from __future__ import absolute_import, unicode_literals import re import time # ISO-8601 date parsing routines written by Fazal Majid. # The ISO 8601 standard is very convoluted and irregular - a full ISO 8601 # parser is beyond the scope of feedparser and would be a worthwhile addition # to the Python library. # A single regular expression cannot parse ISO 8601 date formats into groups # as the standard is highly irregular (for instance is 030104 2003-01-04 or # 0301-04-01), so we use templates instead. # Please note the order in templates is significant because we need a # greedy match. _iso8601_tmpl = ['YYYY-?MM-?DD', 'YYYY-0MM?-?DD', 'YYYY-MM', 'YYYY-?OOO', 'YY-?MM-?DD', 'YY-?OOO', 'YYYY', '-YY-?MM', '-OOO', '-YY', '--MM-?DD', '--MM', '---DD', 'CC', ''] _iso8601_re = [ tmpl.replace( 'YYYY', r'(?P\d{4})').replace( 'YY', r'(?P\d\d)').replace( 'MM', r'(?P[01]\d)').replace( 'DD', r'(?P[0123]\d)').replace( 'OOO', r'(?P[0123]\d\d)').replace( 'CC', r'(?P\d\d$)') + r'(T?(?P\d{2}):(?P\d{2})' + r'(:(?P\d{2}))?' + r'(\.(?P\d+))?' + r'(?P[+-](?P\d{2})(:(?P\d{2}))?|Z)?)?' for tmpl in _iso8601_tmpl] try: del tmpl except NameError: pass _iso8601_matches = [re.compile(regex).match for regex in _iso8601_re] try: del regex except NameError: pass def _parse_date_iso8601(dateString): '''Parse a variety of ISO-8601-compatible formats like 20040105''' m = None for _iso8601_match in _iso8601_matches: m = _iso8601_match(dateString) if m: break if not m: return if m.span() == (0, 0): return params = m.groupdict() ordinal = params.get('ordinal', 0) if ordinal: ordinal = int(ordinal) else: ordinal = 0 year = params.get('year', '--') if not year or year == '--': year = time.gmtime()[0] elif len(year) == 2: # ISO 8601 assumes current century, i.e. 93 -> 2093, NOT 1993 year = 100 * int(time.gmtime()[0] / 100) + int(year) else: year = int(year) month = params.get('month', '-') if not month or month == '-': # ordinals are NOT normalized by mktime, we simulate them # by setting month=1, day=ordinal if ordinal: month = 1 else: month = time.gmtime()[1] month = int(month) day = params.get('day', 0) if not day: # see above if ordinal: day = ordinal elif params.get('century', 0) or \ params.get('year', 0) or params.get('month', 0): day = 1 else: day = time.gmtime()[2] else: day = int(day) # special case of the century - is the first year of the 21st century # 2000 or 2001 ? The debate goes on... if 'century' in params: year = (int(params['century']) - 1) * 100 + 1 # in ISO 8601 most fields are optional for field in ['hour', 'minute', 'second', 'tzhour', 'tzmin']: if not params.get(field, None): params[field] = 0 hour = int(params.get('hour', 0)) minute = int(params.get('minute', 0)) second = int(float(params.get('second', 0))) # weekday is normalized by mktime(), we can ignore it weekday = 0 daylight_savings_flag = -1 tm = [year, month, day, hour, minute, second, weekday, ordinal, daylight_savings_flag] # ISO 8601 time zone adjustments tz = params.get('tz') if tz and tz != 'Z': if tz[0] == '-': tm[3] += int(params.get('tzhour', 0)) tm[4] += int(params.get('tzmin', 0)) elif tz[0] == '+': tm[3] -= int(params.get('tzhour', 0)) tm[4] -= int(params.get('tzmin', 0)) else: return None # Python's time.mktime() is a wrapper around the ANSI C mktime(3c) # which is guaranteed to normalize d/m/y/h/m/s. # Many implementations have bugs, but we'll pretend they don't. return time.localtime(time.mktime(tuple(tm)))