_strptime.py revision cc2b0161257495f859200bce0aea3ed7e646feb3
1"""Strptime-related classes and functions. 2 3CLASSES: 4 LocaleTime -- Discovers and stores locale-specific time information 5 TimeRE -- Creates regexes for pattern matching a string of text containing 6 time information 7 8FUNCTIONS: 9 _getlang -- Figure out what language is being used for the locale 10 strptime -- Calculates the time struct represented by the passed-in string 11 12""" 13import time 14import locale 15import calendar 16from re import compile as re_compile 17from re import IGNORECASE 18from re import escape as re_escape 19from datetime import date as datetime_date 20try: 21 from thread import allocate_lock as _thread_allocate_lock 22except: 23 from dummy_thread import allocate_lock as _thread_allocate_lock 24 25__all__ = ['strptime'] 26 27def _getlang(): 28 # Figure out what the current language is set to. 29 return locale.getlocale(locale.LC_TIME) 30 31class LocaleTime(object): 32 """Stores and handles locale-specific information related to time. 33 34 ATTRIBUTES: 35 f_weekday -- full weekday names (7-item list) 36 a_weekday -- abbreviated weekday names (7-item list) 37 f_month -- full month names (13-item list; dummy value in [0], which 38 is added by code) 39 a_month -- abbreviated month names (13-item list, dummy value in 40 [0], which is added by code) 41 am_pm -- AM/PM representation (2-item list) 42 LC_date_time -- format string for date/time representation (string) 43 LC_date -- format string for date representation (string) 44 LC_time -- format string for time representation (string) 45 timezone -- daylight- and non-daylight-savings timezone representation 46 (2-item list of sets) 47 lang -- Language used by instance (2-item tuple) 48 """ 49 50 def __init__(self): 51 """Set all attributes. 52 53 Order of methods called matters for dependency reasons. 54 55 The locale language is set at the offset and then checked again before 56 exiting. This is to make sure that the attributes were not set with a 57 mix of information from more than one locale. This would most likely 58 happen when using threads where one thread calls a locale-dependent 59 function while another thread changes the locale while the function in 60 the other thread is still running. Proper coding would call for 61 locks to prevent changing the locale while locale-dependent code is 62 running. The check here is done in case someone does not think about 63 doing this. 64 65 Only other possible issue is if someone changed the timezone and did 66 not call tz.tzset . That is an issue for the programmer, though, 67 since changing the timezone is worthless without that call. 68 69 """ 70 self.lang = _getlang() 71 self.__calc_weekday() 72 self.__calc_month() 73 self.__calc_am_pm() 74 self.__calc_timezone() 75 self.__calc_date_time() 76 if _getlang() != self.lang: 77 raise ValueError("locale changed during initialization") 78 79 def __pad(self, seq, front): 80 # Add '' to seq to either the front (is True), else the back. 81 seq = list(seq) 82 if front: 83 seq.insert(0, '') 84 else: 85 seq.append('') 86 return seq 87 88 def __calc_weekday(self): 89 # Set self.a_weekday and self.f_weekday using the calendar 90 # module. 91 a_weekday = [calendar.day_abbr[i].lower() for i in range(7)] 92 f_weekday = [calendar.day_name[i].lower() for i in range(7)] 93 self.a_weekday = a_weekday 94 self.f_weekday = f_weekday 95 96 def __calc_month(self): 97 # Set self.f_month and self.a_month using the calendar module. 98 a_month = [calendar.month_abbr[i].lower() for i in range(13)] 99 f_month = [calendar.month_name[i].lower() for i in range(13)] 100 self.a_month = a_month 101 self.f_month = f_month 102 103 def __calc_am_pm(self): 104 # Set self.am_pm by using time.strftime(). 105 106 # The magic date (1999,3,17,hour,44,55,2,76,0) is not really that 107 # magical; just happened to have used it everywhere else where a 108 # static date was needed. 109 am_pm = [] 110 for hour in (01,22): 111 time_tuple = time.struct_time((1999,3,17,hour,44,55,2,76,0)) 112 am_pm.append(time.strftime("%p", time_tuple).lower()) 113 self.am_pm = am_pm 114 115 def __calc_date_time(self): 116 # Set self.date_time, self.date, & self.time by using 117 # time.strftime(). 118 119 # Use (1999,3,17,22,44,55,2,76,0) for magic date because the amount of 120 # overloaded numbers is minimized. The order in which searches for 121 # values within the format string is very important; it eliminates 122 # possible ambiguity for what something represents. 123 time_tuple = time.struct_time((1999,3,17,22,44,55,2,76,0)) 124 date_time = [None, None, None] 125 date_time[0] = time.strftime("%c", time_tuple).lower() 126 date_time[1] = time.strftime("%x", time_tuple).lower() 127 date_time[2] = time.strftime("%X", time_tuple).lower() 128 replacement_pairs = [('%', '%%'), (self.f_weekday[2], '%A'), 129 (self.f_month[3], '%B'), (self.a_weekday[2], '%a'), 130 (self.a_month[3], '%b'), (self.am_pm[1], '%p'), 131 ('1999', '%Y'), ('99', '%y'), ('22', '%H'), 132 ('44', '%M'), ('55', '%S'), ('76', '%j'), 133 ('17', '%d'), ('03', '%m'), ('3', '%m'), 134 # '3' needed for when no leading zero. 135 ('2', '%w'), ('10', '%I')] 136 replacement_pairs.extend([(tz, "%Z") for tz_values in self.timezone 137 for tz in tz_values]) 138 for offset,directive in ((0,'%c'), (1,'%x'), (2,'%X')): 139 current_format = date_time[offset] 140 for old, new in replacement_pairs: 141 # Must deal with possible lack of locale info 142 # manifesting itself as the empty string (e.g., Swedish's 143 # lack of AM/PM info) or a platform returning a tuple of empty 144 # strings (e.g., MacOS 9 having timezone as ('','')). 145 if old: 146 current_format = current_format.replace(old, new) 147 # If %W is used, then Sunday, 2005-01-03 will fall on week 0 since 148 # 2005-01-03 occurs before the first Monday of the year. Otherwise 149 # %U is used. 150 time_tuple = time.struct_time((1999,1,3,1,1,1,6,3,0)) 151 if '00' in time.strftime(directive, time_tuple): 152 U_W = '%W' 153 else: 154 U_W = '%U' 155 date_time[offset] = current_format.replace('11', U_W) 156 self.LC_date_time = date_time[0] 157 self.LC_date = date_time[1] 158 self.LC_time = date_time[2] 159 160 def __calc_timezone(self): 161 # Set self.timezone by using time.tzname. 162 # Do not worry about possibility of time.tzname[0] == timetzname[1] 163 # and time.daylight; handle that in strptime . 164 try: 165 time.tzset() 166 except AttributeError: 167 pass 168 no_saving = frozenset(["utc", "gmt", time.tzname[0].lower()]) 169 if time.daylight: 170 has_saving = frozenset([time.tzname[1].lower()]) 171 else: 172 has_saving = frozenset() 173 self.timezone = (no_saving, has_saving) 174 175 176class TimeRE(dict): 177 """Handle conversion from format directives to regexes.""" 178 179 def __init__(self, locale_time=None): 180 """Create keys/values. 181 182 Order of execution is important for dependency reasons. 183 184 """ 185 if locale_time: 186 self.locale_time = locale_time 187 else: 188 self.locale_time = LocaleTime() 189 base = super(TimeRE, self) 190 base.__init__({ 191 # The " \d" part of the regex is to make %c from ANSI C work 192 'd': r"(?P<d>3[0-1]|[1-2]\d|0[1-9]|[1-9]| [1-9])", 193 'H': r"(?P<H>2[0-3]|[0-1]\d|\d)", 194 'I': r"(?P<I>1[0-2]|0[1-9]|[1-9])", 195 'j': r"(?P<j>36[0-6]|3[0-5]\d|[1-2]\d\d|0[1-9]\d|00[1-9]|[1-9]\d|0[1-9]|[1-9])", 196 'm': r"(?P<m>1[0-2]|0[1-9]|[1-9])", 197 'M': r"(?P<M>[0-5]\d|\d)", 198 'S': r"(?P<S>6[0-1]|[0-5]\d|\d)", 199 'U': r"(?P<U>5[0-3]|[0-4]\d|\d)", 200 'w': r"(?P<w>[0-6])", 201 # W is set below by using 'U' 202 'y': r"(?P<y>\d\d)", 203 #XXX: Does 'Y' need to worry about having less or more than 204 # 4 digits? 205 'Y': r"(?P<Y>\d\d\d\d)", 206 'A': self.__seqToRE(self.locale_time.f_weekday, 'A'), 207 'a': self.__seqToRE(self.locale_time.a_weekday, 'a'), 208 'B': self.__seqToRE(self.locale_time.f_month[1:], 'B'), 209 'b': self.__seqToRE(self.locale_time.a_month[1:], 'b'), 210 'p': self.__seqToRE(self.locale_time.am_pm, 'p'), 211 'Z': self.__seqToRE((tz for tz_names in self.locale_time.timezone 212 for tz in tz_names), 213 'Z'), 214 '%': '%'}) 215 base.__setitem__('W', base.__getitem__('U').replace('U', 'W')) 216 base.__setitem__('c', self.pattern(self.locale_time.LC_date_time)) 217 base.__setitem__('x', self.pattern(self.locale_time.LC_date)) 218 base.__setitem__('X', self.pattern(self.locale_time.LC_time)) 219 220 def __seqToRE(self, to_convert, directive): 221 """Convert a list to a regex string for matching a directive. 222 223 Want possible matching values to be from longest to shortest. This 224 prevents the possibility of a match occuring for a value that also 225 a substring of a larger value that should have matched (e.g., 'abc' 226 matching when 'abcdef' should have been the match). 227 228 """ 229 to_convert = sorted(to_convert, key=len, reverse=True) 230 for value in to_convert: 231 if value != '': 232 break 233 else: 234 return '' 235 regex = '|'.join(re_escape(stuff) for stuff in to_convert) 236 regex = '(?P<%s>%s' % (directive, regex) 237 return '%s)' % regex 238 239 def pattern(self, format): 240 """Return regex pattern for the format string. 241 242 Need to make sure that any characters that might be interpreted as 243 regex syntax are escaped. 244 245 """ 246 processed_format = '' 247 # The sub() call escapes all characters that might be misconstrued 248 # as regex syntax. Cannot use re.escape since we have to deal with 249 # format directives (%m, etc.). 250 regex_chars = re_compile(r"([\\.^$*+?\(\){}\[\]|])") 251 format = regex_chars.sub(r"\\\1", format) 252 whitespace_replacement = re_compile('\s+') 253 format = whitespace_replacement.sub('\s*', format) 254 while '%' in format: 255 directive_index = format.index('%')+1 256 processed_format = "%s%s%s" % (processed_format, 257 format[:directive_index-1], 258 self[format[directive_index]]) 259 format = format[directive_index+1:] 260 return "%s%s" % (processed_format, format) 261 262 def compile(self, format): 263 """Return a compiled re object for the format string.""" 264 return re_compile(self.pattern(format), IGNORECASE) 265 266_cache_lock = _thread_allocate_lock() 267# DO NOT modify _TimeRE_cache or _regex_cache without acquiring the cache lock 268# first! 269_TimeRE_cache = TimeRE() 270_CACHE_MAX_SIZE = 5 # Max number of regexes stored in _regex_cache 271_regex_cache = {} 272 273def _calc_julian_from_U_or_W(year, week_of_year, day_of_week, week_starts_Mon): 274 """Calculate the Julian day based on the year, week of the year, and day of 275 the week, with week_start_day representing whether the week of the year 276 assumes the week starts on Sunday or Monday (6 or 0).""" 277 first_weekday = datetime_date(year, 1, 1).weekday() 278 # If we are dealing with the %U directive (week starts on Sunday), it's 279 # easier to just shift the view to Sunday being the first day of the 280 # week. 281 if not week_starts_Mon: 282 first_weekday = (first_weekday + 1) % 7 283 day_of_week = (day_of_week + 1) % 7 284 # Need to watch out for a week 0 (when the first day of the year is not 285 # the same as that specified by %U or %W). 286 week_0_length = (7 - first_weekday) % 7 287 if week_of_year == 0: 288 return 1 + day_of_week - first_weekday 289 else: 290 days_to_week = week_0_length + (7 * (week_of_year - 1)) 291 return 1 + days_to_week + day_of_week 292 293 294def strptime(data_string, format="%a %b %d %H:%M:%S %Y"): 295 """Return a time struct based on the input string and the format string.""" 296 global _TimeRE_cache, _regex_cache 297 with _cache_lock: 298 time_re = _TimeRE_cache 299 locale_time = time_re.locale_time 300 if _getlang() != locale_time.lang: 301 _TimeRE_cache = TimeRE() 302 _regex_cache = {} 303 if len(_regex_cache) > _CACHE_MAX_SIZE: 304 _regex_cache.clear() 305 format_regex = _regex_cache.get(format) 306 if not format_regex: 307 try: 308 format_regex = time_re.compile(format) 309 # KeyError raised when a bad format is found; can be specified as 310 # \\, in which case it was a stray % but with a space after it 311 except KeyError as err: 312 bad_directive = err.args[0] 313 if bad_directive == "\\": 314 bad_directive = "%" 315 del err 316 raise ValueError("'%s' is a bad directive in format '%s'" % 317 (bad_directive, format)) 318 # IndexError only occurs when the format string is "%" 319 except IndexError: 320 raise ValueError("stray %% in format '%s'" % format) 321 _regex_cache[format] = format_regex 322 found = format_regex.match(data_string) 323 if not found: 324 raise ValueError("time data %r does not match format %r" % 325 (data_string, format)) 326 if len(data_string) != found.end(): 327 raise ValueError("unconverted data remains: %s" % 328 data_string[found.end():]) 329 year = 1900 330 month = day = 1 331 hour = minute = second = 0 332 tz = -1 333 # Default to -1 to signify that values not known; not critical to have, 334 # though 335 week_of_year = -1 336 week_of_year_start = -1 337 # weekday and julian defaulted to -1 so as to signal need to calculate 338 # values 339 weekday = julian = -1 340 found_dict = found.groupdict() 341 for group_key in found_dict.keys(): 342 # Directives not explicitly handled below: 343 # c, x, X 344 # handled by making out of other directives 345 # U, W 346 # worthless without day of the week 347 if group_key == 'y': 348 year = int(found_dict['y']) 349 # Open Group specification for strptime() states that a %y 350 #value in the range of [00, 68] is in the century 2000, while 351 #[69,99] is in the century 1900 352 if year <= 68: 353 year += 2000 354 else: 355 year += 1900 356 elif group_key == 'Y': 357 year = int(found_dict['Y']) 358 elif group_key == 'm': 359 month = int(found_dict['m']) 360 elif group_key == 'B': 361 month = locale_time.f_month.index(found_dict['B'].lower()) 362 elif group_key == 'b': 363 month = locale_time.a_month.index(found_dict['b'].lower()) 364 elif group_key == 'd': 365 day = int(found_dict['d']) 366 elif group_key == 'H': 367 hour = int(found_dict['H']) 368 elif group_key == 'I': 369 hour = int(found_dict['I']) 370 ampm = found_dict.get('p', '').lower() 371 # If there was no AM/PM indicator, we'll treat this like AM 372 if ampm in ('', locale_time.am_pm[0]): 373 # We're in AM so the hour is correct unless we're 374 # looking at 12 midnight. 375 # 12 midnight == 12 AM == hour 0 376 if hour == 12: 377 hour = 0 378 elif ampm == locale_time.am_pm[1]: 379 # We're in PM so we need to add 12 to the hour unless 380 # we're looking at 12 noon. 381 # 12 noon == 12 PM == hour 12 382 if hour != 12: 383 hour += 12 384 elif group_key == 'M': 385 minute = int(found_dict['M']) 386 elif group_key == 'S': 387 second = int(found_dict['S']) 388 elif group_key == 'A': 389 weekday = locale_time.f_weekday.index(found_dict['A'].lower()) 390 elif group_key == 'a': 391 weekday = locale_time.a_weekday.index(found_dict['a'].lower()) 392 elif group_key == 'w': 393 weekday = int(found_dict['w']) 394 if weekday == 0: 395 weekday = 6 396 else: 397 weekday -= 1 398 elif group_key == 'j': 399 julian = int(found_dict['j']) 400 elif group_key in ('U', 'W'): 401 week_of_year = int(found_dict[group_key]) 402 if group_key == 'U': 403 # U starts week on Sunday. 404 week_of_year_start = 6 405 else: 406 # W starts week on Monday. 407 week_of_year_start = 0 408 elif group_key == 'Z': 409 # Since -1 is default value only need to worry about setting tz if 410 # it can be something other than -1. 411 found_zone = found_dict['Z'].lower() 412 for value, tz_values in enumerate(locale_time.timezone): 413 if found_zone in tz_values: 414 # Deal with bad locale setup where timezone names are the 415 # same and yet time.daylight is true; too ambiguous to 416 # be able to tell what timezone has daylight savings 417 if (time.tzname[0] == time.tzname[1] and 418 time.daylight and found_zone not in ("utc", "gmt")): 419 break 420 else: 421 tz = value 422 break 423 # If we know the week of the year and what day of that week, we can figure 424 # out the Julian day of the year. 425 if julian == -1 and week_of_year != -1 and weekday != -1: 426 week_starts_Mon = True if week_of_year_start == 0 else False 427 julian = _calc_julian_from_U_or_W(year, week_of_year, weekday, 428 week_starts_Mon) 429 # Cannot pre-calculate datetime_date() since can change in Julian 430 # calculation and thus could have different value for the day of the week 431 # calculation. 432 if julian == -1: 433 # Need to add 1 to result since first day of the year is 1, not 0. 434 julian = datetime_date(year, month, day).toordinal() - \ 435 datetime_date(year, 1, 1).toordinal() + 1 436 else: # Assume that if they bothered to include Julian day it will 437 # be accurate. 438 datetime_result = datetime_date.fromordinal((julian - 1) + datetime_date(year, 1, 1).toordinal()) 439 year = datetime_result.year 440 month = datetime_result.month 441 day = datetime_result.day 442 if weekday == -1: 443 weekday = datetime_date(year, month, day).weekday() 444 return time.struct_time((year, month, day, 445 hour, minute, second, 446 weekday, julian, tz)) 447