Add files using upload-large-folder tool

dateutil/parser/__init__.py

@@ -0,0 +1,61 @@
+# -*- coding: utf-8 -*-
+from ._parser import parse, parser, parserinfo, ParserError
+from ._parser import DEFAULTPARSER, DEFAULTTZPARSER
+from ._parser import UnknownTimezoneWarning
+
+from ._parser import __doc__
+
+from .isoparser import isoparser, isoparse
+
+__all__ = ['parse', 'parser', 'parserinfo',
+           'isoparse', 'isoparser',
+           'ParserError',
+           'UnknownTimezoneWarning']
+
+
+###
+# Deprecate portions of the private interface so that downstream code that
+# is improperly relying on it is given *some* notice.
+
+
+def __deprecated_private_func(f):
+    from functools import wraps
+    import warnings
+
+    msg = ('{name} is a private function and may break without warning, '
+           'it will be moved and or renamed in future versions.')
+    msg = msg.format(name=f.__name__)
+
+    @wraps(f)
+    def deprecated_func(*args, **kwargs):
+        warnings.warn(msg, DeprecationWarning)
+        return f(*args, **kwargs)
+
+    return deprecated_func
+
+def __deprecate_private_class(c):
+    import warnings
+
+    msg = ('{name} is a private class and may break without warning, '
+           'it will be moved and or renamed in future versions.')
+    msg = msg.format(name=c.__name__)
+
+    class private_class(c):
+        __doc__ = c.__doc__
+
+        def __init__(self, *args, **kwargs):
+            warnings.warn(msg, DeprecationWarning)
+            super(private_class, self).__init__(*args, **kwargs)
+
+    private_class.__name__ = c.__name__
+
+    return private_class
+
+
+from ._parser import _timelex, _resultbase
+from ._parser import _tzparser, _parsetz
+
+_timelex = __deprecate_private_class(_timelex)
+_tzparser = __deprecate_private_class(_tzparser)
+_resultbase = __deprecate_private_class(_resultbase)
+_parsetz = __deprecated_private_func(_parsetz)

dateutil/parser/_parser.py

@@ -0,0 +1,1613 @@
+# -*- coding: utf-8 -*-
+"""
+This module offers a generic date/time string parser which is able to parse
+most known formats to represent a date and/or time.
+
+This module attempts to be forgiving with regards to unlikely input formats,
+returning a datetime object even for dates which are ambiguous. If an element
+of a date/time stamp is omitted, the following rules are applied:
+
+- If AM or PM is left unspecified, a 24-hour clock is assumed, however, an hour
+  on a 12-hour clock (``0 <= hour <= 12``) *must* be specified if AM or PM is
+  specified.
+- If a time zone is omitted, a timezone-naive datetime is returned.
+
+If any other elements are missing, they are taken from the
+:class:`datetime.datetime` object passed to the parameter ``default``. If this
+results in a day number exceeding the valid number of days per month, the
+value falls back to the end of the month.
+
+Additional resources about date/time string formats can be found below:
+
+- `A summary of the international standard date and time notation
+  <https://www.cl.cam.ac.uk/~mgk25/iso-time.html>`_
+- `W3C Date and Time Formats <https://www.w3.org/TR/NOTE-datetime>`_
+- `Time Formats (Planetary Rings Node) <https://pds-rings.seti.org:443/tools/time_formats.html>`_
+- `CPAN ParseDate module
+  <https://metacpan.org/pod/release/MUIR/Time-modules-2013.0912/lib/Time/ParseDate.pm>`_
+- `Java SimpleDateFormat Class
+  <https://docs.oracle.com/javase/6/docs/api/java/text/SimpleDateFormat.html>`_
+"""
+from __future__ import unicode_literals
+
+import datetime
+import re
+import string
+import time
+import warnings
+
+from calendar import monthrange
+from io import StringIO
+
+import six
+from six import integer_types, text_type
+
+from decimal import Decimal
+
+from warnings import warn
+
+from .. import relativedelta
+from .. import tz
+
+__all__ = ["parse", "parserinfo", "ParserError"]
+
+
+# TODO: pandas.core.tools.datetimes imports this explicitly.  Might be worth
+# making public and/or figuring out if there is something we can
+# take off their plate.
+class _timelex(object):
+    # Fractional seconds are sometimes split by a comma
+    _split_decimal = re.compile("([.,])")
+
+    def __init__(self, instream):
+        if isinstance(instream, (bytes, bytearray)):
+            instream = instream.decode()
+
+        if isinstance(instream, text_type):
+            instream = StringIO(instream)
+        elif getattr(instream, 'read', None) is None:
+            raise TypeError('Parser must be a string or character stream, not '
+                            '{itype}'.format(itype=instream.__class__.__name__))
+
+        self.instream = instream
+        self.charstack = []
+        self.tokenstack = []
+        self.eof = False
+
+    def get_token(self):
+        """
+        This function breaks the time string into lexical units (tokens), which
+        can be parsed by the parser. Lexical units are demarcated by changes in
+        the character set, so any continuous string of letters is considered
+        one unit, any continuous string of numbers is considered one unit.
+
+        The main complication arises from the fact that dots ('.') can be used
+        both as separators (e.g. "Sep.20.2009") or decimal points (e.g.
+        "4:30:21.447"). As such, it is necessary to read the full context of
+        any dot-separated strings before breaking it into tokens; as such, this
+        function maintains a "token stack", for when the ambiguous context
+        demands that multiple tokens be parsed at once.
+        """
+        if self.tokenstack:
+            return self.tokenstack.pop(0)
+
+        seenletters = False
+        token = None
+        state = None
+
+        while not self.eof:
+            # We only realize that we've reached the end of a token when we
+            # find a character that's not part of the current token - since
+            # that character may be part of the next token, it's stored in the
+            # charstack.
+            if self.charstack:
+                nextchar = self.charstack.pop(0)
+            else:
+                nextchar = self.instream.read(1)
+                while nextchar == '\x00':
+                    nextchar = self.instream.read(1)
+
+            if not nextchar:
+                self.eof = True
+                break
+            elif not state:
+                # First character of the token - determines if we're starting
+                # to parse a word, a number or something else.
+                token = nextchar
+                if self.isword(nextchar):
+                    state = 'a'
+                elif self.isnum(nextchar):
+                    state = '0'
+                elif self.isspace(nextchar):
+                    token = ' '
+                    break  # emit token
+                else:
+                    break  # emit token
+            elif state == 'a':
+                # If we've already started reading a word, we keep reading
+                # letters until we find something that's not part of a word.
+                seenletters = True
+                if self.isword(nextchar):
+                    token += nextchar
+                elif nextchar == '.':
+                    token += nextchar
+                    state = 'a.'
+                else:
+                    self.charstack.append(nextchar)
+                    break  # emit token
+            elif state == '0':
+                # If we've already started reading a number, we keep reading
+                # numbers until we find something that doesn't fit.
+                if self.isnum(nextchar):
+                    token += nextchar
+                elif nextchar == '.' or (nextchar == ',' and len(token) >= 2):
+                    token += nextchar
+                    state = '0.'
+                else:
+                    self.charstack.append(nextchar)
+                    break  # emit token
+            elif state == 'a.':
+                # If we've seen some letters and a dot separator, continue
+                # parsing, and the tokens will be broken up later.
+                seenletters = True
+                if nextchar == '.' or self.isword(nextchar):
+                    token += nextchar
+                elif self.isnum(nextchar) and token[-1] == '.':
+                    token += nextchar
+                    state = '0.'
+                else:
+                    self.charstack.append(nextchar)
+                    break  # emit token
+            elif state == '0.':
+                # If we've seen at least one dot separator, keep going, we'll
+                # break up the tokens later.
+                if nextchar == '.' or self.isnum(nextchar):
+                    token += nextchar
+                elif self.isword(nextchar) and token[-1] == '.':
+                    token += nextchar
+                    state = 'a.'
+                else:
+                    self.charstack.append(nextchar)
+                    break  # emit token
+
+        if (state in ('a.', '0.') and (seenletters or token.count('.') > 1 or
+                                       token[-1] in '.,')):
+            l = self._split_decimal.split(token)
+            token = l[0]
+            for tok in l[1:]:
+                if tok:
+                    self.tokenstack.append(tok)
+
+        if state == '0.' and token.count('.') == 0:
+            token = token.replace(',', '.')
+
+        return token
+
+    def __iter__(self):
+        return self
+
+    def __next__(self):
+        token = self.get_token()
+        if token is None:
+            raise StopIteration
+
+        return token
+
+    def next(self):
+        return self.__next__()  # Python 2.x support
+
+    @classmethod
+    def split(cls, s):
+        return list(cls(s))
+
+    @classmethod
+    def isword(cls, nextchar):
+        """ Whether or not the next character is part of a word """
+        return nextchar.isalpha()
+
+    @classmethod
+    def isnum(cls, nextchar):
+        """ Whether the next character is part of a number """
+        return nextchar.isdigit()
+
+    @classmethod
+    def isspace(cls, nextchar):
+        """ Whether the next character is whitespace """
+        return nextchar.isspace()
+
+
+class _resultbase(object):
+
+    def __init__(self):
+        for attr in self.__slots__:
+            setattr(self, attr, None)
+
+    def _repr(self, classname):
+        l = []
+        for attr in self.__slots__:
+            value = getattr(self, attr)
+            if value is not None:
+                l.append("%s=%s" % (attr, repr(value)))
+        return "%s(%s)" % (classname, ", ".join(l))
+
+    def __len__(self):
+        return (sum(getattr(self, attr) is not None
+                    for attr in self.__slots__))
+
+    def __repr__(self):
+        return self._repr(self.__class__.__name__)
+
+
+class parserinfo(object):
+    """
+    Class which handles what inputs are accepted. Subclass this to customize
+    the language and acceptable values for each parameter.
+
+    :param dayfirst:
+        Whether to interpret the first value in an ambiguous 3-integer date
+        (e.g. 01/05/09) as the day (``True``) or month (``False``). If
+        ``yearfirst`` is set to ``True``, this distinguishes between YDM
+        and YMD. Default is ``False``.
+
+    :param yearfirst:
+        Whether to interpret the first value in an ambiguous 3-integer date
+        (e.g. 01/05/09) as the year. If ``True``, the first number is taken
+        to be the year, otherwise the last number is taken to be the year.
+        Default is ``False``.
+    """
+
+    # m from a.m/p.m, t from ISO T separator
+    JUMP = [" ", ".", ",", ";", "-", "/", "'",
+            "at", "on", "and", "ad", "m", "t", "of",
+            "st", "nd", "rd", "th"]
+
+    WEEKDAYS = [("Mon", "Monday"),
+                ("Tue", "Tuesday"),     # TODO: "Tues"
+                ("Wed", "Wednesday"),
+                ("Thu", "Thursday"),    # TODO: "Thurs"
+                ("Fri", "Friday"),
+                ("Sat", "Saturday"),
+                ("Sun", "Sunday")]
+    MONTHS = [("Jan", "January"),
+              ("Feb", "February"),      # TODO: "Febr"
+              ("Mar", "March"),
+              ("Apr", "April"),
+              ("May", "May"),
+              ("Jun", "June"),
+              ("Jul", "July"),
+              ("Aug", "August"),
+              ("Sep", "Sept", "September"),
+              ("Oct", "October"),
+              ("Nov", "November"),
+              ("Dec", "December")]
+    HMS = [("h", "hour", "hours"),
+           ("m", "minute", "minutes"),
+           ("s", "second", "seconds")]
+    AMPM = [("am", "a"),
+            ("pm", "p")]
+    UTCZONE = ["UTC", "GMT", "Z", "z"]
+    PERTAIN = ["of"]
+    TZOFFSET = {}
+    # TODO: ERA = ["AD", "BC", "CE", "BCE", "Stardate",
+    #              "Anno Domini", "Year of Our Lord"]
+
+    def __init__(self, dayfirst=False, yearfirst=False):
+        self._jump = self._convert(self.JUMP)
+        self._weekdays = self._convert(self.WEEKDAYS)
+        self._months = self._convert(self.MONTHS)
+        self._hms = self._convert(self.HMS)
+        self._ampm = self._convert(self.AMPM)
+        self._utczone = self._convert(self.UTCZONE)
+        self._pertain = self._convert(self.PERTAIN)
+
+        self.dayfirst = dayfirst
+        self.yearfirst = yearfirst
+
+        self._year = time.localtime().tm_year
+        self._century = self._year // 100 * 100
+
+    def _convert(self, lst):
+        dct = {}
+        for i, v in enumerate(lst):
+            if isinstance(v, tuple):
+                for v in v:
+                    dct[v.lower()] = i
+            else:
+                dct[v.lower()] = i
+        return dct
+
+    def jump(self, name):
+        return name.lower() in self._jump
+
+    def weekday(self, name):
+        try:
+            return self._weekdays[name.lower()]
+        except KeyError:
+            pass
+        return None
+
+    def month(self, name):
+        try:
+            return self._months[name.lower()] + 1
+        except KeyError:
+            pass
+        return None
+
+    def hms(self, name):
+        try:
+            return self._hms[name.lower()]
+        except KeyError:
+            return None
+
+    def ampm(self, name):
+        try:
+            return self._ampm[name.lower()]
+        except KeyError:
+            return None
+
+    def pertain(self, name):
+        return name.lower() in self._pertain
+
+    def utczone(self, name):
+        return name.lower() in self._utczone
+
+    def tzoffset(self, name):
+        if name in self._utczone:
+            return 0
+
+        return self.TZOFFSET.get(name)
+
+    def convertyear(self, year, century_specified=False):
+        """
+        Converts two-digit years to year within [-50, 49]
+        range of self._year (current local time)
+        """
+
+        # Function contract is that the year is always positive
+        assert year >= 0
+
+        if year < 100 and not century_specified:
+            # assume current century to start
+            year += self._century
+
+            if year >= self._year + 50:  # if too far in future
+                year -= 100
+            elif year < self._year - 50:  # if too far in past
+                year += 100
+
+        return year
+
+    def validate(self, res):
+        # move to info
+        if res.year is not None:
+            res.year = self.convertyear(res.year, res.century_specified)
+
+        if ((res.tzoffset == 0 and not res.tzname) or
+             (res.tzname == 'Z' or res.tzname == 'z')):
+            res.tzname = "UTC"
+            res.tzoffset = 0
+        elif res.tzoffset != 0 and res.tzname and self.utczone(res.tzname):
+            res.tzoffset = 0
+        return True
+
+
+class _ymd(list):
+    def __init__(self, *args, **kwargs):
+        super(self.__class__, self).__init__(*args, **kwargs)
+        self.century_specified = False
+        self.dstridx = None
+        self.mstridx = None
+        self.ystridx = None
+
+    @property
+    def has_year(self):
+        return self.ystridx is not None
+
+    @property
+    def has_month(self):
+        return self.mstridx is not None
+
+    @property
+    def has_day(self):
+        return self.dstridx is not None
+
+    def could_be_day(self, value):
+        if self.has_day:
+            return False
+        elif not self.has_month:
+            return 1 <= value <= 31
+        elif not self.has_year:
+            # Be permissive, assume leap year
+            month = self[self.mstridx]
+            return 1 <= value <= monthrange(2000, month)[1]
+        else:
+            month = self[self.mstridx]
+            year = self[self.ystridx]
+            return 1 <= value <= monthrange(year, month)[1]
+
+    def append(self, val, label=None):
+        if hasattr(val, '__len__'):
+            if val.isdigit() and len(val) > 2:
+                self.century_specified = True
+                if label not in [None, 'Y']:  # pragma: no cover
+                    raise ValueError(label)
+                label = 'Y'
+        elif val > 100:
+            self.century_specified = True
+            if label not in [None, 'Y']:  # pragma: no cover
+                raise ValueError(label)
+            label = 'Y'
+
+        super(self.__class__, self).append(int(val))
+
+        if label == 'M':
+            if self.has_month:
+                raise ValueError('Month is already set')
+            self.mstridx = len(self) - 1
+        elif label == 'D':
+            if self.has_day:
+                raise ValueError('Day is already set')
+            self.dstridx = len(self) - 1
+        elif label == 'Y':
+            if self.has_year:
+                raise ValueError('Year is already set')
+            self.ystridx = len(self) - 1
+
+    def _resolve_from_stridxs(self, strids):
+        """
+        Try to resolve the identities of year/month/day elements using
+        ystridx, mstridx, and dstridx, if enough of these are specified.
+        """
+        if len(self) == 3 and len(strids) == 2:
+            # we can back out the remaining stridx value
+            missing = [x for x in range(3) if x not in strids.values()]
+            key = [x for x in ['y', 'm', 'd'] if x not in strids]
+            assert len(missing) == len(key) == 1
+            key = key[0]
+            val = missing[0]
+            strids[key] = val
+
+        assert len(self) == len(strids)  # otherwise this should not be called
+        out = {key: self[strids[key]] for key in strids}
+        return (out.get('y'), out.get('m'), out.get('d'))
+
+    def resolve_ymd(self, yearfirst, dayfirst):
+        len_ymd = len(self)
+        year, month, day = (None, None, None)
+
+        strids = (('y', self.ystridx),
+                  ('m', self.mstridx),
+                  ('d', self.dstridx))
+
+        strids = {key: val for key, val in strids if val is not None}
+        if (len(self) == len(strids) > 0 or
+                (len(self) == 3 and len(strids) == 2)):
+            return self._resolve_from_stridxs(strids)
+
+        mstridx = self.mstridx
+
+        if len_ymd > 3:
+            raise ValueError("More than three YMD values")
+        elif len_ymd == 1 or (mstridx is not None and len_ymd == 2):
+            # One member, or two members with a month string
+            if mstridx is not None:
+                month = self[mstridx]
+                # since mstridx is 0 or 1, self[mstridx-1] always
+                # looks up the other element
+                other = self[mstridx - 1]
+            else:
+                other = self[0]
+
+            if len_ymd > 1 or mstridx is None:
+                if other > 31:
+                    year = other
+                else:
+                    day = other
+
+        elif len_ymd == 2:
+            # Two members with numbers
+            if self[0] > 31:
+                # 99-01
+                year, month = self
+            elif self[1] > 31:
+                # 01-99
+                month, year = self
+            elif dayfirst and self[1] <= 12:
+                # 13-01
+                day, month = self
+            else:
+                # 01-13
+                month, day = self
+
+        elif len_ymd == 3:
+            # Three members
+            if mstridx == 0:
+                if self[1] > 31:
+                    # Apr-2003-25
+                    month, year, day = self
+                else:
+                    month, day, year = self
+            elif mstridx == 1:
+                if self[0] > 31 or (yearfirst and self[2] <= 31):
+                    # 99-Jan-01
+                    year, month, day = self
+                else:
+                    # 01-Jan-01
+                    # Give precedence to day-first, since
+                    # two-digit years is usually hand-written.
+                    day, month, year = self
+
+            elif mstridx == 2:
+                # WTF!?
+                if self[1] > 31:
+                    # 01-99-Jan
+                    day, year, month = self
+                else:
+                    # 99-01-Jan
+                    year, day, month = self
+
+            else:
+                if (self[0] > 31 or
+                    self.ystridx == 0 or
+                        (yearfirst and self[1] <= 12 and self[2] <= 31)):
+                    # 99-01-01
+                    if dayfirst and self[2] <= 12:
+                        year, day, month = self
+                    else:
+                        year, month, day = self
+                elif self[0] > 12 or (dayfirst and self[1] <= 12):
+                    # 13-01-01
+                    day, month, year = self
+                else:
+                    # 01-13-01
+                    month, day, year = self
+
+        return year, month, day
+
+
+class parser(object):
+    def __init__(self, info=None):
+        self.info = info or parserinfo()
+
+    def parse(self, timestr, default=None,
+              ignoretz=False, tzinfos=None, **kwargs):
+        """
+        Parse the date/time string into a :class:`datetime.datetime` object.
+
+        :param timestr:
+            Any date/time string using the supported formats.
+
+        :param default:
+            The default datetime object, if this is a datetime object and not
+            ``None``, elements specified in ``timestr`` replace elements in the
+            default object.
+
+        :param ignoretz:
+            If set ``True``, time zones in parsed strings are ignored and a
+            naive :class:`datetime.datetime` object is returned.
+
+        :param tzinfos:
+            Additional time zone names / aliases which may be present in the
+            string. This argument maps time zone names (and optionally offsets
+            from those time zones) to time zones. This parameter can be a
+            dictionary with timezone aliases mapping time zone names to time
+            zones or a function taking two parameters (``tzname`` and
+            ``tzoffset``) and returning a time zone.
+
+            The timezones to which the names are mapped can be an integer
+            offset from UTC in seconds or a :class:`tzinfo` object.
+
+            .. doctest::
+               :options: +NORMALIZE_WHITESPACE
+
+                >>> from dateutil.parser import parse
+                >>> from dateutil.tz import gettz
+                >>> tzinfos = {"BRST": -7200, "CST": gettz("America/Chicago")}
+                >>> parse("2012-01-19 17:21:00 BRST", tzinfos=tzinfos)
+                datetime.datetime(2012, 1, 19, 17, 21, tzinfo=tzoffset(u'BRST', -7200))
+                >>> parse("2012-01-19 17:21:00 CST", tzinfos=tzinfos)
+                datetime.datetime(2012, 1, 19, 17, 21,
+                                  tzinfo=tzfile('/usr/share/zoneinfo/America/Chicago'))
+
+            This parameter is ignored if ``ignoretz`` is set.
+
+        :param \\*\\*kwargs:
+            Keyword arguments as passed to ``_parse()``.
+
+        :return:
+            Returns a :class:`datetime.datetime` object or, if the
+            ``fuzzy_with_tokens`` option is ``True``, returns a tuple, the
+            first element being a :class:`datetime.datetime` object, the second
+            a tuple containing the fuzzy tokens.
+
+        :raises ParserError:
+            Raised for invalid or unknown string format, if the provided
+            :class:`tzinfo` is not in a valid format, or if an invalid date
+            would be created.
+
+        :raises TypeError:
+            Raised for non-string or character stream input.
+
+        :raises OverflowError:
+            Raised if the parsed date exceeds the largest valid C integer on
+            your system.
+        """
+
+        if default is None:
+            default = datetime.datetime.now().replace(hour=0, minute=0,
+                                                      second=0, microsecond=0)
+
+        res, skipped_tokens = self._parse(timestr, **kwargs)
+
+        if res is None:
+            raise ParserError("Unknown string format: %s", timestr)
+
+        if len(res) == 0:
+            raise ParserError("String does not contain a date: %s", timestr)
+
+        try:
+            ret = self._build_naive(res, default)
+        except ValueError as e:
+            six.raise_from(ParserError(str(e) + ": %s", timestr), e)
+
+        if not ignoretz:
+            ret = self._build_tzaware(ret, res, tzinfos)
+
+        if kwargs.get('fuzzy_with_tokens', False):
+            return ret, skipped_tokens
+        else:
+            return ret
+
+    class _result(_resultbase):
+        __slots__ = ["year", "month", "day", "weekday",
+                     "hour", "minute", "second", "microsecond",
+                     "tzname", "tzoffset", "ampm","any_unused_tokens"]
+
+    def _parse(self, timestr, dayfirst=None, yearfirst=None, fuzzy=False,
+               fuzzy_with_tokens=False):
+        """
+        Private method which performs the heavy lifting of parsing, called from
+        ``parse()``, which passes on its ``kwargs`` to this function.
+
+        :param timestr:
+            The string to parse.
+
+        :param dayfirst:
+            Whether to interpret the first value in an ambiguous 3-integer date
+            (e.g. 01/05/09) as the day (``True``) or month (``False``). If
+            ``yearfirst`` is set to ``True``, this distinguishes between YDM
+            and YMD. If set to ``None``, this value is retrieved from the
+            current :class:`parserinfo` object (which itself defaults to
+            ``False``).
+
+        :param yearfirst:
+            Whether to interpret the first value in an ambiguous 3-integer date
+            (e.g. 01/05/09) as the year. If ``True``, the first number is taken
+            to be the year, otherwise the last number is taken to be the year.
+            If this is set to ``None``, the value is retrieved from the current
+            :class:`parserinfo` object (which itself defaults to ``False``).
+
+        :param fuzzy:
+            Whether to allow fuzzy parsing, allowing for string like "Today is
+            January 1, 2047 at 8:21:00AM".
+
+        :param fuzzy_with_tokens:
+            If ``True``, ``fuzzy`` is automatically set to True, and the parser
+            will return a tuple where the first element is the parsed
+            :class:`datetime.datetime` datetimestamp and the second element is
+            a tuple containing the portions of the string which were ignored:
+
+            .. doctest::
+
+                >>> from dateutil.parser import parse
+                >>> parse("Today is January 1, 2047 at 8:21:00AM", fuzzy_with_tokens=True)
+                (datetime.datetime(2047, 1, 1, 8, 21), (u'Today is ', u' ', u'at '))
+
+        """
+        if fuzzy_with_tokens:
+            fuzzy = True
+
+        info = self.info
+
+        if dayfirst is None:
+            dayfirst = info.dayfirst
+
+        if yearfirst is None:
+            yearfirst = info.yearfirst
+
+        res = self._result()
+        l = _timelex.split(timestr)         # Splits the timestr into tokens
+
+        skipped_idxs = []
+
+        # year/month/day list
+        ymd = _ymd()
+
+        len_l = len(l)
+        i = 0
+        try:
+            while i < len_l:
+
+                # Check if it's a number
+                value_repr = l[i]
+                try:
+                    value = float(value_repr)
+                except ValueError:
+                    value = None
+
+                if value is not None:
+                    # Numeric token
+                    i = self._parse_numeric_token(l, i, info, ymd, res, fuzzy)
+
+                # Check weekday
+                elif info.weekday(l[i]) is not None:
+                    value = info.weekday(l[i])
+                    res.weekday = value
+
+                # Check month name
+                elif info.month(l[i]) is not None:
+                    value = info.month(l[i])
+                    ymd.append(value, 'M')
+
+                    if i + 1 < len_l:
+                        if l[i + 1] in ('-', '/'):
+                            # Jan-01[-99]
+                            sep = l[i + 1]
+                            ymd.append(l[i + 2])
+
+                            if i + 3 < len_l and l[i + 3] == sep:
+                                # Jan-01-99
+                                ymd.append(l[i + 4])
+                                i += 2
+
+                            i += 2
+
+                        elif (i + 4 < len_l and l[i + 1] == l[i + 3] == ' ' and
+                              info.pertain(l[i + 2])):
+                            # Jan of 01
+                            # In this case, 01 is clearly year
+                            if l[i + 4].isdigit():
+                                # Convert it here to become unambiguous
+                                value = int(l[i + 4])
+                                year = str(info.convertyear(value))
+                                ymd.append(year, 'Y')
+                            else:
+                                # Wrong guess
+                                pass
+                                # TODO: not hit in tests
+                            i += 4
+
+                # Check am/pm
+                elif info.ampm(l[i]) is not None:
+                    value = info.ampm(l[i])
+                    val_is_ampm = self._ampm_valid(res.hour, res.ampm, fuzzy)
+
+                    if val_is_ampm:
+                        res.hour = self._adjust_ampm(res.hour, value)
+                        res.ampm = value
+
+                    elif fuzzy:
+                        skipped_idxs.append(i)
+
+                # Check for a timezone name
+                elif self._could_be_tzname(res.hour, res.tzname, res.tzoffset, l[i]):
+                    res.tzname = l[i]
+                    res.tzoffset = info.tzoffset(res.tzname)
+
+                    # Check for something like GMT+3, or BRST+3. Notice
+                    # that it doesn't mean "I am 3 hours after GMT", but
+                    # "my time +3 is GMT". If found, we reverse the
+                    # logic so that timezone parsing code will get it
+                    # right.
+                    if i + 1 < len_l and l[i + 1] in ('+', '-'):
+                        l[i + 1] = ('+', '-')[l[i + 1] == '+']
+                        res.tzoffset = None
+                        if info.utczone(res.tzname):
+                            # With something like GMT+3, the timezone
+                            # is *not* GMT.
+                            res.tzname = None
+
+                # Check for a numbered timezone
+                elif res.hour is not None and l[i] in ('+', '-'):
+                    signal = (-1, 1)[l[i] == '+']
+                    len_li = len(l[i + 1])
+
+                    # TODO: check that l[i + 1] is integer?
+                    if len_li == 4:
+                        # -0300
+                        hour_offset = int(l[i + 1][:2])
+                        min_offset = int(l[i + 1][2:])
+                    elif i + 2 < len_l and l[i + 2] == ':':
+                        # -03:00
+                        hour_offset = int(l[i + 1])
+                        min_offset = int(l[i + 3])  # TODO: Check that l[i+3] is minute-like?
+                        i += 2
+                    elif len_li <= 2:
+                        # -[0]3
+                        hour_offset = int(l[i + 1][:2])
+                        min_offset = 0
+                    else:
+                        raise ValueError(timestr)
+
+                    res.tzoffset = signal * (hour_offset * 3600 + min_offset * 60)
+
+                    # Look for a timezone name between parenthesis
+                    if (i + 5 < len_l and
+                            info.jump(l[i + 2]) and l[i + 3] == '(' and
+                            l[i + 5] == ')' and
+                            3 <= len(l[i + 4]) and
+                            self._could_be_tzname(res.hour, res.tzname,
+                                                  None, l[i + 4])):
+                        # -0300 (BRST)
+                        res.tzname = l[i + 4]
+                        i += 4
+
+                    i += 1
+
+                # Check jumps
+                elif not (info.jump(l[i]) or fuzzy):
+                    raise ValueError(timestr)
+
+                else:
+                    skipped_idxs.append(i)
+                i += 1
+
+            # Process year/month/day
+            year, month, day = ymd.resolve_ymd(yearfirst, dayfirst)
+
+            res.century_specified = ymd.century_specified
+            res.year = year
+            res.month = month
+            res.day = day
+
+        except (IndexError, ValueError):
+            return None, None
+
+        if not info.validate(res):
+            return None, None
+
+        if fuzzy_with_tokens:
+            skipped_tokens = self._recombine_skipped(l, skipped_idxs)
+            return res, tuple(skipped_tokens)
+        else:
+            return res, None
+
+    def _parse_numeric_token(self, tokens, idx, info, ymd, res, fuzzy):
+        # Token is a number
+        value_repr = tokens[idx]
+        try:
+            value = self._to_decimal(value_repr)
+        except Exception as e:
+            six.raise_from(ValueError('Unknown numeric token'), e)
+
+        len_li = len(value_repr)
+
+        len_l = len(tokens)
+
+        if (len(ymd) == 3 and len_li in (2, 4) and
+            res.hour is None and
+            (idx + 1 >= len_l or
+             (tokens[idx + 1] != ':' and
+              info.hms(tokens[idx + 1]) is None))):
+            # 19990101T23[59]
+            s = tokens[idx]
+            res.hour = int(s[:2])
+
+            if len_li == 4:
+                res.minute = int(s[2:])
+
+        elif len_li == 6 or (len_li > 6 and tokens[idx].find('.') == 6):
+            # YYMMDD or HHMMSS[.ss]
+            s = tokens[idx]
+
+            if not ymd and '.' not in tokens[idx]:
+                ymd.append(s[:2])
+                ymd.append(s[2:4])
+                ymd.append(s[4:])
+            else:
+                # 19990101T235959[.59]
+
+                # TODO: Check if res attributes already set.
+                res.hour = int(s[:2])
+                res.minute = int(s[2:4])
+                res.second, res.microsecond = self._parsems(s[4:])
+
+        elif len_li in (8, 12, 14):
+            # YYYYMMDD
+            s = tokens[idx]
+            ymd.append(s[:4], 'Y')
+            ymd.append(s[4:6])
+            ymd.append(s[6:8])
+
+            if len_li > 8:
+                res.hour = int(s[8:10])
+                res.minute = int(s[10:12])
+
+                if len_li > 12:
+                    res.second = int(s[12:])
+
+        elif self._find_hms_idx(idx, tokens, info, allow_jump=True) is not None:
+            # HH[ ]h or MM[ ]m or SS[.ss][ ]s
+            hms_idx = self._find_hms_idx(idx, tokens, info, allow_jump=True)
+            (idx, hms) = self._parse_hms(idx, tokens, info, hms_idx)
+            if hms is not None:
+                # TODO: checking that hour/minute/second are not
+                # already set?
+                self._assign_hms(res, value_repr, hms)
+
+        elif idx + 2 < len_l and tokens[idx + 1] == ':':
+            # HH:MM[:SS[.ss]]
+            res.hour = int(value)
+            value = self._to_decimal(tokens[idx + 2])  # TODO: try/except for this?
+            (res.minute, res.second) = self._parse_min_sec(value)
+
+            if idx + 4 < len_l and tokens[idx + 3] == ':':
+                res.second, res.microsecond = self._parsems(tokens[idx + 4])
+
+                idx += 2
+
+            idx += 2
+
+        elif idx + 1 < len_l and tokens[idx + 1] in ('-', '/', '.'):
+            sep = tokens[idx + 1]
+            ymd.append(value_repr)
+
+            if idx + 2 < len_l and not info.jump(tokens[idx + 2]):
+                if tokens[idx + 2].isdigit():
+                    # 01-01[-01]
+                    ymd.append(tokens[idx + 2])
+                else:
+                    # 01-Jan[-01]
+                    value = info.month(tokens[idx + 2])
+
+                    if value is not None:
+                        ymd.append(value, 'M')
+                    else:
+                        raise ValueError()
+
+                if idx + 3 < len_l and tokens[idx + 3] == sep:
+                    # We have three members
+                    value = info.month(tokens[idx + 4])
+
+                    if value is not None:
+                        ymd.append(value, 'M')
+                    else:
+                        ymd.append(tokens[idx + 4])
+                    idx += 2
+
+                idx += 1
+            idx += 1
+
+        elif idx + 1 >= len_l or info.jump(tokens[idx + 1]):
+            if idx + 2 < len_l and info.ampm(tokens[idx + 2]) is not None:
+                # 12 am
+                hour = int(value)
+                res.hour = self._adjust_ampm(hour, info.ampm(tokens[idx + 2]))
+                idx += 1
+            else:
+                # Year, month or day
+                ymd.append(value)
+            idx += 1
+
+        elif info.ampm(tokens[idx + 1]) is not None and (0 <= value < 24):
+            # 12am
+            hour = int(value)
+            res.hour = self._adjust_ampm(hour, info.ampm(tokens[idx + 1]))
+            idx += 1
+
+        elif ymd.could_be_day(value):
+            ymd.append(value)
+
+        elif not fuzzy:
+            raise ValueError()
+
+        return idx
+
+    def _find_hms_idx(self, idx, tokens, info, allow_jump):
+        len_l = len(tokens)
+
+        if idx+1 < len_l and info.hms(tokens[idx+1]) is not None:
+            # There is an "h", "m", or "s" label following this token.  We take
+            # assign the upcoming label to the current token.
+            # e.g. the "12" in 12h"
+            hms_idx = idx + 1
+
+        elif (allow_jump and idx+2 < len_l and tokens[idx+1] == ' ' and
+              info.hms(tokens[idx+2]) is not None):
+            # There is a space and then an "h", "m", or "s" label.
+            # e.g. the "12" in "12 h"
+            hms_idx = idx + 2
+
+        elif idx > 0 and info.hms(tokens[idx-1]) is not None:
+            # There is a "h", "m", or "s" preceding this token.  Since neither
+            # of the previous cases was hit, there is no label following this
+            # token, so we use the previous label.
+            # e.g. the "04" in "12h04"
+            hms_idx = idx-1
+
+        elif (1 < idx == len_l-1 and tokens[idx-1] == ' ' and
+              info.hms(tokens[idx-2]) is not None):
+            # If we are looking at the final token, we allow for a
+            # backward-looking check to skip over a space.
+            # TODO: Are we sure this is the right condition here?
+            hms_idx = idx - 2
+
+        else:
+            hms_idx = None
+
+        return hms_idx
+
+    def _assign_hms(self, res, value_repr, hms):
+        # See GH issue #427, fixing float rounding
+        value = self._to_decimal(value_repr)
+
+        if hms == 0:
+            # Hour
+            res.hour = int(value)
+            if value % 1:
+                res.minute = int(60*(value % 1))
+
+        elif hms == 1:
+            (res.minute, res.second) = self._parse_min_sec(value)
+
+        elif hms == 2:
+            (res.second, res.microsecond) = self._parsems(value_repr)
+
+    def _could_be_tzname(self, hour, tzname, tzoffset, token):
+        return (hour is not None and
+                tzname is None and
+                tzoffset is None and
+                len(token) <= 5 and
+                (all(x in string.ascii_uppercase for x in token)
+                 or token in self.info.UTCZONE))
+
+    def _ampm_valid(self, hour, ampm, fuzzy):
+        """
+        For fuzzy parsing, 'a' or 'am' (both valid English words)
+        may erroneously trigger the AM/PM flag. Deal with that
+        here.
+        """
+        val_is_ampm = True
+
+        # If there's already an AM/PM flag, this one isn't one.
+        if fuzzy and ampm is not None:
+            val_is_ampm = False
+
+        # If AM/PM is found and hour is not, raise a ValueError
+        if hour is None:
+            if fuzzy:
+                val_is_ampm = False
+            else:
+                raise ValueError('No hour specified with AM or PM flag.')
+        elif not 0 <= hour <= 12:
+            # If AM/PM is found, it's a 12 hour clock, so raise
+            # an error for invalid range
+            if fuzzy:
+                val_is_ampm = False
+            else:
+                raise ValueError('Invalid hour specified for 12-hour clock.')
+
+        return val_is_ampm
+
+    def _adjust_ampm(self, hour, ampm):
+        if hour < 12 and ampm == 1:
+            hour += 12
+        elif hour == 12 and ampm == 0:
+            hour = 0
+        return hour
+
+    def _parse_min_sec(self, value):
+        # TODO: Every usage of this function sets res.second to the return
+        # value. Are there any cases where second will be returned as None and
+        # we *don't* want to set res.second = None?
+        minute = int(value)
+        second = None
+
+        sec_remainder = value % 1
+        if sec_remainder:
+            second = int(60 * sec_remainder)
+        return (minute, second)
+
+    def _parse_hms(self, idx, tokens, info, hms_idx):
+        # TODO: Is this going to admit a lot of false-positives for when we
+        # just happen to have digits and "h", "m" or "s" characters in non-date
+        # text?  I guess hex hashes won't have that problem, but there's plenty
+        # of random junk out there.
+        if hms_idx is None:
+            hms = None
+            new_idx = idx
+        elif hms_idx > idx:
+            hms = info.hms(tokens[hms_idx])
+            new_idx = hms_idx
+        else:
+            # Looking backwards, increment one.
+            hms = info.hms(tokens[hms_idx]) + 1
+            new_idx = idx
+
+        return (new_idx, hms)
+
+    # ------------------------------------------------------------------
+    # Handling for individual tokens.  These are kept as methods instead
+    #  of functions for the sake of customizability via subclassing.
+
+    def _parsems(self, value):
+        """Parse a I[.F] seconds value into (seconds, microseconds)."""
+        if "." not in value:
+            return int(value), 0
+        else:
+            i, f = value.split(".")
+            return int(i), int(f.ljust(6, "0")[:6])
+
+    def _to_decimal(self, val):
+        try:
+            decimal_value = Decimal(val)
+            # See GH 662, edge case, infinite value should not be converted
+            #  via `_to_decimal`
+            if not decimal_value.is_finite():
+                raise ValueError("Converted decimal value is infinite or NaN")
+        except Exception as e:
+            msg = "Could not convert %s to decimal" % val
+            six.raise_from(ValueError(msg), e)
+        else:
+            return decimal_value
+
+    # ------------------------------------------------------------------
+    # Post-Parsing construction of datetime output.  These are kept as
+    #  methods instead of functions for the sake of customizability via
+    #  subclassing.
+
+    def _build_tzinfo(self, tzinfos, tzname, tzoffset):
+        if callable(tzinfos):
+            tzdata = tzinfos(tzname, tzoffset)
+        else:
+            tzdata = tzinfos.get(tzname)
+        # handle case where tzinfo is paased an options that returns None
+        # eg tzinfos = {'BRST' : None}
+        if isinstance(tzdata, datetime.tzinfo) or tzdata is None:
+            tzinfo = tzdata
+        elif isinstance(tzdata, text_type):
+            tzinfo = tz.tzstr(tzdata)
+        elif isinstance(tzdata, integer_types):
+            tzinfo = tz.tzoffset(tzname, tzdata)
+        else:
+            raise TypeError("Offset must be tzinfo subclass, tz string, "
+                            "or int offset.")
+        return tzinfo
+
+    def _build_tzaware(self, naive, res, tzinfos):
+        if (callable(tzinfos) or (tzinfos and res.tzname in tzinfos)):
+            tzinfo = self._build_tzinfo(tzinfos, res.tzname, res.tzoffset)
+            aware = naive.replace(tzinfo=tzinfo)
+            aware = self._assign_tzname(aware, res.tzname)
+
+        elif res.tzname and res.tzname in time.tzname:
+            aware = naive.replace(tzinfo=tz.tzlocal())
+
+            # Handle ambiguous local datetime
+            aware = self._assign_tzname(aware, res.tzname)
+
+            # This is mostly relevant for winter GMT zones parsed in the UK
+            if (aware.tzname() != res.tzname and
+                    res.tzname in self.info.UTCZONE):
+                aware = aware.replace(tzinfo=tz.UTC)
+
+        elif res.tzoffset == 0:
+            aware = naive.replace(tzinfo=tz.UTC)
+
+        elif res.tzoffset:
+            aware = naive.replace(tzinfo=tz.tzoffset(res.tzname, res.tzoffset))
+
+        elif not res.tzname and not res.tzoffset:
+            # i.e. no timezone information was found.
+            aware = naive
+
+        elif res.tzname:
+            # tz-like string was parsed but we don't know what to do
+            # with it
+            warnings.warn("tzname {tzname} identified but not understood.  "
+                          "Pass `tzinfos` argument in order to correctly "
+                          "return a timezone-aware datetime.  In a future "
+                          "version, this will raise an "
+                          "exception.".format(tzname=res.tzname),
+                          category=UnknownTimezoneWarning)
+            aware = naive
+
+        return aware
+
+    def _build_naive(self, res, default):
+        repl = {}
+        for attr in ("year", "month", "day", "hour",
+                     "minute", "second", "microsecond"):
+            value = getattr(res, attr)
+            if value is not None:
+                repl[attr] = value
+
+        if 'day' not in repl:
+            # If the default day exceeds the last day of the month, fall back
+            # to the end of the month.
+            cyear = default.year if res.year is None else res.year
+            cmonth = default.month if res.month is None else res.month
+            cday = default.day if res.day is None else res.day
+
+            if cday > monthrange(cyear, cmonth)[1]:
+                repl['day'] = monthrange(cyear, cmonth)[1]
+
+        naive = default.replace(**repl)
+
+        if res.weekday is not None and not res.day:
+            naive = naive + relativedelta.relativedelta(weekday=res.weekday)
+
+        return naive
+
+    def _assign_tzname(self, dt, tzname):
+        if dt.tzname() != tzname:
+            new_dt = tz.enfold(dt, fold=1)
+            if new_dt.tzname() == tzname:
+                return new_dt
+
+        return dt
+
+    def _recombine_skipped(self, tokens, skipped_idxs):
+        """
+        >>> tokens = ["foo", " ", "bar", " ", "19June2000", "baz"]
+        >>> skipped_idxs = [0, 1, 2, 5]
+        >>> _recombine_skipped(tokens, skipped_idxs)
+        ["foo bar", "baz"]
+        """
+        skipped_tokens = []
+        for i, idx in enumerate(sorted(skipped_idxs)):
+            if i > 0 and idx - 1 == skipped_idxs[i - 1]:
+                skipped_tokens[-1] = skipped_tokens[-1] + tokens[idx]
+            else:
+                skipped_tokens.append(tokens[idx])
+
+        return skipped_tokens
+
+
+DEFAULTPARSER = parser()
+
+
+def parse(timestr, parserinfo=None, **kwargs):
+    """
+
+    Parse a string in one of the supported formats, using the
+    ``parserinfo`` parameters.
+
+    :param timestr:
+        A string containing a date/time stamp.
+
+    :param parserinfo:
+        A :class:`parserinfo` object containing parameters for the parser.
+        If ``None``, the default arguments to the :class:`parserinfo`
+        constructor are used.
+
+    The ``**kwargs`` parameter takes the following keyword arguments:
+
+    :param default:
+        The default datetime object, if this is a datetime object and not
+        ``None``, elements specified in ``timestr`` replace elements in the
+        default object.
+
+    :param ignoretz:
+        If set ``True``, time zones in parsed strings are ignored and a naive
+        :class:`datetime` object is returned.
+
+    :param tzinfos:
+        Additional time zone names / aliases which may be present in the
+        string. This argument maps time zone names (and optionally offsets
+        from those time zones) to time zones. This parameter can be a
+        dictionary with timezone aliases mapping time zone names to time
+        zones or a function taking two parameters (``tzname`` and
+        ``tzoffset``) and returning a time zone.
+
+        The timezones to which the names are mapped can be an integer
+        offset from UTC in seconds or a :class:`tzinfo` object.
+
+        .. doctest::
+           :options: +NORMALIZE_WHITESPACE
+
+            >>> from dateutil.parser import parse
+            >>> from dateutil.tz import gettz
+            >>> tzinfos = {"BRST": -7200, "CST": gettz("America/Chicago")}
+            >>> parse("2012-01-19 17:21:00 BRST", tzinfos=tzinfos)
+            datetime.datetime(2012, 1, 19, 17, 21, tzinfo=tzoffset(u'BRST', -7200))
+            >>> parse("2012-01-19 17:21:00 CST", tzinfos=tzinfos)
+            datetime.datetime(2012, 1, 19, 17, 21,
+                              tzinfo=tzfile('/usr/share/zoneinfo/America/Chicago'))
+
+        This parameter is ignored if ``ignoretz`` is set.
+
+    :param dayfirst:
+        Whether to interpret the first value in an ambiguous 3-integer date
+        (e.g. 01/05/09) as the day (``True``) or month (``False``). If
+        ``yearfirst`` is set to ``True``, this distinguishes between YDM and
+        YMD. If set to ``None``, this value is retrieved from the current
+        :class:`parserinfo` object (which itself defaults to ``False``).
+
+    :param yearfirst:
+        Whether to interpret the first value in an ambiguous 3-integer date
+        (e.g. 01/05/09) as the year. If ``True``, the first number is taken to
+        be the year, otherwise the last number is taken to be the year. If
+        this is set to ``None``, the value is retrieved from the current
+        :class:`parserinfo` object (which itself defaults to ``False``).
+
+    :param fuzzy:
+        Whether to allow fuzzy parsing, allowing for string like "Today is
+        January 1, 2047 at 8:21:00AM".
+
+    :param fuzzy_with_tokens:
+        If ``True``, ``fuzzy`` is automatically set to True, and the parser
+        will return a tuple where the first element is the parsed
+        :class:`datetime.datetime` datetimestamp and the second element is
+        a tuple containing the portions of the string which were ignored:
+
+        .. doctest::
+
+            >>> from dateutil.parser import parse
+            >>> parse("Today is January 1, 2047 at 8:21:00AM", fuzzy_with_tokens=True)
+            (datetime.datetime(2047, 1, 1, 8, 21), (u'Today is ', u' ', u'at '))
+
+    :return:
+        Returns a :class:`datetime.datetime` object or, if the
+        ``fuzzy_with_tokens`` option is ``True``, returns a tuple, the
+        first element being a :class:`datetime.datetime` object, the second
+        a tuple containing the fuzzy tokens.
+
+    :raises ParserError:
+        Raised for invalid or unknown string formats, if the provided
+        :class:`tzinfo` is not in a valid format, or if an invalid date would
+        be created.
+
+    :raises OverflowError:
+        Raised if the parsed date exceeds the largest valid C integer on
+        your system.
+    """
+    if parserinfo:
+        return parser(parserinfo).parse(timestr, **kwargs)
+    else:
+        return DEFAULTPARSER.parse(timestr, **kwargs)
+
+
+class _tzparser(object):
+
+    class _result(_resultbase):
+
+        __slots__ = ["stdabbr", "stdoffset", "dstabbr", "dstoffset",
+                     "start", "end"]
+
+        class _attr(_resultbase):
+            __slots__ = ["month", "week", "weekday",
+                         "yday", "jyday", "day", "time"]
+
+        def __repr__(self):
+            return self._repr("")
+
+        def __init__(self):
+            _resultbase.__init__(self)
+            self.start = self._attr()
+            self.end = self._attr()
+
+    def parse(self, tzstr):
+        res = self._result()
+        l = [x for x in re.split(r'([,:.]|[a-zA-Z]+|[0-9]+)',tzstr) if x]
+        used_idxs = list()
+        try:
+
+            len_l = len(l)
+
+            i = 0
+            while i < len_l:
+                # BRST+3[BRDT[+2]]
+                j = i
+                while j < len_l and not [x for x in l[j]
+                                         if x in "0123456789:,-+"]:
+                    j += 1
+                if j != i:
+                    if not res.stdabbr:
+                        offattr = "stdoffset"
+                        res.stdabbr = "".join(l[i:j])
+                    else:
+                        offattr = "dstoffset"
+                        res.dstabbr = "".join(l[i:j])
+
+                    for ii in range(j):
+                        used_idxs.append(ii)
+                    i = j
+                    if (i < len_l and (l[i] in ('+', '-') or l[i][0] in
+                                       "0123456789")):
+                        if l[i] in ('+', '-'):
+                            # Yes, that's right.  See the TZ variable
+                            # documentation.
+                            signal = (1, -1)[l[i] == '+']
+                            used_idxs.append(i)
+                            i += 1
+                        else:
+                            signal = -1
+                        len_li = len(l[i])
+                        if len_li == 4:
+                            # -0300
+                            setattr(res, offattr, (int(l[i][:2]) * 3600 +
+                                                   int(l[i][2:]) * 60) * signal)
+                        elif i + 1 < len_l and l[i + 1] == ':':
+                            # -03:00
+                            setattr(res, offattr,
+                                    (int(l[i]) * 3600 +
+                                     int(l[i + 2]) * 60) * signal)
+                            used_idxs.append(i)
+                            i += 2
+                        elif len_li <= 2:
+                            # -[0]3
+                            setattr(res, offattr,
+                                    int(l[i][:2]) * 3600 * signal)
+                        else:
+                            return None
+                        used_idxs.append(i)
+                        i += 1
+                    if res.dstabbr:
+                        break
+                else:
+                    break
+
+
+            if i < len_l:
+                for j in range(i, len_l):
+                    if l[j] == ';':
+                        l[j] = ','
+
+                assert l[i] == ','
+
+                i += 1
+
+            if i >= len_l:
+                pass
+            elif (8 <= l.count(',') <= 9 and
+                  not [y for x in l[i:] if x != ','
+                       for y in x if y not in "0123456789+-"]):
+                # GMT0BST,3,0,30,3600,10,0,26,7200[,3600]
+                for x in (res.start, res.end):
+                    x.month = int(l[i])
+                    used_idxs.append(i)
+                    i += 2
+                    if l[i] == '-':
+                        value = int(l[i + 1]) * -1
+                        used_idxs.append(i)
+                        i += 1
+                    else:
+                        value = int(l[i])
+                    used_idxs.append(i)
+                    i += 2
+                    if value:
+                        x.week = value
+                        x.weekday = (int(l[i]) - 1) % 7
+                    else:
+                        x.day = int(l[i])
+                    used_idxs.append(i)
+                    i += 2
+                    x.time = int(l[i])
+                    used_idxs.append(i)
+                    i += 2
+                if i < len_l:
+                    if l[i] in ('-', '+'):
+                        signal = (-1, 1)[l[i] == "+"]
+                        used_idxs.append(i)
+                        i += 1
+                    else:
+                        signal = 1
+                    used_idxs.append(i)
+                    res.dstoffset = (res.stdoffset + int(l[i]) * signal)
+
+                # This was a made-up format that is not in normal use
+                warn(('Parsed time zone "%s"' % tzstr) +
+                     'is in a non-standard dateutil-specific format, which ' +
+                     'is now deprecated; support for parsing this format ' +
+                     'will be removed in future versions. It is recommended ' +
+                     'that you switch to a standard format like the GNU ' +
+                     'TZ variable format.', tz.DeprecatedTzFormatWarning)
+            elif (l.count(',') == 2 and l[i:].count('/') <= 2 and
+                  not [y for x in l[i:] if x not in (',', '/', 'J', 'M',
+                                                     '.', '-', ':')
+                       for y in x if y not in "0123456789"]):
+                for x in (res.start, res.end):
+                    if l[i] == 'J':
+                        # non-leap year day (1 based)
+                        used_idxs.append(i)
+                        i += 1
+                        x.jyday = int(l[i])
+                    elif l[i] == 'M':
+                        # month[-.]week[-.]weekday
+                        used_idxs.append(i)
+                        i += 1
+                        x.month = int(l[i])
+                        used_idxs.append(i)
+                        i += 1
+                        assert l[i] in ('-', '.')
+                        used_idxs.append(i)
+                        i += 1
+                        x.week = int(l[i])
+                        if x.week == 5:
+                            x.week = -1
+                        used_idxs.append(i)
+                        i += 1
+                        assert l[i] in ('-', '.')
+                        used_idxs.append(i)
+                        i += 1
+                        x.weekday = (int(l[i]) - 1) % 7
+                    else:
+                        # year day (zero based)
+                        x.yday = int(l[i]) + 1
+
+                    used_idxs.append(i)
+                    i += 1
+
+                    if i < len_l and l[i] == '/':
+                        used_idxs.append(i)
+                        i += 1
+                        # start time
+                        len_li = len(l[i])
+                        if len_li == 4:
+                            # -0300
+                            x.time = (int(l[i][:2]) * 3600 +
+                                      int(l[i][2:]) * 60)
+                        elif i + 1 < len_l and l[i + 1] == ':':
+                            # -03:00
+                            x.time = int(l[i]) * 3600 + int(l[i + 2]) * 60
+                            used_idxs.append(i)
+                            i += 2
+                            if i + 1 < len_l and l[i + 1] == ':':
+                                used_idxs.append(i)
+                                i += 2
+                                x.time += int(l[i])
+                        elif len_li <= 2:
+                            # -[0]3
+                            x.time = (int(l[i][:2]) * 3600)
+                        else:
+                            return None
+                        used_idxs.append(i)
+                        i += 1
+
+                    assert i == len_l or l[i] == ','
+
+                    i += 1
+
+                assert i >= len_l
+
+        except (IndexError, ValueError, AssertionError):
+            return None
+
+        unused_idxs = set(range(len_l)).difference(used_idxs)
+        res.any_unused_tokens = not {l[n] for n in unused_idxs}.issubset({",",":"})
+        return res
+
+
+DEFAULTTZPARSER = _tzparser()
+
+
+def _parsetz(tzstr):
+    return DEFAULTTZPARSER.parse(tzstr)
+
+
+class ParserError(ValueError):
+    """Exception subclass used for any failure to parse a datetime string.
+
+    This is a subclass of :py:exc:`ValueError`, and should be raised any time
+    earlier versions of ``dateutil`` would have raised ``ValueError``.
+
+    .. versionadded:: 2.8.1
+    """
+    def __str__(self):
+        try:
+            return self.args[0] % self.args[1:]
+        except (TypeError, IndexError):
+            return super(ParserError, self).__str__()
+
+    def __repr__(self):
+        args = ", ".join("'%s'" % arg for arg in self.args)
+        return "%s(%s)" % (self.__class__.__name__, args)
+
+
+class UnknownTimezoneWarning(RuntimeWarning):
+    """Raised when the parser finds a timezone it cannot parse into a tzinfo.
+
+    .. versionadded:: 2.7.0
+    """
+# vim:ts=4:sw=4:et

dateutil/parser/isoparser.py

@@ -0,0 +1,416 @@
+# -*- coding: utf-8 -*-
+"""
+This module offers a parser for ISO-8601 strings
+
+It is intended to support all valid date, time and datetime formats per the
+ISO-8601 specification.
+
+..versionadded:: 2.7.0
+"""
+from datetime import datetime, timedelta, time, date
+import calendar
+from dateutil import tz
+
+from functools import wraps
+
+import re
+import six
+
+__all__ = ["isoparse", "isoparser"]
+
+
+def _takes_ascii(f):
+    @wraps(f)
+    def func(self, str_in, *args, **kwargs):
+        # If it's a stream, read the whole thing
+        str_in = getattr(str_in, 'read', lambda: str_in)()
+
+        # If it's unicode, turn it into bytes, since ISO-8601 only covers ASCII
+        if isinstance(str_in, six.text_type):
+            # ASCII is the same in UTF-8
+            try:
+                str_in = str_in.encode('ascii')
+            except UnicodeEncodeError as e:
+                msg = 'ISO-8601 strings should contain only ASCII characters'
+                six.raise_from(ValueError(msg), e)
+
+        return f(self, str_in, *args, **kwargs)
+
+    return func
+
+
+class isoparser(object):
+    def __init__(self, sep=None):
+        """
+        :param sep:
+            A single character that separates date and time portions. If
+            ``None``, the parser will accept any single character.
+            For strict ISO-8601 adherence, pass ``'T'``.
+        """
+        if sep is not None:
+            if (len(sep) != 1 or ord(sep) >= 128 or sep in '0123456789'):
+                raise ValueError('Separator must be a single, non-numeric ' +
+                                 'ASCII character')
+
+            sep = sep.encode('ascii')
+
+        self._sep = sep
+
+    @_takes_ascii
+    def isoparse(self, dt_str):
+        """
+        Parse an ISO-8601 datetime string into a :class:`datetime.datetime`.
+
+        An ISO-8601 datetime string consists of a date portion, followed
+        optionally by a time portion - the date and time portions are separated
+        by a single character separator, which is ``T`` in the official
+        standard. Incomplete date formats (such as ``YYYY-MM``) may *not* be
+        combined with a time portion.
+
+        Supported date formats are:
+
+        Common:
+
+        - ``YYYY``
+        - ``YYYY-MM``
+        - ``YYYY-MM-DD`` or ``YYYYMMDD``
+
+        Uncommon:
+
+        - ``YYYY-Www`` or ``YYYYWww`` - ISO week (day defaults to 0)
+        - ``YYYY-Www-D`` or ``YYYYWwwD`` - ISO week and day
+
+        The ISO week and day numbering follows the same logic as
+        :func:`datetime.date.isocalendar`.
+
+        Supported time formats are:
+
+        - ``hh``
+        - ``hh:mm`` or ``hhmm``
+        - ``hh:mm:ss`` or ``hhmmss``
+        - ``hh:mm:ss.ssssss`` (Up to 6 sub-second digits)
+
+        Midnight is a special case for `hh`, as the standard supports both
+        00:00 and 24:00 as a representation. The decimal separator can be
+        either a dot or a comma.
+
+
+        .. caution::
+
+            Support for fractional components other than seconds is part of the
+            ISO-8601 standard, but is not currently implemented in this parser.
+
+        Supported time zone offset formats are:
+
+        - `Z` (UTC)
+        - `±HH:MM`
+        - `±HHMM`
+        - `±HH`
+
+        Offsets will be represented as :class:`dateutil.tz.tzoffset` objects,
+        with the exception of UTC, which will be represented as
+        :class:`dateutil.tz.tzutc`. Time zone offsets equivalent to UTC (such
+        as `+00:00`) will also be represented as :class:`dateutil.tz.tzutc`.
+
+        :param dt_str:
+            A string or stream containing only an ISO-8601 datetime string
+
+        :return:
+            Returns a :class:`datetime.datetime` representing the string.
+            Unspecified components default to their lowest value.
+
+        .. warning::
+
+            As of version 2.7.0, the strictness of the parser should not be
+            considered a stable part of the contract. Any valid ISO-8601 string
+            that parses correctly with the default settings will continue to
+            parse correctly in future versions, but invalid strings that
+            currently fail (e.g. ``2017-01-01T00:00+00:00:00``) are not
+            guaranteed to continue failing in future versions if they encode
+            a valid date.
+
+        .. versionadded:: 2.7.0
+        """
+        components, pos = self._parse_isodate(dt_str)
+
+        if len(dt_str) > pos:
+            if self._sep is None or dt_str[pos:pos + 1] == self._sep:
+                components += self._parse_isotime(dt_str[pos + 1:])
+            else:
+                raise ValueError('String contains unknown ISO components')
+
+        if len(components) > 3 and components[3] == 24:
+            components[3] = 0
+            return datetime(*components) + timedelta(days=1)
+
+        return datetime(*components)
+
+    @_takes_ascii
+    def parse_isodate(self, datestr):
+        """
+        Parse the date portion of an ISO string.
+
+        :param datestr:
+            The string portion of an ISO string, without a separator
+
+        :return:
+            Returns a :class:`datetime.date` object
+        """
+        components, pos = self._parse_isodate(datestr)
+        if pos < len(datestr):
+            raise ValueError('String contains unknown ISO ' +
+                             'components: {!r}'.format(datestr.decode('ascii')))
+        return date(*components)
+
+    @_takes_ascii
+    def parse_isotime(self, timestr):
+        """
+        Parse the time portion of an ISO string.
+
+        :param timestr:
+            The time portion of an ISO string, without a separator
+
+        :return:
+            Returns a :class:`datetime.time` object
+        """
+        components = self._parse_isotime(timestr)
+        if components[0] == 24:
+            components[0] = 0
+        return time(*components)
+
+    @_takes_ascii
+    def parse_tzstr(self, tzstr, zero_as_utc=True):
+        """
+        Parse a valid ISO time zone string.
+
+        See :func:`isoparser.isoparse` for details on supported formats.
+
+        :param tzstr:
+            A string representing an ISO time zone offset
+
+        :param zero_as_utc:
+            Whether to return :class:`dateutil.tz.tzutc` for zero-offset zones
+
+        :return:
+            Returns :class:`dateutil.tz.tzoffset` for offsets and
+            :class:`dateutil.tz.tzutc` for ``Z`` and (if ``zero_as_utc`` is
+            specified) offsets equivalent to UTC.
+        """
+        return self._parse_tzstr(tzstr, zero_as_utc=zero_as_utc)
+
+    # Constants
+    _DATE_SEP = b'-'
+    _TIME_SEP = b':'
+    _FRACTION_REGEX = re.compile(b'[\\.,]([0-9]+)')
+
+    def _parse_isodate(self, dt_str):
+        try:
+            return self._parse_isodate_common(dt_str)
+        except ValueError:
+            return self._parse_isodate_uncommon(dt_str)
+
+    def _parse_isodate_common(self, dt_str):
+        len_str = len(dt_str)
+        components = [1, 1, 1]
+
+        if len_str < 4:
+            raise ValueError('ISO string too short')
+
+        # Year
+        components[0] = int(dt_str[0:4])
+        pos = 4
+        if pos >= len_str:
+            return components, pos
+
+        has_sep = dt_str[pos:pos + 1] == self._DATE_SEP
+        if has_sep:
+            pos += 1
+
+        # Month
+        if len_str - pos < 2:
+            raise ValueError('Invalid common month')
+
+        components[1] = int(dt_str[pos:pos + 2])
+        pos += 2
+
+        if pos >= len_str:
+            if has_sep:
+                return components, pos
+            else:
+                raise ValueError('Invalid ISO format')
+
+        if has_sep:
+            if dt_str[pos:pos + 1] != self._DATE_SEP:
+                raise ValueError('Invalid separator in ISO string')
+            pos += 1
+
+        # Day
+        if len_str - pos < 2:
+            raise ValueError('Invalid common day')
+        components[2] = int(dt_str[pos:pos + 2])
+        return components, pos + 2
+
+    def _parse_isodate_uncommon(self, dt_str):
+        if len(dt_str) < 4:
+            raise ValueError('ISO string too short')
+
+        # All ISO formats start with the year
+        year = int(dt_str[0:4])
+
+        has_sep = dt_str[4:5] == self._DATE_SEP
+
+        pos = 4 + has_sep       # Skip '-' if it's there
+        if dt_str[pos:pos + 1] == b'W':
+            # YYYY-?Www-?D?
+            pos += 1
+            weekno = int(dt_str[pos:pos + 2])
+            pos += 2
+
+            dayno = 1
+            if len(dt_str) > pos:
+                if (dt_str[pos:pos + 1] == self._DATE_SEP) != has_sep:
+                    raise ValueError('Inconsistent use of dash separator')
+
+                pos += has_sep
+
+                dayno = int(dt_str[pos:pos + 1])
+                pos += 1
+
+            base_date = self._calculate_weekdate(year, weekno, dayno)
+        else:
+            # YYYYDDD or YYYY-DDD
+            if len(dt_str) - pos < 3:
+                raise ValueError('Invalid ordinal day')
+
+            ordinal_day = int(dt_str[pos:pos + 3])
+            pos += 3
+
+            if ordinal_day < 1 or ordinal_day > (365 + calendar.isleap(year)):
+                raise ValueError('Invalid ordinal day' +
+                                 ' {} for year {}'.format(ordinal_day, year))
+
+            base_date = date(year, 1, 1) + timedelta(days=ordinal_day - 1)
+
+        components = [base_date.year, base_date.month, base_date.day]
+        return components, pos
+
+    def _calculate_weekdate(self, year, week, day):
+        """
+        Calculate the day of corresponding to the ISO year-week-day calendar.
+
+        This function is effectively the inverse of
+        :func:`datetime.date.isocalendar`.
+
+        :param year:
+            The year in the ISO calendar
+
+        :param week:
+            The week in the ISO calendar - range is [1, 53]
+
+        :param day:
+            The day in the ISO calendar - range is [1 (MON), 7 (SUN)]
+
+        :return:
+            Returns a :class:`datetime.date`
+        """
+        if not 0 < week < 54:
+            raise ValueError('Invalid week: {}'.format(week))
+
+        if not 0 < day < 8:     # Range is 1-7
+            raise ValueError('Invalid weekday: {}'.format(day))
+
+        # Get week 1 for the specific year:
+        jan_4 = date(year, 1, 4)   # Week 1 always has January 4th in it
+        week_1 = jan_4 - timedelta(days=jan_4.isocalendar()[2] - 1)
+
+        # Now add the specific number of weeks and days to get what we want
+        week_offset = (week - 1) * 7 + (day - 1)
+        return week_1 + timedelta(days=week_offset)
+
+    def _parse_isotime(self, timestr):
+        len_str = len(timestr)
+        components = [0, 0, 0, 0, None]
+        pos = 0
+        comp = -1
+
+        if len_str < 2:
+            raise ValueError('ISO time too short')
+
+        has_sep = False
+
+        while pos < len_str and comp < 5:
+            comp += 1
+
+            if timestr[pos:pos + 1] in b'-+Zz':
+                # Detect time zone boundary
+                components[-1] = self._parse_tzstr(timestr[pos:])
+                pos = len_str
+                break
+
+            if comp == 1 and timestr[pos:pos+1] == self._TIME_SEP:
+                has_sep = True
+                pos += 1
+            elif comp == 2 and has_sep:
+                if timestr[pos:pos+1] != self._TIME_SEP:
+                    raise ValueError('Inconsistent use of colon separator')
+                pos += 1
+
+            if comp < 3:
+                # Hour, minute, second
+                components[comp] = int(timestr[pos:pos + 2])
+                pos += 2
+
+            if comp == 3:
+                # Fraction of a second
+                frac = self._FRACTION_REGEX.match(timestr[pos:])
+                if not frac:
+                    continue
+
+                us_str = frac.group(1)[:6]  # Truncate to microseconds
+                components[comp] = int(us_str) * 10**(6 - len(us_str))
+                pos += len(frac.group())
+
+        if pos < len_str:
+            raise ValueError('Unused components in ISO string')
+
+        if components[0] == 24:
+            # Standard supports 00:00 and 24:00 as representations of midnight
+            if any(component != 0 for component in components[1:4]):
+                raise ValueError('Hour may only be 24 at 24:00:00.000')
+
+        return components
+
+    def _parse_tzstr(self, tzstr, zero_as_utc=True):
+        if tzstr == b'Z' or tzstr == b'z':
+            return tz.UTC
+
+        if len(tzstr) not in {3, 5, 6}:
+            raise ValueError('Time zone offset must be 1, 3, 5 or 6 characters')
+
+        if tzstr[0:1] == b'-':
+            mult = -1
+        elif tzstr[0:1] == b'+':
+            mult = 1
+        else:
+            raise ValueError('Time zone offset requires sign')
+
+        hours = int(tzstr[1:3])
+        if len(tzstr) == 3:
+            minutes = 0
+        else:
+            minutes = int(tzstr[(4 if tzstr[3:4] == self._TIME_SEP else 3):])
+
+        if zero_as_utc and hours == 0 and minutes == 0:
+            return tz.UTC
+        else:
+            if minutes > 59:
+                raise ValueError('Invalid minutes in time zone offset')
+
+            if hours > 23:
+                raise ValueError('Invalid hours in time zone offset')
+
+            return tz.tzoffset(None, mult * (hours * 60 + minutes) * 60)
+
+
+DEFAULT_ISOPARSER = isoparser()
+isoparse = DEFAULT_ISOPARSER.isoparse

dateutil/tz/__init__.py

@@ -0,0 +1,12 @@
+# -*- coding: utf-8 -*-
+from .tz import *
+from .tz import __doc__
+
+__all__ = ["tzutc", "tzoffset", "tzlocal", "tzfile", "tzrange",
+           "tzstr", "tzical", "tzwin", "tzwinlocal", "gettz",
+           "enfold", "datetime_ambiguous", "datetime_exists",
+           "resolve_imaginary", "UTC", "DeprecatedTzFormatWarning"]
+
+
+class DeprecatedTzFormatWarning(Warning):
+    """Warning raised when time zones are parsed from deprecated formats."""

dateutil/tz/_common.py

@@ -0,0 +1,419 @@
+from six import PY2
+
+from functools import wraps
+
+from datetime import datetime, timedelta, tzinfo
+
+
+ZERO = timedelta(0)
+
+__all__ = ['tzname_in_python2', 'enfold']
+
+
+def tzname_in_python2(namefunc):
+    """Change unicode output into bytestrings in Python 2
+
+    tzname() API changed in Python 3. It used to return bytes, but was changed
+    to unicode strings
+    """
+    if PY2:
+        @wraps(namefunc)
+        def adjust_encoding(*args, **kwargs):
+            name = namefunc(*args, **kwargs)
+            if name is not None:
+                name = name.encode()
+
+            return name
+
+        return adjust_encoding
+    else:
+        return namefunc
+
+
+# The following is adapted from Alexander Belopolsky's tz library
+# https://github.com/abalkin/tz
+if hasattr(datetime, 'fold'):
+    # This is the pre-python 3.6 fold situation
+    def enfold(dt, fold=1):
+        """
+        Provides a unified interface for assigning the ``fold`` attribute to
+        datetimes both before and after the implementation of PEP-495.
+
+        :param fold:
+            The value for the ``fold`` attribute in the returned datetime. This
+            should be either 0 or 1.
+
+        :return:
+            Returns an object for which ``getattr(dt, 'fold', 0)`` returns
+            ``fold`` for all versions of Python. In versions prior to
+            Python 3.6, this is a ``_DatetimeWithFold`` object, which is a
+            subclass of :py:class:`datetime.datetime` with the ``fold``
+            attribute added, if ``fold`` is 1.
+
+        .. versionadded:: 2.6.0
+        """
+        return dt.replace(fold=fold)
+
+else:
+    class _DatetimeWithFold(datetime):
+        """
+        This is a class designed to provide a PEP 495-compliant interface for
+        Python versions before 3.6. It is used only for dates in a fold, so
+        the ``fold`` attribute is fixed at ``1``.
+
+        .. versionadded:: 2.6.0
+        """
+        __slots__ = ()
+
+        def replace(self, *args, **kwargs):
+            """
+            Return a datetime with the same attributes, except for those
+            attributes given new values by whichever keyword arguments are
+            specified. Note that tzinfo=None can be specified to create a naive
+            datetime from an aware datetime with no conversion of date and time
+            data.
+
+            This is reimplemented in ``_DatetimeWithFold`` because pypy3 will
+            return a ``datetime.datetime`` even if ``fold`` is unchanged.
+            """
+            argnames = (
+                'year', 'month', 'day', 'hour', 'minute', 'second',
+                'microsecond', 'tzinfo'
+            )
+
+            for arg, argname in zip(args, argnames):
+                if argname in kwargs:
+                    raise TypeError('Duplicate argument: {}'.format(argname))
+
+                kwargs[argname] = arg
+
+            for argname in argnames:
+                if argname not in kwargs:
+                    kwargs[argname] = getattr(self, argname)
+
+            dt_class = self.__class__ if kwargs.get('fold', 1) else datetime
+
+            return dt_class(**kwargs)
+
+        @property
+        def fold(self):
+            return 1
+
+    def enfold(dt, fold=1):
+        """
+        Provides a unified interface for assigning the ``fold`` attribute to
+        datetimes both before and after the implementation of PEP-495.
+
+        :param fold:
+            The value for the ``fold`` attribute in the returned datetime. This
+            should be either 0 or 1.
+
+        :return:
+            Returns an object for which ``getattr(dt, 'fold', 0)`` returns
+            ``fold`` for all versions of Python. In versions prior to
+            Python 3.6, this is a ``_DatetimeWithFold`` object, which is a
+            subclass of :py:class:`datetime.datetime` with the ``fold``
+            attribute added, if ``fold`` is 1.
+
+        .. versionadded:: 2.6.0
+        """
+        if getattr(dt, 'fold', 0) == fold:
+            return dt
+
+        args = dt.timetuple()[:6]
+        args += (dt.microsecond, dt.tzinfo)
+
+        if fold:
+            return _DatetimeWithFold(*args)
+        else:
+            return datetime(*args)
+
+
+def _validate_fromutc_inputs(f):
+    """
+    The CPython version of ``fromutc`` checks that the input is a ``datetime``
+    object and that ``self`` is attached as its ``tzinfo``.
+    """
+    @wraps(f)
+    def fromutc(self, dt):
+        if not isinstance(dt, datetime):
+            raise TypeError("fromutc() requires a datetime argument")
+        if dt.tzinfo is not self:
+            raise ValueError("dt.tzinfo is not self")
+
+        return f(self, dt)
+
+    return fromutc
+
+
+class _tzinfo(tzinfo):
+    """
+    Base class for all ``dateutil`` ``tzinfo`` objects.
+    """
+
+    def is_ambiguous(self, dt):
+        """
+        Whether or not the "wall time" of a given datetime is ambiguous in this
+        zone.
+
+        :param dt:
+            A :py:class:`datetime.datetime`, naive or time zone aware.
+
+
+        :return:
+            Returns ``True`` if ambiguous, ``False`` otherwise.
+
+        .. versionadded:: 2.6.0
+        """
+
+        dt = dt.replace(tzinfo=self)
+
+        wall_0 = enfold(dt, fold=0)
+        wall_1 = enfold(dt, fold=1)
+
+        same_offset = wall_0.utcoffset() == wall_1.utcoffset()
+        same_dt = wall_0.replace(tzinfo=None) == wall_1.replace(tzinfo=None)
+
+        return same_dt and not same_offset
+
+    def _fold_status(self, dt_utc, dt_wall):
+        """
+        Determine the fold status of a "wall" datetime, given a representation
+        of the same datetime as a (naive) UTC datetime. This is calculated based
+        on the assumption that ``dt.utcoffset() - dt.dst()`` is constant for all
+        datetimes, and that this offset is the actual number of hours separating
+        ``dt_utc`` and ``dt_wall``.
+
+        :param dt_utc:
+            Representation of the datetime as UTC
+
+        :param dt_wall:
+            Representation of the datetime as "wall time". This parameter must
+            either have a `fold` attribute or have a fold-naive
+            :class:`datetime.tzinfo` attached, otherwise the calculation may
+            fail.
+        """
+        if self.is_ambiguous(dt_wall):
+            delta_wall = dt_wall - dt_utc
+            _fold = int(delta_wall == (dt_utc.utcoffset() - dt_utc.dst()))
+        else:
+            _fold = 0
+
+        return _fold
+
+    def _fold(self, dt):
+        return getattr(dt, 'fold', 0)
+
+    def _fromutc(self, dt):
+        """
+        Given a timezone-aware datetime in a given timezone, calculates a
+        timezone-aware datetime in a new timezone.
+
+        Since this is the one time that we *know* we have an unambiguous
+        datetime object, we take this opportunity to determine whether the
+        datetime is ambiguous and in a "fold" state (e.g. if it's the first
+        occurrence, chronologically, of the ambiguous datetime).
+
+        :param dt:
+            A timezone-aware :class:`datetime.datetime` object.
+        """
+
+        # Re-implement the algorithm from Python's datetime.py
+        dtoff = dt.utcoffset()
+        if dtoff is None:
+            raise ValueError("fromutc() requires a non-None utcoffset() "
+                             "result")
+
+        # The original datetime.py code assumes that `dst()` defaults to
+        # zero during ambiguous times. PEP 495 inverts this presumption, so
+        # for pre-PEP 495 versions of python, we need to tweak the algorithm.
+        dtdst = dt.dst()
+        if dtdst is None:
+            raise ValueError("fromutc() requires a non-None dst() result")
+        delta = dtoff - dtdst
+
+        dt += delta
+        # Set fold=1 so we can default to being in the fold for
+        # ambiguous dates.
+        dtdst = enfold(dt, fold=1).dst()
+        if dtdst is None:
+            raise ValueError("fromutc(): dt.dst gave inconsistent "
+                             "results; cannot convert")
+        return dt + dtdst
+
+    @_validate_fromutc_inputs
+    def fromutc(self, dt):
+        """
+        Given a timezone-aware datetime in a given timezone, calculates a
+        timezone-aware datetime in a new timezone.
+
+        Since this is the one time that we *know* we have an unambiguous
+        datetime object, we take this opportunity to determine whether the
+        datetime is ambiguous and in a "fold" state (e.g. if it's the first
+        occurrence, chronologically, of the ambiguous datetime).
+
+        :param dt:
+            A timezone-aware :class:`datetime.datetime` object.
+        """
+        dt_wall = self._fromutc(dt)
+
+        # Calculate the fold status given the two datetimes.
+        _fold = self._fold_status(dt, dt_wall)
+
+        # Set the default fold value for ambiguous dates
+        return enfold(dt_wall, fold=_fold)
+
+
+class tzrangebase(_tzinfo):
+    """
+    This is an abstract base class for time zones represented by an annual
+    transition into and out of DST. Child classes should implement the following
+    methods:
+
+        * ``__init__(self, *args, **kwargs)``
+        * ``transitions(self, year)`` - this is expected to return a tuple of
+          datetimes representing the DST on and off transitions in standard
+          time.
+
+    A fully initialized ``tzrangebase`` subclass should also provide the
+    following attributes:
+        * ``hasdst``: Boolean whether or not the zone uses DST.
+        * ``_dst_offset`` / ``_std_offset``: :class:`datetime.timedelta` objects
+          representing the respective UTC offsets.
+        * ``_dst_abbr`` / ``_std_abbr``: Strings representing the timezone short
+          abbreviations in DST and STD, respectively.
+        * ``_hasdst``: Whether or not the zone has DST.
+
+    .. versionadded:: 2.6.0
+    """
+    def __init__(self):
+        raise NotImplementedError('tzrangebase is an abstract base class')
+
+    def utcoffset(self, dt):
+        isdst = self._isdst(dt)
+
+        if isdst is None:
+            return None
+        elif isdst:
+            return self._dst_offset
+        else:
+            return self._std_offset
+
+    def dst(self, dt):
+        isdst = self._isdst(dt)
+
+        if isdst is None:
+            return None
+        elif isdst:
+            return self._dst_base_offset
+        else:
+            return ZERO
+
+    @tzname_in_python2
+    def tzname(self, dt):
+        if self._isdst(dt):
+            return self._dst_abbr
+        else:
+            return self._std_abbr
+
+    def fromutc(self, dt):
+        """ Given a datetime in UTC, return local time """
+        if not isinstance(dt, datetime):
+            raise TypeError("fromutc() requires a datetime argument")
+
+        if dt.tzinfo is not self:
+            raise ValueError("dt.tzinfo is not self")
+
+        # Get transitions - if there are none, fixed offset
+        transitions = self.transitions(dt.year)
+        if transitions is None:
+            return dt + self.utcoffset(dt)
+
+        # Get the transition times in UTC
+        dston, dstoff = transitions
+
+        dston -= self._std_offset
+        dstoff -= self._std_offset
+
+        utc_transitions = (dston, dstoff)
+        dt_utc = dt.replace(tzinfo=None)
+
+        isdst = self._naive_isdst(dt_utc, utc_transitions)
+
+        if isdst:
+            dt_wall = dt + self._dst_offset
+        else:
+            dt_wall = dt + self._std_offset
+
+        _fold = int(not isdst and self.is_ambiguous(dt_wall))
+
+        return enfold(dt_wall, fold=_fold)
+
+    def is_ambiguous(self, dt):
+        """
+        Whether or not the "wall time" of a given datetime is ambiguous in this
+        zone.
+
+        :param dt:
+            A :py:class:`datetime.datetime`, naive or time zone aware.
+
+
+        :return:
+            Returns ``True`` if ambiguous, ``False`` otherwise.
+
+        .. versionadded:: 2.6.0
+        """
+        if not self.hasdst:
+            return False
+
+        start, end = self.transitions(dt.year)
+
+        dt = dt.replace(tzinfo=None)
+        return (end <= dt < end + self._dst_base_offset)
+
+    def _isdst(self, dt):
+        if not self.hasdst:
+            return False
+        elif dt is None:
+            return None
+
+        transitions = self.transitions(dt.year)
+
+        if transitions is None:
+            return False
+
+        dt = dt.replace(tzinfo=None)
+
+        isdst = self._naive_isdst(dt, transitions)
+
+        # Handle ambiguous dates
+        if not isdst and self.is_ambiguous(dt):
+            return not self._fold(dt)
+        else:
+            return isdst
+
+    def _naive_isdst(self, dt, transitions):
+        dston, dstoff = transitions
+
+        dt = dt.replace(tzinfo=None)
+
+        if dston < dstoff:
+            isdst = dston <= dt < dstoff
+        else:
+            isdst = not dstoff <= dt < dston
+
+        return isdst
+
+    @property
+    def _dst_base_offset(self):
+        return self._dst_offset - self._std_offset
+
+    __hash__ = None
+
+    def __ne__(self, other):
+        return not (self == other)
+
+    def __repr__(self):
+        return "%s(...)" % self.__class__.__name__
+
+    __reduce__ = object.__reduce__

dateutil/tz/_factories.py

@@ -0,0 +1,80 @@
+from datetime import timedelta
+import weakref
+from collections import OrderedDict
+
+from six.moves import _thread
+
+
+class _TzSingleton(type):
+    def __init__(cls, *args, **kwargs):
+        cls.__instance = None
+        super(_TzSingleton, cls).__init__(*args, **kwargs)
+
+    def __call__(cls):
+        if cls.__instance is None:
+            cls.__instance = super(_TzSingleton, cls).__call__()
+        return cls.__instance
+
+
+class _TzFactory(type):
+    def instance(cls, *args, **kwargs):
+        """Alternate constructor that returns a fresh instance"""
+        return type.__call__(cls, *args, **kwargs)
+
+
+class _TzOffsetFactory(_TzFactory):
+    def __init__(cls, *args, **kwargs):
+        cls.__instances = weakref.WeakValueDictionary()
+        cls.__strong_cache = OrderedDict()
+        cls.__strong_cache_size = 8
+
+        cls._cache_lock = _thread.allocate_lock()
+
+    def __call__(cls, name, offset):
+        if isinstance(offset, timedelta):
+            key = (name, offset.total_seconds())
+        else:
+            key = (name, offset)
+
+        instance = cls.__instances.get(key, None)
+        if instance is None:
+            instance = cls.__instances.setdefault(key,
+                                                  cls.instance(name, offset))
+
+        # This lock may not be necessary in Python 3. See GH issue #901
+        with cls._cache_lock:
+            cls.__strong_cache[key] = cls.__strong_cache.pop(key, instance)
+
+            # Remove an item if the strong cache is overpopulated
+            if len(cls.__strong_cache) > cls.__strong_cache_size:
+                cls.__strong_cache.popitem(last=False)
+
+        return instance
+
+
+class _TzStrFactory(_TzFactory):
+    def __init__(cls, *args, **kwargs):
+        cls.__instances = weakref.WeakValueDictionary()
+        cls.__strong_cache = OrderedDict()
+        cls.__strong_cache_size = 8
+
+        cls.__cache_lock = _thread.allocate_lock()
+
+    def __call__(cls, s, posix_offset=False):
+        key = (s, posix_offset)
+        instance = cls.__instances.get(key, None)
+
+        if instance is None:
+            instance = cls.__instances.setdefault(key,
+                cls.instance(s, posix_offset))
+
+        # This lock may not be necessary in Python 3. See GH issue #901
+        with cls.__cache_lock:
+            cls.__strong_cache[key] = cls.__strong_cache.pop(key, instance)
+
+            # Remove an item if the strong cache is overpopulated
+            if len(cls.__strong_cache) > cls.__strong_cache_size:
+                cls.__strong_cache.popitem(last=False)
+
+        return instance
+

dateutil/tz/tz.py

@@ -0,0 +1,1849 @@
+# -*- coding: utf-8 -*-
+"""
+This module offers timezone implementations subclassing the abstract
+:py:class:`datetime.tzinfo` type. There are classes to handle tzfile format
+files (usually are in :file:`/etc/localtime`, :file:`/usr/share/zoneinfo`,
+etc), TZ environment string (in all known formats), given ranges (with help
+from relative deltas), local machine timezone, fixed offset timezone, and UTC
+timezone.
+"""
+import datetime
+import struct
+import time
+import sys
+import os
+import bisect
+import weakref
+from collections import OrderedDict
+
+import six
+from six import string_types
+from six.moves import _thread
+from ._common import tzname_in_python2, _tzinfo
+from ._common import tzrangebase, enfold
+from ._common import _validate_fromutc_inputs
+
+from ._factories import _TzSingleton, _TzOffsetFactory
+from ._factories import _TzStrFactory
+try:
+    from .win import tzwin, tzwinlocal
+except ImportError:
+    tzwin = tzwinlocal = None
+
+# For warning about rounding tzinfo
+from warnings import warn
+
+ZERO = datetime.timedelta(0)
+EPOCH = datetime.datetime(1970, 1, 1, 0, 0)
+EPOCHORDINAL = EPOCH.toordinal()
+
+
+@six.add_metaclass(_TzSingleton)
+class tzutc(datetime.tzinfo):
+    """
+    This is a tzinfo object that represents the UTC time zone.
+
+    **Examples:**
+
+    .. doctest::
+
+        >>> from datetime import *
+        >>> from dateutil.tz import *
+
+        >>> datetime.now()
+        datetime.datetime(2003, 9, 27, 9, 40, 1, 521290)
+
+        >>> datetime.now(tzutc())
+        datetime.datetime(2003, 9, 27, 12, 40, 12, 156379, tzinfo=tzutc())
+
+        >>> datetime.now(tzutc()).tzname()
+        'UTC'
+
+    .. versionchanged:: 2.7.0
+        ``tzutc()`` is now a singleton, so the result of ``tzutc()`` will
+        always return the same object.
+
+        .. doctest::
+
+            >>> from dateutil.tz import tzutc, UTC
+            >>> tzutc() is tzutc()
+            True
+            >>> tzutc() is UTC
+            True
+    """
+    def utcoffset(self, dt):
+        return ZERO
+
+    def dst(self, dt):
+        return ZERO
+
+    @tzname_in_python2
+    def tzname(self, dt):
+        return "UTC"
+
+    def is_ambiguous(self, dt):
+        """
+        Whether or not the "wall time" of a given datetime is ambiguous in this
+        zone.
+
+        :param dt:
+            A :py:class:`datetime.datetime`, naive or time zone aware.
+
+
+        :return:
+            Returns ``True`` if ambiguous, ``False`` otherwise.
+
+        .. versionadded:: 2.6.0
+        """
+        return False
+
+    @_validate_fromutc_inputs
+    def fromutc(self, dt):
+        """
+        Fast track version of fromutc() returns the original ``dt`` object for
+        any valid :py:class:`datetime.datetime` object.
+        """
+        return dt
+
+    def __eq__(self, other):
+        if not isinstance(other, (tzutc, tzoffset)):
+            return NotImplemented
+
+        return (isinstance(other, tzutc) or
+                (isinstance(other, tzoffset) and other._offset == ZERO))
+
+    __hash__ = None
+
+    def __ne__(self, other):
+        return not (self == other)
+
+    def __repr__(self):
+        return "%s()" % self.__class__.__name__
+
+    __reduce__ = object.__reduce__
+
+
+#: Convenience constant providing a :class:`tzutc()` instance
+#:
+#: .. versionadded:: 2.7.0
+UTC = tzutc()
+
+
+@six.add_metaclass(_TzOffsetFactory)
+class tzoffset(datetime.tzinfo):
+    """
+    A simple class for representing a fixed offset from UTC.
+
+    :param name:
+        The timezone name, to be returned when ``tzname()`` is called.
+    :param offset:
+        The time zone offset in seconds, or (since version 2.6.0, represented
+        as a :py:class:`datetime.timedelta` object).
+    """
+    def __init__(self, name, offset):
+        self._name = name
+
+        try:
+            # Allow a timedelta
+            offset = offset.total_seconds()
+        except (TypeError, AttributeError):
+            pass
+
+        self._offset = datetime.timedelta(seconds=_get_supported_offset(offset))
+
+    def utcoffset(self, dt):
+        return self._offset
+
+    def dst(self, dt):
+        return ZERO
+
+    @tzname_in_python2
+    def tzname(self, dt):
+        return self._name
+
+    @_validate_fromutc_inputs
+    def fromutc(self, dt):
+        return dt + self._offset
+
+    def is_ambiguous(self, dt):
+        """
+        Whether or not the "wall time" of a given datetime is ambiguous in this
+        zone.
+
+        :param dt:
+            A :py:class:`datetime.datetime`, naive or time zone aware.
+        :return:
+            Returns ``True`` if ambiguous, ``False`` otherwise.
+
+        .. versionadded:: 2.6.0
+        """
+        return False
+
+    def __eq__(self, other):
+        if not isinstance(other, tzoffset):
+            return NotImplemented
+
+        return self._offset == other._offset
+
+    __hash__ = None
+
+    def __ne__(self, other):
+        return not (self == other)
+
+    def __repr__(self):
+        return "%s(%s, %s)" % (self.__class__.__name__,
+                               repr(self._name),
+                               int(self._offset.total_seconds()))
+
+    __reduce__ = object.__reduce__
+
+
+class tzlocal(_tzinfo):
+    """
+    A :class:`tzinfo` subclass built around the ``time`` timezone functions.
+    """
+    def __init__(self):
+        super(tzlocal, self).__init__()
+
+        self._std_offset = datetime.timedelta(seconds=-time.timezone)
+        if time.daylight:
+            self._dst_offset = datetime.timedelta(seconds=-time.altzone)
+        else:
+            self._dst_offset = self._std_offset
+
+        self._dst_saved = self._dst_offset - self._std_offset
+        self._hasdst = bool(self._dst_saved)
+        self._tznames = tuple(time.tzname)
+
+    def utcoffset(self, dt):
+        if dt is None and self._hasdst:
+            return None
+
+        if self._isdst(dt):
+            return self._dst_offset
+        else:
+            return self._std_offset
+
+    def dst(self, dt):
+        if dt is None and self._hasdst:
+            return None
+
+        if self._isdst(dt):
+            return self._dst_offset - self._std_offset
+        else:
+            return ZERO
+
+    @tzname_in_python2
+    def tzname(self, dt):
+        return self._tznames[self._isdst(dt)]
+
+    def is_ambiguous(self, dt):
+        """
+        Whether or not the "wall time" of a given datetime is ambiguous in this
+        zone.
+
+        :param dt:
+            A :py:class:`datetime.datetime`, naive or time zone aware.
+
+
+        :return:
+            Returns ``True`` if ambiguous, ``False`` otherwise.
+
+        .. versionadded:: 2.6.0
+        """
+        naive_dst = self._naive_is_dst(dt)
+        return (not naive_dst and
+                (naive_dst != self._naive_is_dst(dt - self._dst_saved)))
+
+    def _naive_is_dst(self, dt):
+        timestamp = _datetime_to_timestamp(dt)
+        return time.localtime(timestamp + time.timezone).tm_isdst
+
+    def _isdst(self, dt, fold_naive=True):
+        # We can't use mktime here. It is unstable when deciding if
+        # the hour near to a change is DST or not.
+        #
+        # timestamp = time.mktime((dt.year, dt.month, dt.day, dt.hour,
+        #                         dt.minute, dt.second, dt.weekday(), 0, -1))
+        # return time.localtime(timestamp).tm_isdst
+        #
+        # The code above yields the following result:
+        #
+        # >>> import tz, datetime
+        # >>> t = tz.tzlocal()
+        # >>> datetime.datetime(2003,2,15,23,tzinfo=t).tzname()
+        # 'BRDT'
+        # >>> datetime.datetime(2003,2,16,0,tzinfo=t).tzname()
+        # 'BRST'
+        # >>> datetime.datetime(2003,2,15,23,tzinfo=t).tzname()
+        # 'BRST'
+        # >>> datetime.datetime(2003,2,15,22,tzinfo=t).tzname()
+        # 'BRDT'
+        # >>> datetime.datetime(2003,2,15,23,tzinfo=t).tzname()
+        # 'BRDT'
+        #
+        # Here is a more stable implementation:
+        #
+        if not self._hasdst:
+            return False
+
+        # Check for ambiguous times:
+        dstval = self._naive_is_dst(dt)
+        fold = getattr(dt, 'fold', None)
+
+        if self.is_ambiguous(dt):
+            if fold is not None:
+                return not self._fold(dt)
+            else:
+                return True
+
+        return dstval
+
+    def __eq__(self, other):
+        if isinstance(other, tzlocal):
+            return (self._std_offset == other._std_offset and
+                    self._dst_offset == other._dst_offset)
+        elif isinstance(other, tzutc):
+            return (not self._hasdst and
+                    self._tznames[0] in {'UTC', 'GMT'} and
+                    self._std_offset == ZERO)
+        elif isinstance(other, tzoffset):
+            return (not self._hasdst and
+                    self._tznames[0] == other._name and
+                    self._std_offset == other._offset)
+        else:
+            return NotImplemented
+
+    __hash__ = None
+
+    def __ne__(self, other):
+        return not (self == other)
+
+    def __repr__(self):
+        return "%s()" % self.__class__.__name__
+
+    __reduce__ = object.__reduce__
+
+
+class _ttinfo(object):
+    __slots__ = ["offset", "delta", "isdst", "abbr",
+                 "isstd", "isgmt", "dstoffset"]
+
+    def __init__(self):
+        for attr in self.__slots__:
+            setattr(self, attr, None)
+
+    def __repr__(self):
+        l = []
+        for attr in self.__slots__:
+            value = getattr(self, attr)
+            if value is not None:
+                l.append("%s=%s" % (attr, repr(value)))
+        return "%s(%s)" % (self.__class__.__name__, ", ".join(l))
+
+    def __eq__(self, other):
+        if not isinstance(other, _ttinfo):
+            return NotImplemented
+
+        return (self.offset == other.offset and
+                self.delta == other.delta and
+                self.isdst == other.isdst and
+                self.abbr == other.abbr and
+                self.isstd == other.isstd and
+                self.isgmt == other.isgmt and
+                self.dstoffset == other.dstoffset)
+
+    __hash__ = None
+
+    def __ne__(self, other):
+        return not (self == other)
+
+    def __getstate__(self):
+        state = {}
+        for name in self.__slots__:
+            state[name] = getattr(self, name, None)
+        return state
+
+    def __setstate__(self, state):
+        for name in self.__slots__:
+            if name in state:
+                setattr(self, name, state[name])
+
+
+class _tzfile(object):
+    """
+    Lightweight class for holding the relevant transition and time zone
+    information read from binary tzfiles.
+    """
+    attrs = ['trans_list', 'trans_list_utc', 'trans_idx', 'ttinfo_list',
+             'ttinfo_std', 'ttinfo_dst', 'ttinfo_before', 'ttinfo_first']
+
+    def __init__(self, **kwargs):
+        for attr in self.attrs:
+            setattr(self, attr, kwargs.get(attr, None))
+
+
+class tzfile(_tzinfo):
+    """
+    This is a ``tzinfo`` subclass that allows one to use the ``tzfile(5)``
+    format timezone files to extract current and historical zone information.
+
+    :param fileobj:
+        This can be an opened file stream or a file name that the time zone
+        information can be read from.
+
+    :param filename:
+        This is an optional parameter specifying the source of the time zone
+        information in the event that ``fileobj`` is a file object. If omitted
+        and ``fileobj`` is a file stream, this parameter will be set either to
+        ``fileobj``'s ``name`` attribute or to ``repr(fileobj)``.
+
+    See `Sources for Time Zone and Daylight Saving Time Data
+    <https://data.iana.org/time-zones/tz-link.html>`_ for more information.
+    Time zone files can be compiled from the `IANA Time Zone database files
+    <https://www.iana.org/time-zones>`_ with the `zic time zone compiler
+    <https://www.freebsd.org/cgi/man.cgi?query=zic&sektion=8>`_
+
+    .. note::
+
+        Only construct a ``tzfile`` directly if you have a specific timezone
+        file on disk that you want to read into a Python ``tzinfo`` object.
+        If you want to get a ``tzfile`` representing a specific IANA zone,
+        (e.g. ``'America/New_York'``), you should call
+        :func:`dateutil.tz.gettz` with the zone identifier.
+
+
+    **Examples:**
+
+    Using the US Eastern time zone as an example, we can see that a ``tzfile``
+    provides time zone information for the standard Daylight Saving offsets:
+
+    .. testsetup:: tzfile
+
+        from dateutil.tz import gettz
+        from datetime import datetime
+
+    .. doctest:: tzfile
+
+        >>> NYC = gettz('America/New_York')
+        >>> NYC
+        tzfile('/usr/share/zoneinfo/America/New_York')
+
+        >>> print(datetime(2016, 1, 3, tzinfo=NYC))     # EST
+        2016-01-03 00:00:00-05:00
+
+        >>> print(datetime(2016, 7, 7, tzinfo=NYC))     # EDT
+        2016-07-07 00:00:00-04:00
+
+
+    The ``tzfile`` structure contains a fully history of the time zone,
+    so historical dates will also have the right offsets. For example, before
+    the adoption of the UTC standards, New York used local solar  mean time:
+
+    .. doctest:: tzfile
+
+       >>> print(datetime(1901, 4, 12, tzinfo=NYC))    # LMT
+       1901-04-12 00:00:00-04:56
+
+    And during World War II, New York was on "Eastern War Time", which was a
+    state of permanent daylight saving time:
+
+    .. doctest:: tzfile
+
+        >>> print(datetime(1944, 2, 7, tzinfo=NYC))    # EWT
+        1944-02-07 00:00:00-04:00
+
+    """
+
+    def __init__(self, fileobj, filename=None):
+        super(tzfile, self).__init__()
+
+        file_opened_here = False
+        if isinstance(fileobj, string_types):
+            self._filename = fileobj
+            fileobj = open(fileobj, 'rb')
+            file_opened_here = True
+        elif filename is not None:
+            self._filename = filename
+        elif hasattr(fileobj, "name"):
+            self._filename = fileobj.name
+        else:
+            self._filename = repr(fileobj)
+
+        if fileobj is not None:
+            if not file_opened_here:
+                fileobj = _nullcontext(fileobj)
+
+            with fileobj as file_stream:
+                tzobj = self._read_tzfile(file_stream)
+
+            self._set_tzdata(tzobj)
+
+    def _set_tzdata(self, tzobj):
+        """ Set the time zone data of this object from a _tzfile object """
+        # Copy the relevant attributes over as private attributes
+        for attr in _tzfile.attrs:
+            setattr(self, '_' + attr, getattr(tzobj, attr))
+
+    def _read_tzfile(self, fileobj):
+        out = _tzfile()
+
+        # From tzfile(5):
+        #
+        # The time zone information files used by tzset(3)
+        # begin with the magic characters "TZif" to identify
+        # them as time zone information files, followed by
+        # sixteen bytes reserved for future use, followed by
+        # six four-byte values of type long, written in a
+        # ``standard'' byte order (the high-order  byte
+        # of the value is written first).
+        if fileobj.read(4).decode() != "TZif":
+            raise ValueError("magic not found")
+
+        fileobj.read(16)
+
+        (
+            # The number of UTC/local indicators stored in the file.
+            ttisgmtcnt,
+
+            # The number of standard/wall indicators stored in the file.
+            ttisstdcnt,
+
+            # The number of leap seconds for which data is
+            # stored in the file.
+            leapcnt,
+
+            # The number of "transition times" for which data
+            # is stored in the file.
+            timecnt,
+
+            # The number of "local time types" for which data
+            # is stored in the file (must not be zero).
+            typecnt,
+
+            # The  number  of  characters  of "time zone
+            # abbreviation strings" stored in the file.
+            charcnt,
+
+        ) = struct.unpack(">6l", fileobj.read(24))
+
+        # The above header is followed by tzh_timecnt four-byte
+        # values  of  type long,  sorted  in ascending order.
+        # These values are written in ``standard'' byte order.
+        # Each is used as a transition time (as  returned  by
+        # time(2)) at which the rules for computing local time
+        # change.
+
+        if timecnt:
+            out.trans_list_utc = list(struct.unpack(">%dl" % timecnt,
+                                                    fileobj.read(timecnt*4)))
+        else:
+            out.trans_list_utc = []
+
+        # Next come tzh_timecnt one-byte values of type unsigned
+        # char; each one tells which of the different types of
+        # ``local time'' types described in the file is associated
+        # with the same-indexed transition time. These values
+        # serve as indices into an array of ttinfo structures that
+        # appears next in the file.
+
+        if timecnt:
+            out.trans_idx = struct.unpack(">%dB" % timecnt,
+                                          fileobj.read(timecnt))
+        else:
+            out.trans_idx = []
+
+        # Each ttinfo structure is written as a four-byte value
+        # for tt_gmtoff  of  type long,  in  a  standard  byte
+        # order, followed  by a one-byte value for tt_isdst
+        # and a one-byte  value  for  tt_abbrind.   In  each
+        # structure, tt_gmtoff  gives  the  number  of
+        # seconds to be added to UTC, tt_isdst tells whether
+        # tm_isdst should be set by  localtime(3),  and
+        # tt_abbrind serves  as an index into the array of
+        # time zone abbreviation characters that follow the
+        # ttinfo structure(s) in the file.
+
+        ttinfo = []
+
+        for i in range(typecnt):
+            ttinfo.append(struct.unpack(">lbb", fileobj.read(6)))
+
+        abbr = fileobj.read(charcnt).decode()
+
+        # Then there are tzh_leapcnt pairs of four-byte
+        # values, written in  standard byte  order;  the
+        # first  value  of  each pair gives the time (as
+        # returned by time(2)) at which a leap second
+        # occurs;  the  second  gives the  total  number of
+        # leap seconds to be applied after the given time.
+        # The pairs of values are sorted in ascending order
+        # by time.
+
+        # Not used, for now (but seek for correct file position)
+        if leapcnt:
+            fileobj.seek(leapcnt * 8, os.SEEK_CUR)
+
+        # Then there are tzh_ttisstdcnt standard/wall
+        # indicators, each stored as a one-byte value;
+        # they tell whether the transition times associated
+        # with local time types were specified as standard
+        # time or wall clock time, and are used when
+        # a time zone file is used in handling POSIX-style
+        # time zone environment variables.
+
+        if ttisstdcnt:
+            isstd = struct.unpack(">%db" % ttisstdcnt,
+                                  fileobj.read(ttisstdcnt))
+
+        # Finally, there are tzh_ttisgmtcnt UTC/local
+        # indicators, each stored as a one-byte value;
+        # they tell whether the transition times associated
+        # with local time types were specified as UTC or
+        # local time, and are used when a time zone file
+        # is used in handling POSIX-style time zone envi-
+        # ronment variables.
+
+        if ttisgmtcnt:
+            isgmt = struct.unpack(">%db" % ttisgmtcnt,
+                                  fileobj.read(ttisgmtcnt))
+
+        # Build ttinfo list
+        out.ttinfo_list = []
+        for i in range(typecnt):
+            gmtoff, isdst, abbrind = ttinfo[i]
+            gmtoff = _get_supported_offset(gmtoff)
+            tti = _ttinfo()
+            tti.offset = gmtoff
+            tti.dstoffset = datetime.timedelta(0)
+            tti.delta = datetime.timedelta(seconds=gmtoff)
+            tti.isdst = isdst
+            tti.abbr = abbr[abbrind:abbr.find('\x00', abbrind)]
+            tti.isstd = (ttisstdcnt > i and isstd[i] != 0)
+            tti.isgmt = (ttisgmtcnt > i and isgmt[i] != 0)
+            out.ttinfo_list.append(tti)
+
+        # Replace ttinfo indexes for ttinfo objects.
+        out.trans_idx = [out.ttinfo_list[idx] for idx in out.trans_idx]
+
+        # Set standard, dst, and before ttinfos. before will be
+        # used when a given time is before any transitions,
+        # and will be set to the first non-dst ttinfo, or to
+        # the first dst, if all of them are dst.
+        out.ttinfo_std = None
+        out.ttinfo_dst = None
+        out.ttinfo_before = None
+        if out.ttinfo_list:
+            if not out.trans_list_utc:
+                out.ttinfo_std = out.ttinfo_first = out.ttinfo_list[0]
+            else:
+                for i in range(timecnt-1, -1, -1):
+                    tti = out.trans_idx[i]
+                    if not out.ttinfo_std and not tti.isdst:
+                        out.ttinfo_std = tti
+                    elif not out.ttinfo_dst and tti.isdst:
+                        out.ttinfo_dst = tti
+
+                    if out.ttinfo_std and out.ttinfo_dst:
+                        break
+                else:
+                    if out.ttinfo_dst and not out.ttinfo_std:
+                        out.ttinfo_std = out.ttinfo_dst
+
+                for tti in out.ttinfo_list:
+                    if not tti.isdst:
+                        out.ttinfo_before = tti
+                        break
+                else:
+                    out.ttinfo_before = out.ttinfo_list[0]
+
+        # Now fix transition times to become relative to wall time.
+        #
+        # I'm not sure about this. In my tests, the tz source file
+        # is setup to wall time, and in the binary file isstd and
+        # isgmt are off, so it should be in wall time. OTOH, it's
+        # always in gmt time. Let me know if you have comments
+        # about this.
+        lastdst = None
+        lastoffset = None
+        lastdstoffset = None
+        lastbaseoffset = None
+        out.trans_list = []
+
+        for i, tti in enumerate(out.trans_idx):
+            offset = tti.offset
+            dstoffset = 0
+
+            if lastdst is not None:
+                if tti.isdst:
+                    if not lastdst:
+                        dstoffset = offset - lastoffset
+
+                    if not dstoffset and lastdstoffset:
+                        dstoffset = lastdstoffset
+
+                    tti.dstoffset = datetime.timedelta(seconds=dstoffset)
+                    lastdstoffset = dstoffset
+
+            # If a time zone changes its base offset during a DST transition,
+            # then you need to adjust by the previous base offset to get the
+            # transition time in local time. Otherwise you use the current
+            # base offset. Ideally, I would have some mathematical proof of
+            # why this is true, but I haven't really thought about it enough.
+            baseoffset = offset - dstoffset
+            adjustment = baseoffset
+            if (lastbaseoffset is not None and baseoffset != lastbaseoffset
+                    and tti.isdst != lastdst):
+                # The base DST has changed
+                adjustment = lastbaseoffset
+
+            lastdst = tti.isdst
+            lastoffset = offset
+            lastbaseoffset = baseoffset
+
+            out.trans_list.append(out.trans_list_utc[i] + adjustment)
+
+        out.trans_idx = tuple(out.trans_idx)
+        out.trans_list = tuple(out.trans_list)
+        out.trans_list_utc = tuple(out.trans_list_utc)
+
+        return out
+
+    def _find_last_transition(self, dt, in_utc=False):
+        # If there's no list, there are no transitions to find
+        if not self._trans_list:
+            return None
+
+        timestamp = _datetime_to_timestamp(dt)
+
+        # Find where the timestamp fits in the transition list - if the
+        # timestamp is a transition time, it's part of the "after" period.
+        trans_list = self._trans_list_utc if in_utc else self._trans_list
+        idx = bisect.bisect_right(trans_list, timestamp)
+
+        # We want to know when the previous transition was, so subtract off 1
+        return idx - 1
+
+    def _get_ttinfo(self, idx):
+        # For no list or after the last transition, default to _ttinfo_std
+        if idx is None or (idx + 1) >= len(self._trans_list):
+            return self._ttinfo_std
+
+        # If there is a list and the time is before it, return _ttinfo_before
+        if idx < 0:
+            return self._ttinfo_before
+
+        return self._trans_idx[idx]
+
+    def _find_ttinfo(self, dt):
+        idx = self._resolve_ambiguous_time(dt)
+
+        return self._get_ttinfo(idx)
+
+    def fromutc(self, dt):
+        """
+        The ``tzfile`` implementation of :py:func:`datetime.tzinfo.fromutc`.
+
+        :param dt:
+            A :py:class:`datetime.datetime` object.
+
+        :raises TypeError:
+            Raised if ``dt`` is not a :py:class:`datetime.datetime` object.
+
+        :raises ValueError:
+            Raised if this is called with a ``dt`` which does not have this
+            ``tzinfo`` attached.
+
+        :return:
+            Returns a :py:class:`datetime.datetime` object representing the
+            wall time in ``self``'s time zone.
+        """
+        # These isinstance checks are in datetime.tzinfo, so we'll preserve
+        # them, even if we don't care about duck typing.
+        if not isinstance(dt, datetime.datetime):
+            raise TypeError("fromutc() requires a datetime argument")
+
+        if dt.tzinfo is not self:
+            raise ValueError("dt.tzinfo is not self")
+
+        # First treat UTC as wall time and get the transition we're in.
+        idx = self._find_last_transition(dt, in_utc=True)
+        tti = self._get_ttinfo(idx)
+
+        dt_out = dt + datetime.timedelta(seconds=tti.offset)
+
+        fold = self.is_ambiguous(dt_out, idx=idx)
+
+        return enfold(dt_out, fold=int(fold))
+
+    def is_ambiguous(self, dt, idx=None):
+        """
+        Whether or not the "wall time" of a given datetime is ambiguous in this
+        zone.
+
+        :param dt:
+            A :py:class:`datetime.datetime`, naive or time zone aware.
+
+
+        :return:
+            Returns ``True`` if ambiguous, ``False`` otherwise.
+
+        .. versionadded:: 2.6.0
+        """
+        if idx is None:
+            idx = self._find_last_transition(dt)
+
+        # Calculate the difference in offsets from current to previous
+        timestamp = _datetime_to_timestamp(dt)
+        tti = self._get_ttinfo(idx)
+
+        if idx is None or idx <= 0:
+            return False
+
+        od = self._get_ttinfo(idx - 1).offset - tti.offset
+        tt = self._trans_list[idx]          # Transition time
+
+        return timestamp < tt + od
+
+    def _resolve_ambiguous_time(self, dt):
+        idx = self._find_last_transition(dt)
+
+        # If we have no transitions, return the index
+        _fold = self._fold(dt)
+        if idx is None or idx == 0:
+            return idx
+
+        # If it's ambiguous and we're in a fold, shift to a different index.
+        idx_offset = int(not _fold and self.is_ambiguous(dt, idx))
+
+        return idx - idx_offset
+
+    def utcoffset(self, dt):
+        if dt is None:
+            return None
+
+        if not self._ttinfo_std:
+            return ZERO
+
+        return self._find_ttinfo(dt).delta
+
+    def dst(self, dt):
+        if dt is None:
+            return None
+
+        if not self._ttinfo_dst:
+            return ZERO
+
+        tti = self._find_ttinfo(dt)
+
+        if not tti.isdst:
+            return ZERO
+
+        # The documentation says that utcoffset()-dst() must
+        # be constant for every dt.
+        return tti.dstoffset
+
+    @tzname_in_python2
+    def tzname(self, dt):
+        if not self._ttinfo_std or dt is None:
+            return None
+        return self._find_ttinfo(dt).abbr
+
+    def __eq__(self, other):
+        if not isinstance(other, tzfile):
+            return NotImplemented
+        return (self._trans_list == other._trans_list and
+                self._trans_idx == other._trans_idx and
+                self._ttinfo_list == other._ttinfo_list)
+
+    __hash__ = None
+
+    def __ne__(self, other):
+        return not (self == other)
+
+    def __repr__(self):
+        return "%s(%s)" % (self.__class__.__name__, repr(self._filename))
+
+    def __reduce__(self):
+        return self.__reduce_ex__(None)
+
+    def __reduce_ex__(self, protocol):
+        return (self.__class__, (None, self._filename), self.__dict__)
+
+
+class tzrange(tzrangebase):
+    """
+    The ``tzrange`` object is a time zone specified by a set of offsets and
+    abbreviations, equivalent to the way the ``TZ`` variable can be specified
+    in POSIX-like systems, but using Python delta objects to specify DST
+    start, end and offsets.
+
+    :param stdabbr:
+        The abbreviation for standard time (e.g. ``'EST'``).
+
+    :param stdoffset:
+        An integer or :class:`datetime.timedelta` object or equivalent
+        specifying the base offset from UTC.
+
+        If unspecified, +00:00 is used.
+
+    :param dstabbr:
+        The abbreviation for DST / "Summer" time (e.g. ``'EDT'``).
+
+        If specified, with no other DST information, DST is assumed to occur
+        and the default behavior or ``dstoffset``, ``start`` and ``end`` is
+        used. If unspecified and no other DST information is specified, it
+        is assumed that this zone has no DST.
+
+        If this is unspecified and other DST information is *is* specified,
+        DST occurs in the zone but the time zone abbreviation is left
+        unchanged.
+
+    :param dstoffset:
+        A an integer or :class:`datetime.timedelta` object or equivalent
+        specifying the UTC offset during DST. If unspecified and any other DST
+        information is specified, it is assumed to be the STD offset +1 hour.
+
+    :param start:
+        A :class:`relativedelta.relativedelta` object or equivalent specifying
+        the time and time of year that daylight savings time starts. To
+        specify, for example, that DST starts at 2AM on the 2nd Sunday in
+        March, pass:
+
+            ``relativedelta(hours=2, month=3, day=1, weekday=SU(+2))``
+
+        If unspecified and any other DST information is specified, the default
+        value is 2 AM on the first Sunday in April.
+
+    :param end:
+        A :class:`relativedelta.relativedelta` object or equivalent
+        representing the time and time of year that daylight savings time
+        ends, with the same specification method as in ``start``. One note is
+        that this should point to the first time in the *standard* zone, so if
+        a transition occurs at 2AM in the DST zone and the clocks are set back
+        1 hour to 1AM, set the ``hours`` parameter to +1.
+
+
+    **Examples:**
+
+    .. testsetup:: tzrange
+
+        from dateutil.tz import tzrange, tzstr
+
+    .. doctest:: tzrange
+
+        >>> tzstr('EST5EDT') == tzrange("EST", -18000, "EDT")
+        True
+
+        >>> from dateutil.relativedelta import *
+        >>> range1 = tzrange("EST", -18000, "EDT")
+        >>> range2 = tzrange("EST", -18000, "EDT", -14400,
+        ...                  relativedelta(hours=+2, month=4, day=1,
+        ...                                weekday=SU(+1)),
+        ...                  relativedelta(hours=+1, month=10, day=31,
+        ...                                weekday=SU(-1)))
+        >>> tzstr('EST5EDT') == range1 == range2
+        True
+
+    """
+    def __init__(self, stdabbr, stdoffset=None,
+                 dstabbr=None, dstoffset=None,
+                 start=None, end=None):
+
+        global relativedelta
+        from dateutil import relativedelta
+
+        self._std_abbr = stdabbr
+        self._dst_abbr = dstabbr
+
+        try:
+            stdoffset = stdoffset.total_seconds()
+        except (TypeError, AttributeError):
+            pass
+
+        try:
+            dstoffset = dstoffset.total_seconds()
+        except (TypeError, AttributeError):
+            pass
+
+        if stdoffset is not None:
+            self._std_offset = datetime.timedelta(seconds=stdoffset)
+        else:
+            self._std_offset = ZERO
+
+        if dstoffset is not None:
+            self._dst_offset = datetime.timedelta(seconds=dstoffset)
+        elif dstabbr and stdoffset is not None:
+            self._dst_offset = self._std_offset + datetime.timedelta(hours=+1)
+        else:
+            self._dst_offset = ZERO
+
+        if dstabbr and start is None:
+            self._start_delta = relativedelta.relativedelta(
+                hours=+2, month=4, day=1, weekday=relativedelta.SU(+1))
+        else:
+            self._start_delta = start
+
+        if dstabbr and end is None:
+            self._end_delta = relativedelta.relativedelta(
+                hours=+1, month=10, day=31, weekday=relativedelta.SU(-1))
+        else:
+            self._end_delta = end
+
+        self._dst_base_offset_ = self._dst_offset - self._std_offset
+        self.hasdst = bool(self._start_delta)
+
+    def transitions(self, year):
+        """
+        For a given year, get the DST on and off transition times, expressed
+        always on the standard time side. For zones with no transitions, this
+        function returns ``None``.
+
+        :param year:
+            The year whose transitions you would like to query.
+
+        :return:
+            Returns a :class:`tuple` of :class:`datetime.datetime` objects,
+            ``(dston, dstoff)`` for zones with an annual DST transition, or
+            ``None`` for fixed offset zones.
+        """
+        if not self.hasdst:
+            return None
+
+        base_year = datetime.datetime(year, 1, 1)
+
+        start = base_year + self._start_delta
+        end = base_year + self._end_delta
+
+        return (start, end)
+
+    def __eq__(self, other):
+        if not isinstance(other, tzrange):
+            return NotImplemented
+
+        return (self._std_abbr == other._std_abbr and
+                self._dst_abbr == other._dst_abbr and
+                self._std_offset == other._std_offset and
+                self._dst_offset == other._dst_offset and
+                self._start_delta == other._start_delta and
+                self._end_delta == other._end_delta)
+
+    @property
+    def _dst_base_offset(self):
+        return self._dst_base_offset_
+
+
+@six.add_metaclass(_TzStrFactory)
+class tzstr(tzrange):
+    """
+    ``tzstr`` objects are time zone objects specified by a time-zone string as
+    it would be passed to a ``TZ`` variable on POSIX-style systems (see
+    the `GNU C Library: TZ Variable`_ for more details).
+
+    There is one notable exception, which is that POSIX-style time zones use an
+    inverted offset format, so normally ``GMT+3`` would be parsed as an offset
+    3 hours *behind* GMT. The ``tzstr`` time zone object will parse this as an
+    offset 3 hours *ahead* of GMT. If you would like to maintain the POSIX
+    behavior, pass a ``True`` value to ``posix_offset``.
+
+    The :class:`tzrange` object provides the same functionality, but is
+    specified using :class:`relativedelta.relativedelta` objects. rather than
+    strings.
+
+    :param s:
+        A time zone string in ``TZ`` variable format. This can be a
+        :class:`bytes` (2.x: :class:`str`), :class:`str` (2.x:
+        :class:`unicode`) or a stream emitting unicode characters
+        (e.g. :class:`StringIO`).
+
+    :param posix_offset:
+        Optional. If set to ``True``, interpret strings such as ``GMT+3`` or
+        ``UTC+3`` as being 3 hours *behind* UTC rather than ahead, per the
+        POSIX standard.
+
+    .. caution::
+
+        Prior to version 2.7.0, this function also supported time zones
+        in the format:
+
+            * ``EST5EDT,4,0,6,7200,10,0,26,7200,3600``
+            * ``EST5EDT,4,1,0,7200,10,-1,0,7200,3600``
+
+        This format is non-standard and has been deprecated; this function
+        will raise a :class:`DeprecatedTZFormatWarning` until
+        support is removed in a future version.
+
+    .. _`GNU C Library: TZ Variable`:
+        https://www.gnu.org/software/libc/manual/html_node/TZ-Variable.html
+    """
+    def __init__(self, s, posix_offset=False):
+        global parser
+        from dateutil.parser import _parser as parser
+
+        self._s = s
+
+        res = parser._parsetz(s)
+        if res is None or res.any_unused_tokens:
+            raise ValueError("unknown string format")
+
+        # Here we break the compatibility with the TZ variable handling.
+        # GMT-3 actually *means* the timezone -3.
+        if res.stdabbr in ("GMT", "UTC") and not posix_offset:
+            res.stdoffset *= -1
+
+        # We must initialize it first, since _delta() needs
+        # _std_offset and _dst_offset set. Use False in start/end
+        # to avoid building it two times.
+        tzrange.__init__(self, res.stdabbr, res.stdoffset,
+                         res.dstabbr, res.dstoffset,
+                         start=False, end=False)
+
+        if not res.dstabbr:
+            self._start_delta = None
+            self._end_delta = None
+        else:
+            self._start_delta = self._delta(res.start)
+            if self._start_delta:
+                self._end_delta = self._delta(res.end, isend=1)
+
+        self.hasdst = bool(self._start_delta)
+
+    def _delta(self, x, isend=0):
+        from dateutil import relativedelta
+        kwargs = {}
+        if x.month is not None:
+            kwargs["month"] = x.month
+            if x.weekday is not None:
+                kwargs["weekday"] = relativedelta.weekday(x.weekday, x.week)
+                if x.week > 0:
+                    kwargs["day"] = 1
+                else:
+                    kwargs["day"] = 31
+            elif x.day:
+                kwargs["day"] = x.day
+        elif x.yday is not None:
+            kwargs["yearday"] = x.yday
+        elif x.jyday is not None:
+            kwargs["nlyearday"] = x.jyday
+        if not kwargs:
+            # Default is to start on first sunday of april, and end
+            # on last sunday of october.
+            if not isend:
+                kwargs["month"] = 4
+                kwargs["day"] = 1
+                kwargs["weekday"] = relativedelta.SU(+1)
+            else:
+                kwargs["month"] = 10
+                kwargs["day"] = 31
+                kwargs["weekday"] = relativedelta.SU(-1)
+        if x.time is not None:
+            kwargs["seconds"] = x.time
+        else:
+            # Default is 2AM.
+            kwargs["seconds"] = 7200
+        if isend:
+            # Convert to standard time, to follow the documented way
+            # of working with the extra hour. See the documentation
+            # of the tzinfo class.
+            delta = self._dst_offset - self._std_offset
+            kwargs["seconds"] -= delta.seconds + delta.days * 86400
+        return relativedelta.relativedelta(**kwargs)
+
+    def __repr__(self):
+        return "%s(%s)" % (self.__class__.__name__, repr(self._s))
+
+
+class _tzicalvtzcomp(object):
+    def __init__(self, tzoffsetfrom, tzoffsetto, isdst,
+                 tzname=None, rrule=None):
+        self.tzoffsetfrom = datetime.timedelta(seconds=tzoffsetfrom)
+        self.tzoffsetto = datetime.timedelta(seconds=tzoffsetto)
+        self.tzoffsetdiff = self.tzoffsetto - self.tzoffsetfrom
+        self.isdst = isdst
+        self.tzname = tzname
+        self.rrule = rrule
+
+
+class _tzicalvtz(_tzinfo):
+    def __init__(self, tzid, comps=[]):
+        super(_tzicalvtz, self).__init__()
+
+        self._tzid = tzid
+        self._comps = comps
+        self._cachedate = []
+        self._cachecomp = []
+        self._cache_lock = _thread.allocate_lock()
+
+    def _find_comp(self, dt):
+        if len(self._comps) == 1:
+            return self._comps[0]
+
+        dt = dt.replace(tzinfo=None)
+
+        try:
+            with self._cache_lock:
+                return self._cachecomp[self._cachedate.index(
+                    (dt, self._fold(dt)))]
+        except ValueError:
+            pass
+
+        lastcompdt = None
+        lastcomp = None
+
+        for comp in self._comps:
+            compdt = self._find_compdt(comp, dt)
+
+            if compdt and (not lastcompdt or lastcompdt < compdt):
+                lastcompdt = compdt
+                lastcomp = comp
+
+        if not lastcomp:
+            # RFC says nothing about what to do when a given
+            # time is before the first onset date. We'll look for the
+            # first standard component, or the first component, if
+            # none is found.
+            for comp in self._comps:
+                if not comp.isdst:
+                    lastcomp = comp
+                    break
+            else:
+                lastcomp = comp[0]
+
+        with self._cache_lock:
+            self._cachedate.insert(0, (dt, self._fold(dt)))
+            self._cachecomp.insert(0, lastcomp)
+
+            if len(self._cachedate) > 10:
+                self._cachedate.pop()
+                self._cachecomp.pop()
+
+        return lastcomp
+
+    def _find_compdt(self, comp, dt):
+        if comp.tzoffsetdiff < ZERO and self._fold(dt):
+            dt -= comp.tzoffsetdiff
+
+        compdt = comp.rrule.before(dt, inc=True)
+
+        return compdt
+
+    def utcoffset(self, dt):
+        if dt is None:
+            return None
+
+        return self._find_comp(dt).tzoffsetto
+
+    def dst(self, dt):
+        comp = self._find_comp(dt)
+        if comp.isdst:
+            return comp.tzoffsetdiff
+        else:
+            return ZERO
+
+    @tzname_in_python2
+    def tzname(self, dt):
+        return self._find_comp(dt).tzname
+
+    def __repr__(self):
+        return "<tzicalvtz %s>" % repr(self._tzid)
+
+    __reduce__ = object.__reduce__
+
+
+class tzical(object):
+    """
+    This object is designed to parse an iCalendar-style ``VTIMEZONE`` structure
+    as set out in `RFC 5545`_ Section 4.6.5 into one or more `tzinfo` objects.
+
+    :param `fileobj`:
+        A file or stream in iCalendar format, which should be UTF-8 encoded
+        with CRLF endings.
+
+    .. _`RFC 5545`: https://tools.ietf.org/html/rfc5545
+    """
+    def __init__(self, fileobj):
+        global rrule
+        from dateutil import rrule
+
+        if isinstance(fileobj, string_types):
+            self._s = fileobj
+            # ical should be encoded in UTF-8 with CRLF
+            fileobj = open(fileobj, 'r')
+        else:
+            self._s = getattr(fileobj, 'name', repr(fileobj))
+            fileobj = _nullcontext(fileobj)
+
+        self._vtz = {}
+
+        with fileobj as fobj:
+            self._parse_rfc(fobj.read())
+
+    def keys(self):
+        """
+        Retrieves the available time zones as a list.
+        """
+        return list(self._vtz.keys())
+
+    def get(self, tzid=None):
+        """
+        Retrieve a :py:class:`datetime.tzinfo` object by its ``tzid``.
+
+        :param tzid:
+            If there is exactly one time zone available, omitting ``tzid``
+            or passing :py:const:`None` value returns it. Otherwise a valid
+            key (which can be retrieved from :func:`keys`) is required.
+
+        :raises ValueError:
+            Raised if ``tzid`` is not specified but there are either more
+            or fewer than 1 zone defined.
+
+        :returns:
+            Returns either a :py:class:`datetime.tzinfo` object representing
+            the relevant time zone or :py:const:`None` if the ``tzid`` was
+            not found.
+        """
+        if tzid is None:
+            if len(self._vtz) == 0:
+                raise ValueError("no timezones defined")
+            elif len(self._vtz) > 1:
+                raise ValueError("more than one timezone available")
+            tzid = next(iter(self._vtz))
+
+        return self._vtz.get(tzid)
+
+    def _parse_offset(self, s):
+        s = s.strip()
+        if not s:
+            raise ValueError("empty offset")
+        if s[0] in ('+', '-'):
+            signal = (-1, +1)[s[0] == '+']
+            s = s[1:]
+        else:
+            signal = +1
+        if len(s) == 4:
+            return (int(s[:2]) * 3600 + int(s[2:]) * 60) * signal
+        elif len(s) == 6:
+            return (int(s[:2]) * 3600 + int(s[2:4]) * 60 + int(s[4:])) * signal
+        else:
+            raise ValueError("invalid offset: " + s)
+
+    def _parse_rfc(self, s):
+        lines = s.splitlines()
+        if not lines:
+            raise ValueError("empty string")
+
+        # Unfold
+        i = 0
+        while i < len(lines):
+            line = lines[i].rstrip()
+            if not line:
+                del lines[i]
+            elif i > 0 and line[0] == " ":
+                lines[i-1] += line[1:]
+                del lines[i]
+            else:
+                i += 1
+
+        tzid = None
+        comps = []
+        invtz = False
+        comptype = None
+        for line in lines:
+            if not line:
+                continue
+            name, value = line.split(':', 1)
+            parms = name.split(';')
+            if not parms:
+                raise ValueError("empty property name")
+            name = parms[0].upper()
+            parms = parms[1:]
+            if invtz:
+                if name == "BEGIN":
+                    if value in ("STANDARD", "DAYLIGHT"):
+                        # Process component
+                        pass
+                    else:
+                        raise ValueError("unknown component: "+value)
+                    comptype = value
+                    founddtstart = False
+                    tzoffsetfrom = None
+                    tzoffsetto = None
+                    rrulelines = []
+                    tzname = None
+                elif name == "END":
+                    if value == "VTIMEZONE":
+                        if comptype:
+                            raise ValueError("component not closed: "+comptype)
+                        if not tzid:
+                            raise ValueError("mandatory TZID not found")
+                        if not comps:
+                            raise ValueError(
+                                "at least one component is needed")
+                        # Process vtimezone
+                        self._vtz[tzid] = _tzicalvtz(tzid, comps)
+                        invtz = False
+                    elif value == comptype:
+                        if not founddtstart:
+                            raise ValueError("mandatory DTSTART not found")
+                        if tzoffsetfrom is None:
+                            raise ValueError(
+                                "mandatory TZOFFSETFROM not found")
+                        if tzoffsetto is None:
+                            raise ValueError(
+                                "mandatory TZOFFSETFROM not found")
+                        # Process component
+                        rr = None
+                        if rrulelines:
+                            rr = rrule.rrulestr("\n".join(rrulelines),
+                                                compatible=True,
+                                                ignoretz=True,
+                                                cache=True)
+                        comp = _tzicalvtzcomp(tzoffsetfrom, tzoffsetto,
+                                              (comptype == "DAYLIGHT"),
+                                              tzname, rr)
+                        comps.append(comp)
+                        comptype = None
+                    else:
+                        raise ValueError("invalid component end: "+value)
+                elif comptype:
+                    if name == "DTSTART":
+                        # DTSTART in VTIMEZONE takes a subset of valid RRULE
+                        # values under RFC 5545.
+                        for parm in parms:
+                            if parm != 'VALUE=DATE-TIME':
+                                msg = ('Unsupported DTSTART param in ' +
+                                       'VTIMEZONE: ' + parm)
+                                raise ValueError(msg)
+                        rrulelines.append(line)
+                        founddtstart = True
+                    elif name in ("RRULE", "RDATE", "EXRULE", "EXDATE"):
+                        rrulelines.append(line)
+                    elif name == "TZOFFSETFROM":
+                        if parms:
+                            raise ValueError(
+                                "unsupported %s parm: %s " % (name, parms[0]))
+                        tzoffsetfrom = self._parse_offset(value)
+                    elif name == "TZOFFSETTO":
+                        if parms:
+                            raise ValueError(
+                                "unsupported TZOFFSETTO parm: "+parms[0])
+                        tzoffsetto = self._parse_offset(value)
+                    elif name == "TZNAME":
+                        if parms:
+                            raise ValueError(
+                                "unsupported TZNAME parm: "+parms[0])
+                        tzname = value
+                    elif name == "COMMENT":
+                        pass
+                    else:
+                        raise ValueError("unsupported property: "+name)
+                else:
+                    if name == "TZID":
+                        if parms:
+                            raise ValueError(
+                                "unsupported TZID parm: "+parms[0])
+                        tzid = value
+                    elif name in ("TZURL", "LAST-MODIFIED", "COMMENT"):
+                        pass
+                    else:
+                        raise ValueError("unsupported property: "+name)
+            elif name == "BEGIN" and value == "VTIMEZONE":
+                tzid = None
+                comps = []
+                invtz = True
+
+    def __repr__(self):
+        return "%s(%s)" % (self.__class__.__name__, repr(self._s))
+
+
+if sys.platform != "win32":
+    TZFILES = ["/etc/localtime", "localtime"]
+    TZPATHS = ["/usr/share/zoneinfo",
+               "/usr/lib/zoneinfo",
+               "/usr/share/lib/zoneinfo",
+               "/etc/zoneinfo"]
+else:
+    TZFILES = []
+    TZPATHS = []
+
+
+def __get_gettz():
+    tzlocal_classes = (tzlocal,)
+    if tzwinlocal is not None:
+        tzlocal_classes += (tzwinlocal,)
+
+    class GettzFunc(object):
+        """
+        Retrieve a time zone object from a string representation
+
+        This function is intended to retrieve the :py:class:`tzinfo` subclass
+        that best represents the time zone that would be used if a POSIX
+        `TZ variable`_ were set to the same value.
+
+        If no argument or an empty string is passed to ``gettz``, local time
+        is returned:
+
+        .. code-block:: python3
+
+            >>> gettz()
+            tzfile('/etc/localtime')
+
+        This function is also the preferred way to map IANA tz database keys
+        to :class:`tzfile` objects:
+
+        .. code-block:: python3
+
+            >>> gettz('Pacific/Kiritimati')
+            tzfile('/usr/share/zoneinfo/Pacific/Kiritimati')
+
+        On Windows, the standard is extended to include the Windows-specific
+        zone names provided by the operating system:
+
+        .. code-block:: python3
+
+            >>> gettz('Egypt Standard Time')
+            tzwin('Egypt Standard Time')
+
+        Passing a GNU ``TZ`` style string time zone specification returns a
+        :class:`tzstr` object:
+
+        .. code-block:: python3
+
+            >>> gettz('AEST-10AEDT-11,M10.1.0/2,M4.1.0/3')
+            tzstr('AEST-10AEDT-11,M10.1.0/2,M4.1.0/3')
+
+        :param name:
+            A time zone name (IANA, or, on Windows, Windows keys), location of
+            a ``tzfile(5)`` zoneinfo file or ``TZ`` variable style time zone
+            specifier. An empty string, no argument or ``None`` is interpreted
+            as local time.
+
+        :return:
+            Returns an instance of one of ``dateutil``'s :py:class:`tzinfo`
+            subclasses.
+
+        .. versionchanged:: 2.7.0
+
+            After version 2.7.0, any two calls to ``gettz`` using the same
+            input strings will return the same object:
+
+            .. code-block:: python3
+
+                >>> tz.gettz('America/Chicago') is tz.gettz('America/Chicago')
+                True
+
+            In addition to improving performance, this ensures that
+            `"same zone" semantics`_ are used for datetimes in the same zone.
+
+
+        .. _`TZ variable`:
+            https://www.gnu.org/software/libc/manual/html_node/TZ-Variable.html
+
+        .. _`"same zone" semantics`:
+            https://blog.ganssle.io/articles/2018/02/aware-datetime-arithmetic.html
+        """
+        def __init__(self):
+
+            self.__instances = weakref.WeakValueDictionary()
+            self.__strong_cache_size = 8
+            self.__strong_cache = OrderedDict()
+            self._cache_lock = _thread.allocate_lock()
+
+        def __call__(self, name=None):
+            with self._cache_lock:
+                rv = self.__instances.get(name, None)
+
+                if rv is None:
+                    rv = self.nocache(name=name)
+                    if not (name is None
+                            or isinstance(rv, tzlocal_classes)
+                            or rv is None):
+                        # tzlocal is slightly more complicated than the other
+                        # time zone providers because it depends on environment
+                        # at construction time, so don't cache that.
+                        #
+                        # We also cannot store weak references to None, so we
+                        # will also not store that.
+                        self.__instances[name] = rv
+                    else:
+                        # No need for strong caching, return immediately
+                        return rv
+
+                self.__strong_cache[name] = self.__strong_cache.pop(name, rv)
+
+                if len(self.__strong_cache) > self.__strong_cache_size:
+                    self.__strong_cache.popitem(last=False)
+
+            return rv
+
+        def set_cache_size(self, size):
+            with self._cache_lock:
+                self.__strong_cache_size = size
+                while len(self.__strong_cache) > size:
+                    self.__strong_cache.popitem(last=False)
+
+        def cache_clear(self):
+            with self._cache_lock:
+                self.__instances = weakref.WeakValueDictionary()
+                self.__strong_cache.clear()
+
+        @staticmethod
+        def nocache(name=None):
+            """A non-cached version of gettz"""
+            tz = None
+            if not name:
+                try:
+                    name = os.environ["TZ"]
+                except KeyError:
+                    pass
+            if name is None or name in ("", ":"):
+                for filepath in TZFILES:
+                    if not os.path.isabs(filepath):
+                        filename = filepath
+                        for path in TZPATHS:
+                            filepath = os.path.join(path, filename)
+                            if os.path.isfile(filepath):
+                                break
+                        else:
+                            continue
+                    if os.path.isfile(filepath):
+                        try:
+                            tz = tzfile(filepath)
+                            break
+                        except (IOError, OSError, ValueError):
+                            pass
+                else:
+                    tz = tzlocal()
+            else:
+                try:
+                    if name.startswith(":"):
+                        name = name[1:]
+                except TypeError as e:
+                    if isinstance(name, bytes):
+                        new_msg = "gettz argument should be str, not bytes"
+                        six.raise_from(TypeError(new_msg), e)
+                    else:
+                        raise
+                if os.path.isabs(name):
+                    if os.path.isfile(name):
+                        tz = tzfile(name)
+                    else:
+                        tz = None
+                else:
+                    for path in TZPATHS:
+                        filepath = os.path.join(path, name)
+                        if not os.path.isfile(filepath):
+                            filepath = filepath.replace(' ', '_')
+                            if not os.path.isfile(filepath):
+                                continue
+                        try:
+                            tz = tzfile(filepath)
+                            break
+                        except (IOError, OSError, ValueError):
+                            pass
+                    else:
+                        tz = None
+                        if tzwin is not None:
+                            try:
+                                tz = tzwin(name)
+                            except (WindowsError, UnicodeEncodeError):
+                                # UnicodeEncodeError is for Python 2.7 compat
+                                tz = None
+
+                        if not tz:
+                            from dateutil.zoneinfo import get_zonefile_instance
+                            tz = get_zonefile_instance().get(name)
+
+                        if not tz:
+                            for c in name:
+                                # name is not a tzstr unless it has at least
+                                # one offset. For short values of "name", an
+                                # explicit for loop seems to be the fastest way
+                                # To determine if a string contains a digit
+                                if c in "0123456789":
+                                    try:
+                                        tz = tzstr(name)
+                                    except ValueError:
+                                        pass
+                                    break
+                            else:
+                                if name in ("GMT", "UTC"):
+                                    tz = UTC
+                                elif name in time.tzname:
+                                    tz = tzlocal()
+            return tz
+
+    return GettzFunc()
+
+
+gettz = __get_gettz()
+del __get_gettz
+
+
+def datetime_exists(dt, tz=None):
+    """
+    Given a datetime and a time zone, determine whether or not a given datetime
+    would fall in a gap.
+
+    :param dt:
+        A :class:`datetime.datetime` (whose time zone will be ignored if ``tz``
+        is provided.)
+
+    :param tz:
+        A :class:`datetime.tzinfo` with support for the ``fold`` attribute. If
+        ``None`` or not provided, the datetime's own time zone will be used.
+
+    :return:
+        Returns a boolean value whether or not the "wall time" exists in
+        ``tz``.
+
+    .. versionadded:: 2.7.0
+    """
+    if tz is None:
+        if dt.tzinfo is None:
+            raise ValueError('Datetime is naive and no time zone provided.')
+        tz = dt.tzinfo
+
+    dt = dt.replace(tzinfo=None)
+
+    # This is essentially a test of whether or not the datetime can survive
+    # a round trip to UTC.
+    dt_rt = dt.replace(tzinfo=tz).astimezone(UTC).astimezone(tz)
+    dt_rt = dt_rt.replace(tzinfo=None)
+
+    return dt == dt_rt
+
+
+def datetime_ambiguous(dt, tz=None):
+    """
+    Given a datetime and a time zone, determine whether or not a given datetime
+    is ambiguous (i.e if there are two times differentiated only by their DST
+    status).
+
+    :param dt:
+        A :class:`datetime.datetime` (whose time zone will be ignored if ``tz``
+        is provided.)
+
+    :param tz:
+        A :class:`datetime.tzinfo` with support for the ``fold`` attribute. If
+        ``None`` or not provided, the datetime's own time zone will be used.
+
+    :return:
+        Returns a boolean value whether or not the "wall time" is ambiguous in
+        ``tz``.
+
+    .. versionadded:: 2.6.0
+    """
+    if tz is None:
+        if dt.tzinfo is None:
+            raise ValueError('Datetime is naive and no time zone provided.')
+
+        tz = dt.tzinfo
+
+    # If a time zone defines its own "is_ambiguous" function, we'll use that.
+    is_ambiguous_fn = getattr(tz, 'is_ambiguous', None)
+    if is_ambiguous_fn is not None:
+        try:
+            return tz.is_ambiguous(dt)
+        except Exception:
+            pass
+
+    # If it doesn't come out and tell us it's ambiguous, we'll just check if
+    # the fold attribute has any effect on this particular date and time.
+    dt = dt.replace(tzinfo=tz)
+    wall_0 = enfold(dt, fold=0)
+    wall_1 = enfold(dt, fold=1)
+
+    same_offset = wall_0.utcoffset() == wall_1.utcoffset()
+    same_dst = wall_0.dst() == wall_1.dst()
+
+    return not (same_offset and same_dst)
+
+
+def resolve_imaginary(dt):
+    """
+    Given a datetime that may be imaginary, return an existing datetime.
+
+    This function assumes that an imaginary datetime represents what the
+    wall time would be in a zone had the offset transition not occurred, so
+    it will always fall forward by the transition's change in offset.
+
+    .. doctest::
+
+        >>> from dateutil import tz
+        >>> from datetime import datetime
+        >>> NYC = tz.gettz('America/New_York')
+        >>> print(tz.resolve_imaginary(datetime(2017, 3, 12, 2, 30, tzinfo=NYC)))
+        2017-03-12 03:30:00-04:00
+
+        >>> KIR = tz.gettz('Pacific/Kiritimati')
+        >>> print(tz.resolve_imaginary(datetime(1995, 1, 1, 12, 30, tzinfo=KIR)))
+        1995-01-02 12:30:00+14:00
+
+    As a note, :func:`datetime.astimezone` is guaranteed to produce a valid,
+    existing datetime, so a round-trip to and from UTC is sufficient to get
+    an extant datetime, however, this generally "falls back" to an earlier time
+    rather than falling forward to the STD side (though no guarantees are made
+    about this behavior).
+
+    :param dt:
+        A :class:`datetime.datetime` which may or may not exist.
+
+    :return:
+        Returns an existing :class:`datetime.datetime`. If ``dt`` was not
+        imaginary, the datetime returned is guaranteed to be the same object
+        passed to the function.
+
+    .. versionadded:: 2.7.0
+    """
+    if dt.tzinfo is not None and not datetime_exists(dt):
+
+        curr_offset = (dt + datetime.timedelta(hours=24)).utcoffset()
+        old_offset = (dt - datetime.timedelta(hours=24)).utcoffset()
+
+        dt += curr_offset - old_offset
+
+    return dt
+
+
+def _datetime_to_timestamp(dt):
+    """
+    Convert a :class:`datetime.datetime` object to an epoch timestamp in
+    seconds since January 1, 1970, ignoring the time zone.
+    """
+    return (dt.replace(tzinfo=None) - EPOCH).total_seconds()
+
+
+if sys.version_info >= (3, 6):
+    def _get_supported_offset(second_offset):
+        return second_offset
+else:
+    def _get_supported_offset(second_offset):
+        # For python pre-3.6, round to full-minutes if that's not the case.
+        # Python's datetime doesn't accept sub-minute timezones. Check
+        # http://python.org/sf/1447945 or https://bugs.python.org/issue5288
+        # for some information.
+        old_offset = second_offset
+        calculated_offset = 60 * ((second_offset + 30) // 60)
+        return calculated_offset
+
+
+try:
+    # Python 3.7 feature
+    from contextlib import nullcontext as _nullcontext
+except ImportError:
+    class _nullcontext(object):
+        """
+        Class for wrapping contexts so that they are passed through in a
+        with statement.
+        """
+        def __init__(self, context):
+            self.context = context
+
+        def __enter__(self):
+            return self.context
+
+        def __exit__(*args, **kwargs):
+            pass
+
+# vim:ts=4:sw=4:et

dateutil/tz/win.py

@@ -0,0 +1,370 @@
+# -*- coding: utf-8 -*-
+"""
+This module provides an interface to the native time zone data on Windows,
+including :py:class:`datetime.tzinfo` implementations.
+
+Attempting to import this module on a non-Windows platform will raise an
+:py:obj:`ImportError`.
+"""
+# This code was originally contributed by Jeffrey Harris.
+import datetime
+import struct
+
+from six.moves import winreg
+from six import text_type
+
+try:
+    import ctypes
+    from ctypes import wintypes
+except ValueError:
+    # ValueError is raised on non-Windows systems for some horrible reason.
+    raise ImportError("Running tzwin on non-Windows system")
+
+from ._common import tzrangebase
+
+__all__ = ["tzwin", "tzwinlocal", "tzres"]
+
+ONEWEEK = datetime.timedelta(7)
+
+TZKEYNAMENT = r"SOFTWARE\Microsoft\Windows NT\CurrentVersion\Time Zones"
+TZKEYNAME9X = r"SOFTWARE\Microsoft\Windows\CurrentVersion\Time Zones"
+TZLOCALKEYNAME = r"SYSTEM\CurrentControlSet\Control\TimeZoneInformation"
+
+
+def _settzkeyname():
+    handle = winreg.ConnectRegistry(None, winreg.HKEY_LOCAL_MACHINE)
+    try:
+        winreg.OpenKey(handle, TZKEYNAMENT).Close()
+        TZKEYNAME = TZKEYNAMENT
+    except WindowsError:
+        TZKEYNAME = TZKEYNAME9X
+    handle.Close()
+    return TZKEYNAME
+
+
+TZKEYNAME = _settzkeyname()
+
+
+class tzres(object):
+    """
+    Class for accessing ``tzres.dll``, which contains timezone name related
+    resources.
+
+    .. versionadded:: 2.5.0
+    """
+    p_wchar = ctypes.POINTER(wintypes.WCHAR)        # Pointer to a wide char
+
+    def __init__(self, tzres_loc='tzres.dll'):
+        # Load the user32 DLL so we can load strings from tzres
+        user32 = ctypes.WinDLL('user32')
+
+        # Specify the LoadStringW function
+        user32.LoadStringW.argtypes = (wintypes.HINSTANCE,
+                                       wintypes.UINT,
+                                       wintypes.LPWSTR,
+                                       ctypes.c_int)
+
+        self.LoadStringW = user32.LoadStringW
+        self._tzres = ctypes.WinDLL(tzres_loc)
+        self.tzres_loc = tzres_loc
+
+    def load_name(self, offset):
+        """
+        Load a timezone name from a DLL offset (integer).
+
+        >>> from dateutil.tzwin import tzres
+        >>> tzr = tzres()
+        >>> print(tzr.load_name(112))
+        'Eastern Standard Time'
+
+        :param offset:
+            A positive integer value referring to a string from the tzres dll.
+
+        .. note::
+
+            Offsets found in the registry are generally of the form
+            ``@tzres.dll,-114``. The offset in this case is 114, not -114.
+
+        """
+        resource = self.p_wchar()
+        lpBuffer = ctypes.cast(ctypes.byref(resource), wintypes.LPWSTR)
+        nchar = self.LoadStringW(self._tzres._handle, offset, lpBuffer, 0)
+        return resource[:nchar]
+
+    def name_from_string(self, tzname_str):
+        """
+        Parse strings as returned from the Windows registry into the time zone
+        name as defined in the registry.
+
+        >>> from dateutil.tzwin import tzres
+        >>> tzr = tzres()
+        >>> print(tzr.name_from_string('@tzres.dll,-251'))
+        'Dateline Daylight Time'
+        >>> print(tzr.name_from_string('Eastern Standard Time'))
+        'Eastern Standard Time'
+
+        :param tzname_str:
+            A timezone name string as returned from a Windows registry key.
+
+        :return:
+            Returns the localized timezone string from tzres.dll if the string
+            is of the form `@tzres.dll,-offset`, else returns the input string.
+        """
+        if not tzname_str.startswith('@'):
+            return tzname_str
+
+        name_splt = tzname_str.split(',-')
+        try:
+            offset = int(name_splt[1])
+        except:
+            raise ValueError("Malformed timezone string.")
+
+        return self.load_name(offset)
+
+
+class tzwinbase(tzrangebase):
+    """tzinfo class based on win32's timezones available in the registry."""
+    def __init__(self):
+        raise NotImplementedError('tzwinbase is an abstract base class')
+
+    def __eq__(self, other):
+        # Compare on all relevant dimensions, including name.
+        if not isinstance(other, tzwinbase):
+            return NotImplemented
+
+        return  (self._std_offset == other._std_offset and
+                 self._dst_offset == other._dst_offset and
+                 self._stddayofweek == other._stddayofweek and
+                 self._dstdayofweek == other._dstdayofweek and
+                 self._stdweeknumber == other._stdweeknumber and
+                 self._dstweeknumber == other._dstweeknumber and
+                 self._stdhour == other._stdhour and
+                 self._dsthour == other._dsthour and
+                 self._stdminute == other._stdminute and
+                 self._dstminute == other._dstminute and
+                 self._std_abbr == other._std_abbr and
+                 self._dst_abbr == other._dst_abbr)
+
+    @staticmethod
+    def list():
+        """Return a list of all time zones known to the system."""
+        with winreg.ConnectRegistry(None, winreg.HKEY_LOCAL_MACHINE) as handle:
+            with winreg.OpenKey(handle, TZKEYNAME) as tzkey:
+                result = [winreg.EnumKey(tzkey, i)
+                          for i in range(winreg.QueryInfoKey(tzkey)[0])]
+        return result
+
+    def display(self):
+        """
+        Return the display name of the time zone.
+        """
+        return self._display
+
+    def transitions(self, year):
+        """
+        For a given year, get the DST on and off transition times, expressed
+        always on the standard time side. For zones with no transitions, this
+        function returns ``None``.
+
+        :param year:
+            The year whose transitions you would like to query.
+
+        :return:
+            Returns a :class:`tuple` of :class:`datetime.datetime` objects,
+            ``(dston, dstoff)`` for zones with an annual DST transition, or
+            ``None`` for fixed offset zones.
+        """
+
+        if not self.hasdst:
+            return None
+
+        dston = picknthweekday(year, self._dstmonth, self._dstdayofweek,
+                               self._dsthour, self._dstminute,
+                               self._dstweeknumber)
+
+        dstoff = picknthweekday(year, self._stdmonth, self._stddayofweek,
+                                self._stdhour, self._stdminute,
+                                self._stdweeknumber)
+
+        # Ambiguous dates default to the STD side
+        dstoff -= self._dst_base_offset
+
+        return dston, dstoff
+
+    def _get_hasdst(self):
+        return self._dstmonth != 0
+
+    @property
+    def _dst_base_offset(self):
+        return self._dst_base_offset_
+
+
+class tzwin(tzwinbase):
+    """
+    Time zone object created from the zone info in the Windows registry
+
+    These are similar to :py:class:`dateutil.tz.tzrange` objects in that
+    the time zone data is provided in the format of a single offset rule
+    for either 0 or 2 time zone transitions per year.
+
+    :param: name
+        The name of a Windows time zone key, e.g. "Eastern Standard Time".
+        The full list of keys can be retrieved with :func:`tzwin.list`.
+    """
+
+    def __init__(self, name):
+        self._name = name
+
+        with winreg.ConnectRegistry(None, winreg.HKEY_LOCAL_MACHINE) as handle:
+            tzkeyname = text_type("{kn}\\{name}").format(kn=TZKEYNAME, name=name)
+            with winreg.OpenKey(handle, tzkeyname) as tzkey:
+                keydict = valuestodict(tzkey)
+
+        self._std_abbr = keydict["Std"]
+        self._dst_abbr = keydict["Dlt"]
+
+        self._display = keydict["Display"]
+
+        # See http://ww_winreg.jsiinc.com/SUBA/tip0300/rh0398.htm
+        tup = struct.unpack("=3l16h", keydict["TZI"])
+        stdoffset = -tup[0]-tup[1]          # Bias + StandardBias * -1
+        dstoffset = stdoffset-tup[2]        # + DaylightBias * -1
+        self._std_offset = datetime.timedelta(minutes=stdoffset)
+        self._dst_offset = datetime.timedelta(minutes=dstoffset)
+
+        # for the meaning see the win32 TIME_ZONE_INFORMATION structure docs
+        # http://msdn.microsoft.com/en-us/library/windows/desktop/ms725481(v=vs.85).aspx
+        (self._stdmonth,
+         self._stddayofweek,   # Sunday = 0
+         self._stdweeknumber,  # Last = 5
+         self._stdhour,
+         self._stdminute) = tup[4:9]
+
+        (self._dstmonth,
+         self._dstdayofweek,   # Sunday = 0
+         self._dstweeknumber,  # Last = 5
+         self._dsthour,
+         self._dstminute) = tup[12:17]
+
+        self._dst_base_offset_ = self._dst_offset - self._std_offset
+        self.hasdst = self._get_hasdst()
+
+    def __repr__(self):
+        return "tzwin(%s)" % repr(self._name)
+
+    def __reduce__(self):
+        return (self.__class__, (self._name,))
+
+
+class tzwinlocal(tzwinbase):
+    """
+    Class representing the local time zone information in the Windows registry
+
+    While :class:`dateutil.tz.tzlocal` makes system calls (via the :mod:`time`
+    module) to retrieve time zone information, ``tzwinlocal`` retrieves the
+    rules directly from the Windows registry and creates an object like
+    :class:`dateutil.tz.tzwin`.
+
+    Because Windows does not have an equivalent of :func:`time.tzset`, on
+    Windows, :class:`dateutil.tz.tzlocal` instances will always reflect the
+    time zone settings *at the time that the process was started*, meaning
+    changes to the machine's time zone settings during the run of a program
+    on Windows will **not** be reflected by :class:`dateutil.tz.tzlocal`.
+    Because ``tzwinlocal`` reads the registry directly, it is unaffected by
+    this issue.
+    """
+    def __init__(self):
+        with winreg.ConnectRegistry(None, winreg.HKEY_LOCAL_MACHINE) as handle:
+            with winreg.OpenKey(handle, TZLOCALKEYNAME) as tzlocalkey:
+                keydict = valuestodict(tzlocalkey)
+
+            self._std_abbr = keydict["StandardName"]
+            self._dst_abbr = keydict["DaylightName"]
+
+            try:
+                tzkeyname = text_type('{kn}\\{sn}').format(kn=TZKEYNAME,
+                                                          sn=self._std_abbr)
+                with winreg.OpenKey(handle, tzkeyname) as tzkey:
+                    _keydict = valuestodict(tzkey)
+                    self._display = _keydict["Display"]
+            except OSError:
+                self._display = None
+
+        stdoffset = -keydict["Bias"]-keydict["StandardBias"]
+        dstoffset = stdoffset-keydict["DaylightBias"]
+
+        self._std_offset = datetime.timedelta(minutes=stdoffset)
+        self._dst_offset = datetime.timedelta(minutes=dstoffset)
+
+        # For reasons unclear, in this particular key, the day of week has been
+        # moved to the END of the SYSTEMTIME structure.
+        tup = struct.unpack("=8h", keydict["StandardStart"])
+
+        (self._stdmonth,
+         self._stdweeknumber,  # Last = 5
+         self._stdhour,
+         self._stdminute) = tup[1:5]
+
+        self._stddayofweek = tup[7]
+
+        tup = struct.unpack("=8h", keydict["DaylightStart"])
+
+        (self._dstmonth,
+         self._dstweeknumber,  # Last = 5
+         self._dsthour,
+         self._dstminute) = tup[1:5]
+
+        self._dstdayofweek = tup[7]
+
+        self._dst_base_offset_ = self._dst_offset - self._std_offset
+        self.hasdst = self._get_hasdst()
+
+    def __repr__(self):
+        return "tzwinlocal()"
+
+    def __str__(self):
+        # str will return the standard name, not the daylight name.
+        return "tzwinlocal(%s)" % repr(self._std_abbr)
+
+    def __reduce__(self):
+        return (self.__class__, ())
+
+
+def picknthweekday(year, month, dayofweek, hour, minute, whichweek):
+    """ dayofweek == 0 means Sunday, whichweek 5 means last instance """
+    first = datetime.datetime(year, month, 1, hour, minute)
+
+    # This will work if dayofweek is ISO weekday (1-7) or Microsoft-style (0-6),
+    # Because 7 % 7 = 0
+    weekdayone = first.replace(day=((dayofweek - first.isoweekday()) % 7) + 1)
+    wd = weekdayone + ((whichweek - 1) * ONEWEEK)
+    if (wd.month != month):
+        wd -= ONEWEEK
+
+    return wd
+
+
+def valuestodict(key):
+    """Convert a registry key's values to a dictionary."""
+    dout = {}
+    size = winreg.QueryInfoKey(key)[1]
+    tz_res = None
+
+    for i in range(size):
+        key_name, value, dtype = winreg.EnumValue(key, i)
+        if dtype == winreg.REG_DWORD or dtype == winreg.REG_DWORD_LITTLE_ENDIAN:
+            # If it's a DWORD (32-bit integer), it's stored as unsigned - convert
+            # that to a proper signed integer
+            if value & (1 << 31):
+                value = value - (1 << 32)
+        elif dtype == winreg.REG_SZ:
+            # If it's a reference to the tzres DLL, load the actual string
+            if value.startswith('@tzres'):
+                tz_res = tz_res or tzres()
+                value = tz_res.name_from_string(value)
+
+            value = value.rstrip('\x00')    # Remove trailing nulls
+
+        dout[key_name] = value
+
+    return dout

dateutil/zoneinfo/__init__.py

@@ -0,0 +1,167 @@
+# -*- coding: utf-8 -*-
+import warnings
+import json
+
+from tarfile import TarFile
+from pkgutil import get_data
+from io import BytesIO
+
+from dateutil.tz import tzfile as _tzfile
+
+__all__ = ["get_zonefile_instance", "gettz", "gettz_db_metadata"]
+
+ZONEFILENAME = "dateutil-zoneinfo.tar.gz"
+METADATA_FN = 'METADATA'
+
+
+class tzfile(_tzfile):
+    def __reduce__(self):
+        return (gettz, (self._filename,))
+
+
+def getzoneinfofile_stream():
+    try:
+        return BytesIO(get_data(__name__, ZONEFILENAME))
+    except IOError as e:  # TODO  switch to FileNotFoundError?
+        warnings.warn("I/O error({0}): {1}".format(e.errno, e.strerror))
+        return None
+
+
+class ZoneInfoFile(object):
+    def __init__(self, zonefile_stream=None):
+        if zonefile_stream is not None:
+            with TarFile.open(fileobj=zonefile_stream) as tf:
+                self.zones = {zf.name: tzfile(tf.extractfile(zf), filename=zf.name)
+                              for zf in tf.getmembers()
+                              if zf.isfile() and zf.name != METADATA_FN}
+                # deal with links: They'll point to their parent object. Less
+                # waste of memory
+                links = {zl.name: self.zones[zl.linkname]
+                         for zl in tf.getmembers() if
+                         zl.islnk() or zl.issym()}
+                self.zones.update(links)
+                try:
+                    metadata_json = tf.extractfile(tf.getmember(METADATA_FN))
+                    metadata_str = metadata_json.read().decode('UTF-8')
+                    self.metadata = json.loads(metadata_str)
+                except KeyError:
+                    # no metadata in tar file
+                    self.metadata = None
+        else:
+            self.zones = {}
+            self.metadata = None
+
+    def get(self, name, default=None):
+        """
+        Wrapper for :func:`ZoneInfoFile.zones.get`. This is a convenience method
+        for retrieving zones from the zone dictionary.
+
+        :param name:
+            The name of the zone to retrieve. (Generally IANA zone names)
+
+        :param default:
+            The value to return in the event of a missing key.
+
+        .. versionadded:: 2.6.0
+
+        """
+        return self.zones.get(name, default)
+
+
+# The current API has gettz as a module function, although in fact it taps into
+# a stateful class. So as a workaround for now, without changing the API, we
+# will create a new "global" class instance the first time a user requests a
+# timezone. Ugly, but adheres to the api.
+#
+# TODO: Remove after deprecation period.
+_CLASS_ZONE_INSTANCE = []
+
+
+def get_zonefile_instance(new_instance=False):
+    """
+    This is a convenience function which provides a :class:`ZoneInfoFile`
+    instance using the data provided by the ``dateutil`` package. By default, it
+    caches a single instance of the ZoneInfoFile object and returns that.
+
+    :param new_instance:
+        If ``True``, a new instance of :class:`ZoneInfoFile` is instantiated and
+        used as the cached instance for the next call. Otherwise, new instances
+        are created only as necessary.
+
+    :return:
+        Returns a :class:`ZoneInfoFile` object.
+
+    .. versionadded:: 2.6
+    """
+    if new_instance:
+        zif = None
+    else:
+        zif = getattr(get_zonefile_instance, '_cached_instance', None)
+
+    if zif is None:
+        zif = ZoneInfoFile(getzoneinfofile_stream())
+
+        get_zonefile_instance._cached_instance = zif
+
+    return zif
+
+
+def gettz(name):
+    """
+    This retrieves a time zone from the local zoneinfo tarball that is packaged
+    with dateutil.
+
+    :param name:
+        An IANA-style time zone name, as found in the zoneinfo file.
+
+    :return:
+        Returns a :class:`dateutil.tz.tzfile` time zone object.
+
+    .. warning::
+        It is generally inadvisable to use this function, and it is only
+        provided for API compatibility with earlier versions. This is *not*
+        equivalent to ``dateutil.tz.gettz()``, which selects an appropriate
+        time zone based on the inputs, favoring system zoneinfo. This is ONLY
+        for accessing the dateutil-specific zoneinfo (which may be out of
+        date compared to the system zoneinfo).
+
+    .. deprecated:: 2.6
+        If you need to use a specific zoneinfofile over the system zoneinfo,
+        instantiate a :class:`dateutil.zoneinfo.ZoneInfoFile` object and call
+        :func:`dateutil.zoneinfo.ZoneInfoFile.get(name)` instead.
+
+        Use :func:`get_zonefile_instance` to retrieve an instance of the
+        dateutil-provided zoneinfo.
+    """
+    warnings.warn("zoneinfo.gettz() will be removed in future versions, "
+                  "to use the dateutil-provided zoneinfo files, instantiate a "
+                  "ZoneInfoFile object and use ZoneInfoFile.zones.get() "
+                  "instead. See the documentation for details.",
+                  DeprecationWarning)
+
+    if len(_CLASS_ZONE_INSTANCE) == 0:
+        _CLASS_ZONE_INSTANCE.append(ZoneInfoFile(getzoneinfofile_stream()))
+    return _CLASS_ZONE_INSTANCE[0].zones.get(name)
+
+
+def gettz_db_metadata():
+    """ Get the zonefile metadata
+
+    See `zonefile_metadata`_
+
+    :returns:
+        A dictionary with the database metadata
+
+    .. deprecated:: 2.6
+        See deprecation warning in :func:`zoneinfo.gettz`. To get metadata,
+        query the attribute ``zoneinfo.ZoneInfoFile.metadata``.
+    """
+    warnings.warn("zoneinfo.gettz_db_metadata() will be removed in future "
+                  "versions, to use the dateutil-provided zoneinfo files, "
+                  "ZoneInfoFile object and query the 'metadata' attribute "
+                  "instead. See the documentation for details.",
+                  DeprecationWarning)
+
+    if len(_CLASS_ZONE_INSTANCE) == 0:
+        _CLASS_ZONE_INSTANCE.append(ZoneInfoFile(getzoneinfofile_stream()))
+    return _CLASS_ZONE_INSTANCE[0].metadata

dateutil/zoneinfo/rebuild.py

@@ -0,0 +1,75 @@
+import logging
+import os
+import tempfile
+import shutil
+import json
+from subprocess import check_call, check_output
+from tarfile import TarFile
+
+from dateutil.zoneinfo import METADATA_FN, ZONEFILENAME
+
+
+def rebuild(filename, tag=None, format="gz", zonegroups=[], metadata=None):
+    """Rebuild the internal timezone info in dateutil/zoneinfo/zoneinfo*tar*
+
+    filename is the timezone tarball from ``ftp.iana.org/tz``.
+
+    """
+    tmpdir = tempfile.mkdtemp()
+    zonedir = os.path.join(tmpdir, "zoneinfo")
+    moduledir = os.path.dirname(__file__)
+    try:
+        with TarFile.open(filename) as tf:
+            for name in zonegroups:
+                tf.extract(name, tmpdir)
+            filepaths = [os.path.join(tmpdir, n) for n in zonegroups]
+
+            _run_zic(zonedir, filepaths)
+
+        # write metadata file
+        with open(os.path.join(zonedir, METADATA_FN), 'w') as f:
+            json.dump(metadata, f, indent=4, sort_keys=True)
+        target = os.path.join(moduledir, ZONEFILENAME)
+        with TarFile.open(target, "w:%s" % format) as tf:
+            for entry in os.listdir(zonedir):
+                entrypath = os.path.join(zonedir, entry)
+                tf.add(entrypath, entry)
+    finally:
+        shutil.rmtree(tmpdir)
+
+
+def _run_zic(zonedir, filepaths):
+    """Calls the ``zic`` compiler in a compatible way to get a "fat" binary.
+
+    Recent versions of ``zic`` default to ``-b slim``, while older versions
+    don't even have the ``-b`` option (but default to "fat" binaries). The
+    current version of dateutil does not support Version 2+ TZif files, which
+    causes problems when used in conjunction with "slim" binaries, so this
+    function is used to ensure that we always get a "fat" binary.
+    """
+
+    try:
+        help_text = check_output(["zic", "--help"])
+    except OSError as e:
+        _print_on_nosuchfile(e)
+        raise
+
+    if b"-b " in help_text:
+        bloat_args = ["-b", "fat"]
+    else:
+        bloat_args = []
+
+    check_call(["zic"] + bloat_args + ["-d", zonedir] + filepaths)
+
+
+def _print_on_nosuchfile(e):
+    """Print helpful troubleshooting message
+
+    e is an exception raised by subprocess.check_call()
+
+    """
+    if e.errno == 2:
+        logging.error(
+            "Could not find zic. Perhaps you need to install "
+            "libc-bin or some other package that provides it, "
+            "or it's not in your PATH?")

pandas/_config/__init__.py

@@ -0,0 +1,45 @@
+"""
+pandas._config is considered explicitly upstream of everything else in pandas,
+should have no intra-pandas dependencies.
+
+importing `dates` and `display` ensures that keys needed by _libs
+are initialized.
+"""
+
+__all__ = [
+    "config",
+    "describe_option",
+    "detect_console_encoding",
+    "get_option",
+    "option_context",
+    "options",
+    "reset_option",
+    "set_option",
+]
+from pandas._config import config
+from pandas._config import dates  # pyright: ignore[reportUnusedImport]  # noqa: F401
+from pandas._config.config import (
+    _global_config,
+    describe_option,
+    get_option,
+    option_context,
+    options,
+    reset_option,
+    set_option,
+)
+from pandas._config.display import detect_console_encoding
+
+
+def using_string_dtype() -> bool:
+    _mode_options = _global_config["future"]
+    return _mode_options["infer_string"]
+
+
+def using_python_scalars() -> bool:
+    _mode_options = _global_config["future"]
+    return _mode_options["python_scalars"]
+
+
+def is_nan_na() -> bool:
+    _mode_options = _global_config["future"]
+    return not _mode_options["distinguish_nan_and_na"]

pandas/_config/config.py

@@ -0,0 +1,954 @@
+"""
+The config module holds package-wide configurables and provides
+a uniform API for working with them.
+
+Overview
+========
+
+This module supports the following requirements:
+- options are referenced using keys in dot.notation, e.g. "x.y.option - z".
+- keys are case-insensitive.
+- functions should accept partial/regex keys, when unambiguous.
+- options can be registered by modules at import time.
+- options can be registered at init-time (via core.config_init)
+- options have a default value, and (optionally) a description and
+  validation function associated with them.
+- options can be deprecated, in which case referencing them
+  should produce a warning.
+- deprecated options can optionally be rerouted to a replacement
+  so that accessing a deprecated option reroutes to a differently
+  named option.
+- options can be reset to their default value.
+- all option can be reset to their default value at once.
+- all options in a certain sub - namespace can be reset at once.
+- the user can set / get / reset or ask for the description of an option.
+- a developer can register and mark an option as deprecated.
+- you can register a callback to be invoked when the option value
+  is set or reset. Changing the stored value is considered misuse, but
+  is not verboten.
+
+Implementation
+==============
+
+- Data is stored using nested dictionaries, and should be accessed
+  through the provided API.
+
+- "Registered options" and "Deprecated options" have metadata associated
+  with them, which are stored in auxiliary dictionaries keyed on the
+  fully-qualified key, e.g. "x.y.z.option".
+
+- the config_init module is imported by the package's __init__.py file.
+  placing any register_option() calls there will ensure those options
+  are available as soon as pandas is loaded. If you use register_option
+  in a module, it will only be available after that module is imported,
+  which you should be aware of.
+
+- `config_prefix` is a context_manager (for use with the `with` keyword)
+  which can save developers some typing, see the docstring.
+
+"""
+
+from __future__ import annotations
+
+from contextlib import contextmanager
+import re
+from typing import (
+    TYPE_CHECKING,
+    Any,
+    NamedTuple,
+    cast,
+)
+import warnings
+
+from pandas._typing import F
+from pandas.util._exceptions import find_stack_level
+
+if TYPE_CHECKING:
+    from collections.abc import (
+        Callable,
+        Generator,
+        Sequence,
+    )
+
+
+class DeprecatedOption(NamedTuple):
+    key: str
+    category: type[Warning]
+    msg: str | None
+    rkey: str | None
+    removal_ver: str | None
+
+
+class RegisteredOption(NamedTuple):
+    key: str
+    defval: Any
+    doc: str
+    validator: Callable[[object], Any] | None
+    cb: Callable[[str], Any] | None
+
+
+# holds deprecated option metadata
+_deprecated_options: dict[str, DeprecatedOption] = {}
+
+# holds registered option metadata
+_registered_options: dict[str, RegisteredOption] = {}
+
+# holds the current values for registered options
+_global_config: dict[str, Any] = {}
+
+# keys which have a special meaning
+_reserved_keys: list[str] = ["all"]
+
+
+class OptionError(AttributeError, KeyError):
+    """
+    Exception raised for pandas.options.
+
+    Backwards compatible with KeyError checks.
+
+    See Also
+    --------
+    options : Access and modify global pandas settings.
+
+    Examples
+    --------
+    >>> pd.options.context
+    Traceback (most recent call last):
+    OptionError: No such option
+    """
+
+    __module__ = "pandas.errors"
+
+
+#
+# User API
+
+
+def _get_single_key(pat: str) -> str:
+    keys = _select_options(pat)
+    if len(keys) == 0:
+        _warn_if_deprecated(pat)
+        raise OptionError(f"No such keys(s): {pat!r}")
+    if len(keys) > 1:
+        raise OptionError("Pattern matched multiple keys")
+    key = keys[0]
+
+    _warn_if_deprecated(key)
+
+    key = _translate_key(key)
+
+    return key
+
+
+def get_option(pat: str) -> Any:
+    """
+    Retrieve the value of the specified option.
+
+    This method allows users to query the current value of a given option
+    in the pandas configuration system. Options control various display,
+    performance, and behavior-related settings within pandas.
+
+    Parameters
+    ----------
+    pat : str
+        Regexp which should match a single option.
+
+        .. warning::
+
+            Partial matches are supported for convenience, but unless you use the
+            full option name (e.g. x.y.z.option_name), your code may break in future
+            versions if new options with similar names are introduced.
+
+    Returns
+    -------
+    Any
+        The value of the option.
+
+    Raises
+    ------
+    OptionError : if no such option exists
+
+    See Also
+    --------
+    set_option : Set the value of the specified option or options.
+    reset_option : Reset one or more options to their default value.
+    describe_option : Print the description for one or more registered options.
+
+    Notes
+    -----
+    For all available options, please view the :ref:`User Guide <options.available>`
+    or use ``pandas.describe_option()``.
+
+    Examples
+    --------
+    >>> pd.get_option("display.max_columns")  # doctest: +SKIP
+    4
+    """
+    key = _get_single_key(pat)
+
+    # walk the nested dict
+    root, k = _get_root(key)
+    return root[k]
+
+
+def set_option(*args) -> None:
+    """
+    Set the value of the specified option or options.
+
+    This method allows fine-grained control over the behavior and display settings
+    of pandas. Options affect various functionalities such as output formatting,
+    display limits, and operational behavior. Settings can be modified at runtime
+    without requiring changes to global configurations or environment variables.
+
+    Parameters
+    ----------
+    *args : str | object | dict
+        Arguments provided in pairs, which will be interpreted as (pattern, value),
+        or as a single dictionary containing multiple option-value pairs.
+        pattern: str
+        Regexp which should match a single option
+        value: object
+        New value of option
+
+        .. warning::
+
+            Partial pattern matches are supported for convenience, but unless you
+            use the full option name (e.g. x.y.z.option_name), your code may break in
+            future versions if new options with similar names are introduced.
+
+    Returns
+    -------
+    None
+        No return value.
+
+    Raises
+    ------
+    ValueError if odd numbers of non-keyword arguments are provided
+    TypeError if keyword arguments are provided
+    OptionError if no such option exists
+
+    See Also
+    --------
+    get_option : Retrieve the value of the specified option.
+    reset_option : Reset one or more options to their default value.
+    describe_option : Print the description for one or more registered options.
+    option_context : Context manager to temporarily set options in a ``with``
+        statement.
+
+    Notes
+    -----
+    For all available options, please view the :ref:`User Guide <options.available>`
+    or use ``pandas.describe_option()``.
+
+    Examples
+    --------
+    Option-Value Pair Input:
+
+    >>> pd.set_option("display.max_columns", 4)
+    >>> df = pd.DataFrame([[1, 2, 3, 4, 5], [6, 7, 8, 9, 10]])
+    >>> df
+    0  1  ...  3   4
+    0  1  2  ...  4   5
+    1  6  7  ...  9  10
+    [2 rows x 5 columns]
+    >>> pd.reset_option("display.max_columns")
+
+    Dictionary Input:
+
+    >>> pd.set_option({"display.max_columns": 4, "display.precision": 1})
+    >>> df = pd.DataFrame([[1, 2, 3, 4, 5], [6, 7, 8, 9, 10]])
+    >>> df
+    0  1  ...  3   4
+    0  1  2  ...  4   5
+    1  6  7  ...  9  10
+    [2 rows x 5 columns]
+    >>> pd.reset_option("display.max_columns")
+    >>> pd.reset_option("display.precision")
+    """
+    # Handle dictionary input
+    if len(args) == 1 and isinstance(args[0], dict):
+        args = tuple(kv for item in args[0].items() for kv in item)
+
+    nargs = len(args)
+    if not nargs or nargs % 2 != 0:
+        raise ValueError("Must provide an even number of non-keyword arguments")
+
+    for k, v in zip(args[::2], args[1::2], strict=True):
+        key = _get_single_key(k)
+
+        opt = _get_registered_option(key)
+        if opt and opt.validator:
+            opt.validator(v)
+
+        # walk the nested dict
+        root, k_root = _get_root(key)
+        root[k_root] = v
+
+        if opt.cb:
+            opt.cb(key)
+
+
+def describe_option(pat: str = "", _print_desc: bool = True) -> str | None:
+    """
+    Print the description for one or more registered options.
+
+    Call with no arguments to get a listing for all registered options.
+
+    Parameters
+    ----------
+    pat : str, default ""
+        String or string regexp pattern.
+        Empty string will return all options.
+        For regexp strings, all matching keys will have their description displayed.
+    _print_desc : bool, default True
+        If True (default) the description(s) will be printed to stdout.
+        Otherwise, the description(s) will be returned as a string
+        (for testing).
+
+    Returns
+    -------
+    None
+        If ``_print_desc=True``.
+    str
+        If the description(s) as a string if ``_print_desc=False``.
+
+    See Also
+    --------
+    get_option : Retrieve the value of the specified option.
+    set_option : Set the value of the specified option or options.
+    reset_option : Reset one or more options to their default value.
+
+    Notes
+    -----
+    For all available options, please view the
+    :ref:`User Guide <options.available>`.
+
+    Examples
+    --------
+    >>> pd.describe_option("display.max_columns")  # doctest: +SKIP
+    display.max_columns : int
+        If max_cols is exceeded, switch to truncate view...
+    """
+    keys = _select_options(pat)
+    if len(keys) == 0:
+        raise OptionError(f"No such keys(s) for {pat=}")
+
+    s = "\n".join([_build_option_description(k) for k in keys])
+
+    if _print_desc:
+        print(s)
+        return None
+    return s
+
+
+def reset_option(pat: str) -> None:
+    """
+    Reset one or more options to their default value.
+
+    This method resets the specified pandas option(s) back to their default
+    values. It allows partial string matching for convenience, but users should
+    exercise caution to avoid unintended resets due to changes in option names
+    in future versions.
+
+    Parameters
+    ----------
+    pat : str/regex
+        If specified only options matching ``pat*`` will be reset.
+        Pass ``"all"`` as argument to reset all options.
+
+        .. warning::
+
+            Partial matches are supported for convenience, but unless you
+            use the full option name (e.g. x.y.z.option_name), your code may break
+            in future versions if new options with similar names are introduced.
+
+    Returns
+    -------
+    None
+        No return value.
+
+    See Also
+    --------
+    get_option : Retrieve the value of the specified option.
+    set_option : Set the value of the specified option or options.
+    describe_option : Print the description for one or more registered options.
+
+    Notes
+    -----
+    For all available options, please view the
+    :ref:`User Guide <options.available>`.
+
+    Examples
+    --------
+    >>> pd.reset_option("display.max_columns")  # doctest: +SKIP
+    """
+    keys = _select_options(pat)
+
+    if len(keys) == 0:
+        raise OptionError(f"No such keys(s) for {pat=}")
+
+    if len(keys) > 1 and len(pat) < 4 and pat != "all":
+        raise ValueError(
+            "You must specify at least 4 characters when "
+            "resetting multiple keys, use the special keyword "
+            '"all" to reset all the options to their default value'
+        )
+
+    for k in keys:
+        set_option(k, _registered_options[k].defval)
+
+
+def get_default_val(pat: str):
+    key = _get_single_key(pat)
+    return _get_registered_option(key).defval
+
+
+class DictWrapper:
+    """provide attribute-style access to a nested dict"""
+
+    d: dict[str, Any]
+
+    def __init__(self, d: dict[str, Any], prefix: str = "") -> None:
+        object.__setattr__(self, "d", d)
+        object.__setattr__(self, "prefix", prefix)
+
+    def __setattr__(self, key: str, val: Any) -> None:
+        prefix = object.__getattribute__(self, "prefix")
+        if prefix:
+            prefix += "."
+        prefix += key
+        # you can't set new keys
+        # can you can't overwrite subtrees
+        if key in self.d and not isinstance(self.d[key], dict):
+            set_option(prefix, val)
+        else:
+            raise OptionError("You can only set the value of existing options")
+
+    def __getattr__(self, key: str):
+        prefix = object.__getattribute__(self, "prefix")
+        if prefix:
+            prefix += "."
+        prefix += key
+        try:
+            v = object.__getattribute__(self, "d")[key]
+        except KeyError as err:
+            raise OptionError("No such option") from err
+        if isinstance(v, dict):
+            return DictWrapper(v, prefix)
+        else:
+            return get_option(prefix)
+
+    def __dir__(self) -> list[str]:
+        return list(self.d.keys())
+
+
+options = DictWrapper(_global_config)
+# DictWrapper defines a custom setattr
+object.__setattr__(options, "__module__", "pandas")
+
+#
+# Functions for use by pandas developers, in addition to User - api
+
+
+@contextmanager
+def option_context(*args) -> Generator[None]:
+    """
+    Context manager to temporarily set options in a ``with`` statement.
+
+    This method allows users to set one or more pandas options temporarily
+    within a controlled block. The previous options' values are restored
+    once the block is exited. This is useful when making temporary adjustments
+    to pandas' behavior without affecting the global state.
+
+    Parameters
+    ----------
+    *args : str | object | dict
+        An even amount of arguments provided in pairs which will be
+        interpreted as (pattern, value) pairs. Alternatively, a single
+        dictionary of {pattern: value} may be provided.
+
+    Returns
+    -------
+    None
+        No return value.
+
+    Yields
+    ------
+    None
+        No yield value.
+
+    See Also
+    --------
+    get_option : Retrieve the value of the specified option.
+    set_option : Set the value of the specified option.
+    reset_option : Reset one or more options to their default value.
+    describe_option : Print the description for one or more registered options.
+
+    Notes
+    -----
+    For all available options, please view the :ref:`User Guide <options.available>`
+    or use ``pandas.describe_option()``.
+
+    Examples
+    --------
+    >>> from pandas import option_context
+    >>> with option_context("display.max_rows", 10, "display.max_columns", 5):
+    ...     pass
+    >>> with option_context({"display.max_rows": 10, "display.max_columns": 5}):
+    ...     pass
+    """
+    if len(args) == 1 and isinstance(args[0], dict):
+        args = tuple(kv for item in args[0].items() for kv in item)
+
+    if len(args) % 2 != 0 or len(args) < 2:
+        raise ValueError(
+            "Provide an even amount of arguments as "
+            "option_context(pat, val, pat, val...)."
+        )
+
+    ops = tuple(zip(args[::2], args[1::2], strict=True))
+    undo: tuple[tuple[Any, Any], ...] = ()
+    try:
+        undo = tuple((pat, get_option(pat)) for pat, val in ops)
+        for pat, val in ops:
+            set_option(pat, val)
+        yield
+    finally:
+        for pat, val in undo:
+            set_option(pat, val)
+
+
+def register_option(
+    key: str,
+    defval: object,
+    doc: str = "",
+    validator: Callable[[object], Any] | None = None,
+    cb: Callable[[str], Any] | None = None,
+) -> None:
+    """
+    Register an option in the package-wide pandas config object
+
+    Parameters
+    ----------
+    key : str
+        Fully-qualified key, e.g. "x.y.option - z".
+    defval : object
+        Default value of the option.
+    doc : str
+        Description of the option.
+    validator : Callable, optional
+        Function of a single argument, should raise `ValueError` if
+        called with a value which is not a legal value for the option.
+    cb
+        a function of a single argument "key", which is called
+        immediately after an option value is set/reset. key is
+        the full name of the option.
+
+    Raises
+    ------
+    ValueError if `validator` is specified and `defval` is not a valid value.
+
+    """
+    import keyword
+    import tokenize
+
+    key = key.lower()
+
+    if key in _registered_options:
+        raise OptionError(f"Option '{key}' has already been registered")
+    if key in _reserved_keys:
+        raise OptionError(f"Option '{key}' is a reserved key")
+
+    # the default value should be legal
+    if validator:
+        validator(defval)
+
+    # walk the nested dict, creating dicts as needed along the path
+    path = key.split(".")
+
+    for k in path:
+        if not re.match("^" + tokenize.Name + "$", k):
+            raise ValueError(f"{k} is not a valid identifier")
+        if keyword.iskeyword(k):
+            raise ValueError(f"{k} is a python keyword")
+
+    cursor = _global_config
+    msg = "Path prefix to option '{option}' is already an option"
+
+    for i, p in enumerate(path[:-1]):
+        if not isinstance(cursor, dict):
+            raise OptionError(msg.format(option=".".join(path[:i])))
+        if p not in cursor:
+            cursor[p] = {}
+        cursor = cursor[p]
+
+    if not isinstance(cursor, dict):
+        raise OptionError(msg.format(option=".".join(path[:-1])))
+
+    cursor[path[-1]] = defval  # initialize
+
+    # save the option metadata
+    _registered_options[key] = RegisteredOption(
+        key=key, defval=defval, doc=doc, validator=validator, cb=cb
+    )
+
+
+def deprecate_option(
+    key: str,
+    category: type[Warning],
+    msg: str | None = None,
+    rkey: str | None = None,
+    removal_ver: str | None = None,
+) -> None:
+    """
+    Mark option `key` as deprecated, if code attempts to access this option,
+    a warning will be produced, using `msg` if given, or a default message
+    if not.
+    if `rkey` is given, any access to the key will be re-routed to `rkey`.
+
+    Neither the existence of `key` nor that if `rkey` is checked. If they
+    do not exist, any subsequence access will fail as usual, after the
+    deprecation warning is given.
+
+    Parameters
+    ----------
+    key : str
+        Name of the option to be deprecated.
+        must be a fully-qualified option name (e.g "x.y.z.rkey").
+    category : Warning
+        Warning class for the deprecation.
+    msg : str, optional
+        Warning message to output when the key is referenced.
+        if no message is given a default message will be emitted.
+    rkey : str, optional
+        Name of an option to reroute access to.
+        If specified, any referenced `key` will be
+        re-routed to `rkey` including set/get/reset.
+        rkey must be a fully-qualified option name (e.g "x.y.z.rkey").
+        used by the default message if no `msg` is specified.
+    removal_ver : str, optional
+        Specifies the version in which this option will
+        be removed. used by the default message if no `msg` is specified.
+
+    Raises
+    ------
+    OptionError
+        If the specified key has already been deprecated.
+    """
+    key = key.lower()
+
+    if key in _deprecated_options:
+        raise OptionError(f"Option '{key}' has already been defined as deprecated.")
+
+    _deprecated_options[key] = DeprecatedOption(key, category, msg, rkey, removal_ver)
+
+
+#
+# functions internal to the module
+
+
+def _select_options(pat: str) -> list[str]:
+    """
+    returns a list of keys matching `pat`
+
+    if pat=="all", returns all registered options
+    """
+    # short-circuit for exact key
+    if pat in _registered_options:
+        return [pat]
+
+    # else look through all of them
+    keys = sorted(_registered_options.keys())
+    if pat == "all":  # reserved key
+        return keys
+
+    return [k for k in keys if re.search(pat, k, re.I)]
+
+
+def _get_root(key: str) -> tuple[dict[str, Any], str]:
+    path = key.split(".")
+    cursor = _global_config
+    for p in path[:-1]:
+        cursor = cursor[p]
+    return cursor, path[-1]
+
+
+def _get_deprecated_option(key: str):
+    """
+    Retrieves the metadata for a deprecated option, if `key` is deprecated.
+
+    Returns
+    -------
+    DeprecatedOption (namedtuple) if key is deprecated, None otherwise
+    """
+    try:
+        d = _deprecated_options[key]
+    except KeyError:
+        return None
+    else:
+        return d
+
+
+def _get_registered_option(key: str):
+    """
+    Retrieves the option metadata if `key` is a registered option.
+
+    Returns
+    -------
+    RegisteredOption (namedtuple) if key is deprecated, None otherwise
+    """
+    return _registered_options.get(key)
+
+
+def _translate_key(key: str) -> str:
+    """
+    if `key` is deprecated and a replacement key defined, will return the
+    replacement key, otherwise returns `key` as-is
+    """
+    d = _get_deprecated_option(key)
+    if d:
+        return d.rkey or key
+    else:
+        return key
+
+
+def _warn_if_deprecated(key: str) -> bool:
+    """
+    Checks if `key` is a deprecated option and if so, prints a warning.
+
+    Returns
+    -------
+    bool - True if `key` is deprecated, False otherwise.
+    """
+    d = _get_deprecated_option(key)
+    if d:
+        if d.msg:
+            warnings.warn(
+                d.msg,
+                d.category,
+                stacklevel=find_stack_level(),
+            )
+        else:
+            msg = f"'{key}' is deprecated"
+            if d.removal_ver:
+                msg += f" and will be removed in {d.removal_ver}"
+            if d.rkey:
+                msg += f", please use '{d.rkey}' instead."
+            else:
+                msg += ", please refrain from using it."
+
+            warnings.warn(
+                msg,
+                d.category,
+                stacklevel=find_stack_level(),
+            )
+        return True
+    return False
+
+
+def _build_option_description(k: str) -> str:
+    """Builds a formatted description of a registered option and prints it"""
+    o = _get_registered_option(k)
+    d = _get_deprecated_option(k)
+
+    s = f"{k} "
+
+    if o.doc:
+        s += "\n".join(o.doc.strip().split("\n"))
+    else:
+        s += "No description available."
+
+    if o:
+        with warnings.catch_warnings():
+            warnings.simplefilter("ignore", FutureWarning)
+            warnings.simplefilter("ignore", DeprecationWarning)
+            s += f"\n    [default: {o.defval}] [currently: {get_option(k)}]"
+
+    if d:
+        rkey = d.rkey or ""
+        s += "\n    (Deprecated"
+        s += f", use `{rkey}` instead."
+        s += ")"
+
+    return s
+
+
+# helpers
+
+
+@contextmanager
+def config_prefix(prefix: str) -> Generator[None]:
+    """
+    contextmanager for multiple invocations of API with a common prefix
+
+    supported API functions: (register / get / set )__option
+
+    Warning: This is not thread - safe, and won't work properly if you import
+    the API functions into your module using the "from x import y" construct.
+
+    Example
+    -------
+    import pandas._config.config as cf
+    with cf.config_prefix("display.font"):
+        cf.register_option("color", "red")
+        cf.register_option("size", " 5 pt")
+        cf.set_option(size, " 6 pt")
+        cf.get_option(size)
+        ...
+
+        etc'
+
+    will register options "display.font.color", "display.font.size", set the
+    value of "display.font.size"... and so on.
+    """
+    # Note: reset_option relies on set_option, and on key directly
+    # it does not fit in to this monkey-patching scheme
+
+    global register_option, get_option, set_option
+
+    def wrap(func: F) -> F:
+        def inner(key: str, *args, **kwds):
+            pkey = f"{prefix}.{key}"
+            return func(pkey, *args, **kwds)
+
+        return cast(F, inner)
+
+    _register_option = register_option
+    _get_option = get_option
+    _set_option = set_option
+    set_option = wrap(set_option)
+    get_option = wrap(get_option)
+    register_option = wrap(register_option)
+    try:
+        yield
+    finally:
+        set_option = _set_option
+        get_option = _get_option
+        register_option = _register_option
+
+
+# These factories and methods are handy for use as the validator
+# arg in register_option
+
+
+def is_type_factory(_type: type[Any]) -> Callable[[Any], None]:
+    """
+
+    Parameters
+    ----------
+    `_type` - a type to be compared against (e.g. type(x) == `_type`)
+
+    Returns
+    -------
+    validator - a function of a single argument x , which raises
+                ValueError if type(x) is not equal to `_type`
+
+    """
+
+    def inner(x) -> None:
+        if type(x) != _type:
+            raise ValueError(f"Value must have type '{_type}'")
+
+    return inner
+
+
+def is_instance_factory(_type: type | tuple[type, ...]) -> Callable[[Any], None]:
+    """
+
+    Parameters
+    ----------
+    `_type` - the type to be checked against
+
+    Returns
+    -------
+    validator - a function of a single argument x , which raises
+                ValueError if x is not an instance of `_type`
+
+    """
+    if isinstance(_type, tuple):
+        type_repr = "|".join(map(str, _type))
+    else:
+        type_repr = f"'{_type}'"
+
+    def inner(x) -> None:
+        if not isinstance(x, _type):
+            raise ValueError(f"Value must be an instance of {type_repr}")
+
+    return inner
+
+
+def is_one_of_factory(legal_values: Sequence) -> Callable[[Any], None]:
+    callables = [c for c in legal_values if callable(c)]
+    legal_values = [c for c in legal_values if not callable(c)]
+
+    def inner(x) -> None:
+        if x not in legal_values:
+            if not any(c(x) for c in callables):
+                uvals = [str(lval) for lval in legal_values]
+                pp_values = "|".join(uvals)
+                msg = f"Value must be one of {pp_values}"
+                if len(callables):
+                    msg += " or a callable"
+                raise ValueError(msg)
+
+    return inner
+
+
+def is_nonnegative_int(value: object) -> None:
+    """
+    Verify that value is None or a positive int.
+
+    Parameters
+    ----------
+    value : None or int
+            The `value` to be checked.
+
+    Raises
+    ------
+    ValueError
+        When the value is not None or is a negative integer
+    """
+    if value is None:
+        return
+
+    elif isinstance(value, int):
+        if value >= 0:
+            return
+
+    msg = "Value must be a nonnegative integer or None"
+    raise ValueError(msg)
+
+
+# common type validators, for convenience
+# usage: register_option(... , validator = is_int)
+is_int = is_type_factory(int)
+is_bool = is_type_factory(bool)
+is_float = is_type_factory(float)
+is_str = is_type_factory(str)
+is_text = is_instance_factory((str, bytes))
+
+
+def is_callable(obj: object) -> bool:
+    """
+
+    Parameters
+    ----------
+    `obj` - the object to be checked
+
+    Returns
+    -------
+    validator - returns True if object is callable
+        raises ValueError otherwise.
+
+    """
+    if not callable(obj):
+        raise ValueError("Value must be a callable")
+    return True
+
+
+# import set_module here would cause circular import
+get_option.__module__ = "pandas"
+set_option.__module__ = "pandas"
+describe_option.__module__ = "pandas"
+reset_option.__module__ = "pandas"
+option_context.__module__ = "pandas"

pandas/_config/dates.py

@@ -0,0 +1,26 @@
+"""
+config for datetime formatting
+"""
+
+from __future__ import annotations
+
+from pandas._config import config as cf
+
+pc_date_dayfirst_doc = """
+: boolean
+    When True, prints and parses dates with the day first, eg 20/01/2005
+"""
+
+pc_date_yearfirst_doc = """
+: boolean
+    When True, prints and parses dates with the year first, eg 2005/01/20
+"""
+
+with cf.config_prefix("display"):
+    # Needed upstream of `_libs` because these are used in tslibs.parsing
+    cf.register_option(
+        "date_dayfirst", False, pc_date_dayfirst_doc, validator=cf.is_bool
+    )
+    cf.register_option(
+        "date_yearfirst", False, pc_date_yearfirst_doc, validator=cf.is_bool
+    )

pandas/_config/display.py

@@ -0,0 +1,62 @@
+"""
+Unopinionated display configuration.
+"""
+
+from __future__ import annotations
+
+import locale
+import sys
+
+from pandas._config import config as cf
+
+# -----------------------------------------------------------------------------
+# Global formatting options
+_initial_defencoding: str | None = None
+
+
+def detect_console_encoding() -> str:
+    """
+    Try to find the most capable encoding supported by the console.
+    slightly modified from the way IPython handles the same issue.
+    """
+    global _initial_defencoding
+
+    encoding = None
+    try:
+        encoding = sys.stdout.encoding or sys.stdin.encoding
+    except (AttributeError, OSError):
+        pass
+
+    # try again for something better
+    if not encoding or "ascii" in encoding.lower():
+        try:
+            encoding = locale.getpreferredencoding()
+        except locale.Error:
+            # can be raised by locale.setlocale(), which is
+            #  called by getpreferredencoding
+            #  (on some systems, see stdlib locale docs)
+            pass
+
+    # when all else fails. this will usually be "ascii"
+    if not encoding or "ascii" in encoding.lower():
+        encoding = sys.getdefaultencoding()
+
+    # GH#3360, save the reported defencoding at import time
+    # MPL backends may change it. Make available for debugging.
+    if not _initial_defencoding:
+        _initial_defencoding = sys.getdefaultencoding()
+
+    return encoding
+
+
+pc_encoding_doc = """
+: str/unicode
+    Defaults to the detected encoding of the console.
+    Specifies the encoding to be used for strings returned by to_string,
+    these are generally strings meant to be displayed on the console.
+"""
+
+with cf.config_prefix("display"):
+    cf.register_option(
+        "encoding", detect_console_encoding(), pc_encoding_doc, validator=cf.is_text
+    )

pandas/_config/localization.py

@@ -0,0 +1,176 @@
+"""
+Helpers for configuring locale settings.
+
+Name `localization` is chosen to avoid overlap with builtin `locale` module.
+"""
+
+from __future__ import annotations
+
+from contextlib import contextmanager
+import locale
+import platform
+import re
+import subprocess
+from typing import (
+    TYPE_CHECKING,
+    cast,
+)
+
+from pandas._config.config import options
+
+if TYPE_CHECKING:
+    from collections.abc import Generator
+
+
+@contextmanager
+def set_locale(
+    new_locale: str | tuple[str, str], lc_var: int = locale.LC_ALL
+) -> Generator[str | tuple[str, str]]:
+    """
+    Context manager for temporarily setting a locale.
+
+    Parameters
+    ----------
+    new_locale : str or tuple
+        A string of the form <language_country>.<encoding>. For example to set
+        the current locale to US English with a UTF8 encoding, you would pass
+        "en_US.UTF-8".
+    lc_var : int, default `locale.LC_ALL`
+        The category of the locale being set.
+
+    Notes
+    -----
+    This is useful when you want to run a particular block of code under a
+    particular locale, without globally setting the locale. This probably isn't
+    thread-safe.
+    """
+    # getlocale is not always compliant with setlocale, use setlocale. GH#46595
+    current_locale = locale.setlocale(lc_var)
+
+    try:
+        locale.setlocale(lc_var, new_locale)
+        normalized_code, normalized_encoding = locale.getlocale()
+        if normalized_code is not None and normalized_encoding is not None:
+            yield f"{normalized_code}.{normalized_encoding}"
+        else:
+            yield new_locale
+    finally:
+        locale.setlocale(lc_var, current_locale)
+
+
+def can_set_locale(lc: str, lc_var: int = locale.LC_ALL) -> bool:
+    """
+    Check to see if we can set a locale, and subsequently get the locale,
+    without raising an Exception.
+
+    Parameters
+    ----------
+    lc : str
+        The locale to attempt to set.
+    lc_var : int, default `locale.LC_ALL`
+        The category of the locale being set.
+
+    Returns
+    -------
+    bool
+        Whether the passed locale can be set
+    """
+    try:
+        with set_locale(lc, lc_var=lc_var):
+            pass
+    except (ValueError, locale.Error):
+        # horrible name for an Exception subclass
+        return False
+    else:
+        return True
+
+
+def _valid_locales(locales: list[str] | str, normalize: bool) -> list[str]:
+    """
+    Return a list of normalized locales that do not throw an ``Exception``
+    when set.
+
+    Parameters
+    ----------
+    locales : str
+        A string where each locale is separated by a newline.
+    normalize : bool
+        Whether to call ``locale.normalize`` on each locale.
+
+    Returns
+    -------
+    valid_locales : list
+        A list of valid locales.
+    """
+    return [
+        loc
+        for loc in (
+            locale.normalize(loc.strip()) if normalize else loc.strip()
+            for loc in locales
+        )
+        if can_set_locale(loc)
+    ]
+
+
+def get_locales(
+    prefix: str | None = None,
+    normalize: bool = True,
+) -> list[str]:
+    """
+    Get all the locales that are available on the system.
+
+    Parameters
+    ----------
+    prefix : str
+        If not ``None`` then return only those locales with the prefix
+        provided. For example to get all English language locales (those that
+        start with ``"en"``), pass ``prefix="en"``.
+    normalize : bool
+        Call ``locale.normalize`` on the resulting list of available locales.
+        If ``True``, only locales that can be set without throwing an
+        ``Exception`` are returned.
+
+    Returns
+    -------
+    locales : list of strings
+        A list of locale strings that can be set with ``locale.setlocale()``.
+        For example::
+
+            locale.setlocale(locale.LC_ALL, locale_string)
+
+    On error will return an empty list (no locale available, e.g. Windows)
+
+    """
+    if platform.system() in ("Linux", "Darwin"):
+        raw_locales = subprocess.check_output(["locale", "-a"])
+    else:
+        # Other platforms e.g. windows platforms don't define "locale -a"
+        #  Note: is_platform_windows causes circular import here
+        return []
+
+    try:
+        # raw_locales is "\n" separated list of locales
+        # it may contain non-decodable parts, so split
+        # extract what we can and then rejoin.
+        split_raw_locales = raw_locales.split(b"\n")
+        out_locales = []
+        for x in split_raw_locales:
+            try:
+                out_locales.append(str(x, encoding=cast(str, options.display.encoding)))
+            except UnicodeError:
+                # 'locale -a' is used to populated 'raw_locales' and on
+                # Redhat 7 Linux (and maybe others) prints locale names
+                # using windows-1252 encoding.  Bug only triggered by
+                # a few special characters and when there is an
+                # extensive list of installed locales.
+                out_locales.append(str(x, encoding="windows-1252"))
+
+    except TypeError:
+        pass
+
+    if prefix is None:
+        return _valid_locales(out_locales, normalize)
+
+    pattern = re.compile(f"{prefix}.*")
+    found = pattern.findall("\n".join(out_locales))
+    return _valid_locales(found, normalize)

pandas/_libs/__init__.py

@@ -0,0 +1,27 @@
+__all__ = [
+    "Interval",
+    "NaT",
+    "NaTType",
+    "OutOfBoundsDatetime",
+    "Period",
+    "Timedelta",
+    "Timestamp",
+    "iNaT",
+]
+
+
+# Below imports needs to happen first to ensure pandas top level
+# module gets monkeypatched with the pandas_datetime_CAPI
+# see pandas_datetime_exec in pd_datetime.c
+import pandas._libs.pandas_parser  # isort: skip # type: ignore[reportUnusedImport]
+import pandas._libs.pandas_datetime  # noqa: F401 # isort: skip # type: ignore[reportUnusedImport]
+from pandas._libs.interval import Interval
+from pandas._libs.tslibs import (
+    NaT,
+    NaTType,
+    OutOfBoundsDatetime,
+    Period,
+    Timedelta,
+    Timestamp,
+    iNaT,
+)

pandas/_libs/algos.pyi

@@ -0,0 +1,443 @@
+from typing import Any
+
+import numpy as np
+
+from pandas._typing import npt
+
+class Infinity:
+    def __eq__(self, other) -> bool: ...
+    def __ne__(self, other) -> bool: ...
+    def __lt__(self, other) -> bool: ...
+    def __le__(self, other) -> bool: ...
+    def __gt__(self, other) -> bool: ...
+    def __ge__(self, other) -> bool: ...
+
+class NegInfinity:
+    def __eq__(self, other) -> bool: ...
+    def __ne__(self, other) -> bool: ...
+    def __lt__(self, other) -> bool: ...
+    def __le__(self, other) -> bool: ...
+    def __gt__(self, other) -> bool: ...
+    def __ge__(self, other) -> bool: ...
+
+def unique_deltas(
+    arr: np.ndarray,  # const int64_t[:]
+) -> np.ndarray: ...  # np.ndarray[np.int64, ndim=1]
+def is_lexsorted(list_of_arrays: list[npt.NDArray[np.int64]]) -> bool: ...
+def groupsort_indexer(
+    index: np.ndarray,  # const int64_t[:]
+    ngroups: int,
+) -> tuple[
+    np.ndarray,  # ndarray[int64_t, ndim=1]
+    np.ndarray,  # ndarray[int64_t, ndim=1]
+]: ...
+def kth_smallest(
+    arr: np.ndarray,  # numeric[:]
+    k: int,
+) -> Any: ...  # numeric
+
+# ----------------------------------------------------------------------
+# Pairwise correlation/covariance
+
+def nancorr(
+    mat: npt.NDArray[np.float64],  # const float64_t[:, :]
+    cov: bool = ...,
+    minp: int | None = ...,
+) -> npt.NDArray[np.float64]: ...  # ndarray[float64_t, ndim=2]
+def nancorr_spearman(
+    mat: npt.NDArray[np.float64],  # ndarray[float64_t, ndim=2]
+    minp: int = ...,
+) -> npt.NDArray[np.float64]: ...  # ndarray[float64_t, ndim=2]
+
+# ----------------------------------------------------------------------
+
+def validate_limit(nobs: int | None, limit=...) -> int: ...
+def get_fill_indexer(
+    mask: npt.NDArray[np.bool_],
+    limit: int | None = None,
+) -> npt.NDArray[np.intp]: ...
+def pad(
+    old: np.ndarray,  # ndarray[numeric_object_t]
+    new: np.ndarray,  # ndarray[numeric_object_t]
+    limit=...,
+) -> npt.NDArray[np.intp]: ...  # np.ndarray[np.intp, ndim=1]
+def pad_inplace(
+    values: np.ndarray,  # numeric_object_t[:]
+    mask: np.ndarray,  # uint8_t[:]
+    limit=...,
+) -> None: ...
+def pad_2d_inplace(
+    values: np.ndarray,  # numeric_object_t[:, :]
+    mask: np.ndarray,  # const uint8_t[:, :]
+    limit=...,
+) -> None: ...
+def backfill(
+    old: np.ndarray,  # ndarray[numeric_object_t]
+    new: np.ndarray,  # ndarray[numeric_object_t]
+    limit=...,
+) -> npt.NDArray[np.intp]: ...  # np.ndarray[np.intp, ndim=1]
+def backfill_inplace(
+    values: np.ndarray,  # numeric_object_t[:]
+    mask: np.ndarray,  # uint8_t[:]
+    limit=...,
+) -> None: ...
+def backfill_2d_inplace(
+    values: np.ndarray,  # numeric_object_t[:, :]
+    mask: np.ndarray,  # const uint8_t[:, :]
+    limit=...,
+) -> None: ...
+def is_monotonic(
+    arr: np.ndarray,  # ndarray[numeric_object_t, ndim=1]
+    timelike: bool,
+) -> tuple[bool, bool, bool]: ...
+
+# ----------------------------------------------------------------------
+# rank_1d, rank_2d
+# ----------------------------------------------------------------------
+
+def rank_1d(
+    values: np.ndarray,  # ndarray[numeric_object_t, ndim=1]
+    labels: np.ndarray | None = ...,  # const int64_t[:]=None
+    is_datetimelike: bool = ...,
+    ties_method=...,
+    ascending: bool = ...,
+    pct: bool = ...,
+    na_option=...,
+    mask: npt.NDArray[np.bool_] | None = ...,
+) -> np.ndarray: ...  # np.ndarray[float64_t, ndim=1]
+def rank_2d(
+    in_arr: np.ndarray,  # ndarray[numeric_object_t, ndim=2]
+    axis: int = ...,
+    is_datetimelike: bool = ...,
+    ties_method=...,
+    ascending: bool = ...,
+    na_option=...,
+    pct: bool = ...,
+) -> np.ndarray: ...  # np.ndarray[float64_t, ndim=1]
+def diff_2d(
+    arr: np.ndarray,  # ndarray[diff_t, ndim=2]
+    out: np.ndarray,  # ndarray[out_t, ndim=2]
+    periods: int,
+    axis: int,
+    datetimelike: bool = ...,
+) -> None: ...
+def ensure_platform_int(arr: object) -> npt.NDArray[np.intp]: ...
+def ensure_object(arr: object) -> npt.NDArray[np.object_]: ...
+def ensure_float64(arr: object) -> npt.NDArray[np.float64]: ...
+def ensure_int8(arr: object) -> npt.NDArray[np.int8]: ...
+def ensure_int16(arr: object) -> npt.NDArray[np.int16]: ...
+def ensure_int32(arr: object) -> npt.NDArray[np.int32]: ...
+def ensure_int64(arr: object) -> npt.NDArray[np.int64]: ...
+def ensure_uint64(arr: object) -> npt.NDArray[np.uint64]: ...
+def take_1d_int8_int8(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_int8_int32(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_int8_int64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_int8_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_int16_int16(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_int16_int32(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_int16_int64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_int16_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_int32_int32(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_int32_int64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_int32_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_int64_int64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_uint16_uint16(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_uint32_uint32(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_uint64_uint64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_int64_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_float32_float32(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_float32_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_float64_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_object_object(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_bool_bool(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_1d_bool_object(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_int8_int8(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_int8_int32(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_int8_int64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_int8_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_int16_int16(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_int16_int32(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_int16_int64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_int16_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_int32_int32(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_int32_int64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_int32_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_int64_int64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_int64_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_uint16_uint16(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_uint32_uint32(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_uint64_uint64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_float32_float32(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_float32_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_float64_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_object_object(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_bool_bool(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis0_bool_object(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_int8_int8(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_int8_int32(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_int8_int64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_int8_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_int16_int16(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_int16_int32(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_int16_int64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_int16_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_int32_int32(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_int32_int64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_int32_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_int64_int64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_uint16_uint16(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_uint32_uint32(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_uint64_uint64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_int64_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_float32_float32(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_float32_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_float64_float64(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_object_object(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_bool_bool(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_axis1_bool_object(
+    values: np.ndarray, indexer: npt.NDArray[np.intp], out: np.ndarray, fill_value=...
+) -> None: ...
+def take_2d_multi_int8_int8(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_int8_int32(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_int8_int64(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_int8_float64(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_int16_int16(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_int16_int32(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_int16_int64(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_int16_float64(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_int32_int32(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_int32_int64(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_int32_float64(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_int64_float64(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_float32_float32(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_float32_float64(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_float64_float64(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_object_object(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_bool_bool(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_bool_object(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...
+def take_2d_multi_int64_int64(
+    values: np.ndarray,
+    indexer: tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]],
+    out: np.ndarray,
+    fill_value=...,
+) -> None: ...

pandas/_libs/arrays.pyi

@@ -0,0 +1,40 @@
+from collections.abc import Sequence
+from typing import Self
+
+import numpy as np
+
+from pandas._typing import (
+    AxisInt,
+    DtypeObj,
+    Shape,
+)
+
+class NDArrayBacked:
+    _dtype: DtypeObj
+    _ndarray: np.ndarray
+    def __init__(self, values: np.ndarray, dtype: DtypeObj) -> None: ...
+    @classmethod
+    def _simple_new(cls, values: np.ndarray, dtype: DtypeObj) -> Self: ...
+    def _from_backing_data(self, values: np.ndarray) -> Self: ...
+    def __setstate__(self, state) -> None: ...
+    def __len__(self) -> int: ...
+    @property
+    def shape(self) -> Shape: ...
+    @property
+    def ndim(self) -> int: ...
+    @property
+    def size(self) -> int: ...
+    @property
+    def nbytes(self) -> int: ...
+    def copy(self, order=...) -> Self: ...
+    def delete(self, loc, axis=...) -> Self: ...
+    def swapaxes(self, axis1, axis2) -> Self: ...
+    def repeat(self, repeats: int | Sequence[int], axis: int | None = ...) -> Self: ...
+    def reshape(self, *args, **kwargs) -> Self: ...
+    def ravel(self, order=...) -> Self: ...
+    @property
+    def T(self) -> Self: ...
+    @classmethod
+    def _concat_same_type(
+        cls, to_concat: Sequence[Self], axis: AxisInt = ...
+    ) -> Self: ...

pandas/_libs/byteswap.pyi

@@ -0,0 +1,5 @@
+def read_float_with_byteswap(data: bytes, offset: int, byteswap: bool) -> float: ...
+def read_double_with_byteswap(data: bytes, offset: int, byteswap: bool) -> float: ...
+def read_uint16_with_byteswap(data: bytes, offset: int, byteswap: bool) -> int: ...
+def read_uint32_with_byteswap(data: bytes, offset: int, byteswap: bool) -> int: ...
+def read_uint64_with_byteswap(data: bytes, offset: int, byteswap: bool) -> int: ...

pandas/_libs/groupby.pyi

@@ -0,0 +1,234 @@
+from typing import Literal
+
+import numpy as np
+
+from pandas._typing import npt
+
+def group_median_float64(
+    out: np.ndarray,  # ndarray[float64_t, ndim=2]
+    counts: npt.NDArray[np.int64],
+    values: np.ndarray,  # ndarray[float64_t, ndim=2]
+    labels: npt.NDArray[np.int64],
+    min_count: int = ...,  # Py_ssize_t
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+    is_datetimelike: bool = ...,  # bint
+    skipna: bool = ...,
+) -> None: ...
+def group_cumprod(
+    out: np.ndarray,  # float64_t[:, ::1]
+    values: np.ndarray,  # const float64_t[:, :]
+    labels: np.ndarray,  # const int64_t[:]
+    ngroups: int,
+    is_datetimelike: bool,
+    skipna: bool = ...,
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+) -> None: ...
+def group_cumsum(
+    out: np.ndarray,  # int64float_t[:, ::1]
+    values: np.ndarray,  # ndarray[int64float_t, ndim=2]
+    labels: np.ndarray,  # const int64_t[:]
+    ngroups: int,
+    is_datetimelike: bool,
+    skipna: bool = ...,
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+) -> None: ...
+def group_shift_indexer(
+    out: np.ndarray,  # int64_t[::1]
+    labels: np.ndarray,  # const int64_t[:]
+    ngroups: int,
+    periods: int,
+) -> None: ...
+def group_fillna_indexer(
+    out: np.ndarray,  # ndarray[intp_t]
+    labels: np.ndarray,  # ndarray[int64_t]
+    mask: npt.NDArray[np.uint8],
+    limit: int,  # int64_t
+    compute_ffill: bool,
+    ngroups: int,
+) -> None: ...
+def group_any_all(
+    out: np.ndarray,  # uint8_t[::1]
+    values: np.ndarray,  # const uint8_t[::1]
+    labels: np.ndarray,  # const int64_t[:]
+    mask: np.ndarray,  # const uint8_t[::1]
+    val_test: Literal["any", "all"],
+    skipna: bool,
+    result_mask: np.ndarray | None,
+) -> None: ...
+def group_sum(
+    out: np.ndarray,  # complexfloatingintuint_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[complexfloatingintuint_t, ndim=2]
+    labels: np.ndarray,  # const intp_t[:]
+    mask: np.ndarray | None,
+    result_mask: np.ndarray | None = ...,
+    min_count: int = ...,
+    is_datetimelike: bool = ...,
+    initial: object = ...,
+    skipna: bool = ...,
+) -> None: ...
+def group_prod(
+    out: np.ndarray,  # int64float_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[int64float_t, ndim=2]
+    labels: np.ndarray,  # const intp_t[:]
+    mask: np.ndarray | None,
+    result_mask: np.ndarray | None = ...,
+    min_count: int = ...,
+    skipna: bool = ...,
+) -> None: ...
+def group_var(
+    out: np.ndarray,  # floating[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[floating, ndim=2]
+    labels: np.ndarray,  # const intp_t[:]
+    min_count: int = ...,  # Py_ssize_t
+    ddof: int = ...,  # int64_t
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+    is_datetimelike: bool = ...,
+    name: str = ...,
+    skipna: bool = ...,
+) -> None: ...
+def group_skew(
+    out: np.ndarray,  # float64_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[float64_T, ndim=2]
+    labels: np.ndarray,  # const intp_t[::1]
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+    skipna: bool = ...,
+) -> None: ...
+def group_kurt(
+    out: np.ndarray,  # float64_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[float64_T, ndim=2]
+    labels: np.ndarray,  # const intp_t[::1]
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+    skipna: bool = ...,
+) -> None: ...
+def group_mean(
+    out: np.ndarray,  # floating[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[floating, ndim=2]
+    labels: np.ndarray,  # const intp_t[:]
+    min_count: int = ...,  # Py_ssize_t
+    is_datetimelike: bool = ...,  # bint
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+    skipna: bool = ...,
+) -> None: ...
+def group_ohlc(
+    out: np.ndarray,  # floatingintuint_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[floatingintuint_t, ndim=2]
+    labels: np.ndarray,  # const intp_t[:]
+    min_count: int = ...,
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+) -> None: ...
+def group_quantile(
+    out: npt.NDArray[np.float64],
+    values: np.ndarray,  # ndarray[numeric, ndim=1]
+    labels: npt.NDArray[np.intp],
+    mask: npt.NDArray[np.uint8],
+    qs: npt.NDArray[np.float64],  # const
+    starts: npt.NDArray[np.int64],
+    ends: npt.NDArray[np.int64],
+    interpolation: Literal["linear", "lower", "higher", "nearest", "midpoint"],
+    result_mask: np.ndarray | None,
+    is_datetimelike: bool,
+) -> None: ...
+def group_last(
+    out: np.ndarray,  # rank_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[rank_t, ndim=2]
+    labels: np.ndarray,  # const int64_t[:]
+    mask: npt.NDArray[np.bool_] | None,
+    result_mask: npt.NDArray[np.bool_] | None = ...,
+    min_count: int = ...,  # Py_ssize_t
+    is_datetimelike: bool = ...,
+    skipna: bool = ...,
+) -> None: ...
+def group_nth(
+    out: np.ndarray,  # rank_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[rank_t, ndim=2]
+    labels: np.ndarray,  # const int64_t[:]
+    mask: npt.NDArray[np.bool_] | None,
+    result_mask: npt.NDArray[np.bool_] | None = ...,
+    min_count: int = ...,  # int64_t
+    rank: int = ...,  # int64_t
+    is_datetimelike: bool = ...,
+    skipna: bool = ...,
+) -> None: ...
+def group_rank(
+    out: np.ndarray,  # float64_t[:, ::1]
+    values: np.ndarray,  # ndarray[rank_t, ndim=2]
+    labels: np.ndarray,  # const int64_t[:]
+    ngroups: int,
+    is_datetimelike: bool,
+    ties_method: Literal["average", "min", "max", "first", "dense"] = ...,
+    ascending: bool = ...,
+    pct: bool = ...,
+    na_option: Literal["keep", "top", "bottom"] = ...,
+    mask: npt.NDArray[np.bool_] | None = ...,
+) -> None: ...
+def group_max(
+    out: np.ndarray,  # groupby_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[groupby_t, ndim=2]
+    labels: np.ndarray,  # const int64_t[:]
+    min_count: int = ...,
+    is_datetimelike: bool = ...,
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+    skipna: bool = ...,
+) -> None: ...
+def group_min(
+    out: np.ndarray,  # groupby_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[groupby_t, ndim=2]
+    labels: np.ndarray,  # const int64_t[:]
+    min_count: int = ...,
+    is_datetimelike: bool = ...,
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+    skipna: bool = ...,
+) -> None: ...
+def group_idxmin_idxmax(
+    out: npt.NDArray[np.intp],
+    counts: npt.NDArray[np.int64],
+    values: np.ndarray,  # ndarray[groupby_t, ndim=2]
+    labels: npt.NDArray[np.intp],
+    min_count: int = ...,
+    is_datetimelike: bool = ...,
+    mask: np.ndarray | None = ...,
+    name: str = ...,
+    skipna: bool = ...,
+    result_mask: np.ndarray | None = ...,
+) -> None: ...
+def group_cummin(
+    out: np.ndarray,  # groupby_t[:, ::1]
+    values: np.ndarray,  # ndarray[groupby_t, ndim=2]
+    labels: np.ndarray,  # const int64_t[:]
+    ngroups: int,
+    is_datetimelike: bool,
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+    skipna: bool = ...,
+) -> None: ...
+def group_cummax(
+    out: np.ndarray,  # groupby_t[:, ::1]
+    values: np.ndarray,  # ndarray[groupby_t, ndim=2]
+    labels: np.ndarray,  # const int64_t[:]
+    ngroups: int,
+    is_datetimelike: bool,
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+    skipna: bool = ...,
+) -> None: ...

pandas/_libs/hashing.pyi

@@ -0,0 +1,9 @@
+import numpy as np
+
+from pandas._typing import npt
+
+def hash_object_array(
+    arr: npt.NDArray[np.object_],
+    key: str,
+    encoding: str = ...,
+) -> npt.NDArray[np.uint64]: ...

pandas/_libs/hashtable.pyi

@@ -0,0 +1,274 @@
+from collections.abc import Hashable
+from typing import (
+    Any,
+    Literal,
+    overload,
+)
+
+import numpy as np
+
+from pandas._typing import npt
+
+def unique_label_indices(
+    labels: np.ndarray,  # const int64_t[:]
+) -> np.ndarray: ...
+
+class Factorizer:
+    count: int
+    uniques: Any
+    def __init__(self, size_hint: int, uses_mask: bool = False) -> None: ...
+    def get_count(self) -> int: ...
+    def factorize(
+        self,
+        values: np.ndarray,
+        na_sentinel=...,
+        na_value=...,
+        mask=...,
+    ) -> npt.NDArray[np.intp]: ...
+    def hash_inner_join(
+        self, values: np.ndarray, mask=...
+    ) -> tuple[np.ndarray, np.ndarray]: ...
+
+class ObjectFactorizer(Factorizer):
+    table: PyObjectHashTable
+    uniques: ObjectVector
+
+class Int64Factorizer(Factorizer):
+    table: Int64HashTable
+    uniques: Int64Vector
+
+class UInt64Factorizer(Factorizer):
+    table: UInt64HashTable
+    uniques: UInt64Vector
+
+class Int32Factorizer(Factorizer):
+    table: Int32HashTable
+    uniques: Int32Vector
+
+class UInt32Factorizer(Factorizer):
+    table: UInt32HashTable
+    uniques: UInt32Vector
+
+class Int16Factorizer(Factorizer):
+    table: Int16HashTable
+    uniques: Int16Vector
+
+class UInt16Factorizer(Factorizer):
+    table: UInt16HashTable
+    uniques: UInt16Vector
+
+class Int8Factorizer(Factorizer):
+    table: Int8HashTable
+    uniques: Int8Vector
+
+class UInt8Factorizer(Factorizer):
+    table: UInt8HashTable
+    uniques: UInt8Vector
+
+class Float64Factorizer(Factorizer):
+    table: Float64HashTable
+    uniques: Float64Vector
+
+class Float32Factorizer(Factorizer):
+    table: Float32HashTable
+    uniques: Float32Vector
+
+class Complex64Factorizer(Factorizer):
+    table: Complex64HashTable
+    uniques: Complex64Vector
+
+class Complex128Factorizer(Factorizer):
+    table: Complex128HashTable
+    uniques: Complex128Vector
+
+class Int64Vector:
+    def __init__(self, *args) -> None: ...
+    def __len__(self) -> int: ...
+    def to_array(self) -> npt.NDArray[np.int64]: ...
+
+class Int32Vector:
+    def __init__(self, *args) -> None: ...
+    def __len__(self) -> int: ...
+    def to_array(self) -> npt.NDArray[np.int32]: ...
+
+class Int16Vector:
+    def __init__(self, *args) -> None: ...
+    def __len__(self) -> int: ...
+    def to_array(self) -> npt.NDArray[np.int16]: ...
+
+class Int8Vector:
+    def __init__(self, *args) -> None: ...
+    def __len__(self) -> int: ...
+    def to_array(self) -> npt.NDArray[np.int8]: ...
+
+class UInt64Vector:
+    def __init__(self, *args) -> None: ...
+    def __len__(self) -> int: ...
+    def to_array(self) -> npt.NDArray[np.uint64]: ...
+
+class UInt32Vector:
+    def __init__(self, *args) -> None: ...
+    def __len__(self) -> int: ...
+    def to_array(self) -> npt.NDArray[np.uint32]: ...
+
+class UInt16Vector:
+    def __init__(self, *args) -> None: ...
+    def __len__(self) -> int: ...
+    def to_array(self) -> npt.NDArray[np.uint16]: ...
+
+class UInt8Vector:
+    def __init__(self, *args) -> None: ...
+    def __len__(self) -> int: ...
+    def to_array(self) -> npt.NDArray[np.uint8]: ...
+
+class Float64Vector:
+    def __init__(self, *args) -> None: ...
+    def __len__(self) -> int: ...
+    def to_array(self) -> npt.NDArray[np.float64]: ...
+
+class Float32Vector:
+    def __init__(self, *args) -> None: ...
+    def __len__(self) -> int: ...
+    def to_array(self) -> npt.NDArray[np.float32]: ...
+
+class Complex128Vector:
+    def __init__(self, *args) -> None: ...
+    def __len__(self) -> int: ...
+    def to_array(self) -> npt.NDArray[np.complex128]: ...
+
+class Complex64Vector:
+    def __init__(self, *args) -> None: ...
+    def __len__(self) -> int: ...
+    def to_array(self) -> npt.NDArray[np.complex64]: ...
+
+class StringVector:
+    def __init__(self, *args) -> None: ...
+    def __len__(self) -> int: ...
+    def to_array(self) -> npt.NDArray[np.object_]: ...
+
+class ObjectVector:
+    def __init__(self, *args) -> None: ...
+    def __len__(self) -> int: ...
+    def to_array(self) -> npt.NDArray[np.object_]: ...
+
+class HashTable:
+    # NB: The base HashTable class does _not_ actually have these methods;
+    #  we are putting them here for the sake of mypy to avoid
+    #  reproducing them in each subclass below.
+    def __init__(self, size_hint: int = ..., uses_mask: bool = ...) -> None: ...
+    def __len__(self) -> int: ...
+    def __contains__(self, key: Hashable) -> bool: ...
+    def sizeof(self, deep: bool = ...) -> int: ...
+    def get_state(self) -> dict[str, int]: ...
+    # TODO: `val/key` type is subclass-specific
+    def get_item(self, val): ...  # TODO: return type?
+    def set_item(self, key, val) -> None: ...
+    def get_na(self): ...  # TODO: return type?
+    def set_na(self, val) -> None: ...
+    def map_locations(
+        self,
+        values: np.ndarray,  # np.ndarray[subclass-specific]
+        mask: npt.NDArray[np.bool_] | None = ...,
+    ) -> None: ...
+    def lookup(
+        self,
+        values: np.ndarray,  # np.ndarray[subclass-specific]
+        mask: npt.NDArray[np.bool_] | None = ...,
+    ) -> npt.NDArray[np.intp]: ...
+    def get_labels(
+        self,
+        values: np.ndarray,  # np.ndarray[subclass-specific]
+        uniques,  # SubclassTypeVector
+        count_prior: int = ...,
+        na_sentinel: int = ...,
+        na_value: object = ...,
+        mask=...,
+    ) -> npt.NDArray[np.intp]: ...
+    @overload
+    def unique(
+        self,
+        values: np.ndarray,  # np.ndarray[subclass-specific]
+        *,
+        return_inverse: Literal[False] = ...,
+        mask: None = ...,
+    ) -> np.ndarray: ...  # np.ndarray[subclass-specific]
+    @overload
+    def unique(
+        self,
+        values: np.ndarray,  # np.ndarray[subclass-specific]
+        *,
+        return_inverse: Literal[True],
+        mask: None = ...,
+    ) -> tuple[np.ndarray, npt.NDArray[np.intp]]: ...  # np.ndarray[subclass-specific]
+    @overload
+    def unique(
+        self,
+        values: np.ndarray,  # np.ndarray[subclass-specific]
+        *,
+        return_inverse: Literal[False] = ...,
+        mask: npt.NDArray[np.bool_],
+    ) -> tuple[
+        np.ndarray,
+        npt.NDArray[np.bool_],
+    ]: ...  # np.ndarray[subclass-specific]
+    def factorize(
+        self,
+        values: np.ndarray,  # np.ndarray[subclass-specific]
+        na_sentinel: int = ...,
+        na_value: object = ...,
+        mask=...,
+        ignore_na: bool = True,
+    ) -> tuple[np.ndarray, npt.NDArray[np.intp]]: ...  # np.ndarray[subclass-specific]
+    def hash_inner_join(
+        self, values: np.ndarray, mask=...
+    ) -> tuple[np.ndarray, np.ndarray]: ...
+
+class Complex128HashTable(HashTable): ...
+class Complex64HashTable(HashTable): ...
+class Float64HashTable(HashTable): ...
+class Float32HashTable(HashTable): ...
+
+class Int64HashTable(HashTable):
+    # Only Int64HashTable has get_labels_groupby, map_keys_to_values
+    def get_labels_groupby(
+        self,
+        values: npt.NDArray[np.int64],  # const int64_t[:]
+    ) -> tuple[npt.NDArray[np.intp], npt.NDArray[np.int64]]: ...
+    def map_keys_to_values(
+        self,
+        keys: npt.NDArray[np.int64],
+        values: npt.NDArray[np.int64],  # const int64_t[:]
+    ) -> None: ...
+
+class Int32HashTable(HashTable): ...
+class Int16HashTable(HashTable): ...
+class Int8HashTable(HashTable): ...
+class UInt64HashTable(HashTable): ...
+class UInt32HashTable(HashTable): ...
+class UInt16HashTable(HashTable): ...
+class UInt8HashTable(HashTable): ...
+class StringHashTable(HashTable): ...
+class PyObjectHashTable(HashTable): ...
+class IntpHashTable(HashTable): ...
+
+def duplicated(
+    values: np.ndarray,
+    keep: Literal["last", "first", False] = ...,
+    mask: npt.NDArray[np.bool_] | None = ...,
+) -> npt.NDArray[np.bool_]: ...
+def mode(
+    values: np.ndarray, dropna: bool, mask: npt.NDArray[np.bool_] | None = ...
+) -> np.ndarray: ...
+def value_count(
+    values: np.ndarray,
+    dropna: bool,
+    mask: npt.NDArray[np.bool_] | None = ...,
+) -> tuple[np.ndarray, npt.NDArray[np.int64], int]: ...  # np.ndarray[same-as-values]
+
+# arr and values should have same dtype
+def ismember(
+    arr: np.ndarray,
+    values: np.ndarray,
+) -> npt.NDArray[np.bool_]: ...
+def object_hash(obj) -> int: ...
+def objects_are_equal(a, b) -> bool: ...

pandas/_libs/index.pyi

@@ -0,0 +1,107 @@
+import numpy as np
+
+from pandas._typing import npt
+
+from pandas import (
+    Index,
+    MultiIndex,
+)
+from pandas.core.arrays import ExtensionArray
+
+multiindex_nulls_shift: int
+
+class IndexEngine:
+    over_size_threshold: bool
+    def __init__(self, values: np.ndarray) -> None: ...
+    def __contains__(self, val: object) -> bool: ...
+
+    # -> int | slice | np.ndarray[bool]
+    def get_loc(self, val: object) -> int | slice | np.ndarray: ...
+    def sizeof(self, deep: bool = ...) -> int: ...
+    def __sizeof__(self) -> int: ...
+    @property
+    def is_unique(self) -> bool: ...
+    @property
+    def is_monotonic_increasing(self) -> bool: ...
+    @property
+    def is_monotonic_decreasing(self) -> bool: ...
+    @property
+    def is_mapping_populated(self) -> bool: ...
+    def clear_mapping(self): ...
+    def get_indexer(self, values: np.ndarray) -> npt.NDArray[np.intp]: ...
+    def get_indexer_non_unique(
+        self,
+        targets: np.ndarray,
+    ) -> tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]]: ...
+
+class MaskedIndexEngine(IndexEngine):
+    def __init__(self, values: object) -> None: ...
+    def get_indexer_non_unique(
+        self, targets: object
+    ) -> tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]]: ...
+
+class Float64Engine(IndexEngine): ...
+class Float32Engine(IndexEngine): ...
+class Complex128Engine(IndexEngine): ...
+class Complex64Engine(IndexEngine): ...
+class Int64Engine(IndexEngine): ...
+class Int32Engine(IndexEngine): ...
+class Int16Engine(IndexEngine): ...
+class Int8Engine(IndexEngine): ...
+class UInt64Engine(IndexEngine): ...
+class UInt32Engine(IndexEngine): ...
+class UInt16Engine(IndexEngine): ...
+class UInt8Engine(IndexEngine): ...
+class ObjectEngine(IndexEngine): ...
+class StringEngine(IndexEngine): ...
+class DatetimeEngine(Int64Engine): ...
+class TimedeltaEngine(DatetimeEngine): ...
+class PeriodEngine(Int64Engine): ...
+class BoolEngine(UInt8Engine): ...
+class MaskedFloat64Engine(MaskedIndexEngine): ...
+class MaskedFloat32Engine(MaskedIndexEngine): ...
+class MaskedComplex128Engine(MaskedIndexEngine): ...
+class MaskedComplex64Engine(MaskedIndexEngine): ...
+class MaskedInt64Engine(MaskedIndexEngine): ...
+class MaskedInt32Engine(MaskedIndexEngine): ...
+class MaskedInt16Engine(MaskedIndexEngine): ...
+class MaskedInt8Engine(MaskedIndexEngine): ...
+class MaskedUInt64Engine(MaskedIndexEngine): ...
+class MaskedUInt32Engine(MaskedIndexEngine): ...
+class MaskedUInt16Engine(MaskedIndexEngine): ...
+class MaskedUInt8Engine(MaskedIndexEngine): ...
+class MaskedBoolEngine(MaskedUInt8Engine): ...
+
+class StringObjectEngine(ObjectEngine):
+    def __init__(self, values: object, na_value) -> None: ...
+
+class BaseMultiIndexCodesEngine:
+    levels: list[np.ndarray]
+    offsets: np.ndarray  # np.ndarray[..., ndim=1]
+
+    def __init__(
+        self,
+        levels: list[Index],  # all entries hashable
+        labels: list[np.ndarray],  # all entries integer-dtyped
+        offsets: np.ndarray,  # np.ndarray[..., ndim=1]
+    ) -> None: ...
+    def get_indexer(self, target: npt.NDArray[np.object_]) -> npt.NDArray[np.intp]: ...
+    def _extract_level_codes(self, target: MultiIndex) -> np.ndarray: ...
+
+class ExtensionEngine:
+    def __init__(self, values: ExtensionArray) -> None: ...
+    def __contains__(self, val: object) -> bool: ...
+    def get_loc(self, val: object) -> int | slice | np.ndarray: ...
+    def get_indexer(self, values: np.ndarray) -> npt.NDArray[np.intp]: ...
+    def get_indexer_non_unique(
+        self,
+        targets: np.ndarray,
+    ) -> tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]]: ...
+    @property
+    def is_unique(self) -> bool: ...
+    @property
+    def is_monotonic_increasing(self) -> bool: ...
+    @property
+    def is_monotonic_decreasing(self) -> bool: ...
+    def sizeof(self, deep: bool = ...) -> int: ...
+    def clear_mapping(self): ...

pandas/_libs/indexing.pyi

@@ -0,0 +1,17 @@
+from typing import (
+    Generic,
+    TypeVar,
+)
+
+from pandas.core.indexing import IndexingMixin
+
+_IndexingMixinT = TypeVar("_IndexingMixinT", bound=IndexingMixin)
+
+class NDFrameIndexerBase(Generic[_IndexingMixinT]):
+    name: str
+    # in practice obj is either a DataFrame or a Series
+    obj: _IndexingMixinT
+
+    def __init__(self, name: str, obj: _IndexingMixinT) -> None: ...
+    @property
+    def ndim(self) -> int: ...

pandas/_libs/internals.pyi

@@ -0,0 +1,96 @@
+from collections.abc import (
+    Iterator,
+    Sequence,
+)
+from typing import (
+    Self,
+    final,
+    overload,
+)
+import weakref
+
+import numpy as np
+
+from pandas._typing import (
+    ArrayLike,
+    npt,
+)
+
+from pandas import Index
+from pandas.core.internals.blocks import Block as B
+
+def slice_len(slc: slice, objlen: int = ...) -> int: ...
+def get_concat_blkno_indexers(
+    blknos_list: list[npt.NDArray[np.intp]],
+) -> list[tuple[npt.NDArray[np.intp], BlockPlacement]]: ...
+def get_blkno_indexers(
+    blknos: np.ndarray,  # int64_t[:]
+    group: bool = ...,
+) -> list[tuple[int, slice | np.ndarray]]: ...
+def get_blkno_placements(
+    blknos: np.ndarray,
+    group: bool = ...,
+) -> Iterator[tuple[int, BlockPlacement]]: ...
+def update_blklocs_and_blknos(
+    blklocs: npt.NDArray[np.intp],
+    blknos: npt.NDArray[np.intp],
+    loc: int,
+    nblocks: int,
+) -> tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]]: ...
+@final
+class BlockPlacement:
+    def __init__(self, val: int | slice | np.ndarray) -> None: ...
+    @property
+    def indexer(self) -> np.ndarray | slice: ...
+    @property
+    def as_array(self) -> np.ndarray: ...
+    @property
+    def as_slice(self) -> slice: ...
+    @property
+    def is_slice_like(self) -> bool: ...
+    @overload
+    def __getitem__(
+        self, loc: slice | Sequence[int] | npt.NDArray[np.intp]
+    ) -> BlockPlacement: ...
+    @overload
+    def __getitem__(self, loc: int) -> int: ...
+    def __iter__(self) -> Iterator[int]: ...
+    def __len__(self) -> int: ...
+    def delete(self, loc) -> BlockPlacement: ...
+    def add(self, other) -> BlockPlacement: ...
+    def append(self, others: list[BlockPlacement]) -> BlockPlacement: ...
+    def tile_for_unstack(self, factor: int) -> npt.NDArray[np.intp]: ...
+
+class Block:
+    _mgr_locs: BlockPlacement
+    ndim: int
+    values: ArrayLike
+    refs: BlockValuesRefs
+    def __init__(
+        self,
+        values: ArrayLike,
+        placement: BlockPlacement,
+        ndim: int,
+        refs: BlockValuesRefs | None = ...,
+    ) -> None: ...
+    def slice_block_rows(self, slicer: slice) -> Self: ...
+
+class BlockManager:
+    blocks: tuple[B, ...]
+    axes: list[Index]
+    _known_consolidated: bool
+    _is_consolidated: bool
+    _blknos: np.ndarray
+    _blklocs: np.ndarray
+    def __init__(
+        self, blocks: tuple[B, ...], axes: list[Index], verify_integrity=...
+    ) -> None: ...
+    def get_slice(self, slobj: slice, axis: int = ...) -> Self: ...
+    def _rebuild_blknos_and_blklocs(self) -> None: ...
+
+class BlockValuesRefs:
+    referenced_blocks: list[weakref.ref]
+    def __init__(self, blk: Block | None = ...) -> None: ...
+    def add_reference(self, blk: Block) -> None: ...
+    def add_index_reference(self, index: Index) -> None: ...
+    def has_reference(self) -> bool: ...

pandas/_libs/interval.pyi

@@ -0,0 +1,174 @@
+from typing import (
+    Any,
+    Generic,
+    TypeVar,
+    overload,
+)
+
+import numpy as np
+import numpy.typing as npt
+
+from pandas._typing import (
+    IntervalClosedType,
+    Timedelta,
+    Timestamp,
+)
+
+VALID_CLOSED: frozenset[str]
+
+_OrderableScalarT = TypeVar("_OrderableScalarT", int, float)
+_OrderableTimesT = TypeVar("_OrderableTimesT", Timestamp, Timedelta)
+_OrderableT = TypeVar("_OrderableT", int, float, Timestamp, Timedelta)
+
+class _LengthDescriptor:
+    @overload
+    def __get__(
+        self, instance: Interval[_OrderableScalarT], owner: Any
+    ) -> _OrderableScalarT: ...
+    @overload
+    def __get__(
+        self, instance: Interval[_OrderableTimesT], owner: Any
+    ) -> Timedelta: ...
+
+class _MidDescriptor:
+    @overload
+    def __get__(self, instance: Interval[_OrderableScalarT], owner: Any) -> float: ...
+    @overload
+    def __get__(
+        self, instance: Interval[_OrderableTimesT], owner: Any
+    ) -> _OrderableTimesT: ...
+
+class IntervalMixin:
+    @property
+    def closed_left(self) -> bool: ...
+    @property
+    def closed_right(self) -> bool: ...
+    @property
+    def open_left(self) -> bool: ...
+    @property
+    def open_right(self) -> bool: ...
+    @property
+    def is_empty(self) -> bool: ...
+    def _check_closed_matches(self, other: IntervalMixin, name: str = ...) -> None: ...
+
+class Interval(IntervalMixin, Generic[_OrderableT]):
+    @property
+    def left(self: Interval[_OrderableT]) -> _OrderableT: ...
+    @property
+    def right(self: Interval[_OrderableT]) -> _OrderableT: ...
+    @property
+    def closed(self) -> IntervalClosedType: ...
+    mid: _MidDescriptor
+    length: _LengthDescriptor
+    def __init__(
+        self,
+        left: _OrderableT,
+        right: _OrderableT,
+        closed: IntervalClosedType = ...,
+    ) -> None: ...
+    def __hash__(self) -> int: ...
+    @overload
+    def __contains__(
+        self: Interval[Timedelta], key: Timedelta | Interval[Timedelta]
+    ) -> bool: ...
+    @overload
+    def __contains__(
+        self: Interval[Timestamp], key: Timestamp | Interval[Timestamp]
+    ) -> bool: ...
+    @overload
+    def __contains__(
+        self: Interval[_OrderableScalarT],
+        key: _OrderableScalarT | Interval[_OrderableScalarT],
+    ) -> bool: ...
+    @overload
+    def __add__(
+        self: Interval[_OrderableTimesT], y: Timedelta
+    ) -> Interval[_OrderableTimesT]: ...
+    @overload
+    def __add__(
+        self: Interval[int], y: _OrderableScalarT
+    ) -> Interval[_OrderableScalarT]: ...
+    @overload
+    def __add__(self: Interval[float], y: float) -> Interval[float]: ...
+    @overload
+    def __radd__(
+        self: Interval[_OrderableTimesT], y: Timedelta
+    ) -> Interval[_OrderableTimesT]: ...
+    @overload
+    def __radd__(
+        self: Interval[int], y: _OrderableScalarT
+    ) -> Interval[_OrderableScalarT]: ...
+    @overload
+    def __radd__(self: Interval[float], y: float) -> Interval[float]: ...
+    @overload
+    def __sub__(
+        self: Interval[_OrderableTimesT], y: Timedelta
+    ) -> Interval[_OrderableTimesT]: ...
+    @overload
+    def __sub__(
+        self: Interval[int], y: _OrderableScalarT
+    ) -> Interval[_OrderableScalarT]: ...
+    @overload
+    def __sub__(self: Interval[float], y: float) -> Interval[float]: ...
+    @overload
+    def __rsub__(
+        self: Interval[_OrderableTimesT], y: Timedelta
+    ) -> Interval[_OrderableTimesT]: ...
+    @overload
+    def __rsub__(
+        self: Interval[int], y: _OrderableScalarT
+    ) -> Interval[_OrderableScalarT]: ...
+    @overload
+    def __rsub__(self: Interval[float], y: float) -> Interval[float]: ...
+    @overload
+    def __mul__(
+        self: Interval[int], y: _OrderableScalarT
+    ) -> Interval[_OrderableScalarT]: ...
+    @overload
+    def __mul__(self: Interval[float], y: float) -> Interval[float]: ...
+    @overload
+    def __rmul__(
+        self: Interval[int], y: _OrderableScalarT
+    ) -> Interval[_OrderableScalarT]: ...
+    @overload
+    def __rmul__(self: Interval[float], y: float) -> Interval[float]: ...
+    @overload
+    def __truediv__(
+        self: Interval[int], y: _OrderableScalarT
+    ) -> Interval[_OrderableScalarT]: ...
+    @overload
+    def __truediv__(self: Interval[float], y: float) -> Interval[float]: ...
+    @overload
+    def __floordiv__(
+        self: Interval[int], y: _OrderableScalarT
+    ) -> Interval[_OrderableScalarT]: ...
+    @overload
+    def __floordiv__(self: Interval[float], y: float) -> Interval[float]: ...
+    def overlaps(self: Interval[_OrderableT], other: Interval[_OrderableT]) -> bool: ...
+
+def intervals_to_interval_bounds(
+    intervals: np.ndarray, validate_closed: bool = ...
+) -> tuple[np.ndarray, np.ndarray, IntervalClosedType]: ...
+
+class IntervalTree(IntervalMixin):
+    def __init__(
+        self,
+        left: np.ndarray,
+        right: np.ndarray,
+        closed: IntervalClosedType = ...,
+        leaf_size: int = ...,
+    ) -> None: ...
+    @property
+    def mid(self) -> np.ndarray: ...
+    @property
+    def length(self) -> np.ndarray: ...
+    def get_indexer(self, target) -> npt.NDArray[np.intp]: ...
+    def get_indexer_non_unique(
+        self, target
+    ) -> tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]]: ...
+    _na_count: int
+    @property
+    def is_overlapping(self) -> bool: ...
+    @property
+    def is_monotonic_increasing(self) -> bool: ...
+    def clear_mapping(self) -> None: ...

pandas/_libs/join.pyi

@@ -0,0 +1,79 @@
+import numpy as np
+
+from pandas._typing import npt
+
+def inner_join(
+    left: np.ndarray,  # const intp_t[:]
+    right: np.ndarray,  # const intp_t[:]
+    max_groups: int,
+    sort: bool = ...,
+) -> tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]]: ...
+def left_outer_join(
+    left: np.ndarray,  # const intp_t[:]
+    right: np.ndarray,  # const intp_t[:]
+    max_groups: int,
+    sort: bool = ...,
+) -> tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]]: ...
+def full_outer_join(
+    left: np.ndarray,  # const intp_t[:]
+    right: np.ndarray,  # const intp_t[:]
+    max_groups: int,
+) -> tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]]: ...
+def ffill_indexer(
+    indexer: np.ndarray,  # const intp_t[:]
+) -> npt.NDArray[np.intp]: ...
+def left_join_indexer_unique(
+    left: np.ndarray,  # ndarray[join_t]
+    right: np.ndarray,  # ndarray[join_t]
+) -> npt.NDArray[np.intp]: ...
+def left_join_indexer(
+    left: np.ndarray,  # ndarray[join_t]
+    right: np.ndarray,  # ndarray[join_t]
+) -> tuple[
+    np.ndarray,  # np.ndarray[join_t]
+    npt.NDArray[np.intp],
+    npt.NDArray[np.intp],
+]: ...
+def inner_join_indexer(
+    left: np.ndarray,  # ndarray[join_t]
+    right: np.ndarray,  # ndarray[join_t]
+) -> tuple[
+    np.ndarray,  # np.ndarray[join_t]
+    npt.NDArray[np.intp],
+    npt.NDArray[np.intp],
+]: ...
+def outer_join_indexer(
+    left: np.ndarray,  # ndarray[join_t]
+    right: np.ndarray,  # ndarray[join_t]
+) -> tuple[
+    np.ndarray,  # np.ndarray[join_t]
+    npt.NDArray[np.intp],
+    npt.NDArray[np.intp],
+]: ...
+def asof_join_backward_on_X_by_Y(
+    left_values: np.ndarray,  # ndarray[numeric_t]
+    right_values: np.ndarray,  # ndarray[numeric_t]
+    left_by_values: np.ndarray,  # const int64_t[:]
+    right_by_values: np.ndarray,  # const int64_t[:]
+    allow_exact_matches: bool = ...,
+    tolerance: np.number | float | None = ...,
+    use_hashtable: bool = ...,
+) -> tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]]: ...
+def asof_join_forward_on_X_by_Y(
+    left_values: np.ndarray,  # ndarray[numeric_t]
+    right_values: np.ndarray,  # ndarray[numeric_t]
+    left_by_values: np.ndarray,  # const int64_t[:]
+    right_by_values: np.ndarray,  # const int64_t[:]
+    allow_exact_matches: bool = ...,
+    tolerance: np.number | float | None = ...,
+    use_hashtable: bool = ...,
+) -> tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]]: ...
+def asof_join_nearest_on_X_by_Y(
+    left_values: np.ndarray,  # ndarray[numeric_t]
+    right_values: np.ndarray,  # ndarray[numeric_t]
+    left_by_values: np.ndarray,  # const int64_t[:]
+    right_by_values: np.ndarray,  # const int64_t[:]
+    allow_exact_matches: bool = ...,
+    tolerance: np.number | float | None = ...,
+    use_hashtable: bool = ...,
+) -> tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]]: ...

pandas/_libs/json.pyi

@@ -0,0 +1,23 @@
+from collections.abc import Callable
+from typing import (
+    Any,
+)
+
+def ujson_dumps(
+    obj: Any,
+    ensure_ascii: bool = ...,
+    double_precision: int = ...,
+    indent: int = ...,
+    orient: str = ...,
+    date_unit: str = ...,
+    iso_dates: bool = ...,
+    default_handler: None
+    | Callable[[Any], str | float | bool | list | dict | None] = ...,
+) -> str: ...
+def ujson_loads(
+    s: str,
+    precise_float: bool = ...,
+    numpy: bool = ...,
+    dtype: None = ...,
+    labelled: bool = ...,
+) -> Any: ...

pandas/_libs/lib.pyi

@@ -0,0 +1,238 @@
+# TODO(npdtypes): Many types specified here can be made more specific/accurate;
+#  the more specific versions are specified in comments
+from collections.abc import (
+    Callable,
+    Generator,
+    Hashable,
+)
+from decimal import Decimal
+from typing import (
+    Any,
+    Final,
+    Literal,
+    TypeAlias,
+    TypeGuard,
+    overload,
+)
+
+import numpy as np
+
+from pandas._typing import (
+    ArrayLike,
+    DtypeObj,
+    npt,
+)
+
+# placeholder until we can specify np.ndarray[object, ndim=2]
+ndarray_obj_2d = np.ndarray
+
+from enum import Enum
+
+class _NoDefault(Enum):
+    no_default = ...
+
+no_default: Final = _NoDefault.no_default
+NoDefault: TypeAlias = Literal[_NoDefault.no_default]
+
+i8max: int
+u8max: int
+
+def is_np_dtype(dtype: object, kinds: str | None = ...) -> TypeGuard[np.dtype]: ...
+def item_from_zerodim(val: object) -> object: ...
+def infer_dtype(value: object, skipna: bool = ...) -> str: ...
+def is_iterator(obj: object) -> bool: ...
+def is_scalar(val: object) -> bool: ...
+def is_list_like(obj: object, allow_sets: bool = ...) -> bool: ...
+def is_pyarrow_array(obj: object) -> bool: ...
+def is_decimal(obj: object) -> TypeGuard[Decimal]: ...
+def is_complex(obj: object) -> TypeGuard[complex]: ...
+def is_bool(obj: object) -> TypeGuard[bool | np.bool_]: ...
+def is_integer(obj: object) -> TypeGuard[int | np.integer]: ...
+def is_int_or_none(obj) -> bool: ...
+def is_float(obj: object) -> TypeGuard[float]: ...
+def is_interval_array(values: np.ndarray) -> bool: ...
+def is_datetime64_array(values: np.ndarray, skipna: bool = True) -> bool: ...
+def is_timedelta_or_timedelta64_array(
+    values: np.ndarray, skipna: bool = True
+) -> bool: ...
+def is_datetime_with_singletz_array(values: np.ndarray) -> bool: ...
+def is_time_array(values: np.ndarray, skipna: bool = ...): ...
+def is_date_array(values: np.ndarray, skipna: bool = ...): ...
+def is_datetime_array(values: np.ndarray, skipna: bool = ...): ...
+def is_string_array(values: np.ndarray, skipna: bool = ...): ...
+def is_float_array(values: np.ndarray, skipna: bool = ...): ...
+def is_integer_array(values: np.ndarray, skipna: bool = ...): ...
+def is_bool_array(values: np.ndarray, skipna: bool = ...): ...
+def fast_multiget(
+    mapping: dict,
+    keys: np.ndarray,  # object[:]
+    default=...,
+) -> ArrayLike: ...
+def fast_unique_multiple_list_gen(gen: Generator, sort: bool = ...) -> list: ...
+@overload
+def map_infer(
+    arr: np.ndarray,
+    f: Callable[[Any], Any],
+    *,
+    convert: Literal[False],
+    ignore_na: bool = ...,
+) -> np.ndarray: ...
+@overload
+def map_infer(
+    arr: np.ndarray,
+    f: Callable[[Any], Any],
+    *,
+    convert: bool = ...,
+    ignore_na: bool = ...,
+) -> ArrayLike: ...
+@overload
+def maybe_convert_objects(
+    objects: npt.NDArray[np.object_],
+    *,
+    try_float: bool = ...,
+    safe: bool = ...,
+    convert_numeric: bool = ...,
+    convert_non_numeric: Literal[False] = ...,
+    convert_to_nullable_dtype: Literal[False] = ...,
+    dtype_if_all_nat: DtypeObj | None = ...,
+) -> npt.NDArray[np.object_ | np.number]: ...
+@overload
+def maybe_convert_objects(
+    objects: npt.NDArray[np.object_],
+    *,
+    try_float: bool = ...,
+    safe: bool = ...,
+    convert_numeric: bool = ...,
+    convert_non_numeric: bool = ...,
+    convert_to_nullable_dtype: Literal[True] = ...,
+    dtype_if_all_nat: DtypeObj | None = ...,
+) -> ArrayLike: ...
+@overload
+def maybe_convert_objects(
+    objects: npt.NDArray[np.object_],
+    *,
+    try_float: bool = ...,
+    safe: bool = ...,
+    convert_numeric: bool = ...,
+    convert_non_numeric: bool = ...,
+    convert_to_nullable_dtype: bool = ...,
+    dtype_if_all_nat: DtypeObj | None = ...,
+) -> ArrayLike: ...
+@overload
+def maybe_convert_numeric(
+    values: npt.NDArray[np.object_],
+    na_values: set,
+    convert_empty: bool = ...,
+    coerce_numeric: bool = ...,
+    convert_to_masked_nullable: Literal[False] = ...,
+) -> tuple[np.ndarray, None]: ...
+@overload
+def maybe_convert_numeric(
+    values: npt.NDArray[np.object_],
+    na_values: set,
+    convert_empty: bool = ...,
+    coerce_numeric: bool = ...,
+    *,
+    convert_to_masked_nullable: Literal[True],
+) -> tuple[np.ndarray, np.ndarray]: ...
+
+# TODO: restrict `arr`?
+def ensure_string_array(
+    arr,
+    na_value: object = ...,
+    convert_na_value: bool = ...,
+    copy: bool = ...,
+    skipna: bool = ...,
+) -> npt.NDArray[np.object_]: ...
+def convert_nans_to_NA(
+    arr: npt.NDArray[np.object_],
+) -> npt.NDArray[np.object_]: ...
+def fast_zip(ndarrays: list) -> npt.NDArray[np.object_]: ...
+
+# TODO: can we be more specific about rows?
+def to_object_array_tuples(rows: object) -> ndarray_obj_2d: ...
+def tuples_to_object_array(
+    tuples: npt.NDArray[np.object_],
+) -> ndarray_obj_2d: ...
+
+# TODO: can we be more specific about rows?
+def to_object_array(rows: object, min_width: int = ...) -> ndarray_obj_2d: ...
+def dicts_to_array(dicts: list, columns: list) -> ndarray_obj_2d: ...
+def maybe_booleans_to_slice(
+    mask: npt.NDArray[np.uint8],
+) -> slice | npt.NDArray[np.uint8]: ...
+def maybe_indices_to_slice(
+    indices: npt.NDArray[np.intp],
+    max_len: int,
+) -> slice | npt.NDArray[np.intp]: ...
+def is_all_arraylike(obj: list) -> bool: ...
+
+# -----------------------------------------------------------------
+# Functions which in reality take memoryviews
+
+def memory_usage_of_objects(arr: np.ndarray) -> int: ...  # object[:]  # np.int64
+@overload
+def map_infer_mask(
+    arr: np.ndarray,
+    f: Callable[[Any], Any],
+    mask: np.ndarray,  # const uint8_t[:]
+    *,
+    convert: Literal[False],
+    na_value: Any = ...,
+    dtype: np.dtype = ...,
+) -> np.ndarray: ...
+@overload
+def map_infer_mask(
+    arr: np.ndarray,
+    f: Callable[[Any], Any],
+    mask: np.ndarray,  # const uint8_t[:]
+    *,
+    convert: bool = ...,
+    na_value: Any = ...,
+    dtype: np.dtype = ...,
+) -> ArrayLike: ...
+def indices_fast(
+    index: npt.NDArray[np.intp],
+    labels: np.ndarray,  # const int64_t[:]
+    keys: list,
+    sorted_labels: list[npt.NDArray[np.int64]],
+) -> dict[Hashable, npt.NDArray[np.intp]]: ...
+def generate_slices(
+    labels: np.ndarray,
+    ngroups: int,  # const intp_t[:]
+) -> tuple[npt.NDArray[np.int64], npt.NDArray[np.int64]]: ...
+def count_level_2d(
+    mask: np.ndarray,  # ndarray[uint8_t, ndim=2, cast=True],
+    labels: np.ndarray,  # const intp_t[:]
+    max_bin: int,
+) -> np.ndarray: ...  # np.ndarray[np.int64, ndim=2]
+def get_level_sorter(
+    codes: np.ndarray,  # const int64_t[:]
+    starts: np.ndarray,  # const intp_t[:]
+) -> np.ndarray: ...  # np.ndarray[np.intp, ndim=1]
+def generate_bins_dt64(
+    values: npt.NDArray[np.int64],
+    binner: np.ndarray,  # const int64_t[:]
+    closed: object = ...,
+    hasnans: bool = ...,
+) -> np.ndarray: ...  # np.ndarray[np.int64, ndim=1]
+def array_equivalent_object(
+    left: npt.NDArray[np.object_],
+    right: npt.NDArray[np.object_],
+) -> bool: ...
+def has_infs(arr: np.ndarray) -> bool: ...  # const floating[:]
+def has_only_ints_or_nan(arr: np.ndarray) -> bool: ...  # const floating[:]
+def get_reverse_indexer(
+    indexer: np.ndarray,  # const intp_t[:]
+    length: int,
+) -> npt.NDArray[np.intp]: ...
+def is_bool_list(obj: list) -> bool: ...
+def dtypes_all_equal(types: list[DtypeObj]) -> bool: ...
+def is_range_indexer(
+    left: np.ndarray,
+    n: int,  # np.ndarray[np.int64, ndim=1]
+) -> bool: ...
+def is_sequence_range(
+    sequence: np.ndarray,
+    step: int,  # np.ndarray[np.int64, ndim=1]
+) -> bool: ...

pandas/_libs/missing.pyi

@@ -0,0 +1,17 @@
+import numpy as np
+from numpy import typing as npt
+
+class NAType:
+    def __new__(cls, *args, **kwargs): ...
+
+NA: NAType
+
+def is_matching_na(
+    left: object, right: object, nan_matches_none: bool = ...
+) -> bool: ...
+def isposinf_scalar(val: object) -> bool: ...
+def isneginf_scalar(val: object) -> bool: ...
+def checknull(val: object) -> bool: ...
+def isnaobj(arr: np.ndarray) -> npt.NDArray[np.bool_]: ...
+def is_numeric_na(values: np.ndarray) -> npt.NDArray[np.bool_]: ...
+def is_pdna_or_none(values: np.ndarray) -> npt.NDArray[np.bool_]: ...

pandas/_libs/ops.pyi

@@ -0,0 +1,53 @@
+from collections.abc import (
+    Callable,
+    Iterable,
+)
+from typing import (
+    Any,
+    Literal,
+    TypeAlias,
+    overload,
+)
+
+import numpy as np
+
+from pandas._typing import npt
+
+_BinOp: TypeAlias = Callable[[Any, Any], Any]
+_BoolOp: TypeAlias = Callable[[Any, Any], bool]
+
+def scalar_compare(
+    values: np.ndarray,  # object[:]
+    val: object,
+    op: _BoolOp,  # {operator.eq, operator.ne, ...}
+) -> npt.NDArray[np.bool_]: ...
+def vec_compare(
+    left: npt.NDArray[np.object_],
+    right: npt.NDArray[np.object_],
+    op: _BoolOp,  # {operator.eq, operator.ne, ...}
+) -> npt.NDArray[np.bool_]: ...
+def scalar_binop(
+    values: np.ndarray,  # object[:]
+    val: object,
+    op: _BinOp,  # binary operator
+) -> np.ndarray: ...
+def vec_binop(
+    left: np.ndarray,  # object[:]
+    right: np.ndarray,  # object[:]
+    op: _BinOp,  # binary operator
+) -> np.ndarray: ...
+@overload
+def maybe_convert_bool(
+    arr: npt.NDArray[np.object_],
+    true_values: Iterable | None = None,
+    false_values: Iterable | None = None,
+    convert_to_masked_nullable: Literal[False] = ...,
+) -> tuple[np.ndarray, None]: ...
+@overload
+def maybe_convert_bool(
+    arr: npt.NDArray[np.object_],
+    true_values: Iterable = ...,
+    false_values: Iterable = ...,
+    *,
+    convert_to_masked_nullable: Literal[True],
+) -> tuple[np.ndarray, np.ndarray]: ...

pandas/_libs/ops_dispatch.pyi

@@ -0,0 +1,5 @@
+import numpy as np
+
+def maybe_dispatch_ufunc_to_dunder_op(
+    self, ufunc: np.ufunc, method: str, *inputs, **kwargs
+): ...

pandas/_libs/parsers.pyi

@@ -0,0 +1,77 @@
+from collections.abc import Hashable
+from typing import (
+    Literal,
+)
+
+import numpy as np
+
+from pandas._typing import (
+    ArrayLike,
+    Dtype,
+    npt,
+)
+
+STR_NA_VALUES: set[str]
+DEFAULT_BUFFER_HEURISTIC: int
+
+def sanitize_objects(
+    values: npt.NDArray[np.object_],
+    na_values: set,
+) -> int: ...
+
+class TextReader:
+    unnamed_cols: set[str]
+    table_width: int  # int64_t
+    leading_cols: int  # int64_t
+    header: list[list[int]]  # non-negative integers
+    def __init__(
+        self,
+        source,
+        delimiter: bytes | str = ...,  # single-character only
+        header=...,
+        header_start: int = ...,  # int64_t
+        header_end: int = ...,  # uint64_t
+        index_col=...,
+        names=...,
+        tokenize_chunksize: int = ...,  # int64_t
+        delim_whitespace: bool = ...,
+        converters=...,
+        skipinitialspace: bool = ...,
+        escapechar: bytes | str | None = ...,  # single-character only
+        doublequote: bool = ...,
+        quotechar: str | bytes | None = ...,  # at most 1 character
+        quoting: int = ...,
+        lineterminator: bytes | str | None = ...,  # at most 1 character
+        comment=...,
+        decimal: bytes | str = ...,  # single-character only
+        thousands: bytes | str | None = ...,  # single-character only
+        dtype: Dtype | dict[Hashable, Dtype] = ...,
+        usecols=...,
+        error_bad_lines: bool = ...,
+        warn_bad_lines: bool = ...,
+        na_filter: bool = ...,
+        na_values=...,
+        na_fvalues=...,
+        keep_default_na: bool = ...,
+        true_values=...,
+        false_values=...,
+        allow_leading_cols: bool = ...,
+        skiprows=...,
+        skipfooter: int = ...,  # int64_t
+        verbose: bool = ...,
+        float_precision: Literal["round_trip", "legacy", "high"] | None = ...,
+        skip_blank_lines: bool = ...,
+        encoding_errors: bytes | str = ...,
+    ) -> None: ...
+    def set_noconvert(self, i: int) -> None: ...
+    def remove_noconvert(self, i: int) -> None: ...
+    def close(self) -> None: ...
+    def read(self, rows: int | None = ...) -> dict[int, ArrayLike]: ...
+    def read_low_memory(self, rows: int | None) -> list[dict[int, ArrayLike]]: ...
+
+# _maybe_upcast, na_values are only exposed for testing
+na_values: dict
+
+def _maybe_upcast(
+    arr, use_dtype_backend: bool = ..., dtype_backend: str = ...
+) -> np.ndarray: ...

pandas/_libs/properties.pyi

@@ -0,0 +1,27 @@
+from collections.abc import Sequence
+from typing import (
+    overload,
+)
+
+from pandas._typing import (
+    AnyArrayLike,
+    DataFrame,
+    Index,
+    Series,
+)
+
+# note: this is a lie to make type checkers happy (they special
+# case property). cache_readonly uses attribute names similar to
+# property (fget) but it does not provide fset and fdel.
+cache_readonly = property
+
+class AxisProperty:
+    axis: int
+    def __init__(self, axis: int = ..., doc: str = ...) -> None: ...
+    @overload
+    def __get__(self, obj: DataFrame | Series, type) -> Index: ...
+    @overload
+    def __get__(self, obj: None, type) -> AxisProperty: ...
+    def __set__(
+        self, obj: DataFrame | Series, value: AnyArrayLike | Sequence
+    ) -> None: ...

pandas/_libs/reshape.pyi

@@ -0,0 +1,16 @@
+import numpy as np
+
+from pandas._typing import npt
+
+def unstack(
+    values: np.ndarray,  # reshape_t[:, :]
+    mask: np.ndarray,  # const uint8_t[:]
+    stride: int,
+    length: int,
+    width: int,
+    new_values: np.ndarray,  # reshape_t[:, :]
+    new_mask: np.ndarray,  # uint8_t[:, :]
+) -> None: ...
+def explode(
+    values: npt.NDArray[np.object_],
+) -> tuple[npt.NDArray[np.object_], npt.NDArray[np.int64]]: ...

pandas/_libs/sas.pyi

@@ -0,0 +1,7 @@
+from pandas.io.sas.sas7bdat import SAS7BDATReader
+
+class Parser:
+    def __init__(self, parser: SAS7BDATReader) -> None: ...
+    def read(self, nrows: int) -> None: ...
+
+def get_subheader_index(signature: bytes) -> int: ...

pandas/_libs/sparse.pyi

@@ -0,0 +1,51 @@
+from typing import Self
+
+import numpy as np
+
+from pandas._typing import (
+    TakeIndexer,
+    npt,
+)
+
+class SparseIndex:
+    length: int
+    npoints: int
+    def __init__(self) -> None: ...
+    @property
+    def ngaps(self) -> int: ...
+    @property
+    def nbytes(self) -> int: ...
+    @property
+    def indices(self) -> npt.NDArray[np.int32]: ...
+    def equals(self, other) -> bool: ...
+    def lookup(self, index: int) -> np.int32: ...
+    def lookup_array(self, indexer: npt.NDArray[np.int32]) -> npt.NDArray[np.int32]: ...
+    def to_int_index(self) -> IntIndex: ...
+    def to_block_index(self) -> BlockIndex: ...
+    def intersect(self, y_: SparseIndex) -> Self: ...
+    def make_union(self, y_: SparseIndex) -> Self: ...
+
+class IntIndex(SparseIndex):
+    indices: npt.NDArray[np.int32]
+    def __init__(
+        self, length: int, indices: TakeIndexer, check_integrity: bool = ...
+    ) -> None: ...
+
+class BlockIndex(SparseIndex):
+    nblocks: int
+    blocs: np.ndarray
+    blengths: np.ndarray
+    def __init__(
+        self, length: int, blocs: np.ndarray, blengths: np.ndarray
+    ) -> None: ...
+
+    # Override to have correct parameters
+    def intersect(self, other: SparseIndex) -> Self: ...
+    def make_union(self, y: SparseIndex) -> Self: ...
+
+def make_mask_object_ndarray(
+    arr: npt.NDArray[np.object_], fill_value
+) -> npt.NDArray[np.bool_]: ...
+def get_blocks(
+    indices: npt.NDArray[np.int32],
+) -> tuple[npt.NDArray[np.int32], npt.NDArray[np.int32]]: ...

pandas/_libs/testing.pyi

@@ -0,0 +1,14 @@
+from collections.abc import Mapping
+
+def assert_dict_equal(a: Mapping, b: Mapping, compare_keys: bool = ...) -> bool: ...
+def assert_almost_equal(
+    a,
+    b,
+    rtol: float = ...,
+    atol: float = ...,
+    check_dtype: bool = ...,
+    obj=...,
+    lobj=...,
+    robj=...,
+    index_values=...,
+) -> bool: ...

pandas/_libs/tslib.pyi

@@ -0,0 +1,33 @@
+from datetime import tzinfo
+
+import numpy as np
+
+from pandas._typing import npt
+
+def format_array_from_datetime(
+    values: npt.NDArray[np.int64],
+    tz: tzinfo | None = ...,
+    format: str | None = ...,
+    na_rep: str | float = ...,
+    reso: int = ...,  # NPY_DATETIMEUNIT
+) -> npt.NDArray[np.object_]: ...
+def first_non_null(values: np.ndarray) -> int: ...
+def array_to_datetime(
+    values: npt.NDArray[np.object_],
+    errors: str = ...,
+    dayfirst: bool = ...,
+    yearfirst: bool = ...,
+    utc: bool = ...,
+    creso: int = ...,
+    unit_for_numerics: str | None = ...,
+) -> tuple[np.ndarray, tzinfo | None]: ...
+
+# returned ndarray may be object dtype or datetime64[ns]
+
+def array_to_datetime_with_tz(
+    values: npt.NDArray[np.object_],
+    tz: tzinfo,
+    dayfirst: bool,
+    yearfirst: bool,
+    creso: int,
+) -> npt.NDArray[np.int64]: ...

pandas/_libs/tslibs/__init__.py

@@ -0,0 +1,89 @@
+__all__ = [
+    "BaseOffset",
+    "Day",
+    "IncompatibleFrequency",
+    "NaT",
+    "NaTType",
+    "OutOfBoundsDatetime",
+    "OutOfBoundsTimedelta",
+    "Period",
+    "Resolution",
+    "Tick",
+    "Timedelta",
+    "Timestamp",
+    "add_overflowsafe",
+    "astype_overflowsafe",
+    "delta_to_nanoseconds",
+    "dt64arr_to_periodarr",
+    "dtypes",
+    "get_resolution",
+    "get_supported_dtype",
+    "get_unit_from_dtype",
+    "guess_datetime_format",
+    "iNaT",
+    "ints_to_pydatetime",
+    "ints_to_pytimedelta",
+    "is_date_array_normalized",
+    "is_supported_dtype",
+    "is_unitless",
+    "localize_pydatetime",
+    "nat_strings",
+    "normalize_i8_timestamps",
+    "periods_per_day",
+    "periods_per_second",
+    "to_offset",
+    "tz_compare",
+    "tz_convert_from_utc",
+    "tz_convert_from_utc_single",
+]
+
+from pandas._libs.tslibs import dtypes
+from pandas._libs.tslibs.conversion import localize_pydatetime
+from pandas._libs.tslibs.dtypes import (
+    Resolution,
+    periods_per_day,
+    periods_per_second,
+)
+from pandas._libs.tslibs.nattype import (
+    NaT,
+    NaTType,
+    iNaT,
+    nat_strings,
+)
+from pandas._libs.tslibs.np_datetime import (
+    OutOfBoundsDatetime,
+    OutOfBoundsTimedelta,
+    add_overflowsafe,
+    astype_overflowsafe,
+    get_supported_dtype,
+    is_supported_dtype,
+    is_unitless,
+    py_get_unit_from_dtype as get_unit_from_dtype,
+)
+from pandas._libs.tslibs.offsets import (
+    BaseOffset,
+    Day,
+    Tick,
+    to_offset,
+)
+from pandas._libs.tslibs.parsing import guess_datetime_format
+from pandas._libs.tslibs.period import (
+    IncompatibleFrequency,
+    Period,
+)
+from pandas._libs.tslibs.timedeltas import (
+    Timedelta,
+    delta_to_nanoseconds,
+    ints_to_pytimedelta,
+)
+from pandas._libs.tslibs.timestamps import Timestamp
+from pandas._libs.tslibs.timezones import tz_compare
+from pandas._libs.tslibs.tzconversion import tz_convert_from_utc_single
+from pandas._libs.tslibs.vectorized import (
+    dt64arr_to_periodarr,
+    get_resolution,
+    ints_to_pydatetime,
+    is_date_array_normalized,
+    normalize_i8_timestamps,
+    tz_convert_from_utc,
+)