_strptime.py revision a783d06f8cb4077dd34567fde418bf6c23073912
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__author__ = "Brett Cannon"
26__email__ = "brett@python.org"
27
28__all__ = ['strptime']
29
30def _getlang():
31    # Figure out what the current language is set to.
32    return locale.getlocale(locale.LC_TIME)
33
34class LocaleTime(object):
35    """Stores and handles locale-specific information related to time.
36
37    ATTRIBUTES:
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
41                    is added by code)
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
49                    (2-item list of sets)
50        lang -- Language used by instance (2-item tuple)
51    """
52
53    def __init__(self):
54        """Set all attributes.
55
56        Order of methods called matters for dependency reasons.
57
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
66        doing this.
67
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.
71
72        """
73        self.lang = _getlang()
74        self.__calc_weekday()
75        self.__calc_month()
76        self.__calc_am_pm()
77        self.__calc_timezone()
78        self.__calc_date_time()
79        if _getlang() != self.lang:
80            raise ValueError("locale changed during initialization")
81
82    def __pad(self, seq, front):
83        # Add '' to seq to either the front (is True), else the back.
84        seq = list(seq)
85        if front:
86            seq.insert(0, '')
87        else:
88            seq.append('')
89        return seq
90
91    def __calc_weekday(self):
92        # Set self.a_weekday and self.f_weekday using the calendar
93        # module.
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
98
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
105
106    def __calc_am_pm(self):
107        # Set self.am_pm by using time.strftime().
108
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.
112        am_pm = []
113        for hour in (01,22):
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())
116        self.am_pm = am_pm
117
118    def __calc_date_time(self):
119        # Set self.date_time, self.date, & self.time by using
120        # time.strftime().
121
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 ('','')).
148                if old:
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
152            # %U is used.
153            time_tuple = time.struct_time((1999,1,3,1,1,1,6,3,0))
154            if '00' in time.strftime(directive, time_tuple):
155                U_W = '%W'
156            else:
157                U_W = '%U'
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]
162
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 .
167        try:
168            time.tzset()
169        except AttributeError:
170            pass
171        no_saving = frozenset(["utc", "gmt", time.tzname[0].lower()])
172        if time.daylight:
173            has_saving = frozenset([time.tzname[1].lower()])
174        else:
175            has_saving = frozenset()
176        self.timezone = (no_saving, has_saving)
177
178
179class TimeRE(dict):
180    """Handle conversion from format directives to regexes."""
181
182    def __init__(self, locale_time=None):
183        """Create keys/values.
184
185        Order of execution is important for dependency reasons.
186
187        """
188        if locale_time:
189            self.locale_time = locale_time
190        else:
191            self.locale_time = LocaleTime()
192        base = super(TimeRE, self)
193        base.__init__({
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'
205            'y': r"(?P<y>\d\d)",
206            #XXX: Does 'Y' need to worry about having less or more than
207            #     4 digits?
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
215                                        for tz in tz_names),
216                                'Z'),
217            '%': '%'})
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))
222
223    def __seqToRE(self, to_convert, directive):
224        """Convert a list to a regex string for matching a directive.
225
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).
230
231        """
232        to_convert = sorted(to_convert, key=len, reverse=True)
233        for value in to_convert:
234            if value != '':
235                break
236        else:
237            return ''
238        regex = '|'.join(re_escape(stuff) for stuff in to_convert)
239        regex = '(?P<%s>%s' % (directive, regex)
240        return '%s)' % regex
241
242    def pattern(self, format):
243        """Return regex pattern for the format string.
244
245        Need to make sure that any characters that might be interpreted as
246        regex syntax are escaped.
247
248        """
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)
257        while '%' in 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)
264
265    def compile(self, format):
266        """Return a compiled re object for the format string."""
267        return re_compile(self.pattern(format), IGNORECASE)
268
269_cache_lock = _thread_allocate_lock()
270# DO NOT modify _TimeRE_cache or _regex_cache without acquiring the cache lock
271# first!
272_TimeRE_cache = TimeRE()
273_CACHE_MAX_SIZE = 5 # Max number of regexes stored in _regex_cache
274_regex_cache = {}
275
276def strptime(data_string, format="%a %b %d %H:%M:%S %Y"):
277    """Return a time struct based on the input string and the format string."""
278    global _TimeRE_cache, _regex_cache
279    _cache_lock.acquire()
280    try:
281        time_re = _TimeRE_cache
282        locale_time = time_re.locale_time
283        if _getlang() != locale_time.lang:
284            _TimeRE_cache = TimeRE()
285            _regex_cache = {}
286        if len(_regex_cache) > _CACHE_MAX_SIZE:
287            _regex_cache.clear()
288        format_regex = _regex_cache.get(format)
289        if not format_regex:
290            format_regex = time_re.compile(format)
291            _regex_cache[format] = format_regex
292    finally:
293        _cache_lock.release()
294    found = format_regex.match(data_string)
295    if not found:
296        raise ValueError("time data did not match format:  data=%s  fmt=%s" %
297                         (data_string, format))
298    if len(data_string) != found.end():
299        raise ValueError("unconverted data remains: %s" %
300                          data_string[found.end():])
301    year = 1900
302    month = day = 1
303    hour = minute = second = 0
304    tz = -1
305    # Default to -1 to signify that values not known; not critical to have,
306    # though
307    week_of_year = -1
308    week_of_year_start = -1
309    # weekday and julian defaulted to -1 so as to signal need to calculate
310    # values
311    weekday = julian = -1
312    found_dict = found.groupdict()
313    for group_key in found_dict.iterkeys():
314        # Directives not explicitly handled below:
315        #   c, x, X
316        #      handled by making out of other directives
317        #   U, W
318        #      worthless without day of the week
319        if group_key == 'y':
320            year = int(found_dict['y'])
321            # Open Group specification for strptime() states that a %y
322            #value in the range of [00, 68] is in the century 2000, while
323            #[69,99] is in the century 1900
324            if year <= 68:
325                year += 2000
326            else:
327                year += 1900
328        elif group_key == 'Y':
329            year = int(found_dict['Y'])
330        elif group_key == 'm':
331            month = int(found_dict['m'])
332        elif group_key == 'B':
333            month = locale_time.f_month.index(found_dict['B'].lower())
334        elif group_key == 'b':
335            month = locale_time.a_month.index(found_dict['b'].lower())
336        elif group_key == 'd':
337            day = int(found_dict['d'])
338        elif group_key == 'H':
339            hour = int(found_dict['H'])
340        elif group_key == 'I':
341            hour = int(found_dict['I'])
342            ampm = found_dict.get('p', '').lower()
343            # If there was no AM/PM indicator, we'll treat this like AM
344            if ampm in ('', locale_time.am_pm[0]):
345                # We're in AM so the hour is correct unless we're
346                # looking at 12 midnight.
347                # 12 midnight == 12 AM == hour 0
348                if hour == 12:
349                    hour = 0
350            elif ampm == locale_time.am_pm[1]:
351                # We're in PM so we need to add 12 to the hour unless
352                # we're looking at 12 noon.
353                # 12 noon == 12 PM == hour 12
354                if hour != 12:
355                    hour += 12
356        elif group_key == 'M':
357            minute = int(found_dict['M'])
358        elif group_key == 'S':
359            second = int(found_dict['S'])
360        elif group_key == 'A':
361            weekday = locale_time.f_weekday.index(found_dict['A'].lower())
362        elif group_key == 'a':
363            weekday = locale_time.a_weekday.index(found_dict['a'].lower())
364        elif group_key == 'w':
365            weekday = int(found_dict['w'])
366            if weekday == 0:
367                weekday = 6
368            else:
369                weekday -= 1
370        elif group_key == 'j':
371            julian = int(found_dict['j'])
372        elif group_key in ('U', 'W'):
373            week_of_year = int(found_dict[group_key])
374            if group_key == 'U':
375                # U starts week on Sunday
376                week_of_year_start = 6
377            else:
378                # W starts week on Monday
379                week_of_year_start = 0
380        elif group_key == 'Z':
381            # Since -1 is default value only need to worry about setting tz if
382            # it can be something other than -1.
383            found_zone = found_dict['Z'].lower()
384            for value, tz_values in enumerate(locale_time.timezone):
385                if found_zone in tz_values:
386                    # Deal with bad locale setup where timezone names are the
387                    # same and yet time.daylight is true; too ambiguous to
388                    # be able to tell what timezone has daylight savings
389                    if (time.tzname[0] == time.tzname[1] and
390                       time.daylight and found_zone not in ("utc", "gmt")):
391                        break
392                    else:
393                        tz = value
394                        break
395    # If we know the week of the year and what day of that week, we can figure
396    # out the Julian day of the year
397    # Calculations below assume 0 is a Monday
398    if julian == -1 and week_of_year != -1 and weekday != -1:
399        # Calculate how many days in week 0
400        first_weekday = datetime_date(year, 1, 1).weekday()
401        preceeding_days = 7 - first_weekday
402        if preceeding_days == 7:
403            preceeding_days = 0
404        # Adjust for U directive so that calculations are not dependent on
405        # directive used to figure out week of year
406        if weekday == 6 and week_of_year_start == 6:
407            week_of_year -= 1
408        # If a year starts and ends on a Monday but a week is specified to
409        # start on a Sunday we need to up the week to counter-balance the fact
410        # that with %W that first Monday starts week 1 while with %U that is
411        # week 0 and thus shifts everything by a week
412        if weekday == 0 and first_weekday == 0 and week_of_year_start == 6:
413            week_of_year += 1
414        # If in week 0, then just figure out how many days from Jan 1 to day of
415        # week specified, else calculate by multiplying week of year by 7,
416        # adding in days in week 0, and the number of days from Monday to the
417        # day of the week
418        if week_of_year == 0:
419            julian = 1 + weekday - first_weekday
420        else:
421            days_to_week = preceeding_days + (7 * (week_of_year - 1))
422            julian = 1 + days_to_week + weekday
423    # Cannot pre-calculate datetime_date() since can change in Julian
424    #calculation and thus could have different value for the day of the week
425    #calculation
426    if julian == -1:
427        # Need to add 1 to result since first day of the year is 1, not 0.
428        julian = datetime_date(year, month, day).toordinal() - \
429                  datetime_date(year, 1, 1).toordinal() + 1
430    else:  # Assume that if they bothered to include Julian day it will
431           #be accurate
432        datetime_result = datetime_date.fromordinal((julian - 1) + datetime_date(year, 1, 1).toordinal())
433        year = datetime_result.year
434        month = datetime_result.month
435        day = datetime_result.day
436    if weekday == -1:
437        weekday = datetime_date(year, month, day).weekday()
438    return time.struct_time((year, month, day,
439                             hour, minute, second,
440                             weekday, julian, tz))
441