Spaces:
Runtime error
Runtime error
| # -*- coding: utf-8 -*- | |
| """ | |
| babel.support | |
| ~~~~~~~~~~~~~ | |
| Several classes and functions that help with integrating and using Babel | |
| in applications. | |
| .. note: the code in this module is not used by Babel itself | |
| :copyright: (c) 2013-2021 by the Babel Team. | |
| :license: BSD, see LICENSE for more details. | |
| """ | |
| import gettext | |
| import locale | |
| from babel.core import Locale | |
| from babel.dates import format_date, format_datetime, format_time, \ | |
| format_timedelta | |
| from babel.numbers import format_number, format_decimal, format_currency, \ | |
| format_percent, format_scientific | |
| from babel._compat import PY2, text_type, text_to_native | |
| class Format(object): | |
| """Wrapper class providing the various date and number formatting functions | |
| bound to a specific locale and time-zone. | |
| >>> from babel.util import UTC | |
| >>> from datetime import date | |
| >>> fmt = Format('en_US', UTC) | |
| >>> fmt.date(date(2007, 4, 1)) | |
| u'Apr 1, 2007' | |
| >>> fmt.decimal(1.2345) | |
| u'1.234' | |
| """ | |
| def __init__(self, locale, tzinfo=None): | |
| """Initialize the formatter. | |
| :param locale: the locale identifier or `Locale` instance | |
| :param tzinfo: the time-zone info (a `tzinfo` instance or `None`) | |
| """ | |
| self.locale = Locale.parse(locale) | |
| self.tzinfo = tzinfo | |
| def date(self, date=None, format='medium'): | |
| """Return a date formatted according to the given pattern. | |
| >>> from datetime import date | |
| >>> fmt = Format('en_US') | |
| >>> fmt.date(date(2007, 4, 1)) | |
| u'Apr 1, 2007' | |
| """ | |
| return format_date(date, format, locale=self.locale) | |
| def datetime(self, datetime=None, format='medium'): | |
| """Return a date and time formatted according to the given pattern. | |
| >>> from datetime import datetime | |
| >>> from pytz import timezone | |
| >>> fmt = Format('en_US', tzinfo=timezone('US/Eastern')) | |
| >>> fmt.datetime(datetime(2007, 4, 1, 15, 30)) | |
| u'Apr 1, 2007, 11:30:00 AM' | |
| """ | |
| return format_datetime(datetime, format, tzinfo=self.tzinfo, | |
| locale=self.locale) | |
| def time(self, time=None, format='medium'): | |
| """Return a time formatted according to the given pattern. | |
| >>> from datetime import datetime | |
| >>> from pytz import timezone | |
| >>> fmt = Format('en_US', tzinfo=timezone('US/Eastern')) | |
| >>> fmt.time(datetime(2007, 4, 1, 15, 30)) | |
| u'11:30:00 AM' | |
| """ | |
| return format_time(time, format, tzinfo=self.tzinfo, locale=self.locale) | |
| def timedelta(self, delta, granularity='second', threshold=.85, | |
| format='long', add_direction=False): | |
| """Return a time delta according to the rules of the given locale. | |
| >>> from datetime import timedelta | |
| >>> fmt = Format('en_US') | |
| >>> fmt.timedelta(timedelta(weeks=11)) | |
| u'3 months' | |
| """ | |
| return format_timedelta(delta, granularity=granularity, | |
| threshold=threshold, | |
| format=format, add_direction=add_direction, | |
| locale=self.locale) | |
| def number(self, number): | |
| """Return an integer number formatted for the locale. | |
| >>> fmt = Format('en_US') | |
| >>> fmt.number(1099) | |
| u'1,099' | |
| """ | |
| return format_number(number, locale=self.locale) | |
| def decimal(self, number, format=None): | |
| """Return a decimal number formatted for the locale. | |
| >>> fmt = Format('en_US') | |
| >>> fmt.decimal(1.2345) | |
| u'1.234' | |
| """ | |
| return format_decimal(number, format, locale=self.locale) | |
| def currency(self, number, currency): | |
| """Return a number in the given currency formatted for the locale. | |
| """ | |
| return format_currency(number, currency, locale=self.locale) | |
| def percent(self, number, format=None): | |
| """Return a number formatted as percentage for the locale. | |
| >>> fmt = Format('en_US') | |
| >>> fmt.percent(0.34) | |
| u'34%' | |
| """ | |
| return format_percent(number, format, locale=self.locale) | |
| def scientific(self, number): | |
| """Return a number formatted using scientific notation for the locale. | |
| """ | |
| return format_scientific(number, locale=self.locale) | |
| class LazyProxy(object): | |
| """Class for proxy objects that delegate to a specified function to evaluate | |
| the actual object. | |
| >>> def greeting(name='world'): | |
| ... return 'Hello, %s!' % name | |
| >>> lazy_greeting = LazyProxy(greeting, name='Joe') | |
| >>> print(lazy_greeting) | |
| Hello, Joe! | |
| >>> u' ' + lazy_greeting | |
| u' Hello, Joe!' | |
| >>> u'(%s)' % lazy_greeting | |
| u'(Hello, Joe!)' | |
| This can be used, for example, to implement lazy translation functions that | |
| delay the actual translation until the string is actually used. The | |
| rationale for such behavior is that the locale of the user may not always | |
| be available. In web applications, you only know the locale when processing | |
| a request. | |
| The proxy implementation attempts to be as complete as possible, so that | |
| the lazy objects should mostly work as expected, for example for sorting: | |
| >>> greetings = [ | |
| ... LazyProxy(greeting, 'world'), | |
| ... LazyProxy(greeting, 'Joe'), | |
| ... LazyProxy(greeting, 'universe'), | |
| ... ] | |
| >>> greetings.sort() | |
| >>> for greeting in greetings: | |
| ... print(greeting) | |
| Hello, Joe! | |
| Hello, universe! | |
| Hello, world! | |
| """ | |
| __slots__ = ['_func', '_args', '_kwargs', '_value', '_is_cache_enabled', '_attribute_error'] | |
| def __init__(self, func, *args, **kwargs): | |
| is_cache_enabled = kwargs.pop('enable_cache', True) | |
| # Avoid triggering our own __setattr__ implementation | |
| object.__setattr__(self, '_func', func) | |
| object.__setattr__(self, '_args', args) | |
| object.__setattr__(self, '_kwargs', kwargs) | |
| object.__setattr__(self, '_is_cache_enabled', is_cache_enabled) | |
| object.__setattr__(self, '_value', None) | |
| object.__setattr__(self, '_attribute_error', None) | |
| def value(self): | |
| if self._value is None: | |
| try: | |
| value = self._func(*self._args, **self._kwargs) | |
| except AttributeError as error: | |
| object.__setattr__(self, '_attribute_error', error) | |
| raise | |
| if not self._is_cache_enabled: | |
| return value | |
| object.__setattr__(self, '_value', value) | |
| return self._value | |
| def __contains__(self, key): | |
| return key in self.value | |
| def __nonzero__(self): | |
| return bool(self.value) | |
| def __dir__(self): | |
| return dir(self.value) | |
| def __iter__(self): | |
| return iter(self.value) | |
| def __len__(self): | |
| return len(self.value) | |
| def __str__(self): | |
| return str(self.value) | |
| def __unicode__(self): | |
| return unicode(self.value) | |
| def __add__(self, other): | |
| return self.value + other | |
| def __radd__(self, other): | |
| return other + self.value | |
| def __mod__(self, other): | |
| return self.value % other | |
| def __rmod__(self, other): | |
| return other % self.value | |
| def __mul__(self, other): | |
| return self.value * other | |
| def __rmul__(self, other): | |
| return other * self.value | |
| def __call__(self, *args, **kwargs): | |
| return self.value(*args, **kwargs) | |
| def __lt__(self, other): | |
| return self.value < other | |
| def __le__(self, other): | |
| return self.value <= other | |
| def __eq__(self, other): | |
| return self.value == other | |
| def __ne__(self, other): | |
| return self.value != other | |
| def __gt__(self, other): | |
| return self.value > other | |
| def __ge__(self, other): | |
| return self.value >= other | |
| def __delattr__(self, name): | |
| delattr(self.value, name) | |
| def __getattr__(self, name): | |
| if self._attribute_error is not None: | |
| raise self._attribute_error | |
| return getattr(self.value, name) | |
| def __setattr__(self, name, value): | |
| setattr(self.value, name, value) | |
| def __delitem__(self, key): | |
| del self.value[key] | |
| def __getitem__(self, key): | |
| return self.value[key] | |
| def __setitem__(self, key, value): | |
| self.value[key] = value | |
| def __copy__(self): | |
| return LazyProxy( | |
| self._func, | |
| enable_cache=self._is_cache_enabled, | |
| *self._args, | |
| **self._kwargs | |
| ) | |
| def __deepcopy__(self, memo): | |
| from copy import deepcopy | |
| return LazyProxy( | |
| deepcopy(self._func, memo), | |
| enable_cache=deepcopy(self._is_cache_enabled, memo), | |
| *deepcopy(self._args, memo), | |
| **deepcopy(self._kwargs, memo) | |
| ) | |
| class NullTranslations(gettext.NullTranslations, object): | |
| DEFAULT_DOMAIN = None | |
| def __init__(self, fp=None): | |
| """Initialize a simple translations class which is not backed by a | |
| real catalog. Behaves similar to gettext.NullTranslations but also | |
| offers Babel's on *gettext methods (e.g. 'dgettext()'). | |
| :param fp: a file-like object (ignored in this class) | |
| """ | |
| # These attributes are set by gettext.NullTranslations when a catalog | |
| # is parsed (fp != None). Ensure that they are always present because | |
| # some *gettext methods (including '.gettext()') rely on the attributes. | |
| self._catalog = {} | |
| self.plural = lambda n: int(n != 1) | |
| super(NullTranslations, self).__init__(fp=fp) | |
| self.files = list(filter(None, [getattr(fp, 'name', None)])) | |
| self.domain = self.DEFAULT_DOMAIN | |
| self._domains = {} | |
| def dgettext(self, domain, message): | |
| """Like ``gettext()``, but look the message up in the specified | |
| domain. | |
| """ | |
| return self._domains.get(domain, self).gettext(message) | |
| def ldgettext(self, domain, message): | |
| """Like ``lgettext()``, but look the message up in the specified | |
| domain. | |
| """ | |
| return self._domains.get(domain, self).lgettext(message) | |
| def udgettext(self, domain, message): | |
| """Like ``ugettext()``, but look the message up in the specified | |
| domain. | |
| """ | |
| return self._domains.get(domain, self).ugettext(message) | |
| # backward compatibility with 0.9 | |
| dugettext = udgettext | |
| def dngettext(self, domain, singular, plural, num): | |
| """Like ``ngettext()``, but look the message up in the specified | |
| domain. | |
| """ | |
| return self._domains.get(domain, self).ngettext(singular, plural, num) | |
| def ldngettext(self, domain, singular, plural, num): | |
| """Like ``lngettext()``, but look the message up in the specified | |
| domain. | |
| """ | |
| return self._domains.get(domain, self).lngettext(singular, plural, num) | |
| def udngettext(self, domain, singular, plural, num): | |
| """Like ``ungettext()`` but look the message up in the specified | |
| domain. | |
| """ | |
| return self._domains.get(domain, self).ungettext(singular, plural, num) | |
| # backward compatibility with 0.9 | |
| dungettext = udngettext | |
| # Most of the downwards code, until it get's included in stdlib, from: | |
| # https://bugs.python.org/file10036/gettext-pgettext.patch | |
| # | |
| # The encoding of a msgctxt and a msgid in a .mo file is | |
| # msgctxt + "\x04" + msgid (gettext version >= 0.15) | |
| CONTEXT_ENCODING = '%s\x04%s' | |
| def pgettext(self, context, message): | |
| """Look up the `context` and `message` id in the catalog and return the | |
| corresponding message string, as an 8-bit string encoded with the | |
| catalog's charset encoding, if known. If there is no entry in the | |
| catalog for the `message` id and `context` , and a fallback has been | |
| set, the look up is forwarded to the fallback's ``pgettext()`` | |
| method. Otherwise, the `message` id is returned. | |
| """ | |
| ctxt_msg_id = self.CONTEXT_ENCODING % (context, message) | |
| missing = object() | |
| tmsg = self._catalog.get(ctxt_msg_id, missing) | |
| if tmsg is missing: | |
| if self._fallback: | |
| return self._fallback.pgettext(context, message) | |
| return message | |
| # Encode the Unicode tmsg back to an 8-bit string, if possible | |
| if self._output_charset: | |
| return text_to_native(tmsg, self._output_charset) | |
| elif self._charset: | |
| return text_to_native(tmsg, self._charset) | |
| return tmsg | |
| def lpgettext(self, context, message): | |
| """Equivalent to ``pgettext()``, but the translation is returned in the | |
| preferred system encoding, if no other encoding was explicitly set with | |
| ``bind_textdomain_codeset()``. | |
| """ | |
| ctxt_msg_id = self.CONTEXT_ENCODING % (context, message) | |
| missing = object() | |
| tmsg = self._catalog.get(ctxt_msg_id, missing) | |
| if tmsg is missing: | |
| if self._fallback: | |
| return self._fallback.lpgettext(context, message) | |
| return message | |
| if self._output_charset: | |
| return tmsg.encode(self._output_charset) | |
| return tmsg.encode(locale.getpreferredencoding()) | |
| def npgettext(self, context, singular, plural, num): | |
| """Do a plural-forms lookup of a message id. `singular` is used as the | |
| message id for purposes of lookup in the catalog, while `num` is used to | |
| determine which plural form to use. The returned message string is an | |
| 8-bit string encoded with the catalog's charset encoding, if known. | |
| If the message id for `context` is not found in the catalog, and a | |
| fallback is specified, the request is forwarded to the fallback's | |
| ``npgettext()`` method. Otherwise, when ``num`` is 1 ``singular`` is | |
| returned, and ``plural`` is returned in all other cases. | |
| """ | |
| ctxt_msg_id = self.CONTEXT_ENCODING % (context, singular) | |
| try: | |
| tmsg = self._catalog[(ctxt_msg_id, self.plural(num))] | |
| if self._output_charset: | |
| return text_to_native(tmsg, self._output_charset) | |
| elif self._charset: | |
| return text_to_native(tmsg, self._charset) | |
| return tmsg | |
| except KeyError: | |
| if self._fallback: | |
| return self._fallback.npgettext(context, singular, plural, num) | |
| if num == 1: | |
| return singular | |
| else: | |
| return plural | |
| def lnpgettext(self, context, singular, plural, num): | |
| """Equivalent to ``npgettext()``, but the translation is returned in the | |
| preferred system encoding, if no other encoding was explicitly set with | |
| ``bind_textdomain_codeset()``. | |
| """ | |
| ctxt_msg_id = self.CONTEXT_ENCODING % (context, singular) | |
| try: | |
| tmsg = self._catalog[(ctxt_msg_id, self.plural(num))] | |
| if self._output_charset: | |
| return tmsg.encode(self._output_charset) | |
| return tmsg.encode(locale.getpreferredencoding()) | |
| except KeyError: | |
| if self._fallback: | |
| return self._fallback.lnpgettext(context, singular, plural, num) | |
| if num == 1: | |
| return singular | |
| else: | |
| return plural | |
| def upgettext(self, context, message): | |
| """Look up the `context` and `message` id in the catalog and return the | |
| corresponding message string, as a Unicode string. If there is no entry | |
| in the catalog for the `message` id and `context`, and a fallback has | |
| been set, the look up is forwarded to the fallback's ``upgettext()`` | |
| method. Otherwise, the `message` id is returned. | |
| """ | |
| ctxt_message_id = self.CONTEXT_ENCODING % (context, message) | |
| missing = object() | |
| tmsg = self._catalog.get(ctxt_message_id, missing) | |
| if tmsg is missing: | |
| if self._fallback: | |
| return self._fallback.upgettext(context, message) | |
| return text_type(message) | |
| return tmsg | |
| def unpgettext(self, context, singular, plural, num): | |
| """Do a plural-forms lookup of a message id. `singular` is used as the | |
| message id for purposes of lookup in the catalog, while `num` is used to | |
| determine which plural form to use. The returned message string is a | |
| Unicode string. | |
| If the message id for `context` is not found in the catalog, and a | |
| fallback is specified, the request is forwarded to the fallback's | |
| ``unpgettext()`` method. Otherwise, when `num` is 1 `singular` is | |
| returned, and `plural` is returned in all other cases. | |
| """ | |
| ctxt_message_id = self.CONTEXT_ENCODING % (context, singular) | |
| try: | |
| tmsg = self._catalog[(ctxt_message_id, self.plural(num))] | |
| except KeyError: | |
| if self._fallback: | |
| return self._fallback.unpgettext(context, singular, plural, num) | |
| if num == 1: | |
| tmsg = text_type(singular) | |
| else: | |
| tmsg = text_type(plural) | |
| return tmsg | |
| def dpgettext(self, domain, context, message): | |
| """Like `pgettext()`, but look the message up in the specified | |
| `domain`. | |
| """ | |
| return self._domains.get(domain, self).pgettext(context, message) | |
| def udpgettext(self, domain, context, message): | |
| """Like `upgettext()`, but look the message up in the specified | |
| `domain`. | |
| """ | |
| return self._domains.get(domain, self).upgettext(context, message) | |
| # backward compatibility with 0.9 | |
| dupgettext = udpgettext | |
| def ldpgettext(self, domain, context, message): | |
| """Equivalent to ``dpgettext()``, but the translation is returned in the | |
| preferred system encoding, if no other encoding was explicitly set with | |
| ``bind_textdomain_codeset()``. | |
| """ | |
| return self._domains.get(domain, self).lpgettext(context, message) | |
| def dnpgettext(self, domain, context, singular, plural, num): | |
| """Like ``npgettext``, but look the message up in the specified | |
| `domain`. | |
| """ | |
| return self._domains.get(domain, self).npgettext(context, singular, | |
| plural, num) | |
| def udnpgettext(self, domain, context, singular, plural, num): | |
| """Like ``unpgettext``, but look the message up in the specified | |
| `domain`. | |
| """ | |
| return self._domains.get(domain, self).unpgettext(context, singular, | |
| plural, num) | |
| # backward compatibility with 0.9 | |
| dunpgettext = udnpgettext | |
| def ldnpgettext(self, domain, context, singular, plural, num): | |
| """Equivalent to ``dnpgettext()``, but the translation is returned in | |
| the preferred system encoding, if no other encoding was explicitly set | |
| with ``bind_textdomain_codeset()``. | |
| """ | |
| return self._domains.get(domain, self).lnpgettext(context, singular, | |
| plural, num) | |
| if not PY2: | |
| ugettext = gettext.NullTranslations.gettext | |
| ungettext = gettext.NullTranslations.ngettext | |
| class Translations(NullTranslations, gettext.GNUTranslations): | |
| """An extended translation catalog class.""" | |
| DEFAULT_DOMAIN = 'messages' | |
| def __init__(self, fp=None, domain=None): | |
| """Initialize the translations catalog. | |
| :param fp: the file-like object the translation should be read from | |
| :param domain: the message domain (default: 'messages') | |
| """ | |
| super(Translations, self).__init__(fp=fp) | |
| self.domain = domain or self.DEFAULT_DOMAIN | |
| if not PY2: | |
| ugettext = gettext.GNUTranslations.gettext | |
| ungettext = gettext.GNUTranslations.ngettext | |
| def load(cls, dirname=None, locales=None, domain=None): | |
| """Load translations from the given directory. | |
| :param dirname: the directory containing the ``MO`` files | |
| :param locales: the list of locales in order of preference (items in | |
| this list can be either `Locale` objects or locale | |
| strings) | |
| :param domain: the message domain (default: 'messages') | |
| """ | |
| if locales is not None: | |
| if not isinstance(locales, (list, tuple)): | |
| locales = [locales] | |
| locales = [str(locale) for locale in locales] | |
| if not domain: | |
| domain = cls.DEFAULT_DOMAIN | |
| filename = gettext.find(domain, dirname, locales) | |
| if not filename: | |
| return NullTranslations() | |
| with open(filename, 'rb') as fp: | |
| return cls(fp=fp, domain=domain) | |
| def __repr__(self): | |
| return '<%s: "%s">' % (type(self).__name__, | |
| self._info.get('project-id-version')) | |
| def add(self, translations, merge=True): | |
| """Add the given translations to the catalog. | |
| If the domain of the translations is different than that of the | |
| current catalog, they are added as a catalog that is only accessible | |
| by the various ``d*gettext`` functions. | |
| :param translations: the `Translations` instance with the messages to | |
| add | |
| :param merge: whether translations for message domains that have | |
| already been added should be merged with the existing | |
| translations | |
| """ | |
| domain = getattr(translations, 'domain', self.DEFAULT_DOMAIN) | |
| if merge and domain == self.domain: | |
| return self.merge(translations) | |
| existing = self._domains.get(domain) | |
| if merge and existing is not None: | |
| existing.merge(translations) | |
| else: | |
| translations.add_fallback(self) | |
| self._domains[domain] = translations | |
| return self | |
| def merge(self, translations): | |
| """Merge the given translations into the catalog. | |
| Message translations in the specified catalog override any messages | |
| with the same identifier in the existing catalog. | |
| :param translations: the `Translations` instance with the messages to | |
| merge | |
| """ | |
| if isinstance(translations, gettext.GNUTranslations): | |
| self._catalog.update(translations._catalog) | |
| if isinstance(translations, Translations): | |
| self.files.extend(translations.files) | |
| return self | |