Hacked By AnonymousFox

Current Path : /proc/thread-self/root/proc/thread-self/root/proc/self/root/opt/alt/python37/lib/python3.7/site-packages/babel/
Upload File :
Current File : //proc/thread-self/root/proc/thread-self/root/proc/self/root/opt/alt/python37/lib/python3.7/site-packages/babel/util.py

# -*- coding: utf-8 -*-
"""
    babel.util
    ~~~~~~~~~~

    Various utility classes and functions.

    :copyright: (c) 2013-2021 by the Babel Team.
    :license: BSD, see LICENSE for more details.
"""

import codecs
import collections
from datetime import timedelta, tzinfo
import os
import re
import textwrap
from babel._compat import izip, imap
import pytz as _pytz
from babel import localtime

missing = object()


def distinct(iterable):
    """Yield all items in an iterable collection that are distinct.

    Unlike when using sets for a similar effect, the original ordering of the
    items in the collection is preserved by this function.

    >>> print(list(distinct([1, 2, 1, 3, 4, 4])))
    [1, 2, 3, 4]
    >>> print(list(distinct('foobar')))
    ['f', 'o', 'b', 'a', 'r']

    :param iterable: the iterable collection providing the data
    """
    seen = set()
    for item in iter(iterable):
        if item not in seen:
            yield item
            seen.add(item)

# Regexp to match python magic encoding line
PYTHON_MAGIC_COMMENT_re = re.compile(
    br'[ \t\f]* \# .* coding[=:][ \t]*([-\w.]+)', re.VERBOSE)


def parse_encoding(fp):
    """Deduce the encoding of a source file from magic comment.

    It does this in the same way as the `Python interpreter`__

    .. __: https://docs.python.org/3.4/reference/lexical_analysis.html#encoding-declarations

    The ``fp`` argument should be a seekable file object.

    (From Jeff Dairiki)
    """
    pos = fp.tell()
    fp.seek(0)
    try:
        line1 = fp.readline()
        has_bom = line1.startswith(codecs.BOM_UTF8)
        if has_bom:
            line1 = line1[len(codecs.BOM_UTF8):]

        m = PYTHON_MAGIC_COMMENT_re.match(line1)
        if not m:
            try:
                import ast
                ast.parse(line1.decode('latin-1'))
            except (ImportError, SyntaxError, UnicodeEncodeError):
                # Either it's a real syntax error, in which case the source is
                # not valid python source, or line2 is a continuation of line1,
                # in which case we don't want to scan line2 for a magic
                # comment.
                pass
            else:
                line2 = fp.readline()
                m = PYTHON_MAGIC_COMMENT_re.match(line2)

        if has_bom:
            if m:
                magic_comment_encoding = m.group(1).decode('latin-1')
                if magic_comment_encoding != 'utf-8':
                    raise SyntaxError(
                        'encoding problem: {0} with BOM'.format(
                            magic_comment_encoding))
            return 'utf-8'
        elif m:
            return m.group(1).decode('latin-1')
        else:
            return None
    finally:
        fp.seek(pos)


PYTHON_FUTURE_IMPORT_re = re.compile(
    r'from\s+__future__\s+import\s+\(*(.+)\)*')


def parse_future_flags(fp, encoding='latin-1'):
    """Parse the compiler flags by :mod:`__future__` from the given Python
    code.
    """
    import __future__
    pos = fp.tell()
    fp.seek(0)
    flags = 0
    try:
        body = fp.read().decode(encoding)

        # Fix up the source to be (hopefully) parsable by regexpen.
        # This will likely do untoward things if the source code itself is broken.

        # (1) Fix `import (\n...` to be `import (...`.
        body = re.sub(r'import\s*\([\r\n]+', 'import (', body)
        # (2) Join line-ending commas with the next line.
        body = re.sub(r',\s*[\r\n]+', ', ', body)
        # (3) Remove backslash line continuations.
        body = re.sub(r'\\\s*[\r\n]+', ' ', body)

        for m in PYTHON_FUTURE_IMPORT_re.finditer(body):
            names = [x.strip().strip('()') for x in m.group(1).split(',')]
            for name in names:
                feature = getattr(__future__, name, None)
                if feature:
                    flags |= feature.compiler_flag
    finally:
        fp.seek(pos)
    return flags


def pathmatch(pattern, filename):
    """Extended pathname pattern matching.

    This function is similar to what is provided by the ``fnmatch`` module in
    the Python standard library, but:

     * can match complete (relative or absolute) path names, and not just file
       names, and
     * also supports a convenience pattern ("**") to match files at any
       directory level.

    Examples:

    >>> pathmatch('**.py', 'bar.py')
    True
    >>> pathmatch('**.py', 'foo/bar/baz.py')
    True
    >>> pathmatch('**.py', 'templates/index.html')
    False

    >>> pathmatch('./foo/**.py', 'foo/bar/baz.py')
    True
    >>> pathmatch('./foo/**.py', 'bar/baz.py')
    False

    >>> pathmatch('^foo/**.py', 'foo/bar/baz.py')
    True
    >>> pathmatch('^foo/**.py', 'bar/baz.py')
    False

    >>> pathmatch('**/templates/*.html', 'templates/index.html')
    True
    >>> pathmatch('**/templates/*.html', 'templates/foo/bar.html')
    False

    :param pattern: the glob pattern
    :param filename: the path name of the file to match against
    """
    symbols = {
        '?': '[^/]',
        '?/': '[^/]/',
        '*': '[^/]+',
        '*/': '[^/]+/',
        '**/': '(?:.+/)*?',
        '**': '(?:.+/)*?[^/]+',
    }

    if pattern.startswith('^'):
        buf = ['^']
        pattern = pattern[1:]
    elif pattern.startswith('./'):
        buf = ['^']
        pattern = pattern[2:]
    else:
        buf = []

    for idx, part in enumerate(re.split('([?*]+/?)', pattern)):
        if idx % 2:
            buf.append(symbols[part])
        elif part:
            buf.append(re.escape(part))
    match = re.match(''.join(buf) + '$', filename.replace(os.sep, '/'))
    return match is not None


class TextWrapper(textwrap.TextWrapper):
    wordsep_re = re.compile(
        r'(\s+|'                                  # any whitespace
        r'(?<=[\w\!\"\'\&\.\,\?])-{2,}(?=\w))'    # em-dash
    )


def wraptext(text, width=70, initial_indent='', subsequent_indent=''):
    """Simple wrapper around the ``textwrap.wrap`` function in the standard
    library. This version does not wrap lines on hyphens in words.

    :param text: the text to wrap
    :param width: the maximum line width
    :param initial_indent: string that will be prepended to the first line of
                           wrapped output
    :param subsequent_indent: string that will be prepended to all lines save
                              the first of wrapped output
    """
    wrapper = TextWrapper(width=width, initial_indent=initial_indent,
                          subsequent_indent=subsequent_indent,
                          break_long_words=False)
    return wrapper.wrap(text)


# TODO (Babel 3.x): Remove this re-export
odict = collections.OrderedDict


class FixedOffsetTimezone(tzinfo):
    """Fixed offset in minutes east from UTC."""

    def __init__(self, offset, name=None):
        self._offset = timedelta(minutes=offset)
        if name is None:
            name = 'Etc/GMT%+d' % offset
        self.zone = name

    def __str__(self):
        return self.zone

    def __repr__(self):
        return '<FixedOffset "%s" %s>' % (self.zone, self._offset)

    def utcoffset(self, dt):
        return self._offset

    def tzname(self, dt):
        return self.zone

    def dst(self, dt):
        return ZERO


# Export the localtime functionality here because that's
# where it was in the past.
UTC = _pytz.utc
LOCALTZ = localtime.LOCALTZ
get_localzone = localtime.get_localzone

STDOFFSET = localtime.STDOFFSET
DSTOFFSET = localtime.DSTOFFSET
DSTDIFF = localtime.DSTDIFF
ZERO = localtime.ZERO

Hacked By AnonymousFox1.0, Coded By AnonymousFox