.month, day.day, time.hour, time.minute\n )\n\n days = self.extractDays(inp)\n times = self.extractTimes(inp)\n return map(merge, zip_longest(days, times, fillvalue=None))\n" ]
import re import datetime try: from itertools import zip_longest except: from itertools import izip_longest as zip_longest from .numbers import NumberService class DateService(object): """Initialize a DateService for extracting dates from text. Args: tz: An optional Pytz timezone. All datetime objects returned will be relative to the supplied timezone, or timezone-less if none is supplied. now: The time to which all returned datetime objects should be relative. For example, if the text is "In 5 hours", the datetime returned will be now + datetime.timedelta(hours=5). Uses datetime.datetime.now() if none is supplied. Returns: A DateService which uses tz and now for all of its computations. """ def __init__(self, tz=None, now=None): self.tz = tz if now: self.now = now else: self.now = datetime.datetime.now(tz=self.tz) __months__ = ['january', 'february', 'march', 'april', 'may', 'june', 'july', 'august', 'september', 'october', 'november', 'december'] __shortMonths__ = ['jan', 'feb', 'mar', 'apr', 'may', 'jun', 'jul', 'aug', 'sept', 'oct', 'nov', 'dec'] __daysOfWeek__ = ['monday', 'tuesday', 'wednesday', 'thursday', 'friday', 'saturday', 'sunday'] __relativeDates__ = ['tomorrow', 'tonight', 'next'] __todayMatches__ = ['tonight', 'today', 'this morning', 'this evening', 'this afternoon'] __tomorrowMatches__ = ['tomorrow', 'next morning', 'next evening', 'next afternoon'] __dateDescriptors__ = { 'one': 1, 'first': 1, 'two': 2, 'second': 2, 'three': 3, 'third': 3, 'four': 4, 'fourth': 4, 'five': 5, 'fifth': 5, 'six': 6, 'sixth': 6, 'seven': 7, 'seventh': 7, 'eight': 8, 'eighth': 8, 'nine': 9, 'ninth': 9, 'ten': 10, 'tenth': 10, 'eleven': 11, 'eleventh': 11, 'twelve': 12, 'twelth': 12, 'thirteen': 13, 'thirteenth': 13, 'fourteen': 14, 'fourteenth': 14, 'fifteen': 15, 'fifteenth': 15, 'sixteen': 16, 'sixteenth': 16, 'seventeen': 17, 'seventeenth': 17, 'eighteen': 18, 'eighteenth': 18, 'nineteen': 19, 'nineteenth': 19, 'twenty': 20, 'twentieth': 20, 'twenty one': 21, 'twenty first': 21, 'twenty two': 22, 'twenty second': 22, 'twenty three': 23, 'twenty third': 23, 'twenty four': 24, 'twenty fourth': 24, 'twenty five': 25, 'twenty fifth': 25, 'twenty six': 26, 'twenty sixth': 26, 'twenty seven': 27, 'twenty seventh': 27, 'twenty eight': 28, 'twenty eighth': 28, 'twenty nine': 29, 'twenty ninth': 29, 'thirty': 30, 'thirtieth': 30, 'thirty one': 31, 'thirty first': 31 } _dayRegex = re.compile( r"""(?ix) ((week|day)s?\ from\ )? ( tomorrow |tonight |today |(next|this)[\ \b](morning|afternoon|evening|Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday) |(Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday) |(Jan(?:uary)?|Feb(?:ruary)?|Mar(?:ch)?|Apr(?:il)?|May|June?|July?|Aug(?:ust)?|Sept(?:ember)?|Oct(?:ober)?|Nov(?:ember)?|Dec(?:ember)?)\ (\w+)((\s|\-)?\w*) ) """) _timeRegex = re.compile( r"""(?ix) .*? ( morning |afternoon |evening |(\d{1,2}\:\d{2})\ ?(am|pm)? |in\ (.+?)\ (hours|minutes)(\ (?:and\ )?(.+?)\ (hours|minutes))? ) .*?""") def _preprocess(self, inp): return inp.replace('-', ' ').lower() def extractDays(self, inp): """Extracts all day-related information from an input string. Ignores any information related to the specific time-of-day. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted date from the input snippet, or an empty list if none found. """ inp = self._preprocess(inp) def extractDayOfWeek(dayMatch): if dayMatch.group(5) in self.__daysOfWeek__: return self.__daysOfWeek__.index(dayMatch.group(5)) elif dayMatch.group(6) in self.__daysOfWeek__: return self.__daysOfWeek__.index(dayMatch.group(6)) def extractMonth(dayMatch): if dayMatch.group(7) in self.__months__: return self.__months__.index(dayMatch.group(7)) + 1 elif dayMatch.group(7) in self.__shortMonths__: return self.__shortMonths__.index(dayMatch.group(7)) + 1 def extractDay(dayMatch): combined = dayMatch.group(8) + dayMatch.group(9) if combined in self.__dateDescriptors__: return self.__dateDescriptors__[combined] elif dayMatch.group(8) in self.__dateDescriptors__: return self.__dateDescriptors__[dayMatch.group(8)] elif int(dayMatch.group(8)) in self.__dateDescriptors__.values(): return int(dayMatch.group(8)) def extractDaysFrom(dayMatch): if not dayMatch.group(1): return 0 def numericalPrefix(dayMatch): # Grab 'three' of 'three weeks from' prefix = inp.split(dayMatch.group(1))[0].strip().split(' ') prefix.reverse() prefix = list(filter(lambda s: s != 'and', prefix)) # Generate best guess number service = NumberService() num = prefix[0] if service.isValid(num): for n in prefix[1:]: inc = n + " " + num if service.isValid(inc): num = inc else: break return service.parse(num) return 1 factor = numericalPrefix(dayMatch) if dayMatch.group(2) == 'week': return factor * 7 elif dayMatch.group(2) == 'day': return factor * 1 def handleMatch(dayMatch): def safe(exp): """For safe evaluation of regex groups""" try: return exp() except: return False days_from = safe(lambda: extractDaysFrom(dayMatch)) today = safe(lambda: dayMatch.group(3) in self.__todayMatches__) tomorrow = safe(lambda: dayMatch.group(3) in self.__tomorrowMatches__) next_week = safe(lambda: dayMatch.group(4) == 'next') day_of_week = safe(lambda: extractDayOfWeek(dayMatch)) month = safe(lambda: extractMonth(dayMatch)) day = safe(lambda: extractDay(dayMatch)) # Convert extracted terms to datetime object if not dayMatch: return None elif today: d = self.now elif tomorrow: d = self.now + datetime.timedelta(days=1) elif type(day_of_week) == int: current_day_of_week = self.now.weekday() num_days_away = (day_of_week - current_day_of_week) % 7 if next_week: num_days_away += 7 d = self.now + \ datetime.timedelta(days=num_days_away) elif month and day: d = datetime.datetime( self.now.year, month, day, self.now.hour, self.now.minute) if days_from: d += datetime.timedelta(days=days_from) return d matches = self._dayRegex.finditer(inp) return [handleMatch(dayMatch) for dayMatch in matches] def extractDay(self, inp): """Returns the first time-related date found in the input string, or None if not found.""" day = self.extractDay(inp) if day: return day[0] return None def extractTimes(self, inp): """Extracts time-related information from an input string. Ignores any information related to the specific date, focusing on the time-of-day. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted times from the input snippet, or an empty list if none found. """ def handleMatch(time): relative = False if not time: return None # Default times: 8am, 12pm, 7pm elif time.group(1) == 'morning': h = 8 m = 0 elif time.group(1) == 'afternoon': h = 12 m = 0 elif time.group(1) == 'evening': h = 19 m = 0 elif time.group(4) and time.group(5): h, m = 0, 0 # Extract hours difference converter = NumberService() try: diff = converter.parse(time.group(4)) except: return None if time.group(5) == 'hours': h += diff else: m += diff # Extract minutes difference if time.group(6): converter = NumberService() try: diff = converter.parse(time.group(7)) except: return None if time.group(8) == 'hours': h += diff else: m += diff relative = True else: # Convert from "HH:MM pm" format t = time.group(2) h, m = int(t.split(':')[0]) % 12, int(t.split(':')[1]) try: if time.group(3) == 'pm': h += 12 except IndexError: pass if relative: return self.now + datetime.timedelta(hours=h, minutes=m) else: return datetime.datetime( self.now.year, self.now.month, self.now.day, h, m ) inp = self._preprocess(inp) return [handleMatch(time) for time in self._timeRegex.finditer(inp)] def extractTime(self, inp): """Returns the first time-related date found in the input string, or None if not found.""" times = self.extractTimes(inp) if times: return times[0] return None def extractDates(self, inp): """Extract semantic date information from an input string. In effect, runs both parseDay and parseTime on the input string and merges the results to produce a comprehensive datetime object. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted dates from the input snippet, or an empty list if not found. """ def merge(param): day, time = param if not (day or time): return None if not day: return time if not time: return day return datetime.datetime( day.year, day.month, day.day, time.hour, time.minute ) days = self.extractDays(inp) times = self.extractTimes(inp) return map(merge, zip_longest(days, times, fillvalue=None)) def extractDate(self, inp): """Returns the first date found in the input string, or None if not found.""" dates = self.extractDates(inp) for date in dates: return date return None def convertDay(self, day, prefix="", weekday=False): """Convert a datetime object representing a day into a human-ready string that can be read, spoken aloud, etc. Args: day (datetime.date): A datetime object to be converted into text. prefix (str): An optional argument that prefixes the converted string. For example, if prefix="in", you'd receive "in two days", rather than "two days", while the method would still return "tomorrow" (rather than "in tomorrow"). weekday (bool): An optional argument that returns "Monday, Oct. 1" if True, rather than "Oct. 1". Returns: A string representation of the input day, ignoring any time-related information. """ def sameDay(d1, d2): d = d1.day == d2.day m = d1.month == d2.month y = d1.year == d2.year return d and m and y tom = self.now + datetime.timedelta(days=1) if sameDay(day, self.now): return "today" elif sameDay(day, tom): return "tomorrow" if weekday: dayString = day.strftime("%A, %B %d") else: dayString = day.strftime("%B %d") # Ex) Remove '0' from 'August 03' if not int(dayString[-2]): dayString = dayString[:-2] + dayString[-1] return prefix + " " + dayString def convertTime(self, time): """Convert a datetime object representing a time into a human-ready string that can be read, spoken aloud, etc. Args: time (datetime.date): A datetime object to be converted into text. Returns: A string representation of the input time, ignoring any day-related information. """ # if ':00', ignore reporting minutes m_format = "" if time.minute: m_format = ":%M" timeString = time.strftime("%I" + m_format + " %p") # if '07:30', cast to '7:30' if not int(timeString[0]): timeString = timeString[1:] return timeString def convertDate(self, date, prefix="", weekday=False): """Convert a datetime object representing into a human-ready string that can be read, spoken aloud, etc. In effect, runs both convertDay and convertTime on the input, merging the results. Args: date (datetime.date): A datetime object to be converted into text. prefix (str): An optional argument that prefixes the converted string. For example, if prefix="in", you'd receive "in two days", rather than "two days", while the method would still return "tomorrow" (rather than "in tomorrow"). weekday (bool): An optional argument that returns "Monday, Oct. 1" if True, rather than "Oct. 1". Returns: A string representation of the input day and time. """ dayString = self.convertDay( date, prefix=prefix, weekday=weekday) timeString = self.convertTime(date) return dayString + " at " + timeString
crm416/semantic
semantic/dates.py
DateService.extractDays
python
def extractDays(self, inp): inp = self._preprocess(inp) def extractDayOfWeek(dayMatch): if dayMatch.group(5) in self.__daysOfWeek__: return self.__daysOfWeek__.index(dayMatch.group(5)) elif dayMatch.group(6) in self.__daysOfWeek__: return self.__daysOfWeek__.index(dayMatch.group(6)) def extractMonth(dayMatch): if dayMatch.group(7) in self.__months__: return self.__months__.index(dayMatch.group(7)) + 1 elif dayMatch.group(7) in self.__shortMonths__: return self.__shortMonths__.index(dayMatch.group(7)) + 1 def extractDay(dayMatch): combined = dayMatch.group(8) + dayMatch.group(9) if combined in self.__dateDescriptors__: return self.__dateDescriptors__[combined] elif dayMatch.group(8) in self.__dateDescriptors__: return self.__dateDescriptors__[dayMatch.group(8)] elif int(dayMatch.group(8)) in self.__dateDescriptors__.values(): return int(dayMatch.group(8)) def extractDaysFrom(dayMatch): if not dayMatch.group(1): return 0 def numericalPrefix(dayMatch): # Grab 'three' of 'three weeks from' prefix = inp.split(dayMatch.group(1))[0].strip().split(' ') prefix.reverse() prefix = list(filter(lambda s: s != 'and', prefix)) # Generate best guess number service = NumberService() num = prefix[0] if service.isValid(num): for n in prefix[1:]: inc = n + " " + num if service.isValid(inc): num = inc else: break return service.parse(num) return 1 factor = numericalPrefix(dayMatch) if dayMatch.group(2) == 'week': return factor * 7 elif dayMatch.group(2) == 'day': return factor * 1 def handleMatch(dayMatch): def safe(exp): """For safe evaluation of regex groups""" try: return exp() except: return False days_from = safe(lambda: extractDaysFrom(dayMatch)) today = safe(lambda: dayMatch.group(3) in self.__todayMatches__) tomorrow = safe(lambda: dayMatch.group(3) in self.__tomorrowMatches__) next_week = safe(lambda: dayMatch.group(4) == 'next') day_of_week = safe(lambda: extractDayOfWeek(dayMatch)) month = safe(lambda: extractMonth(dayMatch)) day = safe(lambda: extractDay(dayMatch)) # Convert extracted terms to datetime object if not dayMatch: return None elif today: d = self.now elif tomorrow: d = self.now + datetime.timedelta(days=1) elif type(day_of_week) == int: current_day_of_week = self.now.weekday() num_days_away = (day_of_week - current_day_of_week) % 7 if next_week: num_days_away += 7 d = self.now + \ datetime.timedelta(days=num_days_away) elif month and day: d = datetime.datetime( self.now.year, month, day, self.now.hour, self.now.minute) if days_from: d += datetime.timedelta(days=days_from) return d matches = self._dayRegex.finditer(inp) return [handleMatch(dayMatch) for dayMatch in matches]
Extracts all day-related information from an input string. Ignores any information related to the specific time-of-day. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted date from the input snippet, or an empty list if none found.
train
https://github.com/crm416/semantic/blob/46deb8fefb3ea58aad2fedc8d0d62f3ee254b8fe/semantic/dates.py#L145-L253
[ "def _preprocess(self, inp):\n return inp.replace('-', ' ').lower()\n" ]
class DateService(object): """Initialize a DateService for extracting dates from text. Args: tz: An optional Pytz timezone. All datetime objects returned will be relative to the supplied timezone, or timezone-less if none is supplied. now: The time to which all returned datetime objects should be relative. For example, if the text is "In 5 hours", the datetime returned will be now + datetime.timedelta(hours=5). Uses datetime.datetime.now() if none is supplied. Returns: A DateService which uses tz and now for all of its computations. """ def __init__(self, tz=None, now=None): self.tz = tz if now: self.now = now else: self.now = datetime.datetime.now(tz=self.tz) __months__ = ['january', 'february', 'march', 'april', 'may', 'june', 'july', 'august', 'september', 'october', 'november', 'december'] __shortMonths__ = ['jan', 'feb', 'mar', 'apr', 'may', 'jun', 'jul', 'aug', 'sept', 'oct', 'nov', 'dec'] __daysOfWeek__ = ['monday', 'tuesday', 'wednesday', 'thursday', 'friday', 'saturday', 'sunday'] __relativeDates__ = ['tomorrow', 'tonight', 'next'] __todayMatches__ = ['tonight', 'today', 'this morning', 'this evening', 'this afternoon'] __tomorrowMatches__ = ['tomorrow', 'next morning', 'next evening', 'next afternoon'] __dateDescriptors__ = { 'one': 1, 'first': 1, 'two': 2, 'second': 2, 'three': 3, 'third': 3, 'four': 4, 'fourth': 4, 'five': 5, 'fifth': 5, 'six': 6, 'sixth': 6, 'seven': 7, 'seventh': 7, 'eight': 8, 'eighth': 8, 'nine': 9, 'ninth': 9, 'ten': 10, 'tenth': 10, 'eleven': 11, 'eleventh': 11, 'twelve': 12, 'twelth': 12, 'thirteen': 13, 'thirteenth': 13, 'fourteen': 14, 'fourteenth': 14, 'fifteen': 15, 'fifteenth': 15, 'sixteen': 16, 'sixteenth': 16, 'seventeen': 17, 'seventeenth': 17, 'eighteen': 18, 'eighteenth': 18, 'nineteen': 19, 'nineteenth': 19, 'twenty': 20, 'twentieth': 20, 'twenty one': 21, 'twenty first': 21, 'twenty two': 22, 'twenty second': 22, 'twenty three': 23, 'twenty third': 23, 'twenty four': 24, 'twenty fourth': 24, 'twenty five': 25, 'twenty fifth': 25, 'twenty six': 26, 'twenty sixth': 26, 'twenty seven': 27, 'twenty seventh': 27, 'twenty eight': 28, 'twenty eighth': 28, 'twenty nine': 29, 'twenty ninth': 29, 'thirty': 30, 'thirtieth': 30, 'thirty one': 31, 'thirty first': 31 } _dayRegex = re.compile( r"""(?ix) ((week|day)s?\ from\ )? ( tomorrow |tonight |today |(next|this)[\ \b](morning|afternoon|evening|Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday) |(Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday) |(Jan(?:uary)?|Feb(?:ruary)?|Mar(?:ch)?|Apr(?:il)?|May|June?|July?|Aug(?:ust)?|Sept(?:ember)?|Oct(?:ober)?|Nov(?:ember)?|Dec(?:ember)?)\ (\w+)((\s|\-)?\w*) ) """) _timeRegex = re.compile( r"""(?ix) .*? ( morning |afternoon |evening |(\d{1,2}\:\d{2})\ ?(am|pm)? |in\ (.+?)\ (hours|minutes)(\ (?:and\ )?(.+?)\ (hours|minutes))? ) .*?""") def _preprocess(self, inp): return inp.replace('-', ' ').lower() def extractDay(self, inp): """Returns the first time-related date found in the input string, or None if not found.""" day = self.extractDay(inp) if day: return day[0] return None def extractTimes(self, inp): """Extracts time-related information from an input string. Ignores any information related to the specific date, focusing on the time-of-day. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted times from the input snippet, or an empty list if none found. """ def handleMatch(time): relative = False if not time: return None # Default times: 8am, 12pm, 7pm elif time.group(1) == 'morning': h = 8 m = 0 elif time.group(1) == 'afternoon': h = 12 m = 0 elif time.group(1) == 'evening': h = 19 m = 0 elif time.group(4) and time.group(5): h, m = 0, 0 # Extract hours difference converter = NumberService() try: diff = converter.parse(time.group(4)) except: return None if time.group(5) == 'hours': h += diff else: m += diff # Extract minutes difference if time.group(6): converter = NumberService() try: diff = converter.parse(time.group(7)) except: return None if time.group(8) == 'hours': h += diff else: m += diff relative = True else: # Convert from "HH:MM pm" format t = time.group(2) h, m = int(t.split(':')[0]) % 12, int(t.split(':')[1]) try: if time.group(3) == 'pm': h += 12 except IndexError: pass if relative: return self.now + datetime.timedelta(hours=h, minutes=m) else: return datetime.datetime( self.now.year, self.now.month, self.now.day, h, m ) inp = self._preprocess(inp) return [handleMatch(time) for time in self._timeRegex.finditer(inp)] def extractTime(self, inp): """Returns the first time-related date found in the input string, or None if not found.""" times = self.extractTimes(inp) if times: return times[0] return None def extractDates(self, inp): """Extract semantic date information from an input string. In effect, runs both parseDay and parseTime on the input string and merges the results to produce a comprehensive datetime object. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted dates from the input snippet, or an empty list if not found. """ def merge(param): day, time = param if not (day or time): return None if not day: return time if not time: return day return datetime.datetime( day.year, day.month, day.day, time.hour, time.minute ) days = self.extractDays(inp) times = self.extractTimes(inp) return map(merge, zip_longest(days, times, fillvalue=None)) def extractDate(self, inp): """Returns the first date found in the input string, or None if not found.""" dates = self.extractDates(inp) for date in dates: return date return None def convertDay(self, day, prefix="", weekday=False): """Convert a datetime object representing a day into a human-ready string that can be read, spoken aloud, etc. Args: day (datetime.date): A datetime object to be converted into text. prefix (str): An optional argument that prefixes the converted string. For example, if prefix="in", you'd receive "in two days", rather than "two days", while the method would still return "tomorrow" (rather than "in tomorrow"). weekday (bool): An optional argument that returns "Monday, Oct. 1" if True, rather than "Oct. 1". Returns: A string representation of the input day, ignoring any time-related information. """ def sameDay(d1, d2): d = d1.day == d2.day m = d1.month == d2.month y = d1.year == d2.year return d and m and y tom = self.now + datetime.timedelta(days=1) if sameDay(day, self.now): return "today" elif sameDay(day, tom): return "tomorrow" if weekday: dayString = day.strftime("%A, %B %d") else: dayString = day.strftime("%B %d") # Ex) Remove '0' from 'August 03' if not int(dayString[-2]): dayString = dayString[:-2] + dayString[-1] return prefix + " " + dayString def convertTime(self, time): """Convert a datetime object representing a time into a human-ready string that can be read, spoken aloud, etc. Args: time (datetime.date): A datetime object to be converted into text. Returns: A string representation of the input time, ignoring any day-related information. """ # if ':00', ignore reporting minutes m_format = "" if time.minute: m_format = ":%M" timeString = time.strftime("%I" + m_format + " %p") # if '07:30', cast to '7:30' if not int(timeString[0]): timeString = timeString[1:] return timeString def convertDate(self, date, prefix="", weekday=False): """Convert a datetime object representing into a human-ready string that can be read, spoken aloud, etc. In effect, runs both convertDay and convertTime on the input, merging the results. Args: date (datetime.date): A datetime object to be converted into text. prefix (str): An optional argument that prefixes the converted string. For example, if prefix="in", you'd receive "in two days", rather than "two days", while the method would still return "tomorrow" (rather than "in tomorrow"). weekday (bool): An optional argument that returns "Monday, Oct. 1" if True, rather than "Oct. 1". Returns: A string representation of the input day and time. """ dayString = self.convertDay( date, prefix=prefix, weekday=weekday) timeString = self.convertTime(date) return dayString + " at " + timeString
crm416/semantic
semantic/dates.py
DateService.extractTimes
python
def extractTimes(self, inp): def handleMatch(time): relative = False if not time: return None # Default times: 8am, 12pm, 7pm elif time.group(1) == 'morning': h = 8 m = 0 elif time.group(1) == 'afternoon': h = 12 m = 0 elif time.group(1) == 'evening': h = 19 m = 0 elif time.group(4) and time.group(5): h, m = 0, 0 # Extract hours difference converter = NumberService() try: diff = converter.parse(time.group(4)) except: return None if time.group(5) == 'hours': h += diff else: m += diff # Extract minutes difference if time.group(6): converter = NumberService() try: diff = converter.parse(time.group(7)) except: return None if time.group(8) == 'hours': h += diff else: m += diff relative = True else: # Convert from "HH:MM pm" format t = time.group(2) h, m = int(t.split(':')[0]) % 12, int(t.split(':')[1]) try: if time.group(3) == 'pm': h += 12 except IndexError: pass if relative: return self.now + datetime.timedelta(hours=h, minutes=m) else: return datetime.datetime( self.now.year, self.now.month, self.now.day, h, m ) inp = self._preprocess(inp) return [handleMatch(time) for time in self._timeRegex.finditer(inp)]
Extracts time-related information from an input string. Ignores any information related to the specific date, focusing on the time-of-day. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted times from the input snippet, or an empty list if none found.
train
https://github.com/crm416/semantic/blob/46deb8fefb3ea58aad2fedc8d0d62f3ee254b8fe/semantic/dates.py#L263-L339
[ "def _preprocess(self, inp):\n return inp.replace('-', ' ').lower()\n" ]
class DateService(object): """Initialize a DateService for extracting dates from text. Args: tz: An optional Pytz timezone. All datetime objects returned will be relative to the supplied timezone, or timezone-less if none is supplied. now: The time to which all returned datetime objects should be relative. For example, if the text is "In 5 hours", the datetime returned will be now + datetime.timedelta(hours=5). Uses datetime.datetime.now() if none is supplied. Returns: A DateService which uses tz and now for all of its computations. """ def __init__(self, tz=None, now=None): self.tz = tz if now: self.now = now else: self.now = datetime.datetime.now(tz=self.tz) __months__ = ['january', 'february', 'march', 'april', 'may', 'june', 'july', 'august', 'september', 'october', 'november', 'december'] __shortMonths__ = ['jan', 'feb', 'mar', 'apr', 'may', 'jun', 'jul', 'aug', 'sept', 'oct', 'nov', 'dec'] __daysOfWeek__ = ['monday', 'tuesday', 'wednesday', 'thursday', 'friday', 'saturday', 'sunday'] __relativeDates__ = ['tomorrow', 'tonight', 'next'] __todayMatches__ = ['tonight', 'today', 'this morning', 'this evening', 'this afternoon'] __tomorrowMatches__ = ['tomorrow', 'next morning', 'next evening', 'next afternoon'] __dateDescriptors__ = { 'one': 1, 'first': 1, 'two': 2, 'second': 2, 'three': 3, 'third': 3, 'four': 4, 'fourth': 4, 'five': 5, 'fifth': 5, 'six': 6, 'sixth': 6, 'seven': 7, 'seventh': 7, 'eight': 8, 'eighth': 8, 'nine': 9, 'ninth': 9, 'ten': 10, 'tenth': 10, 'eleven': 11, 'eleventh': 11, 'twelve': 12, 'twelth': 12, 'thirteen': 13, 'thirteenth': 13, 'fourteen': 14, 'fourteenth': 14, 'fifteen': 15, 'fifteenth': 15, 'sixteen': 16, 'sixteenth': 16, 'seventeen': 17, 'seventeenth': 17, 'eighteen': 18, 'eighteenth': 18, 'nineteen': 19, 'nineteenth': 19, 'twenty': 20, 'twentieth': 20, 'twenty one': 21, 'twenty first': 21, 'twenty two': 22, 'twenty second': 22, 'twenty three': 23, 'twenty third': 23, 'twenty four': 24, 'twenty fourth': 24, 'twenty five': 25, 'twenty fifth': 25, 'twenty six': 26, 'twenty sixth': 26, 'twenty seven': 27, 'twenty seventh': 27, 'twenty eight': 28, 'twenty eighth': 28, 'twenty nine': 29, 'twenty ninth': 29, 'thirty': 30, 'thirtieth': 30, 'thirty one': 31, 'thirty first': 31 } _dayRegex = re.compile( r"""(?ix) ((week|day)s?\ from\ )? ( tomorrow |tonight |today |(next|this)[\ \b](morning|afternoon|evening|Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday) |(Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday) |(Jan(?:uary)?|Feb(?:ruary)?|Mar(?:ch)?|Apr(?:il)?|May|June?|July?|Aug(?:ust)?|Sept(?:ember)?|Oct(?:ober)?|Nov(?:ember)?|Dec(?:ember)?)\ (\w+)((\s|\-)?\w*) ) """) _timeRegex = re.compile( r"""(?ix) .*? ( morning |afternoon |evening |(\d{1,2}\:\d{2})\ ?(am|pm)? |in\ (.+?)\ (hours|minutes)(\ (?:and\ )?(.+?)\ (hours|minutes))? ) .*?""") def _preprocess(self, inp): return inp.replace('-', ' ').lower() def extractDays(self, inp): """Extracts all day-related information from an input string. Ignores any information related to the specific time-of-day. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted date from the input snippet, or an empty list if none found. """ inp = self._preprocess(inp) def extractDayOfWeek(dayMatch): if dayMatch.group(5) in self.__daysOfWeek__: return self.__daysOfWeek__.index(dayMatch.group(5)) elif dayMatch.group(6) in self.__daysOfWeek__: return self.__daysOfWeek__.index(dayMatch.group(6)) def extractMonth(dayMatch): if dayMatch.group(7) in self.__months__: return self.__months__.index(dayMatch.group(7)) + 1 elif dayMatch.group(7) in self.__shortMonths__: return self.__shortMonths__.index(dayMatch.group(7)) + 1 def extractDay(dayMatch): combined = dayMatch.group(8) + dayMatch.group(9) if combined in self.__dateDescriptors__: return self.__dateDescriptors__[combined] elif dayMatch.group(8) in self.__dateDescriptors__: return self.__dateDescriptors__[dayMatch.group(8)] elif int(dayMatch.group(8)) in self.__dateDescriptors__.values(): return int(dayMatch.group(8)) def extractDaysFrom(dayMatch): if not dayMatch.group(1): return 0 def numericalPrefix(dayMatch): # Grab 'three' of 'three weeks from' prefix = inp.split(dayMatch.group(1))[0].strip().split(' ') prefix.reverse() prefix = list(filter(lambda s: s != 'and', prefix)) # Generate best guess number service = NumberService() num = prefix[0] if service.isValid(num): for n in prefix[1:]: inc = n + " " + num if service.isValid(inc): num = inc else: break return service.parse(num) return 1 factor = numericalPrefix(dayMatch) if dayMatch.group(2) == 'week': return factor * 7 elif dayMatch.group(2) == 'day': return factor * 1 def handleMatch(dayMatch): def safe(exp): """For safe evaluation of regex groups""" try: return exp() except: return False days_from = safe(lambda: extractDaysFrom(dayMatch)) today = safe(lambda: dayMatch.group(3) in self.__todayMatches__) tomorrow = safe(lambda: dayMatch.group(3) in self.__tomorrowMatches__) next_week = safe(lambda: dayMatch.group(4) == 'next') day_of_week = safe(lambda: extractDayOfWeek(dayMatch)) month = safe(lambda: extractMonth(dayMatch)) day = safe(lambda: extractDay(dayMatch)) # Convert extracted terms to datetime object if not dayMatch: return None elif today: d = self.now elif tomorrow: d = self.now + datetime.timedelta(days=1) elif type(day_of_week) == int: current_day_of_week = self.now.weekday() num_days_away = (day_of_week - current_day_of_week) % 7 if next_week: num_days_away += 7 d = self.now + \ datetime.timedelta(days=num_days_away) elif month and day: d = datetime.datetime( self.now.year, month, day, self.now.hour, self.now.minute) if days_from: d += datetime.timedelta(days=days_from) return d matches = self._dayRegex.finditer(inp) return [handleMatch(dayMatch) for dayMatch in matches] def extractDay(self, inp): """Returns the first time-related date found in the input string, or None if not found.""" day = self.extractDay(inp) if day: return day[0] return None def extractTime(self, inp): """Returns the first time-related date found in the input string, or None if not found.""" times = self.extractTimes(inp) if times: return times[0] return None def extractDates(self, inp): """Extract semantic date information from an input string. In effect, runs both parseDay and parseTime on the input string and merges the results to produce a comprehensive datetime object. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted dates from the input snippet, or an empty list if not found. """ def merge(param): day, time = param if not (day or time): return None if not day: return time if not time: return day return datetime.datetime( day.year, day.month, day.day, time.hour, time.minute ) days = self.extractDays(inp) times = self.extractTimes(inp) return map(merge, zip_longest(days, times, fillvalue=None)) def extractDate(self, inp): """Returns the first date found in the input string, or None if not found.""" dates = self.extractDates(inp) for date in dates: return date return None def convertDay(self, day, prefix="", weekday=False): """Convert a datetime object representing a day into a human-ready string that can be read, spoken aloud, etc. Args: day (datetime.date): A datetime object to be converted into text. prefix (str): An optional argument that prefixes the converted string. For example, if prefix="in", you'd receive "in two days", rather than "two days", while the method would still return "tomorrow" (rather than "in tomorrow"). weekday (bool): An optional argument that returns "Monday, Oct. 1" if True, rather than "Oct. 1". Returns: A string representation of the input day, ignoring any time-related information. """ def sameDay(d1, d2): d = d1.day == d2.day m = d1.month == d2.month y = d1.year == d2.year return d and m and y tom = self.now + datetime.timedelta(days=1) if sameDay(day, self.now): return "today" elif sameDay(day, tom): return "tomorrow" if weekday: dayString = day.strftime("%A, %B %d") else: dayString = day.strftime("%B %d") # Ex) Remove '0' from 'August 03' if not int(dayString[-2]): dayString = dayString[:-2] + dayString[-1] return prefix + " " + dayString def convertTime(self, time): """Convert a datetime object representing a time into a human-ready string that can be read, spoken aloud, etc. Args: time (datetime.date): A datetime object to be converted into text. Returns: A string representation of the input time, ignoring any day-related information. """ # if ':00', ignore reporting minutes m_format = "" if time.minute: m_format = ":%M" timeString = time.strftime("%I" + m_format + " %p") # if '07:30', cast to '7:30' if not int(timeString[0]): timeString = timeString[1:] return timeString def convertDate(self, date, prefix="", weekday=False): """Convert a datetime object representing into a human-ready string that can be read, spoken aloud, etc. In effect, runs both convertDay and convertTime on the input, merging the results. Args: date (datetime.date): A datetime object to be converted into text. prefix (str): An optional argument that prefixes the converted string. For example, if prefix="in", you'd receive "in two days", rather than "two days", while the method would still return "tomorrow" (rather than "in tomorrow"). weekday (bool): An optional argument that returns "Monday, Oct. 1" if True, rather than "Oct. 1". Returns: A string representation of the input day and time. """ dayString = self.convertDay( date, prefix=prefix, weekday=weekday) timeString = self.convertTime(date) return dayString + " at " + timeString
crm416/semantic
semantic/dates.py
DateService.extractDates
python
def extractDates(self, inp): def merge(param): day, time = param if not (day or time): return None if not day: return time if not time: return day return datetime.datetime( day.year, day.month, day.day, time.hour, time.minute ) days = self.extractDays(inp) times = self.extractTimes(inp) return map(merge, zip_longest(days, times, fillvalue=None))
Extract semantic date information from an input string. In effect, runs both parseDay and parseTime on the input string and merges the results to produce a comprehensive datetime object. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted dates from the input snippet, or an empty list if not found.
train
https://github.com/crm416/semantic/blob/46deb8fefb3ea58aad2fedc8d0d62f3ee254b8fe/semantic/dates.py#L349-L378
[ "def extractDays(self, inp):\n \"\"\"Extracts all day-related information from an input string.\n Ignores any information related to the specific time-of-day.\n\n Args:\n inp (str): Input string to be parsed.\n\n Returns:\n A list of datetime objects containing the extracted date from the\n input snippet, or an empty list if none found.\n \"\"\"\n inp = self._preprocess(inp)\n\n def extractDayOfWeek(dayMatch):\n if dayMatch.group(5) in self.__daysOfWeek__:\n return self.__daysOfWeek__.index(dayMatch.group(5))\n elif dayMatch.group(6) in self.__daysOfWeek__:\n return self.__daysOfWeek__.index(dayMatch.group(6))\n\n def extractMonth(dayMatch):\n if dayMatch.group(7) in self.__months__:\n return self.__months__.index(dayMatch.group(7)) + 1\n elif dayMatch.group(7) in self.__shortMonths__:\n return self.__shortMonths__.index(dayMatch.group(7)) + 1\n\n def extractDay(dayMatch):\n combined = dayMatch.group(8) + dayMatch.group(9)\n if combined in self.__dateDescriptors__:\n return self.__dateDescriptors__[combined]\n elif dayMatch.group(8) in self.__dateDescriptors__:\n return self.__dateDescriptors__[dayMatch.group(8)]\n elif int(dayMatch.group(8)) in self.__dateDescriptors__.values():\n return int(dayMatch.group(8))\n\n def extractDaysFrom(dayMatch):\n if not dayMatch.group(1):\n return 0\n\n def numericalPrefix(dayMatch):\n # Grab 'three' of 'three weeks from'\n prefix = inp.split(dayMatch.group(1))[0].strip().split(' ')\n prefix.reverse()\n prefix = list(filter(lambda s: s != 'and', prefix))\n\n # Generate best guess number\n service = NumberService()\n num = prefix[0]\n if service.isValid(num):\n for n in prefix[1:]:\n inc = n + \" \" + num\n if service.isValid(inc):\n num = inc\n else:\n break\n return service.parse(num)\n return 1\n\n factor = numericalPrefix(dayMatch)\n\n if dayMatch.group(2) == 'week':\n return factor * 7\n elif dayMatch.group(2) == 'day':\n return factor * 1\n\n def handleMatch(dayMatch):\n def safe(exp):\n \"\"\"For safe evaluation of regex groups\"\"\"\n try:\n return exp()\n except:\n return False\n\n days_from = safe(lambda: extractDaysFrom(dayMatch))\n today = safe(lambda: dayMatch.group(3) in self.__todayMatches__)\n tomorrow = safe(lambda: dayMatch.group(3)\n in self.__tomorrowMatches__)\n next_week = safe(lambda: dayMatch.group(4) == 'next')\n day_of_week = safe(lambda: extractDayOfWeek(dayMatch))\n month = safe(lambda: extractMonth(dayMatch))\n day = safe(lambda: extractDay(dayMatch))\n\n # Convert extracted terms to datetime object\n if not dayMatch:\n return None\n elif today:\n d = self.now\n elif tomorrow:\n d = self.now + datetime.timedelta(days=1)\n elif type(day_of_week) == int:\n current_day_of_week = self.now.weekday()\n num_days_away = (day_of_week - current_day_of_week) % 7\n\n if next_week:\n num_days_away += 7\n\n d = self.now + \\\n datetime.timedelta(days=num_days_away)\n elif month and day:\n d = datetime.datetime(\n self.now.year, month, day,\n self.now.hour, self.now.minute)\n\n if days_from:\n d += datetime.timedelta(days=days_from)\n\n return d\n\n matches = self._dayRegex.finditer(inp)\n return [handleMatch(dayMatch) for dayMatch in matches]\n", "def extractTimes(self, inp):\n \"\"\"Extracts time-related information from an input string.\n Ignores any information related to the specific date, focusing\n on the time-of-day.\n\n Args:\n inp (str): Input string to be parsed.\n\n Returns:\n A list of datetime objects containing the extracted times from the\n input snippet, or an empty list if none found.\n \"\"\"\n def handleMatch(time):\n relative = False\n\n if not time:\n return None\n\n # Default times: 8am, 12pm, 7pm\n elif time.group(1) == 'morning':\n h = 8\n m = 0\n elif time.group(1) == 'afternoon':\n h = 12\n m = 0\n elif time.group(1) == 'evening':\n h = 19\n m = 0\n elif time.group(4) and time.group(5):\n h, m = 0, 0\n\n # Extract hours difference\n converter = NumberService()\n try:\n diff = converter.parse(time.group(4))\n except:\n return None\n\n if time.group(5) == 'hours':\n h += diff\n else:\n m += diff\n\n # Extract minutes difference\n if time.group(6):\n converter = NumberService()\n try:\n diff = converter.parse(time.group(7))\n except:\n return None\n\n if time.group(8) == 'hours':\n h += diff\n else:\n m += diff\n\n relative = True\n else:\n # Convert from \"HH:MM pm\" format\n t = time.group(2)\n h, m = int(t.split(':')[0]) % 12, int(t.split(':')[1])\n\n try:\n if time.group(3) == 'pm':\n h += 12\n except IndexError:\n pass\n\n if relative:\n return self.now + datetime.timedelta(hours=h, minutes=m)\n else:\n return datetime.datetime(\n self.now.year, self.now.month, self.now.day, h, m\n )\n\n inp = self._preprocess(inp)\n return [handleMatch(time) for time in self._timeRegex.finditer(inp)]\n" ]
class DateService(object): """Initialize a DateService for extracting dates from text. Args: tz: An optional Pytz timezone. All datetime objects returned will be relative to the supplied timezone, or timezone-less if none is supplied. now: The time to which all returned datetime objects should be relative. For example, if the text is "In 5 hours", the datetime returned will be now + datetime.timedelta(hours=5). Uses datetime.datetime.now() if none is supplied. Returns: A DateService which uses tz and now for all of its computations. """ def __init__(self, tz=None, now=None): self.tz = tz if now: self.now = now else: self.now = datetime.datetime.now(tz=self.tz) __months__ = ['january', 'february', 'march', 'april', 'may', 'june', 'july', 'august', 'september', 'october', 'november', 'december'] __shortMonths__ = ['jan', 'feb', 'mar', 'apr', 'may', 'jun', 'jul', 'aug', 'sept', 'oct', 'nov', 'dec'] __daysOfWeek__ = ['monday', 'tuesday', 'wednesday', 'thursday', 'friday', 'saturday', 'sunday'] __relativeDates__ = ['tomorrow', 'tonight', 'next'] __todayMatches__ = ['tonight', 'today', 'this morning', 'this evening', 'this afternoon'] __tomorrowMatches__ = ['tomorrow', 'next morning', 'next evening', 'next afternoon'] __dateDescriptors__ = { 'one': 1, 'first': 1, 'two': 2, 'second': 2, 'three': 3, 'third': 3, 'four': 4, 'fourth': 4, 'five': 5, 'fifth': 5, 'six': 6, 'sixth': 6, 'seven': 7, 'seventh': 7, 'eight': 8, 'eighth': 8, 'nine': 9, 'ninth': 9, 'ten': 10, 'tenth': 10, 'eleven': 11, 'eleventh': 11, 'twelve': 12, 'twelth': 12, 'thirteen': 13, 'thirteenth': 13, 'fourteen': 14, 'fourteenth': 14, 'fifteen': 15, 'fifteenth': 15, 'sixteen': 16, 'sixteenth': 16, 'seventeen': 17, 'seventeenth': 17, 'eighteen': 18, 'eighteenth': 18, 'nineteen': 19, 'nineteenth': 19, 'twenty': 20, 'twentieth': 20, 'twenty one': 21, 'twenty first': 21, 'twenty two': 22, 'twenty second': 22, 'twenty three': 23, 'twenty third': 23, 'twenty four': 24, 'twenty fourth': 24, 'twenty five': 25, 'twenty fifth': 25, 'twenty six': 26, 'twenty sixth': 26, 'twenty seven': 27, 'twenty seventh': 27, 'twenty eight': 28, 'twenty eighth': 28, 'twenty nine': 29, 'twenty ninth': 29, 'thirty': 30, 'thirtieth': 30, 'thirty one': 31, 'thirty first': 31 } _dayRegex = re.compile( r"""(?ix) ((week|day)s?\ from\ )? ( tomorrow |tonight |today |(next|this)[\ \b](morning|afternoon|evening|Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday) |(Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday) |(Jan(?:uary)?|Feb(?:ruary)?|Mar(?:ch)?|Apr(?:il)?|May|June?|July?|Aug(?:ust)?|Sept(?:ember)?|Oct(?:ober)?|Nov(?:ember)?|Dec(?:ember)?)\ (\w+)((\s|\-)?\w*) ) """) _timeRegex = re.compile( r"""(?ix) .*? ( morning |afternoon |evening |(\d{1,2}\:\d{2})\ ?(am|pm)? |in\ (.+?)\ (hours|minutes)(\ (?:and\ )?(.+?)\ (hours|minutes))? ) .*?""") def _preprocess(self, inp): return inp.replace('-', ' ').lower() def extractDays(self, inp): """Extracts all day-related information from an input string. Ignores any information related to the specific time-of-day. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted date from the input snippet, or an empty list if none found. """ inp = self._preprocess(inp) def extractDayOfWeek(dayMatch): if dayMatch.group(5) in self.__daysOfWeek__: return self.__daysOfWeek__.index(dayMatch.group(5)) elif dayMatch.group(6) in self.__daysOfWeek__: return self.__daysOfWeek__.index(dayMatch.group(6)) def extractMonth(dayMatch): if dayMatch.group(7) in self.__months__: return self.__months__.index(dayMatch.group(7)) + 1 elif dayMatch.group(7) in self.__shortMonths__: return self.__shortMonths__.index(dayMatch.group(7)) + 1 def extractDay(dayMatch): combined = dayMatch.group(8) + dayMatch.group(9) if combined in self.__dateDescriptors__: return self.__dateDescriptors__[combined] elif dayMatch.group(8) in self.__dateDescriptors__: return self.__dateDescriptors__[dayMatch.group(8)] elif int(dayMatch.group(8)) in self.__dateDescriptors__.values(): return int(dayMatch.group(8)) def extractDaysFrom(dayMatch): if not dayMatch.group(1): return 0 def numericalPrefix(dayMatch): # Grab 'three' of 'three weeks from' prefix = inp.split(dayMatch.group(1))[0].strip().split(' ') prefix.reverse() prefix = list(filter(lambda s: s != 'and', prefix)) # Generate best guess number service = NumberService() num = prefix[0] if service.isValid(num): for n in prefix[1:]: inc = n + " " + num if service.isValid(inc): num = inc else: break return service.parse(num) return 1 factor = numericalPrefix(dayMatch) if dayMatch.group(2) == 'week': return factor * 7 elif dayMatch.group(2) == 'day': return factor * 1 def handleMatch(dayMatch): def safe(exp): """For safe evaluation of regex groups""" try: return exp() except: return False days_from = safe(lambda: extractDaysFrom(dayMatch)) today = safe(lambda: dayMatch.group(3) in self.__todayMatches__) tomorrow = safe(lambda: dayMatch.group(3) in self.__tomorrowMatches__) next_week = safe(lambda: dayMatch.group(4) == 'next') day_of_week = safe(lambda: extractDayOfWeek(dayMatch)) month = safe(lambda: extractMonth(dayMatch)) day = safe(lambda: extractDay(dayMatch)) # Convert extracted terms to datetime object if not dayMatch: return None elif today: d = self.now elif tomorrow: d = self.now + datetime.timedelta(days=1) elif type(day_of_week) == int: current_day_of_week = self.now.weekday() num_days_away = (day_of_week - current_day_of_week) % 7 if next_week: num_days_away += 7 d = self.now + \ datetime.timedelta(days=num_days_away) elif month and day: d = datetime.datetime( self.now.year, month, day, self.now.hour, self.now.minute) if days_from: d += datetime.timedelta(days=days_from) return d matches = self._dayRegex.finditer(inp) return [handleMatch(dayMatch) for dayMatch in matches] def extractDay(self, inp): """Returns the first time-related date found in the input string, or None if not found.""" day = self.extractDay(inp) if day: return day[0] return None def extractTimes(self, inp): """Extracts time-related information from an input string. Ignores any information related to the specific date, focusing on the time-of-day. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted times from the input snippet, or an empty list if none found. """ def handleMatch(time): relative = False if not time: return None # Default times: 8am, 12pm, 7pm elif time.group(1) == 'morning': h = 8 m = 0 elif time.group(1) == 'afternoon': h = 12 m = 0 elif time.group(1) == 'evening': h = 19 m = 0 elif time.group(4) and time.group(5): h, m = 0, 0 # Extract hours difference converter = NumberService() try: diff = converter.parse(time.group(4)) except: return None if time.group(5) == 'hours': h += diff else: m += diff # Extract minutes difference if time.group(6): converter = NumberService() try: diff = converter.parse(time.group(7)) except: return None if time.group(8) == 'hours': h += diff else: m += diff relative = True else: # Convert from "HH:MM pm" format t = time.group(2) h, m = int(t.split(':')[0]) % 12, int(t.split(':')[1]) try: if time.group(3) == 'pm': h += 12 except IndexError: pass if relative: return self.now + datetime.timedelta(hours=h, minutes=m) else: return datetime.datetime( self.now.year, self.now.month, self.now.day, h, m ) inp = self._preprocess(inp) return [handleMatch(time) for time in self._timeRegex.finditer(inp)] def extractTime(self, inp): """Returns the first time-related date found in the input string, or None if not found.""" times = self.extractTimes(inp) if times: return times[0] return None def extractDate(self, inp): """Returns the first date found in the input string, or None if not found.""" dates = self.extractDates(inp) for date in dates: return date return None def convertDay(self, day, prefix="", weekday=False): """Convert a datetime object representing a day into a human-ready string that can be read, spoken aloud, etc. Args: day (datetime.date): A datetime object to be converted into text. prefix (str): An optional argument that prefixes the converted string. For example, if prefix="in", you'd receive "in two days", rather than "two days", while the method would still return "tomorrow" (rather than "in tomorrow"). weekday (bool): An optional argument that returns "Monday, Oct. 1" if True, rather than "Oct. 1". Returns: A string representation of the input day, ignoring any time-related information. """ def sameDay(d1, d2): d = d1.day == d2.day m = d1.month == d2.month y = d1.year == d2.year return d and m and y tom = self.now + datetime.timedelta(days=1) if sameDay(day, self.now): return "today" elif sameDay(day, tom): return "tomorrow" if weekday: dayString = day.strftime("%A, %B %d") else: dayString = day.strftime("%B %d") # Ex) Remove '0' from 'August 03' if not int(dayString[-2]): dayString = dayString[:-2] + dayString[-1] return prefix + " " + dayString def convertTime(self, time): """Convert a datetime object representing a time into a human-ready string that can be read, spoken aloud, etc. Args: time (datetime.date): A datetime object to be converted into text. Returns: A string representation of the input time, ignoring any day-related information. """ # if ':00', ignore reporting minutes m_format = "" if time.minute: m_format = ":%M" timeString = time.strftime("%I" + m_format + " %p") # if '07:30', cast to '7:30' if not int(timeString[0]): timeString = timeString[1:] return timeString def convertDate(self, date, prefix="", weekday=False): """Convert a datetime object representing into a human-ready string that can be read, spoken aloud, etc. In effect, runs both convertDay and convertTime on the input, merging the results. Args: date (datetime.date): A datetime object to be converted into text. prefix (str): An optional argument that prefixes the converted string. For example, if prefix="in", you'd receive "in two days", rather than "two days", while the method would still return "tomorrow" (rather than "in tomorrow"). weekday (bool): An optional argument that returns "Monday, Oct. 1" if True, rather than "Oct. 1". Returns: A string representation of the input day and time. """ dayString = self.convertDay( date, prefix=prefix, weekday=weekday) timeString = self.convertTime(date) return dayString + " at " + timeString
crm416/semantic
semantic/dates.py
DateService.extractDate
python
def extractDate(self, inp): dates = self.extractDates(inp) for date in dates: return date return None
Returns the first date found in the input string, or None if not found.
train
https://github.com/crm416/semantic/blob/46deb8fefb3ea58aad2fedc8d0d62f3ee254b8fe/semantic/dates.py#L380-L386
[ "def extractDates(self, inp):\n \"\"\"Extract semantic date information from an input string.\n In effect, runs both parseDay and parseTime on the input\n string and merges the results to produce a comprehensive\n datetime object.\n\n Args:\n inp (str): Input string to be parsed.\n\n Returns:\n A list of datetime objects containing the extracted dates from the\n input snippet, or an empty list if not found.\n \"\"\"\n def merge(param):\n day, time = param\n if not (day or time):\n return None\n\n if not day:\n return time\n if not time:\n return day\n\n return datetime.datetime(\n day.year, day.month, day.day, time.hour, time.minute\n )\n\n days = self.extractDays(inp)\n times = self.extractTimes(inp)\n return map(merge, zip_longest(days, times, fillvalue=None))\n" ]
class DateService(object): """Initialize a DateService for extracting dates from text. Args: tz: An optional Pytz timezone. All datetime objects returned will be relative to the supplied timezone, or timezone-less if none is supplied. now: The time to which all returned datetime objects should be relative. For example, if the text is "In 5 hours", the datetime returned will be now + datetime.timedelta(hours=5). Uses datetime.datetime.now() if none is supplied. Returns: A DateService which uses tz and now for all of its computations. """ def __init__(self, tz=None, now=None): self.tz = tz if now: self.now = now else: self.now = datetime.datetime.now(tz=self.tz) __months__ = ['january', 'february', 'march', 'april', 'may', 'june', 'july', 'august', 'september', 'october', 'november', 'december'] __shortMonths__ = ['jan', 'feb', 'mar', 'apr', 'may', 'jun', 'jul', 'aug', 'sept', 'oct', 'nov', 'dec'] __daysOfWeek__ = ['monday', 'tuesday', 'wednesday', 'thursday', 'friday', 'saturday', 'sunday'] __relativeDates__ = ['tomorrow', 'tonight', 'next'] __todayMatches__ = ['tonight', 'today', 'this morning', 'this evening', 'this afternoon'] __tomorrowMatches__ = ['tomorrow', 'next morning', 'next evening', 'next afternoon'] __dateDescriptors__ = { 'one': 1, 'first': 1, 'two': 2, 'second': 2, 'three': 3, 'third': 3, 'four': 4, 'fourth': 4, 'five': 5, 'fifth': 5, 'six': 6, 'sixth': 6, 'seven': 7, 'seventh': 7, 'eight': 8, 'eighth': 8, 'nine': 9, 'ninth': 9, 'ten': 10, 'tenth': 10, 'eleven': 11, 'eleventh': 11, 'twelve': 12, 'twelth': 12, 'thirteen': 13, 'thirteenth': 13, 'fourteen': 14, 'fourteenth': 14, 'fifteen': 15, 'fifteenth': 15, 'sixteen': 16, 'sixteenth': 16, 'seventeen': 17, 'seventeenth': 17, 'eighteen': 18, 'eighteenth': 18, 'nineteen': 19, 'nineteenth': 19, 'twenty': 20, 'twentieth': 20, 'twenty one': 21, 'twenty first': 21, 'twenty two': 22, 'twenty second': 22, 'twenty three': 23, 'twenty third': 23, 'twenty four': 24, 'twenty fourth': 24, 'twenty five': 25, 'twenty fifth': 25, 'twenty six': 26, 'twenty sixth': 26, 'twenty seven': 27, 'twenty seventh': 27, 'twenty eight': 28, 'twenty eighth': 28, 'twenty nine': 29, 'twenty ninth': 29, 'thirty': 30, 'thirtieth': 30, 'thirty one': 31, 'thirty first': 31 } _dayRegex = re.compile( r"""(?ix) ((week|day)s?\ from\ )? ( tomorrow |tonight |today |(next|this)[\ \b](morning|afternoon|evening|Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday) |(Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday) |(Jan(?:uary)?|Feb(?:ruary)?|Mar(?:ch)?|Apr(?:il)?|May|June?|July?|Aug(?:ust)?|Sept(?:ember)?|Oct(?:ober)?|Nov(?:ember)?|Dec(?:ember)?)\ (\w+)((\s|\-)?\w*) ) """) _timeRegex = re.compile( r"""(?ix) .*? ( morning |afternoon |evening |(\d{1,2}\:\d{2})\ ?(am|pm)? |in\ (.+?)\ (hours|minutes)(\ (?:and\ )?(.+?)\ (hours|minutes))? ) .*?""") def _preprocess(self, inp): return inp.replace('-', ' ').lower() def extractDays(self, inp): """Extracts all day-related information from an input string. Ignores any information related to the specific time-of-day. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted date from the input snippet, or an empty list if none found. """ inp = self._preprocess(inp) def extractDayOfWeek(dayMatch): if dayMatch.group(5) in self.__daysOfWeek__: return self.__daysOfWeek__.index(dayMatch.group(5)) elif dayMatch.group(6) in self.__daysOfWeek__: return self.__daysOfWeek__.index(dayMatch.group(6)) def extractMonth(dayMatch): if dayMatch.group(7) in self.__months__: return self.__months__.index(dayMatch.group(7)) + 1 elif dayMatch.group(7) in self.__shortMonths__: return self.__shortMonths__.index(dayMatch.group(7)) + 1 def extractDay(dayMatch): combined = dayMatch.group(8) + dayMatch.group(9) if combined in self.__dateDescriptors__: return self.__dateDescriptors__[combined] elif dayMatch.group(8) in self.__dateDescriptors__: return self.__dateDescriptors__[dayMatch.group(8)] elif int(dayMatch.group(8)) in self.__dateDescriptors__.values(): return int(dayMatch.group(8)) def extractDaysFrom(dayMatch): if not dayMatch.group(1): return 0 def numericalPrefix(dayMatch): # Grab 'three' of 'three weeks from' prefix = inp.split(dayMatch.group(1))[0].strip().split(' ') prefix.reverse() prefix = list(filter(lambda s: s != 'and', prefix)) # Generate best guess number service = NumberService() num = prefix[0] if service.isValid(num): for n in prefix[1:]: inc = n + " " + num if service.isValid(inc): num = inc else: break return service.parse(num) return 1 factor = numericalPrefix(dayMatch) if dayMatch.group(2) == 'week': return factor * 7 elif dayMatch.group(2) == 'day': return factor * 1 def handleMatch(dayMatch): def safe(exp): """For safe evaluation of regex groups""" try: return exp() except: return False days_from = safe(lambda: extractDaysFrom(dayMatch)) today = safe(lambda: dayMatch.group(3) in self.__todayMatches__) tomorrow = safe(lambda: dayMatch.group(3) in self.__tomorrowMatches__) next_week = safe(lambda: dayMatch.group(4) == 'next') day_of_week = safe(lambda: extractDayOfWeek(dayMatch)) month = safe(lambda: extractMonth(dayMatch)) day = safe(lambda: extractDay(dayMatch)) # Convert extracted terms to datetime object if not dayMatch: return None elif today: d = self.now elif tomorrow: d = self.now + datetime.timedelta(days=1) elif type(day_of_week) == int: current_day_of_week = self.now.weekday() num_days_away = (day_of_week - current_day_of_week) % 7 if next_week: num_days_away += 7 d = self.now + \ datetime.timedelta(days=num_days_away) elif month and day: d = datetime.datetime( self.now.year, month, day, self.now.hour, self.now.minute) if days_from: d += datetime.timedelta(days=days_from) return d matches = self._dayRegex.finditer(inp) return [handleMatch(dayMatch) for dayMatch in matches] def extractDay(self, inp): """Returns the first time-related date found in the input string, or None if not found.""" day = self.extractDay(inp) if day: return day[0] return None def extractTimes(self, inp): """Extracts time-related information from an input string. Ignores any information related to the specific date, focusing on the time-of-day. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted times from the input snippet, or an empty list if none found. """ def handleMatch(time): relative = False if not time: return None # Default times: 8am, 12pm, 7pm elif time.group(1) == 'morning': h = 8 m = 0 elif time.group(1) == 'afternoon': h = 12 m = 0 elif time.group(1) == 'evening': h = 19 m = 0 elif time.group(4) and time.group(5): h, m = 0, 0 # Extract hours difference converter = NumberService() try: diff = converter.parse(time.group(4)) except: return None if time.group(5) == 'hours': h += diff else: m += diff # Extract minutes difference if time.group(6): converter = NumberService() try: diff = converter.parse(time.group(7)) except: return None if time.group(8) == 'hours': h += diff else: m += diff relative = True else: # Convert from "HH:MM pm" format t = time.group(2) h, m = int(t.split(':')[0]) % 12, int(t.split(':')[1]) try: if time.group(3) == 'pm': h += 12 except IndexError: pass if relative: return self.now + datetime.timedelta(hours=h, minutes=m) else: return datetime.datetime( self.now.year, self.now.month, self.now.day, h, m ) inp = self._preprocess(inp) return [handleMatch(time) for time in self._timeRegex.finditer(inp)] def extractTime(self, inp): """Returns the first time-related date found in the input string, or None if not found.""" times = self.extractTimes(inp) if times: return times[0] return None def extractDates(self, inp): """Extract semantic date information from an input string. In effect, runs both parseDay and parseTime on the input string and merges the results to produce a comprehensive datetime object. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted dates from the input snippet, or an empty list if not found. """ def merge(param): day, time = param if not (day or time): return None if not day: return time if not time: return day return datetime.datetime( day.year, day.month, day.day, time.hour, time.minute ) days = self.extractDays(inp) times = self.extractTimes(inp) return map(merge, zip_longest(days, times, fillvalue=None)) def convertDay(self, day, prefix="", weekday=False): """Convert a datetime object representing a day into a human-ready string that can be read, spoken aloud, etc. Args: day (datetime.date): A datetime object to be converted into text. prefix (str): An optional argument that prefixes the converted string. For example, if prefix="in", you'd receive "in two days", rather than "two days", while the method would still return "tomorrow" (rather than "in tomorrow"). weekday (bool): An optional argument that returns "Monday, Oct. 1" if True, rather than "Oct. 1". Returns: A string representation of the input day, ignoring any time-related information. """ def sameDay(d1, d2): d = d1.day == d2.day m = d1.month == d2.month y = d1.year == d2.year return d and m and y tom = self.now + datetime.timedelta(days=1) if sameDay(day, self.now): return "today" elif sameDay(day, tom): return "tomorrow" if weekday: dayString = day.strftime("%A, %B %d") else: dayString = day.strftime("%B %d") # Ex) Remove '0' from 'August 03' if not int(dayString[-2]): dayString = dayString[:-2] + dayString[-1] return prefix + " " + dayString def convertTime(self, time): """Convert a datetime object representing a time into a human-ready string that can be read, spoken aloud, etc. Args: time (datetime.date): A datetime object to be converted into text. Returns: A string representation of the input time, ignoring any day-related information. """ # if ':00', ignore reporting minutes m_format = "" if time.minute: m_format = ":%M" timeString = time.strftime("%I" + m_format + " %p") # if '07:30', cast to '7:30' if not int(timeString[0]): timeString = timeString[1:] return timeString def convertDate(self, date, prefix="", weekday=False): """Convert a datetime object representing into a human-ready string that can be read, spoken aloud, etc. In effect, runs both convertDay and convertTime on the input, merging the results. Args: date (datetime.date): A datetime object to be converted into text. prefix (str): An optional argument that prefixes the converted string. For example, if prefix="in", you'd receive "in two days", rather than "two days", while the method would still return "tomorrow" (rather than "in tomorrow"). weekday (bool): An optional argument that returns "Monday, Oct. 1" if True, rather than "Oct. 1". Returns: A string representation of the input day and time. """ dayString = self.convertDay( date, prefix=prefix, weekday=weekday) timeString = self.convertTime(date) return dayString + " at " + timeString
crm416/semantic
semantic/dates.py
DateService.convertDay
python
def convertDay(self, day, prefix="", weekday=False): def sameDay(d1, d2): d = d1.day == d2.day m = d1.month == d2.month y = d1.year == d2.year return d and m and y tom = self.now + datetime.timedelta(days=1) if sameDay(day, self.now): return "today" elif sameDay(day, tom): return "tomorrow" if weekday: dayString = day.strftime("%A, %B %d") else: dayString = day.strftime("%B %d") # Ex) Remove '0' from 'August 03' if not int(dayString[-2]): dayString = dayString[:-2] + dayString[-1] return prefix + " " + dayString
Convert a datetime object representing a day into a human-ready string that can be read, spoken aloud, etc. Args: day (datetime.date): A datetime object to be converted into text. prefix (str): An optional argument that prefixes the converted string. For example, if prefix="in", you'd receive "in two days", rather than "two days", while the method would still return "tomorrow" (rather than "in tomorrow"). weekday (bool): An optional argument that returns "Monday, Oct. 1" if True, rather than "Oct. 1". Returns: A string representation of the input day, ignoring any time-related information.
train
https://github.com/crm416/semantic/blob/46deb8fefb3ea58aad2fedc8d0d62f3ee254b8fe/semantic/dates.py#L388-L427
[ "def sameDay(d1, d2):\n d = d1.day == d2.day\n m = d1.month == d2.month\n y = d1.year == d2.year\n return d and m and y\n" ]
class DateService(object): """Initialize a DateService for extracting dates from text. Args: tz: An optional Pytz timezone. All datetime objects returned will be relative to the supplied timezone, or timezone-less if none is supplied. now: The time to which all returned datetime objects should be relative. For example, if the text is "In 5 hours", the datetime returned will be now + datetime.timedelta(hours=5). Uses datetime.datetime.now() if none is supplied. Returns: A DateService which uses tz and now for all of its computations. """ def __init__(self, tz=None, now=None): self.tz = tz if now: self.now = now else: self.now = datetime.datetime.now(tz=self.tz) __months__ = ['january', 'february', 'march', 'april', 'may', 'june', 'july', 'august', 'september', 'october', 'november', 'december'] __shortMonths__ = ['jan', 'feb', 'mar', 'apr', 'may', 'jun', 'jul', 'aug', 'sept', 'oct', 'nov', 'dec'] __daysOfWeek__ = ['monday', 'tuesday', 'wednesday', 'thursday', 'friday', 'saturday', 'sunday'] __relativeDates__ = ['tomorrow', 'tonight', 'next'] __todayMatches__ = ['tonight', 'today', 'this morning', 'this evening', 'this afternoon'] __tomorrowMatches__ = ['tomorrow', 'next morning', 'next evening', 'next afternoon'] __dateDescriptors__ = { 'one': 1, 'first': 1, 'two': 2, 'second': 2, 'three': 3, 'third': 3, 'four': 4, 'fourth': 4, 'five': 5, 'fifth': 5, 'six': 6, 'sixth': 6, 'seven': 7, 'seventh': 7, 'eight': 8, 'eighth': 8, 'nine': 9, 'ninth': 9, 'ten': 10, 'tenth': 10, 'eleven': 11, 'eleventh': 11, 'twelve': 12, 'twelth': 12, 'thirteen': 13, 'thirteenth': 13, 'fourteen': 14, 'fourteenth': 14, 'fifteen': 15, 'fifteenth': 15, 'sixteen': 16, 'sixteenth': 16, 'seventeen': 17, 'seventeenth': 17, 'eighteen': 18, 'eighteenth': 18, 'nineteen': 19, 'nineteenth': 19, 'twenty': 20, 'twentieth': 20, 'twenty one': 21, 'twenty first': 21, 'twenty two': 22, 'twenty second': 22, 'twenty three': 23, 'twenty third': 23, 'twenty four': 24, 'twenty fourth': 24, 'twenty five': 25, 'twenty fifth': 25, 'twenty six': 26, 'twenty sixth': 26, 'twenty seven': 27, 'twenty seventh': 27, 'twenty eight': 28, 'twenty eighth': 28, 'twenty nine': 29, 'twenty ninth': 29, 'thirty': 30, 'thirtieth': 30, 'thirty one': 31, 'thirty first': 31 } _dayRegex = re.compile( r"""(?ix) ((week|day)s?\ from\ )? ( tomorrow |tonight |today |(next|this)[\ \b](morning|afternoon|evening|Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday) |(Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday) |(Jan(?:uary)?|Feb(?:ruary)?|Mar(?:ch)?|Apr(?:il)?|May|June?|July?|Aug(?:ust)?|Sept(?:ember)?|Oct(?:ober)?|Nov(?:ember)?|Dec(?:ember)?)\ (\w+)((\s|\-)?\w*) ) """) _timeRegex = re.compile( r"""(?ix) .*? ( morning |afternoon |evening |(\d{1,2}\:\d{2})\ ?(am|pm)? |in\ (.+?)\ (hours|minutes)(\ (?:and\ )?(.+?)\ (hours|minutes))? ) .*?""") def _preprocess(self, inp): return inp.replace('-', ' ').lower() def extractDays(self, inp): """Extracts all day-related information from an input string. Ignores any information related to the specific time-of-day. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted date from the input snippet, or an empty list if none found. """ inp = self._preprocess(inp) def extractDayOfWeek(dayMatch): if dayMatch.group(5) in self.__daysOfWeek__: return self.__daysOfWeek__.index(dayMatch.group(5)) elif dayMatch.group(6) in self.__daysOfWeek__: return self.__daysOfWeek__.index(dayMatch.group(6)) def extractMonth(dayMatch): if dayMatch.group(7) in self.__months__: return self.__months__.index(dayMatch.group(7)) + 1 elif dayMatch.group(7) in self.__shortMonths__: return self.__shortMonths__.index(dayMatch.group(7)) + 1 def extractDay(dayMatch): combined = dayMatch.group(8) + dayMatch.group(9) if combined in self.__dateDescriptors__: return self.__dateDescriptors__[combined] elif dayMatch.group(8) in self.__dateDescriptors__: return self.__dateDescriptors__[dayMatch.group(8)] elif int(dayMatch.group(8)) in self.__dateDescriptors__.values(): return int(dayMatch.group(8)) def extractDaysFrom(dayMatch): if not dayMatch.group(1): return 0 def numericalPrefix(dayMatch): # Grab 'three' of 'three weeks from' prefix = inp.split(dayMatch.group(1))[0].strip().split(' ') prefix.reverse() prefix = list(filter(lambda s: s != 'and', prefix)) # Generate best guess number service = NumberService() num = prefix[0] if service.isValid(num): for n in prefix[1:]: inc = n + " " + num if service.isValid(inc): num = inc else: break return service.parse(num) return 1 factor = numericalPrefix(dayMatch) if dayMatch.group(2) == 'week': return factor * 7 elif dayMatch.group(2) == 'day': return factor * 1 def handleMatch(dayMatch): def safe(exp): """For safe evaluation of regex groups""" try: return exp() except: return False days_from = safe(lambda: extractDaysFrom(dayMatch)) today = safe(lambda: dayMatch.group(3) in self.__todayMatches__) tomorrow = safe(lambda: dayMatch.group(3) in self.__tomorrowMatches__) next_week = safe(lambda: dayMatch.group(4) == 'next') day_of_week = safe(lambda: extractDayOfWeek(dayMatch)) month = safe(lambda: extractMonth(dayMatch)) day = safe(lambda: extractDay(dayMatch)) # Convert extracted terms to datetime object if not dayMatch: return None elif today: d = self.now elif tomorrow: d = self.now + datetime.timedelta(days=1) elif type(day_of_week) == int: current_day_of_week = self.now.weekday() num_days_away = (day_of_week - current_day_of_week) % 7 if next_week: num_days_away += 7 d = self.now + \ datetime.timedelta(days=num_days_away) elif month and day: d = datetime.datetime( self.now.year, month, day, self.now.hour, self.now.minute) if days_from: d += datetime.timedelta(days=days_from) return d matches = self._dayRegex.finditer(inp) return [handleMatch(dayMatch) for dayMatch in matches] def extractDay(self, inp): """Returns the first time-related date found in the input string, or None if not found.""" day = self.extractDay(inp) if day: return day[0] return None def extractTimes(self, inp): """Extracts time-related information from an input string. Ignores any information related to the specific date, focusing on the time-of-day. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted times from the input snippet, or an empty list if none found. """ def handleMatch(time): relative = False if not time: return None # Default times: 8am, 12pm, 7pm elif time.group(1) == 'morning': h = 8 m = 0 elif time.group(1) == 'afternoon': h = 12 m = 0 elif time.group(1) == 'evening': h = 19 m = 0 elif time.group(4) and time.group(5): h, m = 0, 0 # Extract hours difference converter = NumberService() try: diff = converter.parse(time.group(4)) except: return None if time.group(5) == 'hours': h += diff else: m += diff # Extract minutes difference if time.group(6): converter = NumberService() try: diff = converter.parse(time.group(7)) except: return None if time.group(8) == 'hours': h += diff else: m += diff relative = True else: # Convert from "HH:MM pm" format t = time.group(2) h, m = int(t.split(':')[0]) % 12, int(t.split(':')[1]) try: if time.group(3) == 'pm': h += 12 except IndexError: pass if relative: return self.now + datetime.timedelta(hours=h, minutes=m) else: return datetime.datetime( self.now.year, self.now.month, self.now.day, h, m ) inp = self._preprocess(inp) return [handleMatch(time) for time in self._timeRegex.finditer(inp)] def extractTime(self, inp): """Returns the first time-related date found in the input string, or None if not found.""" times = self.extractTimes(inp) if times: return times[0] return None def extractDates(self, inp): """Extract semantic date information from an input string. In effect, runs both parseDay and parseTime on the input string and merges the results to produce a comprehensive datetime object. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted dates from the input snippet, or an empty list if not found. """ def merge(param): day, time = param if not (day or time): return None if not day: return time if not time: return day return datetime.datetime( day.year, day.month, day.day, time.hour, time.minute ) days = self.extractDays(inp) times = self.extractTimes(inp) return map(merge, zip_longest(days, times, fillvalue=None)) def extractDate(self, inp): """Returns the first date found in the input string, or None if not found.""" dates = self.extractDates(inp) for date in dates: return date return None def convertTime(self, time): """Convert a datetime object representing a time into a human-ready string that can be read, spoken aloud, etc. Args: time (datetime.date): A datetime object to be converted into text. Returns: A string representation of the input time, ignoring any day-related information. """ # if ':00', ignore reporting minutes m_format = "" if time.minute: m_format = ":%M" timeString = time.strftime("%I" + m_format + " %p") # if '07:30', cast to '7:30' if not int(timeString[0]): timeString = timeString[1:] return timeString def convertDate(self, date, prefix="", weekday=False): """Convert a datetime object representing into a human-ready string that can be read, spoken aloud, etc. In effect, runs both convertDay and convertTime on the input, merging the results. Args: date (datetime.date): A datetime object to be converted into text. prefix (str): An optional argument that prefixes the converted string. For example, if prefix="in", you'd receive "in two days", rather than "two days", while the method would still return "tomorrow" (rather than "in tomorrow"). weekday (bool): An optional argument that returns "Monday, Oct. 1" if True, rather than "Oct. 1". Returns: A string representation of the input day and time. """ dayString = self.convertDay( date, prefix=prefix, weekday=weekday) timeString = self.convertTime(date) return dayString + " at " + timeString
crm416/semantic
semantic/dates.py
DateService.convertTime
python
def convertTime(self, time): # if ':00', ignore reporting minutes m_format = "" if time.minute: m_format = ":%M" timeString = time.strftime("%I" + m_format + " %p") # if '07:30', cast to '7:30' if not int(timeString[0]): timeString = timeString[1:] return timeString
Convert a datetime object representing a time into a human-ready string that can be read, spoken aloud, etc. Args: time (datetime.date): A datetime object to be converted into text. Returns: A string representation of the input time, ignoring any day-related information.
train
https://github.com/crm416/semantic/blob/46deb8fefb3ea58aad2fedc8d0d62f3ee254b8fe/semantic/dates.py#L429-L451
null
class DateService(object): """Initialize a DateService for extracting dates from text. Args: tz: An optional Pytz timezone. All datetime objects returned will be relative to the supplied timezone, or timezone-less if none is supplied. now: The time to which all returned datetime objects should be relative. For example, if the text is "In 5 hours", the datetime returned will be now + datetime.timedelta(hours=5). Uses datetime.datetime.now() if none is supplied. Returns: A DateService which uses tz and now for all of its computations. """ def __init__(self, tz=None, now=None): self.tz = tz if now: self.now = now else: self.now = datetime.datetime.now(tz=self.tz) __months__ = ['january', 'february', 'march', 'april', 'may', 'june', 'july', 'august', 'september', 'october', 'november', 'december'] __shortMonths__ = ['jan', 'feb', 'mar', 'apr', 'may', 'jun', 'jul', 'aug', 'sept', 'oct', 'nov', 'dec'] __daysOfWeek__ = ['monday', 'tuesday', 'wednesday', 'thursday', 'friday', 'saturday', 'sunday'] __relativeDates__ = ['tomorrow', 'tonight', 'next'] __todayMatches__ = ['tonight', 'today', 'this morning', 'this evening', 'this afternoon'] __tomorrowMatches__ = ['tomorrow', 'next morning', 'next evening', 'next afternoon'] __dateDescriptors__ = { 'one': 1, 'first': 1, 'two': 2, 'second': 2, 'three': 3, 'third': 3, 'four': 4, 'fourth': 4, 'five': 5, 'fifth': 5, 'six': 6, 'sixth': 6, 'seven': 7, 'seventh': 7, 'eight': 8, 'eighth': 8, 'nine': 9, 'ninth': 9, 'ten': 10, 'tenth': 10, 'eleven': 11, 'eleventh': 11, 'twelve': 12, 'twelth': 12, 'thirteen': 13, 'thirteenth': 13, 'fourteen': 14, 'fourteenth': 14, 'fifteen': 15, 'fifteenth': 15, 'sixteen': 16, 'sixteenth': 16, 'seventeen': 17, 'seventeenth': 17, 'eighteen': 18, 'eighteenth': 18, 'nineteen': 19, 'nineteenth': 19, 'twenty': 20, 'twentieth': 20, 'twenty one': 21, 'twenty first': 21, 'twenty two': 22, 'twenty second': 22, 'twenty three': 23, 'twenty third': 23, 'twenty four': 24, 'twenty fourth': 24, 'twenty five': 25, 'twenty fifth': 25, 'twenty six': 26, 'twenty sixth': 26, 'twenty seven': 27, 'twenty seventh': 27, 'twenty eight': 28, 'twenty eighth': 28, 'twenty nine': 29, 'twenty ninth': 29, 'thirty': 30, 'thirtieth': 30, 'thirty one': 31, 'thirty first': 31 } _dayRegex = re.compile( r"""(?ix) ((week|day)s?\ from\ )? ( tomorrow |tonight |today |(next|this)[\ \b](morning|afternoon|evening|Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday) |(Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday) |(Jan(?:uary)?|Feb(?:ruary)?|Mar(?:ch)?|Apr(?:il)?|May|June?|July?|Aug(?:ust)?|Sept(?:ember)?|Oct(?:ober)?|Nov(?:ember)?|Dec(?:ember)?)\ (\w+)((\s|\-)?\w*) ) """) _timeRegex = re.compile( r"""(?ix) .*? ( morning |afternoon |evening |(\d{1,2}\:\d{2})\ ?(am|pm)? |in\ (.+?)\ (hours|minutes)(\ (?:and\ )?(.+?)\ (hours|minutes))? ) .*?""") def _preprocess(self, inp): return inp.replace('-', ' ').lower() def extractDays(self, inp): """Extracts all day-related information from an input string. Ignores any information related to the specific time-of-day. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted date from the input snippet, or an empty list if none found. """ inp = self._preprocess(inp) def extractDayOfWeek(dayMatch): if dayMatch.group(5) in self.__daysOfWeek__: return self.__daysOfWeek__.index(dayMatch.group(5)) elif dayMatch.group(6) in self.__daysOfWeek__: return self.__daysOfWeek__.index(dayMatch.group(6)) def extractMonth(dayMatch): if dayMatch.group(7) in self.__months__: return self.__months__.index(dayMatch.group(7)) + 1 elif dayMatch.group(7) in self.__shortMonths__: return self.__shortMonths__.index(dayMatch.group(7)) + 1 def extractDay(dayMatch): combined = dayMatch.group(8) + dayMatch.group(9) if combined in self.__dateDescriptors__: return self.__dateDescriptors__[combined] elif dayMatch.group(8) in self.__dateDescriptors__: return self.__dateDescriptors__[dayMatch.group(8)] elif int(dayMatch.group(8)) in self.__dateDescriptors__.values(): return int(dayMatch.group(8)) def extractDaysFrom(dayMatch): if not dayMatch.group(1): return 0 def numericalPrefix(dayMatch): # Grab 'three' of 'three weeks from' prefix = inp.split(dayMatch.group(1))[0].strip().split(' ') prefix.reverse() prefix = list(filter(lambda s: s != 'and', prefix)) # Generate best guess number service = NumberService() num = prefix[0] if service.isValid(num): for n in prefix[1:]: inc = n + " " + num if service.isValid(inc): num = inc else: break return service.parse(num) return 1 factor = numericalPrefix(dayMatch) if dayMatch.group(2) == 'week': return factor * 7 elif dayMatch.group(2) == 'day': return factor * 1 def handleMatch(dayMatch): def safe(exp): """For safe evaluation of regex groups""" try: return exp() except: return False days_from = safe(lambda: extractDaysFrom(dayMatch)) today = safe(lambda: dayMatch.group(3) in self.__todayMatches__) tomorrow = safe(lambda: dayMatch.group(3) in self.__tomorrowMatches__) next_week = safe(lambda: dayMatch.group(4) == 'next') day_of_week = safe(lambda: extractDayOfWeek(dayMatch)) month = safe(lambda: extractMonth(dayMatch)) day = safe(lambda: extractDay(dayMatch)) # Convert extracted terms to datetime object if not dayMatch: return None elif today: d = self.now elif tomorrow: d = self.now + datetime.timedelta(days=1) elif type(day_of_week) == int: current_day_of_week = self.now.weekday() num_days_away = (day_of_week - current_day_of_week) % 7 if next_week: num_days_away += 7 d = self.now + \ datetime.timedelta(days=num_days_away) elif month and day: d = datetime.datetime( self.now.year, month, day, self.now.hour, self.now.minute) if days_from: d += datetime.timedelta(days=days_from) return d matches = self._dayRegex.finditer(inp) return [handleMatch(dayMatch) for dayMatch in matches] def extractDay(self, inp): """Returns the first time-related date found in the input string, or None if not found.""" day = self.extractDay(inp) if day: return day[0] return None def extractTimes(self, inp): """Extracts time-related information from an input string. Ignores any information related to the specific date, focusing on the time-of-day. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted times from the input snippet, or an empty list if none found. """ def handleMatch(time): relative = False if not time: return None # Default times: 8am, 12pm, 7pm elif time.group(1) == 'morning': h = 8 m = 0 elif time.group(1) == 'afternoon': h = 12 m = 0 elif time.group(1) == 'evening': h = 19 m = 0 elif time.group(4) and time.group(5): h, m = 0, 0 # Extract hours difference converter = NumberService() try: diff = converter.parse(time.group(4)) except: return None if time.group(5) == 'hours': h += diff else: m += diff # Extract minutes difference if time.group(6): converter = NumberService() try: diff = converter.parse(time.group(7)) except: return None if time.group(8) == 'hours': h += diff else: m += diff relative = True else: # Convert from "HH:MM pm" format t = time.group(2) h, m = int(t.split(':')[0]) % 12, int(t.split(':')[1]) try: if time.group(3) == 'pm': h += 12 except IndexError: pass if relative: return self.now + datetime.timedelta(hours=h, minutes=m) else: return datetime.datetime( self.now.year, self.now.month, self.now.day, h, m ) inp = self._preprocess(inp) return [handleMatch(time) for time in self._timeRegex.finditer(inp)] def extractTime(self, inp): """Returns the first time-related date found in the input string, or None if not found.""" times = self.extractTimes(inp) if times: return times[0] return None def extractDates(self, inp): """Extract semantic date information from an input string. In effect, runs both parseDay and parseTime on the input string and merges the results to produce a comprehensive datetime object. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted dates from the input snippet, or an empty list if not found. """ def merge(param): day, time = param if not (day or time): return None if not day: return time if not time: return day return datetime.datetime( day.year, day.month, day.day, time.hour, time.minute ) days = self.extractDays(inp) times = self.extractTimes(inp) return map(merge, zip_longest(days, times, fillvalue=None)) def extractDate(self, inp): """Returns the first date found in the input string, or None if not found.""" dates = self.extractDates(inp) for date in dates: return date return None def convertDay(self, day, prefix="", weekday=False): """Convert a datetime object representing a day into a human-ready string that can be read, spoken aloud, etc. Args: day (datetime.date): A datetime object to be converted into text. prefix (str): An optional argument that prefixes the converted string. For example, if prefix="in", you'd receive "in two days", rather than "two days", while the method would still return "tomorrow" (rather than "in tomorrow"). weekday (bool): An optional argument that returns "Monday, Oct. 1" if True, rather than "Oct. 1". Returns: A string representation of the input day, ignoring any time-related information. """ def sameDay(d1, d2): d = d1.day == d2.day m = d1.month == d2.month y = d1.year == d2.year return d and m and y tom = self.now + datetime.timedelta(days=1) if sameDay(day, self.now): return "today" elif sameDay(day, tom): return "tomorrow" if weekday: dayString = day.strftime("%A, %B %d") else: dayString = day.strftime("%B %d") # Ex) Remove '0' from 'August 03' if not int(dayString[-2]): dayString = dayString[:-2] + dayString[-1] return prefix + " " + dayString def convertDate(self, date, prefix="", weekday=False): """Convert a datetime object representing into a human-ready string that can be read, spoken aloud, etc. In effect, runs both convertDay and convertTime on the input, merging the results. Args: date (datetime.date): A datetime object to be converted into text. prefix (str): An optional argument that prefixes the converted string. For example, if prefix="in", you'd receive "in two days", rather than "two days", while the method would still return "tomorrow" (rather than "in tomorrow"). weekday (bool): An optional argument that returns "Monday, Oct. 1" if True, rather than "Oct. 1". Returns: A string representation of the input day and time. """ dayString = self.convertDay( date, prefix=prefix, weekday=weekday) timeString = self.convertTime(date) return dayString + " at " + timeString
crm416/semantic
semantic/dates.py
DateService.convertDate
python
def convertDate(self, date, prefix="", weekday=False): dayString = self.convertDay( date, prefix=prefix, weekday=weekday) timeString = self.convertTime(date) return dayString + " at " + timeString
Convert a datetime object representing into a human-ready string that can be read, spoken aloud, etc. In effect, runs both convertDay and convertTime on the input, merging the results. Args: date (datetime.date): A datetime object to be converted into text. prefix (str): An optional argument that prefixes the converted string. For example, if prefix="in", you'd receive "in two days", rather than "two days", while the method would still return "tomorrow" (rather than "in tomorrow"). weekday (bool): An optional argument that returns "Monday, Oct. 1" if True, rather than "Oct. 1". Returns: A string representation of the input day and time.
train
https://github.com/crm416/semantic/blob/46deb8fefb3ea58aad2fedc8d0d62f3ee254b8fe/semantic/dates.py#L453-L473
[ "def convertDay(self, day, prefix=\"\", weekday=False):\n \"\"\"Convert a datetime object representing a day into a human-ready\n string that can be read, spoken aloud, etc.\n\n Args:\n day (datetime.date): A datetime object to be converted into text.\n prefix (str): An optional argument that prefixes the converted\n string. For example, if prefix=\"in\", you'd receive \"in two\n days\", rather than \"two days\", while the method would still\n return \"tomorrow\" (rather than \"in tomorrow\").\n weekday (bool): An optional argument that returns \"Monday, Oct. 1\"\n if True, rather than \"Oct. 1\".\n\n Returns:\n A string representation of the input day, ignoring any time-related\n information.\n \"\"\"\n def sameDay(d1, d2):\n d = d1.day == d2.day\n m = d1.month == d2.month\n y = d1.year == d2.year\n return d and m and y\n\n tom = self.now + datetime.timedelta(days=1)\n\n if sameDay(day, self.now):\n return \"today\"\n elif sameDay(day, tom):\n return \"tomorrow\"\n\n if weekday:\n dayString = day.strftime(\"%A, %B %d\")\n else:\n dayString = day.strftime(\"%B %d\")\n\n # Ex) Remove '0' from 'August 03'\n if not int(dayString[-2]):\n dayString = dayString[:-2] + dayString[-1]\n\n return prefix + \" \" + dayString\n", "def convertTime(self, time):\n \"\"\"Convert a datetime object representing a time into a human-ready\n string that can be read, spoken aloud, etc.\n\n Args:\n time (datetime.date): A datetime object to be converted into text.\n\n Returns:\n A string representation of the input time, ignoring any day-related\n information.\n \"\"\"\n # if ':00', ignore reporting minutes\n m_format = \"\"\n if time.minute:\n m_format = \":%M\"\n\n timeString = time.strftime(\"%I\" + m_format + \" %p\")\n\n # if '07:30', cast to '7:30'\n if not int(timeString[0]):\n timeString = timeString[1:]\n\n return timeString\n" ]
class DateService(object): """Initialize a DateService for extracting dates from text. Args: tz: An optional Pytz timezone. All datetime objects returned will be relative to the supplied timezone, or timezone-less if none is supplied. now: The time to which all returned datetime objects should be relative. For example, if the text is "In 5 hours", the datetime returned will be now + datetime.timedelta(hours=5). Uses datetime.datetime.now() if none is supplied. Returns: A DateService which uses tz and now for all of its computations. """ def __init__(self, tz=None, now=None): self.tz = tz if now: self.now = now else: self.now = datetime.datetime.now(tz=self.tz) __months__ = ['january', 'february', 'march', 'april', 'may', 'june', 'july', 'august', 'september', 'october', 'november', 'december'] __shortMonths__ = ['jan', 'feb', 'mar', 'apr', 'may', 'jun', 'jul', 'aug', 'sept', 'oct', 'nov', 'dec'] __daysOfWeek__ = ['monday', 'tuesday', 'wednesday', 'thursday', 'friday', 'saturday', 'sunday'] __relativeDates__ = ['tomorrow', 'tonight', 'next'] __todayMatches__ = ['tonight', 'today', 'this morning', 'this evening', 'this afternoon'] __tomorrowMatches__ = ['tomorrow', 'next morning', 'next evening', 'next afternoon'] __dateDescriptors__ = { 'one': 1, 'first': 1, 'two': 2, 'second': 2, 'three': 3, 'third': 3, 'four': 4, 'fourth': 4, 'five': 5, 'fifth': 5, 'six': 6, 'sixth': 6, 'seven': 7, 'seventh': 7, 'eight': 8, 'eighth': 8, 'nine': 9, 'ninth': 9, 'ten': 10, 'tenth': 10, 'eleven': 11, 'eleventh': 11, 'twelve': 12, 'twelth': 12, 'thirteen': 13, 'thirteenth': 13, 'fourteen': 14, 'fourteenth': 14, 'fifteen': 15, 'fifteenth': 15, 'sixteen': 16, 'sixteenth': 16, 'seventeen': 17, 'seventeenth': 17, 'eighteen': 18, 'eighteenth': 18, 'nineteen': 19, 'nineteenth': 19, 'twenty': 20, 'twentieth': 20, 'twenty one': 21, 'twenty first': 21, 'twenty two': 22, 'twenty second': 22, 'twenty three': 23, 'twenty third': 23, 'twenty four': 24, 'twenty fourth': 24, 'twenty five': 25, 'twenty fifth': 25, 'twenty six': 26, 'twenty sixth': 26, 'twenty seven': 27, 'twenty seventh': 27, 'twenty eight': 28, 'twenty eighth': 28, 'twenty nine': 29, 'twenty ninth': 29, 'thirty': 30, 'thirtieth': 30, 'thirty one': 31, 'thirty first': 31 } _dayRegex = re.compile( r"""(?ix) ((week|day)s?\ from\ )? ( tomorrow |tonight |today |(next|this)[\ \b](morning|afternoon|evening|Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday) |(Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday) |(Jan(?:uary)?|Feb(?:ruary)?|Mar(?:ch)?|Apr(?:il)?|May|June?|July?|Aug(?:ust)?|Sept(?:ember)?|Oct(?:ober)?|Nov(?:ember)?|Dec(?:ember)?)\ (\w+)((\s|\-)?\w*) ) """) _timeRegex = re.compile( r"""(?ix) .*? ( morning |afternoon |evening |(\d{1,2}\:\d{2})\ ?(am|pm)? |in\ (.+?)\ (hours|minutes)(\ (?:and\ )?(.+?)\ (hours|minutes))? ) .*?""") def _preprocess(self, inp): return inp.replace('-', ' ').lower() def extractDays(self, inp): """Extracts all day-related information from an input string. Ignores any information related to the specific time-of-day. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted date from the input snippet, or an empty list if none found. """ inp = self._preprocess(inp) def extractDayOfWeek(dayMatch): if dayMatch.group(5) in self.__daysOfWeek__: return self.__daysOfWeek__.index(dayMatch.group(5)) elif dayMatch.group(6) in self.__daysOfWeek__: return self.__daysOfWeek__.index(dayMatch.group(6)) def extractMonth(dayMatch): if dayMatch.group(7) in self.__months__: return self.__months__.index(dayMatch.group(7)) + 1 elif dayMatch.group(7) in self.__shortMonths__: return self.__shortMonths__.index(dayMatch.group(7)) + 1 def extractDay(dayMatch): combined = dayMatch.group(8) + dayMatch.group(9) if combined in self.__dateDescriptors__: return self.__dateDescriptors__[combined] elif dayMatch.group(8) in self.__dateDescriptors__: return self.__dateDescriptors__[dayMatch.group(8)] elif int(dayMatch.group(8)) in self.__dateDescriptors__.values(): return int(dayMatch.group(8)) def extractDaysFrom(dayMatch): if not dayMatch.group(1): return 0 def numericalPrefix(dayMatch): # Grab 'three' of 'three weeks from' prefix = inp.split(dayMatch.group(1))[0].strip().split(' ') prefix.reverse() prefix = list(filter(lambda s: s != 'and', prefix)) # Generate best guess number service = NumberService() num = prefix[0] if service.isValid(num): for n in prefix[1:]: inc = n + " " + num if service.isValid(inc): num = inc else: break return service.parse(num) return 1 factor = numericalPrefix(dayMatch) if dayMatch.group(2) == 'week': return factor * 7 elif dayMatch.group(2) == 'day': return factor * 1 def handleMatch(dayMatch): def safe(exp): """For safe evaluation of regex groups""" try: return exp() except: return False days_from = safe(lambda: extractDaysFrom(dayMatch)) today = safe(lambda: dayMatch.group(3) in self.__todayMatches__) tomorrow = safe(lambda: dayMatch.group(3) in self.__tomorrowMatches__) next_week = safe(lambda: dayMatch.group(4) == 'next') day_of_week = safe(lambda: extractDayOfWeek(dayMatch)) month = safe(lambda: extractMonth(dayMatch)) day = safe(lambda: extractDay(dayMatch)) # Convert extracted terms to datetime object if not dayMatch: return None elif today: d = self.now elif tomorrow: d = self.now + datetime.timedelta(days=1) elif type(day_of_week) == int: current_day_of_week = self.now.weekday() num_days_away = (day_of_week - current_day_of_week) % 7 if next_week: num_days_away += 7 d = self.now + \ datetime.timedelta(days=num_days_away) elif month and day: d = datetime.datetime( self.now.year, month, day, self.now.hour, self.now.minute) if days_from: d += datetime.timedelta(days=days_from) return d matches = self._dayRegex.finditer(inp) return [handleMatch(dayMatch) for dayMatch in matches] def extractDay(self, inp): """Returns the first time-related date found in the input string, or None if not found.""" day = self.extractDay(inp) if day: return day[0] return None def extractTimes(self, inp): """Extracts time-related information from an input string. Ignores any information related to the specific date, focusing on the time-of-day. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted times from the input snippet, or an empty list if none found. """ def handleMatch(time): relative = False if not time: return None # Default times: 8am, 12pm, 7pm elif time.group(1) == 'morning': h = 8 m = 0 elif time.group(1) == 'afternoon': h = 12 m = 0 elif time.group(1) == 'evening': h = 19 m = 0 elif time.group(4) and time.group(5): h, m = 0, 0 # Extract hours difference converter = NumberService() try: diff = converter.parse(time.group(4)) except: return None if time.group(5) == 'hours': h += diff else: m += diff # Extract minutes difference if time.group(6): converter = NumberService() try: diff = converter.parse(time.group(7)) except: return None if time.group(8) == 'hours': h += diff else: m += diff relative = True else: # Convert from "HH:MM pm" format t = time.group(2) h, m = int(t.split(':')[0]) % 12, int(t.split(':')[1]) try: if time.group(3) == 'pm': h += 12 except IndexError: pass if relative: return self.now + datetime.timedelta(hours=h, minutes=m) else: return datetime.datetime( self.now.year, self.now.month, self.now.day, h, m ) inp = self._preprocess(inp) return [handleMatch(time) for time in self._timeRegex.finditer(inp)] def extractTime(self, inp): """Returns the first time-related date found in the input string, or None if not found.""" times = self.extractTimes(inp) if times: return times[0] return None def extractDates(self, inp): """Extract semantic date information from an input string. In effect, runs both parseDay and parseTime on the input string and merges the results to produce a comprehensive datetime object. Args: inp (str): Input string to be parsed. Returns: A list of datetime objects containing the extracted dates from the input snippet, or an empty list if not found. """ def merge(param): day, time = param if not (day or time): return None if not day: return time if not time: return day return datetime.datetime( day.year, day.month, day.day, time.hour, time.minute ) days = self.extractDays(inp) times = self.extractTimes(inp) return map(merge, zip_longest(days, times, fillvalue=None)) def extractDate(self, inp): """Returns the first date found in the input string, or None if not found.""" dates = self.extractDates(inp) for date in dates: return date return None def convertDay(self, day, prefix="", weekday=False): """Convert a datetime object representing a day into a human-ready string that can be read, spoken aloud, etc. Args: day (datetime.date): A datetime object to be converted into text. prefix (str): An optional argument that prefixes the converted string. For example, if prefix="in", you'd receive "in two days", rather than "two days", while the method would still return "tomorrow" (rather than "in tomorrow"). weekday (bool): An optional argument that returns "Monday, Oct. 1" if True, rather than "Oct. 1". Returns: A string representation of the input day, ignoring any time-related information. """ def sameDay(d1, d2): d = d1.day == d2.day m = d1.month == d2.month y = d1.year == d2.year return d and m and y tom = self.now + datetime.timedelta(days=1) if sameDay(day, self.now): return "today" elif sameDay(day, tom): return "tomorrow" if weekday: dayString = day.strftime("%A, %B %d") else: dayString = day.strftime("%B %d") # Ex) Remove '0' from 'August 03' if not int(dayString[-2]): dayString = dayString[:-2] + dayString[-1] return prefix + " " + dayString def convertTime(self, time): """Convert a datetime object representing a time into a human-ready string that can be read, spoken aloud, etc. Args: time (datetime.date): A datetime object to be converted into text. Returns: A string representation of the input time, ignoring any day-related information. """ # if ':00', ignore reporting minutes m_format = "" if time.minute: m_format = ":%M" timeString = time.strftime("%I" + m_format + " %p") # if '07:30', cast to '7:30' if not int(timeString[0]): timeString = timeString[1:] return timeString
Parsely/schemato
schemato/validator.py
SchemaValidator.validate
python
def validate(self): log.info("{}\nValidating against {}" .format("-" * 100, self.schema_def.__class__.__name__)) if not self.schema_def: raise ValueError("No schema definition supplied.") self.checked_attributes = [] # TODO - this should maybe choose the actually used namespace, not just # the first one in the list result = ValidationResult(self.allowed_namespaces[0], self.schema_def.__class__.__name__) for subject, predicate, object_ in self.graph: log.info("\nsubj: {subj}\npred: {pred}\n obj: {obj}" .format(subj=subject, pred=predicate, obj=object_.encode('utf-8'))) result.add_error(self._check_triple((subject, predicate, object_))) return result
Iterate over all triples in the graph and validate each one appropriately
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/validator.py#L34-L55
[ "def add_error(self, warning):\n if warning:\n if warning.level == ValidationResult.WARNING:\n self.warnings.append(warning)\n elif warning.level == ValidationResult.ERROR:\n self.errors.append(warning)\n", "def _check_triple(self, triple):\n \"\"\"compare triple to ontology, return error or None\"\"\"\n subj, pred, obj = triple\n\n if self._should_ignore_predicate(pred):\n log.info(\"Ignoring triple with predicate '{}'\"\n .format(self._field_name_from_uri(pred)))\n return\n\n classes = []\n log.warning(\"Possible member %s found\" % pred)\n\n pred = self._expand_qname(pred)\n\n if self._namespace_from_uri(pred) not in self.allowed_namespaces:\n log.info(\"Member %s does not use an allowed namespace\", pred)\n return\n\n instanceof = self._is_instance((subj, pred, obj))\n if type(instanceof) == rt.URIRef:\n instanceof = self._expand_qname(instanceof)\n\n if hasattr(self.schema_def, \"attributes_by_class\") and \\\n not self.schema_def.attributes_by_class:\n log.info(\"Parsed ontology not found. Parsing...\")\n self.schema_def.parse_ontology()\n\n class_invalid = self._validate_class(instanceof)\n if class_invalid:\n log.warning(\"Invalid class %s\" % instanceof)\n return class_invalid\n # TODO - the above sometimes fails when a single object has more than\n # one rdfa type (eg <span property=\"schema:creator rnews:creator\"\n # typeof=\"schema:Person rnews:Person\">\n # Graph chooses the type in an arbitrary order, so it's unreliable\n # eg: http://semanticweb.com/the-impact-of-rdfa_b35003\n\n classes = self._superclasses_for_subject(self.graph, instanceof)\n classes.append(instanceof)\n\n member_invalid = self._validate_member(pred, classes, instanceof)\n if member_invalid:\n log.warning(\"Invalid member of class\")\n return member_invalid\n\n dupe_invalid = self._validate_duplication((subj, pred), instanceof)\n if dupe_invalid:\n log.warning(\"Duplication found\")\n return dupe_invalid\n\n # collect a list of checked attributes\n self.checked_attributes.append((subj, pred))\n log.warning(\"successfully validated triple, no errors\")\n return\n" ]
class SchemaValidator(object): """ASSUMPTIONS: This class knows about the file being validated, but it recieves that file as a graph and a doc_lines It does not perform any parsing logic on the file It recieves a "validatable" graph object and returns errors""" def __init__(self, graph, doc_lines, url=""): super(SchemaValidator, self).__init__() self.schema_def = None self.allowed_namespaces = [] self.graph = graph self.checked_attributes = [] self.stripped_attribute_names = {} log.info("init validator: %s" % self.__class__.__name__) self.doc_lines = doc_lines self.err = partial(_error, doc_lines=self.doc_lines) def _should_ignore_predicate(self, predicate): ignored_predicates = ['type', 'item', 'first', 'rest'] return self._field_name_from_uri(predicate) in ignored_predicates def _check_triple(self, triple): """compare triple to ontology, return error or None""" subj, pred, obj = triple if self._should_ignore_predicate(pred): log.info("Ignoring triple with predicate '{}'" .format(self._field_name_from_uri(pred))) return classes = [] log.warning("Possible member %s found" % pred) pred = self._expand_qname(pred) if self._namespace_from_uri(pred) not in self.allowed_namespaces: log.info("Member %s does not use an allowed namespace", pred) return instanceof = self._is_instance((subj, pred, obj)) if type(instanceof) == rt.URIRef: instanceof = self._expand_qname(instanceof) if hasattr(self.schema_def, "attributes_by_class") and \ not self.schema_def.attributes_by_class: log.info("Parsed ontology not found. Parsing...") self.schema_def.parse_ontology() class_invalid = self._validate_class(instanceof) if class_invalid: log.warning("Invalid class %s" % instanceof) return class_invalid # TODO - the above sometimes fails when a single object has more than # one rdfa type (eg <span property="schema:creator rnews:creator" # typeof="schema:Person rnews:Person"> # Graph chooses the type in an arbitrary order, so it's unreliable # eg: http://semanticweb.com/the-impact-of-rdfa_b35003 classes = self._superclasses_for_subject(self.graph, instanceof) classes.append(instanceof) member_invalid = self._validate_member(pred, classes, instanceof) if member_invalid: log.warning("Invalid member of class") return member_invalid dupe_invalid = self._validate_duplication((subj, pred), instanceof) if dupe_invalid: log.warning("Duplication found") return dupe_invalid # collect a list of checked attributes self.checked_attributes.append((subj, pred)) log.warning("successfully validated triple, no errors") return def _validate_class(self, cl): """return error if class `cl` is not found in the ontology""" if cl not in self.schema_def.attributes_by_class: search_string = self._build_search_string(cl) err = self.err( "{0} - invalid class", self._field_name_from_uri(cl), search_string=search_string) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _get_stripped_attributes(self, member, classes): stripped = [] if member not in self.stripped_attribute_names: for cl in classes: stripped.extend( [self._field_name_from_uri(attr) for attr in self.schema_def.attributes_by_class[cl]]) self.stripped_attribute_names[member] = stripped else: stripped = self.stripped_attribute_names[member] return stripped def _validate_member(self, member, classes, instanceof): """return error if `member` is not a member of any class in `classes` """ log.info("Validating member %s" % member) stripped = self._get_stripped_attributes(member, classes) if self._field_name_from_uri(member) in stripped: all_class_members = sum([self.schema_def.attributes_by_class[cl] for cl in classes], []) if member in all_class_members: return if self._namespace_from_uri(member) in self.allowed_namespaces: err = self.err("Unoficially allowed namespace {0}", self._namespace_from_uri(member)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) else: err = self.err("{0} - invalid member of {1}", self._field_name_from_uri(member), self._field_name_from_uri(instanceof)) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _validate_duplication(self, subj_and_pred, cl): """returns error if we've already seen the member `pred` on `subj`""" subj, pred = subj_and_pred log.info("Validating duplication of member %s" % pred) if (subj, pred) in self.checked_attributes: err = self.err("{0} - duplicated member of {1}", self._field_name_from_uri(pred), self._field_name_from_uri(cl)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) def _superclasses_for_subject(self, graph, typeof): """helper, returns a list of all superclasses of a given class""" # TODO - this might be replacing a fairly simple graph API query where # it doesn't need to classes = [] superclass = typeof while True: found = False for p, o in self.schema_def.ontology[superclass]: if self.schema_def.lexicon['subclass'] == str(p): found = True classes.append(o) superclass = o if not found: break return classes def _is_instance(self, triple): """helper, returns the class type of subj""" subj, pred, obj = triple input_pred_ns = self._namespace_from_uri(self._expand_qname(pred)) triples = self.graph.triples( (subj, rt.URIRef(self.schema_def.lexicon['type']), None) ) if triples: for tr in triples: triple_obj_ns = self._namespace_from_uri( self._expand_qname(tr[2])) if input_pred_ns == triple_obj_ns: # match namespaces return tr[2] # return the object def _field_name_from_uri(self, uri): """helper, returns the name of an attribute (without namespace prefix) """ # TODO - should use graph API uri = str(uri) parts = uri.split('#') if len(parts) == 1: return uri.split('/')[-1] or uri return parts[-1] def _namespace_from_uri(self, uri): """returns the expanded namespace prefix of a uri""" # TODO - this could be helped a bunch with proper use of the graph API # it seems a bit fragile to treat these as simple string-splits uri = str(uri) parts = uri.split('#') if len(parts) == 1: return "%s/" % '/'.join(uri.split('/')[:-1]) return "%s#" % '#'.join(parts[:-1]) def _expand_qname(self, qname): """expand a qualified name's namespace prefix to include the resolved namespace root url""" if type(qname) is not rt.URIRef: raise TypeError("Cannot expand qname of type {}, must be URIRef" .format(type(qname))) for ns in self.graph.namespaces(): if ns[0] == qname.split(':')[0]: return rt.URIRef("%s%s" % (ns[1], qname.split(':')[-1])) return qname def _build_search_string(self, uri): return self._field_name_from_uri(uri)
Parsely/schemato
schemato/validator.py
SchemaValidator._check_triple
python
def _check_triple(self, triple): subj, pred, obj = triple if self._should_ignore_predicate(pred): log.info("Ignoring triple with predicate '{}'" .format(self._field_name_from_uri(pred))) return classes = [] log.warning("Possible member %s found" % pred) pred = self._expand_qname(pred) if self._namespace_from_uri(pred) not in self.allowed_namespaces: log.info("Member %s does not use an allowed namespace", pred) return instanceof = self._is_instance((subj, pred, obj)) if type(instanceof) == rt.URIRef: instanceof = self._expand_qname(instanceof) if hasattr(self.schema_def, "attributes_by_class") and \ not self.schema_def.attributes_by_class: log.info("Parsed ontology not found. Parsing...") self.schema_def.parse_ontology() class_invalid = self._validate_class(instanceof) if class_invalid: log.warning("Invalid class %s" % instanceof) return class_invalid # TODO - the above sometimes fails when a single object has more than # one rdfa type (eg <span property="schema:creator rnews:creator" # typeof="schema:Person rnews:Person"> # Graph chooses the type in an arbitrary order, so it's unreliable # eg: http://semanticweb.com/the-impact-of-rdfa_b35003 classes = self._superclasses_for_subject(self.graph, instanceof) classes.append(instanceof) member_invalid = self._validate_member(pred, classes, instanceof) if member_invalid: log.warning("Invalid member of class") return member_invalid dupe_invalid = self._validate_duplication((subj, pred), instanceof) if dupe_invalid: log.warning("Duplication found") return dupe_invalid # collect a list of checked attributes self.checked_attributes.append((subj, pred)) log.warning("successfully validated triple, no errors") return
compare triple to ontology, return error or None
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/validator.py#L61-L114
[ "def _should_ignore_predicate(self, predicate):\n ignored_predicates = ['type', 'item', 'first', 'rest']\n return self._field_name_from_uri(predicate) in ignored_predicates\n" ]
class SchemaValidator(object): """ASSUMPTIONS: This class knows about the file being validated, but it recieves that file as a graph and a doc_lines It does not perform any parsing logic on the file It recieves a "validatable" graph object and returns errors""" def __init__(self, graph, doc_lines, url=""): super(SchemaValidator, self).__init__() self.schema_def = None self.allowed_namespaces = [] self.graph = graph self.checked_attributes = [] self.stripped_attribute_names = {} log.info("init validator: %s" % self.__class__.__name__) self.doc_lines = doc_lines self.err = partial(_error, doc_lines=self.doc_lines) def validate(self): """Iterate over all triples in the graph and validate each one appropriately """ log.info("{}\nValidating against {}" .format("-" * 100, self.schema_def.__class__.__name__)) if not self.schema_def: raise ValueError("No schema definition supplied.") self.checked_attributes = [] # TODO - this should maybe choose the actually used namespace, not just # the first one in the list result = ValidationResult(self.allowed_namespaces[0], self.schema_def.__class__.__name__) for subject, predicate, object_ in self.graph: log.info("\nsubj: {subj}\npred: {pred}\n obj: {obj}" .format(subj=subject, pred=predicate, obj=object_.encode('utf-8'))) result.add_error(self._check_triple((subject, predicate, object_))) return result def _should_ignore_predicate(self, predicate): ignored_predicates = ['type', 'item', 'first', 'rest'] return self._field_name_from_uri(predicate) in ignored_predicates def _validate_class(self, cl): """return error if class `cl` is not found in the ontology""" if cl not in self.schema_def.attributes_by_class: search_string = self._build_search_string(cl) err = self.err( "{0} - invalid class", self._field_name_from_uri(cl), search_string=search_string) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _get_stripped_attributes(self, member, classes): stripped = [] if member not in self.stripped_attribute_names: for cl in classes: stripped.extend( [self._field_name_from_uri(attr) for attr in self.schema_def.attributes_by_class[cl]]) self.stripped_attribute_names[member] = stripped else: stripped = self.stripped_attribute_names[member] return stripped def _validate_member(self, member, classes, instanceof): """return error if `member` is not a member of any class in `classes` """ log.info("Validating member %s" % member) stripped = self._get_stripped_attributes(member, classes) if self._field_name_from_uri(member) in stripped: all_class_members = sum([self.schema_def.attributes_by_class[cl] for cl in classes], []) if member in all_class_members: return if self._namespace_from_uri(member) in self.allowed_namespaces: err = self.err("Unoficially allowed namespace {0}", self._namespace_from_uri(member)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) else: err = self.err("{0} - invalid member of {1}", self._field_name_from_uri(member), self._field_name_from_uri(instanceof)) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _validate_duplication(self, subj_and_pred, cl): """returns error if we've already seen the member `pred` on `subj`""" subj, pred = subj_and_pred log.info("Validating duplication of member %s" % pred) if (subj, pred) in self.checked_attributes: err = self.err("{0} - duplicated member of {1}", self._field_name_from_uri(pred), self._field_name_from_uri(cl)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) def _superclasses_for_subject(self, graph, typeof): """helper, returns a list of all superclasses of a given class""" # TODO - this might be replacing a fairly simple graph API query where # it doesn't need to classes = [] superclass = typeof while True: found = False for p, o in self.schema_def.ontology[superclass]: if self.schema_def.lexicon['subclass'] == str(p): found = True classes.append(o) superclass = o if not found: break return classes def _is_instance(self, triple): """helper, returns the class type of subj""" subj, pred, obj = triple input_pred_ns = self._namespace_from_uri(self._expand_qname(pred)) triples = self.graph.triples( (subj, rt.URIRef(self.schema_def.lexicon['type']), None) ) if triples: for tr in triples: triple_obj_ns = self._namespace_from_uri( self._expand_qname(tr[2])) if input_pred_ns == triple_obj_ns: # match namespaces return tr[2] # return the object def _field_name_from_uri(self, uri): """helper, returns the name of an attribute (without namespace prefix) """ # TODO - should use graph API uri = str(uri) parts = uri.split('#') if len(parts) == 1: return uri.split('/')[-1] or uri return parts[-1] def _namespace_from_uri(self, uri): """returns the expanded namespace prefix of a uri""" # TODO - this could be helped a bunch with proper use of the graph API # it seems a bit fragile to treat these as simple string-splits uri = str(uri) parts = uri.split('#') if len(parts) == 1: return "%s/" % '/'.join(uri.split('/')[:-1]) return "%s#" % '#'.join(parts[:-1]) def _expand_qname(self, qname): """expand a qualified name's namespace prefix to include the resolved namespace root url""" if type(qname) is not rt.URIRef: raise TypeError("Cannot expand qname of type {}, must be URIRef" .format(type(qname))) for ns in self.graph.namespaces(): if ns[0] == qname.split(':')[0]: return rt.URIRef("%s%s" % (ns[1], qname.split(':')[-1])) return qname def _build_search_string(self, uri): return self._field_name_from_uri(uri)
Parsely/schemato
schemato/validator.py
SchemaValidator._validate_class
python
def _validate_class(self, cl): if cl not in self.schema_def.attributes_by_class: search_string = self._build_search_string(cl) err = self.err( "{0} - invalid class", self._field_name_from_uri(cl), search_string=search_string) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num'])
return error if class `cl` is not found in the ontology
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/validator.py#L116-L124
null
class SchemaValidator(object): """ASSUMPTIONS: This class knows about the file being validated, but it recieves that file as a graph and a doc_lines It does not perform any parsing logic on the file It recieves a "validatable" graph object and returns errors""" def __init__(self, graph, doc_lines, url=""): super(SchemaValidator, self).__init__() self.schema_def = None self.allowed_namespaces = [] self.graph = graph self.checked_attributes = [] self.stripped_attribute_names = {} log.info("init validator: %s" % self.__class__.__name__) self.doc_lines = doc_lines self.err = partial(_error, doc_lines=self.doc_lines) def validate(self): """Iterate over all triples in the graph and validate each one appropriately """ log.info("{}\nValidating against {}" .format("-" * 100, self.schema_def.__class__.__name__)) if not self.schema_def: raise ValueError("No schema definition supplied.") self.checked_attributes = [] # TODO - this should maybe choose the actually used namespace, not just # the first one in the list result = ValidationResult(self.allowed_namespaces[0], self.schema_def.__class__.__name__) for subject, predicate, object_ in self.graph: log.info("\nsubj: {subj}\npred: {pred}\n obj: {obj}" .format(subj=subject, pred=predicate, obj=object_.encode('utf-8'))) result.add_error(self._check_triple((subject, predicate, object_))) return result def _should_ignore_predicate(self, predicate): ignored_predicates = ['type', 'item', 'first', 'rest'] return self._field_name_from_uri(predicate) in ignored_predicates def _check_triple(self, triple): """compare triple to ontology, return error or None""" subj, pred, obj = triple if self._should_ignore_predicate(pred): log.info("Ignoring triple with predicate '{}'" .format(self._field_name_from_uri(pred))) return classes = [] log.warning("Possible member %s found" % pred) pred = self._expand_qname(pred) if self._namespace_from_uri(pred) not in self.allowed_namespaces: log.info("Member %s does not use an allowed namespace", pred) return instanceof = self._is_instance((subj, pred, obj)) if type(instanceof) == rt.URIRef: instanceof = self._expand_qname(instanceof) if hasattr(self.schema_def, "attributes_by_class") and \ not self.schema_def.attributes_by_class: log.info("Parsed ontology not found. Parsing...") self.schema_def.parse_ontology() class_invalid = self._validate_class(instanceof) if class_invalid: log.warning("Invalid class %s" % instanceof) return class_invalid # TODO - the above sometimes fails when a single object has more than # one rdfa type (eg <span property="schema:creator rnews:creator" # typeof="schema:Person rnews:Person"> # Graph chooses the type in an arbitrary order, so it's unreliable # eg: http://semanticweb.com/the-impact-of-rdfa_b35003 classes = self._superclasses_for_subject(self.graph, instanceof) classes.append(instanceof) member_invalid = self._validate_member(pred, classes, instanceof) if member_invalid: log.warning("Invalid member of class") return member_invalid dupe_invalid = self._validate_duplication((subj, pred), instanceof) if dupe_invalid: log.warning("Duplication found") return dupe_invalid # collect a list of checked attributes self.checked_attributes.append((subj, pred)) log.warning("successfully validated triple, no errors") return def _get_stripped_attributes(self, member, classes): stripped = [] if member not in self.stripped_attribute_names: for cl in classes: stripped.extend( [self._field_name_from_uri(attr) for attr in self.schema_def.attributes_by_class[cl]]) self.stripped_attribute_names[member] = stripped else: stripped = self.stripped_attribute_names[member] return stripped def _validate_member(self, member, classes, instanceof): """return error if `member` is not a member of any class in `classes` """ log.info("Validating member %s" % member) stripped = self._get_stripped_attributes(member, classes) if self._field_name_from_uri(member) in stripped: all_class_members = sum([self.schema_def.attributes_by_class[cl] for cl in classes], []) if member in all_class_members: return if self._namespace_from_uri(member) in self.allowed_namespaces: err = self.err("Unoficially allowed namespace {0}", self._namespace_from_uri(member)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) else: err = self.err("{0} - invalid member of {1}", self._field_name_from_uri(member), self._field_name_from_uri(instanceof)) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _validate_duplication(self, subj_and_pred, cl): """returns error if we've already seen the member `pred` on `subj`""" subj, pred = subj_and_pred log.info("Validating duplication of member %s" % pred) if (subj, pred) in self.checked_attributes: err = self.err("{0} - duplicated member of {1}", self._field_name_from_uri(pred), self._field_name_from_uri(cl)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) def _superclasses_for_subject(self, graph, typeof): """helper, returns a list of all superclasses of a given class""" # TODO - this might be replacing a fairly simple graph API query where # it doesn't need to classes = [] superclass = typeof while True: found = False for p, o in self.schema_def.ontology[superclass]: if self.schema_def.lexicon['subclass'] == str(p): found = True classes.append(o) superclass = o if not found: break return classes def _is_instance(self, triple): """helper, returns the class type of subj""" subj, pred, obj = triple input_pred_ns = self._namespace_from_uri(self._expand_qname(pred)) triples = self.graph.triples( (subj, rt.URIRef(self.schema_def.lexicon['type']), None) ) if triples: for tr in triples: triple_obj_ns = self._namespace_from_uri( self._expand_qname(tr[2])) if input_pred_ns == triple_obj_ns: # match namespaces return tr[2] # return the object def _field_name_from_uri(self, uri): """helper, returns the name of an attribute (without namespace prefix) """ # TODO - should use graph API uri = str(uri) parts = uri.split('#') if len(parts) == 1: return uri.split('/')[-1] or uri return parts[-1] def _namespace_from_uri(self, uri): """returns the expanded namespace prefix of a uri""" # TODO - this could be helped a bunch with proper use of the graph API # it seems a bit fragile to treat these as simple string-splits uri = str(uri) parts = uri.split('#') if len(parts) == 1: return "%s/" % '/'.join(uri.split('/')[:-1]) return "%s#" % '#'.join(parts[:-1]) def _expand_qname(self, qname): """expand a qualified name's namespace prefix to include the resolved namespace root url""" if type(qname) is not rt.URIRef: raise TypeError("Cannot expand qname of type {}, must be URIRef" .format(type(qname))) for ns in self.graph.namespaces(): if ns[0] == qname.split(':')[0]: return rt.URIRef("%s%s" % (ns[1], qname.split(':')[-1])) return qname def _build_search_string(self, uri): return self._field_name_from_uri(uri)
Parsely/schemato
schemato/validator.py
SchemaValidator._validate_member
python
def _validate_member(self, member, classes, instanceof): log.info("Validating member %s" % member) stripped = self._get_stripped_attributes(member, classes) if self._field_name_from_uri(member) in stripped: all_class_members = sum([self.schema_def.attributes_by_class[cl] for cl in classes], []) if member in all_class_members: return if self._namespace_from_uri(member) in self.allowed_namespaces: err = self.err("Unoficially allowed namespace {0}", self._namespace_from_uri(member)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) else: err = self.err("{0} - invalid member of {1}", self._field_name_from_uri(member), self._field_name_from_uri(instanceof)) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num'])
return error if `member` is not a member of any class in `classes`
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/validator.py#L138-L160
null
class SchemaValidator(object): """ASSUMPTIONS: This class knows about the file being validated, but it recieves that file as a graph and a doc_lines It does not perform any parsing logic on the file It recieves a "validatable" graph object and returns errors""" def __init__(self, graph, doc_lines, url=""): super(SchemaValidator, self).__init__() self.schema_def = None self.allowed_namespaces = [] self.graph = graph self.checked_attributes = [] self.stripped_attribute_names = {} log.info("init validator: %s" % self.__class__.__name__) self.doc_lines = doc_lines self.err = partial(_error, doc_lines=self.doc_lines) def validate(self): """Iterate over all triples in the graph and validate each one appropriately """ log.info("{}\nValidating against {}" .format("-" * 100, self.schema_def.__class__.__name__)) if not self.schema_def: raise ValueError("No schema definition supplied.") self.checked_attributes = [] # TODO - this should maybe choose the actually used namespace, not just # the first one in the list result = ValidationResult(self.allowed_namespaces[0], self.schema_def.__class__.__name__) for subject, predicate, object_ in self.graph: log.info("\nsubj: {subj}\npred: {pred}\n obj: {obj}" .format(subj=subject, pred=predicate, obj=object_.encode('utf-8'))) result.add_error(self._check_triple((subject, predicate, object_))) return result def _should_ignore_predicate(self, predicate): ignored_predicates = ['type', 'item', 'first', 'rest'] return self._field_name_from_uri(predicate) in ignored_predicates def _check_triple(self, triple): """compare triple to ontology, return error or None""" subj, pred, obj = triple if self._should_ignore_predicate(pred): log.info("Ignoring triple with predicate '{}'" .format(self._field_name_from_uri(pred))) return classes = [] log.warning("Possible member %s found" % pred) pred = self._expand_qname(pred) if self._namespace_from_uri(pred) not in self.allowed_namespaces: log.info("Member %s does not use an allowed namespace", pred) return instanceof = self._is_instance((subj, pred, obj)) if type(instanceof) == rt.URIRef: instanceof = self._expand_qname(instanceof) if hasattr(self.schema_def, "attributes_by_class") and \ not self.schema_def.attributes_by_class: log.info("Parsed ontology not found. Parsing...") self.schema_def.parse_ontology() class_invalid = self._validate_class(instanceof) if class_invalid: log.warning("Invalid class %s" % instanceof) return class_invalid # TODO - the above sometimes fails when a single object has more than # one rdfa type (eg <span property="schema:creator rnews:creator" # typeof="schema:Person rnews:Person"> # Graph chooses the type in an arbitrary order, so it's unreliable # eg: http://semanticweb.com/the-impact-of-rdfa_b35003 classes = self._superclasses_for_subject(self.graph, instanceof) classes.append(instanceof) member_invalid = self._validate_member(pred, classes, instanceof) if member_invalid: log.warning("Invalid member of class") return member_invalid dupe_invalid = self._validate_duplication((subj, pred), instanceof) if dupe_invalid: log.warning("Duplication found") return dupe_invalid # collect a list of checked attributes self.checked_attributes.append((subj, pred)) log.warning("successfully validated triple, no errors") return def _validate_class(self, cl): """return error if class `cl` is not found in the ontology""" if cl not in self.schema_def.attributes_by_class: search_string = self._build_search_string(cl) err = self.err( "{0} - invalid class", self._field_name_from_uri(cl), search_string=search_string) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _get_stripped_attributes(self, member, classes): stripped = [] if member not in self.stripped_attribute_names: for cl in classes: stripped.extend( [self._field_name_from_uri(attr) for attr in self.schema_def.attributes_by_class[cl]]) self.stripped_attribute_names[member] = stripped else: stripped = self.stripped_attribute_names[member] return stripped def _validate_duplication(self, subj_and_pred, cl): """returns error if we've already seen the member `pred` on `subj`""" subj, pred = subj_and_pred log.info("Validating duplication of member %s" % pred) if (subj, pred) in self.checked_attributes: err = self.err("{0} - duplicated member of {1}", self._field_name_from_uri(pred), self._field_name_from_uri(cl)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) def _superclasses_for_subject(self, graph, typeof): """helper, returns a list of all superclasses of a given class""" # TODO - this might be replacing a fairly simple graph API query where # it doesn't need to classes = [] superclass = typeof while True: found = False for p, o in self.schema_def.ontology[superclass]: if self.schema_def.lexicon['subclass'] == str(p): found = True classes.append(o) superclass = o if not found: break return classes def _is_instance(self, triple): """helper, returns the class type of subj""" subj, pred, obj = triple input_pred_ns = self._namespace_from_uri(self._expand_qname(pred)) triples = self.graph.triples( (subj, rt.URIRef(self.schema_def.lexicon['type']), None) ) if triples: for tr in triples: triple_obj_ns = self._namespace_from_uri( self._expand_qname(tr[2])) if input_pred_ns == triple_obj_ns: # match namespaces return tr[2] # return the object def _field_name_from_uri(self, uri): """helper, returns the name of an attribute (without namespace prefix) """ # TODO - should use graph API uri = str(uri) parts = uri.split('#') if len(parts) == 1: return uri.split('/')[-1] or uri return parts[-1] def _namespace_from_uri(self, uri): """returns the expanded namespace prefix of a uri""" # TODO - this could be helped a bunch with proper use of the graph API # it seems a bit fragile to treat these as simple string-splits uri = str(uri) parts = uri.split('#') if len(parts) == 1: return "%s/" % '/'.join(uri.split('/')[:-1]) return "%s#" % '#'.join(parts[:-1]) def _expand_qname(self, qname): """expand a qualified name's namespace prefix to include the resolved namespace root url""" if type(qname) is not rt.URIRef: raise TypeError("Cannot expand qname of type {}, must be URIRef" .format(type(qname))) for ns in self.graph.namespaces(): if ns[0] == qname.split(':')[0]: return rt.URIRef("%s%s" % (ns[1], qname.split(':')[-1])) return qname def _build_search_string(self, uri): return self._field_name_from_uri(uri)
Parsely/schemato
schemato/validator.py
SchemaValidator._validate_duplication
python
def _validate_duplication(self, subj_and_pred, cl): subj, pred = subj_and_pred log.info("Validating duplication of member %s" % pred) if (subj, pred) in self.checked_attributes: err = self.err("{0} - duplicated member of {1}", self._field_name_from_uri(pred), self._field_name_from_uri(cl)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num'])
returns error if we've already seen the member `pred` on `subj`
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/validator.py#L162-L172
null
class SchemaValidator(object): """ASSUMPTIONS: This class knows about the file being validated, but it recieves that file as a graph and a doc_lines It does not perform any parsing logic on the file It recieves a "validatable" graph object and returns errors""" def __init__(self, graph, doc_lines, url=""): super(SchemaValidator, self).__init__() self.schema_def = None self.allowed_namespaces = [] self.graph = graph self.checked_attributes = [] self.stripped_attribute_names = {} log.info("init validator: %s" % self.__class__.__name__) self.doc_lines = doc_lines self.err = partial(_error, doc_lines=self.doc_lines) def validate(self): """Iterate over all triples in the graph and validate each one appropriately """ log.info("{}\nValidating against {}" .format("-" * 100, self.schema_def.__class__.__name__)) if not self.schema_def: raise ValueError("No schema definition supplied.") self.checked_attributes = [] # TODO - this should maybe choose the actually used namespace, not just # the first one in the list result = ValidationResult(self.allowed_namespaces[0], self.schema_def.__class__.__name__) for subject, predicate, object_ in self.graph: log.info("\nsubj: {subj}\npred: {pred}\n obj: {obj}" .format(subj=subject, pred=predicate, obj=object_.encode('utf-8'))) result.add_error(self._check_triple((subject, predicate, object_))) return result def _should_ignore_predicate(self, predicate): ignored_predicates = ['type', 'item', 'first', 'rest'] return self._field_name_from_uri(predicate) in ignored_predicates def _check_triple(self, triple): """compare triple to ontology, return error or None""" subj, pred, obj = triple if self._should_ignore_predicate(pred): log.info("Ignoring triple with predicate '{}'" .format(self._field_name_from_uri(pred))) return classes = [] log.warning("Possible member %s found" % pred) pred = self._expand_qname(pred) if self._namespace_from_uri(pred) not in self.allowed_namespaces: log.info("Member %s does not use an allowed namespace", pred) return instanceof = self._is_instance((subj, pred, obj)) if type(instanceof) == rt.URIRef: instanceof = self._expand_qname(instanceof) if hasattr(self.schema_def, "attributes_by_class") and \ not self.schema_def.attributes_by_class: log.info("Parsed ontology not found. Parsing...") self.schema_def.parse_ontology() class_invalid = self._validate_class(instanceof) if class_invalid: log.warning("Invalid class %s" % instanceof) return class_invalid # TODO - the above sometimes fails when a single object has more than # one rdfa type (eg <span property="schema:creator rnews:creator" # typeof="schema:Person rnews:Person"> # Graph chooses the type in an arbitrary order, so it's unreliable # eg: http://semanticweb.com/the-impact-of-rdfa_b35003 classes = self._superclasses_for_subject(self.graph, instanceof) classes.append(instanceof) member_invalid = self._validate_member(pred, classes, instanceof) if member_invalid: log.warning("Invalid member of class") return member_invalid dupe_invalid = self._validate_duplication((subj, pred), instanceof) if dupe_invalid: log.warning("Duplication found") return dupe_invalid # collect a list of checked attributes self.checked_attributes.append((subj, pred)) log.warning("successfully validated triple, no errors") return def _validate_class(self, cl): """return error if class `cl` is not found in the ontology""" if cl not in self.schema_def.attributes_by_class: search_string = self._build_search_string(cl) err = self.err( "{0} - invalid class", self._field_name_from_uri(cl), search_string=search_string) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _get_stripped_attributes(self, member, classes): stripped = [] if member not in self.stripped_attribute_names: for cl in classes: stripped.extend( [self._field_name_from_uri(attr) for attr in self.schema_def.attributes_by_class[cl]]) self.stripped_attribute_names[member] = stripped else: stripped = self.stripped_attribute_names[member] return stripped def _validate_member(self, member, classes, instanceof): """return error if `member` is not a member of any class in `classes` """ log.info("Validating member %s" % member) stripped = self._get_stripped_attributes(member, classes) if self._field_name_from_uri(member) in stripped: all_class_members = sum([self.schema_def.attributes_by_class[cl] for cl in classes], []) if member in all_class_members: return if self._namespace_from_uri(member) in self.allowed_namespaces: err = self.err("Unoficially allowed namespace {0}", self._namespace_from_uri(member)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) else: err = self.err("{0} - invalid member of {1}", self._field_name_from_uri(member), self._field_name_from_uri(instanceof)) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _superclasses_for_subject(self, graph, typeof): """helper, returns a list of all superclasses of a given class""" # TODO - this might be replacing a fairly simple graph API query where # it doesn't need to classes = [] superclass = typeof while True: found = False for p, o in self.schema_def.ontology[superclass]: if self.schema_def.lexicon['subclass'] == str(p): found = True classes.append(o) superclass = o if not found: break return classes def _is_instance(self, triple): """helper, returns the class type of subj""" subj, pred, obj = triple input_pred_ns = self._namespace_from_uri(self._expand_qname(pred)) triples = self.graph.triples( (subj, rt.URIRef(self.schema_def.lexicon['type']), None) ) if triples: for tr in triples: triple_obj_ns = self._namespace_from_uri( self._expand_qname(tr[2])) if input_pred_ns == triple_obj_ns: # match namespaces return tr[2] # return the object def _field_name_from_uri(self, uri): """helper, returns the name of an attribute (without namespace prefix) """ # TODO - should use graph API uri = str(uri) parts = uri.split('#') if len(parts) == 1: return uri.split('/')[-1] or uri return parts[-1] def _namespace_from_uri(self, uri): """returns the expanded namespace prefix of a uri""" # TODO - this could be helped a bunch with proper use of the graph API # it seems a bit fragile to treat these as simple string-splits uri = str(uri) parts = uri.split('#') if len(parts) == 1: return "%s/" % '/'.join(uri.split('/')[:-1]) return "%s#" % '#'.join(parts[:-1]) def _expand_qname(self, qname): """expand a qualified name's namespace prefix to include the resolved namespace root url""" if type(qname) is not rt.URIRef: raise TypeError("Cannot expand qname of type {}, must be URIRef" .format(type(qname))) for ns in self.graph.namespaces(): if ns[0] == qname.split(':')[0]: return rt.URIRef("%s%s" % (ns[1], qname.split(':')[-1])) return qname def _build_search_string(self, uri): return self._field_name_from_uri(uri)
Parsely/schemato
schemato/validator.py
SchemaValidator._superclasses_for_subject
python
def _superclasses_for_subject(self, graph, typeof): # TODO - this might be replacing a fairly simple graph API query where # it doesn't need to classes = [] superclass = typeof while True: found = False for p, o in self.schema_def.ontology[superclass]: if self.schema_def.lexicon['subclass'] == str(p): found = True classes.append(o) superclass = o if not found: break return classes
helper, returns a list of all superclasses of a given class
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/validator.py#L174-L189
null
class SchemaValidator(object): """ASSUMPTIONS: This class knows about the file being validated, but it recieves that file as a graph and a doc_lines It does not perform any parsing logic on the file It recieves a "validatable" graph object and returns errors""" def __init__(self, graph, doc_lines, url=""): super(SchemaValidator, self).__init__() self.schema_def = None self.allowed_namespaces = [] self.graph = graph self.checked_attributes = [] self.stripped_attribute_names = {} log.info("init validator: %s" % self.__class__.__name__) self.doc_lines = doc_lines self.err = partial(_error, doc_lines=self.doc_lines) def validate(self): """Iterate over all triples in the graph and validate each one appropriately """ log.info("{}\nValidating against {}" .format("-" * 100, self.schema_def.__class__.__name__)) if not self.schema_def: raise ValueError("No schema definition supplied.") self.checked_attributes = [] # TODO - this should maybe choose the actually used namespace, not just # the first one in the list result = ValidationResult(self.allowed_namespaces[0], self.schema_def.__class__.__name__) for subject, predicate, object_ in self.graph: log.info("\nsubj: {subj}\npred: {pred}\n obj: {obj}" .format(subj=subject, pred=predicate, obj=object_.encode('utf-8'))) result.add_error(self._check_triple((subject, predicate, object_))) return result def _should_ignore_predicate(self, predicate): ignored_predicates = ['type', 'item', 'first', 'rest'] return self._field_name_from_uri(predicate) in ignored_predicates def _check_triple(self, triple): """compare triple to ontology, return error or None""" subj, pred, obj = triple if self._should_ignore_predicate(pred): log.info("Ignoring triple with predicate '{}'" .format(self._field_name_from_uri(pred))) return classes = [] log.warning("Possible member %s found" % pred) pred = self._expand_qname(pred) if self._namespace_from_uri(pred) not in self.allowed_namespaces: log.info("Member %s does not use an allowed namespace", pred) return instanceof = self._is_instance((subj, pred, obj)) if type(instanceof) == rt.URIRef: instanceof = self._expand_qname(instanceof) if hasattr(self.schema_def, "attributes_by_class") and \ not self.schema_def.attributes_by_class: log.info("Parsed ontology not found. Parsing...") self.schema_def.parse_ontology() class_invalid = self._validate_class(instanceof) if class_invalid: log.warning("Invalid class %s" % instanceof) return class_invalid # TODO - the above sometimes fails when a single object has more than # one rdfa type (eg <span property="schema:creator rnews:creator" # typeof="schema:Person rnews:Person"> # Graph chooses the type in an arbitrary order, so it's unreliable # eg: http://semanticweb.com/the-impact-of-rdfa_b35003 classes = self._superclasses_for_subject(self.graph, instanceof) classes.append(instanceof) member_invalid = self._validate_member(pred, classes, instanceof) if member_invalid: log.warning("Invalid member of class") return member_invalid dupe_invalid = self._validate_duplication((subj, pred), instanceof) if dupe_invalid: log.warning("Duplication found") return dupe_invalid # collect a list of checked attributes self.checked_attributes.append((subj, pred)) log.warning("successfully validated triple, no errors") return def _validate_class(self, cl): """return error if class `cl` is not found in the ontology""" if cl not in self.schema_def.attributes_by_class: search_string = self._build_search_string(cl) err = self.err( "{0} - invalid class", self._field_name_from_uri(cl), search_string=search_string) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _get_stripped_attributes(self, member, classes): stripped = [] if member not in self.stripped_attribute_names: for cl in classes: stripped.extend( [self._field_name_from_uri(attr) for attr in self.schema_def.attributes_by_class[cl]]) self.stripped_attribute_names[member] = stripped else: stripped = self.stripped_attribute_names[member] return stripped def _validate_member(self, member, classes, instanceof): """return error if `member` is not a member of any class in `classes` """ log.info("Validating member %s" % member) stripped = self._get_stripped_attributes(member, classes) if self._field_name_from_uri(member) in stripped: all_class_members = sum([self.schema_def.attributes_by_class[cl] for cl in classes], []) if member in all_class_members: return if self._namespace_from_uri(member) in self.allowed_namespaces: err = self.err("Unoficially allowed namespace {0}", self._namespace_from_uri(member)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) else: err = self.err("{0} - invalid member of {1}", self._field_name_from_uri(member), self._field_name_from_uri(instanceof)) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _validate_duplication(self, subj_and_pred, cl): """returns error if we've already seen the member `pred` on `subj`""" subj, pred = subj_and_pred log.info("Validating duplication of member %s" % pred) if (subj, pred) in self.checked_attributes: err = self.err("{0} - duplicated member of {1}", self._field_name_from_uri(pred), self._field_name_from_uri(cl)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) def _is_instance(self, triple): """helper, returns the class type of subj""" subj, pred, obj = triple input_pred_ns = self._namespace_from_uri(self._expand_qname(pred)) triples = self.graph.triples( (subj, rt.URIRef(self.schema_def.lexicon['type']), None) ) if triples: for tr in triples: triple_obj_ns = self._namespace_from_uri( self._expand_qname(tr[2])) if input_pred_ns == triple_obj_ns: # match namespaces return tr[2] # return the object def _field_name_from_uri(self, uri): """helper, returns the name of an attribute (without namespace prefix) """ # TODO - should use graph API uri = str(uri) parts = uri.split('#') if len(parts) == 1: return uri.split('/')[-1] or uri return parts[-1] def _namespace_from_uri(self, uri): """returns the expanded namespace prefix of a uri""" # TODO - this could be helped a bunch with proper use of the graph API # it seems a bit fragile to treat these as simple string-splits uri = str(uri) parts = uri.split('#') if len(parts) == 1: return "%s/" % '/'.join(uri.split('/')[:-1]) return "%s#" % '#'.join(parts[:-1]) def _expand_qname(self, qname): """expand a qualified name's namespace prefix to include the resolved namespace root url""" if type(qname) is not rt.URIRef: raise TypeError("Cannot expand qname of type {}, must be URIRef" .format(type(qname))) for ns in self.graph.namespaces(): if ns[0] == qname.split(':')[0]: return rt.URIRef("%s%s" % (ns[1], qname.split(':')[-1])) return qname def _build_search_string(self, uri): return self._field_name_from_uri(uri)
Parsely/schemato
schemato/validator.py
SchemaValidator._is_instance
python
def _is_instance(self, triple): subj, pred, obj = triple input_pred_ns = self._namespace_from_uri(self._expand_qname(pred)) triples = self.graph.triples( (subj, rt.URIRef(self.schema_def.lexicon['type']), None) ) if triples: for tr in triples: triple_obj_ns = self._namespace_from_uri( self._expand_qname(tr[2])) if input_pred_ns == triple_obj_ns: # match namespaces return tr[2]
helper, returns the class type of subj
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/validator.py#L191-L203
null
class SchemaValidator(object): """ASSUMPTIONS: This class knows about the file being validated, but it recieves that file as a graph and a doc_lines It does not perform any parsing logic on the file It recieves a "validatable" graph object and returns errors""" def __init__(self, graph, doc_lines, url=""): super(SchemaValidator, self).__init__() self.schema_def = None self.allowed_namespaces = [] self.graph = graph self.checked_attributes = [] self.stripped_attribute_names = {} log.info("init validator: %s" % self.__class__.__name__) self.doc_lines = doc_lines self.err = partial(_error, doc_lines=self.doc_lines) def validate(self): """Iterate over all triples in the graph and validate each one appropriately """ log.info("{}\nValidating against {}" .format("-" * 100, self.schema_def.__class__.__name__)) if not self.schema_def: raise ValueError("No schema definition supplied.") self.checked_attributes = [] # TODO - this should maybe choose the actually used namespace, not just # the first one in the list result = ValidationResult(self.allowed_namespaces[0], self.schema_def.__class__.__name__) for subject, predicate, object_ in self.graph: log.info("\nsubj: {subj}\npred: {pred}\n obj: {obj}" .format(subj=subject, pred=predicate, obj=object_.encode('utf-8'))) result.add_error(self._check_triple((subject, predicate, object_))) return result def _should_ignore_predicate(self, predicate): ignored_predicates = ['type', 'item', 'first', 'rest'] return self._field_name_from_uri(predicate) in ignored_predicates def _check_triple(self, triple): """compare triple to ontology, return error or None""" subj, pred, obj = triple if self._should_ignore_predicate(pred): log.info("Ignoring triple with predicate '{}'" .format(self._field_name_from_uri(pred))) return classes = [] log.warning("Possible member %s found" % pred) pred = self._expand_qname(pred) if self._namespace_from_uri(pred) not in self.allowed_namespaces: log.info("Member %s does not use an allowed namespace", pred) return instanceof = self._is_instance((subj, pred, obj)) if type(instanceof) == rt.URIRef: instanceof = self._expand_qname(instanceof) if hasattr(self.schema_def, "attributes_by_class") and \ not self.schema_def.attributes_by_class: log.info("Parsed ontology not found. Parsing...") self.schema_def.parse_ontology() class_invalid = self._validate_class(instanceof) if class_invalid: log.warning("Invalid class %s" % instanceof) return class_invalid # TODO - the above sometimes fails when a single object has more than # one rdfa type (eg <span property="schema:creator rnews:creator" # typeof="schema:Person rnews:Person"> # Graph chooses the type in an arbitrary order, so it's unreliable # eg: http://semanticweb.com/the-impact-of-rdfa_b35003 classes = self._superclasses_for_subject(self.graph, instanceof) classes.append(instanceof) member_invalid = self._validate_member(pred, classes, instanceof) if member_invalid: log.warning("Invalid member of class") return member_invalid dupe_invalid = self._validate_duplication((subj, pred), instanceof) if dupe_invalid: log.warning("Duplication found") return dupe_invalid # collect a list of checked attributes self.checked_attributes.append((subj, pred)) log.warning("successfully validated triple, no errors") return def _validate_class(self, cl): """return error if class `cl` is not found in the ontology""" if cl not in self.schema_def.attributes_by_class: search_string = self._build_search_string(cl) err = self.err( "{0} - invalid class", self._field_name_from_uri(cl), search_string=search_string) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _get_stripped_attributes(self, member, classes): stripped = [] if member not in self.stripped_attribute_names: for cl in classes: stripped.extend( [self._field_name_from_uri(attr) for attr in self.schema_def.attributes_by_class[cl]]) self.stripped_attribute_names[member] = stripped else: stripped = self.stripped_attribute_names[member] return stripped def _validate_member(self, member, classes, instanceof): """return error if `member` is not a member of any class in `classes` """ log.info("Validating member %s" % member) stripped = self._get_stripped_attributes(member, classes) if self._field_name_from_uri(member) in stripped: all_class_members = sum([self.schema_def.attributes_by_class[cl] for cl in classes], []) if member in all_class_members: return if self._namespace_from_uri(member) in self.allowed_namespaces: err = self.err("Unoficially allowed namespace {0}", self._namespace_from_uri(member)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) else: err = self.err("{0} - invalid member of {1}", self._field_name_from_uri(member), self._field_name_from_uri(instanceof)) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _validate_duplication(self, subj_and_pred, cl): """returns error if we've already seen the member `pred` on `subj`""" subj, pred = subj_and_pred log.info("Validating duplication of member %s" % pred) if (subj, pred) in self.checked_attributes: err = self.err("{0} - duplicated member of {1}", self._field_name_from_uri(pred), self._field_name_from_uri(cl)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) def _superclasses_for_subject(self, graph, typeof): """helper, returns a list of all superclasses of a given class""" # TODO - this might be replacing a fairly simple graph API query where # it doesn't need to classes = [] superclass = typeof while True: found = False for p, o in self.schema_def.ontology[superclass]: if self.schema_def.lexicon['subclass'] == str(p): found = True classes.append(o) superclass = o if not found: break return classes # return the object def _field_name_from_uri(self, uri): """helper, returns the name of an attribute (without namespace prefix) """ # TODO - should use graph API uri = str(uri) parts = uri.split('#') if len(parts) == 1: return uri.split('/')[-1] or uri return parts[-1] def _namespace_from_uri(self, uri): """returns the expanded namespace prefix of a uri""" # TODO - this could be helped a bunch with proper use of the graph API # it seems a bit fragile to treat these as simple string-splits uri = str(uri) parts = uri.split('#') if len(parts) == 1: return "%s/" % '/'.join(uri.split('/')[:-1]) return "%s#" % '#'.join(parts[:-1]) def _expand_qname(self, qname): """expand a qualified name's namespace prefix to include the resolved namespace root url""" if type(qname) is not rt.URIRef: raise TypeError("Cannot expand qname of type {}, must be URIRef" .format(type(qname))) for ns in self.graph.namespaces(): if ns[0] == qname.split(':')[0]: return rt.URIRef("%s%s" % (ns[1], qname.split(':')[-1])) return qname def _build_search_string(self, uri): return self._field_name_from_uri(uri)
Parsely/schemato
schemato/validator.py
SchemaValidator._field_name_from_uri
python
def _field_name_from_uri(self, uri): # TODO - should use graph API uri = str(uri) parts = uri.split('#') if len(parts) == 1: return uri.split('/')[-1] or uri return parts[-1]
helper, returns the name of an attribute (without namespace prefix)
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/validator.py#L205-L213
null
class SchemaValidator(object): """ASSUMPTIONS: This class knows about the file being validated, but it recieves that file as a graph and a doc_lines It does not perform any parsing logic on the file It recieves a "validatable" graph object and returns errors""" def __init__(self, graph, doc_lines, url=""): super(SchemaValidator, self).__init__() self.schema_def = None self.allowed_namespaces = [] self.graph = graph self.checked_attributes = [] self.stripped_attribute_names = {} log.info("init validator: %s" % self.__class__.__name__) self.doc_lines = doc_lines self.err = partial(_error, doc_lines=self.doc_lines) def validate(self): """Iterate over all triples in the graph and validate each one appropriately """ log.info("{}\nValidating against {}" .format("-" * 100, self.schema_def.__class__.__name__)) if not self.schema_def: raise ValueError("No schema definition supplied.") self.checked_attributes = [] # TODO - this should maybe choose the actually used namespace, not just # the first one in the list result = ValidationResult(self.allowed_namespaces[0], self.schema_def.__class__.__name__) for subject, predicate, object_ in self.graph: log.info("\nsubj: {subj}\npred: {pred}\n obj: {obj}" .format(subj=subject, pred=predicate, obj=object_.encode('utf-8'))) result.add_error(self._check_triple((subject, predicate, object_))) return result def _should_ignore_predicate(self, predicate): ignored_predicates = ['type', 'item', 'first', 'rest'] return self._field_name_from_uri(predicate) in ignored_predicates def _check_triple(self, triple): """compare triple to ontology, return error or None""" subj, pred, obj = triple if self._should_ignore_predicate(pred): log.info("Ignoring triple with predicate '{}'" .format(self._field_name_from_uri(pred))) return classes = [] log.warning("Possible member %s found" % pred) pred = self._expand_qname(pred) if self._namespace_from_uri(pred) not in self.allowed_namespaces: log.info("Member %s does not use an allowed namespace", pred) return instanceof = self._is_instance((subj, pred, obj)) if type(instanceof) == rt.URIRef: instanceof = self._expand_qname(instanceof) if hasattr(self.schema_def, "attributes_by_class") and \ not self.schema_def.attributes_by_class: log.info("Parsed ontology not found. Parsing...") self.schema_def.parse_ontology() class_invalid = self._validate_class(instanceof) if class_invalid: log.warning("Invalid class %s" % instanceof) return class_invalid # TODO - the above sometimes fails when a single object has more than # one rdfa type (eg <span property="schema:creator rnews:creator" # typeof="schema:Person rnews:Person"> # Graph chooses the type in an arbitrary order, so it's unreliable # eg: http://semanticweb.com/the-impact-of-rdfa_b35003 classes = self._superclasses_for_subject(self.graph, instanceof) classes.append(instanceof) member_invalid = self._validate_member(pred, classes, instanceof) if member_invalid: log.warning("Invalid member of class") return member_invalid dupe_invalid = self._validate_duplication((subj, pred), instanceof) if dupe_invalid: log.warning("Duplication found") return dupe_invalid # collect a list of checked attributes self.checked_attributes.append((subj, pred)) log.warning("successfully validated triple, no errors") return def _validate_class(self, cl): """return error if class `cl` is not found in the ontology""" if cl not in self.schema_def.attributes_by_class: search_string = self._build_search_string(cl) err = self.err( "{0} - invalid class", self._field_name_from_uri(cl), search_string=search_string) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _get_stripped_attributes(self, member, classes): stripped = [] if member not in self.stripped_attribute_names: for cl in classes: stripped.extend( [self._field_name_from_uri(attr) for attr in self.schema_def.attributes_by_class[cl]]) self.stripped_attribute_names[member] = stripped else: stripped = self.stripped_attribute_names[member] return stripped def _validate_member(self, member, classes, instanceof): """return error if `member` is not a member of any class in `classes` """ log.info("Validating member %s" % member) stripped = self._get_stripped_attributes(member, classes) if self._field_name_from_uri(member) in stripped: all_class_members = sum([self.schema_def.attributes_by_class[cl] for cl in classes], []) if member in all_class_members: return if self._namespace_from_uri(member) in self.allowed_namespaces: err = self.err("Unoficially allowed namespace {0}", self._namespace_from_uri(member)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) else: err = self.err("{0} - invalid member of {1}", self._field_name_from_uri(member), self._field_name_from_uri(instanceof)) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _validate_duplication(self, subj_and_pred, cl): """returns error if we've already seen the member `pred` on `subj`""" subj, pred = subj_and_pred log.info("Validating duplication of member %s" % pred) if (subj, pred) in self.checked_attributes: err = self.err("{0} - duplicated member of {1}", self._field_name_from_uri(pred), self._field_name_from_uri(cl)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) def _superclasses_for_subject(self, graph, typeof): """helper, returns a list of all superclasses of a given class""" # TODO - this might be replacing a fairly simple graph API query where # it doesn't need to classes = [] superclass = typeof while True: found = False for p, o in self.schema_def.ontology[superclass]: if self.schema_def.lexicon['subclass'] == str(p): found = True classes.append(o) superclass = o if not found: break return classes def _is_instance(self, triple): """helper, returns the class type of subj""" subj, pred, obj = triple input_pred_ns = self._namespace_from_uri(self._expand_qname(pred)) triples = self.graph.triples( (subj, rt.URIRef(self.schema_def.lexicon['type']), None) ) if triples: for tr in triples: triple_obj_ns = self._namespace_from_uri( self._expand_qname(tr[2])) if input_pred_ns == triple_obj_ns: # match namespaces return tr[2] # return the object def _namespace_from_uri(self, uri): """returns the expanded namespace prefix of a uri""" # TODO - this could be helped a bunch with proper use of the graph API # it seems a bit fragile to treat these as simple string-splits uri = str(uri) parts = uri.split('#') if len(parts) == 1: return "%s/" % '/'.join(uri.split('/')[:-1]) return "%s#" % '#'.join(parts[:-1]) def _expand_qname(self, qname): """expand a qualified name's namespace prefix to include the resolved namespace root url""" if type(qname) is not rt.URIRef: raise TypeError("Cannot expand qname of type {}, must be URIRef" .format(type(qname))) for ns in self.graph.namespaces(): if ns[0] == qname.split(':')[0]: return rt.URIRef("%s%s" % (ns[1], qname.split(':')[-1])) return qname def _build_search_string(self, uri): return self._field_name_from_uri(uri)
Parsely/schemato
schemato/validator.py
SchemaValidator._namespace_from_uri
python
def _namespace_from_uri(self, uri): # TODO - this could be helped a bunch with proper use of the graph API # it seems a bit fragile to treat these as simple string-splits uri = str(uri) parts = uri.split('#') if len(parts) == 1: return "%s/" % '/'.join(uri.split('/')[:-1]) return "%s#" % '#'.join(parts[:-1])
returns the expanded namespace prefix of a uri
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/validator.py#L215-L223
null
class SchemaValidator(object): """ASSUMPTIONS: This class knows about the file being validated, but it recieves that file as a graph and a doc_lines It does not perform any parsing logic on the file It recieves a "validatable" graph object and returns errors""" def __init__(self, graph, doc_lines, url=""): super(SchemaValidator, self).__init__() self.schema_def = None self.allowed_namespaces = [] self.graph = graph self.checked_attributes = [] self.stripped_attribute_names = {} log.info("init validator: %s" % self.__class__.__name__) self.doc_lines = doc_lines self.err = partial(_error, doc_lines=self.doc_lines) def validate(self): """Iterate over all triples in the graph and validate each one appropriately """ log.info("{}\nValidating against {}" .format("-" * 100, self.schema_def.__class__.__name__)) if not self.schema_def: raise ValueError("No schema definition supplied.") self.checked_attributes = [] # TODO - this should maybe choose the actually used namespace, not just # the first one in the list result = ValidationResult(self.allowed_namespaces[0], self.schema_def.__class__.__name__) for subject, predicate, object_ in self.graph: log.info("\nsubj: {subj}\npred: {pred}\n obj: {obj}" .format(subj=subject, pred=predicate, obj=object_.encode('utf-8'))) result.add_error(self._check_triple((subject, predicate, object_))) return result def _should_ignore_predicate(self, predicate): ignored_predicates = ['type', 'item', 'first', 'rest'] return self._field_name_from_uri(predicate) in ignored_predicates def _check_triple(self, triple): """compare triple to ontology, return error or None""" subj, pred, obj = triple if self._should_ignore_predicate(pred): log.info("Ignoring triple with predicate '{}'" .format(self._field_name_from_uri(pred))) return classes = [] log.warning("Possible member %s found" % pred) pred = self._expand_qname(pred) if self._namespace_from_uri(pred) not in self.allowed_namespaces: log.info("Member %s does not use an allowed namespace", pred) return instanceof = self._is_instance((subj, pred, obj)) if type(instanceof) == rt.URIRef: instanceof = self._expand_qname(instanceof) if hasattr(self.schema_def, "attributes_by_class") and \ not self.schema_def.attributes_by_class: log.info("Parsed ontology not found. Parsing...") self.schema_def.parse_ontology() class_invalid = self._validate_class(instanceof) if class_invalid: log.warning("Invalid class %s" % instanceof) return class_invalid # TODO - the above sometimes fails when a single object has more than # one rdfa type (eg <span property="schema:creator rnews:creator" # typeof="schema:Person rnews:Person"> # Graph chooses the type in an arbitrary order, so it's unreliable # eg: http://semanticweb.com/the-impact-of-rdfa_b35003 classes = self._superclasses_for_subject(self.graph, instanceof) classes.append(instanceof) member_invalid = self._validate_member(pred, classes, instanceof) if member_invalid: log.warning("Invalid member of class") return member_invalid dupe_invalid = self._validate_duplication((subj, pred), instanceof) if dupe_invalid: log.warning("Duplication found") return dupe_invalid # collect a list of checked attributes self.checked_attributes.append((subj, pred)) log.warning("successfully validated triple, no errors") return def _validate_class(self, cl): """return error if class `cl` is not found in the ontology""" if cl not in self.schema_def.attributes_by_class: search_string = self._build_search_string(cl) err = self.err( "{0} - invalid class", self._field_name_from_uri(cl), search_string=search_string) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _get_stripped_attributes(self, member, classes): stripped = [] if member not in self.stripped_attribute_names: for cl in classes: stripped.extend( [self._field_name_from_uri(attr) for attr in self.schema_def.attributes_by_class[cl]]) self.stripped_attribute_names[member] = stripped else: stripped = self.stripped_attribute_names[member] return stripped def _validate_member(self, member, classes, instanceof): """return error if `member` is not a member of any class in `classes` """ log.info("Validating member %s" % member) stripped = self._get_stripped_attributes(member, classes) if self._field_name_from_uri(member) in stripped: all_class_members = sum([self.schema_def.attributes_by_class[cl] for cl in classes], []) if member in all_class_members: return if self._namespace_from_uri(member) in self.allowed_namespaces: err = self.err("Unoficially allowed namespace {0}", self._namespace_from_uri(member)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) else: err = self.err("{0} - invalid member of {1}", self._field_name_from_uri(member), self._field_name_from_uri(instanceof)) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _validate_duplication(self, subj_and_pred, cl): """returns error if we've already seen the member `pred` on `subj`""" subj, pred = subj_and_pred log.info("Validating duplication of member %s" % pred) if (subj, pred) in self.checked_attributes: err = self.err("{0} - duplicated member of {1}", self._field_name_from_uri(pred), self._field_name_from_uri(cl)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) def _superclasses_for_subject(self, graph, typeof): """helper, returns a list of all superclasses of a given class""" # TODO - this might be replacing a fairly simple graph API query where # it doesn't need to classes = [] superclass = typeof while True: found = False for p, o in self.schema_def.ontology[superclass]: if self.schema_def.lexicon['subclass'] == str(p): found = True classes.append(o) superclass = o if not found: break return classes def _is_instance(self, triple): """helper, returns the class type of subj""" subj, pred, obj = triple input_pred_ns = self._namespace_from_uri(self._expand_qname(pred)) triples = self.graph.triples( (subj, rt.URIRef(self.schema_def.lexicon['type']), None) ) if triples: for tr in triples: triple_obj_ns = self._namespace_from_uri( self._expand_qname(tr[2])) if input_pred_ns == triple_obj_ns: # match namespaces return tr[2] # return the object def _field_name_from_uri(self, uri): """helper, returns the name of an attribute (without namespace prefix) """ # TODO - should use graph API uri = str(uri) parts = uri.split('#') if len(parts) == 1: return uri.split('/')[-1] or uri return parts[-1] def _expand_qname(self, qname): """expand a qualified name's namespace prefix to include the resolved namespace root url""" if type(qname) is not rt.URIRef: raise TypeError("Cannot expand qname of type {}, must be URIRef" .format(type(qname))) for ns in self.graph.namespaces(): if ns[0] == qname.split(':')[0]: return rt.URIRef("%s%s" % (ns[1], qname.split(':')[-1])) return qname def _build_search_string(self, uri): return self._field_name_from_uri(uri)
Parsely/schemato
schemato/validator.py
SchemaValidator._expand_qname
python
def _expand_qname(self, qname): if type(qname) is not rt.URIRef: raise TypeError("Cannot expand qname of type {}, must be URIRef" .format(type(qname))) for ns in self.graph.namespaces(): if ns[0] == qname.split(':')[0]: return rt.URIRef("%s%s" % (ns[1], qname.split(':')[-1])) return qname
expand a qualified name's namespace prefix to include the resolved namespace root url
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/validator.py#L225-L234
null
class SchemaValidator(object): """ASSUMPTIONS: This class knows about the file being validated, but it recieves that file as a graph and a doc_lines It does not perform any parsing logic on the file It recieves a "validatable" graph object and returns errors""" def __init__(self, graph, doc_lines, url=""): super(SchemaValidator, self).__init__() self.schema_def = None self.allowed_namespaces = [] self.graph = graph self.checked_attributes = [] self.stripped_attribute_names = {} log.info("init validator: %s" % self.__class__.__name__) self.doc_lines = doc_lines self.err = partial(_error, doc_lines=self.doc_lines) def validate(self): """Iterate over all triples in the graph and validate each one appropriately """ log.info("{}\nValidating against {}" .format("-" * 100, self.schema_def.__class__.__name__)) if not self.schema_def: raise ValueError("No schema definition supplied.") self.checked_attributes = [] # TODO - this should maybe choose the actually used namespace, not just # the first one in the list result = ValidationResult(self.allowed_namespaces[0], self.schema_def.__class__.__name__) for subject, predicate, object_ in self.graph: log.info("\nsubj: {subj}\npred: {pred}\n obj: {obj}" .format(subj=subject, pred=predicate, obj=object_.encode('utf-8'))) result.add_error(self._check_triple((subject, predicate, object_))) return result def _should_ignore_predicate(self, predicate): ignored_predicates = ['type', 'item', 'first', 'rest'] return self._field_name_from_uri(predicate) in ignored_predicates def _check_triple(self, triple): """compare triple to ontology, return error or None""" subj, pred, obj = triple if self._should_ignore_predicate(pred): log.info("Ignoring triple with predicate '{}'" .format(self._field_name_from_uri(pred))) return classes = [] log.warning("Possible member %s found" % pred) pred = self._expand_qname(pred) if self._namespace_from_uri(pred) not in self.allowed_namespaces: log.info("Member %s does not use an allowed namespace", pred) return instanceof = self._is_instance((subj, pred, obj)) if type(instanceof) == rt.URIRef: instanceof = self._expand_qname(instanceof) if hasattr(self.schema_def, "attributes_by_class") and \ not self.schema_def.attributes_by_class: log.info("Parsed ontology not found. Parsing...") self.schema_def.parse_ontology() class_invalid = self._validate_class(instanceof) if class_invalid: log.warning("Invalid class %s" % instanceof) return class_invalid # TODO - the above sometimes fails when a single object has more than # one rdfa type (eg <span property="schema:creator rnews:creator" # typeof="schema:Person rnews:Person"> # Graph chooses the type in an arbitrary order, so it's unreliable # eg: http://semanticweb.com/the-impact-of-rdfa_b35003 classes = self._superclasses_for_subject(self.graph, instanceof) classes.append(instanceof) member_invalid = self._validate_member(pred, classes, instanceof) if member_invalid: log.warning("Invalid member of class") return member_invalid dupe_invalid = self._validate_duplication((subj, pred), instanceof) if dupe_invalid: log.warning("Duplication found") return dupe_invalid # collect a list of checked attributes self.checked_attributes.append((subj, pred)) log.warning("successfully validated triple, no errors") return def _validate_class(self, cl): """return error if class `cl` is not found in the ontology""" if cl not in self.schema_def.attributes_by_class: search_string = self._build_search_string(cl) err = self.err( "{0} - invalid class", self._field_name_from_uri(cl), search_string=search_string) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _get_stripped_attributes(self, member, classes): stripped = [] if member not in self.stripped_attribute_names: for cl in classes: stripped.extend( [self._field_name_from_uri(attr) for attr in self.schema_def.attributes_by_class[cl]]) self.stripped_attribute_names[member] = stripped else: stripped = self.stripped_attribute_names[member] return stripped def _validate_member(self, member, classes, instanceof): """return error if `member` is not a member of any class in `classes` """ log.info("Validating member %s" % member) stripped = self._get_stripped_attributes(member, classes) if self._field_name_from_uri(member) in stripped: all_class_members = sum([self.schema_def.attributes_by_class[cl] for cl in classes], []) if member in all_class_members: return if self._namespace_from_uri(member) in self.allowed_namespaces: err = self.err("Unoficially allowed namespace {0}", self._namespace_from_uri(member)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) else: err = self.err("{0} - invalid member of {1}", self._field_name_from_uri(member), self._field_name_from_uri(instanceof)) return ValidationWarning(ValidationResult.ERROR, err['err'], err['line'], err['num']) def _validate_duplication(self, subj_and_pred, cl): """returns error if we've already seen the member `pred` on `subj`""" subj, pred = subj_and_pred log.info("Validating duplication of member %s" % pred) if (subj, pred) in self.checked_attributes: err = self.err("{0} - duplicated member of {1}", self._field_name_from_uri(pred), self._field_name_from_uri(cl)) return ValidationWarning(ValidationResult.WARNING, err['err'], err['line'], err['num']) def _superclasses_for_subject(self, graph, typeof): """helper, returns a list of all superclasses of a given class""" # TODO - this might be replacing a fairly simple graph API query where # it doesn't need to classes = [] superclass = typeof while True: found = False for p, o in self.schema_def.ontology[superclass]: if self.schema_def.lexicon['subclass'] == str(p): found = True classes.append(o) superclass = o if not found: break return classes def _is_instance(self, triple): """helper, returns the class type of subj""" subj, pred, obj = triple input_pred_ns = self._namespace_from_uri(self._expand_qname(pred)) triples = self.graph.triples( (subj, rt.URIRef(self.schema_def.lexicon['type']), None) ) if triples: for tr in triples: triple_obj_ns = self._namespace_from_uri( self._expand_qname(tr[2])) if input_pred_ns == triple_obj_ns: # match namespaces return tr[2] # return the object def _field_name_from_uri(self, uri): """helper, returns the name of an attribute (without namespace prefix) """ # TODO - should use graph API uri = str(uri) parts = uri.split('#') if len(parts) == 1: return uri.split('/')[-1] or uri return parts[-1] def _namespace_from_uri(self, uri): """returns the expanded namespace prefix of a uri""" # TODO - this could be helped a bunch with proper use of the graph API # it seems a bit fragile to treat these as simple string-splits uri = str(uri) parts = uri.split('#') if len(parts) == 1: return "%s/" % '/'.join(uri.split('/')[:-1]) return "%s#" % '#'.join(parts[:-1]) def _build_search_string(self, uri): return self._field_name_from_uri(uri)
Parsely/schemato
schemato/schemas/parselypage.py
ParselyPageParser.original_unescape
python
def original_unescape(self, s): if isinstance(s, basestring): return unicode(HTMLParser.unescape(self, s)) elif isinstance(s, list): return [unicode(HTMLParser.unescape(self, item)) for item in s] else: return s
Since we need to use this sometimes
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/schemas/parselypage.py#L35-L42
null
class ParselyPageParser(HTMLParser): def __init__(self): HTMLParser.__init__(self) self.ppage = None def handle_starttag(self, tag, attrs): if self.ppage is not None or tag != "meta": return attrs = dict(attrs) if attrs.get("name") == "parsely-page": ppage = attrs.get("content", attrs.get("value")) if ppage: try: self.ppage = json.loads(ppage) except: raise HTMLParseError("bad ppage") # bad ppage def unescape(self, s): return s
Parsely/schemato
schemato/schemas/parselypage.py
ParselyPageValidator.get_standard
python
def get_standard(self): try: res = urlopen(PARSELY_PAGE_SCHEMA) except: return [] text = res.read() if isinstance(text, bytes): text = text.decode('utf-8') tree = etree.parse(StringIO(text)) stdref = tree.xpath("//div/@about") return [a.split(':')[1] for a in stdref]
get list of allowed parameters
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/schemas/parselypage.py#L56-L67
null
class ParselyPageValidator(SchemaValidator): def __init__(self, graph, doc_lines, url=""): super(ParselyPageValidator, self).__init__(graph, doc_lines, url=url) self.text = '\n'.join([a[0] for a in doc_lines]) self.stdref = self.get_standard() self.url_validator = HttpUrl() self.data = self._get_parselypage(self.text) def _get_parselypage(self, body): """extract the parsely-page meta content from a page""" parser = ParselyPageParser() ret = None try: parser.feed(body) except HTMLParseError: pass # ignore and hope we got ppage if parser.ppage is None: return ret = parser.ppage if ret: ret = {parser.original_unescape(k): parser.original_unescape(v) for k, v in iteritems(ret)} return ret def validate(self): result = ValidationResult("parsely-page", self.__class__.__name__) if self.data: for key in self.data: res = self.check_key(key) if res: result.add_error(res) return result def check_key(self, key): if key not in self.stdref: err = _error("{0} - invalid parsely-page field", key, doc_lines=self.doc_lines) return ValidationWarning( ValidationResult.ERROR, err['err'], err['line'], err['num']) if key in ["link", "image_url"]: if not self.url_validator(self.data[key]): err = _error( "{0} - invalid url for field '{1}'", self.data[key], key, doc_lines=self.doc_lines) return ValidationWarning( ValidationResult.ERROR, err['err'], err['line'], err['num']) return None
Parsely/schemato
schemato/schemas/parselypage.py
ParselyPageValidator._get_parselypage
python
def _get_parselypage(self, body): parser = ParselyPageParser() ret = None try: parser.feed(body) except HTMLParseError: pass # ignore and hope we got ppage if parser.ppage is None: return ret = parser.ppage if ret: ret = {parser.original_unescape(k): parser.original_unescape(v) for k, v in iteritems(ret)} return ret
extract the parsely-page meta content from a page
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/schemas/parselypage.py#L69-L84
null
class ParselyPageValidator(SchemaValidator): def __init__(self, graph, doc_lines, url=""): super(ParselyPageValidator, self).__init__(graph, doc_lines, url=url) self.text = '\n'.join([a[0] for a in doc_lines]) self.stdref = self.get_standard() self.url_validator = HttpUrl() self.data = self._get_parselypage(self.text) def get_standard(self): """get list of allowed parameters""" try: res = urlopen(PARSELY_PAGE_SCHEMA) except: return [] text = res.read() if isinstance(text, bytes): text = text.decode('utf-8') tree = etree.parse(StringIO(text)) stdref = tree.xpath("//div/@about") return [a.split(':')[1] for a in stdref] def validate(self): result = ValidationResult("parsely-page", self.__class__.__name__) if self.data: for key in self.data: res = self.check_key(key) if res: result.add_error(res) return result def check_key(self, key): if key not in self.stdref: err = _error("{0} - invalid parsely-page field", key, doc_lines=self.doc_lines) return ValidationWarning( ValidationResult.ERROR, err['err'], err['line'], err['num']) if key in ["link", "image_url"]: if not self.url_validator(self.data[key]): err = _error( "{0} - invalid url for field '{1}'", self.data[key], key, doc_lines=self.doc_lines) return ValidationWarning( ValidationResult.ERROR, err['err'], err['line'], err['num']) return None
Parsely/schemato
schemato/schemadef.py
SchemaDef._read_schema
python
def _read_schema(self): cache_filename = os.path.join( CACHE_ROOT, "%s.smt" % self._representation) log.info("Attempting to read local schema at %s" % cache_filename) try: if time.time() - os.stat(cache_filename).st_mtime > CACHE_EXPIRY: log.warning("Cache expired, re-pulling") self._pull_schema_definition(cache_filename) except OSError: log.warning("Local schema not found. Pulling from web.") self._pull_schema_definition(cache_filename) else: log.info("Success") return cache_filename
return the local filename of the definition file for this schema if not present or older than expiry, pull the latest version from the web at self._ontology_file
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/schemadef.py#L38-L55
null
class SchemaDef(object): """class to handle the loading and caching of a standard ASSUMPTIONS: This class holds the definition of a standard It does not perform any validation against that standard This class knows nothing about the input file being validated """ errorstring_base = \ "Are you calling parse_ontology from the base SchemaDef class?" def __init__(self): super(SchemaDef, self).__init__() self._ontology_file = "" self.ontology = defaultdict(list) self.attributes_by_class = {} self._ontology_parser_function = None self.lexicon = {} def _pull_schema_definition(self, fname): """download an ontology definition from the web""" std_url = urlopen(self._ontology_file) cached_std = open(fname, "w+") cached_std.write(std_url.read()) cached_std.close() def parse_ontology(self): """place the ontology graph into a set of custom data structures for use by the validator""" start = time.clock() log.info("Parsing ontology file for %s" % self.__class__.__name__) for subj, pred, obj in self._schema_nodes(): if subj not in self.attributes_by_class: if obj == rt.URIRef(self.lexicon['class']) and \ pred == rt.URIRef(self.lexicon['type']): self.attributes_by_class[subj] = [] leaves = [(subj, pred, obj)] if type(obj) == rt.BNode: leaves = deepest_node((subj, pred, obj), self.graph) for s, p, o in leaves: if o not in self.attributes_by_class: self.attributes_by_class[o] = [] if pred == rt.URIRef(self.lexicon['domain']): self.attributes_by_class[o].append(subj) if not self.attributes_by_class: log.info("No nodes found in ontology") log.info("Ontology parsing complete in {}".format( (time.clock() - start) * 1000)) def _schema_nodes(self): """parse self._ontology_file into a graph""" name, ext = os.path.splitext(self._ontology_file) if ext in ['.ttl']: self._ontology_parser_function = \ lambda s: rdflib.Graph().parse(s, format='n3') else: self._ontology_parser_function = \ lambda s: pyRdfa().graph_from_source(s) if not self._ontology_parser_function: raise ValueError( "No function found to parse ontology. %s" % self.errorstring_base) if not self._ontology_file: raise ValueError( "No ontology file specified. %s" % self.errorstring_base) if not self.lexicon: raise ValueError( "No lexicon object assigned. %s" % self.errorstring_base) latest_file = self._read_schema() try: self.graph = self._ontology_parser_function(latest_file) except: raise IOError("Error parsing ontology at %s" % latest_file) for subj, pred, obj in self.graph: self.ontology[subj].append((pred, obj)) yield (subj, pred, obj)
Parsely/schemato
schemato/schemadef.py
SchemaDef._pull_schema_definition
python
def _pull_schema_definition(self, fname): std_url = urlopen(self._ontology_file) cached_std = open(fname, "w+") cached_std.write(std_url.read()) cached_std.close()
download an ontology definition from the web
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/schemadef.py#L57-L62
null
class SchemaDef(object): """class to handle the loading and caching of a standard ASSUMPTIONS: This class holds the definition of a standard It does not perform any validation against that standard This class knows nothing about the input file being validated """ errorstring_base = \ "Are you calling parse_ontology from the base SchemaDef class?" def __init__(self): super(SchemaDef, self).__init__() self._ontology_file = "" self.ontology = defaultdict(list) self.attributes_by_class = {} self._ontology_parser_function = None self.lexicon = {} def _read_schema(self): """return the local filename of the definition file for this schema if not present or older than expiry, pull the latest version from the web at self._ontology_file""" cache_filename = os.path.join( CACHE_ROOT, "%s.smt" % self._representation) log.info("Attempting to read local schema at %s" % cache_filename) try: if time.time() - os.stat(cache_filename).st_mtime > CACHE_EXPIRY: log.warning("Cache expired, re-pulling") self._pull_schema_definition(cache_filename) except OSError: log.warning("Local schema not found. Pulling from web.") self._pull_schema_definition(cache_filename) else: log.info("Success") return cache_filename def parse_ontology(self): """place the ontology graph into a set of custom data structures for use by the validator""" start = time.clock() log.info("Parsing ontology file for %s" % self.__class__.__name__) for subj, pred, obj in self._schema_nodes(): if subj not in self.attributes_by_class: if obj == rt.URIRef(self.lexicon['class']) and \ pred == rt.URIRef(self.lexicon['type']): self.attributes_by_class[subj] = [] leaves = [(subj, pred, obj)] if type(obj) == rt.BNode: leaves = deepest_node((subj, pred, obj), self.graph) for s, p, o in leaves: if o not in self.attributes_by_class: self.attributes_by_class[o] = [] if pred == rt.URIRef(self.lexicon['domain']): self.attributes_by_class[o].append(subj) if not self.attributes_by_class: log.info("No nodes found in ontology") log.info("Ontology parsing complete in {}".format( (time.clock() - start) * 1000)) def _schema_nodes(self): """parse self._ontology_file into a graph""" name, ext = os.path.splitext(self._ontology_file) if ext in ['.ttl']: self._ontology_parser_function = \ lambda s: rdflib.Graph().parse(s, format='n3') else: self._ontology_parser_function = \ lambda s: pyRdfa().graph_from_source(s) if not self._ontology_parser_function: raise ValueError( "No function found to parse ontology. %s" % self.errorstring_base) if not self._ontology_file: raise ValueError( "No ontology file specified. %s" % self.errorstring_base) if not self.lexicon: raise ValueError( "No lexicon object assigned. %s" % self.errorstring_base) latest_file = self._read_schema() try: self.graph = self._ontology_parser_function(latest_file) except: raise IOError("Error parsing ontology at %s" % latest_file) for subj, pred, obj in self.graph: self.ontology[subj].append((pred, obj)) yield (subj, pred, obj)
Parsely/schemato
schemato/schemadef.py
SchemaDef.parse_ontology
python
def parse_ontology(self): start = time.clock() log.info("Parsing ontology file for %s" % self.__class__.__name__) for subj, pred, obj in self._schema_nodes(): if subj not in self.attributes_by_class: if obj == rt.URIRef(self.lexicon['class']) and \ pred == rt.URIRef(self.lexicon['type']): self.attributes_by_class[subj] = [] leaves = [(subj, pred, obj)] if type(obj) == rt.BNode: leaves = deepest_node((subj, pred, obj), self.graph) for s, p, o in leaves: if o not in self.attributes_by_class: self.attributes_by_class[o] = [] if pred == rt.URIRef(self.lexicon['domain']): self.attributes_by_class[o].append(subj) if not self.attributes_by_class: log.info("No nodes found in ontology") log.info("Ontology parsing complete in {}".format( (time.clock() - start) * 1000))
place the ontology graph into a set of custom data structures for use by the validator
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/schemadef.py#L64-L87
[ "def deepest_node((subj, pred, obj), graph):\n \"\"\"recurse down the tree and return a list of the most deeply nested\n child nodes of the given triple\"\"\"\n # i don't fully accept the premise that this docstring presents\n # i'm not a docstring literalist\n to_return = []\n\n def _deepest_node((subj, pred, obj), graph):\n children = []\n if isinstance(obj, rt.BNode):\n for s, p, o in graph:\n if str(s) == str(obj):\n children.append((s, p, o))\n for s, p, o in children:\n s1, p1, o1 = _deepest_node((s, p, o), graph)\n # coupling *smacks hand with ruler*\n if \"rNews\" in str(o1) and (s1, p1, o1) not in to_return:\n to_return.append((s1, p1, o1))\n return (s1, p1, o1)\n else:\n return (subj, pred, obj)\n _deepest_node((subj, pred, obj), graph)\n\n return to_return\n", "def _schema_nodes(self):\n \"\"\"parse self._ontology_file into a graph\"\"\"\n name, ext = os.path.splitext(self._ontology_file)\n if ext in ['.ttl']:\n self._ontology_parser_function = \\\n lambda s: rdflib.Graph().parse(s, format='n3')\n else:\n self._ontology_parser_function = \\\n lambda s: pyRdfa().graph_from_source(s)\n if not self._ontology_parser_function:\n raise ValueError(\n \"No function found to parse ontology. %s\" %\n self.errorstring_base)\n if not self._ontology_file:\n raise ValueError(\n \"No ontology file specified. %s\" % self.errorstring_base)\n if not self.lexicon:\n raise ValueError(\n \"No lexicon object assigned. %s\" % self.errorstring_base)\n\n latest_file = self._read_schema()\n\n try:\n self.graph = self._ontology_parser_function(latest_file)\n except:\n raise IOError(\"Error parsing ontology at %s\" % latest_file)\n\n for subj, pred, obj in self.graph:\n self.ontology[subj].append((pred, obj))\n yield (subj, pred, obj)\n" ]
class SchemaDef(object): """class to handle the loading and caching of a standard ASSUMPTIONS: This class holds the definition of a standard It does not perform any validation against that standard This class knows nothing about the input file being validated """ errorstring_base = \ "Are you calling parse_ontology from the base SchemaDef class?" def __init__(self): super(SchemaDef, self).__init__() self._ontology_file = "" self.ontology = defaultdict(list) self.attributes_by_class = {} self._ontology_parser_function = None self.lexicon = {} def _read_schema(self): """return the local filename of the definition file for this schema if not present or older than expiry, pull the latest version from the web at self._ontology_file""" cache_filename = os.path.join( CACHE_ROOT, "%s.smt" % self._representation) log.info("Attempting to read local schema at %s" % cache_filename) try: if time.time() - os.stat(cache_filename).st_mtime > CACHE_EXPIRY: log.warning("Cache expired, re-pulling") self._pull_schema_definition(cache_filename) except OSError: log.warning("Local schema not found. Pulling from web.") self._pull_schema_definition(cache_filename) else: log.info("Success") return cache_filename def _pull_schema_definition(self, fname): """download an ontology definition from the web""" std_url = urlopen(self._ontology_file) cached_std = open(fname, "w+") cached_std.write(std_url.read()) cached_std.close() def _schema_nodes(self): """parse self._ontology_file into a graph""" name, ext = os.path.splitext(self._ontology_file) if ext in ['.ttl']: self._ontology_parser_function = \ lambda s: rdflib.Graph().parse(s, format='n3') else: self._ontology_parser_function = \ lambda s: pyRdfa().graph_from_source(s) if not self._ontology_parser_function: raise ValueError( "No function found to parse ontology. %s" % self.errorstring_base) if not self._ontology_file: raise ValueError( "No ontology file specified. %s" % self.errorstring_base) if not self.lexicon: raise ValueError( "No lexicon object assigned. %s" % self.errorstring_base) latest_file = self._read_schema() try: self.graph = self._ontology_parser_function(latest_file) except: raise IOError("Error parsing ontology at %s" % latest_file) for subj, pred, obj in self.graph: self.ontology[subj].append((pred, obj)) yield (subj, pred, obj)
Parsely/schemato
schemato/schemadef.py
SchemaDef._schema_nodes
python
def _schema_nodes(self): name, ext = os.path.splitext(self._ontology_file) if ext in ['.ttl']: self._ontology_parser_function = \ lambda s: rdflib.Graph().parse(s, format='n3') else: self._ontology_parser_function = \ lambda s: pyRdfa().graph_from_source(s) if not self._ontology_parser_function: raise ValueError( "No function found to parse ontology. %s" % self.errorstring_base) if not self._ontology_file: raise ValueError( "No ontology file specified. %s" % self.errorstring_base) if not self.lexicon: raise ValueError( "No lexicon object assigned. %s" % self.errorstring_base) latest_file = self._read_schema() try: self.graph = self._ontology_parser_function(latest_file) except: raise IOError("Error parsing ontology at %s" % latest_file) for subj, pred, obj in self.graph: self.ontology[subj].append((pred, obj)) yield (subj, pred, obj)
parse self._ontology_file into a graph
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/schemadef.py#L89-L118
[ "def _read_schema(self):\n \"\"\"return the local filename of the definition file for this schema\n if not present or older than expiry, pull the latest version from\n the web at self._ontology_file\"\"\"\n cache_filename = os.path.join(\n CACHE_ROOT, \"%s.smt\" % self._representation)\n log.info(\"Attempting to read local schema at %s\" % cache_filename)\n try:\n if time.time() - os.stat(cache_filename).st_mtime > CACHE_EXPIRY:\n log.warning(\"Cache expired, re-pulling\")\n self._pull_schema_definition(cache_filename)\n except OSError:\n log.warning(\"Local schema not found. Pulling from web.\")\n self._pull_schema_definition(cache_filename)\n else:\n log.info(\"Success\")\n\n return cache_filename\n" ]
class SchemaDef(object): """class to handle the loading and caching of a standard ASSUMPTIONS: This class holds the definition of a standard It does not perform any validation against that standard This class knows nothing about the input file being validated """ errorstring_base = \ "Are you calling parse_ontology from the base SchemaDef class?" def __init__(self): super(SchemaDef, self).__init__() self._ontology_file = "" self.ontology = defaultdict(list) self.attributes_by_class = {} self._ontology_parser_function = None self.lexicon = {} def _read_schema(self): """return the local filename of the definition file for this schema if not present or older than expiry, pull the latest version from the web at self._ontology_file""" cache_filename = os.path.join( CACHE_ROOT, "%s.smt" % self._representation) log.info("Attempting to read local schema at %s" % cache_filename) try: if time.time() - os.stat(cache_filename).st_mtime > CACHE_EXPIRY: log.warning("Cache expired, re-pulling") self._pull_schema_definition(cache_filename) except OSError: log.warning("Local schema not found. Pulling from web.") self._pull_schema_definition(cache_filename) else: log.info("Success") return cache_filename def _pull_schema_definition(self, fname): """download an ontology definition from the web""" std_url = urlopen(self._ontology_file) cached_std = open(fname, "w+") cached_std.write(std_url.read()) cached_std.close() def parse_ontology(self): """place the ontology graph into a set of custom data structures for use by the validator""" start = time.clock() log.info("Parsing ontology file for %s" % self.__class__.__name__) for subj, pred, obj in self._schema_nodes(): if subj not in self.attributes_by_class: if obj == rt.URIRef(self.lexicon['class']) and \ pred == rt.URIRef(self.lexicon['type']): self.attributes_by_class[subj] = [] leaves = [(subj, pred, obj)] if type(obj) == rt.BNode: leaves = deepest_node((subj, pred, obj), self.graph) for s, p, o in leaves: if o not in self.attributes_by_class: self.attributes_by_class[o] = [] if pred == rt.URIRef(self.lexicon['domain']): self.attributes_by_class[o].append(subj) if not self.attributes_by_class: log.info("No nodes found in ontology") log.info("Ontology parsing complete in {}".format( (time.clock() - start) * 1000))
Parsely/schemato
schemato/utils.py
deepest_node
python
def deepest_node((subj, pred, obj), graph): # i don't fully accept the premise that this docstring presents # i'm not a docstring literalist to_return = [] def _deepest_node((subj, pred, obj), graph): children = [] if isinstance(obj, rt.BNode): for s, p, o in graph: if str(s) == str(obj): children.append((s, p, o)) for s, p, o in children: s1, p1, o1 = _deepest_node((s, p, o), graph) # coupling *smacks hand with ruler* if "rNews" in str(o1) and (s1, p1, o1) not in to_return: to_return.append((s1, p1, o1)) return (s1, p1, o1) else: return (subj, pred, obj) _deepest_node((subj, pred, obj), graph) return to_return
recurse down the tree and return a list of the most deeply nested child nodes of the given triple
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/utils.py#L4-L27
[ "def _deepest_node((subj, pred, obj), graph):\n children = []\n if isinstance(obj, rt.BNode):\n for s, p, o in graph:\n if str(s) == str(obj):\n children.append((s, p, o))\n for s, p, o in children:\n s1, p1, o1 = _deepest_node((s, p, o), graph)\n # coupling *smacks hand with ruler*\n if \"rNews\" in str(o1) and (s1, p1, o1) not in to_return:\n to_return.append((s1, p1, o1))\n return (s1, p1, o1)\n else:\n return (subj, pred, obj)\n" ]
import rdflib.term as rt
Parsely/schemato
schemato/schemato.py
Schemato._document_lines
python
def _document_lines(self, text): inlines = text.split('\n') doc_lines = [(re.sub(r'^ +| +$', '', line), num) for line, num in zip(inlines, xrange(1, len(inlines) + 1))] return doc_lines
helper, get a list of (linetext, linenum) from a string with newlines
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/schemato.py#L64-L72
null
class Schemato(object): def __init__(self, source, url=None, loglevel="ERROR"): """init with a local filepath or a URI""" super(Schemato, self).__init__() self.set_loglevel(loglevel) graph_source = source if url is not None: graph_source = StringIO(source) self.graph = CompoundGraph(graph_source) def _read_stream(source): text, url = self._get_document(source) return url, self._document_lines(text) parsed_url, self.doc_lines = _read_stream(source) self.url = url if url is None: self.url = parsed_url validator = ParselyPageValidator(self.graph, self.doc_lines) self.parsely_page = validator.data def validate(self): self._load_validators() results = [v.validate() for v in self.validators] log.info("returned from validate() : %s", results) for res in results: log.info(res.to_json()) return results def set_loglevel(self, loglevel): if hasattr(logging, loglevel): log.setLevel(loglevel) else: log.setLevel(logging.ERROR) log.error( "Unrecognized loglevel %s, defaulting to ERROR", loglevel) def _load_validators(self): self.validators = set() for entry_point in iter_entry_points('schemato_validators'): validator_fn = entry_point.load() validator = validator_fn(self.graph, self.doc_lines, url=self.url) self.validators.add(validator) def _get_document(self, source): """helper, open a file or url and return the content and identifier""" scheme_url = source if not source.startswith("http"): scheme_url = "http://%s" % source text = source try: text = urllib.urlopen(scheme_url).read() except: pass else: return (text, scheme_url) try: text = open(source, "r").read() except: pass else: return (text, source) return (text, None)
Parsely/schemato
schemato/schemato.py
Schemato._get_document
python
def _get_document(self, source): scheme_url = source if not source.startswith("http"): scheme_url = "http://%s" % source text = source try: text = urllib.urlopen(scheme_url).read() except: pass else: return (text, scheme_url) try: text = open(source, "r").read() except: pass else: return (text, source) return (text, None)
helper, open a file or url and return the content and identifier
train
https://github.com/Parsely/schemato/blob/7002316fbcd52f2e669f8372bf1338c572e3df4b/schemato/schemato.py#L74-L96
null
class Schemato(object): def __init__(self, source, url=None, loglevel="ERROR"): """init with a local filepath or a URI""" super(Schemato, self).__init__() self.set_loglevel(loglevel) graph_source = source if url is not None: graph_source = StringIO(source) self.graph = CompoundGraph(graph_source) def _read_stream(source): text, url = self._get_document(source) return url, self._document_lines(text) parsed_url, self.doc_lines = _read_stream(source) self.url = url if url is None: self.url = parsed_url validator = ParselyPageValidator(self.graph, self.doc_lines) self.parsely_page = validator.data def validate(self): self._load_validators() results = [v.validate() for v in self.validators] log.info("returned from validate() : %s", results) for res in results: log.info(res.to_json()) return results def set_loglevel(self, loglevel): if hasattr(logging, loglevel): log.setLevel(loglevel) else: log.setLevel(logging.ERROR) log.error( "Unrecognized loglevel %s, defaulting to ERROR", loglevel) def _load_validators(self): self.validators = set() for entry_point in iter_entry_points('schemato_validators'): validator_fn = entry_point.load() validator = validator_fn(self.graph, self.doc_lines, url=self.url) self.validators.add(validator) def _document_lines(self, text): """helper, get a list of (linetext, linenum) from a string with newlines """ inlines = text.split('\n') doc_lines = [(re.sub(r'^ +| +$', '', line), num) for line, num in zip(inlines, xrange(1, len(inlines) + 1))] return doc_lines
rycus86/ghost-client
ghost_client/models.py
ModelList.get_page
python
def get_page(self, page_number): if page_number: kwargs = dict(self._list_kwargs) kwargs['limit'] = self.limit kwargs['page'] = page_number return self._controller.list(**kwargs)
:param page_number: The page number to fetch (1-indexed) :return: The requested page fetched from the API for the query
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/models.py#L109-L120
null
class ModelList(list): """ Wrapper around lists returned by the API. Exposes methods related to pagination and wraps each item in their respective model type. """ def __init__(self, data, type_name, controller, list_kwargs, model_type=Model): """ Enhances a regular list. :param data: The original iterable :param type_name: The name of the type as the API knows it :param controller: The controller that returned the list :param list_kwargs: Parameters to use when fetching pages from the API :param model_type: The model type of the items """ super(ModelList, self).__init__(map(model_type, data[type_name])) self.meta = data['meta']['pagination'] self._controller = controller self._list_kwargs = list_kwargs @property def total(self): """ :return: The total number of results available for the query """ return self.meta['total'] @property def pages(self): """ :return: The number of pages available for the query """ return self.meta['pages'] @property def limit(self): """ :return: The limit used for queries """ return self.meta['limit'] def next_page(self): """ :return: The next page fetched from the API for the query """ return self.get_page(self.meta['next']) def prev_page(self): """ :return: The previous page fetched from the API for the query """ return self.get_page(self.meta['prev'])
rycus86/ghost-client
ghost_client/models.py
Controller.list
python
def list(self, **kwargs): return ModelList( self.ghost.execute_get('%s/' % self._type_name, **kwargs), self._type_name, self, kwargs, model_type=self._model_type )
Fetch a list of resources from the API. :param kwargs: Parameters for the request (see from and below https://api.ghost.org/docs/limit) :return: The list of items returned by the API wrapped as `Model` objects with pagination by `ModelList`
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/models.py#L141-L154
null
class Controller(object): """ The API controller dealing with requests for a specific type. """ def __init__(self, ghost, type_name, model_type=Model): """ Initializes a new controller. :param ghost: An instance of the API client :param type_name: The type name as the API knows it :param model_type: The model type to wrap response items as """ self.ghost = ghost self._type_name = type_name self._model_type = model_type def get(self, id=None, slug=None, **kwargs): """ Fetch a resource from the API. Either the `id` or the `slug` has to be present. :param id: The ID of the resource :param slug: The slug of the resource :param kwargs: Parameters for the request (see from and below https://api.ghost.org/docs/limit) :return: The item returned by the API wrapped as a `Model` object """ if id: items = self.ghost.execute_get('%s/%s/' % (self._type_name, id), **kwargs) elif slug: items = self.ghost.execute_get('%s/slug/%s/' % (self._type_name, slug), **kwargs) else: raise GhostException( 500, 'Either the ID or the Slug of the resource needs to be specified' ) return self._model_type(items[self._type_name][0]) def create(self, **kwargs): """ Creates a new resource. :param kwargs: The properties of the resource :return: The created item returned by the API wrapped as a `Model` object """ response = self.ghost.execute_post('%s/' % self._type_name, json={ self._type_name: [ kwargs ] }) return self._model_type(response.get(self._type_name)[0]) def update(self, id, **kwargs): """ Updates an existing resource. :param id: The ID of the resource :param kwargs: The properties of the resource to change :return: The updated item returned by the API wrapped as a `Model` object """ response = self.ghost.execute_put('%s/%s/' % (self._type_name, id), json={ self._type_name: [ kwargs ] }) return self._model_type(response.get(self._type_name)[0]) def delete(self, id): """ Deletes an existing resource. Does not return anything but raises an exception when failed. :param id: The ID of the resource """ self.ghost.execute_delete('%s/%s/' % (self._type_name, id))
rycus86/ghost-client
ghost_client/models.py
Controller.get
python
def get(self, id=None, slug=None, **kwargs): if id: items = self.ghost.execute_get('%s/%s/' % (self._type_name, id), **kwargs) elif slug: items = self.ghost.execute_get('%s/slug/%s/' % (self._type_name, slug), **kwargs) else: raise GhostException( 500, 'Either the ID or the Slug of the resource needs to be specified' ) return self._model_type(items[self._type_name][0])
Fetch a resource from the API. Either the `id` or the `slug` has to be present. :param id: The ID of the resource :param slug: The slug of the resource :param kwargs: Parameters for the request (see from and below https://api.ghost.org/docs/limit) :return: The item returned by the API wrapped as a `Model` object
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/models.py#L156-L180
null
class Controller(object): """ The API controller dealing with requests for a specific type. """ def __init__(self, ghost, type_name, model_type=Model): """ Initializes a new controller. :param ghost: An instance of the API client :param type_name: The type name as the API knows it :param model_type: The model type to wrap response items as """ self.ghost = ghost self._type_name = type_name self._model_type = model_type def list(self, **kwargs): """ Fetch a list of resources from the API. :param kwargs: Parameters for the request (see from and below https://api.ghost.org/docs/limit) :return: The list of items returned by the API wrapped as `Model` objects with pagination by `ModelList` """ return ModelList( self.ghost.execute_get('%s/' % self._type_name, **kwargs), self._type_name, self, kwargs, model_type=self._model_type ) def create(self, **kwargs): """ Creates a new resource. :param kwargs: The properties of the resource :return: The created item returned by the API wrapped as a `Model` object """ response = self.ghost.execute_post('%s/' % self._type_name, json={ self._type_name: [ kwargs ] }) return self._model_type(response.get(self._type_name)[0]) def update(self, id, **kwargs): """ Updates an existing resource. :param id: The ID of the resource :param kwargs: The properties of the resource to change :return: The updated item returned by the API wrapped as a `Model` object """ response = self.ghost.execute_put('%s/%s/' % (self._type_name, id), json={ self._type_name: [ kwargs ] }) return self._model_type(response.get(self._type_name)[0]) def delete(self, id): """ Deletes an existing resource. Does not return anything but raises an exception when failed. :param id: The ID of the resource """ self.ghost.execute_delete('%s/%s/' % (self._type_name, id))
rycus86/ghost-client
ghost_client/models.py
Controller.create
python
def create(self, **kwargs): response = self.ghost.execute_post('%s/' % self._type_name, json={ self._type_name: [ kwargs ] }) return self._model_type(response.get(self._type_name)[0])
Creates a new resource. :param kwargs: The properties of the resource :return: The created item returned by the API wrapped as a `Model` object
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/models.py#L182-L197
null
class Controller(object): """ The API controller dealing with requests for a specific type. """ def __init__(self, ghost, type_name, model_type=Model): """ Initializes a new controller. :param ghost: An instance of the API client :param type_name: The type name as the API knows it :param model_type: The model type to wrap response items as """ self.ghost = ghost self._type_name = type_name self._model_type = model_type def list(self, **kwargs): """ Fetch a list of resources from the API. :param kwargs: Parameters for the request (see from and below https://api.ghost.org/docs/limit) :return: The list of items returned by the API wrapped as `Model` objects with pagination by `ModelList` """ return ModelList( self.ghost.execute_get('%s/' % self._type_name, **kwargs), self._type_name, self, kwargs, model_type=self._model_type ) def get(self, id=None, slug=None, **kwargs): """ Fetch a resource from the API. Either the `id` or the `slug` has to be present. :param id: The ID of the resource :param slug: The slug of the resource :param kwargs: Parameters for the request (see from and below https://api.ghost.org/docs/limit) :return: The item returned by the API wrapped as a `Model` object """ if id: items = self.ghost.execute_get('%s/%s/' % (self._type_name, id), **kwargs) elif slug: items = self.ghost.execute_get('%s/slug/%s/' % (self._type_name, slug), **kwargs) else: raise GhostException( 500, 'Either the ID or the Slug of the resource needs to be specified' ) return self._model_type(items[self._type_name][0]) def update(self, id, **kwargs): """ Updates an existing resource. :param id: The ID of the resource :param kwargs: The properties of the resource to change :return: The updated item returned by the API wrapped as a `Model` object """ response = self.ghost.execute_put('%s/%s/' % (self._type_name, id), json={ self._type_name: [ kwargs ] }) return self._model_type(response.get(self._type_name)[0]) def delete(self, id): """ Deletes an existing resource. Does not return anything but raises an exception when failed. :param id: The ID of the resource """ self.ghost.execute_delete('%s/%s/' % (self._type_name, id))
rycus86/ghost-client
ghost_client/models.py
Controller.update
python
def update(self, id, **kwargs): response = self.ghost.execute_put('%s/%s/' % (self._type_name, id), json={ self._type_name: [ kwargs ] }) return self._model_type(response.get(self._type_name)[0])
Updates an existing resource. :param id: The ID of the resource :param kwargs: The properties of the resource to change :return: The updated item returned by the API wrapped as a `Model` object
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/models.py#L199-L215
null
class Controller(object): """ The API controller dealing with requests for a specific type. """ def __init__(self, ghost, type_name, model_type=Model): """ Initializes a new controller. :param ghost: An instance of the API client :param type_name: The type name as the API knows it :param model_type: The model type to wrap response items as """ self.ghost = ghost self._type_name = type_name self._model_type = model_type def list(self, **kwargs): """ Fetch a list of resources from the API. :param kwargs: Parameters for the request (see from and below https://api.ghost.org/docs/limit) :return: The list of items returned by the API wrapped as `Model` objects with pagination by `ModelList` """ return ModelList( self.ghost.execute_get('%s/' % self._type_name, **kwargs), self._type_name, self, kwargs, model_type=self._model_type ) def get(self, id=None, slug=None, **kwargs): """ Fetch a resource from the API. Either the `id` or the `slug` has to be present. :param id: The ID of the resource :param slug: The slug of the resource :param kwargs: Parameters for the request (see from and below https://api.ghost.org/docs/limit) :return: The item returned by the API wrapped as a `Model` object """ if id: items = self.ghost.execute_get('%s/%s/' % (self._type_name, id), **kwargs) elif slug: items = self.ghost.execute_get('%s/slug/%s/' % (self._type_name, slug), **kwargs) else: raise GhostException( 500, 'Either the ID or the Slug of the resource needs to be specified' ) return self._model_type(items[self._type_name][0]) def create(self, **kwargs): """ Creates a new resource. :param kwargs: The properties of the resource :return: The created item returned by the API wrapped as a `Model` object """ response = self.ghost.execute_post('%s/' % self._type_name, json={ self._type_name: [ kwargs ] }) return self._model_type(response.get(self._type_name)[0]) def delete(self, id): """ Deletes an existing resource. Does not return anything but raises an exception when failed. :param id: The ID of the resource """ self.ghost.execute_delete('%s/%s/' % (self._type_name, id))
rycus86/ghost-client
ghost_client/models.py
PostController.create
python
def create(self, **kwargs): return super(PostController, self).create(**self._with_markdown(kwargs))
Creates a new post. When the `markdown` property is present, it will be automatically converted to `mobiledoc` on v1.+ of the server. :param kwargs: The properties of the post :return: The created `Post` object
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/models.py#L242-L252
[ "def create(self, **kwargs):\n \"\"\"\n Creates a new resource.\n\n :param kwargs: The properties of the resource\n :return: The created item returned by the API\n wrapped as a `Model` object\n \"\"\"\n\n response = self.ghost.execute_post('%s/' % self._type_name, json={\n self._type_name: [\n kwargs\n ]\n })\n\n return self._model_type(response.get(self._type_name)[0])\n", "def _with_markdown(self, kwargs):\n markdown = kwargs.pop('markdown', None)\n\n if markdown:\n if self.ghost.version.startswith('0'):\n # put it back as is for version 0.x\n kwargs['markdown'] = markdown\n\n else:\n updated = dict(kwargs)\n updated['mobiledoc'] = json.dumps({\n \"version\": \"0.3.1\", \"markups\": [], \"atoms\": [],\n \"cards\": [[\"card-markdown\", {\"cardName\": \"card-markdown\", \"markdown\": markdown}]],\n \"sections\": [[10, 0]]})\n return updated\n\n return kwargs\n" ]
class PostController(Controller): """ Controller extension for managing posts. """ def __init__(self, ghost): """ Initialize a new controller for posts. :param ghost: An instance of the API client """ super(PostController, self).__init__(ghost, 'posts', model_type=Post) def update(self, id, **kwargs): """ Updates an existing post. When the `markdown` property is present, it will be automatically converted to `mobiledoc` on v1.+ of the server. :param id: The ID of the existing post :param kwargs: The properties of the post to change :return: The updated `Post` object """ return super(PostController, self).update(id, **self._with_markdown(kwargs)) def _with_markdown(self, kwargs): markdown = kwargs.pop('markdown', None) if markdown: if self.ghost.version.startswith('0'): # put it back as is for version 0.x kwargs['markdown'] = markdown else: updated = dict(kwargs) updated['mobiledoc'] = json.dumps({ "version": "0.3.1", "markups": [], "atoms": [], "cards": [["card-markdown", {"cardName": "card-markdown", "markdown": markdown}]], "sections": [[10, 0]]}) return updated return kwargs
rycus86/ghost-client
ghost_client/models.py
PostController.update
python
def update(self, id, **kwargs): return super(PostController, self).update(id, **self._with_markdown(kwargs))
Updates an existing post. When the `markdown` property is present, it will be automatically converted to `mobiledoc` on v1.+ of the server. :param id: The ID of the existing post :param kwargs: The properties of the post to change :return: The updated `Post` object
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/models.py#L254-L265
[ "def update(self, id, **kwargs):\n \"\"\"\n Updates an existing resource.\n\n :param id: The ID of the resource\n :param kwargs: The properties of the resource to change\n :return: The updated item returned by the API\n wrapped as a `Model` object\n \"\"\"\n\n response = self.ghost.execute_put('%s/%s/' % (self._type_name, id), json={\n self._type_name: [\n kwargs\n ]\n })\n\n return self._model_type(response.get(self._type_name)[0])\n", "def _with_markdown(self, kwargs):\n markdown = kwargs.pop('markdown', None)\n\n if markdown:\n if self.ghost.version.startswith('0'):\n # put it back as is for version 0.x\n kwargs['markdown'] = markdown\n\n else:\n updated = dict(kwargs)\n updated['mobiledoc'] = json.dumps({\n \"version\": \"0.3.1\", \"markups\": [], \"atoms\": [],\n \"cards\": [[\"card-markdown\", {\"cardName\": \"card-markdown\", \"markdown\": markdown}]],\n \"sections\": [[10, 0]]})\n return updated\n\n return kwargs\n" ]
class PostController(Controller): """ Controller extension for managing posts. """ def __init__(self, ghost): """ Initialize a new controller for posts. :param ghost: An instance of the API client """ super(PostController, self).__init__(ghost, 'posts', model_type=Post) def create(self, **kwargs): """ Creates a new post. When the `markdown` property is present, it will be automatically converted to `mobiledoc` on v1.+ of the server. :param kwargs: The properties of the post :return: The created `Post` object """ return super(PostController, self).create(**self._with_markdown(kwargs)) def _with_markdown(self, kwargs): markdown = kwargs.pop('markdown', None) if markdown: if self.ghost.version.startswith('0'): # put it back as is for version 0.x kwargs['markdown'] = markdown else: updated = dict(kwargs) updated['mobiledoc'] = json.dumps({ "version": "0.3.1", "markups": [], "atoms": [], "cards": [["card-markdown", {"cardName": "card-markdown", "markdown": markdown}]], "sections": [[10, 0]]}) return updated return kwargs
rycus86/ghost-client
ghost_client/helpers.py
refresh_session_if_necessary
python
def refresh_session_if_necessary(f): @functools.wraps(f) def wrapped(self, *args, **kwargs): try: result = f(self, *args, **kwargs) except Exception as ex: if hasattr(ex, 'code') and ex.code in (401, 403): self.refresh_session() # retry now result = f(self, *args, **kwargs) else: raise ex return result return wrapped
Decorator to use on methods that are allowed to retry the request after reauthenticating the client. :param f: The original function :return: The decorated function
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/helpers.py#L4-L27
null
import functools
rycus86/ghost-client
ghost_client/api.py
Ghost.from_sqlite
python
def from_sqlite(cls, database_path, base_url, version='auto', client_id='ghost-admin'): import os import sqlite3 fd = os.open(database_path, os.O_RDONLY) connection = sqlite3.connect('/dev/fd/%d' % fd) os.close(fd) try: row = connection.execute( 'SELECT secret FROM clients WHERE slug = ?', (client_id,) ).fetchone() if row: return cls( base_url, version=version, client_id=client_id, client_secret=row[0] ) else: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_secret found for client_id: %s' % client_id }]) finally: connection.close()
Initialize a new Ghost API client, reading the client ID and secret from the SQlite database. :param database_path: The path to the database file. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: The client ID to look for in the database :return: A new Ghost API client instance
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/api.py#L142-L180
null
class Ghost(object): """ API client for the Ghost REST endpoints. See https://api.ghost.org/ for the available parameters. Sample usage: from ghost_client import Ghost # to read the client ID and secret from the database ghost = Ghost.from_sqlite( '/var/lib/ghost/content/data/ghost.db', 'http://localhost:2368' ) # or to use a specific client ID and secret ghost = Ghost( 'http://localhost:2368', client_id='ghost-admin', client_secret='secret_key' ) # log in ghost.login('username', 'password') # print the server's version print(ghost.version) # create a new tag tag = ghost.tags.create(name='API sample') # create a new post using it post = ghost.posts.create( title='Example post', slug='custom-slug', markdown='', # yes, even on v1.+ custom_excerpt='An example post created from Python', tags=[tag] ) # list posts, tags and users posts = ghost.posts.list( status='all', fields=('id', 'title', 'slug'), formats=('html', 'mobiledoc', 'plaintext'), ) tags = ghost.tags.list(fields='name', limit='all') users = ghost.users.list(include='count.posts') # use pagination while posts: for post in posts: print(post) posts = posts.next_page() print(posts.total) print(posts.pages) # update a post & tag updated_post = ghost.posts.update(post.id, title='Updated title') updated_tag = ghost.tags.update(tag.id, name='Updated tag') # note: creating, updating and deleting a user is not allowed by the API # access fields as properties print(post.title) print(post.markdown) # needs formats='mobiledoc' print(post.author.name) # needs include='author' # delete a post & tag ghost.posts.delete(post.id) ghost.tags.delete(tag.id) # upload an image ghost.upload(file_obj=open('sample.png', 'rb')) ghost.upload(file_path='/path/to/image.jpeg', 'rb') ghost.upload(name='image.gif', data=open('local.gif', 'rb').read()) # log out ghost.logout() The logged in credentials will be saved in memory and on HTTP 401 errors the client will attempt to re-authenticate once automatically. Responses are wrapped in `models.ModelList` and `models.Model` types to allow pagination and retrieving fields as properties. """ DEFAULT_VERSION = '1' """ The default version to report when cannot be fetched. """ def __init__( self, base_url, version='auto', client_id=None, client_secret=None, access_token=None, refresh_token=None ): """ Creates a new Ghost API client. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: Self-supplied client ID (optional) :param client_secret: Self-supplied client secret (optional) :param access_token: Self-supplied access token (optional) :param refresh_token: Self-supplied refresh token (optional) """ self.base_url = '%s/ghost/api/v0.1' % base_url self._version = version self._client_id = client_id self._client_secret = client_secret self._access_token = access_token self._refresh_token = refresh_token self._username = None self._password = None if not self._client_id or not self._client_secret: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_id or client_secret given or found' }]) self.posts = PostController(self) self.tags = Controller(self, 'tags') self.users = Controller(self, 'users') @classmethod @property def version(self): """ :return: The version of the server when initialized as 'auto', otherwise the version passed in at initialization """ if self._version != 'auto': return self._version if self._version == 'auto': try: data = self.execute_get('configuration/about/') self._version = data['configuration'][0]['version'] except GhostException: return self.DEFAULT_VERSION return self._version def login(self, username, password): """ Authenticate with the server. :param username: The username of an existing user :param password: The password for the user :return: The authentication response from the REST endpoint """ data = self._authenticate( grant_type='password', username=username, password=password, client_id=self._client_id, client_secret=self._client_secret ) self._username = username self._password = password return data def refresh_session(self): """ Re-authenticate using the refresh token if available. Otherwise log in using the username and password if it was used to authenticate initially. :return: The authentication response or `None` if not available """ if not self._refresh_token: if self._username and self._password: return self.login(self._username, self._password) return return self._authenticate( grant_type='refresh_token', refresh_token=self._refresh_token, client_id=self._client_id, client_secret=self._client_secret ) def _authenticate(self, **kwargs): response = requests.post( '%s/authentication/token' % self.base_url, data=kwargs ) if response.status_code != 200: raise GhostException(response.status_code, response.json().get('errors', [])) data = response.json() self._access_token = data.get('access_token') self._refresh_token = data.get('refresh_token') return data def revoke_access_token(self): """ Revoke the access token currently in use. """ if not self._access_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='access_token', token=self._access_token )) self._access_token = None def revoke_refresh_token(self): """ Revoke the refresh token currently active. """ if not self._refresh_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='refresh_token', token=self._refresh_token )) self._refresh_token = None def logout(self): """ Log out, revoking the access tokens and forgetting the login details if they were given. """ self.revoke_refresh_token() self.revoke_access_token() self._username, self._password = None, None def upload(self, file_obj=None, file_path=None, name=None, data=None): """ Upload an image and return its path on the server. Either `file_obj` or `file_path` or `name` and `data` has to be specified. :param file_obj: A file object to upload :param file_path: A file path to upload from :param name: A file name for uploading :param data: The file content to upload :return: The path of the uploaded file on the server """ close = False if file_obj: file_name, content = os.path.basename(file_obj.name), file_obj elif file_path: file_name, content = os.path.basename(file_path), open(file_path, 'rb') close = True elif name and data: file_name, content = name, data else: raise GhostException( 400, 'Either `file_obj` or `file_path` or ' '`name` and `data` needs to be specified' ) try: content_type, _ = mimetypes.guess_type(file_name) file_arg = (file_name, content, content_type) response = self.execute_post('uploads/', files={'uploadimage': file_arg}) return response finally: if close: content.close() @refresh_session_if_necessary def execute_get(self, resource, **kwargs): """ Execute an HTTP GET request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional query parameters (and optionally headers) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if kwargs: separator = '&' if '?' in url else '?' for key, value in kwargs.items(): if hasattr(value, '__iter__') and type(value) not in six.string_types: url = '%s%s%s=%s' % (url, separator, key, ','.join(value)) else: url = '%s%s%s=%s' % (url, separator, key, value) separator = '&' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token else: separator = '&' if '?' in url else '?' url = '%s%sclient_id=%s&client_secret=%s' % ( url, separator, self._client_id, self._client_secret ) response = requests.get(url, headers=headers) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response.json() def execute_post(self, resource, **kwargs): """ Execute an HTTP POST request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.post, **kwargs).json() def execute_put(self, resource, **kwargs): """ Execute an HTTP PUT request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.put, **kwargs).json() def execute_delete(self, resource, **kwargs): """ Execute an HTTP DELETE request against the API endpoints. This method is meant for internal use. Does not return anything but raises an exception when failed. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) """ self._request(resource, requests.delete, **kwargs) @refresh_session_if_necessary def _request(self, resource, request, **kwargs): if not self._access_token: raise GhostException(401, [{ 'errorType': 'ClientError', 'message': 'Access token not found' }]) url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) if 'json' in kwargs: headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token response = request(url, headers=headers, **kwargs) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response
rycus86/ghost-client
ghost_client/api.py
Ghost.version
python
def version(self): if self._version != 'auto': return self._version if self._version == 'auto': try: data = self.execute_get('configuration/about/') self._version = data['configuration'][0]['version'] except GhostException: return self.DEFAULT_VERSION return self._version
:return: The version of the server when initialized as 'auto', otherwise the version passed in at initialization
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/api.py#L183-L199
null
class Ghost(object): """ API client for the Ghost REST endpoints. See https://api.ghost.org/ for the available parameters. Sample usage: from ghost_client import Ghost # to read the client ID and secret from the database ghost = Ghost.from_sqlite( '/var/lib/ghost/content/data/ghost.db', 'http://localhost:2368' ) # or to use a specific client ID and secret ghost = Ghost( 'http://localhost:2368', client_id='ghost-admin', client_secret='secret_key' ) # log in ghost.login('username', 'password') # print the server's version print(ghost.version) # create a new tag tag = ghost.tags.create(name='API sample') # create a new post using it post = ghost.posts.create( title='Example post', slug='custom-slug', markdown='', # yes, even on v1.+ custom_excerpt='An example post created from Python', tags=[tag] ) # list posts, tags and users posts = ghost.posts.list( status='all', fields=('id', 'title', 'slug'), formats=('html', 'mobiledoc', 'plaintext'), ) tags = ghost.tags.list(fields='name', limit='all') users = ghost.users.list(include='count.posts') # use pagination while posts: for post in posts: print(post) posts = posts.next_page() print(posts.total) print(posts.pages) # update a post & tag updated_post = ghost.posts.update(post.id, title='Updated title') updated_tag = ghost.tags.update(tag.id, name='Updated tag') # note: creating, updating and deleting a user is not allowed by the API # access fields as properties print(post.title) print(post.markdown) # needs formats='mobiledoc' print(post.author.name) # needs include='author' # delete a post & tag ghost.posts.delete(post.id) ghost.tags.delete(tag.id) # upload an image ghost.upload(file_obj=open('sample.png', 'rb')) ghost.upload(file_path='/path/to/image.jpeg', 'rb') ghost.upload(name='image.gif', data=open('local.gif', 'rb').read()) # log out ghost.logout() The logged in credentials will be saved in memory and on HTTP 401 errors the client will attempt to re-authenticate once automatically. Responses are wrapped in `models.ModelList` and `models.Model` types to allow pagination and retrieving fields as properties. """ DEFAULT_VERSION = '1' """ The default version to report when cannot be fetched. """ def __init__( self, base_url, version='auto', client_id=None, client_secret=None, access_token=None, refresh_token=None ): """ Creates a new Ghost API client. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: Self-supplied client ID (optional) :param client_secret: Self-supplied client secret (optional) :param access_token: Self-supplied access token (optional) :param refresh_token: Self-supplied refresh token (optional) """ self.base_url = '%s/ghost/api/v0.1' % base_url self._version = version self._client_id = client_id self._client_secret = client_secret self._access_token = access_token self._refresh_token = refresh_token self._username = None self._password = None if not self._client_id or not self._client_secret: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_id or client_secret given or found' }]) self.posts = PostController(self) self.tags = Controller(self, 'tags') self.users = Controller(self, 'users') @classmethod def from_sqlite(cls, database_path, base_url, version='auto', client_id='ghost-admin'): """ Initialize a new Ghost API client, reading the client ID and secret from the SQlite database. :param database_path: The path to the database file. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: The client ID to look for in the database :return: A new Ghost API client instance """ import os import sqlite3 fd = os.open(database_path, os.O_RDONLY) connection = sqlite3.connect('/dev/fd/%d' % fd) os.close(fd) try: row = connection.execute( 'SELECT secret FROM clients WHERE slug = ?', (client_id,) ).fetchone() if row: return cls( base_url, version=version, client_id=client_id, client_secret=row[0] ) else: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_secret found for client_id: %s' % client_id }]) finally: connection.close() @property def login(self, username, password): """ Authenticate with the server. :param username: The username of an existing user :param password: The password for the user :return: The authentication response from the REST endpoint """ data = self._authenticate( grant_type='password', username=username, password=password, client_id=self._client_id, client_secret=self._client_secret ) self._username = username self._password = password return data def refresh_session(self): """ Re-authenticate using the refresh token if available. Otherwise log in using the username and password if it was used to authenticate initially. :return: The authentication response or `None` if not available """ if not self._refresh_token: if self._username and self._password: return self.login(self._username, self._password) return return self._authenticate( grant_type='refresh_token', refresh_token=self._refresh_token, client_id=self._client_id, client_secret=self._client_secret ) def _authenticate(self, **kwargs): response = requests.post( '%s/authentication/token' % self.base_url, data=kwargs ) if response.status_code != 200: raise GhostException(response.status_code, response.json().get('errors', [])) data = response.json() self._access_token = data.get('access_token') self._refresh_token = data.get('refresh_token') return data def revoke_access_token(self): """ Revoke the access token currently in use. """ if not self._access_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='access_token', token=self._access_token )) self._access_token = None def revoke_refresh_token(self): """ Revoke the refresh token currently active. """ if not self._refresh_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='refresh_token', token=self._refresh_token )) self._refresh_token = None def logout(self): """ Log out, revoking the access tokens and forgetting the login details if they were given. """ self.revoke_refresh_token() self.revoke_access_token() self._username, self._password = None, None def upload(self, file_obj=None, file_path=None, name=None, data=None): """ Upload an image and return its path on the server. Either `file_obj` or `file_path` or `name` and `data` has to be specified. :param file_obj: A file object to upload :param file_path: A file path to upload from :param name: A file name for uploading :param data: The file content to upload :return: The path of the uploaded file on the server """ close = False if file_obj: file_name, content = os.path.basename(file_obj.name), file_obj elif file_path: file_name, content = os.path.basename(file_path), open(file_path, 'rb') close = True elif name and data: file_name, content = name, data else: raise GhostException( 400, 'Either `file_obj` or `file_path` or ' '`name` and `data` needs to be specified' ) try: content_type, _ = mimetypes.guess_type(file_name) file_arg = (file_name, content, content_type) response = self.execute_post('uploads/', files={'uploadimage': file_arg}) return response finally: if close: content.close() @refresh_session_if_necessary def execute_get(self, resource, **kwargs): """ Execute an HTTP GET request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional query parameters (and optionally headers) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if kwargs: separator = '&' if '?' in url else '?' for key, value in kwargs.items(): if hasattr(value, '__iter__') and type(value) not in six.string_types: url = '%s%s%s=%s' % (url, separator, key, ','.join(value)) else: url = '%s%s%s=%s' % (url, separator, key, value) separator = '&' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token else: separator = '&' if '?' in url else '?' url = '%s%sclient_id=%s&client_secret=%s' % ( url, separator, self._client_id, self._client_secret ) response = requests.get(url, headers=headers) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response.json() def execute_post(self, resource, **kwargs): """ Execute an HTTP POST request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.post, **kwargs).json() def execute_put(self, resource, **kwargs): """ Execute an HTTP PUT request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.put, **kwargs).json() def execute_delete(self, resource, **kwargs): """ Execute an HTTP DELETE request against the API endpoints. This method is meant for internal use. Does not return anything but raises an exception when failed. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) """ self._request(resource, requests.delete, **kwargs) @refresh_session_if_necessary def _request(self, resource, request, **kwargs): if not self._access_token: raise GhostException(401, [{ 'errorType': 'ClientError', 'message': 'Access token not found' }]) url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) if 'json' in kwargs: headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token response = request(url, headers=headers, **kwargs) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response
rycus86/ghost-client
ghost_client/api.py
Ghost.login
python
def login(self, username, password): data = self._authenticate( grant_type='password', username=username, password=password, client_id=self._client_id, client_secret=self._client_secret ) self._username = username self._password = password return data
Authenticate with the server. :param username: The username of an existing user :param password: The password for the user :return: The authentication response from the REST endpoint
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/api.py#L201-L221
[ "def _authenticate(self, **kwargs):\n response = requests.post(\n '%s/authentication/token' % self.base_url, data=kwargs\n )\n\n if response.status_code != 200:\n raise GhostException(response.status_code, response.json().get('errors', []))\n\n data = response.json()\n\n self._access_token = data.get('access_token')\n self._refresh_token = data.get('refresh_token')\n\n return data\n" ]
class Ghost(object): """ API client for the Ghost REST endpoints. See https://api.ghost.org/ for the available parameters. Sample usage: from ghost_client import Ghost # to read the client ID and secret from the database ghost = Ghost.from_sqlite( '/var/lib/ghost/content/data/ghost.db', 'http://localhost:2368' ) # or to use a specific client ID and secret ghost = Ghost( 'http://localhost:2368', client_id='ghost-admin', client_secret='secret_key' ) # log in ghost.login('username', 'password') # print the server's version print(ghost.version) # create a new tag tag = ghost.tags.create(name='API sample') # create a new post using it post = ghost.posts.create( title='Example post', slug='custom-slug', markdown='', # yes, even on v1.+ custom_excerpt='An example post created from Python', tags=[tag] ) # list posts, tags and users posts = ghost.posts.list( status='all', fields=('id', 'title', 'slug'), formats=('html', 'mobiledoc', 'plaintext'), ) tags = ghost.tags.list(fields='name', limit='all') users = ghost.users.list(include='count.posts') # use pagination while posts: for post in posts: print(post) posts = posts.next_page() print(posts.total) print(posts.pages) # update a post & tag updated_post = ghost.posts.update(post.id, title='Updated title') updated_tag = ghost.tags.update(tag.id, name='Updated tag') # note: creating, updating and deleting a user is not allowed by the API # access fields as properties print(post.title) print(post.markdown) # needs formats='mobiledoc' print(post.author.name) # needs include='author' # delete a post & tag ghost.posts.delete(post.id) ghost.tags.delete(tag.id) # upload an image ghost.upload(file_obj=open('sample.png', 'rb')) ghost.upload(file_path='/path/to/image.jpeg', 'rb') ghost.upload(name='image.gif', data=open('local.gif', 'rb').read()) # log out ghost.logout() The logged in credentials will be saved in memory and on HTTP 401 errors the client will attempt to re-authenticate once automatically. Responses are wrapped in `models.ModelList` and `models.Model` types to allow pagination and retrieving fields as properties. """ DEFAULT_VERSION = '1' """ The default version to report when cannot be fetched. """ def __init__( self, base_url, version='auto', client_id=None, client_secret=None, access_token=None, refresh_token=None ): """ Creates a new Ghost API client. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: Self-supplied client ID (optional) :param client_secret: Self-supplied client secret (optional) :param access_token: Self-supplied access token (optional) :param refresh_token: Self-supplied refresh token (optional) """ self.base_url = '%s/ghost/api/v0.1' % base_url self._version = version self._client_id = client_id self._client_secret = client_secret self._access_token = access_token self._refresh_token = refresh_token self._username = None self._password = None if not self._client_id or not self._client_secret: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_id or client_secret given or found' }]) self.posts = PostController(self) self.tags = Controller(self, 'tags') self.users = Controller(self, 'users') @classmethod def from_sqlite(cls, database_path, base_url, version='auto', client_id='ghost-admin'): """ Initialize a new Ghost API client, reading the client ID and secret from the SQlite database. :param database_path: The path to the database file. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: The client ID to look for in the database :return: A new Ghost API client instance """ import os import sqlite3 fd = os.open(database_path, os.O_RDONLY) connection = sqlite3.connect('/dev/fd/%d' % fd) os.close(fd) try: row = connection.execute( 'SELECT secret FROM clients WHERE slug = ?', (client_id,) ).fetchone() if row: return cls( base_url, version=version, client_id=client_id, client_secret=row[0] ) else: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_secret found for client_id: %s' % client_id }]) finally: connection.close() @property def version(self): """ :return: The version of the server when initialized as 'auto', otherwise the version passed in at initialization """ if self._version != 'auto': return self._version if self._version == 'auto': try: data = self.execute_get('configuration/about/') self._version = data['configuration'][0]['version'] except GhostException: return self.DEFAULT_VERSION return self._version def refresh_session(self): """ Re-authenticate using the refresh token if available. Otherwise log in using the username and password if it was used to authenticate initially. :return: The authentication response or `None` if not available """ if not self._refresh_token: if self._username and self._password: return self.login(self._username, self._password) return return self._authenticate( grant_type='refresh_token', refresh_token=self._refresh_token, client_id=self._client_id, client_secret=self._client_secret ) def _authenticate(self, **kwargs): response = requests.post( '%s/authentication/token' % self.base_url, data=kwargs ) if response.status_code != 200: raise GhostException(response.status_code, response.json().get('errors', [])) data = response.json() self._access_token = data.get('access_token') self._refresh_token = data.get('refresh_token') return data def revoke_access_token(self): """ Revoke the access token currently in use. """ if not self._access_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='access_token', token=self._access_token )) self._access_token = None def revoke_refresh_token(self): """ Revoke the refresh token currently active. """ if not self._refresh_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='refresh_token', token=self._refresh_token )) self._refresh_token = None def logout(self): """ Log out, revoking the access tokens and forgetting the login details if they were given. """ self.revoke_refresh_token() self.revoke_access_token() self._username, self._password = None, None def upload(self, file_obj=None, file_path=None, name=None, data=None): """ Upload an image and return its path on the server. Either `file_obj` or `file_path` or `name` and `data` has to be specified. :param file_obj: A file object to upload :param file_path: A file path to upload from :param name: A file name for uploading :param data: The file content to upload :return: The path of the uploaded file on the server """ close = False if file_obj: file_name, content = os.path.basename(file_obj.name), file_obj elif file_path: file_name, content = os.path.basename(file_path), open(file_path, 'rb') close = True elif name and data: file_name, content = name, data else: raise GhostException( 400, 'Either `file_obj` or `file_path` or ' '`name` and `data` needs to be specified' ) try: content_type, _ = mimetypes.guess_type(file_name) file_arg = (file_name, content, content_type) response = self.execute_post('uploads/', files={'uploadimage': file_arg}) return response finally: if close: content.close() @refresh_session_if_necessary def execute_get(self, resource, **kwargs): """ Execute an HTTP GET request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional query parameters (and optionally headers) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if kwargs: separator = '&' if '?' in url else '?' for key, value in kwargs.items(): if hasattr(value, '__iter__') and type(value) not in six.string_types: url = '%s%s%s=%s' % (url, separator, key, ','.join(value)) else: url = '%s%s%s=%s' % (url, separator, key, value) separator = '&' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token else: separator = '&' if '?' in url else '?' url = '%s%sclient_id=%s&client_secret=%s' % ( url, separator, self._client_id, self._client_secret ) response = requests.get(url, headers=headers) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response.json() def execute_post(self, resource, **kwargs): """ Execute an HTTP POST request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.post, **kwargs).json() def execute_put(self, resource, **kwargs): """ Execute an HTTP PUT request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.put, **kwargs).json() def execute_delete(self, resource, **kwargs): """ Execute an HTTP DELETE request against the API endpoints. This method is meant for internal use. Does not return anything but raises an exception when failed. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) """ self._request(resource, requests.delete, **kwargs) @refresh_session_if_necessary def _request(self, resource, request, **kwargs): if not self._access_token: raise GhostException(401, [{ 'errorType': 'ClientError', 'message': 'Access token not found' }]) url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) if 'json' in kwargs: headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token response = request(url, headers=headers, **kwargs) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response
rycus86/ghost-client
ghost_client/api.py
Ghost.refresh_session
python
def refresh_session(self): if not self._refresh_token: if self._username and self._password: return self.login(self._username, self._password) return return self._authenticate( grant_type='refresh_token', refresh_token=self._refresh_token, client_id=self._client_id, client_secret=self._client_secret )
Re-authenticate using the refresh token if available. Otherwise log in using the username and password if it was used to authenticate initially. :return: The authentication response or `None` if not available
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/api.py#L223-L243
[ "def login(self, username, password):\n \"\"\"\n Authenticate with the server.\n\n :param username: The username of an existing user\n :param password: The password for the user\n :return: The authentication response from the REST endpoint\n \"\"\"\n\n data = self._authenticate(\n grant_type='password',\n username=username,\n password=password,\n client_id=self._client_id,\n client_secret=self._client_secret\n )\n\n self._username = username\n self._password = password\n\n return data\n", "def _authenticate(self, **kwargs):\n response = requests.post(\n '%s/authentication/token' % self.base_url, data=kwargs\n )\n\n if response.status_code != 200:\n raise GhostException(response.status_code, response.json().get('errors', []))\n\n data = response.json()\n\n self._access_token = data.get('access_token')\n self._refresh_token = data.get('refresh_token')\n\n return data\n" ]
class Ghost(object): """ API client for the Ghost REST endpoints. See https://api.ghost.org/ for the available parameters. Sample usage: from ghost_client import Ghost # to read the client ID and secret from the database ghost = Ghost.from_sqlite( '/var/lib/ghost/content/data/ghost.db', 'http://localhost:2368' ) # or to use a specific client ID and secret ghost = Ghost( 'http://localhost:2368', client_id='ghost-admin', client_secret='secret_key' ) # log in ghost.login('username', 'password') # print the server's version print(ghost.version) # create a new tag tag = ghost.tags.create(name='API sample') # create a new post using it post = ghost.posts.create( title='Example post', slug='custom-slug', markdown='', # yes, even on v1.+ custom_excerpt='An example post created from Python', tags=[tag] ) # list posts, tags and users posts = ghost.posts.list( status='all', fields=('id', 'title', 'slug'), formats=('html', 'mobiledoc', 'plaintext'), ) tags = ghost.tags.list(fields='name', limit='all') users = ghost.users.list(include='count.posts') # use pagination while posts: for post in posts: print(post) posts = posts.next_page() print(posts.total) print(posts.pages) # update a post & tag updated_post = ghost.posts.update(post.id, title='Updated title') updated_tag = ghost.tags.update(tag.id, name='Updated tag') # note: creating, updating and deleting a user is not allowed by the API # access fields as properties print(post.title) print(post.markdown) # needs formats='mobiledoc' print(post.author.name) # needs include='author' # delete a post & tag ghost.posts.delete(post.id) ghost.tags.delete(tag.id) # upload an image ghost.upload(file_obj=open('sample.png', 'rb')) ghost.upload(file_path='/path/to/image.jpeg', 'rb') ghost.upload(name='image.gif', data=open('local.gif', 'rb').read()) # log out ghost.logout() The logged in credentials will be saved in memory and on HTTP 401 errors the client will attempt to re-authenticate once automatically. Responses are wrapped in `models.ModelList` and `models.Model` types to allow pagination and retrieving fields as properties. """ DEFAULT_VERSION = '1' """ The default version to report when cannot be fetched. """ def __init__( self, base_url, version='auto', client_id=None, client_secret=None, access_token=None, refresh_token=None ): """ Creates a new Ghost API client. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: Self-supplied client ID (optional) :param client_secret: Self-supplied client secret (optional) :param access_token: Self-supplied access token (optional) :param refresh_token: Self-supplied refresh token (optional) """ self.base_url = '%s/ghost/api/v0.1' % base_url self._version = version self._client_id = client_id self._client_secret = client_secret self._access_token = access_token self._refresh_token = refresh_token self._username = None self._password = None if not self._client_id or not self._client_secret: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_id or client_secret given or found' }]) self.posts = PostController(self) self.tags = Controller(self, 'tags') self.users = Controller(self, 'users') @classmethod def from_sqlite(cls, database_path, base_url, version='auto', client_id='ghost-admin'): """ Initialize a new Ghost API client, reading the client ID and secret from the SQlite database. :param database_path: The path to the database file. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: The client ID to look for in the database :return: A new Ghost API client instance """ import os import sqlite3 fd = os.open(database_path, os.O_RDONLY) connection = sqlite3.connect('/dev/fd/%d' % fd) os.close(fd) try: row = connection.execute( 'SELECT secret FROM clients WHERE slug = ?', (client_id,) ).fetchone() if row: return cls( base_url, version=version, client_id=client_id, client_secret=row[0] ) else: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_secret found for client_id: %s' % client_id }]) finally: connection.close() @property def version(self): """ :return: The version of the server when initialized as 'auto', otherwise the version passed in at initialization """ if self._version != 'auto': return self._version if self._version == 'auto': try: data = self.execute_get('configuration/about/') self._version = data['configuration'][0]['version'] except GhostException: return self.DEFAULT_VERSION return self._version def login(self, username, password): """ Authenticate with the server. :param username: The username of an existing user :param password: The password for the user :return: The authentication response from the REST endpoint """ data = self._authenticate( grant_type='password', username=username, password=password, client_id=self._client_id, client_secret=self._client_secret ) self._username = username self._password = password return data def _authenticate(self, **kwargs): response = requests.post( '%s/authentication/token' % self.base_url, data=kwargs ) if response.status_code != 200: raise GhostException(response.status_code, response.json().get('errors', [])) data = response.json() self._access_token = data.get('access_token') self._refresh_token = data.get('refresh_token') return data def revoke_access_token(self): """ Revoke the access token currently in use. """ if not self._access_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='access_token', token=self._access_token )) self._access_token = None def revoke_refresh_token(self): """ Revoke the refresh token currently active. """ if not self._refresh_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='refresh_token', token=self._refresh_token )) self._refresh_token = None def logout(self): """ Log out, revoking the access tokens and forgetting the login details if they were given. """ self.revoke_refresh_token() self.revoke_access_token() self._username, self._password = None, None def upload(self, file_obj=None, file_path=None, name=None, data=None): """ Upload an image and return its path on the server. Either `file_obj` or `file_path` or `name` and `data` has to be specified. :param file_obj: A file object to upload :param file_path: A file path to upload from :param name: A file name for uploading :param data: The file content to upload :return: The path of the uploaded file on the server """ close = False if file_obj: file_name, content = os.path.basename(file_obj.name), file_obj elif file_path: file_name, content = os.path.basename(file_path), open(file_path, 'rb') close = True elif name and data: file_name, content = name, data else: raise GhostException( 400, 'Either `file_obj` or `file_path` or ' '`name` and `data` needs to be specified' ) try: content_type, _ = mimetypes.guess_type(file_name) file_arg = (file_name, content, content_type) response = self.execute_post('uploads/', files={'uploadimage': file_arg}) return response finally: if close: content.close() @refresh_session_if_necessary def execute_get(self, resource, **kwargs): """ Execute an HTTP GET request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional query parameters (and optionally headers) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if kwargs: separator = '&' if '?' in url else '?' for key, value in kwargs.items(): if hasattr(value, '__iter__') and type(value) not in six.string_types: url = '%s%s%s=%s' % (url, separator, key, ','.join(value)) else: url = '%s%s%s=%s' % (url, separator, key, value) separator = '&' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token else: separator = '&' if '?' in url else '?' url = '%s%sclient_id=%s&client_secret=%s' % ( url, separator, self._client_id, self._client_secret ) response = requests.get(url, headers=headers) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response.json() def execute_post(self, resource, **kwargs): """ Execute an HTTP POST request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.post, **kwargs).json() def execute_put(self, resource, **kwargs): """ Execute an HTTP PUT request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.put, **kwargs).json() def execute_delete(self, resource, **kwargs): """ Execute an HTTP DELETE request against the API endpoints. This method is meant for internal use. Does not return anything but raises an exception when failed. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) """ self._request(resource, requests.delete, **kwargs) @refresh_session_if_necessary def _request(self, resource, request, **kwargs): if not self._access_token: raise GhostException(401, [{ 'errorType': 'ClientError', 'message': 'Access token not found' }]) url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) if 'json' in kwargs: headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token response = request(url, headers=headers, **kwargs) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response
rycus86/ghost-client
ghost_client/api.py
Ghost.revoke_access_token
python
def revoke_access_token(self): if not self._access_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='access_token', token=self._access_token )) self._access_token = None
Revoke the access token currently in use.
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/api.py#L260-L273
[ "def execute_post(self, resource, **kwargs):\n \"\"\"\n Execute an HTTP POST request against the API endpoints.\n This method is meant for internal use.\n\n :param resource: The last part of the URI\n :param kwargs: Additional parameters for the HTTP call (`request` library)\n :return: The HTTP response as JSON or `GhostException` if unsuccessful\n \"\"\"\n\n return self._request(resource, requests.post, **kwargs).json()\n" ]
class Ghost(object): """ API client for the Ghost REST endpoints. See https://api.ghost.org/ for the available parameters. Sample usage: from ghost_client import Ghost # to read the client ID and secret from the database ghost = Ghost.from_sqlite( '/var/lib/ghost/content/data/ghost.db', 'http://localhost:2368' ) # or to use a specific client ID and secret ghost = Ghost( 'http://localhost:2368', client_id='ghost-admin', client_secret='secret_key' ) # log in ghost.login('username', 'password') # print the server's version print(ghost.version) # create a new tag tag = ghost.tags.create(name='API sample') # create a new post using it post = ghost.posts.create( title='Example post', slug='custom-slug', markdown='', # yes, even on v1.+ custom_excerpt='An example post created from Python', tags=[tag] ) # list posts, tags and users posts = ghost.posts.list( status='all', fields=('id', 'title', 'slug'), formats=('html', 'mobiledoc', 'plaintext'), ) tags = ghost.tags.list(fields='name', limit='all') users = ghost.users.list(include='count.posts') # use pagination while posts: for post in posts: print(post) posts = posts.next_page() print(posts.total) print(posts.pages) # update a post & tag updated_post = ghost.posts.update(post.id, title='Updated title') updated_tag = ghost.tags.update(tag.id, name='Updated tag') # note: creating, updating and deleting a user is not allowed by the API # access fields as properties print(post.title) print(post.markdown) # needs formats='mobiledoc' print(post.author.name) # needs include='author' # delete a post & tag ghost.posts.delete(post.id) ghost.tags.delete(tag.id) # upload an image ghost.upload(file_obj=open('sample.png', 'rb')) ghost.upload(file_path='/path/to/image.jpeg', 'rb') ghost.upload(name='image.gif', data=open('local.gif', 'rb').read()) # log out ghost.logout() The logged in credentials will be saved in memory and on HTTP 401 errors the client will attempt to re-authenticate once automatically. Responses are wrapped in `models.ModelList` and `models.Model` types to allow pagination and retrieving fields as properties. """ DEFAULT_VERSION = '1' """ The default version to report when cannot be fetched. """ def __init__( self, base_url, version='auto', client_id=None, client_secret=None, access_token=None, refresh_token=None ): """ Creates a new Ghost API client. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: Self-supplied client ID (optional) :param client_secret: Self-supplied client secret (optional) :param access_token: Self-supplied access token (optional) :param refresh_token: Self-supplied refresh token (optional) """ self.base_url = '%s/ghost/api/v0.1' % base_url self._version = version self._client_id = client_id self._client_secret = client_secret self._access_token = access_token self._refresh_token = refresh_token self._username = None self._password = None if not self._client_id or not self._client_secret: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_id or client_secret given or found' }]) self.posts = PostController(self) self.tags = Controller(self, 'tags') self.users = Controller(self, 'users') @classmethod def from_sqlite(cls, database_path, base_url, version='auto', client_id='ghost-admin'): """ Initialize a new Ghost API client, reading the client ID and secret from the SQlite database. :param database_path: The path to the database file. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: The client ID to look for in the database :return: A new Ghost API client instance """ import os import sqlite3 fd = os.open(database_path, os.O_RDONLY) connection = sqlite3.connect('/dev/fd/%d' % fd) os.close(fd) try: row = connection.execute( 'SELECT secret FROM clients WHERE slug = ?', (client_id,) ).fetchone() if row: return cls( base_url, version=version, client_id=client_id, client_secret=row[0] ) else: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_secret found for client_id: %s' % client_id }]) finally: connection.close() @property def version(self): """ :return: The version of the server when initialized as 'auto', otherwise the version passed in at initialization """ if self._version != 'auto': return self._version if self._version == 'auto': try: data = self.execute_get('configuration/about/') self._version = data['configuration'][0]['version'] except GhostException: return self.DEFAULT_VERSION return self._version def login(self, username, password): """ Authenticate with the server. :param username: The username of an existing user :param password: The password for the user :return: The authentication response from the REST endpoint """ data = self._authenticate( grant_type='password', username=username, password=password, client_id=self._client_id, client_secret=self._client_secret ) self._username = username self._password = password return data def refresh_session(self): """ Re-authenticate using the refresh token if available. Otherwise log in using the username and password if it was used to authenticate initially. :return: The authentication response or `None` if not available """ if not self._refresh_token: if self._username and self._password: return self.login(self._username, self._password) return return self._authenticate( grant_type='refresh_token', refresh_token=self._refresh_token, client_id=self._client_id, client_secret=self._client_secret ) def _authenticate(self, **kwargs): response = requests.post( '%s/authentication/token' % self.base_url, data=kwargs ) if response.status_code != 200: raise GhostException(response.status_code, response.json().get('errors', [])) data = response.json() self._access_token = data.get('access_token') self._refresh_token = data.get('refresh_token') return data def revoke_refresh_token(self): """ Revoke the refresh token currently active. """ if not self._refresh_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='refresh_token', token=self._refresh_token )) self._refresh_token = None def logout(self): """ Log out, revoking the access tokens and forgetting the login details if they were given. """ self.revoke_refresh_token() self.revoke_access_token() self._username, self._password = None, None def upload(self, file_obj=None, file_path=None, name=None, data=None): """ Upload an image and return its path on the server. Either `file_obj` or `file_path` or `name` and `data` has to be specified. :param file_obj: A file object to upload :param file_path: A file path to upload from :param name: A file name for uploading :param data: The file content to upload :return: The path of the uploaded file on the server """ close = False if file_obj: file_name, content = os.path.basename(file_obj.name), file_obj elif file_path: file_name, content = os.path.basename(file_path), open(file_path, 'rb') close = True elif name and data: file_name, content = name, data else: raise GhostException( 400, 'Either `file_obj` or `file_path` or ' '`name` and `data` needs to be specified' ) try: content_type, _ = mimetypes.guess_type(file_name) file_arg = (file_name, content, content_type) response = self.execute_post('uploads/', files={'uploadimage': file_arg}) return response finally: if close: content.close() @refresh_session_if_necessary def execute_get(self, resource, **kwargs): """ Execute an HTTP GET request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional query parameters (and optionally headers) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if kwargs: separator = '&' if '?' in url else '?' for key, value in kwargs.items(): if hasattr(value, '__iter__') and type(value) not in six.string_types: url = '%s%s%s=%s' % (url, separator, key, ','.join(value)) else: url = '%s%s%s=%s' % (url, separator, key, value) separator = '&' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token else: separator = '&' if '?' in url else '?' url = '%s%sclient_id=%s&client_secret=%s' % ( url, separator, self._client_id, self._client_secret ) response = requests.get(url, headers=headers) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response.json() def execute_post(self, resource, **kwargs): """ Execute an HTTP POST request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.post, **kwargs).json() def execute_put(self, resource, **kwargs): """ Execute an HTTP PUT request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.put, **kwargs).json() def execute_delete(self, resource, **kwargs): """ Execute an HTTP DELETE request against the API endpoints. This method is meant for internal use. Does not return anything but raises an exception when failed. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) """ self._request(resource, requests.delete, **kwargs) @refresh_session_if_necessary def _request(self, resource, request, **kwargs): if not self._access_token: raise GhostException(401, [{ 'errorType': 'ClientError', 'message': 'Access token not found' }]) url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) if 'json' in kwargs: headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token response = request(url, headers=headers, **kwargs) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response
rycus86/ghost-client
ghost_client/api.py
Ghost.revoke_refresh_token
python
def revoke_refresh_token(self): if not self._refresh_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='refresh_token', token=self._refresh_token )) self._refresh_token = None
Revoke the refresh token currently active.
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/api.py#L275-L288
[ "def execute_post(self, resource, **kwargs):\n \"\"\"\n Execute an HTTP POST request against the API endpoints.\n This method is meant for internal use.\n\n :param resource: The last part of the URI\n :param kwargs: Additional parameters for the HTTP call (`request` library)\n :return: The HTTP response as JSON or `GhostException` if unsuccessful\n \"\"\"\n\n return self._request(resource, requests.post, **kwargs).json()\n" ]
class Ghost(object): """ API client for the Ghost REST endpoints. See https://api.ghost.org/ for the available parameters. Sample usage: from ghost_client import Ghost # to read the client ID and secret from the database ghost = Ghost.from_sqlite( '/var/lib/ghost/content/data/ghost.db', 'http://localhost:2368' ) # or to use a specific client ID and secret ghost = Ghost( 'http://localhost:2368', client_id='ghost-admin', client_secret='secret_key' ) # log in ghost.login('username', 'password') # print the server's version print(ghost.version) # create a new tag tag = ghost.tags.create(name='API sample') # create a new post using it post = ghost.posts.create( title='Example post', slug='custom-slug', markdown='', # yes, even on v1.+ custom_excerpt='An example post created from Python', tags=[tag] ) # list posts, tags and users posts = ghost.posts.list( status='all', fields=('id', 'title', 'slug'), formats=('html', 'mobiledoc', 'plaintext'), ) tags = ghost.tags.list(fields='name', limit='all') users = ghost.users.list(include='count.posts') # use pagination while posts: for post in posts: print(post) posts = posts.next_page() print(posts.total) print(posts.pages) # update a post & tag updated_post = ghost.posts.update(post.id, title='Updated title') updated_tag = ghost.tags.update(tag.id, name='Updated tag') # note: creating, updating and deleting a user is not allowed by the API # access fields as properties print(post.title) print(post.markdown) # needs formats='mobiledoc' print(post.author.name) # needs include='author' # delete a post & tag ghost.posts.delete(post.id) ghost.tags.delete(tag.id) # upload an image ghost.upload(file_obj=open('sample.png', 'rb')) ghost.upload(file_path='/path/to/image.jpeg', 'rb') ghost.upload(name='image.gif', data=open('local.gif', 'rb').read()) # log out ghost.logout() The logged in credentials will be saved in memory and on HTTP 401 errors the client will attempt to re-authenticate once automatically. Responses are wrapped in `models.ModelList` and `models.Model` types to allow pagination and retrieving fields as properties. """ DEFAULT_VERSION = '1' """ The default version to report when cannot be fetched. """ def __init__( self, base_url, version='auto', client_id=None, client_secret=None, access_token=None, refresh_token=None ): """ Creates a new Ghost API client. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: Self-supplied client ID (optional) :param client_secret: Self-supplied client secret (optional) :param access_token: Self-supplied access token (optional) :param refresh_token: Self-supplied refresh token (optional) """ self.base_url = '%s/ghost/api/v0.1' % base_url self._version = version self._client_id = client_id self._client_secret = client_secret self._access_token = access_token self._refresh_token = refresh_token self._username = None self._password = None if not self._client_id or not self._client_secret: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_id or client_secret given or found' }]) self.posts = PostController(self) self.tags = Controller(self, 'tags') self.users = Controller(self, 'users') @classmethod def from_sqlite(cls, database_path, base_url, version='auto', client_id='ghost-admin'): """ Initialize a new Ghost API client, reading the client ID and secret from the SQlite database. :param database_path: The path to the database file. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: The client ID to look for in the database :return: A new Ghost API client instance """ import os import sqlite3 fd = os.open(database_path, os.O_RDONLY) connection = sqlite3.connect('/dev/fd/%d' % fd) os.close(fd) try: row = connection.execute( 'SELECT secret FROM clients WHERE slug = ?', (client_id,) ).fetchone() if row: return cls( base_url, version=version, client_id=client_id, client_secret=row[0] ) else: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_secret found for client_id: %s' % client_id }]) finally: connection.close() @property def version(self): """ :return: The version of the server when initialized as 'auto', otherwise the version passed in at initialization """ if self._version != 'auto': return self._version if self._version == 'auto': try: data = self.execute_get('configuration/about/') self._version = data['configuration'][0]['version'] except GhostException: return self.DEFAULT_VERSION return self._version def login(self, username, password): """ Authenticate with the server. :param username: The username of an existing user :param password: The password for the user :return: The authentication response from the REST endpoint """ data = self._authenticate( grant_type='password', username=username, password=password, client_id=self._client_id, client_secret=self._client_secret ) self._username = username self._password = password return data def refresh_session(self): """ Re-authenticate using the refresh token if available. Otherwise log in using the username and password if it was used to authenticate initially. :return: The authentication response or `None` if not available """ if not self._refresh_token: if self._username and self._password: return self.login(self._username, self._password) return return self._authenticate( grant_type='refresh_token', refresh_token=self._refresh_token, client_id=self._client_id, client_secret=self._client_secret ) def _authenticate(self, **kwargs): response = requests.post( '%s/authentication/token' % self.base_url, data=kwargs ) if response.status_code != 200: raise GhostException(response.status_code, response.json().get('errors', [])) data = response.json() self._access_token = data.get('access_token') self._refresh_token = data.get('refresh_token') return data def revoke_access_token(self): """ Revoke the access token currently in use. """ if not self._access_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='access_token', token=self._access_token )) self._access_token = None def logout(self): """ Log out, revoking the access tokens and forgetting the login details if they were given. """ self.revoke_refresh_token() self.revoke_access_token() self._username, self._password = None, None def upload(self, file_obj=None, file_path=None, name=None, data=None): """ Upload an image and return its path on the server. Either `file_obj` or `file_path` or `name` and `data` has to be specified. :param file_obj: A file object to upload :param file_path: A file path to upload from :param name: A file name for uploading :param data: The file content to upload :return: The path of the uploaded file on the server """ close = False if file_obj: file_name, content = os.path.basename(file_obj.name), file_obj elif file_path: file_name, content = os.path.basename(file_path), open(file_path, 'rb') close = True elif name and data: file_name, content = name, data else: raise GhostException( 400, 'Either `file_obj` or `file_path` or ' '`name` and `data` needs to be specified' ) try: content_type, _ = mimetypes.guess_type(file_name) file_arg = (file_name, content, content_type) response = self.execute_post('uploads/', files={'uploadimage': file_arg}) return response finally: if close: content.close() @refresh_session_if_necessary def execute_get(self, resource, **kwargs): """ Execute an HTTP GET request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional query parameters (and optionally headers) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if kwargs: separator = '&' if '?' in url else '?' for key, value in kwargs.items(): if hasattr(value, '__iter__') and type(value) not in six.string_types: url = '%s%s%s=%s' % (url, separator, key, ','.join(value)) else: url = '%s%s%s=%s' % (url, separator, key, value) separator = '&' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token else: separator = '&' if '?' in url else '?' url = '%s%sclient_id=%s&client_secret=%s' % ( url, separator, self._client_id, self._client_secret ) response = requests.get(url, headers=headers) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response.json() def execute_post(self, resource, **kwargs): """ Execute an HTTP POST request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.post, **kwargs).json() def execute_put(self, resource, **kwargs): """ Execute an HTTP PUT request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.put, **kwargs).json() def execute_delete(self, resource, **kwargs): """ Execute an HTTP DELETE request against the API endpoints. This method is meant for internal use. Does not return anything but raises an exception when failed. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) """ self._request(resource, requests.delete, **kwargs) @refresh_session_if_necessary def _request(self, resource, request, **kwargs): if not self._access_token: raise GhostException(401, [{ 'errorType': 'ClientError', 'message': 'Access token not found' }]) url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) if 'json' in kwargs: headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token response = request(url, headers=headers, **kwargs) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response
rycus86/ghost-client
ghost_client/api.py
Ghost.logout
python
def logout(self): self.revoke_refresh_token() self.revoke_access_token() self._username, self._password = None, None
Log out, revoking the access tokens and forgetting the login details if they were given.
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/api.py#L290-L299
[ "def revoke_access_token(self):\n \"\"\"\n Revoke the access token currently in use.\n \"\"\"\n\n if not self._access_token:\n return\n\n self.execute_post('authentication/revoke', json=dict(\n token_type_hint='access_token',\n token=self._access_token\n ))\n\n self._access_token = None\n", "def revoke_refresh_token(self):\n \"\"\"\n Revoke the refresh token currently active.\n \"\"\"\n\n if not self._refresh_token:\n return\n\n self.execute_post('authentication/revoke', json=dict(\n token_type_hint='refresh_token',\n token=self._refresh_token\n ))\n\n self._refresh_token = None\n" ]
class Ghost(object): """ API client for the Ghost REST endpoints. See https://api.ghost.org/ for the available parameters. Sample usage: from ghost_client import Ghost # to read the client ID and secret from the database ghost = Ghost.from_sqlite( '/var/lib/ghost/content/data/ghost.db', 'http://localhost:2368' ) # or to use a specific client ID and secret ghost = Ghost( 'http://localhost:2368', client_id='ghost-admin', client_secret='secret_key' ) # log in ghost.login('username', 'password') # print the server's version print(ghost.version) # create a new tag tag = ghost.tags.create(name='API sample') # create a new post using it post = ghost.posts.create( title='Example post', slug='custom-slug', markdown='', # yes, even on v1.+ custom_excerpt='An example post created from Python', tags=[tag] ) # list posts, tags and users posts = ghost.posts.list( status='all', fields=('id', 'title', 'slug'), formats=('html', 'mobiledoc', 'plaintext'), ) tags = ghost.tags.list(fields='name', limit='all') users = ghost.users.list(include='count.posts') # use pagination while posts: for post in posts: print(post) posts = posts.next_page() print(posts.total) print(posts.pages) # update a post & tag updated_post = ghost.posts.update(post.id, title='Updated title') updated_tag = ghost.tags.update(tag.id, name='Updated tag') # note: creating, updating and deleting a user is not allowed by the API # access fields as properties print(post.title) print(post.markdown) # needs formats='mobiledoc' print(post.author.name) # needs include='author' # delete a post & tag ghost.posts.delete(post.id) ghost.tags.delete(tag.id) # upload an image ghost.upload(file_obj=open('sample.png', 'rb')) ghost.upload(file_path='/path/to/image.jpeg', 'rb') ghost.upload(name='image.gif', data=open('local.gif', 'rb').read()) # log out ghost.logout() The logged in credentials will be saved in memory and on HTTP 401 errors the client will attempt to re-authenticate once automatically. Responses are wrapped in `models.ModelList` and `models.Model` types to allow pagination and retrieving fields as properties. """ DEFAULT_VERSION = '1' """ The default version to report when cannot be fetched. """ def __init__( self, base_url, version='auto', client_id=None, client_secret=None, access_token=None, refresh_token=None ): """ Creates a new Ghost API client. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: Self-supplied client ID (optional) :param client_secret: Self-supplied client secret (optional) :param access_token: Self-supplied access token (optional) :param refresh_token: Self-supplied refresh token (optional) """ self.base_url = '%s/ghost/api/v0.1' % base_url self._version = version self._client_id = client_id self._client_secret = client_secret self._access_token = access_token self._refresh_token = refresh_token self._username = None self._password = None if not self._client_id or not self._client_secret: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_id or client_secret given or found' }]) self.posts = PostController(self) self.tags = Controller(self, 'tags') self.users = Controller(self, 'users') @classmethod def from_sqlite(cls, database_path, base_url, version='auto', client_id='ghost-admin'): """ Initialize a new Ghost API client, reading the client ID and secret from the SQlite database. :param database_path: The path to the database file. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: The client ID to look for in the database :return: A new Ghost API client instance """ import os import sqlite3 fd = os.open(database_path, os.O_RDONLY) connection = sqlite3.connect('/dev/fd/%d' % fd) os.close(fd) try: row = connection.execute( 'SELECT secret FROM clients WHERE slug = ?', (client_id,) ).fetchone() if row: return cls( base_url, version=version, client_id=client_id, client_secret=row[0] ) else: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_secret found for client_id: %s' % client_id }]) finally: connection.close() @property def version(self): """ :return: The version of the server when initialized as 'auto', otherwise the version passed in at initialization """ if self._version != 'auto': return self._version if self._version == 'auto': try: data = self.execute_get('configuration/about/') self._version = data['configuration'][0]['version'] except GhostException: return self.DEFAULT_VERSION return self._version def login(self, username, password): """ Authenticate with the server. :param username: The username of an existing user :param password: The password for the user :return: The authentication response from the REST endpoint """ data = self._authenticate( grant_type='password', username=username, password=password, client_id=self._client_id, client_secret=self._client_secret ) self._username = username self._password = password return data def refresh_session(self): """ Re-authenticate using the refresh token if available. Otherwise log in using the username and password if it was used to authenticate initially. :return: The authentication response or `None` if not available """ if not self._refresh_token: if self._username and self._password: return self.login(self._username, self._password) return return self._authenticate( grant_type='refresh_token', refresh_token=self._refresh_token, client_id=self._client_id, client_secret=self._client_secret ) def _authenticate(self, **kwargs): response = requests.post( '%s/authentication/token' % self.base_url, data=kwargs ) if response.status_code != 200: raise GhostException(response.status_code, response.json().get('errors', [])) data = response.json() self._access_token = data.get('access_token') self._refresh_token = data.get('refresh_token') return data def revoke_access_token(self): """ Revoke the access token currently in use. """ if not self._access_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='access_token', token=self._access_token )) self._access_token = None def revoke_refresh_token(self): """ Revoke the refresh token currently active. """ if not self._refresh_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='refresh_token', token=self._refresh_token )) self._refresh_token = None def upload(self, file_obj=None, file_path=None, name=None, data=None): """ Upload an image and return its path on the server. Either `file_obj` or `file_path` or `name` and `data` has to be specified. :param file_obj: A file object to upload :param file_path: A file path to upload from :param name: A file name for uploading :param data: The file content to upload :return: The path of the uploaded file on the server """ close = False if file_obj: file_name, content = os.path.basename(file_obj.name), file_obj elif file_path: file_name, content = os.path.basename(file_path), open(file_path, 'rb') close = True elif name and data: file_name, content = name, data else: raise GhostException( 400, 'Either `file_obj` or `file_path` or ' '`name` and `data` needs to be specified' ) try: content_type, _ = mimetypes.guess_type(file_name) file_arg = (file_name, content, content_type) response = self.execute_post('uploads/', files={'uploadimage': file_arg}) return response finally: if close: content.close() @refresh_session_if_necessary def execute_get(self, resource, **kwargs): """ Execute an HTTP GET request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional query parameters (and optionally headers) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if kwargs: separator = '&' if '?' in url else '?' for key, value in kwargs.items(): if hasattr(value, '__iter__') and type(value) not in six.string_types: url = '%s%s%s=%s' % (url, separator, key, ','.join(value)) else: url = '%s%s%s=%s' % (url, separator, key, value) separator = '&' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token else: separator = '&' if '?' in url else '?' url = '%s%sclient_id=%s&client_secret=%s' % ( url, separator, self._client_id, self._client_secret ) response = requests.get(url, headers=headers) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response.json() def execute_post(self, resource, **kwargs): """ Execute an HTTP POST request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.post, **kwargs).json() def execute_put(self, resource, **kwargs): """ Execute an HTTP PUT request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.put, **kwargs).json() def execute_delete(self, resource, **kwargs): """ Execute an HTTP DELETE request against the API endpoints. This method is meant for internal use. Does not return anything but raises an exception when failed. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) """ self._request(resource, requests.delete, **kwargs) @refresh_session_if_necessary def _request(self, resource, request, **kwargs): if not self._access_token: raise GhostException(401, [{ 'errorType': 'ClientError', 'message': 'Access token not found' }]) url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) if 'json' in kwargs: headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token response = request(url, headers=headers, **kwargs) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response
rycus86/ghost-client
ghost_client/api.py
Ghost.upload
python
def upload(self, file_obj=None, file_path=None, name=None, data=None): close = False if file_obj: file_name, content = os.path.basename(file_obj.name), file_obj elif file_path: file_name, content = os.path.basename(file_path), open(file_path, 'rb') close = True elif name and data: file_name, content = name, data else: raise GhostException( 400, 'Either `file_obj` or `file_path` or ' '`name` and `data` needs to be specified' ) try: content_type, _ = mimetypes.guess_type(file_name) file_arg = (file_name, content, content_type) response = self.execute_post('uploads/', files={'uploadimage': file_arg}) return response finally: if close: content.close()
Upload an image and return its path on the server. Either `file_obj` or `file_path` or `name` and `data` has to be specified. :param file_obj: A file object to upload :param file_path: A file path to upload from :param name: A file name for uploading :param data: The file content to upload :return: The path of the uploaded file on the server
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/api.py#L301-L343
[ "def execute_post(self, resource, **kwargs):\n \"\"\"\n Execute an HTTP POST request against the API endpoints.\n This method is meant for internal use.\n\n :param resource: The last part of the URI\n :param kwargs: Additional parameters for the HTTP call (`request` library)\n :return: The HTTP response as JSON or `GhostException` if unsuccessful\n \"\"\"\n\n return self._request(resource, requests.post, **kwargs).json()\n" ]
class Ghost(object): """ API client for the Ghost REST endpoints. See https://api.ghost.org/ for the available parameters. Sample usage: from ghost_client import Ghost # to read the client ID and secret from the database ghost = Ghost.from_sqlite( '/var/lib/ghost/content/data/ghost.db', 'http://localhost:2368' ) # or to use a specific client ID and secret ghost = Ghost( 'http://localhost:2368', client_id='ghost-admin', client_secret='secret_key' ) # log in ghost.login('username', 'password') # print the server's version print(ghost.version) # create a new tag tag = ghost.tags.create(name='API sample') # create a new post using it post = ghost.posts.create( title='Example post', slug='custom-slug', markdown='', # yes, even on v1.+ custom_excerpt='An example post created from Python', tags=[tag] ) # list posts, tags and users posts = ghost.posts.list( status='all', fields=('id', 'title', 'slug'), formats=('html', 'mobiledoc', 'plaintext'), ) tags = ghost.tags.list(fields='name', limit='all') users = ghost.users.list(include='count.posts') # use pagination while posts: for post in posts: print(post) posts = posts.next_page() print(posts.total) print(posts.pages) # update a post & tag updated_post = ghost.posts.update(post.id, title='Updated title') updated_tag = ghost.tags.update(tag.id, name='Updated tag') # note: creating, updating and deleting a user is not allowed by the API # access fields as properties print(post.title) print(post.markdown) # needs formats='mobiledoc' print(post.author.name) # needs include='author' # delete a post & tag ghost.posts.delete(post.id) ghost.tags.delete(tag.id) # upload an image ghost.upload(file_obj=open('sample.png', 'rb')) ghost.upload(file_path='/path/to/image.jpeg', 'rb') ghost.upload(name='image.gif', data=open('local.gif', 'rb').read()) # log out ghost.logout() The logged in credentials will be saved in memory and on HTTP 401 errors the client will attempt to re-authenticate once automatically. Responses are wrapped in `models.ModelList` and `models.Model` types to allow pagination and retrieving fields as properties. """ DEFAULT_VERSION = '1' """ The default version to report when cannot be fetched. """ def __init__( self, base_url, version='auto', client_id=None, client_secret=None, access_token=None, refresh_token=None ): """ Creates a new Ghost API client. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: Self-supplied client ID (optional) :param client_secret: Self-supplied client secret (optional) :param access_token: Self-supplied access token (optional) :param refresh_token: Self-supplied refresh token (optional) """ self.base_url = '%s/ghost/api/v0.1' % base_url self._version = version self._client_id = client_id self._client_secret = client_secret self._access_token = access_token self._refresh_token = refresh_token self._username = None self._password = None if not self._client_id or not self._client_secret: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_id or client_secret given or found' }]) self.posts = PostController(self) self.tags = Controller(self, 'tags') self.users = Controller(self, 'users') @classmethod def from_sqlite(cls, database_path, base_url, version='auto', client_id='ghost-admin'): """ Initialize a new Ghost API client, reading the client ID and secret from the SQlite database. :param database_path: The path to the database file. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: The client ID to look for in the database :return: A new Ghost API client instance """ import os import sqlite3 fd = os.open(database_path, os.O_RDONLY) connection = sqlite3.connect('/dev/fd/%d' % fd) os.close(fd) try: row = connection.execute( 'SELECT secret FROM clients WHERE slug = ?', (client_id,) ).fetchone() if row: return cls( base_url, version=version, client_id=client_id, client_secret=row[0] ) else: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_secret found for client_id: %s' % client_id }]) finally: connection.close() @property def version(self): """ :return: The version of the server when initialized as 'auto', otherwise the version passed in at initialization """ if self._version != 'auto': return self._version if self._version == 'auto': try: data = self.execute_get('configuration/about/') self._version = data['configuration'][0]['version'] except GhostException: return self.DEFAULT_VERSION return self._version def login(self, username, password): """ Authenticate with the server. :param username: The username of an existing user :param password: The password for the user :return: The authentication response from the REST endpoint """ data = self._authenticate( grant_type='password', username=username, password=password, client_id=self._client_id, client_secret=self._client_secret ) self._username = username self._password = password return data def refresh_session(self): """ Re-authenticate using the refresh token if available. Otherwise log in using the username and password if it was used to authenticate initially. :return: The authentication response or `None` if not available """ if not self._refresh_token: if self._username and self._password: return self.login(self._username, self._password) return return self._authenticate( grant_type='refresh_token', refresh_token=self._refresh_token, client_id=self._client_id, client_secret=self._client_secret ) def _authenticate(self, **kwargs): response = requests.post( '%s/authentication/token' % self.base_url, data=kwargs ) if response.status_code != 200: raise GhostException(response.status_code, response.json().get('errors', [])) data = response.json() self._access_token = data.get('access_token') self._refresh_token = data.get('refresh_token') return data def revoke_access_token(self): """ Revoke the access token currently in use. """ if not self._access_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='access_token', token=self._access_token )) self._access_token = None def revoke_refresh_token(self): """ Revoke the refresh token currently active. """ if not self._refresh_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='refresh_token', token=self._refresh_token )) self._refresh_token = None def logout(self): """ Log out, revoking the access tokens and forgetting the login details if they were given. """ self.revoke_refresh_token() self.revoke_access_token() self._username, self._password = None, None @refresh_session_if_necessary def execute_get(self, resource, **kwargs): """ Execute an HTTP GET request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional query parameters (and optionally headers) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if kwargs: separator = '&' if '?' in url else '?' for key, value in kwargs.items(): if hasattr(value, '__iter__') and type(value) not in six.string_types: url = '%s%s%s=%s' % (url, separator, key, ','.join(value)) else: url = '%s%s%s=%s' % (url, separator, key, value) separator = '&' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token else: separator = '&' if '?' in url else '?' url = '%s%sclient_id=%s&client_secret=%s' % ( url, separator, self._client_id, self._client_secret ) response = requests.get(url, headers=headers) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response.json() def execute_post(self, resource, **kwargs): """ Execute an HTTP POST request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.post, **kwargs).json() def execute_put(self, resource, **kwargs): """ Execute an HTTP PUT request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.put, **kwargs).json() def execute_delete(self, resource, **kwargs): """ Execute an HTTP DELETE request against the API endpoints. This method is meant for internal use. Does not return anything but raises an exception when failed. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) """ self._request(resource, requests.delete, **kwargs) @refresh_session_if_necessary def _request(self, resource, request, **kwargs): if not self._access_token: raise GhostException(401, [{ 'errorType': 'ClientError', 'message': 'Access token not found' }]) url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) if 'json' in kwargs: headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token response = request(url, headers=headers, **kwargs) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response
rycus86/ghost-client
ghost_client/api.py
Ghost.execute_get
python
def execute_get(self, resource, **kwargs): url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if kwargs: separator = '&' if '?' in url else '?' for key, value in kwargs.items(): if hasattr(value, '__iter__') and type(value) not in six.string_types: url = '%s%s%s=%s' % (url, separator, key, ','.join(value)) else: url = '%s%s%s=%s' % (url, separator, key, value) separator = '&' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token else: separator = '&' if '?' in url else '?' url = '%s%sclient_id=%s&client_secret=%s' % ( url, separator, self._client_id, self._client_secret ) response = requests.get(url, headers=headers) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response.json()
Execute an HTTP GET request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional query parameters (and optionally headers) :return: The HTTP response as JSON or `GhostException` if unsuccessful
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/api.py#L346-L389
null
class Ghost(object): """ API client for the Ghost REST endpoints. See https://api.ghost.org/ for the available parameters. Sample usage: from ghost_client import Ghost # to read the client ID and secret from the database ghost = Ghost.from_sqlite( '/var/lib/ghost/content/data/ghost.db', 'http://localhost:2368' ) # or to use a specific client ID and secret ghost = Ghost( 'http://localhost:2368', client_id='ghost-admin', client_secret='secret_key' ) # log in ghost.login('username', 'password') # print the server's version print(ghost.version) # create a new tag tag = ghost.tags.create(name='API sample') # create a new post using it post = ghost.posts.create( title='Example post', slug='custom-slug', markdown='', # yes, even on v1.+ custom_excerpt='An example post created from Python', tags=[tag] ) # list posts, tags and users posts = ghost.posts.list( status='all', fields=('id', 'title', 'slug'), formats=('html', 'mobiledoc', 'plaintext'), ) tags = ghost.tags.list(fields='name', limit='all') users = ghost.users.list(include='count.posts') # use pagination while posts: for post in posts: print(post) posts = posts.next_page() print(posts.total) print(posts.pages) # update a post & tag updated_post = ghost.posts.update(post.id, title='Updated title') updated_tag = ghost.tags.update(tag.id, name='Updated tag') # note: creating, updating and deleting a user is not allowed by the API # access fields as properties print(post.title) print(post.markdown) # needs formats='mobiledoc' print(post.author.name) # needs include='author' # delete a post & tag ghost.posts.delete(post.id) ghost.tags.delete(tag.id) # upload an image ghost.upload(file_obj=open('sample.png', 'rb')) ghost.upload(file_path='/path/to/image.jpeg', 'rb') ghost.upload(name='image.gif', data=open('local.gif', 'rb').read()) # log out ghost.logout() The logged in credentials will be saved in memory and on HTTP 401 errors the client will attempt to re-authenticate once automatically. Responses are wrapped in `models.ModelList` and `models.Model` types to allow pagination and retrieving fields as properties. """ DEFAULT_VERSION = '1' """ The default version to report when cannot be fetched. """ def __init__( self, base_url, version='auto', client_id=None, client_secret=None, access_token=None, refresh_token=None ): """ Creates a new Ghost API client. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: Self-supplied client ID (optional) :param client_secret: Self-supplied client secret (optional) :param access_token: Self-supplied access token (optional) :param refresh_token: Self-supplied refresh token (optional) """ self.base_url = '%s/ghost/api/v0.1' % base_url self._version = version self._client_id = client_id self._client_secret = client_secret self._access_token = access_token self._refresh_token = refresh_token self._username = None self._password = None if not self._client_id or not self._client_secret: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_id or client_secret given or found' }]) self.posts = PostController(self) self.tags = Controller(self, 'tags') self.users = Controller(self, 'users') @classmethod def from_sqlite(cls, database_path, base_url, version='auto', client_id='ghost-admin'): """ Initialize a new Ghost API client, reading the client ID and secret from the SQlite database. :param database_path: The path to the database file. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: The client ID to look for in the database :return: A new Ghost API client instance """ import os import sqlite3 fd = os.open(database_path, os.O_RDONLY) connection = sqlite3.connect('/dev/fd/%d' % fd) os.close(fd) try: row = connection.execute( 'SELECT secret FROM clients WHERE slug = ?', (client_id,) ).fetchone() if row: return cls( base_url, version=version, client_id=client_id, client_secret=row[0] ) else: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_secret found for client_id: %s' % client_id }]) finally: connection.close() @property def version(self): """ :return: The version of the server when initialized as 'auto', otherwise the version passed in at initialization """ if self._version != 'auto': return self._version if self._version == 'auto': try: data = self.execute_get('configuration/about/') self._version = data['configuration'][0]['version'] except GhostException: return self.DEFAULT_VERSION return self._version def login(self, username, password): """ Authenticate with the server. :param username: The username of an existing user :param password: The password for the user :return: The authentication response from the REST endpoint """ data = self._authenticate( grant_type='password', username=username, password=password, client_id=self._client_id, client_secret=self._client_secret ) self._username = username self._password = password return data def refresh_session(self): """ Re-authenticate using the refresh token if available. Otherwise log in using the username and password if it was used to authenticate initially. :return: The authentication response or `None` if not available """ if not self._refresh_token: if self._username and self._password: return self.login(self._username, self._password) return return self._authenticate( grant_type='refresh_token', refresh_token=self._refresh_token, client_id=self._client_id, client_secret=self._client_secret ) def _authenticate(self, **kwargs): response = requests.post( '%s/authentication/token' % self.base_url, data=kwargs ) if response.status_code != 200: raise GhostException(response.status_code, response.json().get('errors', [])) data = response.json() self._access_token = data.get('access_token') self._refresh_token = data.get('refresh_token') return data def revoke_access_token(self): """ Revoke the access token currently in use. """ if not self._access_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='access_token', token=self._access_token )) self._access_token = None def revoke_refresh_token(self): """ Revoke the refresh token currently active. """ if not self._refresh_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='refresh_token', token=self._refresh_token )) self._refresh_token = None def logout(self): """ Log out, revoking the access tokens and forgetting the login details if they were given. """ self.revoke_refresh_token() self.revoke_access_token() self._username, self._password = None, None def upload(self, file_obj=None, file_path=None, name=None, data=None): """ Upload an image and return its path on the server. Either `file_obj` or `file_path` or `name` and `data` has to be specified. :param file_obj: A file object to upload :param file_path: A file path to upload from :param name: A file name for uploading :param data: The file content to upload :return: The path of the uploaded file on the server """ close = False if file_obj: file_name, content = os.path.basename(file_obj.name), file_obj elif file_path: file_name, content = os.path.basename(file_path), open(file_path, 'rb') close = True elif name and data: file_name, content = name, data else: raise GhostException( 400, 'Either `file_obj` or `file_path` or ' '`name` and `data` needs to be specified' ) try: content_type, _ = mimetypes.guess_type(file_name) file_arg = (file_name, content, content_type) response = self.execute_post('uploads/', files={'uploadimage': file_arg}) return response finally: if close: content.close() @refresh_session_if_necessary def execute_post(self, resource, **kwargs): """ Execute an HTTP POST request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.post, **kwargs).json() def execute_put(self, resource, **kwargs): """ Execute an HTTP PUT request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.put, **kwargs).json() def execute_delete(self, resource, **kwargs): """ Execute an HTTP DELETE request against the API endpoints. This method is meant for internal use. Does not return anything but raises an exception when failed. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) """ self._request(resource, requests.delete, **kwargs) @refresh_session_if_necessary def _request(self, resource, request, **kwargs): if not self._access_token: raise GhostException(401, [{ 'errorType': 'ClientError', 'message': 'Access token not found' }]) url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) if 'json' in kwargs: headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token response = request(url, headers=headers, **kwargs) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response
rycus86/ghost-client
ghost_client/api.py
Ghost.execute_post
python
def execute_post(self, resource, **kwargs): return self._request(resource, requests.post, **kwargs).json()
Execute an HTTP POST request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/api.py#L391-L401
null
class Ghost(object): """ API client for the Ghost REST endpoints. See https://api.ghost.org/ for the available parameters. Sample usage: from ghost_client import Ghost # to read the client ID and secret from the database ghost = Ghost.from_sqlite( '/var/lib/ghost/content/data/ghost.db', 'http://localhost:2368' ) # or to use a specific client ID and secret ghost = Ghost( 'http://localhost:2368', client_id='ghost-admin', client_secret='secret_key' ) # log in ghost.login('username', 'password') # print the server's version print(ghost.version) # create a new tag tag = ghost.tags.create(name='API sample') # create a new post using it post = ghost.posts.create( title='Example post', slug='custom-slug', markdown='', # yes, even on v1.+ custom_excerpt='An example post created from Python', tags=[tag] ) # list posts, tags and users posts = ghost.posts.list( status='all', fields=('id', 'title', 'slug'), formats=('html', 'mobiledoc', 'plaintext'), ) tags = ghost.tags.list(fields='name', limit='all') users = ghost.users.list(include='count.posts') # use pagination while posts: for post in posts: print(post) posts = posts.next_page() print(posts.total) print(posts.pages) # update a post & tag updated_post = ghost.posts.update(post.id, title='Updated title') updated_tag = ghost.tags.update(tag.id, name='Updated tag') # note: creating, updating and deleting a user is not allowed by the API # access fields as properties print(post.title) print(post.markdown) # needs formats='mobiledoc' print(post.author.name) # needs include='author' # delete a post & tag ghost.posts.delete(post.id) ghost.tags.delete(tag.id) # upload an image ghost.upload(file_obj=open('sample.png', 'rb')) ghost.upload(file_path='/path/to/image.jpeg', 'rb') ghost.upload(name='image.gif', data=open('local.gif', 'rb').read()) # log out ghost.logout() The logged in credentials will be saved in memory and on HTTP 401 errors the client will attempt to re-authenticate once automatically. Responses are wrapped in `models.ModelList` and `models.Model` types to allow pagination and retrieving fields as properties. """ DEFAULT_VERSION = '1' """ The default version to report when cannot be fetched. """ def __init__( self, base_url, version='auto', client_id=None, client_secret=None, access_token=None, refresh_token=None ): """ Creates a new Ghost API client. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: Self-supplied client ID (optional) :param client_secret: Self-supplied client secret (optional) :param access_token: Self-supplied access token (optional) :param refresh_token: Self-supplied refresh token (optional) """ self.base_url = '%s/ghost/api/v0.1' % base_url self._version = version self._client_id = client_id self._client_secret = client_secret self._access_token = access_token self._refresh_token = refresh_token self._username = None self._password = None if not self._client_id or not self._client_secret: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_id or client_secret given or found' }]) self.posts = PostController(self) self.tags = Controller(self, 'tags') self.users = Controller(self, 'users') @classmethod def from_sqlite(cls, database_path, base_url, version='auto', client_id='ghost-admin'): """ Initialize a new Ghost API client, reading the client ID and secret from the SQlite database. :param database_path: The path to the database file. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: The client ID to look for in the database :return: A new Ghost API client instance """ import os import sqlite3 fd = os.open(database_path, os.O_RDONLY) connection = sqlite3.connect('/dev/fd/%d' % fd) os.close(fd) try: row = connection.execute( 'SELECT secret FROM clients WHERE slug = ?', (client_id,) ).fetchone() if row: return cls( base_url, version=version, client_id=client_id, client_secret=row[0] ) else: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_secret found for client_id: %s' % client_id }]) finally: connection.close() @property def version(self): """ :return: The version of the server when initialized as 'auto', otherwise the version passed in at initialization """ if self._version != 'auto': return self._version if self._version == 'auto': try: data = self.execute_get('configuration/about/') self._version = data['configuration'][0]['version'] except GhostException: return self.DEFAULT_VERSION return self._version def login(self, username, password): """ Authenticate with the server. :param username: The username of an existing user :param password: The password for the user :return: The authentication response from the REST endpoint """ data = self._authenticate( grant_type='password', username=username, password=password, client_id=self._client_id, client_secret=self._client_secret ) self._username = username self._password = password return data def refresh_session(self): """ Re-authenticate using the refresh token if available. Otherwise log in using the username and password if it was used to authenticate initially. :return: The authentication response or `None` if not available """ if not self._refresh_token: if self._username and self._password: return self.login(self._username, self._password) return return self._authenticate( grant_type='refresh_token', refresh_token=self._refresh_token, client_id=self._client_id, client_secret=self._client_secret ) def _authenticate(self, **kwargs): response = requests.post( '%s/authentication/token' % self.base_url, data=kwargs ) if response.status_code != 200: raise GhostException(response.status_code, response.json().get('errors', [])) data = response.json() self._access_token = data.get('access_token') self._refresh_token = data.get('refresh_token') return data def revoke_access_token(self): """ Revoke the access token currently in use. """ if not self._access_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='access_token', token=self._access_token )) self._access_token = None def revoke_refresh_token(self): """ Revoke the refresh token currently active. """ if not self._refresh_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='refresh_token', token=self._refresh_token )) self._refresh_token = None def logout(self): """ Log out, revoking the access tokens and forgetting the login details if they were given. """ self.revoke_refresh_token() self.revoke_access_token() self._username, self._password = None, None def upload(self, file_obj=None, file_path=None, name=None, data=None): """ Upload an image and return its path on the server. Either `file_obj` or `file_path` or `name` and `data` has to be specified. :param file_obj: A file object to upload :param file_path: A file path to upload from :param name: A file name for uploading :param data: The file content to upload :return: The path of the uploaded file on the server """ close = False if file_obj: file_name, content = os.path.basename(file_obj.name), file_obj elif file_path: file_name, content = os.path.basename(file_path), open(file_path, 'rb') close = True elif name and data: file_name, content = name, data else: raise GhostException( 400, 'Either `file_obj` or `file_path` or ' '`name` and `data` needs to be specified' ) try: content_type, _ = mimetypes.guess_type(file_name) file_arg = (file_name, content, content_type) response = self.execute_post('uploads/', files={'uploadimage': file_arg}) return response finally: if close: content.close() @refresh_session_if_necessary def execute_get(self, resource, **kwargs): """ Execute an HTTP GET request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional query parameters (and optionally headers) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if kwargs: separator = '&' if '?' in url else '?' for key, value in kwargs.items(): if hasattr(value, '__iter__') and type(value) not in six.string_types: url = '%s%s%s=%s' % (url, separator, key, ','.join(value)) else: url = '%s%s%s=%s' % (url, separator, key, value) separator = '&' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token else: separator = '&' if '?' in url else '?' url = '%s%sclient_id=%s&client_secret=%s' % ( url, separator, self._client_id, self._client_secret ) response = requests.get(url, headers=headers) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response.json() def execute_put(self, resource, **kwargs): """ Execute an HTTP PUT request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.put, **kwargs).json() def execute_delete(self, resource, **kwargs): """ Execute an HTTP DELETE request against the API endpoints. This method is meant for internal use. Does not return anything but raises an exception when failed. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) """ self._request(resource, requests.delete, **kwargs) @refresh_session_if_necessary def _request(self, resource, request, **kwargs): if not self._access_token: raise GhostException(401, [{ 'errorType': 'ClientError', 'message': 'Access token not found' }]) url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) if 'json' in kwargs: headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token response = request(url, headers=headers, **kwargs) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response
rycus86/ghost-client
ghost_client/api.py
Ghost.execute_put
python
def execute_put(self, resource, **kwargs): return self._request(resource, requests.put, **kwargs).json()
Execute an HTTP PUT request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/api.py#L403-L413
null
class Ghost(object): """ API client for the Ghost REST endpoints. See https://api.ghost.org/ for the available parameters. Sample usage: from ghost_client import Ghost # to read the client ID and secret from the database ghost = Ghost.from_sqlite( '/var/lib/ghost/content/data/ghost.db', 'http://localhost:2368' ) # or to use a specific client ID and secret ghost = Ghost( 'http://localhost:2368', client_id='ghost-admin', client_secret='secret_key' ) # log in ghost.login('username', 'password') # print the server's version print(ghost.version) # create a new tag tag = ghost.tags.create(name='API sample') # create a new post using it post = ghost.posts.create( title='Example post', slug='custom-slug', markdown='', # yes, even on v1.+ custom_excerpt='An example post created from Python', tags=[tag] ) # list posts, tags and users posts = ghost.posts.list( status='all', fields=('id', 'title', 'slug'), formats=('html', 'mobiledoc', 'plaintext'), ) tags = ghost.tags.list(fields='name', limit='all') users = ghost.users.list(include='count.posts') # use pagination while posts: for post in posts: print(post) posts = posts.next_page() print(posts.total) print(posts.pages) # update a post & tag updated_post = ghost.posts.update(post.id, title='Updated title') updated_tag = ghost.tags.update(tag.id, name='Updated tag') # note: creating, updating and deleting a user is not allowed by the API # access fields as properties print(post.title) print(post.markdown) # needs formats='mobiledoc' print(post.author.name) # needs include='author' # delete a post & tag ghost.posts.delete(post.id) ghost.tags.delete(tag.id) # upload an image ghost.upload(file_obj=open('sample.png', 'rb')) ghost.upload(file_path='/path/to/image.jpeg', 'rb') ghost.upload(name='image.gif', data=open('local.gif', 'rb').read()) # log out ghost.logout() The logged in credentials will be saved in memory and on HTTP 401 errors the client will attempt to re-authenticate once automatically. Responses are wrapped in `models.ModelList` and `models.Model` types to allow pagination and retrieving fields as properties. """ DEFAULT_VERSION = '1' """ The default version to report when cannot be fetched. """ def __init__( self, base_url, version='auto', client_id=None, client_secret=None, access_token=None, refresh_token=None ): """ Creates a new Ghost API client. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: Self-supplied client ID (optional) :param client_secret: Self-supplied client secret (optional) :param access_token: Self-supplied access token (optional) :param refresh_token: Self-supplied refresh token (optional) """ self.base_url = '%s/ghost/api/v0.1' % base_url self._version = version self._client_id = client_id self._client_secret = client_secret self._access_token = access_token self._refresh_token = refresh_token self._username = None self._password = None if not self._client_id or not self._client_secret: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_id or client_secret given or found' }]) self.posts = PostController(self) self.tags = Controller(self, 'tags') self.users = Controller(self, 'users') @classmethod def from_sqlite(cls, database_path, base_url, version='auto', client_id='ghost-admin'): """ Initialize a new Ghost API client, reading the client ID and secret from the SQlite database. :param database_path: The path to the database file. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: The client ID to look for in the database :return: A new Ghost API client instance """ import os import sqlite3 fd = os.open(database_path, os.O_RDONLY) connection = sqlite3.connect('/dev/fd/%d' % fd) os.close(fd) try: row = connection.execute( 'SELECT secret FROM clients WHERE slug = ?', (client_id,) ).fetchone() if row: return cls( base_url, version=version, client_id=client_id, client_secret=row[0] ) else: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_secret found for client_id: %s' % client_id }]) finally: connection.close() @property def version(self): """ :return: The version of the server when initialized as 'auto', otherwise the version passed in at initialization """ if self._version != 'auto': return self._version if self._version == 'auto': try: data = self.execute_get('configuration/about/') self._version = data['configuration'][0]['version'] except GhostException: return self.DEFAULT_VERSION return self._version def login(self, username, password): """ Authenticate with the server. :param username: The username of an existing user :param password: The password for the user :return: The authentication response from the REST endpoint """ data = self._authenticate( grant_type='password', username=username, password=password, client_id=self._client_id, client_secret=self._client_secret ) self._username = username self._password = password return data def refresh_session(self): """ Re-authenticate using the refresh token if available. Otherwise log in using the username and password if it was used to authenticate initially. :return: The authentication response or `None` if not available """ if not self._refresh_token: if self._username and self._password: return self.login(self._username, self._password) return return self._authenticate( grant_type='refresh_token', refresh_token=self._refresh_token, client_id=self._client_id, client_secret=self._client_secret ) def _authenticate(self, **kwargs): response = requests.post( '%s/authentication/token' % self.base_url, data=kwargs ) if response.status_code != 200: raise GhostException(response.status_code, response.json().get('errors', [])) data = response.json() self._access_token = data.get('access_token') self._refresh_token = data.get('refresh_token') return data def revoke_access_token(self): """ Revoke the access token currently in use. """ if not self._access_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='access_token', token=self._access_token )) self._access_token = None def revoke_refresh_token(self): """ Revoke the refresh token currently active. """ if not self._refresh_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='refresh_token', token=self._refresh_token )) self._refresh_token = None def logout(self): """ Log out, revoking the access tokens and forgetting the login details if they were given. """ self.revoke_refresh_token() self.revoke_access_token() self._username, self._password = None, None def upload(self, file_obj=None, file_path=None, name=None, data=None): """ Upload an image and return its path on the server. Either `file_obj` or `file_path` or `name` and `data` has to be specified. :param file_obj: A file object to upload :param file_path: A file path to upload from :param name: A file name for uploading :param data: The file content to upload :return: The path of the uploaded file on the server """ close = False if file_obj: file_name, content = os.path.basename(file_obj.name), file_obj elif file_path: file_name, content = os.path.basename(file_path), open(file_path, 'rb') close = True elif name and data: file_name, content = name, data else: raise GhostException( 400, 'Either `file_obj` or `file_path` or ' '`name` and `data` needs to be specified' ) try: content_type, _ = mimetypes.guess_type(file_name) file_arg = (file_name, content, content_type) response = self.execute_post('uploads/', files={'uploadimage': file_arg}) return response finally: if close: content.close() @refresh_session_if_necessary def execute_get(self, resource, **kwargs): """ Execute an HTTP GET request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional query parameters (and optionally headers) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if kwargs: separator = '&' if '?' in url else '?' for key, value in kwargs.items(): if hasattr(value, '__iter__') and type(value) not in six.string_types: url = '%s%s%s=%s' % (url, separator, key, ','.join(value)) else: url = '%s%s%s=%s' % (url, separator, key, value) separator = '&' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token else: separator = '&' if '?' in url else '?' url = '%s%sclient_id=%s&client_secret=%s' % ( url, separator, self._client_id, self._client_secret ) response = requests.get(url, headers=headers) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response.json() def execute_post(self, resource, **kwargs): """ Execute an HTTP POST request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.post, **kwargs).json() def execute_delete(self, resource, **kwargs): """ Execute an HTTP DELETE request against the API endpoints. This method is meant for internal use. Does not return anything but raises an exception when failed. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) """ self._request(resource, requests.delete, **kwargs) @refresh_session_if_necessary def _request(self, resource, request, **kwargs): if not self._access_token: raise GhostException(401, [{ 'errorType': 'ClientError', 'message': 'Access token not found' }]) url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) if 'json' in kwargs: headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token response = request(url, headers=headers, **kwargs) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response
rycus86/ghost-client
ghost_client/api.py
Ghost.execute_delete
python
def execute_delete(self, resource, **kwargs): self._request(resource, requests.delete, **kwargs)
Execute an HTTP DELETE request against the API endpoints. This method is meant for internal use. Does not return anything but raises an exception when failed. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library)
train
https://github.com/rycus86/ghost-client/blob/863d332801d2c1b8e7ad4573c7b16db78a7f8c8d/ghost_client/api.py#L415-L425
null
class Ghost(object): """ API client for the Ghost REST endpoints. See https://api.ghost.org/ for the available parameters. Sample usage: from ghost_client import Ghost # to read the client ID and secret from the database ghost = Ghost.from_sqlite( '/var/lib/ghost/content/data/ghost.db', 'http://localhost:2368' ) # or to use a specific client ID and secret ghost = Ghost( 'http://localhost:2368', client_id='ghost-admin', client_secret='secret_key' ) # log in ghost.login('username', 'password') # print the server's version print(ghost.version) # create a new tag tag = ghost.tags.create(name='API sample') # create a new post using it post = ghost.posts.create( title='Example post', slug='custom-slug', markdown='', # yes, even on v1.+ custom_excerpt='An example post created from Python', tags=[tag] ) # list posts, tags and users posts = ghost.posts.list( status='all', fields=('id', 'title', 'slug'), formats=('html', 'mobiledoc', 'plaintext'), ) tags = ghost.tags.list(fields='name', limit='all') users = ghost.users.list(include='count.posts') # use pagination while posts: for post in posts: print(post) posts = posts.next_page() print(posts.total) print(posts.pages) # update a post & tag updated_post = ghost.posts.update(post.id, title='Updated title') updated_tag = ghost.tags.update(tag.id, name='Updated tag') # note: creating, updating and deleting a user is not allowed by the API # access fields as properties print(post.title) print(post.markdown) # needs formats='mobiledoc' print(post.author.name) # needs include='author' # delete a post & tag ghost.posts.delete(post.id) ghost.tags.delete(tag.id) # upload an image ghost.upload(file_obj=open('sample.png', 'rb')) ghost.upload(file_path='/path/to/image.jpeg', 'rb') ghost.upload(name='image.gif', data=open('local.gif', 'rb').read()) # log out ghost.logout() The logged in credentials will be saved in memory and on HTTP 401 errors the client will attempt to re-authenticate once automatically. Responses are wrapped in `models.ModelList` and `models.Model` types to allow pagination and retrieving fields as properties. """ DEFAULT_VERSION = '1' """ The default version to report when cannot be fetched. """ def __init__( self, base_url, version='auto', client_id=None, client_secret=None, access_token=None, refresh_token=None ): """ Creates a new Ghost API client. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: Self-supplied client ID (optional) :param client_secret: Self-supplied client secret (optional) :param access_token: Self-supplied access token (optional) :param refresh_token: Self-supplied refresh token (optional) """ self.base_url = '%s/ghost/api/v0.1' % base_url self._version = version self._client_id = client_id self._client_secret = client_secret self._access_token = access_token self._refresh_token = refresh_token self._username = None self._password = None if not self._client_id or not self._client_secret: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_id or client_secret given or found' }]) self.posts = PostController(self) self.tags = Controller(self, 'tags') self.users = Controller(self, 'users') @classmethod def from_sqlite(cls, database_path, base_url, version='auto', client_id='ghost-admin'): """ Initialize a new Ghost API client, reading the client ID and secret from the SQlite database. :param database_path: The path to the database file. :param base_url: The base url of the server :param version: The server version to use (default: `auto`) :param client_id: The client ID to look for in the database :return: A new Ghost API client instance """ import os import sqlite3 fd = os.open(database_path, os.O_RDONLY) connection = sqlite3.connect('/dev/fd/%d' % fd) os.close(fd) try: row = connection.execute( 'SELECT secret FROM clients WHERE slug = ?', (client_id,) ).fetchone() if row: return cls( base_url, version=version, client_id=client_id, client_secret=row[0] ) else: raise GhostException(401, [{ 'errorType': 'InternalError', 'message': 'No client_secret found for client_id: %s' % client_id }]) finally: connection.close() @property def version(self): """ :return: The version of the server when initialized as 'auto', otherwise the version passed in at initialization """ if self._version != 'auto': return self._version if self._version == 'auto': try: data = self.execute_get('configuration/about/') self._version = data['configuration'][0]['version'] except GhostException: return self.DEFAULT_VERSION return self._version def login(self, username, password): """ Authenticate with the server. :param username: The username of an existing user :param password: The password for the user :return: The authentication response from the REST endpoint """ data = self._authenticate( grant_type='password', username=username, password=password, client_id=self._client_id, client_secret=self._client_secret ) self._username = username self._password = password return data def refresh_session(self): """ Re-authenticate using the refresh token if available. Otherwise log in using the username and password if it was used to authenticate initially. :return: The authentication response or `None` if not available """ if not self._refresh_token: if self._username and self._password: return self.login(self._username, self._password) return return self._authenticate( grant_type='refresh_token', refresh_token=self._refresh_token, client_id=self._client_id, client_secret=self._client_secret ) def _authenticate(self, **kwargs): response = requests.post( '%s/authentication/token' % self.base_url, data=kwargs ) if response.status_code != 200: raise GhostException(response.status_code, response.json().get('errors', [])) data = response.json() self._access_token = data.get('access_token') self._refresh_token = data.get('refresh_token') return data def revoke_access_token(self): """ Revoke the access token currently in use. """ if not self._access_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='access_token', token=self._access_token )) self._access_token = None def revoke_refresh_token(self): """ Revoke the refresh token currently active. """ if not self._refresh_token: return self.execute_post('authentication/revoke', json=dict( token_type_hint='refresh_token', token=self._refresh_token )) self._refresh_token = None def logout(self): """ Log out, revoking the access tokens and forgetting the login details if they were given. """ self.revoke_refresh_token() self.revoke_access_token() self._username, self._password = None, None def upload(self, file_obj=None, file_path=None, name=None, data=None): """ Upload an image and return its path on the server. Either `file_obj` or `file_path` or `name` and `data` has to be specified. :param file_obj: A file object to upload :param file_path: A file path to upload from :param name: A file name for uploading :param data: The file content to upload :return: The path of the uploaded file on the server """ close = False if file_obj: file_name, content = os.path.basename(file_obj.name), file_obj elif file_path: file_name, content = os.path.basename(file_path), open(file_path, 'rb') close = True elif name and data: file_name, content = name, data else: raise GhostException( 400, 'Either `file_obj` or `file_path` or ' '`name` and `data` needs to be specified' ) try: content_type, _ = mimetypes.guess_type(file_name) file_arg = (file_name, content, content_type) response = self.execute_post('uploads/', files={'uploadimage': file_arg}) return response finally: if close: content.close() @refresh_session_if_necessary def execute_get(self, resource, **kwargs): """ Execute an HTTP GET request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional query parameters (and optionally headers) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if kwargs: separator = '&' if '?' in url else '?' for key, value in kwargs.items(): if hasattr(value, '__iter__') and type(value) not in six.string_types: url = '%s%s%s=%s' % (url, separator, key, ','.join(value)) else: url = '%s%s%s=%s' % (url, separator, key, value) separator = '&' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token else: separator = '&' if '?' in url else '?' url = '%s%sclient_id=%s&client_secret=%s' % ( url, separator, self._client_id, self._client_secret ) response = requests.get(url, headers=headers) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response.json() def execute_post(self, resource, **kwargs): """ Execute an HTTP POST request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.post, **kwargs).json() def execute_put(self, resource, **kwargs): """ Execute an HTTP PUT request against the API endpoints. This method is meant for internal use. :param resource: The last part of the URI :param kwargs: Additional parameters for the HTTP call (`request` library) :return: The HTTP response as JSON or `GhostException` if unsuccessful """ return self._request(resource, requests.put, **kwargs).json() @refresh_session_if_necessary def _request(self, resource, request, **kwargs): if not self._access_token: raise GhostException(401, [{ 'errorType': 'ClientError', 'message': 'Access token not found' }]) url = '%s/%s' % (self.base_url, resource) headers = kwargs.pop('headers', dict()) if 'json' in kwargs: headers['Accept'] = 'application/json' headers['Content-Type'] = 'application/json' if self._access_token: headers['Authorization'] = 'Bearer %s' % self._access_token response = request(url, headers=headers, **kwargs) if response.status_code // 100 != 2: raise GhostException(response.status_code, response.json().get('errors', [])) return response
vmonaco/pohmm
pohmm/classification.py
PohmmClassifier.fit
python
def fit(self, labels, samples, pstates): assert len(labels) == len(samples) == len(pstates) for label in set(labels): label_samples = [s for l,s in zip(labels, samples) if l == label] label_pstates = [p for l,p in zip(labels, pstates) if l == label] pohmm = self.pohmm_factory() pohmm.fit(label_samples, label_pstates) self.pohmms[label] = pohmm return self
Fit the classifier with labels y and observations X
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/classification.py#L16-L30
null
class PohmmClassifier(object): """ Train a POHMM for each label and make predictions by choosing the maximum likelihood model """ def __init__(self, pohmm_factory): self.pohmms = {} self.pohmm_factory = pohmm_factory return def fit_df(self, labels, dfs, pstate_col=PSTATE_COL): """ Fit the classifier with labels y and DataFrames dfs """ assert len(labels) == len(dfs) for label in set(labels): label_dfs = [s for l,s in zip(labels, dfs) if l == label] pohmm = self.pohmm_factory() pohmm.fit_df(label_dfs, pstate_col=pstate_col) self.pohmms[label] = pohmm return self def predict(self, sample, pstates): """ Predict the class label of X """ scores = {} for label, pohmm in self.pohmms.items(): scores[label] = pohmm.score(sample, pstates) max_score_label = max(scores.items(), key=itemgetter(1))[0] return max_score_label, scores def predict_df(self, df, pstate_col=PSTATE_COL): """ Predict the class label of DataFrame df """ scores = {} for label, pohmm in self.pohmms.items(): scores[label] = pohmm.score_df(df, pstate_col=pstate_col) max_score_label = max(scores.items(), key=itemgetter(1))[0] return max_score_label, scores
vmonaco/pohmm
pohmm/classification.py
PohmmClassifier.fit_df
python
def fit_df(self, labels, dfs, pstate_col=PSTATE_COL): assert len(labels) == len(dfs) for label in set(labels): label_dfs = [s for l,s in zip(labels, dfs) if l == label] pohmm = self.pohmm_factory() pohmm.fit_df(label_dfs, pstate_col=pstate_col) self.pohmms[label] = pohmm return self
Fit the classifier with labels y and DataFrames dfs
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/classification.py#L32-L45
[ "def keystroke_model():\n \"\"\"Generates a 2-state model with lognormal emissions and frequency smoothing\"\"\"\n model = Pohmm(n_hidden_states=2,\n init_spread=2,\n emissions=['lognormal', 'lognormal'],\n smoothing='freq',\n init_method='obs',\n thresh=1)\n return model\n" ]
class PohmmClassifier(object): """ Train a POHMM for each label and make predictions by choosing the maximum likelihood model """ def __init__(self, pohmm_factory): self.pohmms = {} self.pohmm_factory = pohmm_factory return def fit(self, labels, samples, pstates): """ Fit the classifier with labels y and observations X """ assert len(labels) == len(samples) == len(pstates) for label in set(labels): label_samples = [s for l,s in zip(labels, samples) if l == label] label_pstates = [p for l,p in zip(labels, pstates) if l == label] pohmm = self.pohmm_factory() pohmm.fit(label_samples, label_pstates) self.pohmms[label] = pohmm return self def predict(self, sample, pstates): """ Predict the class label of X """ scores = {} for label, pohmm in self.pohmms.items(): scores[label] = pohmm.score(sample, pstates) max_score_label = max(scores.items(), key=itemgetter(1))[0] return max_score_label, scores def predict_df(self, df, pstate_col=PSTATE_COL): """ Predict the class label of DataFrame df """ scores = {} for label, pohmm in self.pohmms.items(): scores[label] = pohmm.score_df(df, pstate_col=pstate_col) max_score_label = max(scores.items(), key=itemgetter(1))[0] return max_score_label, scores
vmonaco/pohmm
pohmm/classification.py
PohmmClassifier.predict
python
def predict(self, sample, pstates): scores = {} for label, pohmm in self.pohmms.items(): scores[label] = pohmm.score(sample, pstates) max_score_label = max(scores.items(), key=itemgetter(1))[0] return max_score_label, scores
Predict the class label of X
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/classification.py#L47-L55
null
class PohmmClassifier(object): """ Train a POHMM for each label and make predictions by choosing the maximum likelihood model """ def __init__(self, pohmm_factory): self.pohmms = {} self.pohmm_factory = pohmm_factory return def fit(self, labels, samples, pstates): """ Fit the classifier with labels y and observations X """ assert len(labels) == len(samples) == len(pstates) for label in set(labels): label_samples = [s for l,s in zip(labels, samples) if l == label] label_pstates = [p for l,p in zip(labels, pstates) if l == label] pohmm = self.pohmm_factory() pohmm.fit(label_samples, label_pstates) self.pohmms[label] = pohmm return self def fit_df(self, labels, dfs, pstate_col=PSTATE_COL): """ Fit the classifier with labels y and DataFrames dfs """ assert len(labels) == len(dfs) for label in set(labels): label_dfs = [s for l,s in zip(labels, dfs) if l == label] pohmm = self.pohmm_factory() pohmm.fit_df(label_dfs, pstate_col=pstate_col) self.pohmms[label] = pohmm return self def predict_df(self, df, pstate_col=PSTATE_COL): """ Predict the class label of DataFrame df """ scores = {} for label, pohmm in self.pohmms.items(): scores[label] = pohmm.score_df(df, pstate_col=pstate_col) max_score_label = max(scores.items(), key=itemgetter(1))[0] return max_score_label, scores
vmonaco/pohmm
pohmm/classification.py
PohmmClassifier.predict_df
python
def predict_df(self, df, pstate_col=PSTATE_COL): scores = {} for label, pohmm in self.pohmms.items(): scores[label] = pohmm.score_df(df, pstate_col=pstate_col) max_score_label = max(scores.items(), key=itemgetter(1))[0] return max_score_label, scores
Predict the class label of DataFrame df
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/classification.py#L57-L65
null
class PohmmClassifier(object): """ Train a POHMM for each label and make predictions by choosing the maximum likelihood model """ def __init__(self, pohmm_factory): self.pohmms = {} self.pohmm_factory = pohmm_factory return def fit(self, labels, samples, pstates): """ Fit the classifier with labels y and observations X """ assert len(labels) == len(samples) == len(pstates) for label in set(labels): label_samples = [s for l,s in zip(labels, samples) if l == label] label_pstates = [p for l,p in zip(labels, pstates) if l == label] pohmm = self.pohmm_factory() pohmm.fit(label_samples, label_pstates) self.pohmms[label] = pohmm return self def fit_df(self, labels, dfs, pstate_col=PSTATE_COL): """ Fit the classifier with labels y and DataFrames dfs """ assert len(labels) == len(dfs) for label in set(labels): label_dfs = [s for l,s in zip(labels, dfs) if l == label] pohmm = self.pohmm_factory() pohmm.fit_df(label_dfs, pstate_col=pstate_col) self.pohmms[label] = pohmm return self def predict(self, sample, pstates): """ Predict the class label of X """ scores = {} for label, pohmm in self.pohmms.items(): scores[label] = pohmm.score(sample, pstates) max_score_label = max(scores.items(), key=itemgetter(1))[0] return max_score_label, scores
vmonaco/pohmm
pohmm/pohmm.py
Pohmm.rand
python
def rand(self, unique_pstates, random_state=None): self._init_pstates(unique_pstates) self._init_random(random_state=random_state) self._compute_marginals() return self
Randomize the POHMM parameters
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/pohmm.py#L662-L669
[ "def _init_pstates(self, unique_pstates):\n # Map events to a unique index. The unknown p-state is at idx 0\n self.e = defaultdict(int)\n self.e.update(dict(zip(np.sort(unique_pstates), range(1, len(unique_pstates) + 1))))\n self.er = {v: k for k, v in self.e.items()} # Reverse lookup\n self.er[0] = UNKNOWN_PSTATE\n self.n_partial_states = len(self.e.keys()) + 1 # Add one for the unknown state\n return\n", "def _init_random(self, random_state=None):\n\n if random_state is None:\n random_state = self.random_state\n random_state = check_random_state(random_state)\n\n self.pstate_startprob = np.zeros(self.n_partial_states)\n self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states))\n self.pstate_steadyprob = np.zeros(self.n_partial_states)\n\n self.pstate_startprob[1:] = gen_stochastic_matrix(size=self.n_partial_states - 1, random_state=random_state)\n self.pstate_transmat[1:, 1:] = gen_stochastic_matrix(\n size=(self.n_partial_states - 1, self.n_partial_states - 1), random_state=random_state)\n self.pstate_steadyprob[1:] = steadystate(self.pstate_transmat[1:, 1:])\n\n self.startprob = gen_stochastic_matrix(size=(self.n_partial_states, self.n_hidden_states),\n random_state=random_state)\n\n transmat = np.zeros((self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states))\n for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)):\n transmat[i, j] = gen_stochastic_matrix(size=(self.n_hidden_states, self.n_hidden_states),\n random_state=random_state)\n self.transmat = normalize(transmat, axis=3)\n\n # Initialize emission parameters\n for feature_name, feature_distr in zip(self.emission_name, self.emission_distr):\n if feature_distr == 'normal':\n self.emission[feature_name]['mu'] = random_state.uniform(*_RANDINIT[feature_distr]['mu'], size=(\n self.n_partial_states, self.n_hidden_states))\n self.emission[feature_name]['sigma'] = random_state.uniform(*_RANDINIT[feature_distr]['sigma'], size=(\n self.n_partial_states, self.n_hidden_states))\n if feature_distr == 'lognormal':\n self.emission[feature_name]['logmu'] = random_state.uniform(*_RANDINIT[feature_distr]['logmu'], size=(\n self.n_partial_states, self.n_hidden_states))\n self.emission[feature_name]['logsigma'] = random_state.uniform(*_RANDINIT[feature_distr]['logsigma'],\n size=(self.n_partial_states,\n self.n_hidden_states))\n\n if self.emission_distr[0] == 'normal':\n self.emission[self.emission_name[0]]['mu'] = np.sort(self.emission[self.emission_name[0]]['mu'], axis=1)\n elif self.emission_distr[0] == 'lognormal':\n self.emission[self.emission_name[0]]['logmu'] = np.sort(self.emission[self.emission_name[0]]['logmu'],\n axis=1)\n\n return\n", "def _compute_marginals(self):\n # TODO: cythonize some of this\n\n # Start prob, weighted by p-state start probs\n self.startprob[0] = (self.pstate_startprob[1:, np.newaxis] * self.startprob[1:]).sum(axis=0)\n\n # Use the p-state transmat and transmat to get the full transmat\n full_transmat = ph2full(self.pstate_transmat[1:, 1:], self.transmat[1:, 1:])\n full_steadyprob = steadystate(full_transmat)\n\n # Steady state probas are determined by the full trans mat, need to be updated\n steadyprob = np.zeros(shape=(self.n_partial_states, self.n_hidden_states))\n steadyprob[0] = full_steadyprob.reshape(-1, self.n_hidden_states).sum(axis=0)\n for i in range(self.n_partial_states - 1):\n steadyprob[i + 1] = normalize(\n full_steadyprob[i * self.n_hidden_states:i * self.n_hidden_states + self.n_hidden_states])\n\n self.steadyprob = steadyprob\n\n # Update the transations to/from the marginal state\n transmat = self.transmat\n # Group the hidden states within each partial state\n for hidx1, hidx2 in product(range(self.n_hidden_states), range(self.n_hidden_states)):\n transmat[0, 0][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, hidx2::self.n_hidden_states].sum()\n for pidx in range(self.n_partial_states - 1):\n transmat[pidx + 1, 0][hidx1, hidx2] = full_transmat[pidx * self.n_hidden_states + hidx1,\n hidx2::self.n_hidden_states].sum()\n transmat[0, pidx + 1][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states,\n pidx * self.n_hidden_states + hidx2].sum()\n self.transmat = normalize(transmat, axis=3)\n\n pweights = self.pstate_steadyprob[1:, np.newaxis]\n # Update emission parameters\n for feature_name, feature_distr in zip(self.emission_name, self.emission_distr):\n if feature_distr == 'normal':\n # Marginal state is a mixture of normals\n mu = self.emission[feature_name]['mu'][1:, :]\n sigma = self.emission[feature_name]['sigma'][1:, :]\n\n # Weighted mean and var\n mu_0 = (pweights * mu).sum(axis=0)\n self.emission[feature_name]['mu'][0, :] = mu_0\n self.emission[feature_name]['sigma'][0, :] = np.sqrt(\n (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0))\n\n if feature_distr == 'lognormal':\n # Marginal state is a mixture of normals\n mu = self.emission[feature_name]['logmu'][1:, :]\n sigma = self.emission[feature_name]['logsigma'][1:, :]\n\n # Weighted mean and var\n mu_0 = (pweights * mu).sum(axis=0)\n self.emission[feature_name]['logmu'][0, :] = mu_0\n self.emission[feature_name]['logsigma'][0, :] = np.sqrt(\n (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0))\n\n return\n" ]
class Pohmm(object): """ Partially observable hidden Markov model """ def __init__(self, n_hidden_states=2, emissions=['normal'], max_iter=1000, thresh=1e-6, init_method='obs', init_spread=2, smoothing=None, random_state=None): if type(n_hidden_states) is int: self.n_hidden_states = n_hidden_states else: raise Exception('Wrong type for n_hidden_states. Must be int') if type(emissions[0]) is tuple: emission_name, emission_distr = zip(*emissions) elif type(emissions[0]) is str: emission_name, emission_distr = np.arange(len(emissions)), emissions for distr in emission_distr: if distr not in _DISTRIBUTIONS.keys(): raise ValueError('Emission distribution must be one of', _DISTRIBUTIONS.keys()) self.emission_name = emission_name self.emission_distr = emission_distr self.emission_name_distr = dict(zip(emission_name, emission_distr)) self.n_features = len(emissions) # Set up the emission parameters # emission: {'feature':{'param': np.array(shape=(n_partial_states, n_hidden_states))}} self.emission = defaultdict(dict) for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: self.emission[name][param] = None self.emission = dict(self.emission) assert max_iter >= 0 assert thresh >= 0 self.max_iter = max_iter self.thresh = thresh assert init_spread >= 0 self.init_spread = init_spread if init_method not in _INIT_METHODS: raise ValueError('init_method must be one of', _INIT_METHODS) self.init_method = init_method if smoothing is None: smoothing = {'transmat': None, 'startprob': None} for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: smoothing['%s:%s' % (name, param)] = None elif type(smoothing) is str: s = smoothing smoothing = {'transmat': s, 'startprob': s} for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: smoothing['%s:%s' % (name, param)] = s elif type(smoothing) is dict: assert 'transmat' in smoothing.keys() assert 'startprob' in smoothing.keys() for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: assert param in smoothing.keys() or '%s:%s' % (name, param) in smoothing.keys() if param in smoothing.keys() and '%s:%s' % (name, param) not in smoothing.keys(): smoothing['%s:%s' % (name, param)] = smoothing[param] else: raise Exception('Wrong type for smoothing. Must be None, str, or dict') self.smoothing = smoothing self.random_state = random_state # Number of unique partial states is unknown until fit self.n_partial_states = None # Results after fitting the model self.logprob_ = None self.n_iter_performed_ = None self.logprob_delta_ = None # Mapping between p-states and a unique index # Defaults to 0 for unknown or missing p-states self.e = defaultdict(int) def _get_startprob(self): return np.exp(self._log_startprob) def _set_startprob(self, startprob): if startprob is None: startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states else: startprob = np.asarray(startprob, dtype=np.float) # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(startprob): startprob = normalize(startprob, axis=1) if len(startprob) != self.n_partial_states: raise ValueError('startprob must have length n_partial_states') if not np.allclose(np.sum(startprob, axis=1), 1.0): raise ValueError('startprob must sum to 1.0') self._log_startprob = np.log(np.asarray(startprob).copy()) startprob = property(_get_startprob, _set_startprob) def _get_steadyprob(self): return np.exp(self._log_steadyprob) def _set_steadyprob(self, steadyprob): if steadyprob is None: steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states else: steadyprob = np.asarray(steadyprob, dtype=np.float) # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(steadyprob): steadyprob = normalize(steadyprob, axis=1) if len(steadyprob) != self.n_partial_states: raise ValueError('steadyprob must have length n_partial_states') if not np.allclose(np.sum(steadyprob, axis=1), 1.0): raise ValueError('steadyprob must sum to 1.0') self._log_steadyprob = np.log(np.asarray(steadyprob).copy()) steadyprob = property(_get_steadyprob, _set_steadyprob) def _get_transmat(self): return np.exp(self._log_transmat) def _set_transmat(self, transmat): if transmat is None: transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) / self.n_hidden_states # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(transmat): transmat = normalize(transmat, axis=3) if (np.asarray(transmat).shape != (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)): raise ValueError('transmat must have shape ' '(n_partial_states,n_partial_states,n_hidden_states,n_hidden_states)') if not np.all(np.allclose(np.sum(transmat, axis=3), 1.0)): raise ValueError('Rows of transmat must sum to 1.0') self._log_transmat = np.log(np.asarray(transmat).copy()) underflow_idx = np.isnan(self._log_transmat) self._log_transmat[underflow_idx] = NEGINF transmat = property(_get_transmat, _set_transmat) def _compute_log_likelihood(self, obs, pstates_idx): q = np.zeros(shape=(len(obs), self.n_hidden_states, self.n_features)) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': mu = self.emission[feature_name]['mu'][pstates_idx] sigma = self.emission[feature_name]['sigma'][pstates_idx] for j in range(self.n_hidden_states): q[:, j, col] = np.log( np.maximum(MIN_PROBA, stats.norm.pdf(obs[:, col], loc=mu[:, j], scale=sigma[:, j]))) if feature_distr == 'lognormal': logmu = self.emission[feature_name]['logmu'][pstates_idx] logsigma = self.emission[feature_name]['logsigma'][pstates_idx] for j in range(self.n_hidden_states): q[:, j, col] = np.log(np.maximum(MIN_PROBA, stats.lognorm.pdf(obs[:, col], logsigma[:, j], loc=0, scale=np.exp(logmu[:, j])))) q = q.sum(axis=2) return q def _generate_sample_from_state(self, hidden_state, pstates_idx, random_state=None): sample = np.zeros(self.n_features) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': mu = self.emission[feature_name]['mu'][pstates_idx][hidden_state] sigma = self.emission[feature_name]['sigma'][pstates_idx][hidden_state] sample[col] = stats.norm.rvs(loc=mu, scale=sigma, random_state=random_state) if feature_distr == 'lognormal': logmu = self.emission[feature_name]['logmu'][pstates_idx][hidden_state] logsigma = self.emission[feature_name]['logsigma'][pstates_idx][hidden_state] sample[col] = stats.lognorm.rvs(logsigma, loc=0, scale=np.exp(logmu), random_state=random_state) return sample def _init_pstates(self, unique_pstates): # Map events to a unique index. The unknown p-state is at idx 0 self.e = defaultdict(int) self.e.update(dict(zip(np.sort(unique_pstates), range(1, len(unique_pstates) + 1)))) self.er = {v: k for k, v in self.e.items()} # Reverse lookup self.er[0] = UNKNOWN_PSTATE self.n_partial_states = len(self.e.keys()) + 1 # Add one for the unknown state return def _init_pstate_freqs(self, pstates_idx): # Partial state frequencies self.pstate_freq = Counter([idx for seq in pstates_idx for idx in seq]) self.pstate_trans_freq = Counter([(idx1, idx2) for seq in pstates_idx for idx1, idx2 in zip(seq[:-1], seq[1:])]) # Store freqs for the meta state self.pstate_freq[0] = len(np.concatenate(pstates_idx)) for seq in pstates_idx: self.pstate_startprob[seq[0]] += 1 self.pstate_steadyprob += np.bincount(seq, minlength=self.n_partial_states) for idx1, idx2 in zip(seq[:-1], seq[1:]): self.pstate_trans_freq[(0, 0)] += 1 self.pstate_trans_freq[(idx1, 0)] += 1 self.pstate_trans_freq[(0, idx2)] += 1 self.pstate_transmat[idx1, idx2] += 1 # TODO: separate probas from freqs # Normalize to get the probabilities, ignore the meta state at idx 0 self.pstate_startprob[1:] = normalize(self.pstate_startprob[1:]) self.pstate_transmat[1:, 1:] = normalize(self.pstate_transmat[1:, 1:], axis=1) self.pstate_steadyprob[1:] = normalize(self.pstate_steadyprob[1:]) return def _init_from_obs(self, obs, pstates_idx): # Partial state probabilities self.pstate_startprob = np.zeros(self.n_partial_states) self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states)) self.pstate_steadyprob = np.zeros(self.n_partial_states) # obs should be (N*T, n_features) # N is the number of samples # T is the size of each sample obs = np.concatenate(obs) pstates_idx = np.concatenate(pstates_idx) # Initialize starting and transition probas self.startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states self.steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states self.transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) / self.n_hidden_states # Initialize emission parameters # Hidden states are ordered by the first feature feature1 = self.emission_name[0] for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': self.emission[feature_name]['mu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['sigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) for idx in range(1, self.n_partial_states): idx_pstate = (pstates_idx == idx) if not np.any(idx_pstate): idx_pstate = np.arange(len(pstates_idx)) if feature_name == feature1: self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA) self.emission[feature_name]['mu'][idx] = obs[idx_pstate, col].mean() + \ obs[:, col].std() * np.linspace( -self.init_spread, self.init_spread, self.n_hidden_states) else: self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA) self.emission[feature_name]['mu'][idx, :] = obs[idx_pstate, col].mean() if feature_distr == 'lognormal': self.emission[feature_name]['logmu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['logsigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) for idx in range(1, self.n_partial_states): idx_pstate = (pstates_idx == idx) if not np.any(idx_pstate): idx_pstate = np.arange(len(pstates_idx)) if feature_name == feature1: self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(), MIN_PROBA) self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() + np.log( obs[:, col]).std() * np.linspace(-self.init_spread, self.init_spread, self.n_hidden_states) else: self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(), MIN_PROBA) self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() return def _init_random(self, random_state=None): if random_state is None: random_state = self.random_state random_state = check_random_state(random_state) self.pstate_startprob = np.zeros(self.n_partial_states) self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states)) self.pstate_steadyprob = np.zeros(self.n_partial_states) self.pstate_startprob[1:] = gen_stochastic_matrix(size=self.n_partial_states - 1, random_state=random_state) self.pstate_transmat[1:, 1:] = gen_stochastic_matrix( size=(self.n_partial_states - 1, self.n_partial_states - 1), random_state=random_state) self.pstate_steadyprob[1:] = steadystate(self.pstate_transmat[1:, 1:]) self.startprob = gen_stochastic_matrix(size=(self.n_partial_states, self.n_hidden_states), random_state=random_state) transmat = np.zeros((self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)): transmat[i, j] = gen_stochastic_matrix(size=(self.n_hidden_states, self.n_hidden_states), random_state=random_state) self.transmat = normalize(transmat, axis=3) # Initialize emission parameters for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): if feature_distr == 'normal': self.emission[feature_name]['mu'] = random_state.uniform(*_RANDINIT[feature_distr]['mu'], size=( self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['sigma'] = random_state.uniform(*_RANDINIT[feature_distr]['sigma'], size=( self.n_partial_states, self.n_hidden_states)) if feature_distr == 'lognormal': self.emission[feature_name]['logmu'] = random_state.uniform(*_RANDINIT[feature_distr]['logmu'], size=( self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['logsigma'] = random_state.uniform(*_RANDINIT[feature_distr]['logsigma'], size=(self.n_partial_states, self.n_hidden_states)) if self.emission_distr[0] == 'normal': self.emission[self.emission_name[0]]['mu'] = np.sort(self.emission[self.emission_name[0]]['mu'], axis=1) elif self.emission_distr[0] == 'lognormal': self.emission[self.emission_name[0]]['logmu'] = np.sort(self.emission[self.emission_name[0]]['logmu'], axis=1) return def _smooth(self): self._compute_marginals() startprob = self.startprob for j in range(1, self.n_partial_states): if 'freq' == self.smoothing['startprob']: w_ = 1 / (1 + self.pstate_freq[j]) w_j = 1 - w_ elif 'proba' == self.smoothing['startprob']: w_j = self.pstate_steadyprob[j] w_ = 1 - w_j elif 'exp' == self.smoothing['startprob']: w_ = np.exp(-self.pstate_freq[j]) w_j = 1 - w_ elif 'fixed' == self.smoothing['startprob']: w_ = 1 w_j = 0 elif self.smoothing['startprob'] is None: w_ = 0 w_j = 1 else: raise Exception('Wrong smoothing for startprob: ' + self.smoothing['startprob']) startprob[j] = w_j * self.startprob[j] + w_ * self.startprob[0] self.startprob = startprob transmat = self.transmat for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)): if 'freq' == self.smoothing['transmat']: w_i0 = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[0, j]) w_0j = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[i, 0]) w_ij = 1 - (w_i0 + w_0j) w_ = 0 elif 'proba' == self.smoothing['transmat']: denom = self.pstate_transmat[i, j] + self.pstate_transmat[i, :].sum() + self.pstate_transmat[:, j].sum() w_i0 = self.pstate_transmat[i, :].sum() / denom w_0j = self.pstate_transmat[:, j].sum() / denom w_ij = self.pstate_transmat[i, j] / denom w_ = 0 elif 'exp' == self.smoothing['transmat']: w_i0 = self.pstate_trans_freq[i, 0] * np.exp(-self.pstate_trans_freq[i, j]) w_0j = self.pstate_trans_freq[0, j] * np.exp(-self.pstate_trans_freq[i, j]) w_ = self.pstate_trans_freq[0, 0] * np.exp( -(self.pstate_trans_freq[i, 0] + self.pstate_trans_freq[0, j])) w_ij = self.pstate_trans_freq[i, j] w_ij, w_i0, w_0j, w_ = normalize(np.array([w_ij, w_i0, w_0j, w_])) elif 'fixed' == self.smoothing['transmat']: w_i0 = 0 w_0j = 0 w_ = 1 w_ij = 0 elif self.smoothing['transmat'] is None: w_i0 = 0 w_0j = 0 w_ = 0 w_ij = 1 else: raise Exception('Wrong smoothing for transmat: ' + self.smoothing['transmat']) assert (w_i0 + w_0j + w_ + w_ij) - 1 < TOLERANCE transmat[i, j] = w_ij * self.transmat[i, j] + w_i0 * self.transmat[i, 0] + w_0j * self.transmat[0, j] + w_ * \ self.transmat[ 0, 0] self.transmat = transmat assert np.all(self.startprob.sum(axis=1) - 1 < TOLERANCE) assert np.all(self.steadyprob.sum(axis=1) - 1 < TOLERANCE) assert np.all(self.transmat.sum(axis=3) - 1 < TOLERANCE) for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[feature_distr]: key = '%s:%s' % (feature_name, param) for j in range(1, self.n_partial_states): if 'freq' == self.smoothing[key]: w_ = 1 / (1 + self.pstate_freq[j]) w_j = 1 - w_ elif 'proba' == self.smoothing[key]: w_j = self.pstate_steadyprob[j] w_ = 1 - w_j elif 'exp' == self.smoothing[key]: w_ = np.exp(-self.pstate_freq[j]) w_j = 1 - w_ elif 'fixed' == self.smoothing[key]: w_ = 1 w_j = 0 elif self.smoothing[key] is None: w_ = 0 w_j = 1 else: raise Exception('Wrong smoothing for ' + key) self.emission[feature_name][param][j] = w_j * self.emission[feature_name][param][j] + w_ * \ self.emission[ feature_name][ param][0] return def _initialize_sufficient_statistics(self): stats = { 'nobs': 0, 'post': np.zeros((self.n_partial_states, self.n_hidden_states)), 'obs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'obs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'lnobs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'lnobs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'start': np.zeros((self.n_partial_states, self.n_hidden_states)), 'steady': np.zeros((self.n_partial_states, self.n_hidden_states)), 'trans': np.zeros( (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) } return stats def _accumulate_sufficient_statistics(self, stats, obs, pstates_idx, framelogprob, posteriors, fwdlattice, bwdlattice): stats['nobs'] += 1 n_observations, n_hidden_states = framelogprob.shape stats['start'][0] += posteriors[0] for i in range(self.n_partial_states): if len(np.where(pstates_idx == i)[0]) > 0: stats['start'][i] += posteriors[np.where(pstates_idx == i)[0].min()] if n_observations > 1: lneta = np.zeros((n_observations - 1, n_hidden_states, n_hidden_states)) _hmmc._compute_lneta(n_observations, n_hidden_states, pstates_idx, fwdlattice, self._log_transmat, bwdlattice, framelogprob, lneta) for i, j in unique_rows(np.c_[pstates_idx[:-1], pstates_idx[1:]]): if ((pstates_idx[:-1] == i) & (pstates_idx[1:] == j)).sum() > 0: stats['trans'][i, j] += np.exp( logsumexp(lneta[(pstates_idx[:-1] == i) & (pstates_idx[1:] == j)], axis=0)) for i in range(self.n_partial_states): stats['post'][i] += posteriors[pstates_idx == i].sum(axis=0) stats['obs'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i]) stats['obs**2'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i] ** 2) stats['lnobs'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i])) stats['lnobs**2'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i]) ** 2) return def _compute_marginals(self): # TODO: cythonize some of this # Start prob, weighted by p-state start probs self.startprob[0] = (self.pstate_startprob[1:, np.newaxis] * self.startprob[1:]).sum(axis=0) # Use the p-state transmat and transmat to get the full transmat full_transmat = ph2full(self.pstate_transmat[1:, 1:], self.transmat[1:, 1:]) full_steadyprob = steadystate(full_transmat) # Steady state probas are determined by the full trans mat, need to be updated steadyprob = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) steadyprob[0] = full_steadyprob.reshape(-1, self.n_hidden_states).sum(axis=0) for i in range(self.n_partial_states - 1): steadyprob[i + 1] = normalize( full_steadyprob[i * self.n_hidden_states:i * self.n_hidden_states + self.n_hidden_states]) self.steadyprob = steadyprob # Update the transations to/from the marginal state transmat = self.transmat # Group the hidden states within each partial state for hidx1, hidx2 in product(range(self.n_hidden_states), range(self.n_hidden_states)): transmat[0, 0][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, hidx2::self.n_hidden_states].sum() for pidx in range(self.n_partial_states - 1): transmat[pidx + 1, 0][hidx1, hidx2] = full_transmat[pidx * self.n_hidden_states + hidx1, hidx2::self.n_hidden_states].sum() transmat[0, pidx + 1][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, pidx * self.n_hidden_states + hidx2].sum() self.transmat = normalize(transmat, axis=3) pweights = self.pstate_steadyprob[1:, np.newaxis] # Update emission parameters for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): if feature_distr == 'normal': # Marginal state is a mixture of normals mu = self.emission[feature_name]['mu'][1:, :] sigma = self.emission[feature_name]['sigma'][1:, :] # Weighted mean and var mu_0 = (pweights * mu).sum(axis=0) self.emission[feature_name]['mu'][0, :] = mu_0 self.emission[feature_name]['sigma'][0, :] = np.sqrt( (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0)) if feature_distr == 'lognormal': # Marginal state is a mixture of normals mu = self.emission[feature_name]['logmu'][1:, :] sigma = self.emission[feature_name]['logsigma'][1:, :] # Weighted mean and var mu_0 = (pweights * mu).sum(axis=0) self.emission[feature_name]['logmu'][0, :] = mu_0 self.emission[feature_name]['logsigma'][0, :] = np.sqrt( (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0)) return def _do_mstep(self, stats): self.startprob = normalize(np.maximum(stats['start'], MIN_PROBA), axis=1) self.transmat = normalize(np.maximum(stats['trans'], MIN_PROBA), axis=3) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': denom = np.maximum(stats['post'], MIN_PROBA) mu = stats['obs'][:, :, col] / denom cv_num = (stats['obs**2'][:, :, col] - 2 * mu * stats['obs'][:, :, col] + mu ** 2 * denom) sigma = np.sqrt(cv_num / denom) sigma[np.isnan(sigma)] = MIN_PROBA self.emission[feature_name]['mu'] = mu self.emission[feature_name]['sigma'] = np.maximum(sigma, MIN_PROBA) if feature_distr == 'lognormal': denom = np.maximum(stats['post'], MIN_PROBA) mu = stats['lnobs'][:, :, col] / denom cv_num = (stats['lnobs**2'][:, :, col] - 2 * mu * stats['lnobs'][:, :, col] + mu ** 2 * denom) sigma = np.sqrt(cv_num / denom) sigma[np.isnan(sigma)] = MIN_PROBA self.emission[feature_name]['logmu'] = mu self.emission[feature_name]['logsigma'] = np.maximum(sigma, MIN_PROBA) return def _do_forward_pass(self, framelogprob, event_idx): n_observations, n_hidden_states = framelogprob.shape fwdlattice = np.zeros((n_observations, n_hidden_states)) _hmmc._forward(n_observations, n_hidden_states, event_idx, self._log_startprob, self._log_transmat, framelogprob, fwdlattice) return logsumexp(fwdlattice[-1]), fwdlattice def _do_backward_pass(self, framelogprob, event_idx): n_observations, n_hidden_states = framelogprob.shape bwdlattice = np.zeros((n_observations, n_hidden_states)) _hmmc._backward(n_observations, n_hidden_states, event_idx, self._log_startprob, self._log_transmat, framelogprob, bwdlattice) return bwdlattice def _do_viterbi_pass(self, framelogprob, event_idx): n_observations, n_components = framelogprob.shape state_sequence, logprob = _hmmc._viterbi( n_observations, n_components, event_idx, self._log_startprob, self._log_transmat, framelogprob) return logprob, state_sequence def fit(self, obs, pstates, unique_pstates=None): """ Estimate model parameters. """ obs = [np.array(o) for o in obs] pstates = [np.array(p) for p in pstates] # List or array of observation sequences assert len(obs) == len(pstates) assert obs[0].ndim == 2 assert pstates[0].ndim == 1 if unique_pstates is not None: self._init_pstates(unique_pstates) else: self._init_pstates(list(set(np.concatenate(pstates)))) # Map the partial states to a unique index pstates_idx = [np.array([self.e[p] for p in seq]) for seq in pstates] if self.init_method == 'rand': self._init_random() elif self.init_method == 'obs': self._init_from_obs(obs, pstates_idx) self._init_pstate_freqs(pstates_idx) self._smooth() logprob = [] for i in range(self.max_iter): # Expectation step stats = self._initialize_sufficient_statistics() curr_logprob = 0 for obs_i, pstates_idx_i in zip(obs, pstates_idx): framelogprob = self._compute_log_likelihood(obs_i, pstates_idx_i) lpr, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx_i) bwdlattice = self._do_backward_pass(framelogprob, pstates_idx_i) gamma = fwdlattice + bwdlattice posteriors = np.exp(gamma.T - logsumexp(gamma, axis=1)).T curr_logprob += lpr self._accumulate_sufficient_statistics(stats, obs_i, pstates_idx_i, framelogprob, posteriors, fwdlattice, bwdlattice) logprob.append(curr_logprob) self.logprob_ = curr_logprob # Check for convergence. self.n_iter_performed_ = i if i > 0: self.logprob_delta = logprob[-1] - logprob[-2] if self.logprob_delta < self.thresh: break # Maximization step self._do_mstep(stats) # Mix the parameters self._smooth() return self def score(self, obs, pstates): """ Compute the log probability under the model. """ pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) logprob, _ = self._do_forward_pass(framelogprob, pstates_idx) return logprob def score_events(self, obs, pstates): """ Compute the log probability of each event under the model. """ pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) _, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx) L = logsumexp(fwdlattice, axis=1) return np.concatenate([L[[0]], np.diff(L)]) def predict_states(self, obs, pstates): pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) viterbi_logprob, state_sequence = self._do_viterbi_pass(framelogprob, pstates_idx) return viterbi_logprob, state_sequence def predict(self, obs, pstates, next_pstate=None): """ Predict the next observation """ assert len(obs) == len(pstates) pstates_idx = np.array([self.e[ei] for ei in pstates]) next_pstate_idx = self.e[next_pstate] if len(obs) == 0: # No history, use the starting probas next_hstate_prob = self.startprob[next_pstate_idx] else: # With a history, determine the hidden state posteriors using # the last posteriors and transition matrix framelogprob = self._compute_log_likelihood(obs, pstates_idx) _, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx) next_hstate_prob = np.zeros(self.n_hidden_states) alpha_n = fwdlattice[-1] vmax = alpha_n.max(axis=0) alpha_n = np.exp(alpha_n - vmax) alpha_n = alpha_n / alpha_n.sum() trans = self.transmat[pstates_idx[-1], next_pstate_idx] for i in range(self.n_hidden_states): next_hstate_prob[i] = np.sum([alpha_n[j] * trans[j, i] for j in range(self.n_hidden_states)]) assert next_hstate_prob.sum() - 1 < TOLERANCE # Make the prediction prediction = np.array( [self.expected_value(feature, pstate=next_pstate, hstate_prob=next_hstate_prob) for feature in self.emission_name]) # next_hstate = np.argmax(next_hstate_prob) # prediction = np.array( # [self.expected_value(feature, pstate=next_pstate, hstate=next_hstate) for feature in # self.emission_name]) return prediction def gen_pstates_idx(self, n, random_state=None): if random_state is None: random_state = self.random_state random_state = check_random_state(random_state) startprob_cdf = np.cumsum(self.pstate_startprob) transmat_cdf = np.cumsum(self.pstate_transmat, 1) # Initial state. rand = random_state.rand() curr_pstate = (startprob_cdf > rand).argmax() pstates = [curr_pstate] for _ in range(1, n): rand = random_state.rand() curr_pstate = (transmat_cdf[curr_pstate] > rand).argmax() pstates.append(curr_pstate) return np.array(pstates, dtype=int) def sample(self, pstates=None, n_obs=None, random_state=None): """ """ random_state = check_random_state(random_state) if pstates is None and n_obs is None: raise Exception('Must provide either pstates or n_obs') if pstates is not None and n_obs is not None: raise Exception('Must provide either pstates or n_obs but not both') gen_pstates = False rand = random_state.rand() if pstates is None: gen_pstates = True pstartprob_cdf = np.cumsum(self.pstate_startprob) ptransmat_cdf = np.cumsum(self.pstate_transmat, 1) # Initial pstate currpstate = (pstartprob_cdf > rand).argmax() pstates_idx = [currpstate] pstates = [self.er[currpstate]] else: n_obs = len(pstates) pstates_idx = np.array([self.e[p] for p in pstates]) startprob_pdf = self.startprob[pstates_idx[0]] startprob_cdf = np.cumsum(startprob_pdf) transmat_cdf = np.cumsum(self.transmat[0, pstates_idx[0]], 1) # Initial hidden state rand = random_state.rand() currstate = (startprob_cdf > rand).argmax() hidden_states = [currstate] obs = [self._generate_sample_from_state(currstate, pstates_idx[0], random_state)] for i in range(1, n_obs): rand = random_state.rand() if gen_pstates: currpstate = (ptransmat_cdf[currpstate] > rand).argmax() pstates_idx.append(currpstate) pstates.append(self.er[currpstate]) transmat_cdf = np.cumsum(self.transmat[pstates_idx[i - 1], pstates_idx[i]], 1) rand = random_state.rand() currstate = (transmat_cdf[currstate] > rand).argmax() hidden_states.append(currstate) obs.append(self._generate_sample_from_state(currstate, pstates_idx[i], random_state)) return np.array(obs), np.array(pstates), np.array(hidden_states, dtype=int) def fit_df(self, dfs, pstate_col=PSTATE_COL): """ Convenience function to fit a model from a list of dataframes """ obs_cols = list(self.emission_name) obs = [df[df.columns.difference([pstate_col])][obs_cols].values for df in dfs] pstates = [df[pstate_col].values for df in dfs] return self.fit(obs, pstates) def score_df(self, df, pstate_col=PSTATE_COL): """ """ obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values return self.score(obs, pstates) def score_events_df(self, df, pstate_col=PSTATE_COL, score_col='score'): """ """ df = df.copy() obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values df[score_col] = self.score_events(obs, pstates) return df def predict_states_df(self, df, pstate_col=PSTATE_COL, hstate_col=HSTATE_COL): df = df.copy() obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values _, df[hstate_col] = self.predict_states(obs, pstates) return df def predict_df(self, df, next_pstate=None, pstate_col=PSTATE_COL): obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values return self.predict(obs, pstates, next_pstate) def sample_df(self, pstates=None, n_obs=None, random_state=None, pstate_col=PSTATE_COL, hstate_col=HSTATE_COL): """ Convenience function to generate samples a model and create a dataframe """ try: import pandas as pd except Exception as e: raise e obs, pstates, hstates = self.sample(pstates, n_obs, random_state) items = [] if pstate_col is not None: items.append((pstate_col, pstates)) if hstate_col is not None: items.append((hstate_col, hstates)) items = items + [(self.emission_name[i], obs[:, i]) for i in range(self.n_features)] df = pd.DataFrame.from_items(items) return df def __str__(self): pstates = sorted(self.e.keys()) sep = '-' * 80 + '\n' sep2 = '_' * 40 + '\n' out = 'POHMM\n' out += 'H-states: %d\n' % self.n_hidden_states out += 'P-states: (%d) %s\n' % (len(pstates), str(pstates)) out += 'Emission: %s\n' % (self.emission_distr) out += sep out += 'Transition probas\n' out += sep2 out += '. -> .\n%s\n' % str(self.transmat[0, 0]) for pstate in pstates: out += sep2 out += '%s -> .\n%s\n' % (pstate, str(self.transmat[self.e[pstate], 0])) out += sep2 out += '. -> %s\n%s\n' % (pstate, str(self.transmat[0, self.e[pstate]])) for pstate1, pstate2 in product(pstates, pstates): out += sep2 out += '%s -> %s\n%s\n' % (pstate1, pstate2, str(self.transmat[self.e[pstate1], self.e[pstate2]])) out += sep out += 'Starting probas\n' out += '.: %s\n' % str(self.startprob[0]) for pstate in pstates: out += '%s: %s\n' % (pstate, str(self.startprob[self.e[pstate]])) out += sep out += 'Steady probas\n' out += '.: %s\n' % str(self.steadyprob[0]) for pstate in pstates: out += '%s: %s\n' % (pstate, str(self.steadyprob[self.e[pstate]])) out += sep out += 'Emissions\n' for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): out += sep2 out += 'Feature %s: %s\n' % (feature_name, feature_distr) for param in _DISTRIBUTIONS[feature_distr]: out += '.: %s = %s\n' % (param, str(self.emission[feature_name][param][0])) for pstate in pstates: out += '%s: %s = %s\n' % (pstate, param, str(self.emission[feature_name][param][self.e[pstate]])) out += sep return out def expected_value(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): """ Determine the joint maximum likelihood estimate """ # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': return np.sum(w * expected_lognormal(self.emission[feature]['logsigma'].flatten(), self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': return np.sum(w * expected_normal(self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten())) elif hstate is None: # Marginalize hstate if self.emission_name_distr[feature] == 'lognormal': return np.sum(hstate_prob * expected_lognormal(self.emission[feature]['logsigma'][pstate_idx, :], self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': return np.sum(hstate_prob * expected_normal(self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :])) elif pstate is None: # Marginalize pstate if self.emission_name_distr[feature] == 'lognormal': return np.sum(pstate_prob * expected_lognormal(self.emission[feature]['logsigma'][:, hstate], self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': return np.sum(pstate_prob * expected_normal(self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate])) else: if self.emission_name_distr[feature] == 'lognormal': return expected_lognormal(self.emission[feature]['logsigma'][pstate_idx, hstate], self.emission[feature]['logmu'][pstate_idx, hstate]) elif self.emission_name_distr[feature] == 'normal': return expected_normal(self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) return def pdf_fn(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'].flatten(), 0, np.exp(self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten()) elif hstate is None: # Marginalize hstate w = hstate_prob if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][pstate_idx, :], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :]) elif pstate is None: # Marginalize pstate w = pstate_prob if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][:, hstate], 0, np.exp(self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate]) else: w = 1 if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][pstate_idx, hstate], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, hstate])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) def fn(x): if np.isscalar(x): p = np.sum(w * pdf(x)) else: x = np.array(x) p = np.zeros(len(x)) for i, xi in enumerate(x): p[i] = np.sum(w * pdf(xi)) return p return fn def cdf_fn(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'].flatten(), 0, np.exp(self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten()) elif hstate is None: # Marginalize hstate w = hstate_prob if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][pstate_idx, :], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :]) elif pstate is None: # Marginalize pstate w = pstate_prob if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][:, hstate], 0, np.exp(self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate]) else: w = 1 if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][pstate_idx, hstate], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, hstate])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) def fn(x): if np.isscalar(x): p = np.sum(w * cdf(x)) else: x = np.array(x) p = np.zeros(len(x)) for i, xi in enumerate(x): p[i] = np.sum(w * cdf(xi)) return p return fn def params(self, pstates=None): if pstates is None: pstates = [None] + sorted( set(self.er.values())) # TODO: self.e caches any unknown value, maybe it shouldn't? params = [] # emission parameters for hstate, pstate_label in product(range(self.n_hidden_states), pstates): for feature, distr in zip(self.emission_name, self.emission_distr): for feature_param in _DISTRIBUTIONS[distr]: params.append(self.emission[feature][feature_param][self.e[pstate_label], hstate]) # transition parameters, diagonals only assuming 2 state for hstate, pstate_label in product(range(self.n_hidden_states), pstates): params.append(self.transmat[self.e[pstate_label], self.e[pstate_label], hstate, hstate]) return np.array(params)
vmonaco/pohmm
pohmm/pohmm.py
Pohmm.fit
python
def fit(self, obs, pstates, unique_pstates=None): obs = [np.array(o) for o in obs] pstates = [np.array(p) for p in pstates] # List or array of observation sequences assert len(obs) == len(pstates) assert obs[0].ndim == 2 assert pstates[0].ndim == 1 if unique_pstates is not None: self._init_pstates(unique_pstates) else: self._init_pstates(list(set(np.concatenate(pstates)))) # Map the partial states to a unique index pstates_idx = [np.array([self.e[p] for p in seq]) for seq in pstates] if self.init_method == 'rand': self._init_random() elif self.init_method == 'obs': self._init_from_obs(obs, pstates_idx) self._init_pstate_freqs(pstates_idx) self._smooth() logprob = [] for i in range(self.max_iter): # Expectation step stats = self._initialize_sufficient_statistics() curr_logprob = 0 for obs_i, pstates_idx_i in zip(obs, pstates_idx): framelogprob = self._compute_log_likelihood(obs_i, pstates_idx_i) lpr, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx_i) bwdlattice = self._do_backward_pass(framelogprob, pstates_idx_i) gamma = fwdlattice + bwdlattice posteriors = np.exp(gamma.T - logsumexp(gamma, axis=1)).T curr_logprob += lpr self._accumulate_sufficient_statistics(stats, obs_i, pstates_idx_i, framelogprob, posteriors, fwdlattice, bwdlattice) logprob.append(curr_logprob) self.logprob_ = curr_logprob # Check for convergence. self.n_iter_performed_ = i if i > 0: self.logprob_delta = logprob[-1] - logprob[-2] if self.logprob_delta < self.thresh: break # Maximization step self._do_mstep(stats) # Mix the parameters self._smooth() return self
Estimate model parameters.
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/pohmm.py#L671-L731
[ "def logsumexp(arr, axis=0):\n \"\"\"\n Computes the sum of arr assuming arr is in the log domain.\n\n Returns log(sum(exp(arr))) while minimizing the possibility of\n over/underflow.\n \"\"\"\n arr = np.rollaxis(arr, axis)\n # Use the max to normalize, as with the log this is what accumulates\n # the less errors\n vmax = arr.max(axis=0)\n out = np.log(np.sum(np.exp(arr - vmax), axis=0))\n out += vmax\n return out\n", "def _compute_log_likelihood(self, obs, pstates_idx):\n\n q = np.zeros(shape=(len(obs), self.n_hidden_states, self.n_features))\n\n for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)):\n if feature_distr == 'normal':\n mu = self.emission[feature_name]['mu'][pstates_idx]\n sigma = self.emission[feature_name]['sigma'][pstates_idx]\n for j in range(self.n_hidden_states):\n q[:, j, col] = np.log(\n np.maximum(MIN_PROBA, stats.norm.pdf(obs[:, col], loc=mu[:, j], scale=sigma[:, j])))\n if feature_distr == 'lognormal':\n logmu = self.emission[feature_name]['logmu'][pstates_idx]\n logsigma = self.emission[feature_name]['logsigma'][pstates_idx]\n for j in range(self.n_hidden_states):\n q[:, j, col] = np.log(np.maximum(MIN_PROBA,\n stats.lognorm.pdf(obs[:, col], logsigma[:, j], loc=0,\n scale=np.exp(logmu[:, j]))))\n\n q = q.sum(axis=2)\n return q\n", "def _init_pstates(self, unique_pstates):\n # Map events to a unique index. The unknown p-state is at idx 0\n self.e = defaultdict(int)\n self.e.update(dict(zip(np.sort(unique_pstates), range(1, len(unique_pstates) + 1))))\n self.er = {v: k for k, v in self.e.items()} # Reverse lookup\n self.er[0] = UNKNOWN_PSTATE\n self.n_partial_states = len(self.e.keys()) + 1 # Add one for the unknown state\n return\n", "def _init_pstate_freqs(self, pstates_idx):\n # Partial state frequencies\n self.pstate_freq = Counter([idx for seq in pstates_idx for idx in seq])\n self.pstate_trans_freq = Counter([(idx1, idx2) for seq in pstates_idx for idx1, idx2 in zip(seq[:-1], seq[1:])])\n\n # Store freqs for the meta state\n self.pstate_freq[0] = len(np.concatenate(pstates_idx))\n\n for seq in pstates_idx:\n self.pstate_startprob[seq[0]] += 1\n self.pstate_steadyprob += np.bincount(seq, minlength=self.n_partial_states)\n for idx1, idx2 in zip(seq[:-1], seq[1:]):\n self.pstate_trans_freq[(0, 0)] += 1\n self.pstate_trans_freq[(idx1, 0)] += 1\n self.pstate_trans_freq[(0, idx2)] += 1\n\n self.pstate_transmat[idx1, idx2] += 1\n\n # TODO: separate probas from freqs\n # Normalize to get the probabilities, ignore the meta state at idx 0\n self.pstate_startprob[1:] = normalize(self.pstate_startprob[1:])\n self.pstate_transmat[1:, 1:] = normalize(self.pstate_transmat[1:, 1:], axis=1)\n self.pstate_steadyprob[1:] = normalize(self.pstate_steadyprob[1:])\n\n return\n", "def _init_from_obs(self, obs, pstates_idx):\n # Partial state probabilities\n self.pstate_startprob = np.zeros(self.n_partial_states)\n self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states))\n self.pstate_steadyprob = np.zeros(self.n_partial_states)\n\n # obs should be (N*T, n_features)\n # N is the number of samples\n # T is the size of each sample \n obs = np.concatenate(obs)\n pstates_idx = np.concatenate(pstates_idx)\n\n # Initialize starting and transition probas\n self.startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states\n self.steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states\n self.transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states,\n self.n_hidden_states)) / self.n_hidden_states\n\n # Initialize emission parameters\n # Hidden states are ordered by the first feature\n feature1 = self.emission_name[0]\n for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)):\n if feature_distr == 'normal':\n self.emission[feature_name]['mu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states))\n self.emission[feature_name]['sigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states))\n for idx in range(1, self.n_partial_states):\n idx_pstate = (pstates_idx == idx)\n\n if not np.any(idx_pstate):\n idx_pstate = np.arange(len(pstates_idx))\n\n if feature_name == feature1:\n self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA)\n self.emission[feature_name]['mu'][idx] = obs[idx_pstate, col].mean() + \\\n obs[:, col].std() * np.linspace(\n -self.init_spread, self.init_spread,\n self.n_hidden_states)\n else:\n self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA)\n self.emission[feature_name]['mu'][idx, :] = obs[idx_pstate, col].mean()\n\n if feature_distr == 'lognormal':\n self.emission[feature_name]['logmu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states))\n self.emission[feature_name]['logsigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states))\n\n for idx in range(1, self.n_partial_states):\n idx_pstate = (pstates_idx == idx)\n\n if not np.any(idx_pstate):\n idx_pstate = np.arange(len(pstates_idx))\n\n if feature_name == feature1:\n self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(),\n MIN_PROBA)\n self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() + np.log(\n obs[:, col]).std() * np.linspace(-self.init_spread, self.init_spread,\n self.n_hidden_states)\n else:\n self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(),\n MIN_PROBA)\n self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean()\n\n return\n", "def _init_random(self, random_state=None):\n\n if random_state is None:\n random_state = self.random_state\n random_state = check_random_state(random_state)\n\n self.pstate_startprob = np.zeros(self.n_partial_states)\n self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states))\n self.pstate_steadyprob = np.zeros(self.n_partial_states)\n\n self.pstate_startprob[1:] = gen_stochastic_matrix(size=self.n_partial_states - 1, random_state=random_state)\n self.pstate_transmat[1:, 1:] = gen_stochastic_matrix(\n size=(self.n_partial_states - 1, self.n_partial_states - 1), random_state=random_state)\n self.pstate_steadyprob[1:] = steadystate(self.pstate_transmat[1:, 1:])\n\n self.startprob = gen_stochastic_matrix(size=(self.n_partial_states, self.n_hidden_states),\n random_state=random_state)\n\n transmat = np.zeros((self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states))\n for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)):\n transmat[i, j] = gen_stochastic_matrix(size=(self.n_hidden_states, self.n_hidden_states),\n random_state=random_state)\n self.transmat = normalize(transmat, axis=3)\n\n # Initialize emission parameters\n for feature_name, feature_distr in zip(self.emission_name, self.emission_distr):\n if feature_distr == 'normal':\n self.emission[feature_name]['mu'] = random_state.uniform(*_RANDINIT[feature_distr]['mu'], size=(\n self.n_partial_states, self.n_hidden_states))\n self.emission[feature_name]['sigma'] = random_state.uniform(*_RANDINIT[feature_distr]['sigma'], size=(\n self.n_partial_states, self.n_hidden_states))\n if feature_distr == 'lognormal':\n self.emission[feature_name]['logmu'] = random_state.uniform(*_RANDINIT[feature_distr]['logmu'], size=(\n self.n_partial_states, self.n_hidden_states))\n self.emission[feature_name]['logsigma'] = random_state.uniform(*_RANDINIT[feature_distr]['logsigma'],\n size=(self.n_partial_states,\n self.n_hidden_states))\n\n if self.emission_distr[0] == 'normal':\n self.emission[self.emission_name[0]]['mu'] = np.sort(self.emission[self.emission_name[0]]['mu'], axis=1)\n elif self.emission_distr[0] == 'lognormal':\n self.emission[self.emission_name[0]]['logmu'] = np.sort(self.emission[self.emission_name[0]]['logmu'],\n axis=1)\n\n return\n", "def _smooth(self):\n self._compute_marginals()\n\n startprob = self.startprob\n for j in range(1, self.n_partial_states):\n if 'freq' == self.smoothing['startprob']:\n w_ = 1 / (1 + self.pstate_freq[j])\n w_j = 1 - w_\n elif 'proba' == self.smoothing['startprob']:\n w_j = self.pstate_steadyprob[j]\n w_ = 1 - w_j\n elif 'exp' == self.smoothing['startprob']:\n w_ = np.exp(-self.pstate_freq[j])\n w_j = 1 - w_\n elif 'fixed' == self.smoothing['startprob']:\n w_ = 1\n w_j = 0\n elif self.smoothing['startprob'] is None:\n w_ = 0\n w_j = 1\n else:\n raise Exception('Wrong smoothing for startprob: ' + self.smoothing['startprob'])\n startprob[j] = w_j * self.startprob[j] + w_ * self.startprob[0]\n self.startprob = startprob\n\n transmat = self.transmat\n for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)):\n if 'freq' == self.smoothing['transmat']:\n w_i0 = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[0, j])\n w_0j = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[i, 0])\n w_ij = 1 - (w_i0 + w_0j)\n w_ = 0\n elif 'proba' == self.smoothing['transmat']:\n denom = self.pstate_transmat[i, j] + self.pstate_transmat[i, :].sum() + self.pstate_transmat[:, j].sum()\n w_i0 = self.pstate_transmat[i, :].sum() / denom\n w_0j = self.pstate_transmat[:, j].sum() / denom\n w_ij = self.pstate_transmat[i, j] / denom\n w_ = 0\n elif 'exp' == self.smoothing['transmat']:\n w_i0 = self.pstate_trans_freq[i, 0] * np.exp(-self.pstate_trans_freq[i, j])\n w_0j = self.pstate_trans_freq[0, j] * np.exp(-self.pstate_trans_freq[i, j])\n w_ = self.pstate_trans_freq[0, 0] * np.exp(\n -(self.pstate_trans_freq[i, 0] + self.pstate_trans_freq[0, j]))\n w_ij = self.pstate_trans_freq[i, j]\n w_ij, w_i0, w_0j, w_ = normalize(np.array([w_ij, w_i0, w_0j, w_]))\n elif 'fixed' == self.smoothing['transmat']:\n w_i0 = 0\n w_0j = 0\n w_ = 1\n w_ij = 0\n elif self.smoothing['transmat'] is None:\n w_i0 = 0\n w_0j = 0\n w_ = 0\n w_ij = 1\n else:\n raise Exception('Wrong smoothing for transmat: ' + self.smoothing['transmat'])\n\n assert (w_i0 + w_0j + w_ + w_ij) - 1 < TOLERANCE\n transmat[i, j] = w_ij * self.transmat[i, j] + w_i0 * self.transmat[i, 0] + w_0j * self.transmat[0, j] + w_ * \\\n self.transmat[\n 0, 0]\n self.transmat = transmat\n\n assert np.all(self.startprob.sum(axis=1) - 1 < TOLERANCE)\n assert np.all(self.steadyprob.sum(axis=1) - 1 < TOLERANCE)\n assert np.all(self.transmat.sum(axis=3) - 1 < TOLERANCE)\n\n for feature_name, feature_distr in zip(self.emission_name, self.emission_distr):\n for param in _DISTRIBUTIONS[feature_distr]:\n key = '%s:%s' % (feature_name, param)\n for j in range(1, self.n_partial_states):\n if 'freq' == self.smoothing[key]:\n w_ = 1 / (1 + self.pstate_freq[j])\n w_j = 1 - w_\n elif 'proba' == self.smoothing[key]:\n w_j = self.pstate_steadyprob[j]\n w_ = 1 - w_j\n elif 'exp' == self.smoothing[key]:\n w_ = np.exp(-self.pstate_freq[j])\n w_j = 1 - w_\n elif 'fixed' == self.smoothing[key]:\n w_ = 1\n w_j = 0\n elif self.smoothing[key] is None:\n w_ = 0\n w_j = 1\n else:\n raise Exception('Wrong smoothing for ' + key)\n self.emission[feature_name][param][j] = w_j * self.emission[feature_name][param][j] + w_ * \\\n self.emission[\n feature_name][\n param][0]\n\n return\n", "def _initialize_sufficient_statistics(self):\n stats = {\n 'nobs': 0,\n 'post': np.zeros((self.n_partial_states, self.n_hidden_states)),\n 'obs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)),\n 'obs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)),\n 'lnobs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)),\n 'lnobs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)),\n 'start': np.zeros((self.n_partial_states, self.n_hidden_states)),\n 'steady': np.zeros((self.n_partial_states, self.n_hidden_states)),\n 'trans': np.zeros(\n (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states))\n }\n\n return stats\n", "def _accumulate_sufficient_statistics(self, stats, obs, pstates_idx, framelogprob,\n posteriors, fwdlattice, bwdlattice):\n stats['nobs'] += 1\n n_observations, n_hidden_states = framelogprob.shape\n\n stats['start'][0] += posteriors[0]\n for i in range(self.n_partial_states):\n if len(np.where(pstates_idx == i)[0]) > 0:\n stats['start'][i] += posteriors[np.where(pstates_idx == i)[0].min()]\n\n if n_observations > 1:\n lneta = np.zeros((n_observations - 1, n_hidden_states, n_hidden_states))\n _hmmc._compute_lneta(n_observations, n_hidden_states, pstates_idx, fwdlattice,\n self._log_transmat, bwdlattice, framelogprob,\n lneta)\n\n for i, j in unique_rows(np.c_[pstates_idx[:-1], pstates_idx[1:]]):\n if ((pstates_idx[:-1] == i) & (pstates_idx[1:] == j)).sum() > 0:\n stats['trans'][i, j] += np.exp(\n logsumexp(lneta[(pstates_idx[:-1] == i) & (pstates_idx[1:] == j)], axis=0))\n\n for i in range(self.n_partial_states):\n stats['post'][i] += posteriors[pstates_idx == i].sum(axis=0)\n stats['obs'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i])\n stats['obs**2'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i] ** 2)\n stats['lnobs'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i]))\n stats['lnobs**2'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i]) ** 2)\n\n return\n", "def _do_mstep(self, stats):\n self.startprob = normalize(np.maximum(stats['start'], MIN_PROBA), axis=1)\n self.transmat = normalize(np.maximum(stats['trans'], MIN_PROBA), axis=3)\n\n for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)):\n if feature_distr == 'normal':\n denom = np.maximum(stats['post'], MIN_PROBA)\n mu = stats['obs'][:, :, col] / denom\n cv_num = (stats['obs**2'][:, :, col]\n - 2 * mu * stats['obs'][:, :, col]\n + mu ** 2 * denom)\n sigma = np.sqrt(cv_num / denom)\n sigma[np.isnan(sigma)] = MIN_PROBA\n self.emission[feature_name]['mu'] = mu\n self.emission[feature_name]['sigma'] = np.maximum(sigma, MIN_PROBA)\n if feature_distr == 'lognormal':\n denom = np.maximum(stats['post'], MIN_PROBA)\n mu = stats['lnobs'][:, :, col] / denom\n cv_num = (stats['lnobs**2'][:, :, col]\n - 2 * mu * stats['lnobs'][:, :, col]\n + mu ** 2 * denom)\n sigma = np.sqrt(cv_num / denom)\n sigma[np.isnan(sigma)] = MIN_PROBA\n self.emission[feature_name]['logmu'] = mu\n self.emission[feature_name]['logsigma'] = np.maximum(sigma, MIN_PROBA)\n\n return\n", "def _do_forward_pass(self, framelogprob, event_idx):\n n_observations, n_hidden_states = framelogprob.shape\n fwdlattice = np.zeros((n_observations, n_hidden_states))\n _hmmc._forward(n_observations, n_hidden_states,\n event_idx, self._log_startprob,\n self._log_transmat, framelogprob, fwdlattice)\n\n return logsumexp(fwdlattice[-1]), fwdlattice\n", "def _do_backward_pass(self, framelogprob, event_idx):\n n_observations, n_hidden_states = framelogprob.shape\n bwdlattice = np.zeros((n_observations, n_hidden_states))\n _hmmc._backward(n_observations, n_hidden_states,\n event_idx, self._log_startprob,\n self._log_transmat, framelogprob, bwdlattice)\n return bwdlattice\n" ]
class Pohmm(object): """ Partially observable hidden Markov model """ def __init__(self, n_hidden_states=2, emissions=['normal'], max_iter=1000, thresh=1e-6, init_method='obs', init_spread=2, smoothing=None, random_state=None): if type(n_hidden_states) is int: self.n_hidden_states = n_hidden_states else: raise Exception('Wrong type for n_hidden_states. Must be int') if type(emissions[0]) is tuple: emission_name, emission_distr = zip(*emissions) elif type(emissions[0]) is str: emission_name, emission_distr = np.arange(len(emissions)), emissions for distr in emission_distr: if distr not in _DISTRIBUTIONS.keys(): raise ValueError('Emission distribution must be one of', _DISTRIBUTIONS.keys()) self.emission_name = emission_name self.emission_distr = emission_distr self.emission_name_distr = dict(zip(emission_name, emission_distr)) self.n_features = len(emissions) # Set up the emission parameters # emission: {'feature':{'param': np.array(shape=(n_partial_states, n_hidden_states))}} self.emission = defaultdict(dict) for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: self.emission[name][param] = None self.emission = dict(self.emission) assert max_iter >= 0 assert thresh >= 0 self.max_iter = max_iter self.thresh = thresh assert init_spread >= 0 self.init_spread = init_spread if init_method not in _INIT_METHODS: raise ValueError('init_method must be one of', _INIT_METHODS) self.init_method = init_method if smoothing is None: smoothing = {'transmat': None, 'startprob': None} for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: smoothing['%s:%s' % (name, param)] = None elif type(smoothing) is str: s = smoothing smoothing = {'transmat': s, 'startprob': s} for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: smoothing['%s:%s' % (name, param)] = s elif type(smoothing) is dict: assert 'transmat' in smoothing.keys() assert 'startprob' in smoothing.keys() for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: assert param in smoothing.keys() or '%s:%s' % (name, param) in smoothing.keys() if param in smoothing.keys() and '%s:%s' % (name, param) not in smoothing.keys(): smoothing['%s:%s' % (name, param)] = smoothing[param] else: raise Exception('Wrong type for smoothing. Must be None, str, or dict') self.smoothing = smoothing self.random_state = random_state # Number of unique partial states is unknown until fit self.n_partial_states = None # Results after fitting the model self.logprob_ = None self.n_iter_performed_ = None self.logprob_delta_ = None # Mapping between p-states and a unique index # Defaults to 0 for unknown or missing p-states self.e = defaultdict(int) def _get_startprob(self): return np.exp(self._log_startprob) def _set_startprob(self, startprob): if startprob is None: startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states else: startprob = np.asarray(startprob, dtype=np.float) # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(startprob): startprob = normalize(startprob, axis=1) if len(startprob) != self.n_partial_states: raise ValueError('startprob must have length n_partial_states') if not np.allclose(np.sum(startprob, axis=1), 1.0): raise ValueError('startprob must sum to 1.0') self._log_startprob = np.log(np.asarray(startprob).copy()) startprob = property(_get_startprob, _set_startprob) def _get_steadyprob(self): return np.exp(self._log_steadyprob) def _set_steadyprob(self, steadyprob): if steadyprob is None: steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states else: steadyprob = np.asarray(steadyprob, dtype=np.float) # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(steadyprob): steadyprob = normalize(steadyprob, axis=1) if len(steadyprob) != self.n_partial_states: raise ValueError('steadyprob must have length n_partial_states') if not np.allclose(np.sum(steadyprob, axis=1), 1.0): raise ValueError('steadyprob must sum to 1.0') self._log_steadyprob = np.log(np.asarray(steadyprob).copy()) steadyprob = property(_get_steadyprob, _set_steadyprob) def _get_transmat(self): return np.exp(self._log_transmat) def _set_transmat(self, transmat): if transmat is None: transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) / self.n_hidden_states # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(transmat): transmat = normalize(transmat, axis=3) if (np.asarray(transmat).shape != (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)): raise ValueError('transmat must have shape ' '(n_partial_states,n_partial_states,n_hidden_states,n_hidden_states)') if not np.all(np.allclose(np.sum(transmat, axis=3), 1.0)): raise ValueError('Rows of transmat must sum to 1.0') self._log_transmat = np.log(np.asarray(transmat).copy()) underflow_idx = np.isnan(self._log_transmat) self._log_transmat[underflow_idx] = NEGINF transmat = property(_get_transmat, _set_transmat) def _compute_log_likelihood(self, obs, pstates_idx): q = np.zeros(shape=(len(obs), self.n_hidden_states, self.n_features)) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': mu = self.emission[feature_name]['mu'][pstates_idx] sigma = self.emission[feature_name]['sigma'][pstates_idx] for j in range(self.n_hidden_states): q[:, j, col] = np.log( np.maximum(MIN_PROBA, stats.norm.pdf(obs[:, col], loc=mu[:, j], scale=sigma[:, j]))) if feature_distr == 'lognormal': logmu = self.emission[feature_name]['logmu'][pstates_idx] logsigma = self.emission[feature_name]['logsigma'][pstates_idx] for j in range(self.n_hidden_states): q[:, j, col] = np.log(np.maximum(MIN_PROBA, stats.lognorm.pdf(obs[:, col], logsigma[:, j], loc=0, scale=np.exp(logmu[:, j])))) q = q.sum(axis=2) return q def _generate_sample_from_state(self, hidden_state, pstates_idx, random_state=None): sample = np.zeros(self.n_features) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': mu = self.emission[feature_name]['mu'][pstates_idx][hidden_state] sigma = self.emission[feature_name]['sigma'][pstates_idx][hidden_state] sample[col] = stats.norm.rvs(loc=mu, scale=sigma, random_state=random_state) if feature_distr == 'lognormal': logmu = self.emission[feature_name]['logmu'][pstates_idx][hidden_state] logsigma = self.emission[feature_name]['logsigma'][pstates_idx][hidden_state] sample[col] = stats.lognorm.rvs(logsigma, loc=0, scale=np.exp(logmu), random_state=random_state) return sample def _init_pstates(self, unique_pstates): # Map events to a unique index. The unknown p-state is at idx 0 self.e = defaultdict(int) self.e.update(dict(zip(np.sort(unique_pstates), range(1, len(unique_pstates) + 1)))) self.er = {v: k for k, v in self.e.items()} # Reverse lookup self.er[0] = UNKNOWN_PSTATE self.n_partial_states = len(self.e.keys()) + 1 # Add one for the unknown state return def _init_pstate_freqs(self, pstates_idx): # Partial state frequencies self.pstate_freq = Counter([idx for seq in pstates_idx for idx in seq]) self.pstate_trans_freq = Counter([(idx1, idx2) for seq in pstates_idx for idx1, idx2 in zip(seq[:-1], seq[1:])]) # Store freqs for the meta state self.pstate_freq[0] = len(np.concatenate(pstates_idx)) for seq in pstates_idx: self.pstate_startprob[seq[0]] += 1 self.pstate_steadyprob += np.bincount(seq, minlength=self.n_partial_states) for idx1, idx2 in zip(seq[:-1], seq[1:]): self.pstate_trans_freq[(0, 0)] += 1 self.pstate_trans_freq[(idx1, 0)] += 1 self.pstate_trans_freq[(0, idx2)] += 1 self.pstate_transmat[idx1, idx2] += 1 # TODO: separate probas from freqs # Normalize to get the probabilities, ignore the meta state at idx 0 self.pstate_startprob[1:] = normalize(self.pstate_startprob[1:]) self.pstate_transmat[1:, 1:] = normalize(self.pstate_transmat[1:, 1:], axis=1) self.pstate_steadyprob[1:] = normalize(self.pstate_steadyprob[1:]) return def _init_from_obs(self, obs, pstates_idx): # Partial state probabilities self.pstate_startprob = np.zeros(self.n_partial_states) self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states)) self.pstate_steadyprob = np.zeros(self.n_partial_states) # obs should be (N*T, n_features) # N is the number of samples # T is the size of each sample obs = np.concatenate(obs) pstates_idx = np.concatenate(pstates_idx) # Initialize starting and transition probas self.startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states self.steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states self.transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) / self.n_hidden_states # Initialize emission parameters # Hidden states are ordered by the first feature feature1 = self.emission_name[0] for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': self.emission[feature_name]['mu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['sigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) for idx in range(1, self.n_partial_states): idx_pstate = (pstates_idx == idx) if not np.any(idx_pstate): idx_pstate = np.arange(len(pstates_idx)) if feature_name == feature1: self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA) self.emission[feature_name]['mu'][idx] = obs[idx_pstate, col].mean() + \ obs[:, col].std() * np.linspace( -self.init_spread, self.init_spread, self.n_hidden_states) else: self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA) self.emission[feature_name]['mu'][idx, :] = obs[idx_pstate, col].mean() if feature_distr == 'lognormal': self.emission[feature_name]['logmu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['logsigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) for idx in range(1, self.n_partial_states): idx_pstate = (pstates_idx == idx) if not np.any(idx_pstate): idx_pstate = np.arange(len(pstates_idx)) if feature_name == feature1: self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(), MIN_PROBA) self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() + np.log( obs[:, col]).std() * np.linspace(-self.init_spread, self.init_spread, self.n_hidden_states) else: self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(), MIN_PROBA) self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() return def _init_random(self, random_state=None): if random_state is None: random_state = self.random_state random_state = check_random_state(random_state) self.pstate_startprob = np.zeros(self.n_partial_states) self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states)) self.pstate_steadyprob = np.zeros(self.n_partial_states) self.pstate_startprob[1:] = gen_stochastic_matrix(size=self.n_partial_states - 1, random_state=random_state) self.pstate_transmat[1:, 1:] = gen_stochastic_matrix( size=(self.n_partial_states - 1, self.n_partial_states - 1), random_state=random_state) self.pstate_steadyprob[1:] = steadystate(self.pstate_transmat[1:, 1:]) self.startprob = gen_stochastic_matrix(size=(self.n_partial_states, self.n_hidden_states), random_state=random_state) transmat = np.zeros((self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)): transmat[i, j] = gen_stochastic_matrix(size=(self.n_hidden_states, self.n_hidden_states), random_state=random_state) self.transmat = normalize(transmat, axis=3) # Initialize emission parameters for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): if feature_distr == 'normal': self.emission[feature_name]['mu'] = random_state.uniform(*_RANDINIT[feature_distr]['mu'], size=( self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['sigma'] = random_state.uniform(*_RANDINIT[feature_distr]['sigma'], size=( self.n_partial_states, self.n_hidden_states)) if feature_distr == 'lognormal': self.emission[feature_name]['logmu'] = random_state.uniform(*_RANDINIT[feature_distr]['logmu'], size=( self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['logsigma'] = random_state.uniform(*_RANDINIT[feature_distr]['logsigma'], size=(self.n_partial_states, self.n_hidden_states)) if self.emission_distr[0] == 'normal': self.emission[self.emission_name[0]]['mu'] = np.sort(self.emission[self.emission_name[0]]['mu'], axis=1) elif self.emission_distr[0] == 'lognormal': self.emission[self.emission_name[0]]['logmu'] = np.sort(self.emission[self.emission_name[0]]['logmu'], axis=1) return def _smooth(self): self._compute_marginals() startprob = self.startprob for j in range(1, self.n_partial_states): if 'freq' == self.smoothing['startprob']: w_ = 1 / (1 + self.pstate_freq[j]) w_j = 1 - w_ elif 'proba' == self.smoothing['startprob']: w_j = self.pstate_steadyprob[j] w_ = 1 - w_j elif 'exp' == self.smoothing['startprob']: w_ = np.exp(-self.pstate_freq[j]) w_j = 1 - w_ elif 'fixed' == self.smoothing['startprob']: w_ = 1 w_j = 0 elif self.smoothing['startprob'] is None: w_ = 0 w_j = 1 else: raise Exception('Wrong smoothing for startprob: ' + self.smoothing['startprob']) startprob[j] = w_j * self.startprob[j] + w_ * self.startprob[0] self.startprob = startprob transmat = self.transmat for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)): if 'freq' == self.smoothing['transmat']: w_i0 = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[0, j]) w_0j = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[i, 0]) w_ij = 1 - (w_i0 + w_0j) w_ = 0 elif 'proba' == self.smoothing['transmat']: denom = self.pstate_transmat[i, j] + self.pstate_transmat[i, :].sum() + self.pstate_transmat[:, j].sum() w_i0 = self.pstate_transmat[i, :].sum() / denom w_0j = self.pstate_transmat[:, j].sum() / denom w_ij = self.pstate_transmat[i, j] / denom w_ = 0 elif 'exp' == self.smoothing['transmat']: w_i0 = self.pstate_trans_freq[i, 0] * np.exp(-self.pstate_trans_freq[i, j]) w_0j = self.pstate_trans_freq[0, j] * np.exp(-self.pstate_trans_freq[i, j]) w_ = self.pstate_trans_freq[0, 0] * np.exp( -(self.pstate_trans_freq[i, 0] + self.pstate_trans_freq[0, j])) w_ij = self.pstate_trans_freq[i, j] w_ij, w_i0, w_0j, w_ = normalize(np.array([w_ij, w_i0, w_0j, w_])) elif 'fixed' == self.smoothing['transmat']: w_i0 = 0 w_0j = 0 w_ = 1 w_ij = 0 elif self.smoothing['transmat'] is None: w_i0 = 0 w_0j = 0 w_ = 0 w_ij = 1 else: raise Exception('Wrong smoothing for transmat: ' + self.smoothing['transmat']) assert (w_i0 + w_0j + w_ + w_ij) - 1 < TOLERANCE transmat[i, j] = w_ij * self.transmat[i, j] + w_i0 * self.transmat[i, 0] + w_0j * self.transmat[0, j] + w_ * \ self.transmat[ 0, 0] self.transmat = transmat assert np.all(self.startprob.sum(axis=1) - 1 < TOLERANCE) assert np.all(self.steadyprob.sum(axis=1) - 1 < TOLERANCE) assert np.all(self.transmat.sum(axis=3) - 1 < TOLERANCE) for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[feature_distr]: key = '%s:%s' % (feature_name, param) for j in range(1, self.n_partial_states): if 'freq' == self.smoothing[key]: w_ = 1 / (1 + self.pstate_freq[j]) w_j = 1 - w_ elif 'proba' == self.smoothing[key]: w_j = self.pstate_steadyprob[j] w_ = 1 - w_j elif 'exp' == self.smoothing[key]: w_ = np.exp(-self.pstate_freq[j]) w_j = 1 - w_ elif 'fixed' == self.smoothing[key]: w_ = 1 w_j = 0 elif self.smoothing[key] is None: w_ = 0 w_j = 1 else: raise Exception('Wrong smoothing for ' + key) self.emission[feature_name][param][j] = w_j * self.emission[feature_name][param][j] + w_ * \ self.emission[ feature_name][ param][0] return def _initialize_sufficient_statistics(self): stats = { 'nobs': 0, 'post': np.zeros((self.n_partial_states, self.n_hidden_states)), 'obs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'obs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'lnobs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'lnobs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'start': np.zeros((self.n_partial_states, self.n_hidden_states)), 'steady': np.zeros((self.n_partial_states, self.n_hidden_states)), 'trans': np.zeros( (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) } return stats def _accumulate_sufficient_statistics(self, stats, obs, pstates_idx, framelogprob, posteriors, fwdlattice, bwdlattice): stats['nobs'] += 1 n_observations, n_hidden_states = framelogprob.shape stats['start'][0] += posteriors[0] for i in range(self.n_partial_states): if len(np.where(pstates_idx == i)[0]) > 0: stats['start'][i] += posteriors[np.where(pstates_idx == i)[0].min()] if n_observations > 1: lneta = np.zeros((n_observations - 1, n_hidden_states, n_hidden_states)) _hmmc._compute_lneta(n_observations, n_hidden_states, pstates_idx, fwdlattice, self._log_transmat, bwdlattice, framelogprob, lneta) for i, j in unique_rows(np.c_[pstates_idx[:-1], pstates_idx[1:]]): if ((pstates_idx[:-1] == i) & (pstates_idx[1:] == j)).sum() > 0: stats['trans'][i, j] += np.exp( logsumexp(lneta[(pstates_idx[:-1] == i) & (pstates_idx[1:] == j)], axis=0)) for i in range(self.n_partial_states): stats['post'][i] += posteriors[pstates_idx == i].sum(axis=0) stats['obs'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i]) stats['obs**2'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i] ** 2) stats['lnobs'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i])) stats['lnobs**2'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i]) ** 2) return def _compute_marginals(self): # TODO: cythonize some of this # Start prob, weighted by p-state start probs self.startprob[0] = (self.pstate_startprob[1:, np.newaxis] * self.startprob[1:]).sum(axis=0) # Use the p-state transmat and transmat to get the full transmat full_transmat = ph2full(self.pstate_transmat[1:, 1:], self.transmat[1:, 1:]) full_steadyprob = steadystate(full_transmat) # Steady state probas are determined by the full trans mat, need to be updated steadyprob = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) steadyprob[0] = full_steadyprob.reshape(-1, self.n_hidden_states).sum(axis=0) for i in range(self.n_partial_states - 1): steadyprob[i + 1] = normalize( full_steadyprob[i * self.n_hidden_states:i * self.n_hidden_states + self.n_hidden_states]) self.steadyprob = steadyprob # Update the transations to/from the marginal state transmat = self.transmat # Group the hidden states within each partial state for hidx1, hidx2 in product(range(self.n_hidden_states), range(self.n_hidden_states)): transmat[0, 0][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, hidx2::self.n_hidden_states].sum() for pidx in range(self.n_partial_states - 1): transmat[pidx + 1, 0][hidx1, hidx2] = full_transmat[pidx * self.n_hidden_states + hidx1, hidx2::self.n_hidden_states].sum() transmat[0, pidx + 1][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, pidx * self.n_hidden_states + hidx2].sum() self.transmat = normalize(transmat, axis=3) pweights = self.pstate_steadyprob[1:, np.newaxis] # Update emission parameters for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): if feature_distr == 'normal': # Marginal state is a mixture of normals mu = self.emission[feature_name]['mu'][1:, :] sigma = self.emission[feature_name]['sigma'][1:, :] # Weighted mean and var mu_0 = (pweights * mu).sum(axis=0) self.emission[feature_name]['mu'][0, :] = mu_0 self.emission[feature_name]['sigma'][0, :] = np.sqrt( (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0)) if feature_distr == 'lognormal': # Marginal state is a mixture of normals mu = self.emission[feature_name]['logmu'][1:, :] sigma = self.emission[feature_name]['logsigma'][1:, :] # Weighted mean and var mu_0 = (pweights * mu).sum(axis=0) self.emission[feature_name]['logmu'][0, :] = mu_0 self.emission[feature_name]['logsigma'][0, :] = np.sqrt( (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0)) return def _do_mstep(self, stats): self.startprob = normalize(np.maximum(stats['start'], MIN_PROBA), axis=1) self.transmat = normalize(np.maximum(stats['trans'], MIN_PROBA), axis=3) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': denom = np.maximum(stats['post'], MIN_PROBA) mu = stats['obs'][:, :, col] / denom cv_num = (stats['obs**2'][:, :, col] - 2 * mu * stats['obs'][:, :, col] + mu ** 2 * denom) sigma = np.sqrt(cv_num / denom) sigma[np.isnan(sigma)] = MIN_PROBA self.emission[feature_name]['mu'] = mu self.emission[feature_name]['sigma'] = np.maximum(sigma, MIN_PROBA) if feature_distr == 'lognormal': denom = np.maximum(stats['post'], MIN_PROBA) mu = stats['lnobs'][:, :, col] / denom cv_num = (stats['lnobs**2'][:, :, col] - 2 * mu * stats['lnobs'][:, :, col] + mu ** 2 * denom) sigma = np.sqrt(cv_num / denom) sigma[np.isnan(sigma)] = MIN_PROBA self.emission[feature_name]['logmu'] = mu self.emission[feature_name]['logsigma'] = np.maximum(sigma, MIN_PROBA) return def _do_forward_pass(self, framelogprob, event_idx): n_observations, n_hidden_states = framelogprob.shape fwdlattice = np.zeros((n_observations, n_hidden_states)) _hmmc._forward(n_observations, n_hidden_states, event_idx, self._log_startprob, self._log_transmat, framelogprob, fwdlattice) return logsumexp(fwdlattice[-1]), fwdlattice def _do_backward_pass(self, framelogprob, event_idx): n_observations, n_hidden_states = framelogprob.shape bwdlattice = np.zeros((n_observations, n_hidden_states)) _hmmc._backward(n_observations, n_hidden_states, event_idx, self._log_startprob, self._log_transmat, framelogprob, bwdlattice) return bwdlattice def _do_viterbi_pass(self, framelogprob, event_idx): n_observations, n_components = framelogprob.shape state_sequence, logprob = _hmmc._viterbi( n_observations, n_components, event_idx, self._log_startprob, self._log_transmat, framelogprob) return logprob, state_sequence def rand(self, unique_pstates, random_state=None): """ Randomize the POHMM parameters """ self._init_pstates(unique_pstates) self._init_random(random_state=random_state) self._compute_marginals() return self def score(self, obs, pstates): """ Compute the log probability under the model. """ pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) logprob, _ = self._do_forward_pass(framelogprob, pstates_idx) return logprob def score_events(self, obs, pstates): """ Compute the log probability of each event under the model. """ pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) _, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx) L = logsumexp(fwdlattice, axis=1) return np.concatenate([L[[0]], np.diff(L)]) def predict_states(self, obs, pstates): pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) viterbi_logprob, state_sequence = self._do_viterbi_pass(framelogprob, pstates_idx) return viterbi_logprob, state_sequence def predict(self, obs, pstates, next_pstate=None): """ Predict the next observation """ assert len(obs) == len(pstates) pstates_idx = np.array([self.e[ei] for ei in pstates]) next_pstate_idx = self.e[next_pstate] if len(obs) == 0: # No history, use the starting probas next_hstate_prob = self.startprob[next_pstate_idx] else: # With a history, determine the hidden state posteriors using # the last posteriors and transition matrix framelogprob = self._compute_log_likelihood(obs, pstates_idx) _, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx) next_hstate_prob = np.zeros(self.n_hidden_states) alpha_n = fwdlattice[-1] vmax = alpha_n.max(axis=0) alpha_n = np.exp(alpha_n - vmax) alpha_n = alpha_n / alpha_n.sum() trans = self.transmat[pstates_idx[-1], next_pstate_idx] for i in range(self.n_hidden_states): next_hstate_prob[i] = np.sum([alpha_n[j] * trans[j, i] for j in range(self.n_hidden_states)]) assert next_hstate_prob.sum() - 1 < TOLERANCE # Make the prediction prediction = np.array( [self.expected_value(feature, pstate=next_pstate, hstate_prob=next_hstate_prob) for feature in self.emission_name]) # next_hstate = np.argmax(next_hstate_prob) # prediction = np.array( # [self.expected_value(feature, pstate=next_pstate, hstate=next_hstate) for feature in # self.emission_name]) return prediction def gen_pstates_idx(self, n, random_state=None): if random_state is None: random_state = self.random_state random_state = check_random_state(random_state) startprob_cdf = np.cumsum(self.pstate_startprob) transmat_cdf = np.cumsum(self.pstate_transmat, 1) # Initial state. rand = random_state.rand() curr_pstate = (startprob_cdf > rand).argmax() pstates = [curr_pstate] for _ in range(1, n): rand = random_state.rand() curr_pstate = (transmat_cdf[curr_pstate] > rand).argmax() pstates.append(curr_pstate) return np.array(pstates, dtype=int) def sample(self, pstates=None, n_obs=None, random_state=None): """ """ random_state = check_random_state(random_state) if pstates is None and n_obs is None: raise Exception('Must provide either pstates or n_obs') if pstates is not None and n_obs is not None: raise Exception('Must provide either pstates or n_obs but not both') gen_pstates = False rand = random_state.rand() if pstates is None: gen_pstates = True pstartprob_cdf = np.cumsum(self.pstate_startprob) ptransmat_cdf = np.cumsum(self.pstate_transmat, 1) # Initial pstate currpstate = (pstartprob_cdf > rand).argmax() pstates_idx = [currpstate] pstates = [self.er[currpstate]] else: n_obs = len(pstates) pstates_idx = np.array([self.e[p] for p in pstates]) startprob_pdf = self.startprob[pstates_idx[0]] startprob_cdf = np.cumsum(startprob_pdf) transmat_cdf = np.cumsum(self.transmat[0, pstates_idx[0]], 1) # Initial hidden state rand = random_state.rand() currstate = (startprob_cdf > rand).argmax() hidden_states = [currstate] obs = [self._generate_sample_from_state(currstate, pstates_idx[0], random_state)] for i in range(1, n_obs): rand = random_state.rand() if gen_pstates: currpstate = (ptransmat_cdf[currpstate] > rand).argmax() pstates_idx.append(currpstate) pstates.append(self.er[currpstate]) transmat_cdf = np.cumsum(self.transmat[pstates_idx[i - 1], pstates_idx[i]], 1) rand = random_state.rand() currstate = (transmat_cdf[currstate] > rand).argmax() hidden_states.append(currstate) obs.append(self._generate_sample_from_state(currstate, pstates_idx[i], random_state)) return np.array(obs), np.array(pstates), np.array(hidden_states, dtype=int) def fit_df(self, dfs, pstate_col=PSTATE_COL): """ Convenience function to fit a model from a list of dataframes """ obs_cols = list(self.emission_name) obs = [df[df.columns.difference([pstate_col])][obs_cols].values for df in dfs] pstates = [df[pstate_col].values for df in dfs] return self.fit(obs, pstates) def score_df(self, df, pstate_col=PSTATE_COL): """ """ obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values return self.score(obs, pstates) def score_events_df(self, df, pstate_col=PSTATE_COL, score_col='score'): """ """ df = df.copy() obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values df[score_col] = self.score_events(obs, pstates) return df def predict_states_df(self, df, pstate_col=PSTATE_COL, hstate_col=HSTATE_COL): df = df.copy() obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values _, df[hstate_col] = self.predict_states(obs, pstates) return df def predict_df(self, df, next_pstate=None, pstate_col=PSTATE_COL): obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values return self.predict(obs, pstates, next_pstate) def sample_df(self, pstates=None, n_obs=None, random_state=None, pstate_col=PSTATE_COL, hstate_col=HSTATE_COL): """ Convenience function to generate samples a model and create a dataframe """ try: import pandas as pd except Exception as e: raise e obs, pstates, hstates = self.sample(pstates, n_obs, random_state) items = [] if pstate_col is not None: items.append((pstate_col, pstates)) if hstate_col is not None: items.append((hstate_col, hstates)) items = items + [(self.emission_name[i], obs[:, i]) for i in range(self.n_features)] df = pd.DataFrame.from_items(items) return df def __str__(self): pstates = sorted(self.e.keys()) sep = '-' * 80 + '\n' sep2 = '_' * 40 + '\n' out = 'POHMM\n' out += 'H-states: %d\n' % self.n_hidden_states out += 'P-states: (%d) %s\n' % (len(pstates), str(pstates)) out += 'Emission: %s\n' % (self.emission_distr) out += sep out += 'Transition probas\n' out += sep2 out += '. -> .\n%s\n' % str(self.transmat[0, 0]) for pstate in pstates: out += sep2 out += '%s -> .\n%s\n' % (pstate, str(self.transmat[self.e[pstate], 0])) out += sep2 out += '. -> %s\n%s\n' % (pstate, str(self.transmat[0, self.e[pstate]])) for pstate1, pstate2 in product(pstates, pstates): out += sep2 out += '%s -> %s\n%s\n' % (pstate1, pstate2, str(self.transmat[self.e[pstate1], self.e[pstate2]])) out += sep out += 'Starting probas\n' out += '.: %s\n' % str(self.startprob[0]) for pstate in pstates: out += '%s: %s\n' % (pstate, str(self.startprob[self.e[pstate]])) out += sep out += 'Steady probas\n' out += '.: %s\n' % str(self.steadyprob[0]) for pstate in pstates: out += '%s: %s\n' % (pstate, str(self.steadyprob[self.e[pstate]])) out += sep out += 'Emissions\n' for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): out += sep2 out += 'Feature %s: %s\n' % (feature_name, feature_distr) for param in _DISTRIBUTIONS[feature_distr]: out += '.: %s = %s\n' % (param, str(self.emission[feature_name][param][0])) for pstate in pstates: out += '%s: %s = %s\n' % (pstate, param, str(self.emission[feature_name][param][self.e[pstate]])) out += sep return out def expected_value(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): """ Determine the joint maximum likelihood estimate """ # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': return np.sum(w * expected_lognormal(self.emission[feature]['logsigma'].flatten(), self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': return np.sum(w * expected_normal(self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten())) elif hstate is None: # Marginalize hstate if self.emission_name_distr[feature] == 'lognormal': return np.sum(hstate_prob * expected_lognormal(self.emission[feature]['logsigma'][pstate_idx, :], self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': return np.sum(hstate_prob * expected_normal(self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :])) elif pstate is None: # Marginalize pstate if self.emission_name_distr[feature] == 'lognormal': return np.sum(pstate_prob * expected_lognormal(self.emission[feature]['logsigma'][:, hstate], self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': return np.sum(pstate_prob * expected_normal(self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate])) else: if self.emission_name_distr[feature] == 'lognormal': return expected_lognormal(self.emission[feature]['logsigma'][pstate_idx, hstate], self.emission[feature]['logmu'][pstate_idx, hstate]) elif self.emission_name_distr[feature] == 'normal': return expected_normal(self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) return def pdf_fn(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'].flatten(), 0, np.exp(self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten()) elif hstate is None: # Marginalize hstate w = hstate_prob if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][pstate_idx, :], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :]) elif pstate is None: # Marginalize pstate w = pstate_prob if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][:, hstate], 0, np.exp(self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate]) else: w = 1 if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][pstate_idx, hstate], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, hstate])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) def fn(x): if np.isscalar(x): p = np.sum(w * pdf(x)) else: x = np.array(x) p = np.zeros(len(x)) for i, xi in enumerate(x): p[i] = np.sum(w * pdf(xi)) return p return fn def cdf_fn(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'].flatten(), 0, np.exp(self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten()) elif hstate is None: # Marginalize hstate w = hstate_prob if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][pstate_idx, :], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :]) elif pstate is None: # Marginalize pstate w = pstate_prob if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][:, hstate], 0, np.exp(self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate]) else: w = 1 if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][pstate_idx, hstate], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, hstate])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) def fn(x): if np.isscalar(x): p = np.sum(w * cdf(x)) else: x = np.array(x) p = np.zeros(len(x)) for i, xi in enumerate(x): p[i] = np.sum(w * cdf(xi)) return p return fn def params(self, pstates=None): if pstates is None: pstates = [None] + sorted( set(self.er.values())) # TODO: self.e caches any unknown value, maybe it shouldn't? params = [] # emission parameters for hstate, pstate_label in product(range(self.n_hidden_states), pstates): for feature, distr in zip(self.emission_name, self.emission_distr): for feature_param in _DISTRIBUTIONS[distr]: params.append(self.emission[feature][feature_param][self.e[pstate_label], hstate]) # transition parameters, diagonals only assuming 2 state for hstate, pstate_label in product(range(self.n_hidden_states), pstates): params.append(self.transmat[self.e[pstate_label], self.e[pstate_label], hstate, hstate]) return np.array(params)
vmonaco/pohmm
pohmm/pohmm.py
Pohmm.score
python
def score(self, obs, pstates): pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) logprob, _ = self._do_forward_pass(framelogprob, pstates_idx) return logprob
Compute the log probability under the model.
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/pohmm.py#L733-L740
[ "def _compute_log_likelihood(self, obs, pstates_idx):\n\n q = np.zeros(shape=(len(obs), self.n_hidden_states, self.n_features))\n\n for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)):\n if feature_distr == 'normal':\n mu = self.emission[feature_name]['mu'][pstates_idx]\n sigma = self.emission[feature_name]['sigma'][pstates_idx]\n for j in range(self.n_hidden_states):\n q[:, j, col] = np.log(\n np.maximum(MIN_PROBA, stats.norm.pdf(obs[:, col], loc=mu[:, j], scale=sigma[:, j])))\n if feature_distr == 'lognormal':\n logmu = self.emission[feature_name]['logmu'][pstates_idx]\n logsigma = self.emission[feature_name]['logsigma'][pstates_idx]\n for j in range(self.n_hidden_states):\n q[:, j, col] = np.log(np.maximum(MIN_PROBA,\n stats.lognorm.pdf(obs[:, col], logsigma[:, j], loc=0,\n scale=np.exp(logmu[:, j]))))\n\n q = q.sum(axis=2)\n return q\n", "def _do_forward_pass(self, framelogprob, event_idx):\n n_observations, n_hidden_states = framelogprob.shape\n fwdlattice = np.zeros((n_observations, n_hidden_states))\n _hmmc._forward(n_observations, n_hidden_states,\n event_idx, self._log_startprob,\n self._log_transmat, framelogprob, fwdlattice)\n\n return logsumexp(fwdlattice[-1]), fwdlattice\n" ]
class Pohmm(object): """ Partially observable hidden Markov model """ def __init__(self, n_hidden_states=2, emissions=['normal'], max_iter=1000, thresh=1e-6, init_method='obs', init_spread=2, smoothing=None, random_state=None): if type(n_hidden_states) is int: self.n_hidden_states = n_hidden_states else: raise Exception('Wrong type for n_hidden_states. Must be int') if type(emissions[0]) is tuple: emission_name, emission_distr = zip(*emissions) elif type(emissions[0]) is str: emission_name, emission_distr = np.arange(len(emissions)), emissions for distr in emission_distr: if distr not in _DISTRIBUTIONS.keys(): raise ValueError('Emission distribution must be one of', _DISTRIBUTIONS.keys()) self.emission_name = emission_name self.emission_distr = emission_distr self.emission_name_distr = dict(zip(emission_name, emission_distr)) self.n_features = len(emissions) # Set up the emission parameters # emission: {'feature':{'param': np.array(shape=(n_partial_states, n_hidden_states))}} self.emission = defaultdict(dict) for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: self.emission[name][param] = None self.emission = dict(self.emission) assert max_iter >= 0 assert thresh >= 0 self.max_iter = max_iter self.thresh = thresh assert init_spread >= 0 self.init_spread = init_spread if init_method not in _INIT_METHODS: raise ValueError('init_method must be one of', _INIT_METHODS) self.init_method = init_method if smoothing is None: smoothing = {'transmat': None, 'startprob': None} for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: smoothing['%s:%s' % (name, param)] = None elif type(smoothing) is str: s = smoothing smoothing = {'transmat': s, 'startprob': s} for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: smoothing['%s:%s' % (name, param)] = s elif type(smoothing) is dict: assert 'transmat' in smoothing.keys() assert 'startprob' in smoothing.keys() for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: assert param in smoothing.keys() or '%s:%s' % (name, param) in smoothing.keys() if param in smoothing.keys() and '%s:%s' % (name, param) not in smoothing.keys(): smoothing['%s:%s' % (name, param)] = smoothing[param] else: raise Exception('Wrong type for smoothing. Must be None, str, or dict') self.smoothing = smoothing self.random_state = random_state # Number of unique partial states is unknown until fit self.n_partial_states = None # Results after fitting the model self.logprob_ = None self.n_iter_performed_ = None self.logprob_delta_ = None # Mapping between p-states and a unique index # Defaults to 0 for unknown or missing p-states self.e = defaultdict(int) def _get_startprob(self): return np.exp(self._log_startprob) def _set_startprob(self, startprob): if startprob is None: startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states else: startprob = np.asarray(startprob, dtype=np.float) # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(startprob): startprob = normalize(startprob, axis=1) if len(startprob) != self.n_partial_states: raise ValueError('startprob must have length n_partial_states') if not np.allclose(np.sum(startprob, axis=1), 1.0): raise ValueError('startprob must sum to 1.0') self._log_startprob = np.log(np.asarray(startprob).copy()) startprob = property(_get_startprob, _set_startprob) def _get_steadyprob(self): return np.exp(self._log_steadyprob) def _set_steadyprob(self, steadyprob): if steadyprob is None: steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states else: steadyprob = np.asarray(steadyprob, dtype=np.float) # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(steadyprob): steadyprob = normalize(steadyprob, axis=1) if len(steadyprob) != self.n_partial_states: raise ValueError('steadyprob must have length n_partial_states') if not np.allclose(np.sum(steadyprob, axis=1), 1.0): raise ValueError('steadyprob must sum to 1.0') self._log_steadyprob = np.log(np.asarray(steadyprob).copy()) steadyprob = property(_get_steadyprob, _set_steadyprob) def _get_transmat(self): return np.exp(self._log_transmat) def _set_transmat(self, transmat): if transmat is None: transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) / self.n_hidden_states # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(transmat): transmat = normalize(transmat, axis=3) if (np.asarray(transmat).shape != (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)): raise ValueError('transmat must have shape ' '(n_partial_states,n_partial_states,n_hidden_states,n_hidden_states)') if not np.all(np.allclose(np.sum(transmat, axis=3), 1.0)): raise ValueError('Rows of transmat must sum to 1.0') self._log_transmat = np.log(np.asarray(transmat).copy()) underflow_idx = np.isnan(self._log_transmat) self._log_transmat[underflow_idx] = NEGINF transmat = property(_get_transmat, _set_transmat) def _compute_log_likelihood(self, obs, pstates_idx): q = np.zeros(shape=(len(obs), self.n_hidden_states, self.n_features)) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': mu = self.emission[feature_name]['mu'][pstates_idx] sigma = self.emission[feature_name]['sigma'][pstates_idx] for j in range(self.n_hidden_states): q[:, j, col] = np.log( np.maximum(MIN_PROBA, stats.norm.pdf(obs[:, col], loc=mu[:, j], scale=sigma[:, j]))) if feature_distr == 'lognormal': logmu = self.emission[feature_name]['logmu'][pstates_idx] logsigma = self.emission[feature_name]['logsigma'][pstates_idx] for j in range(self.n_hidden_states): q[:, j, col] = np.log(np.maximum(MIN_PROBA, stats.lognorm.pdf(obs[:, col], logsigma[:, j], loc=0, scale=np.exp(logmu[:, j])))) q = q.sum(axis=2) return q def _generate_sample_from_state(self, hidden_state, pstates_idx, random_state=None): sample = np.zeros(self.n_features) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': mu = self.emission[feature_name]['mu'][pstates_idx][hidden_state] sigma = self.emission[feature_name]['sigma'][pstates_idx][hidden_state] sample[col] = stats.norm.rvs(loc=mu, scale=sigma, random_state=random_state) if feature_distr == 'lognormal': logmu = self.emission[feature_name]['logmu'][pstates_idx][hidden_state] logsigma = self.emission[feature_name]['logsigma'][pstates_idx][hidden_state] sample[col] = stats.lognorm.rvs(logsigma, loc=0, scale=np.exp(logmu), random_state=random_state) return sample def _init_pstates(self, unique_pstates): # Map events to a unique index. The unknown p-state is at idx 0 self.e = defaultdict(int) self.e.update(dict(zip(np.sort(unique_pstates), range(1, len(unique_pstates) + 1)))) self.er = {v: k for k, v in self.e.items()} # Reverse lookup self.er[0] = UNKNOWN_PSTATE self.n_partial_states = len(self.e.keys()) + 1 # Add one for the unknown state return def _init_pstate_freqs(self, pstates_idx): # Partial state frequencies self.pstate_freq = Counter([idx for seq in pstates_idx for idx in seq]) self.pstate_trans_freq = Counter([(idx1, idx2) for seq in pstates_idx for idx1, idx2 in zip(seq[:-1], seq[1:])]) # Store freqs for the meta state self.pstate_freq[0] = len(np.concatenate(pstates_idx)) for seq in pstates_idx: self.pstate_startprob[seq[0]] += 1 self.pstate_steadyprob += np.bincount(seq, minlength=self.n_partial_states) for idx1, idx2 in zip(seq[:-1], seq[1:]): self.pstate_trans_freq[(0, 0)] += 1 self.pstate_trans_freq[(idx1, 0)] += 1 self.pstate_trans_freq[(0, idx2)] += 1 self.pstate_transmat[idx1, idx2] += 1 # TODO: separate probas from freqs # Normalize to get the probabilities, ignore the meta state at idx 0 self.pstate_startprob[1:] = normalize(self.pstate_startprob[1:]) self.pstate_transmat[1:, 1:] = normalize(self.pstate_transmat[1:, 1:], axis=1) self.pstate_steadyprob[1:] = normalize(self.pstate_steadyprob[1:]) return def _init_from_obs(self, obs, pstates_idx): # Partial state probabilities self.pstate_startprob = np.zeros(self.n_partial_states) self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states)) self.pstate_steadyprob = np.zeros(self.n_partial_states) # obs should be (N*T, n_features) # N is the number of samples # T is the size of each sample obs = np.concatenate(obs) pstates_idx = np.concatenate(pstates_idx) # Initialize starting and transition probas self.startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states self.steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states self.transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) / self.n_hidden_states # Initialize emission parameters # Hidden states are ordered by the first feature feature1 = self.emission_name[0] for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': self.emission[feature_name]['mu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['sigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) for idx in range(1, self.n_partial_states): idx_pstate = (pstates_idx == idx) if not np.any(idx_pstate): idx_pstate = np.arange(len(pstates_idx)) if feature_name == feature1: self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA) self.emission[feature_name]['mu'][idx] = obs[idx_pstate, col].mean() + \ obs[:, col].std() * np.linspace( -self.init_spread, self.init_spread, self.n_hidden_states) else: self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA) self.emission[feature_name]['mu'][idx, :] = obs[idx_pstate, col].mean() if feature_distr == 'lognormal': self.emission[feature_name]['logmu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['logsigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) for idx in range(1, self.n_partial_states): idx_pstate = (pstates_idx == idx) if not np.any(idx_pstate): idx_pstate = np.arange(len(pstates_idx)) if feature_name == feature1: self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(), MIN_PROBA) self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() + np.log( obs[:, col]).std() * np.linspace(-self.init_spread, self.init_spread, self.n_hidden_states) else: self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(), MIN_PROBA) self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() return def _init_random(self, random_state=None): if random_state is None: random_state = self.random_state random_state = check_random_state(random_state) self.pstate_startprob = np.zeros(self.n_partial_states) self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states)) self.pstate_steadyprob = np.zeros(self.n_partial_states) self.pstate_startprob[1:] = gen_stochastic_matrix(size=self.n_partial_states - 1, random_state=random_state) self.pstate_transmat[1:, 1:] = gen_stochastic_matrix( size=(self.n_partial_states - 1, self.n_partial_states - 1), random_state=random_state) self.pstate_steadyprob[1:] = steadystate(self.pstate_transmat[1:, 1:]) self.startprob = gen_stochastic_matrix(size=(self.n_partial_states, self.n_hidden_states), random_state=random_state) transmat = np.zeros((self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)): transmat[i, j] = gen_stochastic_matrix(size=(self.n_hidden_states, self.n_hidden_states), random_state=random_state) self.transmat = normalize(transmat, axis=3) # Initialize emission parameters for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): if feature_distr == 'normal': self.emission[feature_name]['mu'] = random_state.uniform(*_RANDINIT[feature_distr]['mu'], size=( self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['sigma'] = random_state.uniform(*_RANDINIT[feature_distr]['sigma'], size=( self.n_partial_states, self.n_hidden_states)) if feature_distr == 'lognormal': self.emission[feature_name]['logmu'] = random_state.uniform(*_RANDINIT[feature_distr]['logmu'], size=( self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['logsigma'] = random_state.uniform(*_RANDINIT[feature_distr]['logsigma'], size=(self.n_partial_states, self.n_hidden_states)) if self.emission_distr[0] == 'normal': self.emission[self.emission_name[0]]['mu'] = np.sort(self.emission[self.emission_name[0]]['mu'], axis=1) elif self.emission_distr[0] == 'lognormal': self.emission[self.emission_name[0]]['logmu'] = np.sort(self.emission[self.emission_name[0]]['logmu'], axis=1) return def _smooth(self): self._compute_marginals() startprob = self.startprob for j in range(1, self.n_partial_states): if 'freq' == self.smoothing['startprob']: w_ = 1 / (1 + self.pstate_freq[j]) w_j = 1 - w_ elif 'proba' == self.smoothing['startprob']: w_j = self.pstate_steadyprob[j] w_ = 1 - w_j elif 'exp' == self.smoothing['startprob']: w_ = np.exp(-self.pstate_freq[j]) w_j = 1 - w_ elif 'fixed' == self.smoothing['startprob']: w_ = 1 w_j = 0 elif self.smoothing['startprob'] is None: w_ = 0 w_j = 1 else: raise Exception('Wrong smoothing for startprob: ' + self.smoothing['startprob']) startprob[j] = w_j * self.startprob[j] + w_ * self.startprob[0] self.startprob = startprob transmat = self.transmat for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)): if 'freq' == self.smoothing['transmat']: w_i0 = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[0, j]) w_0j = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[i, 0]) w_ij = 1 - (w_i0 + w_0j) w_ = 0 elif 'proba' == self.smoothing['transmat']: denom = self.pstate_transmat[i, j] + self.pstate_transmat[i, :].sum() + self.pstate_transmat[:, j].sum() w_i0 = self.pstate_transmat[i, :].sum() / denom w_0j = self.pstate_transmat[:, j].sum() / denom w_ij = self.pstate_transmat[i, j] / denom w_ = 0 elif 'exp' == self.smoothing['transmat']: w_i0 = self.pstate_trans_freq[i, 0] * np.exp(-self.pstate_trans_freq[i, j]) w_0j = self.pstate_trans_freq[0, j] * np.exp(-self.pstate_trans_freq[i, j]) w_ = self.pstate_trans_freq[0, 0] * np.exp( -(self.pstate_trans_freq[i, 0] + self.pstate_trans_freq[0, j])) w_ij = self.pstate_trans_freq[i, j] w_ij, w_i0, w_0j, w_ = normalize(np.array([w_ij, w_i0, w_0j, w_])) elif 'fixed' == self.smoothing['transmat']: w_i0 = 0 w_0j = 0 w_ = 1 w_ij = 0 elif self.smoothing['transmat'] is None: w_i0 = 0 w_0j = 0 w_ = 0 w_ij = 1 else: raise Exception('Wrong smoothing for transmat: ' + self.smoothing['transmat']) assert (w_i0 + w_0j + w_ + w_ij) - 1 < TOLERANCE transmat[i, j] = w_ij * self.transmat[i, j] + w_i0 * self.transmat[i, 0] + w_0j * self.transmat[0, j] + w_ * \ self.transmat[ 0, 0] self.transmat = transmat assert np.all(self.startprob.sum(axis=1) - 1 < TOLERANCE) assert np.all(self.steadyprob.sum(axis=1) - 1 < TOLERANCE) assert np.all(self.transmat.sum(axis=3) - 1 < TOLERANCE) for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[feature_distr]: key = '%s:%s' % (feature_name, param) for j in range(1, self.n_partial_states): if 'freq' == self.smoothing[key]: w_ = 1 / (1 + self.pstate_freq[j]) w_j = 1 - w_ elif 'proba' == self.smoothing[key]: w_j = self.pstate_steadyprob[j] w_ = 1 - w_j elif 'exp' == self.smoothing[key]: w_ = np.exp(-self.pstate_freq[j]) w_j = 1 - w_ elif 'fixed' == self.smoothing[key]: w_ = 1 w_j = 0 elif self.smoothing[key] is None: w_ = 0 w_j = 1 else: raise Exception('Wrong smoothing for ' + key) self.emission[feature_name][param][j] = w_j * self.emission[feature_name][param][j] + w_ * \ self.emission[ feature_name][ param][0] return def _initialize_sufficient_statistics(self): stats = { 'nobs': 0, 'post': np.zeros((self.n_partial_states, self.n_hidden_states)), 'obs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'obs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'lnobs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'lnobs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'start': np.zeros((self.n_partial_states, self.n_hidden_states)), 'steady': np.zeros((self.n_partial_states, self.n_hidden_states)), 'trans': np.zeros( (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) } return stats def _accumulate_sufficient_statistics(self, stats, obs, pstates_idx, framelogprob, posteriors, fwdlattice, bwdlattice): stats['nobs'] += 1 n_observations, n_hidden_states = framelogprob.shape stats['start'][0] += posteriors[0] for i in range(self.n_partial_states): if len(np.where(pstates_idx == i)[0]) > 0: stats['start'][i] += posteriors[np.where(pstates_idx == i)[0].min()] if n_observations > 1: lneta = np.zeros((n_observations - 1, n_hidden_states, n_hidden_states)) _hmmc._compute_lneta(n_observations, n_hidden_states, pstates_idx, fwdlattice, self._log_transmat, bwdlattice, framelogprob, lneta) for i, j in unique_rows(np.c_[pstates_idx[:-1], pstates_idx[1:]]): if ((pstates_idx[:-1] == i) & (pstates_idx[1:] == j)).sum() > 0: stats['trans'][i, j] += np.exp( logsumexp(lneta[(pstates_idx[:-1] == i) & (pstates_idx[1:] == j)], axis=0)) for i in range(self.n_partial_states): stats['post'][i] += posteriors[pstates_idx == i].sum(axis=0) stats['obs'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i]) stats['obs**2'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i] ** 2) stats['lnobs'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i])) stats['lnobs**2'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i]) ** 2) return def _compute_marginals(self): # TODO: cythonize some of this # Start prob, weighted by p-state start probs self.startprob[0] = (self.pstate_startprob[1:, np.newaxis] * self.startprob[1:]).sum(axis=0) # Use the p-state transmat and transmat to get the full transmat full_transmat = ph2full(self.pstate_transmat[1:, 1:], self.transmat[1:, 1:]) full_steadyprob = steadystate(full_transmat) # Steady state probas are determined by the full trans mat, need to be updated steadyprob = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) steadyprob[0] = full_steadyprob.reshape(-1, self.n_hidden_states).sum(axis=0) for i in range(self.n_partial_states - 1): steadyprob[i + 1] = normalize( full_steadyprob[i * self.n_hidden_states:i * self.n_hidden_states + self.n_hidden_states]) self.steadyprob = steadyprob # Update the transations to/from the marginal state transmat = self.transmat # Group the hidden states within each partial state for hidx1, hidx2 in product(range(self.n_hidden_states), range(self.n_hidden_states)): transmat[0, 0][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, hidx2::self.n_hidden_states].sum() for pidx in range(self.n_partial_states - 1): transmat[pidx + 1, 0][hidx1, hidx2] = full_transmat[pidx * self.n_hidden_states + hidx1, hidx2::self.n_hidden_states].sum() transmat[0, pidx + 1][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, pidx * self.n_hidden_states + hidx2].sum() self.transmat = normalize(transmat, axis=3) pweights = self.pstate_steadyprob[1:, np.newaxis] # Update emission parameters for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): if feature_distr == 'normal': # Marginal state is a mixture of normals mu = self.emission[feature_name]['mu'][1:, :] sigma = self.emission[feature_name]['sigma'][1:, :] # Weighted mean and var mu_0 = (pweights * mu).sum(axis=0) self.emission[feature_name]['mu'][0, :] = mu_0 self.emission[feature_name]['sigma'][0, :] = np.sqrt( (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0)) if feature_distr == 'lognormal': # Marginal state is a mixture of normals mu = self.emission[feature_name]['logmu'][1:, :] sigma = self.emission[feature_name]['logsigma'][1:, :] # Weighted mean and var mu_0 = (pweights * mu).sum(axis=0) self.emission[feature_name]['logmu'][0, :] = mu_0 self.emission[feature_name]['logsigma'][0, :] = np.sqrt( (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0)) return def _do_mstep(self, stats): self.startprob = normalize(np.maximum(stats['start'], MIN_PROBA), axis=1) self.transmat = normalize(np.maximum(stats['trans'], MIN_PROBA), axis=3) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': denom = np.maximum(stats['post'], MIN_PROBA) mu = stats['obs'][:, :, col] / denom cv_num = (stats['obs**2'][:, :, col] - 2 * mu * stats['obs'][:, :, col] + mu ** 2 * denom) sigma = np.sqrt(cv_num / denom) sigma[np.isnan(sigma)] = MIN_PROBA self.emission[feature_name]['mu'] = mu self.emission[feature_name]['sigma'] = np.maximum(sigma, MIN_PROBA) if feature_distr == 'lognormal': denom = np.maximum(stats['post'], MIN_PROBA) mu = stats['lnobs'][:, :, col] / denom cv_num = (stats['lnobs**2'][:, :, col] - 2 * mu * stats['lnobs'][:, :, col] + mu ** 2 * denom) sigma = np.sqrt(cv_num / denom) sigma[np.isnan(sigma)] = MIN_PROBA self.emission[feature_name]['logmu'] = mu self.emission[feature_name]['logsigma'] = np.maximum(sigma, MIN_PROBA) return def _do_forward_pass(self, framelogprob, event_idx): n_observations, n_hidden_states = framelogprob.shape fwdlattice = np.zeros((n_observations, n_hidden_states)) _hmmc._forward(n_observations, n_hidden_states, event_idx, self._log_startprob, self._log_transmat, framelogprob, fwdlattice) return logsumexp(fwdlattice[-1]), fwdlattice def _do_backward_pass(self, framelogprob, event_idx): n_observations, n_hidden_states = framelogprob.shape bwdlattice = np.zeros((n_observations, n_hidden_states)) _hmmc._backward(n_observations, n_hidden_states, event_idx, self._log_startprob, self._log_transmat, framelogprob, bwdlattice) return bwdlattice def _do_viterbi_pass(self, framelogprob, event_idx): n_observations, n_components = framelogprob.shape state_sequence, logprob = _hmmc._viterbi( n_observations, n_components, event_idx, self._log_startprob, self._log_transmat, framelogprob) return logprob, state_sequence def rand(self, unique_pstates, random_state=None): """ Randomize the POHMM parameters """ self._init_pstates(unique_pstates) self._init_random(random_state=random_state) self._compute_marginals() return self def fit(self, obs, pstates, unique_pstates=None): """ Estimate model parameters. """ obs = [np.array(o) for o in obs] pstates = [np.array(p) for p in pstates] # List or array of observation sequences assert len(obs) == len(pstates) assert obs[0].ndim == 2 assert pstates[0].ndim == 1 if unique_pstates is not None: self._init_pstates(unique_pstates) else: self._init_pstates(list(set(np.concatenate(pstates)))) # Map the partial states to a unique index pstates_idx = [np.array([self.e[p] for p in seq]) for seq in pstates] if self.init_method == 'rand': self._init_random() elif self.init_method == 'obs': self._init_from_obs(obs, pstates_idx) self._init_pstate_freqs(pstates_idx) self._smooth() logprob = [] for i in range(self.max_iter): # Expectation step stats = self._initialize_sufficient_statistics() curr_logprob = 0 for obs_i, pstates_idx_i in zip(obs, pstates_idx): framelogprob = self._compute_log_likelihood(obs_i, pstates_idx_i) lpr, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx_i) bwdlattice = self._do_backward_pass(framelogprob, pstates_idx_i) gamma = fwdlattice + bwdlattice posteriors = np.exp(gamma.T - logsumexp(gamma, axis=1)).T curr_logprob += lpr self._accumulate_sufficient_statistics(stats, obs_i, pstates_idx_i, framelogprob, posteriors, fwdlattice, bwdlattice) logprob.append(curr_logprob) self.logprob_ = curr_logprob # Check for convergence. self.n_iter_performed_ = i if i > 0: self.logprob_delta = logprob[-1] - logprob[-2] if self.logprob_delta < self.thresh: break # Maximization step self._do_mstep(stats) # Mix the parameters self._smooth() return self def score_events(self, obs, pstates): """ Compute the log probability of each event under the model. """ pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) _, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx) L = logsumexp(fwdlattice, axis=1) return np.concatenate([L[[0]], np.diff(L)]) def predict_states(self, obs, pstates): pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) viterbi_logprob, state_sequence = self._do_viterbi_pass(framelogprob, pstates_idx) return viterbi_logprob, state_sequence def predict(self, obs, pstates, next_pstate=None): """ Predict the next observation """ assert len(obs) == len(pstates) pstates_idx = np.array([self.e[ei] for ei in pstates]) next_pstate_idx = self.e[next_pstate] if len(obs) == 0: # No history, use the starting probas next_hstate_prob = self.startprob[next_pstate_idx] else: # With a history, determine the hidden state posteriors using # the last posteriors and transition matrix framelogprob = self._compute_log_likelihood(obs, pstates_idx) _, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx) next_hstate_prob = np.zeros(self.n_hidden_states) alpha_n = fwdlattice[-1] vmax = alpha_n.max(axis=0) alpha_n = np.exp(alpha_n - vmax) alpha_n = alpha_n / alpha_n.sum() trans = self.transmat[pstates_idx[-1], next_pstate_idx] for i in range(self.n_hidden_states): next_hstate_prob[i] = np.sum([alpha_n[j] * trans[j, i] for j in range(self.n_hidden_states)]) assert next_hstate_prob.sum() - 1 < TOLERANCE # Make the prediction prediction = np.array( [self.expected_value(feature, pstate=next_pstate, hstate_prob=next_hstate_prob) for feature in self.emission_name]) # next_hstate = np.argmax(next_hstate_prob) # prediction = np.array( # [self.expected_value(feature, pstate=next_pstate, hstate=next_hstate) for feature in # self.emission_name]) return prediction def gen_pstates_idx(self, n, random_state=None): if random_state is None: random_state = self.random_state random_state = check_random_state(random_state) startprob_cdf = np.cumsum(self.pstate_startprob) transmat_cdf = np.cumsum(self.pstate_transmat, 1) # Initial state. rand = random_state.rand() curr_pstate = (startprob_cdf > rand).argmax() pstates = [curr_pstate] for _ in range(1, n): rand = random_state.rand() curr_pstate = (transmat_cdf[curr_pstate] > rand).argmax() pstates.append(curr_pstate) return np.array(pstates, dtype=int) def sample(self, pstates=None, n_obs=None, random_state=None): """ """ random_state = check_random_state(random_state) if pstates is None and n_obs is None: raise Exception('Must provide either pstates or n_obs') if pstates is not None and n_obs is not None: raise Exception('Must provide either pstates or n_obs but not both') gen_pstates = False rand = random_state.rand() if pstates is None: gen_pstates = True pstartprob_cdf = np.cumsum(self.pstate_startprob) ptransmat_cdf = np.cumsum(self.pstate_transmat, 1) # Initial pstate currpstate = (pstartprob_cdf > rand).argmax() pstates_idx = [currpstate] pstates = [self.er[currpstate]] else: n_obs = len(pstates) pstates_idx = np.array([self.e[p] for p in pstates]) startprob_pdf = self.startprob[pstates_idx[0]] startprob_cdf = np.cumsum(startprob_pdf) transmat_cdf = np.cumsum(self.transmat[0, pstates_idx[0]], 1) # Initial hidden state rand = random_state.rand() currstate = (startprob_cdf > rand).argmax() hidden_states = [currstate] obs = [self._generate_sample_from_state(currstate, pstates_idx[0], random_state)] for i in range(1, n_obs): rand = random_state.rand() if gen_pstates: currpstate = (ptransmat_cdf[currpstate] > rand).argmax() pstates_idx.append(currpstate) pstates.append(self.er[currpstate]) transmat_cdf = np.cumsum(self.transmat[pstates_idx[i - 1], pstates_idx[i]], 1) rand = random_state.rand() currstate = (transmat_cdf[currstate] > rand).argmax() hidden_states.append(currstate) obs.append(self._generate_sample_from_state(currstate, pstates_idx[i], random_state)) return np.array(obs), np.array(pstates), np.array(hidden_states, dtype=int) def fit_df(self, dfs, pstate_col=PSTATE_COL): """ Convenience function to fit a model from a list of dataframes """ obs_cols = list(self.emission_name) obs = [df[df.columns.difference([pstate_col])][obs_cols].values for df in dfs] pstates = [df[pstate_col].values for df in dfs] return self.fit(obs, pstates) def score_df(self, df, pstate_col=PSTATE_COL): """ """ obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values return self.score(obs, pstates) def score_events_df(self, df, pstate_col=PSTATE_COL, score_col='score'): """ """ df = df.copy() obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values df[score_col] = self.score_events(obs, pstates) return df def predict_states_df(self, df, pstate_col=PSTATE_COL, hstate_col=HSTATE_COL): df = df.copy() obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values _, df[hstate_col] = self.predict_states(obs, pstates) return df def predict_df(self, df, next_pstate=None, pstate_col=PSTATE_COL): obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values return self.predict(obs, pstates, next_pstate) def sample_df(self, pstates=None, n_obs=None, random_state=None, pstate_col=PSTATE_COL, hstate_col=HSTATE_COL): """ Convenience function to generate samples a model and create a dataframe """ try: import pandas as pd except Exception as e: raise e obs, pstates, hstates = self.sample(pstates, n_obs, random_state) items = [] if pstate_col is not None: items.append((pstate_col, pstates)) if hstate_col is not None: items.append((hstate_col, hstates)) items = items + [(self.emission_name[i], obs[:, i]) for i in range(self.n_features)] df = pd.DataFrame.from_items(items) return df def __str__(self): pstates = sorted(self.e.keys()) sep = '-' * 80 + '\n' sep2 = '_' * 40 + '\n' out = 'POHMM\n' out += 'H-states: %d\n' % self.n_hidden_states out += 'P-states: (%d) %s\n' % (len(pstates), str(pstates)) out += 'Emission: %s\n' % (self.emission_distr) out += sep out += 'Transition probas\n' out += sep2 out += '. -> .\n%s\n' % str(self.transmat[0, 0]) for pstate in pstates: out += sep2 out += '%s -> .\n%s\n' % (pstate, str(self.transmat[self.e[pstate], 0])) out += sep2 out += '. -> %s\n%s\n' % (pstate, str(self.transmat[0, self.e[pstate]])) for pstate1, pstate2 in product(pstates, pstates): out += sep2 out += '%s -> %s\n%s\n' % (pstate1, pstate2, str(self.transmat[self.e[pstate1], self.e[pstate2]])) out += sep out += 'Starting probas\n' out += '.: %s\n' % str(self.startprob[0]) for pstate in pstates: out += '%s: %s\n' % (pstate, str(self.startprob[self.e[pstate]])) out += sep out += 'Steady probas\n' out += '.: %s\n' % str(self.steadyprob[0]) for pstate in pstates: out += '%s: %s\n' % (pstate, str(self.steadyprob[self.e[pstate]])) out += sep out += 'Emissions\n' for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): out += sep2 out += 'Feature %s: %s\n' % (feature_name, feature_distr) for param in _DISTRIBUTIONS[feature_distr]: out += '.: %s = %s\n' % (param, str(self.emission[feature_name][param][0])) for pstate in pstates: out += '%s: %s = %s\n' % (pstate, param, str(self.emission[feature_name][param][self.e[pstate]])) out += sep return out def expected_value(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): """ Determine the joint maximum likelihood estimate """ # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': return np.sum(w * expected_lognormal(self.emission[feature]['logsigma'].flatten(), self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': return np.sum(w * expected_normal(self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten())) elif hstate is None: # Marginalize hstate if self.emission_name_distr[feature] == 'lognormal': return np.sum(hstate_prob * expected_lognormal(self.emission[feature]['logsigma'][pstate_idx, :], self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': return np.sum(hstate_prob * expected_normal(self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :])) elif pstate is None: # Marginalize pstate if self.emission_name_distr[feature] == 'lognormal': return np.sum(pstate_prob * expected_lognormal(self.emission[feature]['logsigma'][:, hstate], self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': return np.sum(pstate_prob * expected_normal(self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate])) else: if self.emission_name_distr[feature] == 'lognormal': return expected_lognormal(self.emission[feature]['logsigma'][pstate_idx, hstate], self.emission[feature]['logmu'][pstate_idx, hstate]) elif self.emission_name_distr[feature] == 'normal': return expected_normal(self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) return def pdf_fn(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'].flatten(), 0, np.exp(self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten()) elif hstate is None: # Marginalize hstate w = hstate_prob if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][pstate_idx, :], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :]) elif pstate is None: # Marginalize pstate w = pstate_prob if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][:, hstate], 0, np.exp(self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate]) else: w = 1 if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][pstate_idx, hstate], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, hstate])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) def fn(x): if np.isscalar(x): p = np.sum(w * pdf(x)) else: x = np.array(x) p = np.zeros(len(x)) for i, xi in enumerate(x): p[i] = np.sum(w * pdf(xi)) return p return fn def cdf_fn(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'].flatten(), 0, np.exp(self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten()) elif hstate is None: # Marginalize hstate w = hstate_prob if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][pstate_idx, :], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :]) elif pstate is None: # Marginalize pstate w = pstate_prob if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][:, hstate], 0, np.exp(self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate]) else: w = 1 if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][pstate_idx, hstate], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, hstate])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) def fn(x): if np.isscalar(x): p = np.sum(w * cdf(x)) else: x = np.array(x) p = np.zeros(len(x)) for i, xi in enumerate(x): p[i] = np.sum(w * cdf(xi)) return p return fn def params(self, pstates=None): if pstates is None: pstates = [None] + sorted( set(self.er.values())) # TODO: self.e caches any unknown value, maybe it shouldn't? params = [] # emission parameters for hstate, pstate_label in product(range(self.n_hidden_states), pstates): for feature, distr in zip(self.emission_name, self.emission_distr): for feature_param in _DISTRIBUTIONS[distr]: params.append(self.emission[feature][feature_param][self.e[pstate_label], hstate]) # transition parameters, diagonals only assuming 2 state for hstate, pstate_label in product(range(self.n_hidden_states), pstates): params.append(self.transmat[self.e[pstate_label], self.e[pstate_label], hstate, hstate]) return np.array(params)
vmonaco/pohmm
pohmm/pohmm.py
Pohmm.score_events
python
def score_events(self, obs, pstates): pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) _, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx) L = logsumexp(fwdlattice, axis=1) return np.concatenate([L[[0]], np.diff(L)])
Compute the log probability of each event under the model.
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/pohmm.py#L742-L751
[ "def logsumexp(arr, axis=0):\n \"\"\"\n Computes the sum of arr assuming arr is in the log domain.\n\n Returns log(sum(exp(arr))) while minimizing the possibility of\n over/underflow.\n \"\"\"\n arr = np.rollaxis(arr, axis)\n # Use the max to normalize, as with the log this is what accumulates\n # the less errors\n vmax = arr.max(axis=0)\n out = np.log(np.sum(np.exp(arr - vmax), axis=0))\n out += vmax\n return out\n", "def _compute_log_likelihood(self, obs, pstates_idx):\n\n q = np.zeros(shape=(len(obs), self.n_hidden_states, self.n_features))\n\n for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)):\n if feature_distr == 'normal':\n mu = self.emission[feature_name]['mu'][pstates_idx]\n sigma = self.emission[feature_name]['sigma'][pstates_idx]\n for j in range(self.n_hidden_states):\n q[:, j, col] = np.log(\n np.maximum(MIN_PROBA, stats.norm.pdf(obs[:, col], loc=mu[:, j], scale=sigma[:, j])))\n if feature_distr == 'lognormal':\n logmu = self.emission[feature_name]['logmu'][pstates_idx]\n logsigma = self.emission[feature_name]['logsigma'][pstates_idx]\n for j in range(self.n_hidden_states):\n q[:, j, col] = np.log(np.maximum(MIN_PROBA,\n stats.lognorm.pdf(obs[:, col], logsigma[:, j], loc=0,\n scale=np.exp(logmu[:, j]))))\n\n q = q.sum(axis=2)\n return q\n", "def _do_forward_pass(self, framelogprob, event_idx):\n n_observations, n_hidden_states = framelogprob.shape\n fwdlattice = np.zeros((n_observations, n_hidden_states))\n _hmmc._forward(n_observations, n_hidden_states,\n event_idx, self._log_startprob,\n self._log_transmat, framelogprob, fwdlattice)\n\n return logsumexp(fwdlattice[-1]), fwdlattice\n" ]
class Pohmm(object): """ Partially observable hidden Markov model """ def __init__(self, n_hidden_states=2, emissions=['normal'], max_iter=1000, thresh=1e-6, init_method='obs', init_spread=2, smoothing=None, random_state=None): if type(n_hidden_states) is int: self.n_hidden_states = n_hidden_states else: raise Exception('Wrong type for n_hidden_states. Must be int') if type(emissions[0]) is tuple: emission_name, emission_distr = zip(*emissions) elif type(emissions[0]) is str: emission_name, emission_distr = np.arange(len(emissions)), emissions for distr in emission_distr: if distr not in _DISTRIBUTIONS.keys(): raise ValueError('Emission distribution must be one of', _DISTRIBUTIONS.keys()) self.emission_name = emission_name self.emission_distr = emission_distr self.emission_name_distr = dict(zip(emission_name, emission_distr)) self.n_features = len(emissions) # Set up the emission parameters # emission: {'feature':{'param': np.array(shape=(n_partial_states, n_hidden_states))}} self.emission = defaultdict(dict) for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: self.emission[name][param] = None self.emission = dict(self.emission) assert max_iter >= 0 assert thresh >= 0 self.max_iter = max_iter self.thresh = thresh assert init_spread >= 0 self.init_spread = init_spread if init_method not in _INIT_METHODS: raise ValueError('init_method must be one of', _INIT_METHODS) self.init_method = init_method if smoothing is None: smoothing = {'transmat': None, 'startprob': None} for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: smoothing['%s:%s' % (name, param)] = None elif type(smoothing) is str: s = smoothing smoothing = {'transmat': s, 'startprob': s} for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: smoothing['%s:%s' % (name, param)] = s elif type(smoothing) is dict: assert 'transmat' in smoothing.keys() assert 'startprob' in smoothing.keys() for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: assert param in smoothing.keys() or '%s:%s' % (name, param) in smoothing.keys() if param in smoothing.keys() and '%s:%s' % (name, param) not in smoothing.keys(): smoothing['%s:%s' % (name, param)] = smoothing[param] else: raise Exception('Wrong type for smoothing. Must be None, str, or dict') self.smoothing = smoothing self.random_state = random_state # Number of unique partial states is unknown until fit self.n_partial_states = None # Results after fitting the model self.logprob_ = None self.n_iter_performed_ = None self.logprob_delta_ = None # Mapping between p-states and a unique index # Defaults to 0 for unknown or missing p-states self.e = defaultdict(int) def _get_startprob(self): return np.exp(self._log_startprob) def _set_startprob(self, startprob): if startprob is None: startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states else: startprob = np.asarray(startprob, dtype=np.float) # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(startprob): startprob = normalize(startprob, axis=1) if len(startprob) != self.n_partial_states: raise ValueError('startprob must have length n_partial_states') if not np.allclose(np.sum(startprob, axis=1), 1.0): raise ValueError('startprob must sum to 1.0') self._log_startprob = np.log(np.asarray(startprob).copy()) startprob = property(_get_startprob, _set_startprob) def _get_steadyprob(self): return np.exp(self._log_steadyprob) def _set_steadyprob(self, steadyprob): if steadyprob is None: steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states else: steadyprob = np.asarray(steadyprob, dtype=np.float) # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(steadyprob): steadyprob = normalize(steadyprob, axis=1) if len(steadyprob) != self.n_partial_states: raise ValueError('steadyprob must have length n_partial_states') if not np.allclose(np.sum(steadyprob, axis=1), 1.0): raise ValueError('steadyprob must sum to 1.0') self._log_steadyprob = np.log(np.asarray(steadyprob).copy()) steadyprob = property(_get_steadyprob, _set_steadyprob) def _get_transmat(self): return np.exp(self._log_transmat) def _set_transmat(self, transmat): if transmat is None: transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) / self.n_hidden_states # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(transmat): transmat = normalize(transmat, axis=3) if (np.asarray(transmat).shape != (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)): raise ValueError('transmat must have shape ' '(n_partial_states,n_partial_states,n_hidden_states,n_hidden_states)') if not np.all(np.allclose(np.sum(transmat, axis=3), 1.0)): raise ValueError('Rows of transmat must sum to 1.0') self._log_transmat = np.log(np.asarray(transmat).copy()) underflow_idx = np.isnan(self._log_transmat) self._log_transmat[underflow_idx] = NEGINF transmat = property(_get_transmat, _set_transmat) def _compute_log_likelihood(self, obs, pstates_idx): q = np.zeros(shape=(len(obs), self.n_hidden_states, self.n_features)) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': mu = self.emission[feature_name]['mu'][pstates_idx] sigma = self.emission[feature_name]['sigma'][pstates_idx] for j in range(self.n_hidden_states): q[:, j, col] = np.log( np.maximum(MIN_PROBA, stats.norm.pdf(obs[:, col], loc=mu[:, j], scale=sigma[:, j]))) if feature_distr == 'lognormal': logmu = self.emission[feature_name]['logmu'][pstates_idx] logsigma = self.emission[feature_name]['logsigma'][pstates_idx] for j in range(self.n_hidden_states): q[:, j, col] = np.log(np.maximum(MIN_PROBA, stats.lognorm.pdf(obs[:, col], logsigma[:, j], loc=0, scale=np.exp(logmu[:, j])))) q = q.sum(axis=2) return q def _generate_sample_from_state(self, hidden_state, pstates_idx, random_state=None): sample = np.zeros(self.n_features) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': mu = self.emission[feature_name]['mu'][pstates_idx][hidden_state] sigma = self.emission[feature_name]['sigma'][pstates_idx][hidden_state] sample[col] = stats.norm.rvs(loc=mu, scale=sigma, random_state=random_state) if feature_distr == 'lognormal': logmu = self.emission[feature_name]['logmu'][pstates_idx][hidden_state] logsigma = self.emission[feature_name]['logsigma'][pstates_idx][hidden_state] sample[col] = stats.lognorm.rvs(logsigma, loc=0, scale=np.exp(logmu), random_state=random_state) return sample def _init_pstates(self, unique_pstates): # Map events to a unique index. The unknown p-state is at idx 0 self.e = defaultdict(int) self.e.update(dict(zip(np.sort(unique_pstates), range(1, len(unique_pstates) + 1)))) self.er = {v: k for k, v in self.e.items()} # Reverse lookup self.er[0] = UNKNOWN_PSTATE self.n_partial_states = len(self.e.keys()) + 1 # Add one for the unknown state return def _init_pstate_freqs(self, pstates_idx): # Partial state frequencies self.pstate_freq = Counter([idx for seq in pstates_idx for idx in seq]) self.pstate_trans_freq = Counter([(idx1, idx2) for seq in pstates_idx for idx1, idx2 in zip(seq[:-1], seq[1:])]) # Store freqs for the meta state self.pstate_freq[0] = len(np.concatenate(pstates_idx)) for seq in pstates_idx: self.pstate_startprob[seq[0]] += 1 self.pstate_steadyprob += np.bincount(seq, minlength=self.n_partial_states) for idx1, idx2 in zip(seq[:-1], seq[1:]): self.pstate_trans_freq[(0, 0)] += 1 self.pstate_trans_freq[(idx1, 0)] += 1 self.pstate_trans_freq[(0, idx2)] += 1 self.pstate_transmat[idx1, idx2] += 1 # TODO: separate probas from freqs # Normalize to get the probabilities, ignore the meta state at idx 0 self.pstate_startprob[1:] = normalize(self.pstate_startprob[1:]) self.pstate_transmat[1:, 1:] = normalize(self.pstate_transmat[1:, 1:], axis=1) self.pstate_steadyprob[1:] = normalize(self.pstate_steadyprob[1:]) return def _init_from_obs(self, obs, pstates_idx): # Partial state probabilities self.pstate_startprob = np.zeros(self.n_partial_states) self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states)) self.pstate_steadyprob = np.zeros(self.n_partial_states) # obs should be (N*T, n_features) # N is the number of samples # T is the size of each sample obs = np.concatenate(obs) pstates_idx = np.concatenate(pstates_idx) # Initialize starting and transition probas self.startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states self.steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states self.transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) / self.n_hidden_states # Initialize emission parameters # Hidden states are ordered by the first feature feature1 = self.emission_name[0] for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': self.emission[feature_name]['mu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['sigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) for idx in range(1, self.n_partial_states): idx_pstate = (pstates_idx == idx) if not np.any(idx_pstate): idx_pstate = np.arange(len(pstates_idx)) if feature_name == feature1: self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA) self.emission[feature_name]['mu'][idx] = obs[idx_pstate, col].mean() + \ obs[:, col].std() * np.linspace( -self.init_spread, self.init_spread, self.n_hidden_states) else: self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA) self.emission[feature_name]['mu'][idx, :] = obs[idx_pstate, col].mean() if feature_distr == 'lognormal': self.emission[feature_name]['logmu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['logsigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) for idx in range(1, self.n_partial_states): idx_pstate = (pstates_idx == idx) if not np.any(idx_pstate): idx_pstate = np.arange(len(pstates_idx)) if feature_name == feature1: self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(), MIN_PROBA) self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() + np.log( obs[:, col]).std() * np.linspace(-self.init_spread, self.init_spread, self.n_hidden_states) else: self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(), MIN_PROBA) self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() return def _init_random(self, random_state=None): if random_state is None: random_state = self.random_state random_state = check_random_state(random_state) self.pstate_startprob = np.zeros(self.n_partial_states) self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states)) self.pstate_steadyprob = np.zeros(self.n_partial_states) self.pstate_startprob[1:] = gen_stochastic_matrix(size=self.n_partial_states - 1, random_state=random_state) self.pstate_transmat[1:, 1:] = gen_stochastic_matrix( size=(self.n_partial_states - 1, self.n_partial_states - 1), random_state=random_state) self.pstate_steadyprob[1:] = steadystate(self.pstate_transmat[1:, 1:]) self.startprob = gen_stochastic_matrix(size=(self.n_partial_states, self.n_hidden_states), random_state=random_state) transmat = np.zeros((self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)): transmat[i, j] = gen_stochastic_matrix(size=(self.n_hidden_states, self.n_hidden_states), random_state=random_state) self.transmat = normalize(transmat, axis=3) # Initialize emission parameters for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): if feature_distr == 'normal': self.emission[feature_name]['mu'] = random_state.uniform(*_RANDINIT[feature_distr]['mu'], size=( self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['sigma'] = random_state.uniform(*_RANDINIT[feature_distr]['sigma'], size=( self.n_partial_states, self.n_hidden_states)) if feature_distr == 'lognormal': self.emission[feature_name]['logmu'] = random_state.uniform(*_RANDINIT[feature_distr]['logmu'], size=( self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['logsigma'] = random_state.uniform(*_RANDINIT[feature_distr]['logsigma'], size=(self.n_partial_states, self.n_hidden_states)) if self.emission_distr[0] == 'normal': self.emission[self.emission_name[0]]['mu'] = np.sort(self.emission[self.emission_name[0]]['mu'], axis=1) elif self.emission_distr[0] == 'lognormal': self.emission[self.emission_name[0]]['logmu'] = np.sort(self.emission[self.emission_name[0]]['logmu'], axis=1) return def _smooth(self): self._compute_marginals() startprob = self.startprob for j in range(1, self.n_partial_states): if 'freq' == self.smoothing['startprob']: w_ = 1 / (1 + self.pstate_freq[j]) w_j = 1 - w_ elif 'proba' == self.smoothing['startprob']: w_j = self.pstate_steadyprob[j] w_ = 1 - w_j elif 'exp' == self.smoothing['startprob']: w_ = np.exp(-self.pstate_freq[j]) w_j = 1 - w_ elif 'fixed' == self.smoothing['startprob']: w_ = 1 w_j = 0 elif self.smoothing['startprob'] is None: w_ = 0 w_j = 1 else: raise Exception('Wrong smoothing for startprob: ' + self.smoothing['startprob']) startprob[j] = w_j * self.startprob[j] + w_ * self.startprob[0] self.startprob = startprob transmat = self.transmat for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)): if 'freq' == self.smoothing['transmat']: w_i0 = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[0, j]) w_0j = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[i, 0]) w_ij = 1 - (w_i0 + w_0j) w_ = 0 elif 'proba' == self.smoothing['transmat']: denom = self.pstate_transmat[i, j] + self.pstate_transmat[i, :].sum() + self.pstate_transmat[:, j].sum() w_i0 = self.pstate_transmat[i, :].sum() / denom w_0j = self.pstate_transmat[:, j].sum() / denom w_ij = self.pstate_transmat[i, j] / denom w_ = 0 elif 'exp' == self.smoothing['transmat']: w_i0 = self.pstate_trans_freq[i, 0] * np.exp(-self.pstate_trans_freq[i, j]) w_0j = self.pstate_trans_freq[0, j] * np.exp(-self.pstate_trans_freq[i, j]) w_ = self.pstate_trans_freq[0, 0] * np.exp( -(self.pstate_trans_freq[i, 0] + self.pstate_trans_freq[0, j])) w_ij = self.pstate_trans_freq[i, j] w_ij, w_i0, w_0j, w_ = normalize(np.array([w_ij, w_i0, w_0j, w_])) elif 'fixed' == self.smoothing['transmat']: w_i0 = 0 w_0j = 0 w_ = 1 w_ij = 0 elif self.smoothing['transmat'] is None: w_i0 = 0 w_0j = 0 w_ = 0 w_ij = 1 else: raise Exception('Wrong smoothing for transmat: ' + self.smoothing['transmat']) assert (w_i0 + w_0j + w_ + w_ij) - 1 < TOLERANCE transmat[i, j] = w_ij * self.transmat[i, j] + w_i0 * self.transmat[i, 0] + w_0j * self.transmat[0, j] + w_ * \ self.transmat[ 0, 0] self.transmat = transmat assert np.all(self.startprob.sum(axis=1) - 1 < TOLERANCE) assert np.all(self.steadyprob.sum(axis=1) - 1 < TOLERANCE) assert np.all(self.transmat.sum(axis=3) - 1 < TOLERANCE) for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[feature_distr]: key = '%s:%s' % (feature_name, param) for j in range(1, self.n_partial_states): if 'freq' == self.smoothing[key]: w_ = 1 / (1 + self.pstate_freq[j]) w_j = 1 - w_ elif 'proba' == self.smoothing[key]: w_j = self.pstate_steadyprob[j] w_ = 1 - w_j elif 'exp' == self.smoothing[key]: w_ = np.exp(-self.pstate_freq[j]) w_j = 1 - w_ elif 'fixed' == self.smoothing[key]: w_ = 1 w_j = 0 elif self.smoothing[key] is None: w_ = 0 w_j = 1 else: raise Exception('Wrong smoothing for ' + key) self.emission[feature_name][param][j] = w_j * self.emission[feature_name][param][j] + w_ * \ self.emission[ feature_name][ param][0] return def _initialize_sufficient_statistics(self): stats = { 'nobs': 0, 'post': np.zeros((self.n_partial_states, self.n_hidden_states)), 'obs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'obs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'lnobs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'lnobs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'start': np.zeros((self.n_partial_states, self.n_hidden_states)), 'steady': np.zeros((self.n_partial_states, self.n_hidden_states)), 'trans': np.zeros( (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) } return stats def _accumulate_sufficient_statistics(self, stats, obs, pstates_idx, framelogprob, posteriors, fwdlattice, bwdlattice): stats['nobs'] += 1 n_observations, n_hidden_states = framelogprob.shape stats['start'][0] += posteriors[0] for i in range(self.n_partial_states): if len(np.where(pstates_idx == i)[0]) > 0: stats['start'][i] += posteriors[np.where(pstates_idx == i)[0].min()] if n_observations > 1: lneta = np.zeros((n_observations - 1, n_hidden_states, n_hidden_states)) _hmmc._compute_lneta(n_observations, n_hidden_states, pstates_idx, fwdlattice, self._log_transmat, bwdlattice, framelogprob, lneta) for i, j in unique_rows(np.c_[pstates_idx[:-1], pstates_idx[1:]]): if ((pstates_idx[:-1] == i) & (pstates_idx[1:] == j)).sum() > 0: stats['trans'][i, j] += np.exp( logsumexp(lneta[(pstates_idx[:-1] == i) & (pstates_idx[1:] == j)], axis=0)) for i in range(self.n_partial_states): stats['post'][i] += posteriors[pstates_idx == i].sum(axis=0) stats['obs'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i]) stats['obs**2'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i] ** 2) stats['lnobs'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i])) stats['lnobs**2'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i]) ** 2) return def _compute_marginals(self): # TODO: cythonize some of this # Start prob, weighted by p-state start probs self.startprob[0] = (self.pstate_startprob[1:, np.newaxis] * self.startprob[1:]).sum(axis=0) # Use the p-state transmat and transmat to get the full transmat full_transmat = ph2full(self.pstate_transmat[1:, 1:], self.transmat[1:, 1:]) full_steadyprob = steadystate(full_transmat) # Steady state probas are determined by the full trans mat, need to be updated steadyprob = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) steadyprob[0] = full_steadyprob.reshape(-1, self.n_hidden_states).sum(axis=0) for i in range(self.n_partial_states - 1): steadyprob[i + 1] = normalize( full_steadyprob[i * self.n_hidden_states:i * self.n_hidden_states + self.n_hidden_states]) self.steadyprob = steadyprob # Update the transations to/from the marginal state transmat = self.transmat # Group the hidden states within each partial state for hidx1, hidx2 in product(range(self.n_hidden_states), range(self.n_hidden_states)): transmat[0, 0][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, hidx2::self.n_hidden_states].sum() for pidx in range(self.n_partial_states - 1): transmat[pidx + 1, 0][hidx1, hidx2] = full_transmat[pidx * self.n_hidden_states + hidx1, hidx2::self.n_hidden_states].sum() transmat[0, pidx + 1][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, pidx * self.n_hidden_states + hidx2].sum() self.transmat = normalize(transmat, axis=3) pweights = self.pstate_steadyprob[1:, np.newaxis] # Update emission parameters for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): if feature_distr == 'normal': # Marginal state is a mixture of normals mu = self.emission[feature_name]['mu'][1:, :] sigma = self.emission[feature_name]['sigma'][1:, :] # Weighted mean and var mu_0 = (pweights * mu).sum(axis=0) self.emission[feature_name]['mu'][0, :] = mu_0 self.emission[feature_name]['sigma'][0, :] = np.sqrt( (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0)) if feature_distr == 'lognormal': # Marginal state is a mixture of normals mu = self.emission[feature_name]['logmu'][1:, :] sigma = self.emission[feature_name]['logsigma'][1:, :] # Weighted mean and var mu_0 = (pweights * mu).sum(axis=0) self.emission[feature_name]['logmu'][0, :] = mu_0 self.emission[feature_name]['logsigma'][0, :] = np.sqrt( (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0)) return def _do_mstep(self, stats): self.startprob = normalize(np.maximum(stats['start'], MIN_PROBA), axis=1) self.transmat = normalize(np.maximum(stats['trans'], MIN_PROBA), axis=3) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': denom = np.maximum(stats['post'], MIN_PROBA) mu = stats['obs'][:, :, col] / denom cv_num = (stats['obs**2'][:, :, col] - 2 * mu * stats['obs'][:, :, col] + mu ** 2 * denom) sigma = np.sqrt(cv_num / denom) sigma[np.isnan(sigma)] = MIN_PROBA self.emission[feature_name]['mu'] = mu self.emission[feature_name]['sigma'] = np.maximum(sigma, MIN_PROBA) if feature_distr == 'lognormal': denom = np.maximum(stats['post'], MIN_PROBA) mu = stats['lnobs'][:, :, col] / denom cv_num = (stats['lnobs**2'][:, :, col] - 2 * mu * stats['lnobs'][:, :, col] + mu ** 2 * denom) sigma = np.sqrt(cv_num / denom) sigma[np.isnan(sigma)] = MIN_PROBA self.emission[feature_name]['logmu'] = mu self.emission[feature_name]['logsigma'] = np.maximum(sigma, MIN_PROBA) return def _do_forward_pass(self, framelogprob, event_idx): n_observations, n_hidden_states = framelogprob.shape fwdlattice = np.zeros((n_observations, n_hidden_states)) _hmmc._forward(n_observations, n_hidden_states, event_idx, self._log_startprob, self._log_transmat, framelogprob, fwdlattice) return logsumexp(fwdlattice[-1]), fwdlattice def _do_backward_pass(self, framelogprob, event_idx): n_observations, n_hidden_states = framelogprob.shape bwdlattice = np.zeros((n_observations, n_hidden_states)) _hmmc._backward(n_observations, n_hidden_states, event_idx, self._log_startprob, self._log_transmat, framelogprob, bwdlattice) return bwdlattice def _do_viterbi_pass(self, framelogprob, event_idx): n_observations, n_components = framelogprob.shape state_sequence, logprob = _hmmc._viterbi( n_observations, n_components, event_idx, self._log_startprob, self._log_transmat, framelogprob) return logprob, state_sequence def rand(self, unique_pstates, random_state=None): """ Randomize the POHMM parameters """ self._init_pstates(unique_pstates) self._init_random(random_state=random_state) self._compute_marginals() return self def fit(self, obs, pstates, unique_pstates=None): """ Estimate model parameters. """ obs = [np.array(o) for o in obs] pstates = [np.array(p) for p in pstates] # List or array of observation sequences assert len(obs) == len(pstates) assert obs[0].ndim == 2 assert pstates[0].ndim == 1 if unique_pstates is not None: self._init_pstates(unique_pstates) else: self._init_pstates(list(set(np.concatenate(pstates)))) # Map the partial states to a unique index pstates_idx = [np.array([self.e[p] for p in seq]) for seq in pstates] if self.init_method == 'rand': self._init_random() elif self.init_method == 'obs': self._init_from_obs(obs, pstates_idx) self._init_pstate_freqs(pstates_idx) self._smooth() logprob = [] for i in range(self.max_iter): # Expectation step stats = self._initialize_sufficient_statistics() curr_logprob = 0 for obs_i, pstates_idx_i in zip(obs, pstates_idx): framelogprob = self._compute_log_likelihood(obs_i, pstates_idx_i) lpr, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx_i) bwdlattice = self._do_backward_pass(framelogprob, pstates_idx_i) gamma = fwdlattice + bwdlattice posteriors = np.exp(gamma.T - logsumexp(gamma, axis=1)).T curr_logprob += lpr self._accumulate_sufficient_statistics(stats, obs_i, pstates_idx_i, framelogprob, posteriors, fwdlattice, bwdlattice) logprob.append(curr_logprob) self.logprob_ = curr_logprob # Check for convergence. self.n_iter_performed_ = i if i > 0: self.logprob_delta = logprob[-1] - logprob[-2] if self.logprob_delta < self.thresh: break # Maximization step self._do_mstep(stats) # Mix the parameters self._smooth() return self def score(self, obs, pstates): """ Compute the log probability under the model. """ pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) logprob, _ = self._do_forward_pass(framelogprob, pstates_idx) return logprob def predict_states(self, obs, pstates): pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) viterbi_logprob, state_sequence = self._do_viterbi_pass(framelogprob, pstates_idx) return viterbi_logprob, state_sequence def predict(self, obs, pstates, next_pstate=None): """ Predict the next observation """ assert len(obs) == len(pstates) pstates_idx = np.array([self.e[ei] for ei in pstates]) next_pstate_idx = self.e[next_pstate] if len(obs) == 0: # No history, use the starting probas next_hstate_prob = self.startprob[next_pstate_idx] else: # With a history, determine the hidden state posteriors using # the last posteriors and transition matrix framelogprob = self._compute_log_likelihood(obs, pstates_idx) _, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx) next_hstate_prob = np.zeros(self.n_hidden_states) alpha_n = fwdlattice[-1] vmax = alpha_n.max(axis=0) alpha_n = np.exp(alpha_n - vmax) alpha_n = alpha_n / alpha_n.sum() trans = self.transmat[pstates_idx[-1], next_pstate_idx] for i in range(self.n_hidden_states): next_hstate_prob[i] = np.sum([alpha_n[j] * trans[j, i] for j in range(self.n_hidden_states)]) assert next_hstate_prob.sum() - 1 < TOLERANCE # Make the prediction prediction = np.array( [self.expected_value(feature, pstate=next_pstate, hstate_prob=next_hstate_prob) for feature in self.emission_name]) # next_hstate = np.argmax(next_hstate_prob) # prediction = np.array( # [self.expected_value(feature, pstate=next_pstate, hstate=next_hstate) for feature in # self.emission_name]) return prediction def gen_pstates_idx(self, n, random_state=None): if random_state is None: random_state = self.random_state random_state = check_random_state(random_state) startprob_cdf = np.cumsum(self.pstate_startprob) transmat_cdf = np.cumsum(self.pstate_transmat, 1) # Initial state. rand = random_state.rand() curr_pstate = (startprob_cdf > rand).argmax() pstates = [curr_pstate] for _ in range(1, n): rand = random_state.rand() curr_pstate = (transmat_cdf[curr_pstate] > rand).argmax() pstates.append(curr_pstate) return np.array(pstates, dtype=int) def sample(self, pstates=None, n_obs=None, random_state=None): """ """ random_state = check_random_state(random_state) if pstates is None and n_obs is None: raise Exception('Must provide either pstates or n_obs') if pstates is not None and n_obs is not None: raise Exception('Must provide either pstates or n_obs but not both') gen_pstates = False rand = random_state.rand() if pstates is None: gen_pstates = True pstartprob_cdf = np.cumsum(self.pstate_startprob) ptransmat_cdf = np.cumsum(self.pstate_transmat, 1) # Initial pstate currpstate = (pstartprob_cdf > rand).argmax() pstates_idx = [currpstate] pstates = [self.er[currpstate]] else: n_obs = len(pstates) pstates_idx = np.array([self.e[p] for p in pstates]) startprob_pdf = self.startprob[pstates_idx[0]] startprob_cdf = np.cumsum(startprob_pdf) transmat_cdf = np.cumsum(self.transmat[0, pstates_idx[0]], 1) # Initial hidden state rand = random_state.rand() currstate = (startprob_cdf > rand).argmax() hidden_states = [currstate] obs = [self._generate_sample_from_state(currstate, pstates_idx[0], random_state)] for i in range(1, n_obs): rand = random_state.rand() if gen_pstates: currpstate = (ptransmat_cdf[currpstate] > rand).argmax() pstates_idx.append(currpstate) pstates.append(self.er[currpstate]) transmat_cdf = np.cumsum(self.transmat[pstates_idx[i - 1], pstates_idx[i]], 1) rand = random_state.rand() currstate = (transmat_cdf[currstate] > rand).argmax() hidden_states.append(currstate) obs.append(self._generate_sample_from_state(currstate, pstates_idx[i], random_state)) return np.array(obs), np.array(pstates), np.array(hidden_states, dtype=int) def fit_df(self, dfs, pstate_col=PSTATE_COL): """ Convenience function to fit a model from a list of dataframes """ obs_cols = list(self.emission_name) obs = [df[df.columns.difference([pstate_col])][obs_cols].values for df in dfs] pstates = [df[pstate_col].values for df in dfs] return self.fit(obs, pstates) def score_df(self, df, pstate_col=PSTATE_COL): """ """ obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values return self.score(obs, pstates) def score_events_df(self, df, pstate_col=PSTATE_COL, score_col='score'): """ """ df = df.copy() obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values df[score_col] = self.score_events(obs, pstates) return df def predict_states_df(self, df, pstate_col=PSTATE_COL, hstate_col=HSTATE_COL): df = df.copy() obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values _, df[hstate_col] = self.predict_states(obs, pstates) return df def predict_df(self, df, next_pstate=None, pstate_col=PSTATE_COL): obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values return self.predict(obs, pstates, next_pstate) def sample_df(self, pstates=None, n_obs=None, random_state=None, pstate_col=PSTATE_COL, hstate_col=HSTATE_COL): """ Convenience function to generate samples a model and create a dataframe """ try: import pandas as pd except Exception as e: raise e obs, pstates, hstates = self.sample(pstates, n_obs, random_state) items = [] if pstate_col is not None: items.append((pstate_col, pstates)) if hstate_col is not None: items.append((hstate_col, hstates)) items = items + [(self.emission_name[i], obs[:, i]) for i in range(self.n_features)] df = pd.DataFrame.from_items(items) return df def __str__(self): pstates = sorted(self.e.keys()) sep = '-' * 80 + '\n' sep2 = '_' * 40 + '\n' out = 'POHMM\n' out += 'H-states: %d\n' % self.n_hidden_states out += 'P-states: (%d) %s\n' % (len(pstates), str(pstates)) out += 'Emission: %s\n' % (self.emission_distr) out += sep out += 'Transition probas\n' out += sep2 out += '. -> .\n%s\n' % str(self.transmat[0, 0]) for pstate in pstates: out += sep2 out += '%s -> .\n%s\n' % (pstate, str(self.transmat[self.e[pstate], 0])) out += sep2 out += '. -> %s\n%s\n' % (pstate, str(self.transmat[0, self.e[pstate]])) for pstate1, pstate2 in product(pstates, pstates): out += sep2 out += '%s -> %s\n%s\n' % (pstate1, pstate2, str(self.transmat[self.e[pstate1], self.e[pstate2]])) out += sep out += 'Starting probas\n' out += '.: %s\n' % str(self.startprob[0]) for pstate in pstates: out += '%s: %s\n' % (pstate, str(self.startprob[self.e[pstate]])) out += sep out += 'Steady probas\n' out += '.: %s\n' % str(self.steadyprob[0]) for pstate in pstates: out += '%s: %s\n' % (pstate, str(self.steadyprob[self.e[pstate]])) out += sep out += 'Emissions\n' for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): out += sep2 out += 'Feature %s: %s\n' % (feature_name, feature_distr) for param in _DISTRIBUTIONS[feature_distr]: out += '.: %s = %s\n' % (param, str(self.emission[feature_name][param][0])) for pstate in pstates: out += '%s: %s = %s\n' % (pstate, param, str(self.emission[feature_name][param][self.e[pstate]])) out += sep return out def expected_value(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): """ Determine the joint maximum likelihood estimate """ # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': return np.sum(w * expected_lognormal(self.emission[feature]['logsigma'].flatten(), self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': return np.sum(w * expected_normal(self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten())) elif hstate is None: # Marginalize hstate if self.emission_name_distr[feature] == 'lognormal': return np.sum(hstate_prob * expected_lognormal(self.emission[feature]['logsigma'][pstate_idx, :], self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': return np.sum(hstate_prob * expected_normal(self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :])) elif pstate is None: # Marginalize pstate if self.emission_name_distr[feature] == 'lognormal': return np.sum(pstate_prob * expected_lognormal(self.emission[feature]['logsigma'][:, hstate], self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': return np.sum(pstate_prob * expected_normal(self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate])) else: if self.emission_name_distr[feature] == 'lognormal': return expected_lognormal(self.emission[feature]['logsigma'][pstate_idx, hstate], self.emission[feature]['logmu'][pstate_idx, hstate]) elif self.emission_name_distr[feature] == 'normal': return expected_normal(self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) return def pdf_fn(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'].flatten(), 0, np.exp(self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten()) elif hstate is None: # Marginalize hstate w = hstate_prob if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][pstate_idx, :], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :]) elif pstate is None: # Marginalize pstate w = pstate_prob if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][:, hstate], 0, np.exp(self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate]) else: w = 1 if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][pstate_idx, hstate], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, hstate])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) def fn(x): if np.isscalar(x): p = np.sum(w * pdf(x)) else: x = np.array(x) p = np.zeros(len(x)) for i, xi in enumerate(x): p[i] = np.sum(w * pdf(xi)) return p return fn def cdf_fn(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'].flatten(), 0, np.exp(self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten()) elif hstate is None: # Marginalize hstate w = hstate_prob if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][pstate_idx, :], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :]) elif pstate is None: # Marginalize pstate w = pstate_prob if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][:, hstate], 0, np.exp(self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate]) else: w = 1 if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][pstate_idx, hstate], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, hstate])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) def fn(x): if np.isscalar(x): p = np.sum(w * cdf(x)) else: x = np.array(x) p = np.zeros(len(x)) for i, xi in enumerate(x): p[i] = np.sum(w * cdf(xi)) return p return fn def params(self, pstates=None): if pstates is None: pstates = [None] + sorted( set(self.er.values())) # TODO: self.e caches any unknown value, maybe it shouldn't? params = [] # emission parameters for hstate, pstate_label in product(range(self.n_hidden_states), pstates): for feature, distr in zip(self.emission_name, self.emission_distr): for feature_param in _DISTRIBUTIONS[distr]: params.append(self.emission[feature][feature_param][self.e[pstate_label], hstate]) # transition parameters, diagonals only assuming 2 state for hstate, pstate_label in product(range(self.n_hidden_states), pstates): params.append(self.transmat[self.e[pstate_label], self.e[pstate_label], hstate, hstate]) return np.array(params)
vmonaco/pohmm
pohmm/pohmm.py
Pohmm.predict
python
def predict(self, obs, pstates, next_pstate=None): assert len(obs) == len(pstates) pstates_idx = np.array([self.e[ei] for ei in pstates]) next_pstate_idx = self.e[next_pstate] if len(obs) == 0: # No history, use the starting probas next_hstate_prob = self.startprob[next_pstate_idx] else: # With a history, determine the hidden state posteriors using # the last posteriors and transition matrix framelogprob = self._compute_log_likelihood(obs, pstates_idx) _, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx) next_hstate_prob = np.zeros(self.n_hidden_states) alpha_n = fwdlattice[-1] vmax = alpha_n.max(axis=0) alpha_n = np.exp(alpha_n - vmax) alpha_n = alpha_n / alpha_n.sum() trans = self.transmat[pstates_idx[-1], next_pstate_idx] for i in range(self.n_hidden_states): next_hstate_prob[i] = np.sum([alpha_n[j] * trans[j, i] for j in range(self.n_hidden_states)]) assert next_hstate_prob.sum() - 1 < TOLERANCE # Make the prediction prediction = np.array( [self.expected_value(feature, pstate=next_pstate, hstate_prob=next_hstate_prob) for feature in self.emission_name]) # next_hstate = np.argmax(next_hstate_prob) # prediction = np.array( # [self.expected_value(feature, pstate=next_pstate, hstate=next_hstate) for feature in # self.emission_name]) return prediction
Predict the next observation
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/pohmm.py#L759-L800
[ "def _compute_log_likelihood(self, obs, pstates_idx):\n\n q = np.zeros(shape=(len(obs), self.n_hidden_states, self.n_features))\n\n for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)):\n if feature_distr == 'normal':\n mu = self.emission[feature_name]['mu'][pstates_idx]\n sigma = self.emission[feature_name]['sigma'][pstates_idx]\n for j in range(self.n_hidden_states):\n q[:, j, col] = np.log(\n np.maximum(MIN_PROBA, stats.norm.pdf(obs[:, col], loc=mu[:, j], scale=sigma[:, j])))\n if feature_distr == 'lognormal':\n logmu = self.emission[feature_name]['logmu'][pstates_idx]\n logsigma = self.emission[feature_name]['logsigma'][pstates_idx]\n for j in range(self.n_hidden_states):\n q[:, j, col] = np.log(np.maximum(MIN_PROBA,\n stats.lognorm.pdf(obs[:, col], logsigma[:, j], loc=0,\n scale=np.exp(logmu[:, j]))))\n\n q = q.sum(axis=2)\n return q\n", "def _do_forward_pass(self, framelogprob, event_idx):\n n_observations, n_hidden_states = framelogprob.shape\n fwdlattice = np.zeros((n_observations, n_hidden_states))\n _hmmc._forward(n_observations, n_hidden_states,\n event_idx, self._log_startprob,\n self._log_transmat, framelogprob, fwdlattice)\n\n return logsumexp(fwdlattice[-1]), fwdlattice\n" ]
class Pohmm(object): """ Partially observable hidden Markov model """ def __init__(self, n_hidden_states=2, emissions=['normal'], max_iter=1000, thresh=1e-6, init_method='obs', init_spread=2, smoothing=None, random_state=None): if type(n_hidden_states) is int: self.n_hidden_states = n_hidden_states else: raise Exception('Wrong type for n_hidden_states. Must be int') if type(emissions[0]) is tuple: emission_name, emission_distr = zip(*emissions) elif type(emissions[0]) is str: emission_name, emission_distr = np.arange(len(emissions)), emissions for distr in emission_distr: if distr not in _DISTRIBUTIONS.keys(): raise ValueError('Emission distribution must be one of', _DISTRIBUTIONS.keys()) self.emission_name = emission_name self.emission_distr = emission_distr self.emission_name_distr = dict(zip(emission_name, emission_distr)) self.n_features = len(emissions) # Set up the emission parameters # emission: {'feature':{'param': np.array(shape=(n_partial_states, n_hidden_states))}} self.emission = defaultdict(dict) for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: self.emission[name][param] = None self.emission = dict(self.emission) assert max_iter >= 0 assert thresh >= 0 self.max_iter = max_iter self.thresh = thresh assert init_spread >= 0 self.init_spread = init_spread if init_method not in _INIT_METHODS: raise ValueError('init_method must be one of', _INIT_METHODS) self.init_method = init_method if smoothing is None: smoothing = {'transmat': None, 'startprob': None} for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: smoothing['%s:%s' % (name, param)] = None elif type(smoothing) is str: s = smoothing smoothing = {'transmat': s, 'startprob': s} for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: smoothing['%s:%s' % (name, param)] = s elif type(smoothing) is dict: assert 'transmat' in smoothing.keys() assert 'startprob' in smoothing.keys() for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: assert param in smoothing.keys() or '%s:%s' % (name, param) in smoothing.keys() if param in smoothing.keys() and '%s:%s' % (name, param) not in smoothing.keys(): smoothing['%s:%s' % (name, param)] = smoothing[param] else: raise Exception('Wrong type for smoothing. Must be None, str, or dict') self.smoothing = smoothing self.random_state = random_state # Number of unique partial states is unknown until fit self.n_partial_states = None # Results after fitting the model self.logprob_ = None self.n_iter_performed_ = None self.logprob_delta_ = None # Mapping between p-states and a unique index # Defaults to 0 for unknown or missing p-states self.e = defaultdict(int) def _get_startprob(self): return np.exp(self._log_startprob) def _set_startprob(self, startprob): if startprob is None: startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states else: startprob = np.asarray(startprob, dtype=np.float) # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(startprob): startprob = normalize(startprob, axis=1) if len(startprob) != self.n_partial_states: raise ValueError('startprob must have length n_partial_states') if not np.allclose(np.sum(startprob, axis=1), 1.0): raise ValueError('startprob must sum to 1.0') self._log_startprob = np.log(np.asarray(startprob).copy()) startprob = property(_get_startprob, _set_startprob) def _get_steadyprob(self): return np.exp(self._log_steadyprob) def _set_steadyprob(self, steadyprob): if steadyprob is None: steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states else: steadyprob = np.asarray(steadyprob, dtype=np.float) # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(steadyprob): steadyprob = normalize(steadyprob, axis=1) if len(steadyprob) != self.n_partial_states: raise ValueError('steadyprob must have length n_partial_states') if not np.allclose(np.sum(steadyprob, axis=1), 1.0): raise ValueError('steadyprob must sum to 1.0') self._log_steadyprob = np.log(np.asarray(steadyprob).copy()) steadyprob = property(_get_steadyprob, _set_steadyprob) def _get_transmat(self): return np.exp(self._log_transmat) def _set_transmat(self, transmat): if transmat is None: transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) / self.n_hidden_states # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(transmat): transmat = normalize(transmat, axis=3) if (np.asarray(transmat).shape != (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)): raise ValueError('transmat must have shape ' '(n_partial_states,n_partial_states,n_hidden_states,n_hidden_states)') if not np.all(np.allclose(np.sum(transmat, axis=3), 1.0)): raise ValueError('Rows of transmat must sum to 1.0') self._log_transmat = np.log(np.asarray(transmat).copy()) underflow_idx = np.isnan(self._log_transmat) self._log_transmat[underflow_idx] = NEGINF transmat = property(_get_transmat, _set_transmat) def _compute_log_likelihood(self, obs, pstates_idx): q = np.zeros(shape=(len(obs), self.n_hidden_states, self.n_features)) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': mu = self.emission[feature_name]['mu'][pstates_idx] sigma = self.emission[feature_name]['sigma'][pstates_idx] for j in range(self.n_hidden_states): q[:, j, col] = np.log( np.maximum(MIN_PROBA, stats.norm.pdf(obs[:, col], loc=mu[:, j], scale=sigma[:, j]))) if feature_distr == 'lognormal': logmu = self.emission[feature_name]['logmu'][pstates_idx] logsigma = self.emission[feature_name]['logsigma'][pstates_idx] for j in range(self.n_hidden_states): q[:, j, col] = np.log(np.maximum(MIN_PROBA, stats.lognorm.pdf(obs[:, col], logsigma[:, j], loc=0, scale=np.exp(logmu[:, j])))) q = q.sum(axis=2) return q def _generate_sample_from_state(self, hidden_state, pstates_idx, random_state=None): sample = np.zeros(self.n_features) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': mu = self.emission[feature_name]['mu'][pstates_idx][hidden_state] sigma = self.emission[feature_name]['sigma'][pstates_idx][hidden_state] sample[col] = stats.norm.rvs(loc=mu, scale=sigma, random_state=random_state) if feature_distr == 'lognormal': logmu = self.emission[feature_name]['logmu'][pstates_idx][hidden_state] logsigma = self.emission[feature_name]['logsigma'][pstates_idx][hidden_state] sample[col] = stats.lognorm.rvs(logsigma, loc=0, scale=np.exp(logmu), random_state=random_state) return sample def _init_pstates(self, unique_pstates): # Map events to a unique index. The unknown p-state is at idx 0 self.e = defaultdict(int) self.e.update(dict(zip(np.sort(unique_pstates), range(1, len(unique_pstates) + 1)))) self.er = {v: k for k, v in self.e.items()} # Reverse lookup self.er[0] = UNKNOWN_PSTATE self.n_partial_states = len(self.e.keys()) + 1 # Add one for the unknown state return def _init_pstate_freqs(self, pstates_idx): # Partial state frequencies self.pstate_freq = Counter([idx for seq in pstates_idx for idx in seq]) self.pstate_trans_freq = Counter([(idx1, idx2) for seq in pstates_idx for idx1, idx2 in zip(seq[:-1], seq[1:])]) # Store freqs for the meta state self.pstate_freq[0] = len(np.concatenate(pstates_idx)) for seq in pstates_idx: self.pstate_startprob[seq[0]] += 1 self.pstate_steadyprob += np.bincount(seq, minlength=self.n_partial_states) for idx1, idx2 in zip(seq[:-1], seq[1:]): self.pstate_trans_freq[(0, 0)] += 1 self.pstate_trans_freq[(idx1, 0)] += 1 self.pstate_trans_freq[(0, idx2)] += 1 self.pstate_transmat[idx1, idx2] += 1 # TODO: separate probas from freqs # Normalize to get the probabilities, ignore the meta state at idx 0 self.pstate_startprob[1:] = normalize(self.pstate_startprob[1:]) self.pstate_transmat[1:, 1:] = normalize(self.pstate_transmat[1:, 1:], axis=1) self.pstate_steadyprob[1:] = normalize(self.pstate_steadyprob[1:]) return def _init_from_obs(self, obs, pstates_idx): # Partial state probabilities self.pstate_startprob = np.zeros(self.n_partial_states) self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states)) self.pstate_steadyprob = np.zeros(self.n_partial_states) # obs should be (N*T, n_features) # N is the number of samples # T is the size of each sample obs = np.concatenate(obs) pstates_idx = np.concatenate(pstates_idx) # Initialize starting and transition probas self.startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states self.steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states self.transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) / self.n_hidden_states # Initialize emission parameters # Hidden states are ordered by the first feature feature1 = self.emission_name[0] for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': self.emission[feature_name]['mu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['sigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) for idx in range(1, self.n_partial_states): idx_pstate = (pstates_idx == idx) if not np.any(idx_pstate): idx_pstate = np.arange(len(pstates_idx)) if feature_name == feature1: self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA) self.emission[feature_name]['mu'][idx] = obs[idx_pstate, col].mean() + \ obs[:, col].std() * np.linspace( -self.init_spread, self.init_spread, self.n_hidden_states) else: self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA) self.emission[feature_name]['mu'][idx, :] = obs[idx_pstate, col].mean() if feature_distr == 'lognormal': self.emission[feature_name]['logmu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['logsigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) for idx in range(1, self.n_partial_states): idx_pstate = (pstates_idx == idx) if not np.any(idx_pstate): idx_pstate = np.arange(len(pstates_idx)) if feature_name == feature1: self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(), MIN_PROBA) self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() + np.log( obs[:, col]).std() * np.linspace(-self.init_spread, self.init_spread, self.n_hidden_states) else: self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(), MIN_PROBA) self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() return def _init_random(self, random_state=None): if random_state is None: random_state = self.random_state random_state = check_random_state(random_state) self.pstate_startprob = np.zeros(self.n_partial_states) self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states)) self.pstate_steadyprob = np.zeros(self.n_partial_states) self.pstate_startprob[1:] = gen_stochastic_matrix(size=self.n_partial_states - 1, random_state=random_state) self.pstate_transmat[1:, 1:] = gen_stochastic_matrix( size=(self.n_partial_states - 1, self.n_partial_states - 1), random_state=random_state) self.pstate_steadyprob[1:] = steadystate(self.pstate_transmat[1:, 1:]) self.startprob = gen_stochastic_matrix(size=(self.n_partial_states, self.n_hidden_states), random_state=random_state) transmat = np.zeros((self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)): transmat[i, j] = gen_stochastic_matrix(size=(self.n_hidden_states, self.n_hidden_states), random_state=random_state) self.transmat = normalize(transmat, axis=3) # Initialize emission parameters for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): if feature_distr == 'normal': self.emission[feature_name]['mu'] = random_state.uniform(*_RANDINIT[feature_distr]['mu'], size=( self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['sigma'] = random_state.uniform(*_RANDINIT[feature_distr]['sigma'], size=( self.n_partial_states, self.n_hidden_states)) if feature_distr == 'lognormal': self.emission[feature_name]['logmu'] = random_state.uniform(*_RANDINIT[feature_distr]['logmu'], size=( self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['logsigma'] = random_state.uniform(*_RANDINIT[feature_distr]['logsigma'], size=(self.n_partial_states, self.n_hidden_states)) if self.emission_distr[0] == 'normal': self.emission[self.emission_name[0]]['mu'] = np.sort(self.emission[self.emission_name[0]]['mu'], axis=1) elif self.emission_distr[0] == 'lognormal': self.emission[self.emission_name[0]]['logmu'] = np.sort(self.emission[self.emission_name[0]]['logmu'], axis=1) return def _smooth(self): self._compute_marginals() startprob = self.startprob for j in range(1, self.n_partial_states): if 'freq' == self.smoothing['startprob']: w_ = 1 / (1 + self.pstate_freq[j]) w_j = 1 - w_ elif 'proba' == self.smoothing['startprob']: w_j = self.pstate_steadyprob[j] w_ = 1 - w_j elif 'exp' == self.smoothing['startprob']: w_ = np.exp(-self.pstate_freq[j]) w_j = 1 - w_ elif 'fixed' == self.smoothing['startprob']: w_ = 1 w_j = 0 elif self.smoothing['startprob'] is None: w_ = 0 w_j = 1 else: raise Exception('Wrong smoothing for startprob: ' + self.smoothing['startprob']) startprob[j] = w_j * self.startprob[j] + w_ * self.startprob[0] self.startprob = startprob transmat = self.transmat for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)): if 'freq' == self.smoothing['transmat']: w_i0 = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[0, j]) w_0j = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[i, 0]) w_ij = 1 - (w_i0 + w_0j) w_ = 0 elif 'proba' == self.smoothing['transmat']: denom = self.pstate_transmat[i, j] + self.pstate_transmat[i, :].sum() + self.pstate_transmat[:, j].sum() w_i0 = self.pstate_transmat[i, :].sum() / denom w_0j = self.pstate_transmat[:, j].sum() / denom w_ij = self.pstate_transmat[i, j] / denom w_ = 0 elif 'exp' == self.smoothing['transmat']: w_i0 = self.pstate_trans_freq[i, 0] * np.exp(-self.pstate_trans_freq[i, j]) w_0j = self.pstate_trans_freq[0, j] * np.exp(-self.pstate_trans_freq[i, j]) w_ = self.pstate_trans_freq[0, 0] * np.exp( -(self.pstate_trans_freq[i, 0] + self.pstate_trans_freq[0, j])) w_ij = self.pstate_trans_freq[i, j] w_ij, w_i0, w_0j, w_ = normalize(np.array([w_ij, w_i0, w_0j, w_])) elif 'fixed' == self.smoothing['transmat']: w_i0 = 0 w_0j = 0 w_ = 1 w_ij = 0 elif self.smoothing['transmat'] is None: w_i0 = 0 w_0j = 0 w_ = 0 w_ij = 1 else: raise Exception('Wrong smoothing for transmat: ' + self.smoothing['transmat']) assert (w_i0 + w_0j + w_ + w_ij) - 1 < TOLERANCE transmat[i, j] = w_ij * self.transmat[i, j] + w_i0 * self.transmat[i, 0] + w_0j * self.transmat[0, j] + w_ * \ self.transmat[ 0, 0] self.transmat = transmat assert np.all(self.startprob.sum(axis=1) - 1 < TOLERANCE) assert np.all(self.steadyprob.sum(axis=1) - 1 < TOLERANCE) assert np.all(self.transmat.sum(axis=3) - 1 < TOLERANCE) for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[feature_distr]: key = '%s:%s' % (feature_name, param) for j in range(1, self.n_partial_states): if 'freq' == self.smoothing[key]: w_ = 1 / (1 + self.pstate_freq[j]) w_j = 1 - w_ elif 'proba' == self.smoothing[key]: w_j = self.pstate_steadyprob[j] w_ = 1 - w_j elif 'exp' == self.smoothing[key]: w_ = np.exp(-self.pstate_freq[j]) w_j = 1 - w_ elif 'fixed' == self.smoothing[key]: w_ = 1 w_j = 0 elif self.smoothing[key] is None: w_ = 0 w_j = 1 else: raise Exception('Wrong smoothing for ' + key) self.emission[feature_name][param][j] = w_j * self.emission[feature_name][param][j] + w_ * \ self.emission[ feature_name][ param][0] return def _initialize_sufficient_statistics(self): stats = { 'nobs': 0, 'post': np.zeros((self.n_partial_states, self.n_hidden_states)), 'obs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'obs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'lnobs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'lnobs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'start': np.zeros((self.n_partial_states, self.n_hidden_states)), 'steady': np.zeros((self.n_partial_states, self.n_hidden_states)), 'trans': np.zeros( (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) } return stats def _accumulate_sufficient_statistics(self, stats, obs, pstates_idx, framelogprob, posteriors, fwdlattice, bwdlattice): stats['nobs'] += 1 n_observations, n_hidden_states = framelogprob.shape stats['start'][0] += posteriors[0] for i in range(self.n_partial_states): if len(np.where(pstates_idx == i)[0]) > 0: stats['start'][i] += posteriors[np.where(pstates_idx == i)[0].min()] if n_observations > 1: lneta = np.zeros((n_observations - 1, n_hidden_states, n_hidden_states)) _hmmc._compute_lneta(n_observations, n_hidden_states, pstates_idx, fwdlattice, self._log_transmat, bwdlattice, framelogprob, lneta) for i, j in unique_rows(np.c_[pstates_idx[:-1], pstates_idx[1:]]): if ((pstates_idx[:-1] == i) & (pstates_idx[1:] == j)).sum() > 0: stats['trans'][i, j] += np.exp( logsumexp(lneta[(pstates_idx[:-1] == i) & (pstates_idx[1:] == j)], axis=0)) for i in range(self.n_partial_states): stats['post'][i] += posteriors[pstates_idx == i].sum(axis=0) stats['obs'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i]) stats['obs**2'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i] ** 2) stats['lnobs'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i])) stats['lnobs**2'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i]) ** 2) return def _compute_marginals(self): # TODO: cythonize some of this # Start prob, weighted by p-state start probs self.startprob[0] = (self.pstate_startprob[1:, np.newaxis] * self.startprob[1:]).sum(axis=0) # Use the p-state transmat and transmat to get the full transmat full_transmat = ph2full(self.pstate_transmat[1:, 1:], self.transmat[1:, 1:]) full_steadyprob = steadystate(full_transmat) # Steady state probas are determined by the full trans mat, need to be updated steadyprob = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) steadyprob[0] = full_steadyprob.reshape(-1, self.n_hidden_states).sum(axis=0) for i in range(self.n_partial_states - 1): steadyprob[i + 1] = normalize( full_steadyprob[i * self.n_hidden_states:i * self.n_hidden_states + self.n_hidden_states]) self.steadyprob = steadyprob # Update the transations to/from the marginal state transmat = self.transmat # Group the hidden states within each partial state for hidx1, hidx2 in product(range(self.n_hidden_states), range(self.n_hidden_states)): transmat[0, 0][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, hidx2::self.n_hidden_states].sum() for pidx in range(self.n_partial_states - 1): transmat[pidx + 1, 0][hidx1, hidx2] = full_transmat[pidx * self.n_hidden_states + hidx1, hidx2::self.n_hidden_states].sum() transmat[0, pidx + 1][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, pidx * self.n_hidden_states + hidx2].sum() self.transmat = normalize(transmat, axis=3) pweights = self.pstate_steadyprob[1:, np.newaxis] # Update emission parameters for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): if feature_distr == 'normal': # Marginal state is a mixture of normals mu = self.emission[feature_name]['mu'][1:, :] sigma = self.emission[feature_name]['sigma'][1:, :] # Weighted mean and var mu_0 = (pweights * mu).sum(axis=0) self.emission[feature_name]['mu'][0, :] = mu_0 self.emission[feature_name]['sigma'][0, :] = np.sqrt( (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0)) if feature_distr == 'lognormal': # Marginal state is a mixture of normals mu = self.emission[feature_name]['logmu'][1:, :] sigma = self.emission[feature_name]['logsigma'][1:, :] # Weighted mean and var mu_0 = (pweights * mu).sum(axis=0) self.emission[feature_name]['logmu'][0, :] = mu_0 self.emission[feature_name]['logsigma'][0, :] = np.sqrt( (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0)) return def _do_mstep(self, stats): self.startprob = normalize(np.maximum(stats['start'], MIN_PROBA), axis=1) self.transmat = normalize(np.maximum(stats['trans'], MIN_PROBA), axis=3) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': denom = np.maximum(stats['post'], MIN_PROBA) mu = stats['obs'][:, :, col] / denom cv_num = (stats['obs**2'][:, :, col] - 2 * mu * stats['obs'][:, :, col] + mu ** 2 * denom) sigma = np.sqrt(cv_num / denom) sigma[np.isnan(sigma)] = MIN_PROBA self.emission[feature_name]['mu'] = mu self.emission[feature_name]['sigma'] = np.maximum(sigma, MIN_PROBA) if feature_distr == 'lognormal': denom = np.maximum(stats['post'], MIN_PROBA) mu = stats['lnobs'][:, :, col] / denom cv_num = (stats['lnobs**2'][:, :, col] - 2 * mu * stats['lnobs'][:, :, col] + mu ** 2 * denom) sigma = np.sqrt(cv_num / denom) sigma[np.isnan(sigma)] = MIN_PROBA self.emission[feature_name]['logmu'] = mu self.emission[feature_name]['logsigma'] = np.maximum(sigma, MIN_PROBA) return def _do_forward_pass(self, framelogprob, event_idx): n_observations, n_hidden_states = framelogprob.shape fwdlattice = np.zeros((n_observations, n_hidden_states)) _hmmc._forward(n_observations, n_hidden_states, event_idx, self._log_startprob, self._log_transmat, framelogprob, fwdlattice) return logsumexp(fwdlattice[-1]), fwdlattice def _do_backward_pass(self, framelogprob, event_idx): n_observations, n_hidden_states = framelogprob.shape bwdlattice = np.zeros((n_observations, n_hidden_states)) _hmmc._backward(n_observations, n_hidden_states, event_idx, self._log_startprob, self._log_transmat, framelogprob, bwdlattice) return bwdlattice def _do_viterbi_pass(self, framelogprob, event_idx): n_observations, n_components = framelogprob.shape state_sequence, logprob = _hmmc._viterbi( n_observations, n_components, event_idx, self._log_startprob, self._log_transmat, framelogprob) return logprob, state_sequence def rand(self, unique_pstates, random_state=None): """ Randomize the POHMM parameters """ self._init_pstates(unique_pstates) self._init_random(random_state=random_state) self._compute_marginals() return self def fit(self, obs, pstates, unique_pstates=None): """ Estimate model parameters. """ obs = [np.array(o) for o in obs] pstates = [np.array(p) for p in pstates] # List or array of observation sequences assert len(obs) == len(pstates) assert obs[0].ndim == 2 assert pstates[0].ndim == 1 if unique_pstates is not None: self._init_pstates(unique_pstates) else: self._init_pstates(list(set(np.concatenate(pstates)))) # Map the partial states to a unique index pstates_idx = [np.array([self.e[p] for p in seq]) for seq in pstates] if self.init_method == 'rand': self._init_random() elif self.init_method == 'obs': self._init_from_obs(obs, pstates_idx) self._init_pstate_freqs(pstates_idx) self._smooth() logprob = [] for i in range(self.max_iter): # Expectation step stats = self._initialize_sufficient_statistics() curr_logprob = 0 for obs_i, pstates_idx_i in zip(obs, pstates_idx): framelogprob = self._compute_log_likelihood(obs_i, pstates_idx_i) lpr, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx_i) bwdlattice = self._do_backward_pass(framelogprob, pstates_idx_i) gamma = fwdlattice + bwdlattice posteriors = np.exp(gamma.T - logsumexp(gamma, axis=1)).T curr_logprob += lpr self._accumulate_sufficient_statistics(stats, obs_i, pstates_idx_i, framelogprob, posteriors, fwdlattice, bwdlattice) logprob.append(curr_logprob) self.logprob_ = curr_logprob # Check for convergence. self.n_iter_performed_ = i if i > 0: self.logprob_delta = logprob[-1] - logprob[-2] if self.logprob_delta < self.thresh: break # Maximization step self._do_mstep(stats) # Mix the parameters self._smooth() return self def score(self, obs, pstates): """ Compute the log probability under the model. """ pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) logprob, _ = self._do_forward_pass(framelogprob, pstates_idx) return logprob def score_events(self, obs, pstates): """ Compute the log probability of each event under the model. """ pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) _, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx) L = logsumexp(fwdlattice, axis=1) return np.concatenate([L[[0]], np.diff(L)]) def predict_states(self, obs, pstates): pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) viterbi_logprob, state_sequence = self._do_viterbi_pass(framelogprob, pstates_idx) return viterbi_logprob, state_sequence def gen_pstates_idx(self, n, random_state=None): if random_state is None: random_state = self.random_state random_state = check_random_state(random_state) startprob_cdf = np.cumsum(self.pstate_startprob) transmat_cdf = np.cumsum(self.pstate_transmat, 1) # Initial state. rand = random_state.rand() curr_pstate = (startprob_cdf > rand).argmax() pstates = [curr_pstate] for _ in range(1, n): rand = random_state.rand() curr_pstate = (transmat_cdf[curr_pstate] > rand).argmax() pstates.append(curr_pstate) return np.array(pstates, dtype=int) def sample(self, pstates=None, n_obs=None, random_state=None): """ """ random_state = check_random_state(random_state) if pstates is None and n_obs is None: raise Exception('Must provide either pstates or n_obs') if pstates is not None and n_obs is not None: raise Exception('Must provide either pstates or n_obs but not both') gen_pstates = False rand = random_state.rand() if pstates is None: gen_pstates = True pstartprob_cdf = np.cumsum(self.pstate_startprob) ptransmat_cdf = np.cumsum(self.pstate_transmat, 1) # Initial pstate currpstate = (pstartprob_cdf > rand).argmax() pstates_idx = [currpstate] pstates = [self.er[currpstate]] else: n_obs = len(pstates) pstates_idx = np.array([self.e[p] for p in pstates]) startprob_pdf = self.startprob[pstates_idx[0]] startprob_cdf = np.cumsum(startprob_pdf) transmat_cdf = np.cumsum(self.transmat[0, pstates_idx[0]], 1) # Initial hidden state rand = random_state.rand() currstate = (startprob_cdf > rand).argmax() hidden_states = [currstate] obs = [self._generate_sample_from_state(currstate, pstates_idx[0], random_state)] for i in range(1, n_obs): rand = random_state.rand() if gen_pstates: currpstate = (ptransmat_cdf[currpstate] > rand).argmax() pstates_idx.append(currpstate) pstates.append(self.er[currpstate]) transmat_cdf = np.cumsum(self.transmat[pstates_idx[i - 1], pstates_idx[i]], 1) rand = random_state.rand() currstate = (transmat_cdf[currstate] > rand).argmax() hidden_states.append(currstate) obs.append(self._generate_sample_from_state(currstate, pstates_idx[i], random_state)) return np.array(obs), np.array(pstates), np.array(hidden_states, dtype=int) def fit_df(self, dfs, pstate_col=PSTATE_COL): """ Convenience function to fit a model from a list of dataframes """ obs_cols = list(self.emission_name) obs = [df[df.columns.difference([pstate_col])][obs_cols].values for df in dfs] pstates = [df[pstate_col].values for df in dfs] return self.fit(obs, pstates) def score_df(self, df, pstate_col=PSTATE_COL): """ """ obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values return self.score(obs, pstates) def score_events_df(self, df, pstate_col=PSTATE_COL, score_col='score'): """ """ df = df.copy() obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values df[score_col] = self.score_events(obs, pstates) return df def predict_states_df(self, df, pstate_col=PSTATE_COL, hstate_col=HSTATE_COL): df = df.copy() obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values _, df[hstate_col] = self.predict_states(obs, pstates) return df def predict_df(self, df, next_pstate=None, pstate_col=PSTATE_COL): obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values return self.predict(obs, pstates, next_pstate) def sample_df(self, pstates=None, n_obs=None, random_state=None, pstate_col=PSTATE_COL, hstate_col=HSTATE_COL): """ Convenience function to generate samples a model and create a dataframe """ try: import pandas as pd except Exception as e: raise e obs, pstates, hstates = self.sample(pstates, n_obs, random_state) items = [] if pstate_col is not None: items.append((pstate_col, pstates)) if hstate_col is not None: items.append((hstate_col, hstates)) items = items + [(self.emission_name[i], obs[:, i]) for i in range(self.n_features)] df = pd.DataFrame.from_items(items) return df def __str__(self): pstates = sorted(self.e.keys()) sep = '-' * 80 + '\n' sep2 = '_' * 40 + '\n' out = 'POHMM\n' out += 'H-states: %d\n' % self.n_hidden_states out += 'P-states: (%d) %s\n' % (len(pstates), str(pstates)) out += 'Emission: %s\n' % (self.emission_distr) out += sep out += 'Transition probas\n' out += sep2 out += '. -> .\n%s\n' % str(self.transmat[0, 0]) for pstate in pstates: out += sep2 out += '%s -> .\n%s\n' % (pstate, str(self.transmat[self.e[pstate], 0])) out += sep2 out += '. -> %s\n%s\n' % (pstate, str(self.transmat[0, self.e[pstate]])) for pstate1, pstate2 in product(pstates, pstates): out += sep2 out += '%s -> %s\n%s\n' % (pstate1, pstate2, str(self.transmat[self.e[pstate1], self.e[pstate2]])) out += sep out += 'Starting probas\n' out += '.: %s\n' % str(self.startprob[0]) for pstate in pstates: out += '%s: %s\n' % (pstate, str(self.startprob[self.e[pstate]])) out += sep out += 'Steady probas\n' out += '.: %s\n' % str(self.steadyprob[0]) for pstate in pstates: out += '%s: %s\n' % (pstate, str(self.steadyprob[self.e[pstate]])) out += sep out += 'Emissions\n' for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): out += sep2 out += 'Feature %s: %s\n' % (feature_name, feature_distr) for param in _DISTRIBUTIONS[feature_distr]: out += '.: %s = %s\n' % (param, str(self.emission[feature_name][param][0])) for pstate in pstates: out += '%s: %s = %s\n' % (pstate, param, str(self.emission[feature_name][param][self.e[pstate]])) out += sep return out def expected_value(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): """ Determine the joint maximum likelihood estimate """ # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': return np.sum(w * expected_lognormal(self.emission[feature]['logsigma'].flatten(), self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': return np.sum(w * expected_normal(self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten())) elif hstate is None: # Marginalize hstate if self.emission_name_distr[feature] == 'lognormal': return np.sum(hstate_prob * expected_lognormal(self.emission[feature]['logsigma'][pstate_idx, :], self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': return np.sum(hstate_prob * expected_normal(self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :])) elif pstate is None: # Marginalize pstate if self.emission_name_distr[feature] == 'lognormal': return np.sum(pstate_prob * expected_lognormal(self.emission[feature]['logsigma'][:, hstate], self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': return np.sum(pstate_prob * expected_normal(self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate])) else: if self.emission_name_distr[feature] == 'lognormal': return expected_lognormal(self.emission[feature]['logsigma'][pstate_idx, hstate], self.emission[feature]['logmu'][pstate_idx, hstate]) elif self.emission_name_distr[feature] == 'normal': return expected_normal(self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) return def pdf_fn(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'].flatten(), 0, np.exp(self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten()) elif hstate is None: # Marginalize hstate w = hstate_prob if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][pstate_idx, :], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :]) elif pstate is None: # Marginalize pstate w = pstate_prob if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][:, hstate], 0, np.exp(self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate]) else: w = 1 if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][pstate_idx, hstate], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, hstate])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) def fn(x): if np.isscalar(x): p = np.sum(w * pdf(x)) else: x = np.array(x) p = np.zeros(len(x)) for i, xi in enumerate(x): p[i] = np.sum(w * pdf(xi)) return p return fn def cdf_fn(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'].flatten(), 0, np.exp(self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten()) elif hstate is None: # Marginalize hstate w = hstate_prob if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][pstate_idx, :], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :]) elif pstate is None: # Marginalize pstate w = pstate_prob if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][:, hstate], 0, np.exp(self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate]) else: w = 1 if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][pstate_idx, hstate], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, hstate])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) def fn(x): if np.isscalar(x): p = np.sum(w * cdf(x)) else: x = np.array(x) p = np.zeros(len(x)) for i, xi in enumerate(x): p[i] = np.sum(w * cdf(xi)) return p return fn def params(self, pstates=None): if pstates is None: pstates = [None] + sorted( set(self.er.values())) # TODO: self.e caches any unknown value, maybe it shouldn't? params = [] # emission parameters for hstate, pstate_label in product(range(self.n_hidden_states), pstates): for feature, distr in zip(self.emission_name, self.emission_distr): for feature_param in _DISTRIBUTIONS[distr]: params.append(self.emission[feature][feature_param][self.e[pstate_label], hstate]) # transition parameters, diagonals only assuming 2 state for hstate, pstate_label in product(range(self.n_hidden_states), pstates): params.append(self.transmat[self.e[pstate_label], self.e[pstate_label], hstate, hstate]) return np.array(params)
vmonaco/pohmm
pohmm/pohmm.py
Pohmm.fit_df
python
def fit_df(self, dfs, pstate_col=PSTATE_COL): obs_cols = list(self.emission_name) obs = [df[df.columns.difference([pstate_col])][obs_cols].values for df in dfs] pstates = [df[pstate_col].values for df in dfs] return self.fit(obs, pstates)
Convenience function to fit a model from a list of dataframes
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/pohmm.py#L876-L883
[ "def fit(self, obs, pstates, unique_pstates=None):\n \"\"\"\n Estimate model parameters.\n \"\"\"\n obs = [np.array(o) for o in obs]\n pstates = [np.array(p) for p in pstates]\n\n # List or array of observation sequences\n assert len(obs) == len(pstates)\n assert obs[0].ndim == 2\n assert pstates[0].ndim == 1\n\n if unique_pstates is not None:\n self._init_pstates(unique_pstates)\n else:\n self._init_pstates(list(set(np.concatenate(pstates))))\n\n # Map the partial states to a unique index\n pstates_idx = [np.array([self.e[p] for p in seq]) for seq in pstates]\n\n if self.init_method == 'rand':\n self._init_random()\n elif self.init_method == 'obs':\n self._init_from_obs(obs, pstates_idx)\n\n self._init_pstate_freqs(pstates_idx)\n self._smooth()\n\n logprob = []\n for i in range(self.max_iter):\n # Expectation step\n stats = self._initialize_sufficient_statistics()\n curr_logprob = 0\n for obs_i, pstates_idx_i in zip(obs, pstates_idx):\n framelogprob = self._compute_log_likelihood(obs_i, pstates_idx_i)\n lpr, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx_i)\n bwdlattice = self._do_backward_pass(framelogprob, pstates_idx_i)\n gamma = fwdlattice + bwdlattice\n posteriors = np.exp(gamma.T - logsumexp(gamma, axis=1)).T\n curr_logprob += lpr\n\n self._accumulate_sufficient_statistics(stats, obs_i,\n pstates_idx_i, framelogprob, posteriors, fwdlattice, bwdlattice)\n\n logprob.append(curr_logprob)\n self.logprob_ = curr_logprob\n\n # Check for convergence.\n self.n_iter_performed_ = i\n if i > 0:\n self.logprob_delta = logprob[-1] - logprob[-2]\n if self.logprob_delta < self.thresh:\n break\n\n # Maximization step\n self._do_mstep(stats)\n\n # Mix the parameters\n self._smooth()\n\n return self\n" ]
class Pohmm(object): """ Partially observable hidden Markov model """ def __init__(self, n_hidden_states=2, emissions=['normal'], max_iter=1000, thresh=1e-6, init_method='obs', init_spread=2, smoothing=None, random_state=None): if type(n_hidden_states) is int: self.n_hidden_states = n_hidden_states else: raise Exception('Wrong type for n_hidden_states. Must be int') if type(emissions[0]) is tuple: emission_name, emission_distr = zip(*emissions) elif type(emissions[0]) is str: emission_name, emission_distr = np.arange(len(emissions)), emissions for distr in emission_distr: if distr not in _DISTRIBUTIONS.keys(): raise ValueError('Emission distribution must be one of', _DISTRIBUTIONS.keys()) self.emission_name = emission_name self.emission_distr = emission_distr self.emission_name_distr = dict(zip(emission_name, emission_distr)) self.n_features = len(emissions) # Set up the emission parameters # emission: {'feature':{'param': np.array(shape=(n_partial_states, n_hidden_states))}} self.emission = defaultdict(dict) for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: self.emission[name][param] = None self.emission = dict(self.emission) assert max_iter >= 0 assert thresh >= 0 self.max_iter = max_iter self.thresh = thresh assert init_spread >= 0 self.init_spread = init_spread if init_method not in _INIT_METHODS: raise ValueError('init_method must be one of', _INIT_METHODS) self.init_method = init_method if smoothing is None: smoothing = {'transmat': None, 'startprob': None} for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: smoothing['%s:%s' % (name, param)] = None elif type(smoothing) is str: s = smoothing smoothing = {'transmat': s, 'startprob': s} for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: smoothing['%s:%s' % (name, param)] = s elif type(smoothing) is dict: assert 'transmat' in smoothing.keys() assert 'startprob' in smoothing.keys() for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: assert param in smoothing.keys() or '%s:%s' % (name, param) in smoothing.keys() if param in smoothing.keys() and '%s:%s' % (name, param) not in smoothing.keys(): smoothing['%s:%s' % (name, param)] = smoothing[param] else: raise Exception('Wrong type for smoothing. Must be None, str, or dict') self.smoothing = smoothing self.random_state = random_state # Number of unique partial states is unknown until fit self.n_partial_states = None # Results after fitting the model self.logprob_ = None self.n_iter_performed_ = None self.logprob_delta_ = None # Mapping between p-states and a unique index # Defaults to 0 for unknown or missing p-states self.e = defaultdict(int) def _get_startprob(self): return np.exp(self._log_startprob) def _set_startprob(self, startprob): if startprob is None: startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states else: startprob = np.asarray(startprob, dtype=np.float) # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(startprob): startprob = normalize(startprob, axis=1) if len(startprob) != self.n_partial_states: raise ValueError('startprob must have length n_partial_states') if not np.allclose(np.sum(startprob, axis=1), 1.0): raise ValueError('startprob must sum to 1.0') self._log_startprob = np.log(np.asarray(startprob).copy()) startprob = property(_get_startprob, _set_startprob) def _get_steadyprob(self): return np.exp(self._log_steadyprob) def _set_steadyprob(self, steadyprob): if steadyprob is None: steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states else: steadyprob = np.asarray(steadyprob, dtype=np.float) # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(steadyprob): steadyprob = normalize(steadyprob, axis=1) if len(steadyprob) != self.n_partial_states: raise ValueError('steadyprob must have length n_partial_states') if not np.allclose(np.sum(steadyprob, axis=1), 1.0): raise ValueError('steadyprob must sum to 1.0') self._log_steadyprob = np.log(np.asarray(steadyprob).copy()) steadyprob = property(_get_steadyprob, _set_steadyprob) def _get_transmat(self): return np.exp(self._log_transmat) def _set_transmat(self, transmat): if transmat is None: transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) / self.n_hidden_states # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(transmat): transmat = normalize(transmat, axis=3) if (np.asarray(transmat).shape != (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)): raise ValueError('transmat must have shape ' '(n_partial_states,n_partial_states,n_hidden_states,n_hidden_states)') if not np.all(np.allclose(np.sum(transmat, axis=3), 1.0)): raise ValueError('Rows of transmat must sum to 1.0') self._log_transmat = np.log(np.asarray(transmat).copy()) underflow_idx = np.isnan(self._log_transmat) self._log_transmat[underflow_idx] = NEGINF transmat = property(_get_transmat, _set_transmat) def _compute_log_likelihood(self, obs, pstates_idx): q = np.zeros(shape=(len(obs), self.n_hidden_states, self.n_features)) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': mu = self.emission[feature_name]['mu'][pstates_idx] sigma = self.emission[feature_name]['sigma'][pstates_idx] for j in range(self.n_hidden_states): q[:, j, col] = np.log( np.maximum(MIN_PROBA, stats.norm.pdf(obs[:, col], loc=mu[:, j], scale=sigma[:, j]))) if feature_distr == 'lognormal': logmu = self.emission[feature_name]['logmu'][pstates_idx] logsigma = self.emission[feature_name]['logsigma'][pstates_idx] for j in range(self.n_hidden_states): q[:, j, col] = np.log(np.maximum(MIN_PROBA, stats.lognorm.pdf(obs[:, col], logsigma[:, j], loc=0, scale=np.exp(logmu[:, j])))) q = q.sum(axis=2) return q def _generate_sample_from_state(self, hidden_state, pstates_idx, random_state=None): sample = np.zeros(self.n_features) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': mu = self.emission[feature_name]['mu'][pstates_idx][hidden_state] sigma = self.emission[feature_name]['sigma'][pstates_idx][hidden_state] sample[col] = stats.norm.rvs(loc=mu, scale=sigma, random_state=random_state) if feature_distr == 'lognormal': logmu = self.emission[feature_name]['logmu'][pstates_idx][hidden_state] logsigma = self.emission[feature_name]['logsigma'][pstates_idx][hidden_state] sample[col] = stats.lognorm.rvs(logsigma, loc=0, scale=np.exp(logmu), random_state=random_state) return sample def _init_pstates(self, unique_pstates): # Map events to a unique index. The unknown p-state is at idx 0 self.e = defaultdict(int) self.e.update(dict(zip(np.sort(unique_pstates), range(1, len(unique_pstates) + 1)))) self.er = {v: k for k, v in self.e.items()} # Reverse lookup self.er[0] = UNKNOWN_PSTATE self.n_partial_states = len(self.e.keys()) + 1 # Add one for the unknown state return def _init_pstate_freqs(self, pstates_idx): # Partial state frequencies self.pstate_freq = Counter([idx for seq in pstates_idx for idx in seq]) self.pstate_trans_freq = Counter([(idx1, idx2) for seq in pstates_idx for idx1, idx2 in zip(seq[:-1], seq[1:])]) # Store freqs for the meta state self.pstate_freq[0] = len(np.concatenate(pstates_idx)) for seq in pstates_idx: self.pstate_startprob[seq[0]] += 1 self.pstate_steadyprob += np.bincount(seq, minlength=self.n_partial_states) for idx1, idx2 in zip(seq[:-1], seq[1:]): self.pstate_trans_freq[(0, 0)] += 1 self.pstate_trans_freq[(idx1, 0)] += 1 self.pstate_trans_freq[(0, idx2)] += 1 self.pstate_transmat[idx1, idx2] += 1 # TODO: separate probas from freqs # Normalize to get the probabilities, ignore the meta state at idx 0 self.pstate_startprob[1:] = normalize(self.pstate_startprob[1:]) self.pstate_transmat[1:, 1:] = normalize(self.pstate_transmat[1:, 1:], axis=1) self.pstate_steadyprob[1:] = normalize(self.pstate_steadyprob[1:]) return def _init_from_obs(self, obs, pstates_idx): # Partial state probabilities self.pstate_startprob = np.zeros(self.n_partial_states) self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states)) self.pstate_steadyprob = np.zeros(self.n_partial_states) # obs should be (N*T, n_features) # N is the number of samples # T is the size of each sample obs = np.concatenate(obs) pstates_idx = np.concatenate(pstates_idx) # Initialize starting and transition probas self.startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states self.steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states self.transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) / self.n_hidden_states # Initialize emission parameters # Hidden states are ordered by the first feature feature1 = self.emission_name[0] for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': self.emission[feature_name]['mu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['sigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) for idx in range(1, self.n_partial_states): idx_pstate = (pstates_idx == idx) if not np.any(idx_pstate): idx_pstate = np.arange(len(pstates_idx)) if feature_name == feature1: self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA) self.emission[feature_name]['mu'][idx] = obs[idx_pstate, col].mean() + \ obs[:, col].std() * np.linspace( -self.init_spread, self.init_spread, self.n_hidden_states) else: self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA) self.emission[feature_name]['mu'][idx, :] = obs[idx_pstate, col].mean() if feature_distr == 'lognormal': self.emission[feature_name]['logmu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['logsigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) for idx in range(1, self.n_partial_states): idx_pstate = (pstates_idx == idx) if not np.any(idx_pstate): idx_pstate = np.arange(len(pstates_idx)) if feature_name == feature1: self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(), MIN_PROBA) self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() + np.log( obs[:, col]).std() * np.linspace(-self.init_spread, self.init_spread, self.n_hidden_states) else: self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(), MIN_PROBA) self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() return def _init_random(self, random_state=None): if random_state is None: random_state = self.random_state random_state = check_random_state(random_state) self.pstate_startprob = np.zeros(self.n_partial_states) self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states)) self.pstate_steadyprob = np.zeros(self.n_partial_states) self.pstate_startprob[1:] = gen_stochastic_matrix(size=self.n_partial_states - 1, random_state=random_state) self.pstate_transmat[1:, 1:] = gen_stochastic_matrix( size=(self.n_partial_states - 1, self.n_partial_states - 1), random_state=random_state) self.pstate_steadyprob[1:] = steadystate(self.pstate_transmat[1:, 1:]) self.startprob = gen_stochastic_matrix(size=(self.n_partial_states, self.n_hidden_states), random_state=random_state) transmat = np.zeros((self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)): transmat[i, j] = gen_stochastic_matrix(size=(self.n_hidden_states, self.n_hidden_states), random_state=random_state) self.transmat = normalize(transmat, axis=3) # Initialize emission parameters for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): if feature_distr == 'normal': self.emission[feature_name]['mu'] = random_state.uniform(*_RANDINIT[feature_distr]['mu'], size=( self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['sigma'] = random_state.uniform(*_RANDINIT[feature_distr]['sigma'], size=( self.n_partial_states, self.n_hidden_states)) if feature_distr == 'lognormal': self.emission[feature_name]['logmu'] = random_state.uniform(*_RANDINIT[feature_distr]['logmu'], size=( self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['logsigma'] = random_state.uniform(*_RANDINIT[feature_distr]['logsigma'], size=(self.n_partial_states, self.n_hidden_states)) if self.emission_distr[0] == 'normal': self.emission[self.emission_name[0]]['mu'] = np.sort(self.emission[self.emission_name[0]]['mu'], axis=1) elif self.emission_distr[0] == 'lognormal': self.emission[self.emission_name[0]]['logmu'] = np.sort(self.emission[self.emission_name[0]]['logmu'], axis=1) return def _smooth(self): self._compute_marginals() startprob = self.startprob for j in range(1, self.n_partial_states): if 'freq' == self.smoothing['startprob']: w_ = 1 / (1 + self.pstate_freq[j]) w_j = 1 - w_ elif 'proba' == self.smoothing['startprob']: w_j = self.pstate_steadyprob[j] w_ = 1 - w_j elif 'exp' == self.smoothing['startprob']: w_ = np.exp(-self.pstate_freq[j]) w_j = 1 - w_ elif 'fixed' == self.smoothing['startprob']: w_ = 1 w_j = 0 elif self.smoothing['startprob'] is None: w_ = 0 w_j = 1 else: raise Exception('Wrong smoothing for startprob: ' + self.smoothing['startprob']) startprob[j] = w_j * self.startprob[j] + w_ * self.startprob[0] self.startprob = startprob transmat = self.transmat for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)): if 'freq' == self.smoothing['transmat']: w_i0 = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[0, j]) w_0j = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[i, 0]) w_ij = 1 - (w_i0 + w_0j) w_ = 0 elif 'proba' == self.smoothing['transmat']: denom = self.pstate_transmat[i, j] + self.pstate_transmat[i, :].sum() + self.pstate_transmat[:, j].sum() w_i0 = self.pstate_transmat[i, :].sum() / denom w_0j = self.pstate_transmat[:, j].sum() / denom w_ij = self.pstate_transmat[i, j] / denom w_ = 0 elif 'exp' == self.smoothing['transmat']: w_i0 = self.pstate_trans_freq[i, 0] * np.exp(-self.pstate_trans_freq[i, j]) w_0j = self.pstate_trans_freq[0, j] * np.exp(-self.pstate_trans_freq[i, j]) w_ = self.pstate_trans_freq[0, 0] * np.exp( -(self.pstate_trans_freq[i, 0] + self.pstate_trans_freq[0, j])) w_ij = self.pstate_trans_freq[i, j] w_ij, w_i0, w_0j, w_ = normalize(np.array([w_ij, w_i0, w_0j, w_])) elif 'fixed' == self.smoothing['transmat']: w_i0 = 0 w_0j = 0 w_ = 1 w_ij = 0 elif self.smoothing['transmat'] is None: w_i0 = 0 w_0j = 0 w_ = 0 w_ij = 1 else: raise Exception('Wrong smoothing for transmat: ' + self.smoothing['transmat']) assert (w_i0 + w_0j + w_ + w_ij) - 1 < TOLERANCE transmat[i, j] = w_ij * self.transmat[i, j] + w_i0 * self.transmat[i, 0] + w_0j * self.transmat[0, j] + w_ * \ self.transmat[ 0, 0] self.transmat = transmat assert np.all(self.startprob.sum(axis=1) - 1 < TOLERANCE) assert np.all(self.steadyprob.sum(axis=1) - 1 < TOLERANCE) assert np.all(self.transmat.sum(axis=3) - 1 < TOLERANCE) for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[feature_distr]: key = '%s:%s' % (feature_name, param) for j in range(1, self.n_partial_states): if 'freq' == self.smoothing[key]: w_ = 1 / (1 + self.pstate_freq[j]) w_j = 1 - w_ elif 'proba' == self.smoothing[key]: w_j = self.pstate_steadyprob[j] w_ = 1 - w_j elif 'exp' == self.smoothing[key]: w_ = np.exp(-self.pstate_freq[j]) w_j = 1 - w_ elif 'fixed' == self.smoothing[key]: w_ = 1 w_j = 0 elif self.smoothing[key] is None: w_ = 0 w_j = 1 else: raise Exception('Wrong smoothing for ' + key) self.emission[feature_name][param][j] = w_j * self.emission[feature_name][param][j] + w_ * \ self.emission[ feature_name][ param][0] return def _initialize_sufficient_statistics(self): stats = { 'nobs': 0, 'post': np.zeros((self.n_partial_states, self.n_hidden_states)), 'obs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'obs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'lnobs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'lnobs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'start': np.zeros((self.n_partial_states, self.n_hidden_states)), 'steady': np.zeros((self.n_partial_states, self.n_hidden_states)), 'trans': np.zeros( (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) } return stats def _accumulate_sufficient_statistics(self, stats, obs, pstates_idx, framelogprob, posteriors, fwdlattice, bwdlattice): stats['nobs'] += 1 n_observations, n_hidden_states = framelogprob.shape stats['start'][0] += posteriors[0] for i in range(self.n_partial_states): if len(np.where(pstates_idx == i)[0]) > 0: stats['start'][i] += posteriors[np.where(pstates_idx == i)[0].min()] if n_observations > 1: lneta = np.zeros((n_observations - 1, n_hidden_states, n_hidden_states)) _hmmc._compute_lneta(n_observations, n_hidden_states, pstates_idx, fwdlattice, self._log_transmat, bwdlattice, framelogprob, lneta) for i, j in unique_rows(np.c_[pstates_idx[:-1], pstates_idx[1:]]): if ((pstates_idx[:-1] == i) & (pstates_idx[1:] == j)).sum() > 0: stats['trans'][i, j] += np.exp( logsumexp(lneta[(pstates_idx[:-1] == i) & (pstates_idx[1:] == j)], axis=0)) for i in range(self.n_partial_states): stats['post'][i] += posteriors[pstates_idx == i].sum(axis=0) stats['obs'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i]) stats['obs**2'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i] ** 2) stats['lnobs'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i])) stats['lnobs**2'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i]) ** 2) return def _compute_marginals(self): # TODO: cythonize some of this # Start prob, weighted by p-state start probs self.startprob[0] = (self.pstate_startprob[1:, np.newaxis] * self.startprob[1:]).sum(axis=0) # Use the p-state transmat and transmat to get the full transmat full_transmat = ph2full(self.pstate_transmat[1:, 1:], self.transmat[1:, 1:]) full_steadyprob = steadystate(full_transmat) # Steady state probas are determined by the full trans mat, need to be updated steadyprob = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) steadyprob[0] = full_steadyprob.reshape(-1, self.n_hidden_states).sum(axis=0) for i in range(self.n_partial_states - 1): steadyprob[i + 1] = normalize( full_steadyprob[i * self.n_hidden_states:i * self.n_hidden_states + self.n_hidden_states]) self.steadyprob = steadyprob # Update the transations to/from the marginal state transmat = self.transmat # Group the hidden states within each partial state for hidx1, hidx2 in product(range(self.n_hidden_states), range(self.n_hidden_states)): transmat[0, 0][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, hidx2::self.n_hidden_states].sum() for pidx in range(self.n_partial_states - 1): transmat[pidx + 1, 0][hidx1, hidx2] = full_transmat[pidx * self.n_hidden_states + hidx1, hidx2::self.n_hidden_states].sum() transmat[0, pidx + 1][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, pidx * self.n_hidden_states + hidx2].sum() self.transmat = normalize(transmat, axis=3) pweights = self.pstate_steadyprob[1:, np.newaxis] # Update emission parameters for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): if feature_distr == 'normal': # Marginal state is a mixture of normals mu = self.emission[feature_name]['mu'][1:, :] sigma = self.emission[feature_name]['sigma'][1:, :] # Weighted mean and var mu_0 = (pweights * mu).sum(axis=0) self.emission[feature_name]['mu'][0, :] = mu_0 self.emission[feature_name]['sigma'][0, :] = np.sqrt( (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0)) if feature_distr == 'lognormal': # Marginal state is a mixture of normals mu = self.emission[feature_name]['logmu'][1:, :] sigma = self.emission[feature_name]['logsigma'][1:, :] # Weighted mean and var mu_0 = (pweights * mu).sum(axis=0) self.emission[feature_name]['logmu'][0, :] = mu_0 self.emission[feature_name]['logsigma'][0, :] = np.sqrt( (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0)) return def _do_mstep(self, stats): self.startprob = normalize(np.maximum(stats['start'], MIN_PROBA), axis=1) self.transmat = normalize(np.maximum(stats['trans'], MIN_PROBA), axis=3) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': denom = np.maximum(stats['post'], MIN_PROBA) mu = stats['obs'][:, :, col] / denom cv_num = (stats['obs**2'][:, :, col] - 2 * mu * stats['obs'][:, :, col] + mu ** 2 * denom) sigma = np.sqrt(cv_num / denom) sigma[np.isnan(sigma)] = MIN_PROBA self.emission[feature_name]['mu'] = mu self.emission[feature_name]['sigma'] = np.maximum(sigma, MIN_PROBA) if feature_distr == 'lognormal': denom = np.maximum(stats['post'], MIN_PROBA) mu = stats['lnobs'][:, :, col] / denom cv_num = (stats['lnobs**2'][:, :, col] - 2 * mu * stats['lnobs'][:, :, col] + mu ** 2 * denom) sigma = np.sqrt(cv_num / denom) sigma[np.isnan(sigma)] = MIN_PROBA self.emission[feature_name]['logmu'] = mu self.emission[feature_name]['logsigma'] = np.maximum(sigma, MIN_PROBA) return def _do_forward_pass(self, framelogprob, event_idx): n_observations, n_hidden_states = framelogprob.shape fwdlattice = np.zeros((n_observations, n_hidden_states)) _hmmc._forward(n_observations, n_hidden_states, event_idx, self._log_startprob, self._log_transmat, framelogprob, fwdlattice) return logsumexp(fwdlattice[-1]), fwdlattice def _do_backward_pass(self, framelogprob, event_idx): n_observations, n_hidden_states = framelogprob.shape bwdlattice = np.zeros((n_observations, n_hidden_states)) _hmmc._backward(n_observations, n_hidden_states, event_idx, self._log_startprob, self._log_transmat, framelogprob, bwdlattice) return bwdlattice def _do_viterbi_pass(self, framelogprob, event_idx): n_observations, n_components = framelogprob.shape state_sequence, logprob = _hmmc._viterbi( n_observations, n_components, event_idx, self._log_startprob, self._log_transmat, framelogprob) return logprob, state_sequence def rand(self, unique_pstates, random_state=None): """ Randomize the POHMM parameters """ self._init_pstates(unique_pstates) self._init_random(random_state=random_state) self._compute_marginals() return self def fit(self, obs, pstates, unique_pstates=None): """ Estimate model parameters. """ obs = [np.array(o) for o in obs] pstates = [np.array(p) for p in pstates] # List or array of observation sequences assert len(obs) == len(pstates) assert obs[0].ndim == 2 assert pstates[0].ndim == 1 if unique_pstates is not None: self._init_pstates(unique_pstates) else: self._init_pstates(list(set(np.concatenate(pstates)))) # Map the partial states to a unique index pstates_idx = [np.array([self.e[p] for p in seq]) for seq in pstates] if self.init_method == 'rand': self._init_random() elif self.init_method == 'obs': self._init_from_obs(obs, pstates_idx) self._init_pstate_freqs(pstates_idx) self._smooth() logprob = [] for i in range(self.max_iter): # Expectation step stats = self._initialize_sufficient_statistics() curr_logprob = 0 for obs_i, pstates_idx_i in zip(obs, pstates_idx): framelogprob = self._compute_log_likelihood(obs_i, pstates_idx_i) lpr, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx_i) bwdlattice = self._do_backward_pass(framelogprob, pstates_idx_i) gamma = fwdlattice + bwdlattice posteriors = np.exp(gamma.T - logsumexp(gamma, axis=1)).T curr_logprob += lpr self._accumulate_sufficient_statistics(stats, obs_i, pstates_idx_i, framelogprob, posteriors, fwdlattice, bwdlattice) logprob.append(curr_logprob) self.logprob_ = curr_logprob # Check for convergence. self.n_iter_performed_ = i if i > 0: self.logprob_delta = logprob[-1] - logprob[-2] if self.logprob_delta < self.thresh: break # Maximization step self._do_mstep(stats) # Mix the parameters self._smooth() return self def score(self, obs, pstates): """ Compute the log probability under the model. """ pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) logprob, _ = self._do_forward_pass(framelogprob, pstates_idx) return logprob def score_events(self, obs, pstates): """ Compute the log probability of each event under the model. """ pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) _, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx) L = logsumexp(fwdlattice, axis=1) return np.concatenate([L[[0]], np.diff(L)]) def predict_states(self, obs, pstates): pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) viterbi_logprob, state_sequence = self._do_viterbi_pass(framelogprob, pstates_idx) return viterbi_logprob, state_sequence def predict(self, obs, pstates, next_pstate=None): """ Predict the next observation """ assert len(obs) == len(pstates) pstates_idx = np.array([self.e[ei] for ei in pstates]) next_pstate_idx = self.e[next_pstate] if len(obs) == 0: # No history, use the starting probas next_hstate_prob = self.startprob[next_pstate_idx] else: # With a history, determine the hidden state posteriors using # the last posteriors and transition matrix framelogprob = self._compute_log_likelihood(obs, pstates_idx) _, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx) next_hstate_prob = np.zeros(self.n_hidden_states) alpha_n = fwdlattice[-1] vmax = alpha_n.max(axis=0) alpha_n = np.exp(alpha_n - vmax) alpha_n = alpha_n / alpha_n.sum() trans = self.transmat[pstates_idx[-1], next_pstate_idx] for i in range(self.n_hidden_states): next_hstate_prob[i] = np.sum([alpha_n[j] * trans[j, i] for j in range(self.n_hidden_states)]) assert next_hstate_prob.sum() - 1 < TOLERANCE # Make the prediction prediction = np.array( [self.expected_value(feature, pstate=next_pstate, hstate_prob=next_hstate_prob) for feature in self.emission_name]) # next_hstate = np.argmax(next_hstate_prob) # prediction = np.array( # [self.expected_value(feature, pstate=next_pstate, hstate=next_hstate) for feature in # self.emission_name]) return prediction def gen_pstates_idx(self, n, random_state=None): if random_state is None: random_state = self.random_state random_state = check_random_state(random_state) startprob_cdf = np.cumsum(self.pstate_startprob) transmat_cdf = np.cumsum(self.pstate_transmat, 1) # Initial state. rand = random_state.rand() curr_pstate = (startprob_cdf > rand).argmax() pstates = [curr_pstate] for _ in range(1, n): rand = random_state.rand() curr_pstate = (transmat_cdf[curr_pstate] > rand).argmax() pstates.append(curr_pstate) return np.array(pstates, dtype=int) def sample(self, pstates=None, n_obs=None, random_state=None): """ """ random_state = check_random_state(random_state) if pstates is None and n_obs is None: raise Exception('Must provide either pstates or n_obs') if pstates is not None and n_obs is not None: raise Exception('Must provide either pstates or n_obs but not both') gen_pstates = False rand = random_state.rand() if pstates is None: gen_pstates = True pstartprob_cdf = np.cumsum(self.pstate_startprob) ptransmat_cdf = np.cumsum(self.pstate_transmat, 1) # Initial pstate currpstate = (pstartprob_cdf > rand).argmax() pstates_idx = [currpstate] pstates = [self.er[currpstate]] else: n_obs = len(pstates) pstates_idx = np.array([self.e[p] for p in pstates]) startprob_pdf = self.startprob[pstates_idx[0]] startprob_cdf = np.cumsum(startprob_pdf) transmat_cdf = np.cumsum(self.transmat[0, pstates_idx[0]], 1) # Initial hidden state rand = random_state.rand() currstate = (startprob_cdf > rand).argmax() hidden_states = [currstate] obs = [self._generate_sample_from_state(currstate, pstates_idx[0], random_state)] for i in range(1, n_obs): rand = random_state.rand() if gen_pstates: currpstate = (ptransmat_cdf[currpstate] > rand).argmax() pstates_idx.append(currpstate) pstates.append(self.er[currpstate]) transmat_cdf = np.cumsum(self.transmat[pstates_idx[i - 1], pstates_idx[i]], 1) rand = random_state.rand() currstate = (transmat_cdf[currstate] > rand).argmax() hidden_states.append(currstate) obs.append(self._generate_sample_from_state(currstate, pstates_idx[i], random_state)) return np.array(obs), np.array(pstates), np.array(hidden_states, dtype=int) def score_df(self, df, pstate_col=PSTATE_COL): """ """ obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values return self.score(obs, pstates) def score_events_df(self, df, pstate_col=PSTATE_COL, score_col='score'): """ """ df = df.copy() obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values df[score_col] = self.score_events(obs, pstates) return df def predict_states_df(self, df, pstate_col=PSTATE_COL, hstate_col=HSTATE_COL): df = df.copy() obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values _, df[hstate_col] = self.predict_states(obs, pstates) return df def predict_df(self, df, next_pstate=None, pstate_col=PSTATE_COL): obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values return self.predict(obs, pstates, next_pstate) def sample_df(self, pstates=None, n_obs=None, random_state=None, pstate_col=PSTATE_COL, hstate_col=HSTATE_COL): """ Convenience function to generate samples a model and create a dataframe """ try: import pandas as pd except Exception as e: raise e obs, pstates, hstates = self.sample(pstates, n_obs, random_state) items = [] if pstate_col is not None: items.append((pstate_col, pstates)) if hstate_col is not None: items.append((hstate_col, hstates)) items = items + [(self.emission_name[i], obs[:, i]) for i in range(self.n_features)] df = pd.DataFrame.from_items(items) return df def __str__(self): pstates = sorted(self.e.keys()) sep = '-' * 80 + '\n' sep2 = '_' * 40 + '\n' out = 'POHMM\n' out += 'H-states: %d\n' % self.n_hidden_states out += 'P-states: (%d) %s\n' % (len(pstates), str(pstates)) out += 'Emission: %s\n' % (self.emission_distr) out += sep out += 'Transition probas\n' out += sep2 out += '. -> .\n%s\n' % str(self.transmat[0, 0]) for pstate in pstates: out += sep2 out += '%s -> .\n%s\n' % (pstate, str(self.transmat[self.e[pstate], 0])) out += sep2 out += '. -> %s\n%s\n' % (pstate, str(self.transmat[0, self.e[pstate]])) for pstate1, pstate2 in product(pstates, pstates): out += sep2 out += '%s -> %s\n%s\n' % (pstate1, pstate2, str(self.transmat[self.e[pstate1], self.e[pstate2]])) out += sep out += 'Starting probas\n' out += '.: %s\n' % str(self.startprob[0]) for pstate in pstates: out += '%s: %s\n' % (pstate, str(self.startprob[self.e[pstate]])) out += sep out += 'Steady probas\n' out += '.: %s\n' % str(self.steadyprob[0]) for pstate in pstates: out += '%s: %s\n' % (pstate, str(self.steadyprob[self.e[pstate]])) out += sep out += 'Emissions\n' for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): out += sep2 out += 'Feature %s: %s\n' % (feature_name, feature_distr) for param in _DISTRIBUTIONS[feature_distr]: out += '.: %s = %s\n' % (param, str(self.emission[feature_name][param][0])) for pstate in pstates: out += '%s: %s = %s\n' % (pstate, param, str(self.emission[feature_name][param][self.e[pstate]])) out += sep return out def expected_value(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): """ Determine the joint maximum likelihood estimate """ # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': return np.sum(w * expected_lognormal(self.emission[feature]['logsigma'].flatten(), self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': return np.sum(w * expected_normal(self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten())) elif hstate is None: # Marginalize hstate if self.emission_name_distr[feature] == 'lognormal': return np.sum(hstate_prob * expected_lognormal(self.emission[feature]['logsigma'][pstate_idx, :], self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': return np.sum(hstate_prob * expected_normal(self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :])) elif pstate is None: # Marginalize pstate if self.emission_name_distr[feature] == 'lognormal': return np.sum(pstate_prob * expected_lognormal(self.emission[feature]['logsigma'][:, hstate], self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': return np.sum(pstate_prob * expected_normal(self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate])) else: if self.emission_name_distr[feature] == 'lognormal': return expected_lognormal(self.emission[feature]['logsigma'][pstate_idx, hstate], self.emission[feature]['logmu'][pstate_idx, hstate]) elif self.emission_name_distr[feature] == 'normal': return expected_normal(self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) return def pdf_fn(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'].flatten(), 0, np.exp(self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten()) elif hstate is None: # Marginalize hstate w = hstate_prob if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][pstate_idx, :], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :]) elif pstate is None: # Marginalize pstate w = pstate_prob if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][:, hstate], 0, np.exp(self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate]) else: w = 1 if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][pstate_idx, hstate], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, hstate])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) def fn(x): if np.isscalar(x): p = np.sum(w * pdf(x)) else: x = np.array(x) p = np.zeros(len(x)) for i, xi in enumerate(x): p[i] = np.sum(w * pdf(xi)) return p return fn def cdf_fn(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'].flatten(), 0, np.exp(self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten()) elif hstate is None: # Marginalize hstate w = hstate_prob if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][pstate_idx, :], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :]) elif pstate is None: # Marginalize pstate w = pstate_prob if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][:, hstate], 0, np.exp(self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate]) else: w = 1 if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][pstate_idx, hstate], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, hstate])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) def fn(x): if np.isscalar(x): p = np.sum(w * cdf(x)) else: x = np.array(x) p = np.zeros(len(x)) for i, xi in enumerate(x): p[i] = np.sum(w * cdf(xi)) return p return fn def params(self, pstates=None): if pstates is None: pstates = [None] + sorted( set(self.er.values())) # TODO: self.e caches any unknown value, maybe it shouldn't? params = [] # emission parameters for hstate, pstate_label in product(range(self.n_hidden_states), pstates): for feature, distr in zip(self.emission_name, self.emission_distr): for feature_param in _DISTRIBUTIONS[distr]: params.append(self.emission[feature][feature_param][self.e[pstate_label], hstate]) # transition parameters, diagonals only assuming 2 state for hstate, pstate_label in product(range(self.n_hidden_states), pstates): params.append(self.transmat[self.e[pstate_label], self.e[pstate_label], hstate, hstate]) return np.array(params)
vmonaco/pohmm
pohmm/pohmm.py
Pohmm.sample_df
python
def sample_df(self, pstates=None, n_obs=None, random_state=None, pstate_col=PSTATE_COL, hstate_col=HSTATE_COL): try: import pandas as pd except Exception as e: raise e obs, pstates, hstates = self.sample(pstates, n_obs, random_state) items = [] if pstate_col is not None: items.append((pstate_col, pstates)) if hstate_col is not None: items.append((hstate_col, hstates)) items = items + [(self.emission_name[i], obs[:, i]) for i in range(self.n_features)] df = pd.DataFrame.from_items(items) return df
Convenience function to generate samples a model and create a dataframe
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/pohmm.py#L919-L939
[ "def sample(self, pstates=None, n_obs=None, random_state=None):\n \"\"\"\n\n \"\"\"\n random_state = check_random_state(random_state)\n\n if pstates is None and n_obs is None:\n raise Exception('Must provide either pstates or n_obs')\n\n if pstates is not None and n_obs is not None:\n raise Exception('Must provide either pstates or n_obs but not both')\n\n gen_pstates = False\n rand = random_state.rand()\n if pstates is None:\n gen_pstates = True\n pstartprob_cdf = np.cumsum(self.pstate_startprob)\n ptransmat_cdf = np.cumsum(self.pstate_transmat, 1)\n\n # Initial pstate\n currpstate = (pstartprob_cdf > rand).argmax()\n pstates_idx = [currpstate]\n pstates = [self.er[currpstate]]\n else:\n n_obs = len(pstates)\n pstates_idx = np.array([self.e[p] for p in pstates])\n\n startprob_pdf = self.startprob[pstates_idx[0]]\n startprob_cdf = np.cumsum(startprob_pdf)\n transmat_cdf = np.cumsum(self.transmat[0, pstates_idx[0]], 1)\n\n # Initial hidden state\n rand = random_state.rand()\n currstate = (startprob_cdf > rand).argmax()\n hidden_states = [currstate]\n obs = [self._generate_sample_from_state(currstate, pstates_idx[0], random_state)]\n\n for i in range(1, n_obs):\n rand = random_state.rand()\n\n if gen_pstates:\n currpstate = (ptransmat_cdf[currpstate] > rand).argmax()\n pstates_idx.append(currpstate)\n pstates.append(self.er[currpstate])\n\n transmat_cdf = np.cumsum(self.transmat[pstates_idx[i - 1], pstates_idx[i]], 1)\n rand = random_state.rand()\n currstate = (transmat_cdf[currstate] > rand).argmax()\n hidden_states.append(currstate)\n obs.append(self._generate_sample_from_state(currstate, pstates_idx[i], random_state))\n\n return np.array(obs), np.array(pstates), np.array(hidden_states, dtype=int)\n" ]
class Pohmm(object): """ Partially observable hidden Markov model """ def __init__(self, n_hidden_states=2, emissions=['normal'], max_iter=1000, thresh=1e-6, init_method='obs', init_spread=2, smoothing=None, random_state=None): if type(n_hidden_states) is int: self.n_hidden_states = n_hidden_states else: raise Exception('Wrong type for n_hidden_states. Must be int') if type(emissions[0]) is tuple: emission_name, emission_distr = zip(*emissions) elif type(emissions[0]) is str: emission_name, emission_distr = np.arange(len(emissions)), emissions for distr in emission_distr: if distr not in _DISTRIBUTIONS.keys(): raise ValueError('Emission distribution must be one of', _DISTRIBUTIONS.keys()) self.emission_name = emission_name self.emission_distr = emission_distr self.emission_name_distr = dict(zip(emission_name, emission_distr)) self.n_features = len(emissions) # Set up the emission parameters # emission: {'feature':{'param': np.array(shape=(n_partial_states, n_hidden_states))}} self.emission = defaultdict(dict) for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: self.emission[name][param] = None self.emission = dict(self.emission) assert max_iter >= 0 assert thresh >= 0 self.max_iter = max_iter self.thresh = thresh assert init_spread >= 0 self.init_spread = init_spread if init_method not in _INIT_METHODS: raise ValueError('init_method must be one of', _INIT_METHODS) self.init_method = init_method if smoothing is None: smoothing = {'transmat': None, 'startprob': None} for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: smoothing['%s:%s' % (name, param)] = None elif type(smoothing) is str: s = smoothing smoothing = {'transmat': s, 'startprob': s} for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: smoothing['%s:%s' % (name, param)] = s elif type(smoothing) is dict: assert 'transmat' in smoothing.keys() assert 'startprob' in smoothing.keys() for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: assert param in smoothing.keys() or '%s:%s' % (name, param) in smoothing.keys() if param in smoothing.keys() and '%s:%s' % (name, param) not in smoothing.keys(): smoothing['%s:%s' % (name, param)] = smoothing[param] else: raise Exception('Wrong type for smoothing. Must be None, str, or dict') self.smoothing = smoothing self.random_state = random_state # Number of unique partial states is unknown until fit self.n_partial_states = None # Results after fitting the model self.logprob_ = None self.n_iter_performed_ = None self.logprob_delta_ = None # Mapping between p-states and a unique index # Defaults to 0 for unknown or missing p-states self.e = defaultdict(int) def _get_startprob(self): return np.exp(self._log_startprob) def _set_startprob(self, startprob): if startprob is None: startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states else: startprob = np.asarray(startprob, dtype=np.float) # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(startprob): startprob = normalize(startprob, axis=1) if len(startprob) != self.n_partial_states: raise ValueError('startprob must have length n_partial_states') if not np.allclose(np.sum(startprob, axis=1), 1.0): raise ValueError('startprob must sum to 1.0') self._log_startprob = np.log(np.asarray(startprob).copy()) startprob = property(_get_startprob, _set_startprob) def _get_steadyprob(self): return np.exp(self._log_steadyprob) def _set_steadyprob(self, steadyprob): if steadyprob is None: steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states else: steadyprob = np.asarray(steadyprob, dtype=np.float) # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(steadyprob): steadyprob = normalize(steadyprob, axis=1) if len(steadyprob) != self.n_partial_states: raise ValueError('steadyprob must have length n_partial_states') if not np.allclose(np.sum(steadyprob, axis=1), 1.0): raise ValueError('steadyprob must sum to 1.0') self._log_steadyprob = np.log(np.asarray(steadyprob).copy()) steadyprob = property(_get_steadyprob, _set_steadyprob) def _get_transmat(self): return np.exp(self._log_transmat) def _set_transmat(self, transmat): if transmat is None: transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) / self.n_hidden_states # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(transmat): transmat = normalize(transmat, axis=3) if (np.asarray(transmat).shape != (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)): raise ValueError('transmat must have shape ' '(n_partial_states,n_partial_states,n_hidden_states,n_hidden_states)') if not np.all(np.allclose(np.sum(transmat, axis=3), 1.0)): raise ValueError('Rows of transmat must sum to 1.0') self._log_transmat = np.log(np.asarray(transmat).copy()) underflow_idx = np.isnan(self._log_transmat) self._log_transmat[underflow_idx] = NEGINF transmat = property(_get_transmat, _set_transmat) def _compute_log_likelihood(self, obs, pstates_idx): q = np.zeros(shape=(len(obs), self.n_hidden_states, self.n_features)) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': mu = self.emission[feature_name]['mu'][pstates_idx] sigma = self.emission[feature_name]['sigma'][pstates_idx] for j in range(self.n_hidden_states): q[:, j, col] = np.log( np.maximum(MIN_PROBA, stats.norm.pdf(obs[:, col], loc=mu[:, j], scale=sigma[:, j]))) if feature_distr == 'lognormal': logmu = self.emission[feature_name]['logmu'][pstates_idx] logsigma = self.emission[feature_name]['logsigma'][pstates_idx] for j in range(self.n_hidden_states): q[:, j, col] = np.log(np.maximum(MIN_PROBA, stats.lognorm.pdf(obs[:, col], logsigma[:, j], loc=0, scale=np.exp(logmu[:, j])))) q = q.sum(axis=2) return q def _generate_sample_from_state(self, hidden_state, pstates_idx, random_state=None): sample = np.zeros(self.n_features) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': mu = self.emission[feature_name]['mu'][pstates_idx][hidden_state] sigma = self.emission[feature_name]['sigma'][pstates_idx][hidden_state] sample[col] = stats.norm.rvs(loc=mu, scale=sigma, random_state=random_state) if feature_distr == 'lognormal': logmu = self.emission[feature_name]['logmu'][pstates_idx][hidden_state] logsigma = self.emission[feature_name]['logsigma'][pstates_idx][hidden_state] sample[col] = stats.lognorm.rvs(logsigma, loc=0, scale=np.exp(logmu), random_state=random_state) return sample def _init_pstates(self, unique_pstates): # Map events to a unique index. The unknown p-state is at idx 0 self.e = defaultdict(int) self.e.update(dict(zip(np.sort(unique_pstates), range(1, len(unique_pstates) + 1)))) self.er = {v: k for k, v in self.e.items()} # Reverse lookup self.er[0] = UNKNOWN_PSTATE self.n_partial_states = len(self.e.keys()) + 1 # Add one for the unknown state return def _init_pstate_freqs(self, pstates_idx): # Partial state frequencies self.pstate_freq = Counter([idx for seq in pstates_idx for idx in seq]) self.pstate_trans_freq = Counter([(idx1, idx2) for seq in pstates_idx for idx1, idx2 in zip(seq[:-1], seq[1:])]) # Store freqs for the meta state self.pstate_freq[0] = len(np.concatenate(pstates_idx)) for seq in pstates_idx: self.pstate_startprob[seq[0]] += 1 self.pstate_steadyprob += np.bincount(seq, minlength=self.n_partial_states) for idx1, idx2 in zip(seq[:-1], seq[1:]): self.pstate_trans_freq[(0, 0)] += 1 self.pstate_trans_freq[(idx1, 0)] += 1 self.pstate_trans_freq[(0, idx2)] += 1 self.pstate_transmat[idx1, idx2] += 1 # TODO: separate probas from freqs # Normalize to get the probabilities, ignore the meta state at idx 0 self.pstate_startprob[1:] = normalize(self.pstate_startprob[1:]) self.pstate_transmat[1:, 1:] = normalize(self.pstate_transmat[1:, 1:], axis=1) self.pstate_steadyprob[1:] = normalize(self.pstate_steadyprob[1:]) return def _init_from_obs(self, obs, pstates_idx): # Partial state probabilities self.pstate_startprob = np.zeros(self.n_partial_states) self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states)) self.pstate_steadyprob = np.zeros(self.n_partial_states) # obs should be (N*T, n_features) # N is the number of samples # T is the size of each sample obs = np.concatenate(obs) pstates_idx = np.concatenate(pstates_idx) # Initialize starting and transition probas self.startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states self.steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states self.transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) / self.n_hidden_states # Initialize emission parameters # Hidden states are ordered by the first feature feature1 = self.emission_name[0] for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': self.emission[feature_name]['mu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['sigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) for idx in range(1, self.n_partial_states): idx_pstate = (pstates_idx == idx) if not np.any(idx_pstate): idx_pstate = np.arange(len(pstates_idx)) if feature_name == feature1: self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA) self.emission[feature_name]['mu'][idx] = obs[idx_pstate, col].mean() + \ obs[:, col].std() * np.linspace( -self.init_spread, self.init_spread, self.n_hidden_states) else: self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA) self.emission[feature_name]['mu'][idx, :] = obs[idx_pstate, col].mean() if feature_distr == 'lognormal': self.emission[feature_name]['logmu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['logsigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) for idx in range(1, self.n_partial_states): idx_pstate = (pstates_idx == idx) if not np.any(idx_pstate): idx_pstate = np.arange(len(pstates_idx)) if feature_name == feature1: self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(), MIN_PROBA) self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() + np.log( obs[:, col]).std() * np.linspace(-self.init_spread, self.init_spread, self.n_hidden_states) else: self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(), MIN_PROBA) self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() return def _init_random(self, random_state=None): if random_state is None: random_state = self.random_state random_state = check_random_state(random_state) self.pstate_startprob = np.zeros(self.n_partial_states) self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states)) self.pstate_steadyprob = np.zeros(self.n_partial_states) self.pstate_startprob[1:] = gen_stochastic_matrix(size=self.n_partial_states - 1, random_state=random_state) self.pstate_transmat[1:, 1:] = gen_stochastic_matrix( size=(self.n_partial_states - 1, self.n_partial_states - 1), random_state=random_state) self.pstate_steadyprob[1:] = steadystate(self.pstate_transmat[1:, 1:]) self.startprob = gen_stochastic_matrix(size=(self.n_partial_states, self.n_hidden_states), random_state=random_state) transmat = np.zeros((self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)): transmat[i, j] = gen_stochastic_matrix(size=(self.n_hidden_states, self.n_hidden_states), random_state=random_state) self.transmat = normalize(transmat, axis=3) # Initialize emission parameters for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): if feature_distr == 'normal': self.emission[feature_name]['mu'] = random_state.uniform(*_RANDINIT[feature_distr]['mu'], size=( self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['sigma'] = random_state.uniform(*_RANDINIT[feature_distr]['sigma'], size=( self.n_partial_states, self.n_hidden_states)) if feature_distr == 'lognormal': self.emission[feature_name]['logmu'] = random_state.uniform(*_RANDINIT[feature_distr]['logmu'], size=( self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['logsigma'] = random_state.uniform(*_RANDINIT[feature_distr]['logsigma'], size=(self.n_partial_states, self.n_hidden_states)) if self.emission_distr[0] == 'normal': self.emission[self.emission_name[0]]['mu'] = np.sort(self.emission[self.emission_name[0]]['mu'], axis=1) elif self.emission_distr[0] == 'lognormal': self.emission[self.emission_name[0]]['logmu'] = np.sort(self.emission[self.emission_name[0]]['logmu'], axis=1) return def _smooth(self): self._compute_marginals() startprob = self.startprob for j in range(1, self.n_partial_states): if 'freq' == self.smoothing['startprob']: w_ = 1 / (1 + self.pstate_freq[j]) w_j = 1 - w_ elif 'proba' == self.smoothing['startprob']: w_j = self.pstate_steadyprob[j] w_ = 1 - w_j elif 'exp' == self.smoothing['startprob']: w_ = np.exp(-self.pstate_freq[j]) w_j = 1 - w_ elif 'fixed' == self.smoothing['startprob']: w_ = 1 w_j = 0 elif self.smoothing['startprob'] is None: w_ = 0 w_j = 1 else: raise Exception('Wrong smoothing for startprob: ' + self.smoothing['startprob']) startprob[j] = w_j * self.startprob[j] + w_ * self.startprob[0] self.startprob = startprob transmat = self.transmat for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)): if 'freq' == self.smoothing['transmat']: w_i0 = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[0, j]) w_0j = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[i, 0]) w_ij = 1 - (w_i0 + w_0j) w_ = 0 elif 'proba' == self.smoothing['transmat']: denom = self.pstate_transmat[i, j] + self.pstate_transmat[i, :].sum() + self.pstate_transmat[:, j].sum() w_i0 = self.pstate_transmat[i, :].sum() / denom w_0j = self.pstate_transmat[:, j].sum() / denom w_ij = self.pstate_transmat[i, j] / denom w_ = 0 elif 'exp' == self.smoothing['transmat']: w_i0 = self.pstate_trans_freq[i, 0] * np.exp(-self.pstate_trans_freq[i, j]) w_0j = self.pstate_trans_freq[0, j] * np.exp(-self.pstate_trans_freq[i, j]) w_ = self.pstate_trans_freq[0, 0] * np.exp( -(self.pstate_trans_freq[i, 0] + self.pstate_trans_freq[0, j])) w_ij = self.pstate_trans_freq[i, j] w_ij, w_i0, w_0j, w_ = normalize(np.array([w_ij, w_i0, w_0j, w_])) elif 'fixed' == self.smoothing['transmat']: w_i0 = 0 w_0j = 0 w_ = 1 w_ij = 0 elif self.smoothing['transmat'] is None: w_i0 = 0 w_0j = 0 w_ = 0 w_ij = 1 else: raise Exception('Wrong smoothing for transmat: ' + self.smoothing['transmat']) assert (w_i0 + w_0j + w_ + w_ij) - 1 < TOLERANCE transmat[i, j] = w_ij * self.transmat[i, j] + w_i0 * self.transmat[i, 0] + w_0j * self.transmat[0, j] + w_ * \ self.transmat[ 0, 0] self.transmat = transmat assert np.all(self.startprob.sum(axis=1) - 1 < TOLERANCE) assert np.all(self.steadyprob.sum(axis=1) - 1 < TOLERANCE) assert np.all(self.transmat.sum(axis=3) - 1 < TOLERANCE) for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[feature_distr]: key = '%s:%s' % (feature_name, param) for j in range(1, self.n_partial_states): if 'freq' == self.smoothing[key]: w_ = 1 / (1 + self.pstate_freq[j]) w_j = 1 - w_ elif 'proba' == self.smoothing[key]: w_j = self.pstate_steadyprob[j] w_ = 1 - w_j elif 'exp' == self.smoothing[key]: w_ = np.exp(-self.pstate_freq[j]) w_j = 1 - w_ elif 'fixed' == self.smoothing[key]: w_ = 1 w_j = 0 elif self.smoothing[key] is None: w_ = 0 w_j = 1 else: raise Exception('Wrong smoothing for ' + key) self.emission[feature_name][param][j] = w_j * self.emission[feature_name][param][j] + w_ * \ self.emission[ feature_name][ param][0] return def _initialize_sufficient_statistics(self): stats = { 'nobs': 0, 'post': np.zeros((self.n_partial_states, self.n_hidden_states)), 'obs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'obs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'lnobs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'lnobs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'start': np.zeros((self.n_partial_states, self.n_hidden_states)), 'steady': np.zeros((self.n_partial_states, self.n_hidden_states)), 'trans': np.zeros( (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) } return stats def _accumulate_sufficient_statistics(self, stats, obs, pstates_idx, framelogprob, posteriors, fwdlattice, bwdlattice): stats['nobs'] += 1 n_observations, n_hidden_states = framelogprob.shape stats['start'][0] += posteriors[0] for i in range(self.n_partial_states): if len(np.where(pstates_idx == i)[0]) > 0: stats['start'][i] += posteriors[np.where(pstates_idx == i)[0].min()] if n_observations > 1: lneta = np.zeros((n_observations - 1, n_hidden_states, n_hidden_states)) _hmmc._compute_lneta(n_observations, n_hidden_states, pstates_idx, fwdlattice, self._log_transmat, bwdlattice, framelogprob, lneta) for i, j in unique_rows(np.c_[pstates_idx[:-1], pstates_idx[1:]]): if ((pstates_idx[:-1] == i) & (pstates_idx[1:] == j)).sum() > 0: stats['trans'][i, j] += np.exp( logsumexp(lneta[(pstates_idx[:-1] == i) & (pstates_idx[1:] == j)], axis=0)) for i in range(self.n_partial_states): stats['post'][i] += posteriors[pstates_idx == i].sum(axis=0) stats['obs'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i]) stats['obs**2'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i] ** 2) stats['lnobs'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i])) stats['lnobs**2'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i]) ** 2) return def _compute_marginals(self): # TODO: cythonize some of this # Start prob, weighted by p-state start probs self.startprob[0] = (self.pstate_startprob[1:, np.newaxis] * self.startprob[1:]).sum(axis=0) # Use the p-state transmat and transmat to get the full transmat full_transmat = ph2full(self.pstate_transmat[1:, 1:], self.transmat[1:, 1:]) full_steadyprob = steadystate(full_transmat) # Steady state probas are determined by the full trans mat, need to be updated steadyprob = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) steadyprob[0] = full_steadyprob.reshape(-1, self.n_hidden_states).sum(axis=0) for i in range(self.n_partial_states - 1): steadyprob[i + 1] = normalize( full_steadyprob[i * self.n_hidden_states:i * self.n_hidden_states + self.n_hidden_states]) self.steadyprob = steadyprob # Update the transations to/from the marginal state transmat = self.transmat # Group the hidden states within each partial state for hidx1, hidx2 in product(range(self.n_hidden_states), range(self.n_hidden_states)): transmat[0, 0][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, hidx2::self.n_hidden_states].sum() for pidx in range(self.n_partial_states - 1): transmat[pidx + 1, 0][hidx1, hidx2] = full_transmat[pidx * self.n_hidden_states + hidx1, hidx2::self.n_hidden_states].sum() transmat[0, pidx + 1][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, pidx * self.n_hidden_states + hidx2].sum() self.transmat = normalize(transmat, axis=3) pweights = self.pstate_steadyprob[1:, np.newaxis] # Update emission parameters for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): if feature_distr == 'normal': # Marginal state is a mixture of normals mu = self.emission[feature_name]['mu'][1:, :] sigma = self.emission[feature_name]['sigma'][1:, :] # Weighted mean and var mu_0 = (pweights * mu).sum(axis=0) self.emission[feature_name]['mu'][0, :] = mu_0 self.emission[feature_name]['sigma'][0, :] = np.sqrt( (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0)) if feature_distr == 'lognormal': # Marginal state is a mixture of normals mu = self.emission[feature_name]['logmu'][1:, :] sigma = self.emission[feature_name]['logsigma'][1:, :] # Weighted mean and var mu_0 = (pweights * mu).sum(axis=0) self.emission[feature_name]['logmu'][0, :] = mu_0 self.emission[feature_name]['logsigma'][0, :] = np.sqrt( (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0)) return def _do_mstep(self, stats): self.startprob = normalize(np.maximum(stats['start'], MIN_PROBA), axis=1) self.transmat = normalize(np.maximum(stats['trans'], MIN_PROBA), axis=3) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': denom = np.maximum(stats['post'], MIN_PROBA) mu = stats['obs'][:, :, col] / denom cv_num = (stats['obs**2'][:, :, col] - 2 * mu * stats['obs'][:, :, col] + mu ** 2 * denom) sigma = np.sqrt(cv_num / denom) sigma[np.isnan(sigma)] = MIN_PROBA self.emission[feature_name]['mu'] = mu self.emission[feature_name]['sigma'] = np.maximum(sigma, MIN_PROBA) if feature_distr == 'lognormal': denom = np.maximum(stats['post'], MIN_PROBA) mu = stats['lnobs'][:, :, col] / denom cv_num = (stats['lnobs**2'][:, :, col] - 2 * mu * stats['lnobs'][:, :, col] + mu ** 2 * denom) sigma = np.sqrt(cv_num / denom) sigma[np.isnan(sigma)] = MIN_PROBA self.emission[feature_name]['logmu'] = mu self.emission[feature_name]['logsigma'] = np.maximum(sigma, MIN_PROBA) return def _do_forward_pass(self, framelogprob, event_idx): n_observations, n_hidden_states = framelogprob.shape fwdlattice = np.zeros((n_observations, n_hidden_states)) _hmmc._forward(n_observations, n_hidden_states, event_idx, self._log_startprob, self._log_transmat, framelogprob, fwdlattice) return logsumexp(fwdlattice[-1]), fwdlattice def _do_backward_pass(self, framelogprob, event_idx): n_observations, n_hidden_states = framelogprob.shape bwdlattice = np.zeros((n_observations, n_hidden_states)) _hmmc._backward(n_observations, n_hidden_states, event_idx, self._log_startprob, self._log_transmat, framelogprob, bwdlattice) return bwdlattice def _do_viterbi_pass(self, framelogprob, event_idx): n_observations, n_components = framelogprob.shape state_sequence, logprob = _hmmc._viterbi( n_observations, n_components, event_idx, self._log_startprob, self._log_transmat, framelogprob) return logprob, state_sequence def rand(self, unique_pstates, random_state=None): """ Randomize the POHMM parameters """ self._init_pstates(unique_pstates) self._init_random(random_state=random_state) self._compute_marginals() return self def fit(self, obs, pstates, unique_pstates=None): """ Estimate model parameters. """ obs = [np.array(o) for o in obs] pstates = [np.array(p) for p in pstates] # List or array of observation sequences assert len(obs) == len(pstates) assert obs[0].ndim == 2 assert pstates[0].ndim == 1 if unique_pstates is not None: self._init_pstates(unique_pstates) else: self._init_pstates(list(set(np.concatenate(pstates)))) # Map the partial states to a unique index pstates_idx = [np.array([self.e[p] for p in seq]) for seq in pstates] if self.init_method == 'rand': self._init_random() elif self.init_method == 'obs': self._init_from_obs(obs, pstates_idx) self._init_pstate_freqs(pstates_idx) self._smooth() logprob = [] for i in range(self.max_iter): # Expectation step stats = self._initialize_sufficient_statistics() curr_logprob = 0 for obs_i, pstates_idx_i in zip(obs, pstates_idx): framelogprob = self._compute_log_likelihood(obs_i, pstates_idx_i) lpr, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx_i) bwdlattice = self._do_backward_pass(framelogprob, pstates_idx_i) gamma = fwdlattice + bwdlattice posteriors = np.exp(gamma.T - logsumexp(gamma, axis=1)).T curr_logprob += lpr self._accumulate_sufficient_statistics(stats, obs_i, pstates_idx_i, framelogprob, posteriors, fwdlattice, bwdlattice) logprob.append(curr_logprob) self.logprob_ = curr_logprob # Check for convergence. self.n_iter_performed_ = i if i > 0: self.logprob_delta = logprob[-1] - logprob[-2] if self.logprob_delta < self.thresh: break # Maximization step self._do_mstep(stats) # Mix the parameters self._smooth() return self def score(self, obs, pstates): """ Compute the log probability under the model. """ pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) logprob, _ = self._do_forward_pass(framelogprob, pstates_idx) return logprob def score_events(self, obs, pstates): """ Compute the log probability of each event under the model. """ pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) _, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx) L = logsumexp(fwdlattice, axis=1) return np.concatenate([L[[0]], np.diff(L)]) def predict_states(self, obs, pstates): pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) viterbi_logprob, state_sequence = self._do_viterbi_pass(framelogprob, pstates_idx) return viterbi_logprob, state_sequence def predict(self, obs, pstates, next_pstate=None): """ Predict the next observation """ assert len(obs) == len(pstates) pstates_idx = np.array([self.e[ei] for ei in pstates]) next_pstate_idx = self.e[next_pstate] if len(obs) == 0: # No history, use the starting probas next_hstate_prob = self.startprob[next_pstate_idx] else: # With a history, determine the hidden state posteriors using # the last posteriors and transition matrix framelogprob = self._compute_log_likelihood(obs, pstates_idx) _, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx) next_hstate_prob = np.zeros(self.n_hidden_states) alpha_n = fwdlattice[-1] vmax = alpha_n.max(axis=0) alpha_n = np.exp(alpha_n - vmax) alpha_n = alpha_n / alpha_n.sum() trans = self.transmat[pstates_idx[-1], next_pstate_idx] for i in range(self.n_hidden_states): next_hstate_prob[i] = np.sum([alpha_n[j] * trans[j, i] for j in range(self.n_hidden_states)]) assert next_hstate_prob.sum() - 1 < TOLERANCE # Make the prediction prediction = np.array( [self.expected_value(feature, pstate=next_pstate, hstate_prob=next_hstate_prob) for feature in self.emission_name]) # next_hstate = np.argmax(next_hstate_prob) # prediction = np.array( # [self.expected_value(feature, pstate=next_pstate, hstate=next_hstate) for feature in # self.emission_name]) return prediction def gen_pstates_idx(self, n, random_state=None): if random_state is None: random_state = self.random_state random_state = check_random_state(random_state) startprob_cdf = np.cumsum(self.pstate_startprob) transmat_cdf = np.cumsum(self.pstate_transmat, 1) # Initial state. rand = random_state.rand() curr_pstate = (startprob_cdf > rand).argmax() pstates = [curr_pstate] for _ in range(1, n): rand = random_state.rand() curr_pstate = (transmat_cdf[curr_pstate] > rand).argmax() pstates.append(curr_pstate) return np.array(pstates, dtype=int) def sample(self, pstates=None, n_obs=None, random_state=None): """ """ random_state = check_random_state(random_state) if pstates is None and n_obs is None: raise Exception('Must provide either pstates or n_obs') if pstates is not None and n_obs is not None: raise Exception('Must provide either pstates or n_obs but not both') gen_pstates = False rand = random_state.rand() if pstates is None: gen_pstates = True pstartprob_cdf = np.cumsum(self.pstate_startprob) ptransmat_cdf = np.cumsum(self.pstate_transmat, 1) # Initial pstate currpstate = (pstartprob_cdf > rand).argmax() pstates_idx = [currpstate] pstates = [self.er[currpstate]] else: n_obs = len(pstates) pstates_idx = np.array([self.e[p] for p in pstates]) startprob_pdf = self.startprob[pstates_idx[0]] startprob_cdf = np.cumsum(startprob_pdf) transmat_cdf = np.cumsum(self.transmat[0, pstates_idx[0]], 1) # Initial hidden state rand = random_state.rand() currstate = (startprob_cdf > rand).argmax() hidden_states = [currstate] obs = [self._generate_sample_from_state(currstate, pstates_idx[0], random_state)] for i in range(1, n_obs): rand = random_state.rand() if gen_pstates: currpstate = (ptransmat_cdf[currpstate] > rand).argmax() pstates_idx.append(currpstate) pstates.append(self.er[currpstate]) transmat_cdf = np.cumsum(self.transmat[pstates_idx[i - 1], pstates_idx[i]], 1) rand = random_state.rand() currstate = (transmat_cdf[currstate] > rand).argmax() hidden_states.append(currstate) obs.append(self._generate_sample_from_state(currstate, pstates_idx[i], random_state)) return np.array(obs), np.array(pstates), np.array(hidden_states, dtype=int) def fit_df(self, dfs, pstate_col=PSTATE_COL): """ Convenience function to fit a model from a list of dataframes """ obs_cols = list(self.emission_name) obs = [df[df.columns.difference([pstate_col])][obs_cols].values for df in dfs] pstates = [df[pstate_col].values for df in dfs] return self.fit(obs, pstates) def score_df(self, df, pstate_col=PSTATE_COL): """ """ obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values return self.score(obs, pstates) def score_events_df(self, df, pstate_col=PSTATE_COL, score_col='score'): """ """ df = df.copy() obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values df[score_col] = self.score_events(obs, pstates) return df def predict_states_df(self, df, pstate_col=PSTATE_COL, hstate_col=HSTATE_COL): df = df.copy() obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values _, df[hstate_col] = self.predict_states(obs, pstates) return df def predict_df(self, df, next_pstate=None, pstate_col=PSTATE_COL): obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values return self.predict(obs, pstates, next_pstate) def __str__(self): pstates = sorted(self.e.keys()) sep = '-' * 80 + '\n' sep2 = '_' * 40 + '\n' out = 'POHMM\n' out += 'H-states: %d\n' % self.n_hidden_states out += 'P-states: (%d) %s\n' % (len(pstates), str(pstates)) out += 'Emission: %s\n' % (self.emission_distr) out += sep out += 'Transition probas\n' out += sep2 out += '. -> .\n%s\n' % str(self.transmat[0, 0]) for pstate in pstates: out += sep2 out += '%s -> .\n%s\n' % (pstate, str(self.transmat[self.e[pstate], 0])) out += sep2 out += '. -> %s\n%s\n' % (pstate, str(self.transmat[0, self.e[pstate]])) for pstate1, pstate2 in product(pstates, pstates): out += sep2 out += '%s -> %s\n%s\n' % (pstate1, pstate2, str(self.transmat[self.e[pstate1], self.e[pstate2]])) out += sep out += 'Starting probas\n' out += '.: %s\n' % str(self.startprob[0]) for pstate in pstates: out += '%s: %s\n' % (pstate, str(self.startprob[self.e[pstate]])) out += sep out += 'Steady probas\n' out += '.: %s\n' % str(self.steadyprob[0]) for pstate in pstates: out += '%s: %s\n' % (pstate, str(self.steadyprob[self.e[pstate]])) out += sep out += 'Emissions\n' for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): out += sep2 out += 'Feature %s: %s\n' % (feature_name, feature_distr) for param in _DISTRIBUTIONS[feature_distr]: out += '.: %s = %s\n' % (param, str(self.emission[feature_name][param][0])) for pstate in pstates: out += '%s: %s = %s\n' % (pstate, param, str(self.emission[feature_name][param][self.e[pstate]])) out += sep return out def expected_value(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): """ Determine the joint maximum likelihood estimate """ # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': return np.sum(w * expected_lognormal(self.emission[feature]['logsigma'].flatten(), self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': return np.sum(w * expected_normal(self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten())) elif hstate is None: # Marginalize hstate if self.emission_name_distr[feature] == 'lognormal': return np.sum(hstate_prob * expected_lognormal(self.emission[feature]['logsigma'][pstate_idx, :], self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': return np.sum(hstate_prob * expected_normal(self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :])) elif pstate is None: # Marginalize pstate if self.emission_name_distr[feature] == 'lognormal': return np.sum(pstate_prob * expected_lognormal(self.emission[feature]['logsigma'][:, hstate], self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': return np.sum(pstate_prob * expected_normal(self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate])) else: if self.emission_name_distr[feature] == 'lognormal': return expected_lognormal(self.emission[feature]['logsigma'][pstate_idx, hstate], self.emission[feature]['logmu'][pstate_idx, hstate]) elif self.emission_name_distr[feature] == 'normal': return expected_normal(self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) return def pdf_fn(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'].flatten(), 0, np.exp(self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten()) elif hstate is None: # Marginalize hstate w = hstate_prob if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][pstate_idx, :], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :]) elif pstate is None: # Marginalize pstate w = pstate_prob if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][:, hstate], 0, np.exp(self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate]) else: w = 1 if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][pstate_idx, hstate], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, hstate])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) def fn(x): if np.isscalar(x): p = np.sum(w * pdf(x)) else: x = np.array(x) p = np.zeros(len(x)) for i, xi in enumerate(x): p[i] = np.sum(w * pdf(xi)) return p return fn def cdf_fn(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'].flatten(), 0, np.exp(self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten()) elif hstate is None: # Marginalize hstate w = hstate_prob if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][pstate_idx, :], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :]) elif pstate is None: # Marginalize pstate w = pstate_prob if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][:, hstate], 0, np.exp(self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate]) else: w = 1 if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][pstate_idx, hstate], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, hstate])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) def fn(x): if np.isscalar(x): p = np.sum(w * cdf(x)) else: x = np.array(x) p = np.zeros(len(x)) for i, xi in enumerate(x): p[i] = np.sum(w * cdf(xi)) return p return fn def params(self, pstates=None): if pstates is None: pstates = [None] + sorted( set(self.er.values())) # TODO: self.e caches any unknown value, maybe it shouldn't? params = [] # emission parameters for hstate, pstate_label in product(range(self.n_hidden_states), pstates): for feature, distr in zip(self.emission_name, self.emission_distr): for feature_param in _DISTRIBUTIONS[distr]: params.append(self.emission[feature][feature_param][self.e[pstate_label], hstate]) # transition parameters, diagonals only assuming 2 state for hstate, pstate_label in product(range(self.n_hidden_states), pstates): params.append(self.transmat[self.e[pstate_label], self.e[pstate_label], hstate, hstate]) return np.array(params)
vmonaco/pohmm
pohmm/pohmm.py
Pohmm.expected_value
python
def expected_value(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': return np.sum(w * expected_lognormal(self.emission[feature]['logsigma'].flatten(), self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': return np.sum(w * expected_normal(self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten())) elif hstate is None: # Marginalize hstate if self.emission_name_distr[feature] == 'lognormal': return np.sum(hstate_prob * expected_lognormal(self.emission[feature]['logsigma'][pstate_idx, :], self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': return np.sum(hstate_prob * expected_normal(self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :])) elif pstate is None: # Marginalize pstate if self.emission_name_distr[feature] == 'lognormal': return np.sum(pstate_prob * expected_lognormal(self.emission[feature]['logsigma'][:, hstate], self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': return np.sum(pstate_prob * expected_normal(self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate])) else: if self.emission_name_distr[feature] == 'lognormal': return expected_lognormal(self.emission[feature]['logsigma'][pstate_idx, hstate], self.emission[feature]['logmu'][pstate_idx, hstate]) elif self.emission_name_distr[feature] == 'normal': return expected_normal(self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) return
Determine the joint maximum likelihood estimate
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/pohmm.py#L989-L1049
[ "def expected_lognormal(logsigma, logmu):\n return np.exp(logmu + (logsigma ** 2) / 2)\n", "def expected_normal(logmu, logsigma):\n return logmu\n" ]
class Pohmm(object): """ Partially observable hidden Markov model """ def __init__(self, n_hidden_states=2, emissions=['normal'], max_iter=1000, thresh=1e-6, init_method='obs', init_spread=2, smoothing=None, random_state=None): if type(n_hidden_states) is int: self.n_hidden_states = n_hidden_states else: raise Exception('Wrong type for n_hidden_states. Must be int') if type(emissions[0]) is tuple: emission_name, emission_distr = zip(*emissions) elif type(emissions[0]) is str: emission_name, emission_distr = np.arange(len(emissions)), emissions for distr in emission_distr: if distr not in _DISTRIBUTIONS.keys(): raise ValueError('Emission distribution must be one of', _DISTRIBUTIONS.keys()) self.emission_name = emission_name self.emission_distr = emission_distr self.emission_name_distr = dict(zip(emission_name, emission_distr)) self.n_features = len(emissions) # Set up the emission parameters # emission: {'feature':{'param': np.array(shape=(n_partial_states, n_hidden_states))}} self.emission = defaultdict(dict) for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: self.emission[name][param] = None self.emission = dict(self.emission) assert max_iter >= 0 assert thresh >= 0 self.max_iter = max_iter self.thresh = thresh assert init_spread >= 0 self.init_spread = init_spread if init_method not in _INIT_METHODS: raise ValueError('init_method must be one of', _INIT_METHODS) self.init_method = init_method if smoothing is None: smoothing = {'transmat': None, 'startprob': None} for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: smoothing['%s:%s' % (name, param)] = None elif type(smoothing) is str: s = smoothing smoothing = {'transmat': s, 'startprob': s} for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: smoothing['%s:%s' % (name, param)] = s elif type(smoothing) is dict: assert 'transmat' in smoothing.keys() assert 'startprob' in smoothing.keys() for name, distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[distr]: assert param in smoothing.keys() or '%s:%s' % (name, param) in smoothing.keys() if param in smoothing.keys() and '%s:%s' % (name, param) not in smoothing.keys(): smoothing['%s:%s' % (name, param)] = smoothing[param] else: raise Exception('Wrong type for smoothing. Must be None, str, or dict') self.smoothing = smoothing self.random_state = random_state # Number of unique partial states is unknown until fit self.n_partial_states = None # Results after fitting the model self.logprob_ = None self.n_iter_performed_ = None self.logprob_delta_ = None # Mapping between p-states and a unique index # Defaults to 0 for unknown or missing p-states self.e = defaultdict(int) def _get_startprob(self): return np.exp(self._log_startprob) def _set_startprob(self, startprob): if startprob is None: startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states else: startprob = np.asarray(startprob, dtype=np.float) # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(startprob): startprob = normalize(startprob, axis=1) if len(startprob) != self.n_partial_states: raise ValueError('startprob must have length n_partial_states') if not np.allclose(np.sum(startprob, axis=1), 1.0): raise ValueError('startprob must sum to 1.0') self._log_startprob = np.log(np.asarray(startprob).copy()) startprob = property(_get_startprob, _set_startprob) def _get_steadyprob(self): return np.exp(self._log_steadyprob) def _set_steadyprob(self, steadyprob): if steadyprob is None: steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states else: steadyprob = np.asarray(steadyprob, dtype=np.float) # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(steadyprob): steadyprob = normalize(steadyprob, axis=1) if len(steadyprob) != self.n_partial_states: raise ValueError('steadyprob must have length n_partial_states') if not np.allclose(np.sum(steadyprob, axis=1), 1.0): raise ValueError('steadyprob must sum to 1.0') self._log_steadyprob = np.log(np.asarray(steadyprob).copy()) steadyprob = property(_get_steadyprob, _set_steadyprob) def _get_transmat(self): return np.exp(self._log_transmat) def _set_transmat(self, transmat): if transmat is None: transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) / self.n_hidden_states # check if there exists a component whose value is exactly zero # if so, add a small number and re-normalize if not np.alltrue(transmat): transmat = normalize(transmat, axis=3) if (np.asarray(transmat).shape != (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)): raise ValueError('transmat must have shape ' '(n_partial_states,n_partial_states,n_hidden_states,n_hidden_states)') if not np.all(np.allclose(np.sum(transmat, axis=3), 1.0)): raise ValueError('Rows of transmat must sum to 1.0') self._log_transmat = np.log(np.asarray(transmat).copy()) underflow_idx = np.isnan(self._log_transmat) self._log_transmat[underflow_idx] = NEGINF transmat = property(_get_transmat, _set_transmat) def _compute_log_likelihood(self, obs, pstates_idx): q = np.zeros(shape=(len(obs), self.n_hidden_states, self.n_features)) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': mu = self.emission[feature_name]['mu'][pstates_idx] sigma = self.emission[feature_name]['sigma'][pstates_idx] for j in range(self.n_hidden_states): q[:, j, col] = np.log( np.maximum(MIN_PROBA, stats.norm.pdf(obs[:, col], loc=mu[:, j], scale=sigma[:, j]))) if feature_distr == 'lognormal': logmu = self.emission[feature_name]['logmu'][pstates_idx] logsigma = self.emission[feature_name]['logsigma'][pstates_idx] for j in range(self.n_hidden_states): q[:, j, col] = np.log(np.maximum(MIN_PROBA, stats.lognorm.pdf(obs[:, col], logsigma[:, j], loc=0, scale=np.exp(logmu[:, j])))) q = q.sum(axis=2) return q def _generate_sample_from_state(self, hidden_state, pstates_idx, random_state=None): sample = np.zeros(self.n_features) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': mu = self.emission[feature_name]['mu'][pstates_idx][hidden_state] sigma = self.emission[feature_name]['sigma'][pstates_idx][hidden_state] sample[col] = stats.norm.rvs(loc=mu, scale=sigma, random_state=random_state) if feature_distr == 'lognormal': logmu = self.emission[feature_name]['logmu'][pstates_idx][hidden_state] logsigma = self.emission[feature_name]['logsigma'][pstates_idx][hidden_state] sample[col] = stats.lognorm.rvs(logsigma, loc=0, scale=np.exp(logmu), random_state=random_state) return sample def _init_pstates(self, unique_pstates): # Map events to a unique index. The unknown p-state is at idx 0 self.e = defaultdict(int) self.e.update(dict(zip(np.sort(unique_pstates), range(1, len(unique_pstates) + 1)))) self.er = {v: k for k, v in self.e.items()} # Reverse lookup self.er[0] = UNKNOWN_PSTATE self.n_partial_states = len(self.e.keys()) + 1 # Add one for the unknown state return def _init_pstate_freqs(self, pstates_idx): # Partial state frequencies self.pstate_freq = Counter([idx for seq in pstates_idx for idx in seq]) self.pstate_trans_freq = Counter([(idx1, idx2) for seq in pstates_idx for idx1, idx2 in zip(seq[:-1], seq[1:])]) # Store freqs for the meta state self.pstate_freq[0] = len(np.concatenate(pstates_idx)) for seq in pstates_idx: self.pstate_startprob[seq[0]] += 1 self.pstate_steadyprob += np.bincount(seq, minlength=self.n_partial_states) for idx1, idx2 in zip(seq[:-1], seq[1:]): self.pstate_trans_freq[(0, 0)] += 1 self.pstate_trans_freq[(idx1, 0)] += 1 self.pstate_trans_freq[(0, idx2)] += 1 self.pstate_transmat[idx1, idx2] += 1 # TODO: separate probas from freqs # Normalize to get the probabilities, ignore the meta state at idx 0 self.pstate_startprob[1:] = normalize(self.pstate_startprob[1:]) self.pstate_transmat[1:, 1:] = normalize(self.pstate_transmat[1:, 1:], axis=1) self.pstate_steadyprob[1:] = normalize(self.pstate_steadyprob[1:]) return def _init_from_obs(self, obs, pstates_idx): # Partial state probabilities self.pstate_startprob = np.zeros(self.n_partial_states) self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states)) self.pstate_steadyprob = np.zeros(self.n_partial_states) # obs should be (N*T, n_features) # N is the number of samples # T is the size of each sample obs = np.concatenate(obs) pstates_idx = np.concatenate(pstates_idx) # Initialize starting and transition probas self.startprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states self.steadyprob = np.ones(shape=(self.n_partial_states, self.n_hidden_states)) / self.n_hidden_states self.transmat = np.ones(shape=(self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) / self.n_hidden_states # Initialize emission parameters # Hidden states are ordered by the first feature feature1 = self.emission_name[0] for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': self.emission[feature_name]['mu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['sigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) for idx in range(1, self.n_partial_states): idx_pstate = (pstates_idx == idx) if not np.any(idx_pstate): idx_pstate = np.arange(len(pstates_idx)) if feature_name == feature1: self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA) self.emission[feature_name]['mu'][idx] = obs[idx_pstate, col].mean() + \ obs[:, col].std() * np.linspace( -self.init_spread, self.init_spread, self.n_hidden_states) else: self.emission[feature_name]['sigma'][idx, :] = np.maximum(obs[idx_pstate, col].std(), MIN_PROBA) self.emission[feature_name]['mu'][idx, :] = obs[idx_pstate, col].mean() if feature_distr == 'lognormal': self.emission[feature_name]['logmu'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['logsigma'] = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) for idx in range(1, self.n_partial_states): idx_pstate = (pstates_idx == idx) if not np.any(idx_pstate): idx_pstate = np.arange(len(pstates_idx)) if feature_name == feature1: self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(), MIN_PROBA) self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() + np.log( obs[:, col]).std() * np.linspace(-self.init_spread, self.init_spread, self.n_hidden_states) else: self.emission[feature_name]['logsigma'][idx, :] = np.maximum(np.log(obs[idx_pstate, col]).std(), MIN_PROBA) self.emission[feature_name]['logmu'][idx] = np.log(obs[idx_pstate, col]).mean() return def _init_random(self, random_state=None): if random_state is None: random_state = self.random_state random_state = check_random_state(random_state) self.pstate_startprob = np.zeros(self.n_partial_states) self.pstate_transmat = np.zeros((self.n_partial_states, self.n_partial_states)) self.pstate_steadyprob = np.zeros(self.n_partial_states) self.pstate_startprob[1:] = gen_stochastic_matrix(size=self.n_partial_states - 1, random_state=random_state) self.pstate_transmat[1:, 1:] = gen_stochastic_matrix( size=(self.n_partial_states - 1, self.n_partial_states - 1), random_state=random_state) self.pstate_steadyprob[1:] = steadystate(self.pstate_transmat[1:, 1:]) self.startprob = gen_stochastic_matrix(size=(self.n_partial_states, self.n_hidden_states), random_state=random_state) transmat = np.zeros((self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)): transmat[i, j] = gen_stochastic_matrix(size=(self.n_hidden_states, self.n_hidden_states), random_state=random_state) self.transmat = normalize(transmat, axis=3) # Initialize emission parameters for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): if feature_distr == 'normal': self.emission[feature_name]['mu'] = random_state.uniform(*_RANDINIT[feature_distr]['mu'], size=( self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['sigma'] = random_state.uniform(*_RANDINIT[feature_distr]['sigma'], size=( self.n_partial_states, self.n_hidden_states)) if feature_distr == 'lognormal': self.emission[feature_name]['logmu'] = random_state.uniform(*_RANDINIT[feature_distr]['logmu'], size=( self.n_partial_states, self.n_hidden_states)) self.emission[feature_name]['logsigma'] = random_state.uniform(*_RANDINIT[feature_distr]['logsigma'], size=(self.n_partial_states, self.n_hidden_states)) if self.emission_distr[0] == 'normal': self.emission[self.emission_name[0]]['mu'] = np.sort(self.emission[self.emission_name[0]]['mu'], axis=1) elif self.emission_distr[0] == 'lognormal': self.emission[self.emission_name[0]]['logmu'] = np.sort(self.emission[self.emission_name[0]]['logmu'], axis=1) return def _smooth(self): self._compute_marginals() startprob = self.startprob for j in range(1, self.n_partial_states): if 'freq' == self.smoothing['startprob']: w_ = 1 / (1 + self.pstate_freq[j]) w_j = 1 - w_ elif 'proba' == self.smoothing['startprob']: w_j = self.pstate_steadyprob[j] w_ = 1 - w_j elif 'exp' == self.smoothing['startprob']: w_ = np.exp(-self.pstate_freq[j]) w_j = 1 - w_ elif 'fixed' == self.smoothing['startprob']: w_ = 1 w_j = 0 elif self.smoothing['startprob'] is None: w_ = 0 w_j = 1 else: raise Exception('Wrong smoothing for startprob: ' + self.smoothing['startprob']) startprob[j] = w_j * self.startprob[j] + w_ * self.startprob[0] self.startprob = startprob transmat = self.transmat for i, j in product(range(1, self.n_partial_states), range(1, self.n_partial_states)): if 'freq' == self.smoothing['transmat']: w_i0 = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[0, j]) w_0j = 1 / (1 + self.pstate_trans_freq[i, j] + self.pstate_trans_freq[i, 0]) w_ij = 1 - (w_i0 + w_0j) w_ = 0 elif 'proba' == self.smoothing['transmat']: denom = self.pstate_transmat[i, j] + self.pstate_transmat[i, :].sum() + self.pstate_transmat[:, j].sum() w_i0 = self.pstate_transmat[i, :].sum() / denom w_0j = self.pstate_transmat[:, j].sum() / denom w_ij = self.pstate_transmat[i, j] / denom w_ = 0 elif 'exp' == self.smoothing['transmat']: w_i0 = self.pstate_trans_freq[i, 0] * np.exp(-self.pstate_trans_freq[i, j]) w_0j = self.pstate_trans_freq[0, j] * np.exp(-self.pstate_trans_freq[i, j]) w_ = self.pstate_trans_freq[0, 0] * np.exp( -(self.pstate_trans_freq[i, 0] + self.pstate_trans_freq[0, j])) w_ij = self.pstate_trans_freq[i, j] w_ij, w_i0, w_0j, w_ = normalize(np.array([w_ij, w_i0, w_0j, w_])) elif 'fixed' == self.smoothing['transmat']: w_i0 = 0 w_0j = 0 w_ = 1 w_ij = 0 elif self.smoothing['transmat'] is None: w_i0 = 0 w_0j = 0 w_ = 0 w_ij = 1 else: raise Exception('Wrong smoothing for transmat: ' + self.smoothing['transmat']) assert (w_i0 + w_0j + w_ + w_ij) - 1 < TOLERANCE transmat[i, j] = w_ij * self.transmat[i, j] + w_i0 * self.transmat[i, 0] + w_0j * self.transmat[0, j] + w_ * \ self.transmat[ 0, 0] self.transmat = transmat assert np.all(self.startprob.sum(axis=1) - 1 < TOLERANCE) assert np.all(self.steadyprob.sum(axis=1) - 1 < TOLERANCE) assert np.all(self.transmat.sum(axis=3) - 1 < TOLERANCE) for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): for param in _DISTRIBUTIONS[feature_distr]: key = '%s:%s' % (feature_name, param) for j in range(1, self.n_partial_states): if 'freq' == self.smoothing[key]: w_ = 1 / (1 + self.pstate_freq[j]) w_j = 1 - w_ elif 'proba' == self.smoothing[key]: w_j = self.pstate_steadyprob[j] w_ = 1 - w_j elif 'exp' == self.smoothing[key]: w_ = np.exp(-self.pstate_freq[j]) w_j = 1 - w_ elif 'fixed' == self.smoothing[key]: w_ = 1 w_j = 0 elif self.smoothing[key] is None: w_ = 0 w_j = 1 else: raise Exception('Wrong smoothing for ' + key) self.emission[feature_name][param][j] = w_j * self.emission[feature_name][param][j] + w_ * \ self.emission[ feature_name][ param][0] return def _initialize_sufficient_statistics(self): stats = { 'nobs': 0, 'post': np.zeros((self.n_partial_states, self.n_hidden_states)), 'obs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'obs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'lnobs': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'lnobs**2': np.zeros((self.n_partial_states, self.n_hidden_states, self.n_features)), 'start': np.zeros((self.n_partial_states, self.n_hidden_states)), 'steady': np.zeros((self.n_partial_states, self.n_hidden_states)), 'trans': np.zeros( (self.n_partial_states, self.n_partial_states, self.n_hidden_states, self.n_hidden_states)) } return stats def _accumulate_sufficient_statistics(self, stats, obs, pstates_idx, framelogprob, posteriors, fwdlattice, bwdlattice): stats['nobs'] += 1 n_observations, n_hidden_states = framelogprob.shape stats['start'][0] += posteriors[0] for i in range(self.n_partial_states): if len(np.where(pstates_idx == i)[0]) > 0: stats['start'][i] += posteriors[np.where(pstates_idx == i)[0].min()] if n_observations > 1: lneta = np.zeros((n_observations - 1, n_hidden_states, n_hidden_states)) _hmmc._compute_lneta(n_observations, n_hidden_states, pstates_idx, fwdlattice, self._log_transmat, bwdlattice, framelogprob, lneta) for i, j in unique_rows(np.c_[pstates_idx[:-1], pstates_idx[1:]]): if ((pstates_idx[:-1] == i) & (pstates_idx[1:] == j)).sum() > 0: stats['trans'][i, j] += np.exp( logsumexp(lneta[(pstates_idx[:-1] == i) & (pstates_idx[1:] == j)], axis=0)) for i in range(self.n_partial_states): stats['post'][i] += posteriors[pstates_idx == i].sum(axis=0) stats['obs'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i]) stats['obs**2'][i] += np.dot(posteriors[pstates_idx == i].T, obs[pstates_idx == i] ** 2) stats['lnobs'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i])) stats['lnobs**2'][i] += np.dot(posteriors[pstates_idx == i].T, np.log(obs[pstates_idx == i]) ** 2) return def _compute_marginals(self): # TODO: cythonize some of this # Start prob, weighted by p-state start probs self.startprob[0] = (self.pstate_startprob[1:, np.newaxis] * self.startprob[1:]).sum(axis=0) # Use the p-state transmat and transmat to get the full transmat full_transmat = ph2full(self.pstate_transmat[1:, 1:], self.transmat[1:, 1:]) full_steadyprob = steadystate(full_transmat) # Steady state probas are determined by the full trans mat, need to be updated steadyprob = np.zeros(shape=(self.n_partial_states, self.n_hidden_states)) steadyprob[0] = full_steadyprob.reshape(-1, self.n_hidden_states).sum(axis=0) for i in range(self.n_partial_states - 1): steadyprob[i + 1] = normalize( full_steadyprob[i * self.n_hidden_states:i * self.n_hidden_states + self.n_hidden_states]) self.steadyprob = steadyprob # Update the transations to/from the marginal state transmat = self.transmat # Group the hidden states within each partial state for hidx1, hidx2 in product(range(self.n_hidden_states), range(self.n_hidden_states)): transmat[0, 0][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, hidx2::self.n_hidden_states].sum() for pidx in range(self.n_partial_states - 1): transmat[pidx + 1, 0][hidx1, hidx2] = full_transmat[pidx * self.n_hidden_states + hidx1, hidx2::self.n_hidden_states].sum() transmat[0, pidx + 1][hidx1, hidx2] = full_transmat[hidx1::self.n_hidden_states, pidx * self.n_hidden_states + hidx2].sum() self.transmat = normalize(transmat, axis=3) pweights = self.pstate_steadyprob[1:, np.newaxis] # Update emission parameters for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): if feature_distr == 'normal': # Marginal state is a mixture of normals mu = self.emission[feature_name]['mu'][1:, :] sigma = self.emission[feature_name]['sigma'][1:, :] # Weighted mean and var mu_0 = (pweights * mu).sum(axis=0) self.emission[feature_name]['mu'][0, :] = mu_0 self.emission[feature_name]['sigma'][0, :] = np.sqrt( (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0)) if feature_distr == 'lognormal': # Marginal state is a mixture of normals mu = self.emission[feature_name]['logmu'][1:, :] sigma = self.emission[feature_name]['logsigma'][1:, :] # Weighted mean and var mu_0 = (pweights * mu).sum(axis=0) self.emission[feature_name]['logmu'][0, :] = mu_0 self.emission[feature_name]['logsigma'][0, :] = np.sqrt( (pweights * ((mu - mu_0) ** 2 + sigma ** 2)).sum(axis=0)) return def _do_mstep(self, stats): self.startprob = normalize(np.maximum(stats['start'], MIN_PROBA), axis=1) self.transmat = normalize(np.maximum(stats['trans'], MIN_PROBA), axis=3) for col, (feature_name, feature_distr) in enumerate(zip(self.emission_name, self.emission_distr)): if feature_distr == 'normal': denom = np.maximum(stats['post'], MIN_PROBA) mu = stats['obs'][:, :, col] / denom cv_num = (stats['obs**2'][:, :, col] - 2 * mu * stats['obs'][:, :, col] + mu ** 2 * denom) sigma = np.sqrt(cv_num / denom) sigma[np.isnan(sigma)] = MIN_PROBA self.emission[feature_name]['mu'] = mu self.emission[feature_name]['sigma'] = np.maximum(sigma, MIN_PROBA) if feature_distr == 'lognormal': denom = np.maximum(stats['post'], MIN_PROBA) mu = stats['lnobs'][:, :, col] / denom cv_num = (stats['lnobs**2'][:, :, col] - 2 * mu * stats['lnobs'][:, :, col] + mu ** 2 * denom) sigma = np.sqrt(cv_num / denom) sigma[np.isnan(sigma)] = MIN_PROBA self.emission[feature_name]['logmu'] = mu self.emission[feature_name]['logsigma'] = np.maximum(sigma, MIN_PROBA) return def _do_forward_pass(self, framelogprob, event_idx): n_observations, n_hidden_states = framelogprob.shape fwdlattice = np.zeros((n_observations, n_hidden_states)) _hmmc._forward(n_observations, n_hidden_states, event_idx, self._log_startprob, self._log_transmat, framelogprob, fwdlattice) return logsumexp(fwdlattice[-1]), fwdlattice def _do_backward_pass(self, framelogprob, event_idx): n_observations, n_hidden_states = framelogprob.shape bwdlattice = np.zeros((n_observations, n_hidden_states)) _hmmc._backward(n_observations, n_hidden_states, event_idx, self._log_startprob, self._log_transmat, framelogprob, bwdlattice) return bwdlattice def _do_viterbi_pass(self, framelogprob, event_idx): n_observations, n_components = framelogprob.shape state_sequence, logprob = _hmmc._viterbi( n_observations, n_components, event_idx, self._log_startprob, self._log_transmat, framelogprob) return logprob, state_sequence def rand(self, unique_pstates, random_state=None): """ Randomize the POHMM parameters """ self._init_pstates(unique_pstates) self._init_random(random_state=random_state) self._compute_marginals() return self def fit(self, obs, pstates, unique_pstates=None): """ Estimate model parameters. """ obs = [np.array(o) for o in obs] pstates = [np.array(p) for p in pstates] # List or array of observation sequences assert len(obs) == len(pstates) assert obs[0].ndim == 2 assert pstates[0].ndim == 1 if unique_pstates is not None: self._init_pstates(unique_pstates) else: self._init_pstates(list(set(np.concatenate(pstates)))) # Map the partial states to a unique index pstates_idx = [np.array([self.e[p] for p in seq]) for seq in pstates] if self.init_method == 'rand': self._init_random() elif self.init_method == 'obs': self._init_from_obs(obs, pstates_idx) self._init_pstate_freqs(pstates_idx) self._smooth() logprob = [] for i in range(self.max_iter): # Expectation step stats = self._initialize_sufficient_statistics() curr_logprob = 0 for obs_i, pstates_idx_i in zip(obs, pstates_idx): framelogprob = self._compute_log_likelihood(obs_i, pstates_idx_i) lpr, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx_i) bwdlattice = self._do_backward_pass(framelogprob, pstates_idx_i) gamma = fwdlattice + bwdlattice posteriors = np.exp(gamma.T - logsumexp(gamma, axis=1)).T curr_logprob += lpr self._accumulate_sufficient_statistics(stats, obs_i, pstates_idx_i, framelogprob, posteriors, fwdlattice, bwdlattice) logprob.append(curr_logprob) self.logprob_ = curr_logprob # Check for convergence. self.n_iter_performed_ = i if i > 0: self.logprob_delta = logprob[-1] - logprob[-2] if self.logprob_delta < self.thresh: break # Maximization step self._do_mstep(stats) # Mix the parameters self._smooth() return self def score(self, obs, pstates): """ Compute the log probability under the model. """ pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) logprob, _ = self._do_forward_pass(framelogprob, pstates_idx) return logprob def score_events(self, obs, pstates): """ Compute the log probability of each event under the model. """ pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) _, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx) L = logsumexp(fwdlattice, axis=1) return np.concatenate([L[[0]], np.diff(L)]) def predict_states(self, obs, pstates): pstates_idx = np.array([self.e[p] for p in pstates]) framelogprob = self._compute_log_likelihood(obs, pstates_idx) viterbi_logprob, state_sequence = self._do_viterbi_pass(framelogprob, pstates_idx) return viterbi_logprob, state_sequence def predict(self, obs, pstates, next_pstate=None): """ Predict the next observation """ assert len(obs) == len(pstates) pstates_idx = np.array([self.e[ei] for ei in pstates]) next_pstate_idx = self.e[next_pstate] if len(obs) == 0: # No history, use the starting probas next_hstate_prob = self.startprob[next_pstate_idx] else: # With a history, determine the hidden state posteriors using # the last posteriors and transition matrix framelogprob = self._compute_log_likelihood(obs, pstates_idx) _, fwdlattice = self._do_forward_pass(framelogprob, pstates_idx) next_hstate_prob = np.zeros(self.n_hidden_states) alpha_n = fwdlattice[-1] vmax = alpha_n.max(axis=0) alpha_n = np.exp(alpha_n - vmax) alpha_n = alpha_n / alpha_n.sum() trans = self.transmat[pstates_idx[-1], next_pstate_idx] for i in range(self.n_hidden_states): next_hstate_prob[i] = np.sum([alpha_n[j] * trans[j, i] for j in range(self.n_hidden_states)]) assert next_hstate_prob.sum() - 1 < TOLERANCE # Make the prediction prediction = np.array( [self.expected_value(feature, pstate=next_pstate, hstate_prob=next_hstate_prob) for feature in self.emission_name]) # next_hstate = np.argmax(next_hstate_prob) # prediction = np.array( # [self.expected_value(feature, pstate=next_pstate, hstate=next_hstate) for feature in # self.emission_name]) return prediction def gen_pstates_idx(self, n, random_state=None): if random_state is None: random_state = self.random_state random_state = check_random_state(random_state) startprob_cdf = np.cumsum(self.pstate_startprob) transmat_cdf = np.cumsum(self.pstate_transmat, 1) # Initial state. rand = random_state.rand() curr_pstate = (startprob_cdf > rand).argmax() pstates = [curr_pstate] for _ in range(1, n): rand = random_state.rand() curr_pstate = (transmat_cdf[curr_pstate] > rand).argmax() pstates.append(curr_pstate) return np.array(pstates, dtype=int) def sample(self, pstates=None, n_obs=None, random_state=None): """ """ random_state = check_random_state(random_state) if pstates is None and n_obs is None: raise Exception('Must provide either pstates or n_obs') if pstates is not None and n_obs is not None: raise Exception('Must provide either pstates or n_obs but not both') gen_pstates = False rand = random_state.rand() if pstates is None: gen_pstates = True pstartprob_cdf = np.cumsum(self.pstate_startprob) ptransmat_cdf = np.cumsum(self.pstate_transmat, 1) # Initial pstate currpstate = (pstartprob_cdf > rand).argmax() pstates_idx = [currpstate] pstates = [self.er[currpstate]] else: n_obs = len(pstates) pstates_idx = np.array([self.e[p] for p in pstates]) startprob_pdf = self.startprob[pstates_idx[0]] startprob_cdf = np.cumsum(startprob_pdf) transmat_cdf = np.cumsum(self.transmat[0, pstates_idx[0]], 1) # Initial hidden state rand = random_state.rand() currstate = (startprob_cdf > rand).argmax() hidden_states = [currstate] obs = [self._generate_sample_from_state(currstate, pstates_idx[0], random_state)] for i in range(1, n_obs): rand = random_state.rand() if gen_pstates: currpstate = (ptransmat_cdf[currpstate] > rand).argmax() pstates_idx.append(currpstate) pstates.append(self.er[currpstate]) transmat_cdf = np.cumsum(self.transmat[pstates_idx[i - 1], pstates_idx[i]], 1) rand = random_state.rand() currstate = (transmat_cdf[currstate] > rand).argmax() hidden_states.append(currstate) obs.append(self._generate_sample_from_state(currstate, pstates_idx[i], random_state)) return np.array(obs), np.array(pstates), np.array(hidden_states, dtype=int) def fit_df(self, dfs, pstate_col=PSTATE_COL): """ Convenience function to fit a model from a list of dataframes """ obs_cols = list(self.emission_name) obs = [df[df.columns.difference([pstate_col])][obs_cols].values for df in dfs] pstates = [df[pstate_col].values for df in dfs] return self.fit(obs, pstates) def score_df(self, df, pstate_col=PSTATE_COL): """ """ obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values return self.score(obs, pstates) def score_events_df(self, df, pstate_col=PSTATE_COL, score_col='score'): """ """ df = df.copy() obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values df[score_col] = self.score_events(obs, pstates) return df def predict_states_df(self, df, pstate_col=PSTATE_COL, hstate_col=HSTATE_COL): df = df.copy() obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values _, df[hstate_col] = self.predict_states(obs, pstates) return df def predict_df(self, df, next_pstate=None, pstate_col=PSTATE_COL): obs_cols = list(self.emission_name) obs = df[df.columns.difference([pstate_col])][obs_cols].values pstates = df[pstate_col].values return self.predict(obs, pstates, next_pstate) def sample_df(self, pstates=None, n_obs=None, random_state=None, pstate_col=PSTATE_COL, hstate_col=HSTATE_COL): """ Convenience function to generate samples a model and create a dataframe """ try: import pandas as pd except Exception as e: raise e obs, pstates, hstates = self.sample(pstates, n_obs, random_state) items = [] if pstate_col is not None: items.append((pstate_col, pstates)) if hstate_col is not None: items.append((hstate_col, hstates)) items = items + [(self.emission_name[i], obs[:, i]) for i in range(self.n_features)] df = pd.DataFrame.from_items(items) return df def __str__(self): pstates = sorted(self.e.keys()) sep = '-' * 80 + '\n' sep2 = '_' * 40 + '\n' out = 'POHMM\n' out += 'H-states: %d\n' % self.n_hidden_states out += 'P-states: (%d) %s\n' % (len(pstates), str(pstates)) out += 'Emission: %s\n' % (self.emission_distr) out += sep out += 'Transition probas\n' out += sep2 out += '. -> .\n%s\n' % str(self.transmat[0, 0]) for pstate in pstates: out += sep2 out += '%s -> .\n%s\n' % (pstate, str(self.transmat[self.e[pstate], 0])) out += sep2 out += '. -> %s\n%s\n' % (pstate, str(self.transmat[0, self.e[pstate]])) for pstate1, pstate2 in product(pstates, pstates): out += sep2 out += '%s -> %s\n%s\n' % (pstate1, pstate2, str(self.transmat[self.e[pstate1], self.e[pstate2]])) out += sep out += 'Starting probas\n' out += '.: %s\n' % str(self.startprob[0]) for pstate in pstates: out += '%s: %s\n' % (pstate, str(self.startprob[self.e[pstate]])) out += sep out += 'Steady probas\n' out += '.: %s\n' % str(self.steadyprob[0]) for pstate in pstates: out += '%s: %s\n' % (pstate, str(self.steadyprob[self.e[pstate]])) out += sep out += 'Emissions\n' for feature_name, feature_distr in zip(self.emission_name, self.emission_distr): out += sep2 out += 'Feature %s: %s\n' % (feature_name, feature_distr) for param in _DISTRIBUTIONS[feature_distr]: out += '.: %s = %s\n' % (param, str(self.emission[feature_name][param][0])) for pstate in pstates: out += '%s: %s = %s\n' % (pstate, param, str(self.emission[feature_name][param][self.e[pstate]])) out += sep return out def pdf_fn(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'].flatten(), 0, np.exp(self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten()) elif hstate is None: # Marginalize hstate w = hstate_prob if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][pstate_idx, :], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :]) elif pstate is None: # Marginalize pstate w = pstate_prob if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][:, hstate], 0, np.exp(self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate]) else: w = 1 if self.emission_name_distr[feature] == 'lognormal': pdf = lambda x: stats.lognorm.pdf(x, self.emission[feature]['logsigma'][pstate_idx, hstate], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, hstate])) elif self.emission_name_distr[feature] == 'normal': pdf = lambda x: stats.norm.pdf(x, self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) def fn(x): if np.isscalar(x): p = np.sum(w * pdf(x)) else: x = np.array(x) p = np.zeros(len(x)) for i, xi in enumerate(x): p[i] = np.sum(w * pdf(xi)) return p return fn def cdf_fn(self, feature=None, pstate=None, hstate=None, pstate_prob=None, hstate_prob=None): # Use the first feature by default if feature is None: feature = self.emission_name[0] # Will default to marginal pstate if pstate is unknown or None pstate_idx = self.e[pstate] if pstate is not None and pstate_prob is not None: raise Exception('Must provide either pstate or pstate_proba but not both') if hstate is not None and hstate_prob is not None: raise Exception('Must provide either hstate or hstate_proba but not both') # Marginalize pstate using the steady state probas if pstate_prob is None: pstate_prob = self.pstate_steadyprob # Marginalize hstate using the steady state probas if hstate_prob is None: hstate_prob = self.steadyprob[pstate_idx] if pstate is None and hstate is None: # Marginalize both pstate and hstate w = (pstate_prob[:, np.newaxis] * hstate_prob).flatten() if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'].flatten(), 0, np.exp(self.emission[feature]['logmu'].flatten())) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'].flatten(), self.emission[feature]['sigma'].flatten()) elif hstate is None: # Marginalize hstate w = hstate_prob if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][pstate_idx, :], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, :])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][pstate_idx, :], self.emission[feature]['sigma'][pstate_idx, :]) elif pstate is None: # Marginalize pstate w = pstate_prob if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][:, hstate], 0, np.exp(self.emission[feature]['logmu'][:, hstate])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][:, hstate], self.emission[feature]['sigma'][:, hstate]) else: w = 1 if self.emission_name_distr[feature] == 'lognormal': cdf = lambda x: stats.lognorm.cdf(x, self.emission[feature]['logsigma'][pstate_idx, hstate], 0, np.exp(self.emission[feature]['logmu'][pstate_idx, hstate])) elif self.emission_name_distr[feature] == 'normal': cdf = lambda x: stats.norm.cdf(x, self.emission[feature]['mu'][pstate_idx, hstate], self.emission[feature]['sigma'][pstate_idx, hstate]) def fn(x): if np.isscalar(x): p = np.sum(w * cdf(x)) else: x = np.array(x) p = np.zeros(len(x)) for i, xi in enumerate(x): p[i] = np.sum(w * cdf(xi)) return p return fn def params(self, pstates=None): if pstates is None: pstates = [None] + sorted( set(self.er.values())) # TODO: self.e caches any unknown value, maybe it shouldn't? params = [] # emission parameters for hstate, pstate_label in product(range(self.n_hidden_states), pstates): for feature, distr in zip(self.emission_name, self.emission_distr): for feature_param in _DISTRIBUTIONS[distr]: params.append(self.emission[feature][feature_param][self.e[pstate_label], hstate]) # transition parameters, diagonals only assuming 2 state for hstate, pstate_label in product(range(self.n_hidden_states), pstates): params.append(self.transmat[self.e[pstate_label], self.e[pstate_label], hstate, hstate]) return np.array(params)
vmonaco/pohmm
examples/keystroke.py
stratified_kfold
python
def stratified_kfold(df, n_folds): sessions = pd.DataFrame.from_records(list(df.index.unique())).groupby(0).apply(lambda x: x[1].unique()) sessions.apply(lambda x: np.random.shuffle(x)) folds = [] for i in range(n_folds): idx = sessions.apply(lambda x: pd.Series(x[i * (len(x) / n_folds):(i + 1) * (len(x) / n_folds)])) idx = pd.DataFrame(idx.stack().reset_index(level=1, drop=True)).set_index(0, append=True).index.values folds.append(df.loc[idx]) return folds
Create stratified k-folds from an indexed dataframe
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/examples/keystroke.py#L15-L26
null
import numpy as np import pandas as pd from scipy import interp from sklearn.metrics import auc from sklearn.metrics.ranking import _binary_clf_curve from pohmm import Pohmm, PohmmClassifier # CMU Keystroke Dynamics Benchmark Dataset # See: http://www.cs.cmu.edu/~keystroke/ # Kevin S. Killourhy and Roy A. Maxion. "Comparing Anomaly Detectors for Keystroke Dynamics" DATASET_URL = 'http://www.cs.cmu.edu/~keystroke/DSL-StrongPasswordData.csv' def user_folds(df, target): users = df.index.get_level_values(0).unique() return [df.loc[u].reset_index().set_index([target, 'session']) for u in users] def preprocess(df): """Convert the CMU dataset from row vectors into time/duration row observations""" def process_row(idx_row): idx, row = idx_row # press-press latency tau = 1000 * row[4::3].astype(float).values tau = np.r_[np.median(tau), tau] # duration duration = 1000 * row[3::3].astype(float).values keyname = list('.tie5Roanl') + ['enter'] return pd.DataFrame.from_items([ ('user', [row['subject']] * 11), ('session', [row['sessionIndex'] * 100 + row['rep']] * 11), ('tau', tau), ('duration', duration), ('event', keyname) ]) df = pd.concat(map(process_row, df.iterrows())).set_index(['user', 'session']) return df def roc_curve(y_true, y_score): fps, tps, thresholds = _binary_clf_curve( y_true, y_score, pos_label=None, sample_weight=None) if tps.size == 0 or fps[0] != 0: # Add an extra threshold position if necessary tps = np.r_[0, tps] fps = np.r_[0, fps] thresholds = np.r_[thresholds[0] + 1e-2, thresholds] fpr = fps / fps[-1] tpr = tps / tps[-1] return fpr, 1 - tpr, thresholds def ROC(scores): # Generate an ROC curve for each fold, ordered by increasing threshold roc = scores.groupby('user').apply(lambda x: pd.DataFrame(np.c_[roc_curve(x['genuine'], x['score'])][::-1], columns=['far', 'frr', 'threshold'])) # interpolate to get the same threshold values in each fold thresholds = np.sort(roc['threshold'].unique()) roc = roc.groupby(level='user').apply(lambda x: pd.DataFrame(np.c_[thresholds, interp(thresholds, x['threshold'], x['far']), interp(thresholds, x['threshold'], x['frr'])], columns=['threshold', 'far', 'frr'])) roc = roc.reset_index(level=1, drop=True).reset_index() return roc def EER(roc): far, frr = roc['far'].values, roc['frr'].values def perp(a): b = np.empty_like(a) b[0] = -a[1] b[1] = a[0] return b # line segment a given by endpoints a1, a2 # line segment b given by endpoints b1, b2 def seg_intersect(a1, a2, b1, b2): da = a2 - a1 db = b2 - b1 dp = a1 - b1 dap = perp(da) denom = np.dot(dap, db) num = np.dot(dap, dp) return (num / denom) * db + b1 d = far <= frr idx = np.diff(d).nonzero()[0][0] return seg_intersect(np.array([idx, far[idx]]), np.array([idx + 1, far[idx + 1]]), np.array([idx, frr[idx]]), np.array([idx + 1, frr[idx + 1]]))[1] def AUC(roc): return auc(roc['frr'].values, roc['far'].values) def keystroke_model(): """Generates a 2-state model with lognormal emissions and frequency smoothing""" model = Pohmm(n_hidden_states=2, init_spread=2, emissions=['lognormal', 'lognormal'], smoothing='freq', init_method='obs', thresh=1) return model def identification(df, n_folds=10, seed=1234): # Obtain identification results using k-fold cross validation np.random.seed(seed) folds = stratified_kfold(df, n_folds) identification_results = [] for i in range(n_folds): print('Fold %d of %d' % (i + 1, n_folds)) test_idx, test_samples = zip(*folds[i].groupby(level=[0, 1])) train_idx, train_samples = zip(*pd.concat(folds[:i] + folds[i + 1:]).groupby(level=[0, 1])) test_labels, _ = zip(*test_idx) train_labels, _ = zip(*train_idx) cl = PohmmClassifier(keystroke_model) cl.fit_df(train_labels, train_samples) for test_label, test_sample in zip(test_labels, test_samples): result, _ = cl.predict_df(test_sample) identification_results.append((i, test_label, result)) identification_results = pd.DataFrame.from_records(identification_results, columns=['fold', 'label', 'prediction']) acc_summary = identification_results.groupby('fold').apply( lambda x: (x['label'] == x['prediction']).sum() / len(x)).describe() print('Identification summary') print('ACC: %.3f +/- %.3f' % (acc_summary['mean'], acc_summary['std'])) return def verification(df): verification_results = [] users = set(df.index.get_level_values(level='user').unique()) for genuine_user in users: impostor_users = users.difference([genuine_user]) genuine_samples = df.loc[genuine_user] _, genuine_samples = zip(*genuine_samples.groupby(level='session')) train, test = genuine_samples[150:200], genuine_samples[200:] pohmm = keystroke_model() pohmm.fit_df(train) # train_scores = np.array([pohmm.score_df(sample) for sample in train]) scores = [] for sample in test: score = pohmm.score_df(sample) scores.append(score) verification_results.append((genuine_user, True, score)) for imposter_user in impostor_users: _, impostor_samples = zip(*df.loc[imposter_user].groupby(level='session')) for sample in impostor_samples[:5]: score = pohmm.score_df(sample) scores.append(score) verification_results.append((genuine_user, False, score)) verification_results = pd.DataFrame.from_records(verification_results, columns=['user', 'genuine', 'score']) verification_ROC = verification_results.groupby('user').apply(ROC) eer_summary = verification_ROC.groupby('user').apply(EER).describe() auc_summary = verification_ROC.groupby('user').apply(AUC).describe() print('Verification summary') print('EER: %.3f +/- %.3f' % (eer_summary['mean'], eer_summary['std'])) print('AUC: %.3f +/- %.3f' % (auc_summary['mean'], auc_summary['std'])) return if __name__ == '__main__': print('This example takes about 15 minutes to run on an Intel i5...') # Download and preprocess the CMU dataset df = pd.read_csv(DATASET_URL) df = preprocess(df) # Verification results obtained using the 4th session as training data, # sessions 5-8 as genuine and reps 1-5 as impostor verification(df) # Identification results obtained by 10-fold stratified cross validation using only the last session identification(df.groupby(level=0).apply(lambda x: x[-(11 * 50):]).reset_index(level=0, drop=True))
vmonaco/pohmm
examples/keystroke.py
preprocess
python
def preprocess(df): def process_row(idx_row): idx, row = idx_row # press-press latency tau = 1000 * row[4::3].astype(float).values tau = np.r_[np.median(tau), tau] # duration duration = 1000 * row[3::3].astype(float).values keyname = list('.tie5Roanl') + ['enter'] return pd.DataFrame.from_items([ ('user', [row['subject']] * 11), ('session', [row['sessionIndex'] * 100 + row['rep']] * 11), ('tau', tau), ('duration', duration), ('event', keyname) ]) df = pd.concat(map(process_row, df.iterrows())).set_index(['user', 'session']) return df
Convert the CMU dataset from row vectors into time/duration row observations
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/examples/keystroke.py#L34-L58
null
import numpy as np import pandas as pd from scipy import interp from sklearn.metrics import auc from sklearn.metrics.ranking import _binary_clf_curve from pohmm import Pohmm, PohmmClassifier # CMU Keystroke Dynamics Benchmark Dataset # See: http://www.cs.cmu.edu/~keystroke/ # Kevin S. Killourhy and Roy A. Maxion. "Comparing Anomaly Detectors for Keystroke Dynamics" DATASET_URL = 'http://www.cs.cmu.edu/~keystroke/DSL-StrongPasswordData.csv' def stratified_kfold(df, n_folds): """ Create stratified k-folds from an indexed dataframe """ sessions = pd.DataFrame.from_records(list(df.index.unique())).groupby(0).apply(lambda x: x[1].unique()) sessions.apply(lambda x: np.random.shuffle(x)) folds = [] for i in range(n_folds): idx = sessions.apply(lambda x: pd.Series(x[i * (len(x) / n_folds):(i + 1) * (len(x) / n_folds)])) idx = pd.DataFrame(idx.stack().reset_index(level=1, drop=True)).set_index(0, append=True).index.values folds.append(df.loc[idx]) return folds def user_folds(df, target): users = df.index.get_level_values(0).unique() return [df.loc[u].reset_index().set_index([target, 'session']) for u in users] def roc_curve(y_true, y_score): fps, tps, thresholds = _binary_clf_curve( y_true, y_score, pos_label=None, sample_weight=None) if tps.size == 0 or fps[0] != 0: # Add an extra threshold position if necessary tps = np.r_[0, tps] fps = np.r_[0, fps] thresholds = np.r_[thresholds[0] + 1e-2, thresholds] fpr = fps / fps[-1] tpr = tps / tps[-1] return fpr, 1 - tpr, thresholds def ROC(scores): # Generate an ROC curve for each fold, ordered by increasing threshold roc = scores.groupby('user').apply(lambda x: pd.DataFrame(np.c_[roc_curve(x['genuine'], x['score'])][::-1], columns=['far', 'frr', 'threshold'])) # interpolate to get the same threshold values in each fold thresholds = np.sort(roc['threshold'].unique()) roc = roc.groupby(level='user').apply(lambda x: pd.DataFrame(np.c_[thresholds, interp(thresholds, x['threshold'], x['far']), interp(thresholds, x['threshold'], x['frr'])], columns=['threshold', 'far', 'frr'])) roc = roc.reset_index(level=1, drop=True).reset_index() return roc def EER(roc): far, frr = roc['far'].values, roc['frr'].values def perp(a): b = np.empty_like(a) b[0] = -a[1] b[1] = a[0] return b # line segment a given by endpoints a1, a2 # line segment b given by endpoints b1, b2 def seg_intersect(a1, a2, b1, b2): da = a2 - a1 db = b2 - b1 dp = a1 - b1 dap = perp(da) denom = np.dot(dap, db) num = np.dot(dap, dp) return (num / denom) * db + b1 d = far <= frr idx = np.diff(d).nonzero()[0][0] return seg_intersect(np.array([idx, far[idx]]), np.array([idx + 1, far[idx + 1]]), np.array([idx, frr[idx]]), np.array([idx + 1, frr[idx + 1]]))[1] def AUC(roc): return auc(roc['frr'].values, roc['far'].values) def keystroke_model(): """Generates a 2-state model with lognormal emissions and frequency smoothing""" model = Pohmm(n_hidden_states=2, init_spread=2, emissions=['lognormal', 'lognormal'], smoothing='freq', init_method='obs', thresh=1) return model def identification(df, n_folds=10, seed=1234): # Obtain identification results using k-fold cross validation np.random.seed(seed) folds = stratified_kfold(df, n_folds) identification_results = [] for i in range(n_folds): print('Fold %d of %d' % (i + 1, n_folds)) test_idx, test_samples = zip(*folds[i].groupby(level=[0, 1])) train_idx, train_samples = zip(*pd.concat(folds[:i] + folds[i + 1:]).groupby(level=[0, 1])) test_labels, _ = zip(*test_idx) train_labels, _ = zip(*train_idx) cl = PohmmClassifier(keystroke_model) cl.fit_df(train_labels, train_samples) for test_label, test_sample in zip(test_labels, test_samples): result, _ = cl.predict_df(test_sample) identification_results.append((i, test_label, result)) identification_results = pd.DataFrame.from_records(identification_results, columns=['fold', 'label', 'prediction']) acc_summary = identification_results.groupby('fold').apply( lambda x: (x['label'] == x['prediction']).sum() / len(x)).describe() print('Identification summary') print('ACC: %.3f +/- %.3f' % (acc_summary['mean'], acc_summary['std'])) return def verification(df): verification_results = [] users = set(df.index.get_level_values(level='user').unique()) for genuine_user in users: impostor_users = users.difference([genuine_user]) genuine_samples = df.loc[genuine_user] _, genuine_samples = zip(*genuine_samples.groupby(level='session')) train, test = genuine_samples[150:200], genuine_samples[200:] pohmm = keystroke_model() pohmm.fit_df(train) # train_scores = np.array([pohmm.score_df(sample) for sample in train]) scores = [] for sample in test: score = pohmm.score_df(sample) scores.append(score) verification_results.append((genuine_user, True, score)) for imposter_user in impostor_users: _, impostor_samples = zip(*df.loc[imposter_user].groupby(level='session')) for sample in impostor_samples[:5]: score = pohmm.score_df(sample) scores.append(score) verification_results.append((genuine_user, False, score)) verification_results = pd.DataFrame.from_records(verification_results, columns=['user', 'genuine', 'score']) verification_ROC = verification_results.groupby('user').apply(ROC) eer_summary = verification_ROC.groupby('user').apply(EER).describe() auc_summary = verification_ROC.groupby('user').apply(AUC).describe() print('Verification summary') print('EER: %.3f +/- %.3f' % (eer_summary['mean'], eer_summary['std'])) print('AUC: %.3f +/- %.3f' % (auc_summary['mean'], auc_summary['std'])) return if __name__ == '__main__': print('This example takes about 15 minutes to run on an Intel i5...') # Download and preprocess the CMU dataset df = pd.read_csv(DATASET_URL) df = preprocess(df) # Verification results obtained using the 4th session as training data, # sessions 5-8 as genuine and reps 1-5 as impostor verification(df) # Identification results obtained by 10-fold stratified cross validation using only the last session identification(df.groupby(level=0).apply(lambda x: x[-(11 * 50):]).reset_index(level=0, drop=True))
vmonaco/pohmm
examples/keystroke.py
keystroke_model
python
def keystroke_model(): model = Pohmm(n_hidden_states=2, init_spread=2, emissions=['lognormal', 'lognormal'], smoothing='freq', init_method='obs', thresh=1) return model
Generates a 2-state model with lognormal emissions and frequency smoothing
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/examples/keystroke.py#L123-L131
null
import numpy as np import pandas as pd from scipy import interp from sklearn.metrics import auc from sklearn.metrics.ranking import _binary_clf_curve from pohmm import Pohmm, PohmmClassifier # CMU Keystroke Dynamics Benchmark Dataset # See: http://www.cs.cmu.edu/~keystroke/ # Kevin S. Killourhy and Roy A. Maxion. "Comparing Anomaly Detectors for Keystroke Dynamics" DATASET_URL = 'http://www.cs.cmu.edu/~keystroke/DSL-StrongPasswordData.csv' def stratified_kfold(df, n_folds): """ Create stratified k-folds from an indexed dataframe """ sessions = pd.DataFrame.from_records(list(df.index.unique())).groupby(0).apply(lambda x: x[1].unique()) sessions.apply(lambda x: np.random.shuffle(x)) folds = [] for i in range(n_folds): idx = sessions.apply(lambda x: pd.Series(x[i * (len(x) / n_folds):(i + 1) * (len(x) / n_folds)])) idx = pd.DataFrame(idx.stack().reset_index(level=1, drop=True)).set_index(0, append=True).index.values folds.append(df.loc[idx]) return folds def user_folds(df, target): users = df.index.get_level_values(0).unique() return [df.loc[u].reset_index().set_index([target, 'session']) for u in users] def preprocess(df): """Convert the CMU dataset from row vectors into time/duration row observations""" def process_row(idx_row): idx, row = idx_row # press-press latency tau = 1000 * row[4::3].astype(float).values tau = np.r_[np.median(tau), tau] # duration duration = 1000 * row[3::3].astype(float).values keyname = list('.tie5Roanl') + ['enter'] return pd.DataFrame.from_items([ ('user', [row['subject']] * 11), ('session', [row['sessionIndex'] * 100 + row['rep']] * 11), ('tau', tau), ('duration', duration), ('event', keyname) ]) df = pd.concat(map(process_row, df.iterrows())).set_index(['user', 'session']) return df def roc_curve(y_true, y_score): fps, tps, thresholds = _binary_clf_curve( y_true, y_score, pos_label=None, sample_weight=None) if tps.size == 0 or fps[0] != 0: # Add an extra threshold position if necessary tps = np.r_[0, tps] fps = np.r_[0, fps] thresholds = np.r_[thresholds[0] + 1e-2, thresholds] fpr = fps / fps[-1] tpr = tps / tps[-1] return fpr, 1 - tpr, thresholds def ROC(scores): # Generate an ROC curve for each fold, ordered by increasing threshold roc = scores.groupby('user').apply(lambda x: pd.DataFrame(np.c_[roc_curve(x['genuine'], x['score'])][::-1], columns=['far', 'frr', 'threshold'])) # interpolate to get the same threshold values in each fold thresholds = np.sort(roc['threshold'].unique()) roc = roc.groupby(level='user').apply(lambda x: pd.DataFrame(np.c_[thresholds, interp(thresholds, x['threshold'], x['far']), interp(thresholds, x['threshold'], x['frr'])], columns=['threshold', 'far', 'frr'])) roc = roc.reset_index(level=1, drop=True).reset_index() return roc def EER(roc): far, frr = roc['far'].values, roc['frr'].values def perp(a): b = np.empty_like(a) b[0] = -a[1] b[1] = a[0] return b # line segment a given by endpoints a1, a2 # line segment b given by endpoints b1, b2 def seg_intersect(a1, a2, b1, b2): da = a2 - a1 db = b2 - b1 dp = a1 - b1 dap = perp(da) denom = np.dot(dap, db) num = np.dot(dap, dp) return (num / denom) * db + b1 d = far <= frr idx = np.diff(d).nonzero()[0][0] return seg_intersect(np.array([idx, far[idx]]), np.array([idx + 1, far[idx + 1]]), np.array([idx, frr[idx]]), np.array([idx + 1, frr[idx + 1]]))[1] def AUC(roc): return auc(roc['frr'].values, roc['far'].values) def identification(df, n_folds=10, seed=1234): # Obtain identification results using k-fold cross validation np.random.seed(seed) folds = stratified_kfold(df, n_folds) identification_results = [] for i in range(n_folds): print('Fold %d of %d' % (i + 1, n_folds)) test_idx, test_samples = zip(*folds[i].groupby(level=[0, 1])) train_idx, train_samples = zip(*pd.concat(folds[:i] + folds[i + 1:]).groupby(level=[0, 1])) test_labels, _ = zip(*test_idx) train_labels, _ = zip(*train_idx) cl = PohmmClassifier(keystroke_model) cl.fit_df(train_labels, train_samples) for test_label, test_sample in zip(test_labels, test_samples): result, _ = cl.predict_df(test_sample) identification_results.append((i, test_label, result)) identification_results = pd.DataFrame.from_records(identification_results, columns=['fold', 'label', 'prediction']) acc_summary = identification_results.groupby('fold').apply( lambda x: (x['label'] == x['prediction']).sum() / len(x)).describe() print('Identification summary') print('ACC: %.3f +/- %.3f' % (acc_summary['mean'], acc_summary['std'])) return def verification(df): verification_results = [] users = set(df.index.get_level_values(level='user').unique()) for genuine_user in users: impostor_users = users.difference([genuine_user]) genuine_samples = df.loc[genuine_user] _, genuine_samples = zip(*genuine_samples.groupby(level='session')) train, test = genuine_samples[150:200], genuine_samples[200:] pohmm = keystroke_model() pohmm.fit_df(train) # train_scores = np.array([pohmm.score_df(sample) for sample in train]) scores = [] for sample in test: score = pohmm.score_df(sample) scores.append(score) verification_results.append((genuine_user, True, score)) for imposter_user in impostor_users: _, impostor_samples = zip(*df.loc[imposter_user].groupby(level='session')) for sample in impostor_samples[:5]: score = pohmm.score_df(sample) scores.append(score) verification_results.append((genuine_user, False, score)) verification_results = pd.DataFrame.from_records(verification_results, columns=['user', 'genuine', 'score']) verification_ROC = verification_results.groupby('user').apply(ROC) eer_summary = verification_ROC.groupby('user').apply(EER).describe() auc_summary = verification_ROC.groupby('user').apply(AUC).describe() print('Verification summary') print('EER: %.3f +/- %.3f' % (eer_summary['mean'], eer_summary['std'])) print('AUC: %.3f +/- %.3f' % (auc_summary['mean'], auc_summary['std'])) return if __name__ == '__main__': print('This example takes about 15 minutes to run on an Intel i5...') # Download and preprocess the CMU dataset df = pd.read_csv(DATASET_URL) df = preprocess(df) # Verification results obtained using the 4th session as training data, # sessions 5-8 as genuine and reps 1-5 as impostor verification(df) # Identification results obtained by 10-fold stratified cross validation using only the last session identification(df.groupby(level=0).apply(lambda x: x[-(11 * 50):]).reset_index(level=0, drop=True))
vmonaco/pohmm
pohmm/utils.py
normalize
python
def normalize(A, axis=None, inplace=False): if not inplace: A = A.copy() A += np.finfo(float).eps Asum = A.sum(axis) if axis and A.ndim > 1: # Make sure we don't divide by zero. Asum[Asum == 0] = 1 shape = list(A.shape) shape[axis] = 1 Asum.shape = shape A /= Asum return A
Normalize the input array so that it sums to 1. Parameters ---------- A: array, shape (n_samples, n_features) Non-normalized input data. axis: int Dimension along which normalization is performed. Returns ------- normalized_A: array, shape (n_samples, n_features) A with values normalized (summing to 1) along the prescribed axis
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/utils.py#L40-L68
null
import numbers import numpy as np from itertools import product def check_random_state(seed): """Turn seed into a np.random.RandomState instance If seed is None, return the RandomState singleton used by np.random. If seed is an int, return a new RandomState instance seeded with seed. If seed is already a RandomState instance, return it. Otherwise raise ValueError. """ if seed is None or seed is np.random: return np.random.mtrand._rand if isinstance(seed, (numbers.Integral, np.integer)): return np.random.RandomState(seed) if isinstance(seed, np.random.RandomState): return seed raise ValueError('%r cannot be used to seed a numpy.random.RandomState' ' instance' % seed) def logsumexp(arr, axis=0): """ Computes the sum of arr assuming arr is in the log domain. Returns log(sum(exp(arr))) while minimizing the possibility of over/underflow. """ arr = np.rollaxis(arr, axis) # Use the max to normalize, as with the log this is what accumulates # the less errors vmax = arr.max(axis=0) out = np.log(np.sum(np.exp(arr - vmax), axis=0)) out += vmax return out def ph2full(ptrans, htrans): """ Convert a p-state transition matrix and h-state matrices to the full transation matrix The full transmat hase N=n_pstates*n_hstates states """ n_pstates = len(ptrans) n_hstates = len(htrans[0, 0]) N = n_pstates * n_hstates trans = np.zeros((N, N)) for pidx in range(n_pstates): for hidx in range(n_hstates): trans[pidx * n_hstates + hidx] = (ptrans[pidx, :, np.newaxis] * htrans[pidx, :, hidx]).flatten() return trans def full2ph(trans, n_pstates): """ Convert a full transmat to the respective p-state and h-state transmats """ n_hstates = len(trans) / n_pstates htrans = np.zeros((n_pstates, n_pstates, n_hstates, n_hstates)) for pidx1, pidx2 in product(range(n_pstates), range(n_pstates)): idx1 = pidx1 * n_hstates idx2 = pidx2 * n_hstates htrans[pidx1, pidx2] = trans[idx1:idx1 + n_hstates, idx2:idx2 + n_hstates] ptrans = normalize(htrans.sum(axis=-1).sum(axis=-1), axis=1) htrans = normalize(htrans, axis=3) return ptrans, htrans def gen_stochastic_matrix(size, random_state=None): """ Generate a unfiformly-random stochastic array or matrix """ if not type(size) is tuple: size = (1, size) assert len(size) == 2 n = random_state.uniform(size=(size[0], size[1] - 1)) n = np.concatenate([np.zeros((size[0], 1)), n, np.ones((size[0], 1))], axis=1) A = np.diff(np.sort(n)) return A.squeeze() def steadystate(A, max_iter=100): """ Empirically determine the steady state probabilities from a stochastic matrix """ P = np.linalg.matrix_power(A, max_iter) # Determine the unique rows in A v = [] for i in range(len(P)): if not np.any([np.allclose(P[i], vi, ) for vi in v]): v.append(P[i]) return normalize(np.sum(v, axis=0)) def unique_rows(a): """ Get the unique row values in matrix a """ a = np.ascontiguousarray(a) unique_a = np.unique(a.view([('', a.dtype)] * a.shape[1])) return unique_a.view(a.dtype).reshape((unique_a.shape[0], a.shape[1])) def expected_lognormal(logsigma, logmu): return np.exp(logmu + (logsigma ** 2) / 2) def expected_normal(logmu, logsigma): return logmu
vmonaco/pohmm
pohmm/utils.py
ph2full
python
def ph2full(ptrans, htrans): n_pstates = len(ptrans) n_hstates = len(htrans[0, 0]) N = n_pstates * n_hstates trans = np.zeros((N, N)) for pidx in range(n_pstates): for hidx in range(n_hstates): trans[pidx * n_hstates + hidx] = (ptrans[pidx, :, np.newaxis] * htrans[pidx, :, hidx]).flatten() return trans
Convert a p-state transition matrix and h-state matrices to the full transation matrix The full transmat hase N=n_pstates*n_hstates states
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/utils.py#L71-L85
null
import numbers import numpy as np from itertools import product def check_random_state(seed): """Turn seed into a np.random.RandomState instance If seed is None, return the RandomState singleton used by np.random. If seed is an int, return a new RandomState instance seeded with seed. If seed is already a RandomState instance, return it. Otherwise raise ValueError. """ if seed is None or seed is np.random: return np.random.mtrand._rand if isinstance(seed, (numbers.Integral, np.integer)): return np.random.RandomState(seed) if isinstance(seed, np.random.RandomState): return seed raise ValueError('%r cannot be used to seed a numpy.random.RandomState' ' instance' % seed) def logsumexp(arr, axis=0): """ Computes the sum of arr assuming arr is in the log domain. Returns log(sum(exp(arr))) while minimizing the possibility of over/underflow. """ arr = np.rollaxis(arr, axis) # Use the max to normalize, as with the log this is what accumulates # the less errors vmax = arr.max(axis=0) out = np.log(np.sum(np.exp(arr - vmax), axis=0)) out += vmax return out def normalize(A, axis=None, inplace=False): """ Normalize the input array so that it sums to 1. Parameters ---------- A: array, shape (n_samples, n_features) Non-normalized input data. axis: int Dimension along which normalization is performed. Returns ------- normalized_A: array, shape (n_samples, n_features) A with values normalized (summing to 1) along the prescribed axis """ if not inplace: A = A.copy() A += np.finfo(float).eps Asum = A.sum(axis) if axis and A.ndim > 1: # Make sure we don't divide by zero. Asum[Asum == 0] = 1 shape = list(A.shape) shape[axis] = 1 Asum.shape = shape A /= Asum return A def full2ph(trans, n_pstates): """ Convert a full transmat to the respective p-state and h-state transmats """ n_hstates = len(trans) / n_pstates htrans = np.zeros((n_pstates, n_pstates, n_hstates, n_hstates)) for pidx1, pidx2 in product(range(n_pstates), range(n_pstates)): idx1 = pidx1 * n_hstates idx2 = pidx2 * n_hstates htrans[pidx1, pidx2] = trans[idx1:idx1 + n_hstates, idx2:idx2 + n_hstates] ptrans = normalize(htrans.sum(axis=-1).sum(axis=-1), axis=1) htrans = normalize(htrans, axis=3) return ptrans, htrans def gen_stochastic_matrix(size, random_state=None): """ Generate a unfiformly-random stochastic array or matrix """ if not type(size) is tuple: size = (1, size) assert len(size) == 2 n = random_state.uniform(size=(size[0], size[1] - 1)) n = np.concatenate([np.zeros((size[0], 1)), n, np.ones((size[0], 1))], axis=1) A = np.diff(np.sort(n)) return A.squeeze() def steadystate(A, max_iter=100): """ Empirically determine the steady state probabilities from a stochastic matrix """ P = np.linalg.matrix_power(A, max_iter) # Determine the unique rows in A v = [] for i in range(len(P)): if not np.any([np.allclose(P[i], vi, ) for vi in v]): v.append(P[i]) return normalize(np.sum(v, axis=0)) def unique_rows(a): """ Get the unique row values in matrix a """ a = np.ascontiguousarray(a) unique_a = np.unique(a.view([('', a.dtype)] * a.shape[1])) return unique_a.view(a.dtype).reshape((unique_a.shape[0], a.shape[1])) def expected_lognormal(logsigma, logmu): return np.exp(logmu + (logsigma ** 2) / 2) def expected_normal(logmu, logsigma): return logmu
vmonaco/pohmm
pohmm/utils.py
full2ph
python
def full2ph(trans, n_pstates): n_hstates = len(trans) / n_pstates htrans = np.zeros((n_pstates, n_pstates, n_hstates, n_hstates)) for pidx1, pidx2 in product(range(n_pstates), range(n_pstates)): idx1 = pidx1 * n_hstates idx2 = pidx2 * n_hstates htrans[pidx1, pidx2] = trans[idx1:idx1 + n_hstates, idx2:idx2 + n_hstates] ptrans = normalize(htrans.sum(axis=-1).sum(axis=-1), axis=1) htrans = normalize(htrans, axis=3) return ptrans, htrans
Convert a full transmat to the respective p-state and h-state transmats
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/utils.py#L88-L103
[ "def normalize(A, axis=None, inplace=False):\n \"\"\"\n Normalize the input array so that it sums to 1.\n\n Parameters\n ----------\n A: array, shape (n_samples, n_features)\n Non-normalized input data.\n axis: int\n Dimension along which normalization is performed.\n\n Returns\n -------\n normalized_A: array, shape (n_samples, n_features)\n A with values normalized (summing to 1) along the prescribed axis\n \"\"\"\n if not inplace:\n A = A.copy()\n\n A += np.finfo(float).eps\n Asum = A.sum(axis)\n if axis and A.ndim > 1:\n # Make sure we don't divide by zero.\n Asum[Asum == 0] = 1\n shape = list(A.shape)\n shape[axis] = 1\n Asum.shape = shape\n A /= Asum\n return A\n" ]
import numbers import numpy as np from itertools import product def check_random_state(seed): """Turn seed into a np.random.RandomState instance If seed is None, return the RandomState singleton used by np.random. If seed is an int, return a new RandomState instance seeded with seed. If seed is already a RandomState instance, return it. Otherwise raise ValueError. """ if seed is None or seed is np.random: return np.random.mtrand._rand if isinstance(seed, (numbers.Integral, np.integer)): return np.random.RandomState(seed) if isinstance(seed, np.random.RandomState): return seed raise ValueError('%r cannot be used to seed a numpy.random.RandomState' ' instance' % seed) def logsumexp(arr, axis=0): """ Computes the sum of arr assuming arr is in the log domain. Returns log(sum(exp(arr))) while minimizing the possibility of over/underflow. """ arr = np.rollaxis(arr, axis) # Use the max to normalize, as with the log this is what accumulates # the less errors vmax = arr.max(axis=0) out = np.log(np.sum(np.exp(arr - vmax), axis=0)) out += vmax return out def normalize(A, axis=None, inplace=False): """ Normalize the input array so that it sums to 1. Parameters ---------- A: array, shape (n_samples, n_features) Non-normalized input data. axis: int Dimension along which normalization is performed. Returns ------- normalized_A: array, shape (n_samples, n_features) A with values normalized (summing to 1) along the prescribed axis """ if not inplace: A = A.copy() A += np.finfo(float).eps Asum = A.sum(axis) if axis and A.ndim > 1: # Make sure we don't divide by zero. Asum[Asum == 0] = 1 shape = list(A.shape) shape[axis] = 1 Asum.shape = shape A /= Asum return A def ph2full(ptrans, htrans): """ Convert a p-state transition matrix and h-state matrices to the full transation matrix The full transmat hase N=n_pstates*n_hstates states """ n_pstates = len(ptrans) n_hstates = len(htrans[0, 0]) N = n_pstates * n_hstates trans = np.zeros((N, N)) for pidx in range(n_pstates): for hidx in range(n_hstates): trans[pidx * n_hstates + hidx] = (ptrans[pidx, :, np.newaxis] * htrans[pidx, :, hidx]).flatten() return trans def gen_stochastic_matrix(size, random_state=None): """ Generate a unfiformly-random stochastic array or matrix """ if not type(size) is tuple: size = (1, size) assert len(size) == 2 n = random_state.uniform(size=(size[0], size[1] - 1)) n = np.concatenate([np.zeros((size[0], 1)), n, np.ones((size[0], 1))], axis=1) A = np.diff(np.sort(n)) return A.squeeze() def steadystate(A, max_iter=100): """ Empirically determine the steady state probabilities from a stochastic matrix """ P = np.linalg.matrix_power(A, max_iter) # Determine the unique rows in A v = [] for i in range(len(P)): if not np.any([np.allclose(P[i], vi, ) for vi in v]): v.append(P[i]) return normalize(np.sum(v, axis=0)) def unique_rows(a): """ Get the unique row values in matrix a """ a = np.ascontiguousarray(a) unique_a = np.unique(a.view([('', a.dtype)] * a.shape[1])) return unique_a.view(a.dtype).reshape((unique_a.shape[0], a.shape[1])) def expected_lognormal(logsigma, logmu): return np.exp(logmu + (logsigma ** 2) / 2) def expected_normal(logmu, logsigma): return logmu
vmonaco/pohmm
pohmm/utils.py
gen_stochastic_matrix
python
def gen_stochastic_matrix(size, random_state=None): if not type(size) is tuple: size = (1, size) assert len(size) == 2 n = random_state.uniform(size=(size[0], size[1] - 1)) n = np.concatenate([np.zeros((size[0], 1)), n, np.ones((size[0], 1))], axis=1) A = np.diff(np.sort(n)) return A.squeeze()
Generate a unfiformly-random stochastic array or matrix
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/utils.py#L106-L119
null
import numbers import numpy as np from itertools import product def check_random_state(seed): """Turn seed into a np.random.RandomState instance If seed is None, return the RandomState singleton used by np.random. If seed is an int, return a new RandomState instance seeded with seed. If seed is already a RandomState instance, return it. Otherwise raise ValueError. """ if seed is None or seed is np.random: return np.random.mtrand._rand if isinstance(seed, (numbers.Integral, np.integer)): return np.random.RandomState(seed) if isinstance(seed, np.random.RandomState): return seed raise ValueError('%r cannot be used to seed a numpy.random.RandomState' ' instance' % seed) def logsumexp(arr, axis=0): """ Computes the sum of arr assuming arr is in the log domain. Returns log(sum(exp(arr))) while minimizing the possibility of over/underflow. """ arr = np.rollaxis(arr, axis) # Use the max to normalize, as with the log this is what accumulates # the less errors vmax = arr.max(axis=0) out = np.log(np.sum(np.exp(arr - vmax), axis=0)) out += vmax return out def normalize(A, axis=None, inplace=False): """ Normalize the input array so that it sums to 1. Parameters ---------- A: array, shape (n_samples, n_features) Non-normalized input data. axis: int Dimension along which normalization is performed. Returns ------- normalized_A: array, shape (n_samples, n_features) A with values normalized (summing to 1) along the prescribed axis """ if not inplace: A = A.copy() A += np.finfo(float).eps Asum = A.sum(axis) if axis and A.ndim > 1: # Make sure we don't divide by zero. Asum[Asum == 0] = 1 shape = list(A.shape) shape[axis] = 1 Asum.shape = shape A /= Asum return A def ph2full(ptrans, htrans): """ Convert a p-state transition matrix and h-state matrices to the full transation matrix The full transmat hase N=n_pstates*n_hstates states """ n_pstates = len(ptrans) n_hstates = len(htrans[0, 0]) N = n_pstates * n_hstates trans = np.zeros((N, N)) for pidx in range(n_pstates): for hidx in range(n_hstates): trans[pidx * n_hstates + hidx] = (ptrans[pidx, :, np.newaxis] * htrans[pidx, :, hidx]).flatten() return trans def full2ph(trans, n_pstates): """ Convert a full transmat to the respective p-state and h-state transmats """ n_hstates = len(trans) / n_pstates htrans = np.zeros((n_pstates, n_pstates, n_hstates, n_hstates)) for pidx1, pidx2 in product(range(n_pstates), range(n_pstates)): idx1 = pidx1 * n_hstates idx2 = pidx2 * n_hstates htrans[pidx1, pidx2] = trans[idx1:idx1 + n_hstates, idx2:idx2 + n_hstates] ptrans = normalize(htrans.sum(axis=-1).sum(axis=-1), axis=1) htrans = normalize(htrans, axis=3) return ptrans, htrans def steadystate(A, max_iter=100): """ Empirically determine the steady state probabilities from a stochastic matrix """ P = np.linalg.matrix_power(A, max_iter) # Determine the unique rows in A v = [] for i in range(len(P)): if not np.any([np.allclose(P[i], vi, ) for vi in v]): v.append(P[i]) return normalize(np.sum(v, axis=0)) def unique_rows(a): """ Get the unique row values in matrix a """ a = np.ascontiguousarray(a) unique_a = np.unique(a.view([('', a.dtype)] * a.shape[1])) return unique_a.view(a.dtype).reshape((unique_a.shape[0], a.shape[1])) def expected_lognormal(logsigma, logmu): return np.exp(logmu + (logsigma ** 2) / 2) def expected_normal(logmu, logsigma): return logmu
vmonaco/pohmm
pohmm/utils.py
steadystate
python
def steadystate(A, max_iter=100): P = np.linalg.matrix_power(A, max_iter) # Determine the unique rows in A v = [] for i in range(len(P)): if not np.any([np.allclose(P[i], vi, ) for vi in v]): v.append(P[i]) return normalize(np.sum(v, axis=0))
Empirically determine the steady state probabilities from a stochastic matrix
train
https://github.com/vmonaco/pohmm/blob/c00f8a62d3005a171d424549a55d46c421859ae9/pohmm/utils.py#L122-L134
[ "def normalize(A, axis=None, inplace=False):\n \"\"\"\n Normalize the input array so that it sums to 1.\n\n Parameters\n ----------\n A: array, shape (n_samples, n_features)\n Non-normalized input data.\n axis: int\n Dimension along which normalization is performed.\n\n Returns\n -------\n normalized_A: array, shape (n_samples, n_features)\n A with values normalized (summing to 1) along the prescribed axis\n \"\"\"\n if not inplace:\n A = A.copy()\n\n A += np.finfo(float).eps\n Asum = A.sum(axis)\n if axis and A.ndim > 1:\n # Make sure we don't divide by zero.\n Asum[Asum == 0] = 1\n shape = list(A.shape)\n shape[axis] = 1\n Asum.shape = shape\n A /= Asum\n return A\n" ]
import numbers import numpy as np from itertools import product def check_random_state(seed): """Turn seed into a np.random.RandomState instance If seed is None, return the RandomState singleton used by np.random. If seed is an int, return a new RandomState instance seeded with seed. If seed is already a RandomState instance, return it. Otherwise raise ValueError. """ if seed is None or seed is np.random: return np.random.mtrand._rand if isinstance(seed, (numbers.Integral, np.integer)): return np.random.RandomState(seed) if isinstance(seed, np.random.RandomState): return seed raise ValueError('%r cannot be used to seed a numpy.random.RandomState' ' instance' % seed) def logsumexp(arr, axis=0): """ Computes the sum of arr assuming arr is in the log domain. Returns log(sum(exp(arr))) while minimizing the possibility of over/underflow. """ arr = np.rollaxis(arr, axis) # Use the max to normalize, as with the log this is what accumulates # the less errors vmax = arr.max(axis=0) out = np.log(np.sum(np.exp(arr - vmax), axis=0)) out += vmax return out def normalize(A, axis=None, inplace=False): """ Normalize the input array so that it sums to 1. Parameters ---------- A: array, shape (n_samples, n_features) Non-normalized input data. axis: int Dimension along which normalization is performed. Returns ------- normalized_A: array, shape (n_samples, n_features) A with values normalized (summing to 1) along the prescribed axis """ if not inplace: A = A.copy() A += np.finfo(float).eps Asum = A.sum(axis) if axis and A.ndim > 1: # Make sure we don't divide by zero. Asum[Asum == 0] = 1 shape = list(A.shape) shape[axis] = 1 Asum.shape = shape A /= Asum return A def ph2full(ptrans, htrans): """ Convert a p-state transition matrix and h-state matrices to the full transation matrix The full transmat hase N=n_pstates*n_hstates states """ n_pstates = len(ptrans) n_hstates = len(htrans[0, 0]) N = n_pstates * n_hstates trans = np.zeros((N, N)) for pidx in range(n_pstates): for hidx in range(n_hstates): trans[pidx * n_hstates + hidx] = (ptrans[pidx, :, np.newaxis] * htrans[pidx, :, hidx]).flatten() return trans def full2ph(trans, n_pstates): """ Convert a full transmat to the respective p-state and h-state transmats """ n_hstates = len(trans) / n_pstates htrans = np.zeros((n_pstates, n_pstates, n_hstates, n_hstates)) for pidx1, pidx2 in product(range(n_pstates), range(n_pstates)): idx1 = pidx1 * n_hstates idx2 = pidx2 * n_hstates htrans[pidx1, pidx2] = trans[idx1:idx1 + n_hstates, idx2:idx2 + n_hstates] ptrans = normalize(htrans.sum(axis=-1).sum(axis=-1), axis=1) htrans = normalize(htrans, axis=3) return ptrans, htrans def gen_stochastic_matrix(size, random_state=None): """ Generate a unfiformly-random stochastic array or matrix """ if not type(size) is tuple: size = (1, size) assert len(size) == 2 n = random_state.uniform(size=(size[0], size[1] - 1)) n = np.concatenate([np.zeros((size[0], 1)), n, np.ones((size[0], 1))], axis=1) A = np.diff(np.sort(n)) return A.squeeze() def unique_rows(a): """ Get the unique row values in matrix a """ a = np.ascontiguousarray(a) unique_a = np.unique(a.view([('', a.dtype)] * a.shape[1])) return unique_a.view(a.dtype).reshape((unique_a.shape[0], a.shape[1])) def expected_lognormal(logsigma, logmu): return np.exp(logmu + (logsigma ** 2) / 2) def expected_normal(logmu, logsigma): return logmu
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.__get_header_with_auth
python
def __get_header_with_auth(self): auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header
This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L54-L77
[ "def __get_header(self):\n header = dict()\n header['Content-Type'] = 'application/json'\n header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0'\n\n if self.language:\n header['Accept-Language'] = self.language\n\n return header\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.login
python
def login(self): auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!')
This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L79-L99
[ "def run_request(request_type, url, retries=5, data=None, headers=None):\n func = __request_factory(request_type)\n\n for attempt in range(retries+1):\n try:\n response = func(url, data=data, headers=headers)\n return response\n except RequestException:\n warnings.warn('Got error on request for attemp %d - %s' %\n (attempt, 'retry is possible' if attempt < retries else 'no retry'))\n return None\n", "def __get_header(self):\n header = dict()\n header['Content-Type'] = 'application/json'\n header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0'\n\n if self.language:\n header['Accept-Language'] = self.language\n\n return header\n", "def parse_raw_response(self, raw_response):\n\n if raw_response.status_code == 200:\n return json.loads(make_str_content(raw_response.content))\n else:\n return self.__handle_error(raw_response)\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.search_series
python
def search_series(self, name=None, imdb_id=None, zap2it_id=None): arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L102-L120
[ "def run_request(request_type, url, retries=5, data=None, headers=None):\n func = __request_factory(request_type)\n\n for attempt in range(retries+1):\n try:\n response = func(url, data=data, headers=headers)\n return response\n except RequestException:\n warnings.warn('Got error on request for attemp %d - %s' %\n (attempt, 'retry is possible' if attempt < retries else 'no retry'))\n return None\n", "def query_param_string_from_option_args(a2q_dict, args_dict):\n \"\"\"\n From a dictionary of arguments to query string parameters, loops through ad arguments list and makes a query string.\n :param a2q_dict: a dictionary containing argument_name > query string parameter name.\n :param args_dict: a dictionary containing the argument_name > argument_value\n :return: a query string.\n \"\"\"\n\n name_value_pairs = dict()\n for ak in a2q_dict.keys():\n value = args_dict[ak]\n if value != None:\n name_value_pairs[a2q_dict[ak]] = str(value)\n\n return urllib.urlencode(name_value_pairs)\n", "def __get_header_with_auth(self):\n \"\"\"\n This private method returns the HTTP heder filled with the Authorization information with the user token.\n The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB\n (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over\n 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24\n hours have passed since the token generation, a login is performed to generate a new one, instead.\n\n :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls.\n \"\"\"\n auth_header = self.__get_header()\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS)\n\n if datetime.now() > token_renew_time:\n token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION)\n if datetime.now() < token_max_time:\n self.__refresh_token()\n else:\n self.login()\n\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n\n return auth_header\n", "def parse_raw_response(self, raw_response):\n\n if raw_response.status_code == 200:\n return json.loads(make_str_content(raw_response.content))\n else:\n return self.__handle_error(raw_response)\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.get_series_actors
python
def get_series_actors(self, series_id): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L137-L148
[ "def run_request(request_type, url, retries=5, data=None, headers=None):\n func = __request_factory(request_type)\n\n for attempt in range(retries+1):\n try:\n response = func(url, data=data, headers=headers)\n return response\n except RequestException:\n warnings.warn('Got error on request for attemp %d - %s' %\n (attempt, 'retry is possible' if attempt < retries else 'no retry'))\n return None\n", "def __get_header_with_auth(self):\n \"\"\"\n This private method returns the HTTP heder filled with the Authorization information with the user token.\n The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB\n (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over\n 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24\n hours have passed since the token generation, a login is performed to generate a new one, instead.\n\n :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls.\n \"\"\"\n auth_header = self.__get_header()\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS)\n\n if datetime.now() > token_renew_time:\n token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION)\n if datetime.now() < token_max_time:\n self.__refresh_token()\n else:\n self.login()\n\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n\n return auth_header\n", "def parse_raw_response(self, raw_response):\n\n if raw_response.status_code == 200:\n return json.loads(make_str_content(raw_response.content))\n else:\n return self.__handle_error(raw_response)\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.get_series_episodes
python
def get_series_episodes(self, series_id, page=1): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L151-L164
[ "def run_request(request_type, url, retries=5, data=None, headers=None):\n func = __request_factory(request_type)\n\n for attempt in range(retries+1):\n try:\n response = func(url, data=data, headers=headers)\n return response\n except RequestException:\n warnings.warn('Got error on request for attemp %d - %s' %\n (attempt, 'retry is possible' if attempt < retries else 'no retry'))\n return None\n", "def query_param_string_from_option_args(a2q_dict, args_dict):\n \"\"\"\n From a dictionary of arguments to query string parameters, loops through ad arguments list and makes a query string.\n :param a2q_dict: a dictionary containing argument_name > query string parameter name.\n :param args_dict: a dictionary containing the argument_name > argument_value\n :return: a query string.\n \"\"\"\n\n name_value_pairs = dict()\n for ak in a2q_dict.keys():\n value = args_dict[ak]\n if value != None:\n name_value_pairs[a2q_dict[ak]] = str(value)\n\n return urllib.urlencode(name_value_pairs)\n", "def __get_header_with_auth(self):\n \"\"\"\n This private method returns the HTTP heder filled with the Authorization information with the user token.\n The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB\n (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over\n 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24\n hours have passed since the token generation, a login is performed to generate a new one, instead.\n\n :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls.\n \"\"\"\n auth_header = self.__get_header()\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS)\n\n if datetime.now() > token_renew_time:\n token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION)\n if datetime.now() < token_max_time:\n self.__refresh_token()\n else:\n self.login()\n\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n\n return auth_header\n", "def parse_raw_response(self, raw_response):\n\n if raw_response.status_code == 200:\n return json.loads(make_str_content(raw_response.content))\n else:\n return self.__handle_error(raw_response)\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.get_series_episodes
python
def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L167-L194
[ "def run_request(request_type, url, retries=5, data=None, headers=None):\n func = __request_factory(request_type)\n\n for attempt in range(retries+1):\n try:\n response = func(url, data=data, headers=headers)\n return response\n except RequestException:\n warnings.warn('Got error on request for attemp %d - %s' %\n (attempt, 'retry is possible' if attempt < retries else 'no retry'))\n return None\n", "def query_param_string_from_option_args(a2q_dict, args_dict):\n \"\"\"\n From a dictionary of arguments to query string parameters, loops through ad arguments list and makes a query string.\n :param a2q_dict: a dictionary containing argument_name > query string parameter name.\n :param args_dict: a dictionary containing the argument_name > argument_value\n :return: a query string.\n \"\"\"\n\n name_value_pairs = dict()\n for ak in a2q_dict.keys():\n value = args_dict[ak]\n if value != None:\n name_value_pairs[a2q_dict[ak]] = str(value)\n\n return urllib.urlencode(name_value_pairs)\n", "def __get_header_with_auth(self):\n \"\"\"\n This private method returns the HTTP heder filled with the Authorization information with the user token.\n The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB\n (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over\n 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24\n hours have passed since the token generation, a login is performed to generate a new one, instead.\n\n :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls.\n \"\"\"\n auth_header = self.__get_header()\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS)\n\n if datetime.now() > token_renew_time:\n token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION)\n if datetime.now() < token_max_time:\n self.__refresh_token()\n else:\n self.login()\n\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n\n return auth_header\n", "def parse_raw_response(self, raw_response):\n\n if raw_response.status_code == 200:\n return json.loads(make_str_content(raw_response.content))\n else:\n return self.__handle_error(raw_response)\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.get_series_images
python
def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id)
Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L226-L249
[ "def run_request(request_type, url, retries=5, data=None, headers=None):\n func = __request_factory(request_type)\n\n for attempt in range(retries+1):\n try:\n response = func(url, data=data, headers=headers)\n return response\n except RequestException:\n warnings.warn('Got error on request for attemp %d - %s' %\n (attempt, 'retry is possible' if attempt < retries else 'no retry'))\n return None\n", "def query_param_string_from_option_args(a2q_dict, args_dict):\n \"\"\"\n From a dictionary of arguments to query string parameters, loops through ad arguments list and makes a query string.\n :param a2q_dict: a dictionary containing argument_name > query string parameter name.\n :param args_dict: a dictionary containing the argument_name > argument_value\n :return: a query string.\n \"\"\"\n\n name_value_pairs = dict()\n for ak in a2q_dict.keys():\n value = args_dict[ak]\n if value != None:\n name_value_pairs[a2q_dict[ak]] = str(value)\n\n return urllib.urlencode(name_value_pairs)\n", "def __get_header_with_auth(self):\n \"\"\"\n This private method returns the HTTP heder filled with the Authorization information with the user token.\n The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB\n (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over\n 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24\n hours have passed since the token generation, a login is performed to generate a new one, instead.\n\n :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls.\n \"\"\"\n auth_header = self.__get_header()\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS)\n\n if datetime.now() > token_renew_time:\n token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION)\n if datetime.now() < token_max_time:\n self.__refresh_token()\n else:\n self.login()\n\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n\n return auth_header\n", "def parse_raw_response(self, raw_response):\n\n if raw_response.status_code == 200:\n return json.loads(make_str_content(raw_response.content))\n else:\n return self.__handle_error(raw_response)\n", "def inner(instance, *args, **kwargs):\n if not instance.is_authenticated:\n raise UserNotLoggedInException('Authentication is required!')\n else:\n return func(instance, *args, **kwargs)\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.get_updated
python
def get_updated(self, from_time, to_time=None): arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L252-L271
[ "def run_request(request_type, url, retries=5, data=None, headers=None):\n func = __request_factory(request_type)\n\n for attempt in range(retries+1):\n try:\n response = func(url, data=data, headers=headers)\n return response\n except RequestException:\n warnings.warn('Got error on request for attemp %d - %s' %\n (attempt, 'retry is possible' if attempt < retries else 'no retry'))\n return None\n", "def query_param_string_from_option_args(a2q_dict, args_dict):\n \"\"\"\n From a dictionary of arguments to query string parameters, loops through ad arguments list and makes a query string.\n :param a2q_dict: a dictionary containing argument_name > query string parameter name.\n :param args_dict: a dictionary containing the argument_name > argument_value\n :return: a query string.\n \"\"\"\n\n name_value_pairs = dict()\n for ak in a2q_dict.keys():\n value = args_dict[ak]\n if value != None:\n name_value_pairs[a2q_dict[ak]] = str(value)\n\n return urllib.urlencode(name_value_pairs)\n", "def __get_header_with_auth(self):\n \"\"\"\n This private method returns the HTTP heder filled with the Authorization information with the user token.\n The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB\n (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over\n 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24\n hours have passed since the token generation, a login is performed to generate a new one, instead.\n\n :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls.\n \"\"\"\n auth_header = self.__get_header()\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS)\n\n if datetime.now() > token_renew_time:\n token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION)\n if datetime.now() < token_max_time:\n self.__refresh_token()\n else:\n self.login()\n\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n\n return auth_header\n", "def parse_raw_response(self, raw_response):\n\n if raw_response.status_code == 200:\n return json.loads(make_str_content(raw_response.content))\n else:\n return self.__handle_error(raw_response)\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.get_user
python
def get_user(self): return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth()))
Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L274-L282
[ "def run_request(request_type, url, retries=5, data=None, headers=None):\n func = __request_factory(request_type)\n\n for attempt in range(retries+1):\n try:\n response = func(url, data=data, headers=headers)\n return response\n except RequestException:\n warnings.warn('Got error on request for attemp %d - %s' %\n (attempt, 'retry is possible' if attempt < retries else 'no retry'))\n return None\n", "def __get_header_with_auth(self):\n \"\"\"\n This private method returns the HTTP heder filled with the Authorization information with the user token.\n The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB\n (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over\n 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24\n hours have passed since the token generation, a login is performed to generate a new one, instead.\n\n :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls.\n \"\"\"\n auth_header = self.__get_header()\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS)\n\n if datetime.now() > token_renew_time:\n token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION)\n if datetime.now() < token_max_time:\n self.__refresh_token()\n else:\n self.login()\n\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n\n return auth_header\n", "def parse_raw_response(self, raw_response):\n\n if raw_response.status_code == 200:\n return json.loads(make_str_content(raw_response.content))\n else:\n return self.__handle_error(raw_response)\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.get_user_favorites
python
def get_user_favorites(self): return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth()))
Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L285-L293
[ "def run_request(request_type, url, retries=5, data=None, headers=None):\n func = __request_factory(request_type)\n\n for attempt in range(retries+1):\n try:\n response = func(url, data=data, headers=headers)\n return response\n except RequestException:\n warnings.warn('Got error on request for attemp %d - %s' %\n (attempt, 'retry is possible' if attempt < retries else 'no retry'))\n return None\n", "def __get_header_with_auth(self):\n \"\"\"\n This private method returns the HTTP heder filled with the Authorization information with the user token.\n The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB\n (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over\n 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24\n hours have passed since the token generation, a login is performed to generate a new one, instead.\n\n :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls.\n \"\"\"\n auth_header = self.__get_header()\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS)\n\n if datetime.now() > token_renew_time:\n token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION)\n if datetime.now() < token_max_time:\n self.__refresh_token()\n else:\n self.login()\n\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n\n return auth_header\n", "def parse_raw_response(self, raw_response):\n\n if raw_response.status_code == 200:\n return json.loads(make_str_content(raw_response.content))\n else:\n return self.__handle_error(raw_response)\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.delete_user_favorite
python
def delete_user_favorite(self, series_id): return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth()))
Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L296-L306
[ "def run_request(request_type, url, retries=5, data=None, headers=None):\n func = __request_factory(request_type)\n\n for attempt in range(retries+1):\n try:\n response = func(url, data=data, headers=headers)\n return response\n except RequestException:\n warnings.warn('Got error on request for attemp %d - %s' %\n (attempt, 'retry is possible' if attempt < retries else 'no retry'))\n return None\n", "def __get_header_with_auth(self):\n \"\"\"\n This private method returns the HTTP heder filled with the Authorization information with the user token.\n The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB\n (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over\n 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24\n hours have passed since the token generation, a login is performed to generate a new one, instead.\n\n :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls.\n \"\"\"\n auth_header = self.__get_header()\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS)\n\n if datetime.now() > token_renew_time:\n token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION)\n if datetime.now() < token_max_time:\n self.__refresh_token()\n else:\n self.login()\n\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n\n return auth_header\n", "def parse_raw_response(self, raw_response):\n\n if raw_response.status_code == 200:\n return json.loads(make_str_content(raw_response.content))\n else:\n return self.__handle_error(raw_response)\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.__get_user_ratings
python
def __get_user_ratings(self): return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth()))
Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L322-L330
null
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.get_user_ratings
python
def get_user_ratings(self, item_type=None): if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings()
Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L333-L348
[ "def run_request(request_type, url, retries=5, data=None, headers=None):\n func = __request_factory(request_type)\n\n for attempt in range(retries+1):\n try:\n response = func(url, data=data, headers=headers)\n return response\n except RequestException:\n warnings.warn('Got error on request for attemp %d - %s' %\n (attempt, 'retry is possible' if attempt < retries else 'no retry'))\n return None\n", "def __get_header_with_auth(self):\n \"\"\"\n This private method returns the HTTP heder filled with the Authorization information with the user token.\n The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB\n (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over\n 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24\n hours have passed since the token generation, a login is performed to generate a new one, instead.\n\n :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls.\n \"\"\"\n auth_header = self.__get_header()\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS)\n\n if datetime.now() > token_renew_time:\n token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION)\n if datetime.now() < token_max_time:\n self.__refresh_token()\n else:\n self.login()\n\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n\n return auth_header\n", "def parse_raw_response(self, raw_response):\n\n if raw_response.status_code == 200:\n return json.loads(make_str_content(raw_response.content))\n else:\n return self.__handle_error(raw_response)\n", "def inner(instance, *args, **kwargs):\n if not instance.is_authenticated:\n raise UserNotLoggedInException('Authentication is required!')\n else:\n return func(instance, *args, **kwargs)\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.add_user_rating
python
def add_user_rating(self, item_type, item_id, item_rating): raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return:
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L351-L366
[ "def run_request(request_type, url, retries=5, data=None, headers=None):\n func = __request_factory(request_type)\n\n for attempt in range(retries+1):\n try:\n response = func(url, data=data, headers=headers)\n return response\n except RequestException:\n warnings.warn('Got error on request for attemp %d - %s' %\n (attempt, 'retry is possible' if attempt < retries else 'no retry'))\n return None\n", "def __get_header_with_auth(self):\n \"\"\"\n This private method returns the HTTP heder filled with the Authorization information with the user token.\n The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB\n (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over\n 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24\n hours have passed since the token generation, a login is performed to generate a new one, instead.\n\n :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls.\n \"\"\"\n auth_header = self.__get_header()\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS)\n\n if datetime.now() > token_renew_time:\n token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION)\n if datetime.now() < token_max_time:\n self.__refresh_token()\n else:\n self.login()\n\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n\n return auth_header\n", "def parse_raw_response(self, raw_response):\n\n if raw_response.status_code == 200:\n return json.loads(make_str_content(raw_response.content))\n else:\n return self.__handle_error(raw_response)\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.delete_user_rating
python
def delete_user_rating(self, item_type, item_id): raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L369-L382
[ "def run_request(request_type, url, retries=5, data=None, headers=None):\n func = __request_factory(request_type)\n\n for attempt in range(retries+1):\n try:\n response = func(url, data=data, headers=headers)\n return response\n except RequestException:\n warnings.warn('Got error on request for attemp %d - %s' %\n (attempt, 'retry is possible' if attempt < retries else 'no retry'))\n return None\n", "def __get_header_with_auth(self):\n \"\"\"\n This private method returns the HTTP heder filled with the Authorization information with the user token.\n The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB\n (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over\n 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24\n hours have passed since the token generation, a login is performed to generate a new one, instead.\n\n :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls.\n \"\"\"\n auth_header = self.__get_header()\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS)\n\n if datetime.now() > token_renew_time:\n token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION)\n if datetime.now() < token_max_time:\n self.__refresh_token()\n else:\n self.login()\n\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n\n return auth_header\n", "def parse_raw_response(self, raw_response):\n\n if raw_response.status_code == 200:\n return json.loads(make_str_content(raw_response.content))\n else:\n return self.__handle_error(raw_response)\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.get_episode
python
def get_episode(self, episode_id): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L385-L396
[ "def run_request(request_type, url, retries=5, data=None, headers=None):\n func = __request_factory(request_type)\n\n for attempt in range(retries+1):\n try:\n response = func(url, data=data, headers=headers)\n return response\n except RequestException:\n warnings.warn('Got error on request for attemp %d - %s' %\n (attempt, 'retry is possible' if attempt < retries else 'no retry'))\n return None\n", "def __get_header_with_auth(self):\n \"\"\"\n This private method returns the HTTP heder filled with the Authorization information with the user token.\n The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB\n (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over\n 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24\n hours have passed since the token generation, a login is performed to generate a new one, instead.\n\n :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls.\n \"\"\"\n auth_header = self.__get_header()\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS)\n\n if datetime.now() > token_renew_time:\n token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION)\n if datetime.now() < token_max_time:\n self.__refresh_token()\n else:\n self.login()\n\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n\n return auth_header\n", "def parse_raw_response(self, raw_response):\n\n if raw_response.status_code == 200:\n return json.loads(make_str_content(raw_response.content))\n else:\n return self.__handle_error(raw_response)\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.get_languages
python
def get_languages(self): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L399-L409
[ "def run_request(request_type, url, retries=5, data=None, headers=None):\n func = __request_factory(request_type)\n\n for attempt in range(retries+1):\n try:\n response = func(url, data=data, headers=headers)\n return response\n except RequestException:\n warnings.warn('Got error on request for attemp %d - %s' %\n (attempt, 'retry is possible' if attempt < retries else 'no retry'))\n return None\n", "def __get_header_with_auth(self):\n \"\"\"\n This private method returns the HTTP heder filled with the Authorization information with the user token.\n The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB\n (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over\n 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24\n hours have passed since the token generation, a login is performed to generate a new one, instead.\n\n :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls.\n \"\"\"\n auth_header = self.__get_header()\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS)\n\n if datetime.now() > token_renew_time:\n token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION)\n if datetime.now() < token_max_time:\n self.__refresh_token()\n else:\n self.login()\n\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n\n return auth_header\n", "def parse_raw_response(self, raw_response):\n\n if raw_response.status_code == 200:\n return json.loads(make_str_content(raw_response.content))\n else:\n return self.__handle_error(raw_response)\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required @authentication_required def get_language(self, language_id): """ Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
thilux/tvdb_client
tvdb_client/clients/ApiV2Client.py
ApiV2Client.get_language
python
def get_language(self, language_id): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages/%d' % language_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response)
Retrieves information about the language of the given id. :param language_id: The TheTVDB Id of the language. :return: a python dictionary with either the result of the search or an error from TheTVDB.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/clients/ApiV2Client.py#L412-L423
[ "def run_request(request_type, url, retries=5, data=None, headers=None):\n func = __request_factory(request_type)\n\n for attempt in range(retries+1):\n try:\n response = func(url, data=data, headers=headers)\n return response\n except RequestException:\n warnings.warn('Got error on request for attemp %d - %s' %\n (attempt, 'retry is possible' if attempt < retries else 'no retry'))\n return None\n", "def __get_header_with_auth(self):\n \"\"\"\n This private method returns the HTTP heder filled with the Authorization information with the user token.\n The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB\n (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over\n 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24\n hours have passed since the token generation, a login is performed to generate a new one, instead.\n\n :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls.\n \"\"\"\n auth_header = self.__get_header()\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS)\n\n if datetime.now() > token_renew_time:\n token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION)\n if datetime.now() < token_max_time:\n self.__refresh_token()\n else:\n self.login()\n\n auth_header['Authorization'] = 'Bearer %s' % self.__token\n\n return auth_header\n", "def parse_raw_response(self, raw_response):\n\n if raw_response.status_code == 200:\n return json.loads(make_str_content(raw_response.content))\n else:\n return self.__handle_error(raw_response)\n" ]
class ApiV2Client(BaseClient): """ This is the python library implementation of the TheTVDB API V2. Details of the APIs is documented in the swagger page maintained by TheTVDB that can be found on this address: https://api.thetvdb.com/swagger The only APIs not represented by methods in this library are the ones that only returns parameters information of the actual API. """ API_BASE_URL = 'https://api.thetvdb.com' TOKEN_DURATION_SECONDS = 23 * 3600 # 23 Hours TOKEN_MAX_DURATION = 24 * 3600 # 24 Hours def __init__(self, username, api_key, account_identifier, language=None): self.username = username self.api_key = api_key self.account_identifier = account_identifier self.is_authenticated = False self.__token = None self.__auth_time = 0 self.language = language def __get_header(self): header = dict() header['Content-Type'] = 'application/json' header['User-Agent'] = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0' if self.language: header['Accept-Language'] = self.language return header def __refresh_token(self): headers = self.__get_header() headers['Authorization'] = 'Bearer %s' % self.__token resp = requests_util.run_request('get', self.API_BASE_URL + '/refresh_token', headers=headers) if resp.status_code == 200: token_resp = self.parse_raw_response(resp) self.__token = token_resp['token'] self.__auth_time = datetime.now() def __get_header_with_auth(self): """ This private method returns the HTTP heder filled with the Authorization information with the user token. The token validity is monitored whenever this function is called, so according to the swagger page of TheTVDB (https://api.thetvdb.com/swagger) the tokens are valid for 24 hours, therefore if a token is generated for over 23 hours already, this function will also perform a token refresh using TheTVDB refresh_token API. If over 24 hours have passed since the token generation, a login is performed to generate a new one, instead. :return: A python dictionary representing the HTTP header to be used in TheTVDB API calls. """ auth_header = self.__get_header() auth_header['Authorization'] = 'Bearer %s' % self.__token token_renew_time = self.__auth_time + timedelta(seconds=self.TOKEN_DURATION_SECONDS) if datetime.now() > token_renew_time: token_max_time = self.__auth_time + timedelta(seconds=self.TOKEN_MAX_DURATION) if datetime.now() < token_max_time: self.__refresh_token() else: self.login() auth_header['Authorization'] = 'Bearer %s' % self.__token return auth_header def login(self): """ This method performs the login on TheTVDB given the api key, user name and account identifier. :return: None """ auth_data = dict() auth_data['apikey'] = self.api_key auth_data['username'] = self.username auth_data['userkey'] = self.account_identifier auth_resp = requests_util.run_request('post', self.API_BASE_URL + '/login', data=json.dumps(auth_data), headers=self.__get_header()) if auth_resp.status_code == 200: auth_resp_data = self.parse_raw_response(auth_resp) self.__token = auth_resp_data['token'] self.__auth_time = datetime.now() self.is_authenticated = True else: raise AuthenticationFailedException('Authentication failed!') @authentication_required def search_series(self, name=None, imdb_id=None, zap2it_id=None): """ Searchs for a series in TheTVDB by either its name, imdb_id or zap2it_id. :param name: the name of the series to look for :param imdb_id: the IMDB id of the series to look for :param zap2it_id: the zap2it id of the series to look for. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'name': 'name', 'imdb_id': 'imdbId', 'zap2it_id': 'zap2itId'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', '%s%s?%s' % (self.API_BASE_URL, '/search/series', query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series(self, series_id): """ Retrieves the information of a series from TheTVDB given the series ID. :param series_id: the id of the series on TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_actors(self, series_id): """ Retrieves the information on the actors of a particular series, given its TheTVDB id. :param series_id: the TheTVDB id of the series :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/actors' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, page=1): """ Retrieves all episodes for a particular series given its TheTVDB id. It retrieves a maximum of 100 results per page. :param series_id: The TheTVDB id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes?page=%d' % (series_id, page), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes(self, series_id, episode_number=None, aired_season=None, aired_episode=None, dvd_season=None, dvd_episode=None, imdb_id=None, page=1): """ Retrieves all episodes for a particular series given its TheTVDB and filtered by additional optional details. :param series_id: The TheTVDB id of the series. :param episode_number: The optional absolute episode number. :param aired_season: The optional aired season number. :param aired_episode: The optional aired episode number. :param dvd_season: The optional DVD season number. :param dvd_episode: The optional DVD episode number. :param imdb_id: The optional IMDB Id of the series. :param page: The page number. If none is provided, 1 is used by default. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'episode_number': 'absoluteNumber', 'aired_season': 'airedSeason', 'aired_episode': 'airedEpisode', 'dvd_season': 'dvdSeason', 'dvd_episode': 'dvdEpisode', 'imdb_id': 'imdbId', 'page': 'page'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_episodes_summary(self, series_id): """ Retrieves the summary of the episodes and seasons of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/episodes/summary' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def __get_series_images(self, series_id): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images' % series_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_series_images(self, series_id, image_type=None, resolution=None, sub_key=None): """ Retrieves the url to images (posters, fanart) for the series, seasons and episodes of a series given its TheTVDB id and filtered by additional parameters. :param series_id: The TheTVDB id of the series. :param image_type: The optional type of image: posters, fanart, thumbnail, etc. :param resolution: The optional resolution: i.e. 1280x1024 :param sub_key: The optional subkey: graphical, text. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'image_type': 'keyType', 'resolution': 'resolution', 'sub_key': 'subKey'} query_string = utils.query_param_string_from_option_args(optional_parameters, arguments) if len(query_string): raw_response = requests_util.run_request('get', self.API_BASE_URL + '/series/%d/images/query?%s' % (series_id, query_string), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) else: return self.__get_series_images(series_id) @authentication_required def get_updated(self, from_time, to_time=None): """ Retrives a list of series that have changed on TheTVDB since a provided from time parameter and optionally to an specified to time. :param from_time: An epoch representation of the date from which to restrict the query to. :param to_time: An optional epcoh representation of the date to which to restrict the query to. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ arguments = locals() optional_parameters = {'to_time': 'toTime'} query_string = 'fromTime=%s&%s' % (from_time, utils.query_param_string_from_option_args(optional_parameters, arguments)) raw_response = requests_util.run_request('get', self.API_BASE_URL + '/uodated/query?%s' % query_string, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_user(self): """ Retrieves information about the user currently using the api. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user', headers=self.__get_header_with_auth())) @authentication_required def get_user_favorites(self): """ Retrieves the list of tv series the current user has flagged as favorite. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/favorites', headers=self.__get_header_with_auth())) @authentication_required def delete_user_favorite(self, series_id): """ Deletes the series of the provided id from the favorites list of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('delete', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def add_user_favorite(self, series_id): """ Added the series related to the series id provided to the list of favorites of the current user. :param series_id: The TheTVDB id of the series. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('put', self.API_BASE_URL + '/user/favorites/%d' % series_id, headers=self.__get_header_with_auth())) @authentication_required def __get_user_ratings(self): """ Returns a list of the ratings provided by the current user. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ return self.parse_raw_response(requests_util.run_request('get', self.API_BASE_URL + '/user/ratings', headers=self.__get_header_with_auth())) @authentication_required def get_user_ratings(self, item_type=None): """ Returns a list of the ratings for the type of item provided, for the current user. :param item_type: One of: series, episode or banner. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ if item_type: query_string = 'itemType=%s' % item_type return self.parse_raw_response( requests_util.run_request('get', self.API_BASE_URL + '/user/ratings/qeury?%s' % query_string, headers=self.__get_header_with_auth())) else: return self.__get_user_ratings() @authentication_required def add_user_rating(self, item_type, item_id, item_rating): """ Adds the rating for the item indicated for the current user. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB id of the item. :param item_rating: The rating from 0 to 10. :return: """ raw_response = requests_util.run_request('put', self.API_BASE_URL + '/user/ratings/%s/%d/%d' % (item_type, item_id, item_rating), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def delete_user_rating(self, item_type, item_id): """ Deletes from the list of rating of the current user, the rating provided for the specified element type. :param item_type: One of: series, episode, banner. :param item_id: The TheTVDB Id of the item. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('delete', self.API_BASE_URL + '/user/ratings/%s/%d' % (item_type, item_id), headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_episode(self, episode_id): """ Returns the full information of the episode belonging to the Id provided. :param episode_id: The TheTVDB id of the episode. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/episodes/%d' % episode_id, headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required def get_languages(self): """ Returns a list of all language options available in TheTVDB. :return: a python dictionary with either the result of the search or an error from TheTVDB. """ raw_response = requests_util.run_request('get', self.API_BASE_URL + '/languages', headers=self.__get_header_with_auth()) return self.parse_raw_response(raw_response) @authentication_required
thilux/tvdb_client
tvdb_client/utils/utils.py
query_param_string_from_option_args
python
def query_param_string_from_option_args(a2q_dict, args_dict): name_value_pairs = dict() for ak in a2q_dict.keys(): value = args_dict[ak] if value != None: name_value_pairs[a2q_dict[ak]] = str(value) return urllib.urlencode(name_value_pairs)
From a dictionary of arguments to query string parameters, loops through ad arguments list and makes a query string. :param a2q_dict: a dictionary containing argument_name > query string parameter name. :param args_dict: a dictionary containing the argument_name > argument_value :return: a query string.
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/utils/utils.py#L6-L20
null
# encoding=latin-1 __author__ = 'tsantana' import urllib def make_str_content(content): """ In python3+ requests.Response.content returns bytes instead of ol'good str. :param content: requests.Response.content :return: str representation of the requests.Response.content data """ if not isinstance(content, str): content = str(content.decode()) return content
thilux/tvdb_client
tvdb_client/utils/utils.py
make_str_content
python
def make_str_content(content): if not isinstance(content, str): content = str(content.decode()) return content
In python3+ requests.Response.content returns bytes instead of ol'good str. :param content: requests.Response.content :return: str representation of the requests.Response.content data
train
https://github.com/thilux/tvdb_client/blob/2d5106f260367c0abe1284683697874df6343f78/tvdb_client/utils/utils.py#L23-L31
null
# encoding=latin-1 __author__ = 'tsantana' import urllib def query_param_string_from_option_args(a2q_dict, args_dict): """ From a dictionary of arguments to query string parameters, loops through ad arguments list and makes a query string. :param a2q_dict: a dictionary containing argument_name > query string parameter name. :param args_dict: a dictionary containing the argument_name > argument_value :return: a query string. """ name_value_pairs = dict() for ak in a2q_dict.keys(): value = args_dict[ak] if value != None: name_value_pairs[a2q_dict[ak]] = str(value) return urllib.urlencode(name_value_pairs)
ChrisBeaumont/smother
smother/diff.py
parse_intervals
python
def parse_intervals(diff_report): for patch in diff_report.patch_set: try: old_pf = diff_report.old_file(patch.source_file) new_pf = diff_report.new_file(patch.target_file) except InvalidPythonFile: continue for hunk in patch: for line in hunk: if line.line_type == LINE_TYPE_ADDED: idx = line.target_line_no yield ContextInterval(new_pf.filename, new_pf.context(idx)) elif line.line_type == LINE_TYPE_REMOVED: idx = line.source_line_no yield ContextInterval(old_pf.filename, old_pf.context(idx)) elif line.line_type in (LINE_TYPE_EMPTY, LINE_TYPE_CONTEXT): pass else: raise AssertionError("Unexpected line type: %s" % line)
Parse a diff into an iterator of Intervals.
train
https://github.com/ChrisBeaumont/smother/blob/65d1ea6ae0060d213b0dcbb983c5aa8e7fee07bb/smother/diff.py#L31-L54
null
""" Diff parser """ from abc import ABCMeta from abc import abstractmethod from abc import abstractproperty import six from functools import wraps from more_itertools import unique_justseen from unidiff.constants import LINE_TYPE_ADDED from unidiff.constants import LINE_TYPE_CONTEXT from unidiff.constants import LINE_TYPE_EMPTY from unidiff.constants import LINE_TYPE_REMOVED from smother.interval import ContextInterval from smother.python import InvalidPythonFile def dedup(func): @wraps(func) def wrapper(*args, **kwargs): return unique_justseen(func(*args, **kwargs)) return wrapper @dedup @six.add_metaclass(ABCMeta) class DiffReporter(object): @abstractproperty def patch_set(self): pass @abstractmethod def old_file(self, path): """ Given an old file name from a diff report, return a PythonFile """ pass @abstractmethod def new_file(self, path): """ Given a new file name from a diff report, return a PythonFile """ pass def changed_intervals(self): return parse_intervals(self)
ChrisBeaumont/smother
smother/python.py
Visitor._filldown
python
def _filldown(self, lineno): if self.line > lineno: # XXX decorated functions make us jump backwards. # understand this more return self.lines.extend( self.current_context for _ in range(self.line, lineno)) self.line = lineno
Copy current_context into `lines` down up until lineno
train
https://github.com/ChrisBeaumont/smother/blob/65d1ea6ae0060d213b0dcbb983c5aa8e7fee07bb/smother/python.py#L50-L61
null
class Visitor(NodeVisitor): """ Walk a module's ast. Build a `lines` list whose value at each index is a context block name for the corresponding source code line. """ def __init__(self, prefix=''): """ Parameters ---------- prefix : str The name to give to the module-level context. """ self.line = 1 # which line (1-based) do we populate next? # a stack of nested contexts self.context = [] self.prefix = prefix self.current_context = prefix self.lines = [] def _update_current_context(self): if self.prefix and self.context: self.current_context = self.prefix + ':' + '.'.join(self.context) elif self.prefix: self.current_context = self.prefix elif self.context: self.current_context = '.'.join(self.context) else: self.current_context = '' def _add_section(self, node): """ Register the current node as a new context block """ self._filldown(node.lineno) # push a new context onto stack self.context.append(node.name) self._update_current_context() for _ in map(self.visit, iter_child_nodes(node)): pass # restore current context self.context.pop() self._update_current_context() def generic_visit(self, node): if hasattr(node, 'lineno'): self._filldown(node.lineno + 1) for _ in map(self.visit, iter_child_nodes(node)): pass def visit_Module(self, node): # noqa # need to manually insert one line for empty modules like __init__.py if not node.body: self.lines = [self.current_context] else: self.generic_visit(node) visit_ClassDef = _add_section visit_FunctionDef = _add_section visit_AsyncFunctionDef = _add_section
ChrisBeaumont/smother
smother/python.py
Visitor._add_section
python
def _add_section(self, node): self._filldown(node.lineno) # push a new context onto stack self.context.append(node.name) self._update_current_context() for _ in map(self.visit, iter_child_nodes(node)): pass # restore current context self.context.pop() self._update_current_context()
Register the current node as a new context block
train
https://github.com/ChrisBeaumont/smother/blob/65d1ea6ae0060d213b0dcbb983c5aa8e7fee07bb/smother/python.py#L63-L78
null
class Visitor(NodeVisitor): """ Walk a module's ast. Build a `lines` list whose value at each index is a context block name for the corresponding source code line. """ def __init__(self, prefix=''): """ Parameters ---------- prefix : str The name to give to the module-level context. """ self.line = 1 # which line (1-based) do we populate next? # a stack of nested contexts self.context = [] self.prefix = prefix self.current_context = prefix self.lines = [] def _update_current_context(self): if self.prefix and self.context: self.current_context = self.prefix + ':' + '.'.join(self.context) elif self.prefix: self.current_context = self.prefix elif self.context: self.current_context = '.'.join(self.context) else: self.current_context = '' def _filldown(self, lineno): """ Copy current_context into `lines` down up until lineno """ if self.line > lineno: # XXX decorated functions make us jump backwards. # understand this more return self.lines.extend( self.current_context for _ in range(self.line, lineno)) self.line = lineno def generic_visit(self, node): if hasattr(node, 'lineno'): self._filldown(node.lineno + 1) for _ in map(self.visit, iter_child_nodes(node)): pass def visit_Module(self, node): # noqa # need to manually insert one line for empty modules like __init__.py if not node.body: self.lines = [self.current_context] else: self.generic_visit(node) visit_ClassDef = _add_section visit_FunctionDef = _add_section visit_AsyncFunctionDef = _add_section
ChrisBeaumont/smother
smother/python.py
PythonFile._module_name
python
def _module_name(filename): absfile = os.path.abspath(filename) match = filename for base in [''] + sys.path: base = os.path.abspath(base) if absfile.startswith(base): match = absfile[len(base):] break return SUFFIX_RE.sub('', match).lstrip('/').replace('/', '.')
Try to find a module name for a file path by stripping off a prefix found in sys.modules.
train
https://github.com/ChrisBeaumont/smother/blob/65d1ea6ae0060d213b0dcbb983c5aa8e7fee07bb/smother/python.py#L140-L155
null
class PythonFile(object): """ A file of python source. """ def __init__(self, filename, source=None, prefix=None): """ Parameters ---------- filename : str The path to the file source : str (optional) The contents of the file. Will be read from `filename` if not provided. prefix : str (optional) Name to give to the outermost context in the file. If not provided, will be the "." form of filename (ie a/b/c.py -> a.b.c) """ self.filename = filename if prefix is None: self.prefix = self._module_name(filename) else: self.prefix = prefix if source is None: with open(filename) as infile: self.source = infile.read() else: self.source = source try: self.ast = parse(self.source) except SyntaxError: raise InvalidPythonFile(self.filename) visitor = Visitor(prefix=self.prefix) visitor.visit(self.ast) self.lines = visitor.lines @staticmethod @classmethod def from_modulename(cls, module_name): """ Build a PythonFile given a dotted module name like a.b.c """ # XXX make this more robust (pyc files? zip archives? etc) slug = module_name.replace('.', '/') paths = [slug + '.py', slug + '/__init__.py'] # always search from current directory for base in [''] + sys.path: for path in paths: fullpath = os.path.join(base, path) if os.path.exists(fullpath): return cls(fullpath, prefix=module_name) else: raise ValueError("Module not found: %s" % module_name) @property def line_count(self): return len(self.lines) def context_range(self, context): """ Return the 1-offset, right-open range of lines spanned by a particular context name. Parameters ---------- context : str Raises ------ ValueError, if context is not present in the file. """ if not context.startswith(self.prefix): context = self.prefix + '.' + context lo = hi = None for idx, line_context in enumerate(self.lines, 1): # context is hierarchical -- context spans itself # and any suffix. if line_context.startswith(context): lo = lo or idx hi = idx if lo is None: raise ValueError("Context %s does not exist in file %s" % (context, self.filename)) return lo, hi + 1 def context(self, line): """ Return the context for a given 1-offset line number. """ # XXX due to a limitation in Visitor, # non-python code after the last python code # in a file is not added to self.lines, so we # have to guard against IndexErrors. idx = line - 1 if idx >= len(self.lines): return self.prefix return self.lines[idx]
ChrisBeaumont/smother
smother/python.py
PythonFile.from_modulename
python
def from_modulename(cls, module_name): # XXX make this more robust (pyc files? zip archives? etc) slug = module_name.replace('.', '/') paths = [slug + '.py', slug + '/__init__.py'] # always search from current directory for base in [''] + sys.path: for path in paths: fullpath = os.path.join(base, path) if os.path.exists(fullpath): return cls(fullpath, prefix=module_name) else: raise ValueError("Module not found: %s" % module_name)
Build a PythonFile given a dotted module name like a.b.c
train
https://github.com/ChrisBeaumont/smother/blob/65d1ea6ae0060d213b0dcbb983c5aa8e7fee07bb/smother/python.py#L158-L173
null
class PythonFile(object): """ A file of python source. """ def __init__(self, filename, source=None, prefix=None): """ Parameters ---------- filename : str The path to the file source : str (optional) The contents of the file. Will be read from `filename` if not provided. prefix : str (optional) Name to give to the outermost context in the file. If not provided, will be the "." form of filename (ie a/b/c.py -> a.b.c) """ self.filename = filename if prefix is None: self.prefix = self._module_name(filename) else: self.prefix = prefix if source is None: with open(filename) as infile: self.source = infile.read() else: self.source = source try: self.ast = parse(self.source) except SyntaxError: raise InvalidPythonFile(self.filename) visitor = Visitor(prefix=self.prefix) visitor.visit(self.ast) self.lines = visitor.lines @staticmethod def _module_name(filename): """ Try to find a module name for a file path by stripping off a prefix found in sys.modules. """ absfile = os.path.abspath(filename) match = filename for base in [''] + sys.path: base = os.path.abspath(base) if absfile.startswith(base): match = absfile[len(base):] break return SUFFIX_RE.sub('', match).lstrip('/').replace('/', '.') @classmethod @property def line_count(self): return len(self.lines) def context_range(self, context): """ Return the 1-offset, right-open range of lines spanned by a particular context name. Parameters ---------- context : str Raises ------ ValueError, if context is not present in the file. """ if not context.startswith(self.prefix): context = self.prefix + '.' + context lo = hi = None for idx, line_context in enumerate(self.lines, 1): # context is hierarchical -- context spans itself # and any suffix. if line_context.startswith(context): lo = lo or idx hi = idx if lo is None: raise ValueError("Context %s does not exist in file %s" % (context, self.filename)) return lo, hi + 1 def context(self, line): """ Return the context for a given 1-offset line number. """ # XXX due to a limitation in Visitor, # non-python code after the last python code # in a file is not added to self.lines, so we # have to guard against IndexErrors. idx = line - 1 if idx >= len(self.lines): return self.prefix return self.lines[idx]