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