_strptime.py revision ab68fcaee3c39ad21cf3a62f6d299e79a2c1fd56
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__ = [] 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 if time.tzname != self.tzname or time.daylight != self.daylight: 79 raise ValueError("timezone changed during initialization") 80 81 def __pad(self, seq, front): 82 # Add '' to seq to either the front (is True), else the back. 83 seq = list(seq) 84 if front: 85 seq.insert(0, '') 86 else: 87 seq.append('') 88 return seq 89 90 def __calc_weekday(self): 91 # Set self.a_weekday and self.f_weekday using the calendar 92 # module. 93 a_weekday = [calendar.day_abbr[i].lower() for i in range(7)] 94 f_weekday = [calendar.day_name[i].lower() for i in range(7)] 95 self.a_weekday = a_weekday 96 self.f_weekday = f_weekday 97 98 def __calc_month(self): 99 # Set self.f_month and self.a_month using the calendar module. 100 a_month = [calendar.month_abbr[i].lower() for i in range(13)] 101 f_month = [calendar.month_name[i].lower() for i in range(13)] 102 self.a_month = a_month 103 self.f_month = f_month 104 105 def __calc_am_pm(self): 106 # Set self.am_pm by using time.strftime(). 107 108 # The magic date (1999,3,17,hour,44,55,2,76,0) is not really that 109 # magical; just happened to have used it everywhere else where a 110 # static date was needed. 111 am_pm = [] 112 for hour in (01,22): 113 time_tuple = time.struct_time((1999,3,17,hour,44,55,2,76,0)) 114 am_pm.append(time.strftime("%p", time_tuple).lower()) 115 self.am_pm = am_pm 116 117 def __calc_date_time(self): 118 # Set self.date_time, self.date, & self.time by using 119 # time.strftime(). 120 121 # Use (1999,3,17,22,44,55,2,76,0) for magic date because the amount of 122 # overloaded numbers is minimized. The order in which searches for 123 # values within the format string is very important; it eliminates 124 # possible ambiguity for what something represents. 125 time_tuple = time.struct_time((1999,3,17,22,44,55,2,76,0)) 126 date_time = [None, None, None] 127 date_time[0] = time.strftime("%c", time_tuple).lower() 128 date_time[1] = time.strftime("%x", time_tuple).lower() 129 date_time[2] = time.strftime("%X", time_tuple).lower() 130 replacement_pairs = [('%', '%%'), (self.f_weekday[2], '%A'), 131 (self.f_month[3], '%B'), (self.a_weekday[2], '%a'), 132 (self.a_month[3], '%b'), (self.am_pm[1], '%p'), 133 ('1999', '%Y'), ('99', '%y'), ('22', '%H'), 134 ('44', '%M'), ('55', '%S'), ('76', '%j'), 135 ('17', '%d'), ('03', '%m'), ('3', '%m'), 136 # '3' needed for when no leading zero. 137 ('2', '%w'), ('10', '%I')] 138 replacement_pairs.extend([(tz, "%Z") for tz_values in self.timezone 139 for tz in tz_values]) 140 for offset,directive in ((0,'%c'), (1,'%x'), (2,'%X')): 141 current_format = date_time[offset] 142 for old, new in replacement_pairs: 143 # Must deal with possible lack of locale info 144 # manifesting itself as the empty string (e.g., Swedish's 145 # lack of AM/PM info) or a platform returning a tuple of empty 146 # strings (e.g., MacOS 9 having timezone as ('','')). 147 if old: 148 current_format = current_format.replace(old, new) 149 # If %W is used, then Sunday, 2005-01-03 will fall on week 0 since 150 # 2005-01-03 occurs before the first Monday of the year. Otherwise 151 # %U is used. 152 time_tuple = time.struct_time((1999,1,3,1,1,1,6,3,0)) 153 if '00' in time.strftime(directive, time_tuple): 154 U_W = '%W' 155 else: 156 U_W = '%U' 157 date_time[offset] = current_format.replace('11', U_W) 158 self.LC_date_time = date_time[0] 159 self.LC_date = date_time[1] 160 self.LC_time = date_time[2] 161 162 def __calc_timezone(self): 163 # Set self.timezone by using time.tzname. 164 # Do not worry about possibility of time.tzname[0] == time.tzname[1] 165 # and time.daylight; handle that in strptime. 166 try: 167 time.tzset() 168 except AttributeError: 169 pass 170 self.tzname = time.tzname 171 self.daylight = time.daylight 172 no_saving = frozenset(["utc", "gmt", self.tzname[0].lower()]) 173 if self.daylight: 174 has_saving = frozenset([self.tzname[1].lower()]) 175 else: 176 has_saving = frozenset() 177 self.timezone = (no_saving, has_saving) 178 179 180class TimeRE(dict): 181 """Handle conversion from format directives to regexes.""" 182 183 def __init__(self, locale_time=None): 184 """Create keys/values. 185 186 Order of execution is important for dependency reasons. 187 188 """ 189 if locale_time: 190 self.locale_time = locale_time 191 else: 192 self.locale_time = LocaleTime() 193 base = super(TimeRE, self) 194 base.__init__({ 195 # The " \d" part of the regex is to make %c from ANSI C work 196 'd': r"(?P<d>3[0-1]|[1-2]\d|0[1-9]|[1-9]| [1-9])", 197 'f': r"(?P<f>[0-9]{1,6})", 198 'H': r"(?P<H>2[0-3]|[0-1]\d|\d)", 199 'I': r"(?P<I>1[0-2]|0[1-9]|[1-9])", 200 '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])", 201 'm': r"(?P<m>1[0-2]|0[1-9]|[1-9])", 202 'M': r"(?P<M>[0-5]\d|\d)", 203 'S': r"(?P<S>6[0-1]|[0-5]\d|\d)", 204 'U': r"(?P<U>5[0-3]|[0-4]\d|\d)", 205 'w': r"(?P<w>[0-6])", 206 # W is set below by using 'U' 207 'y': r"(?P<y>\d\d)", 208 #XXX: Does 'Y' need to worry about having less or more than 209 # 4 digits? 210 'Y': r"(?P<Y>\d\d\d\d)", 211 'A': self.__seqToRE(self.locale_time.f_weekday, 'A'), 212 'a': self.__seqToRE(self.locale_time.a_weekday, 'a'), 213 'B': self.__seqToRE(self.locale_time.f_month[1:], 'B'), 214 'b': self.__seqToRE(self.locale_time.a_month[1:], 'b'), 215 'p': self.__seqToRE(self.locale_time.am_pm, 'p'), 216 'Z': self.__seqToRE((tz for tz_names in self.locale_time.timezone 217 for tz in tz_names), 218 'Z'), 219 '%': '%'}) 220 base.__setitem__('W', base.__getitem__('U').replace('U', 'W')) 221 base.__setitem__('c', self.pattern(self.locale_time.LC_date_time)) 222 base.__setitem__('x', self.pattern(self.locale_time.LC_date)) 223 base.__setitem__('X', self.pattern(self.locale_time.LC_time)) 224 225 def __seqToRE(self, to_convert, directive): 226 """Convert a list to a regex string for matching a directive. 227 228 Want possible matching values to be from longest to shortest. This 229 prevents the possibility of a match occurring for a value that also 230 a substring of a larger value that should have matched (e.g., 'abc' 231 matching when 'abcdef' should have been the match). 232 233 """ 234 to_convert = sorted(to_convert, key=len, reverse=True) 235 for value in to_convert: 236 if value != '': 237 break 238 else: 239 return '' 240 regex = '|'.join(re_escape(stuff) for stuff in to_convert) 241 regex = '(?P<%s>%s' % (directive, regex) 242 return '%s)' % regex 243 244 def pattern(self, format): 245 """Return regex pattern for the format string. 246 247 Need to make sure that any characters that might be interpreted as 248 regex syntax are escaped. 249 250 """ 251 processed_format = '' 252 # The sub() call escapes all characters that might be misconstrued 253 # as regex syntax. Cannot use re.escape since we have to deal with 254 # format directives (%m, etc.). 255 regex_chars = re_compile(r"([\\.^$*+?\(\){}\[\]|])") 256 format = regex_chars.sub(r"\\\1", format) 257 whitespace_replacement = re_compile('\s+') 258 format = whitespace_replacement.sub('\s+', format) 259 while '%' in format: 260 directive_index = format.index('%')+1 261 processed_format = "%s%s%s" % (processed_format, 262 format[:directive_index-1], 263 self[format[directive_index]]) 264 format = format[directive_index+1:] 265 return "%s%s" % (processed_format, format) 266 267 def compile(self, format): 268 """Return a compiled re object for the format string.""" 269 return re_compile(self.pattern(format), IGNORECASE) 270 271_cache_lock = _thread_allocate_lock() 272# DO NOT modify _TimeRE_cache or _regex_cache without acquiring the cache lock 273# first! 274_TimeRE_cache = TimeRE() 275_CACHE_MAX_SIZE = 5 # Max number of regexes stored in _regex_cache 276_regex_cache = {} 277 278def _calc_julian_from_U_or_W(year, week_of_year, day_of_week, week_starts_Mon): 279 """Calculate the Julian day based on the year, week of the year, and day of 280 the week, with week_start_day representing whether the week of the year 281 assumes the week starts on Sunday or Monday (6 or 0).""" 282 first_weekday = datetime_date(year, 1, 1).weekday() 283 # If we are dealing with the %U directive (week starts on Sunday), it's 284 # easier to just shift the view to Sunday being the first day of the 285 # week. 286 if not week_starts_Mon: 287 first_weekday = (first_weekday + 1) % 7 288 day_of_week = (day_of_week + 1) % 7 289 # Need to watch out for a week 0 (when the first day of the year is not 290 # the same as that specified by %U or %W). 291 week_0_length = (7 - first_weekday) % 7 292 if week_of_year == 0: 293 return 1 + day_of_week - first_weekday 294 else: 295 days_to_week = week_0_length + (7 * (week_of_year - 1)) 296 return 1 + days_to_week + day_of_week 297 298 299def _strptime(data_string, format="%a %b %d %H:%M:%S %Y"): 300 """Return a time struct based on the input string and the format string.""" 301 global _TimeRE_cache, _regex_cache 302 with _cache_lock: 303 locale_time = _TimeRE_cache.locale_time 304 if (_getlang() != locale_time.lang or 305 time.tzname != locale_time.tzname or 306 time.daylight != locale_time.daylight): 307 _TimeRE_cache = TimeRE() 308 _regex_cache.clear() 309 locale_time = _TimeRE_cache.locale_time 310 if len(_regex_cache) > _CACHE_MAX_SIZE: 311 _regex_cache.clear() 312 format_regex = _regex_cache.get(format) 313 if not format_regex: 314 try: 315 format_regex = _TimeRE_cache.compile(format) 316 # KeyError raised when a bad format is found; can be specified as 317 # \\, in which case it was a stray % but with a space after it 318 except KeyError, err: 319 bad_directive = err.args[0] 320 if bad_directive == "\\": 321 bad_directive = "%" 322 del err 323 raise ValueError("'%s' is a bad directive in format '%s'" % 324 (bad_directive, format)) 325 # IndexError only occurs when the format string is "%" 326 except IndexError: 327 raise ValueError("stray %% in format '%s'" % format) 328 _regex_cache[format] = format_regex 329 found = format_regex.match(data_string) 330 if not found: 331 raise ValueError("time data %r does not match format %r" % 332 (data_string, format)) 333 if len(data_string) != found.end(): 334 raise ValueError("unconverted data remains: %s" % 335 data_string[found.end():]) 336 337 year = None 338 month = day = 1 339 hour = minute = second = fraction = 0 340 tz = -1 341 # Default to -1 to signify that values not known; not critical to have, 342 # though 343 week_of_year = -1 344 week_of_year_start = -1 345 # weekday and julian defaulted to None so as to signal need to calculate 346 # values 347 weekday = julian = None 348 found_dict = found.groupdict() 349 for group_key in found_dict.iterkeys(): 350 # Directives not explicitly handled below: 351 # c, x, X 352 # handled by making out of other directives 353 # U, W 354 # worthless without day of the week 355 if group_key == 'y': 356 year = int(found_dict['y']) 357 # Open Group specification for strptime() states that a %y 358 #value in the range of [00, 68] is in the century 2000, while 359 #[69,99] is in the century 1900 360 if year <= 68: 361 year += 2000 362 else: 363 year += 1900 364 elif group_key == 'Y': 365 year = int(found_dict['Y']) 366 elif group_key == 'm': 367 month = int(found_dict['m']) 368 elif group_key == 'B': 369 month = locale_time.f_month.index(found_dict['B'].lower()) 370 elif group_key == 'b': 371 month = locale_time.a_month.index(found_dict['b'].lower()) 372 elif group_key == 'd': 373 day = int(found_dict['d']) 374 elif group_key == 'H': 375 hour = int(found_dict['H']) 376 elif group_key == 'I': 377 hour = int(found_dict['I']) 378 ampm = found_dict.get('p', '').lower() 379 # If there was no AM/PM indicator, we'll treat this like AM 380 if ampm in ('', locale_time.am_pm[0]): 381 # We're in AM so the hour is correct unless we're 382 # looking at 12 midnight. 383 # 12 midnight == 12 AM == hour 0 384 if hour == 12: 385 hour = 0 386 elif ampm == locale_time.am_pm[1]: 387 # We're in PM so we need to add 12 to the hour unless 388 # we're looking at 12 noon. 389 # 12 noon == 12 PM == hour 12 390 if hour != 12: 391 hour += 12 392 elif group_key == 'M': 393 minute = int(found_dict['M']) 394 elif group_key == 'S': 395 second = int(found_dict['S']) 396 elif group_key == 'f': 397 s = found_dict['f'] 398 # Pad to always return microseconds. 399 s += "0" * (6 - len(s)) 400 fraction = int(s) 401 elif group_key == 'A': 402 weekday = locale_time.f_weekday.index(found_dict['A'].lower()) 403 elif group_key == 'a': 404 weekday = locale_time.a_weekday.index(found_dict['a'].lower()) 405 elif group_key == 'w': 406 weekday = int(found_dict['w']) 407 if weekday == 0: 408 weekday = 6 409 else: 410 weekday -= 1 411 elif group_key == 'j': 412 julian = int(found_dict['j']) 413 elif group_key in ('U', 'W'): 414 week_of_year = int(found_dict[group_key]) 415 if group_key == 'U': 416 # U starts week on Sunday. 417 week_of_year_start = 6 418 else: 419 # W starts week on Monday. 420 week_of_year_start = 0 421 elif group_key == 'Z': 422 # Since -1 is default value only need to worry about setting tz if 423 # it can be something other than -1. 424 found_zone = found_dict['Z'].lower() 425 for value, tz_values in enumerate(locale_time.timezone): 426 if found_zone in tz_values: 427 # Deal with bad locale setup where timezone names are the 428 # same and yet time.daylight is true; too ambiguous to 429 # be able to tell what timezone has daylight savings 430 if (time.tzname[0] == time.tzname[1] and 431 time.daylight and found_zone not in ("utc", "gmt")): 432 break 433 else: 434 tz = value 435 break 436 leap_year_fix = False 437 if year is None and month == 2 and day == 29: 438 year = 1904 # 1904 is first leap year of 20th century 439 leap_year_fix = True 440 elif year is None: 441 year = 1900 442 # If we know the week of the year and what day of that week, we can figure 443 # out the Julian day of the year. 444 if julian is None and week_of_year != -1 and weekday is not None: 445 week_starts_Mon = True if week_of_year_start == 0 else False 446 julian = _calc_julian_from_U_or_W(year, week_of_year, weekday, 447 week_starts_Mon) 448 # Cannot pre-calculate datetime_date() since can change in Julian 449 # calculation and thus could have different value for the day of the week 450 # calculation. 451 if julian is None: 452 # Need to add 1 to result since first day of the year is 1, not 0. 453 julian = datetime_date(year, month, day).toordinal() - \ 454 datetime_date(year, 1, 1).toordinal() + 1 455 else: # Assume that if they bothered to include Julian day it will 456 # be accurate. 457 datetime_result = datetime_date.fromordinal((julian - 1) + datetime_date(year, 1, 1).toordinal()) 458 year = datetime_result.year 459 month = datetime_result.month 460 day = datetime_result.day 461 if weekday is None: 462 weekday = datetime_date(year, month, day).weekday() 463 if leap_year_fix: 464 # the caller didn't supply a year but asked for Feb 29th. We couldn't 465 # use the default of 1900 for computations. We set it back to ensure 466 # that February 29th is smaller than March 1st. 467 year = 1900 468 469 return (time.struct_time((year, month, day, 470 hour, minute, second, 471 weekday, julian, tz)), fraction) 472 473def _strptime_time(data_string, format="%a %b %d %H:%M:%S %Y"): 474 return _strptime(data_string, format)[0] 475