1 # -*- coding: utf-8 -*-
6 Jinja extensions allow to add custom tags similar to the way django custom
7 tags work. By default two example extensions exist: an i18n and a cache
10 :copyright: (c) 2010 by the Jinja Team.
13 from collections import deque
14 from jinja2 import nodes
15 from jinja2.defaults import *
16 from jinja2.environment import get_spontaneous_environment
17 from jinja2.runtime import Undefined, concat
18 from jinja2.exceptions import TemplateAssertionError, TemplateSyntaxError
19 from jinja2.utils import contextfunction, import_string, Markup, next
22 # the only real useful gettext functions for a Jinja template. Note
23 # that ugettext must be assigned to gettext as Jinja doesn't support
24 # non unicode strings.
25 GETTEXT_FUNCTIONS = ('_', 'gettext', 'ngettext')
28 class ExtensionRegistry(type):
29 """Gives the extension an unique identifier."""
31 def __new__(cls, name, bases, d):
32 rv = type.__new__(cls, name, bases, d)
33 rv.identifier = rv.__module__ + '.' + rv.__name__
37 class Extension(object):
38 """Extensions can be used to add extra functionality to the Jinja template
39 system at the parser level. Custom extensions are bound to an environment
40 but may not store environment specific data on `self`. The reason for
41 this is that an extension can be bound to another environment (for
42 overlays) by creating a copy and reassigning the `environment` attribute.
44 As extensions are created by the environment they cannot accept any
45 arguments for configuration. One may want to work around that by using
46 a factory function, but that is not possible as extensions are identified
47 by their import name. The correct way to configure the extension is
48 storing the configuration values on the environment. Because this way the
49 environment ends up acting as central configuration storage the
50 attributes may clash which is why extensions have to ensure that the names
51 they choose for configuration are not too generic. ``prefix`` for example
52 is a terrible name, ``fragment_cache_prefix`` on the other hand is a good
53 name as includes the name of the extension (fragment cache).
55 __metaclass__ = ExtensionRegistry
57 #: if this extension parses this is the list of tags it's listening to.
60 #: the priority of that extension. This is especially useful for
61 #: extensions that preprocess values. A lower value means higher
64 #: .. versionadded:: 2.4
67 def __init__(self, environment):
68 self.environment = environment
70 def bind(self, environment):
71 """Create a copy of this extension bound to another environment."""
72 rv = object.__new__(self.__class__)
73 rv.__dict__.update(self.__dict__)
74 rv.environment = environment
77 def preprocess(self, source, name, filename=None):
78 """This method is called before the actual lexing and can be used to
79 preprocess the source. The `filename` is optional. The return value
80 must be the preprocessed source.
84 def filter_stream(self, stream):
85 """It's passed a :class:`~jinja2.lexer.TokenStream` that can be used
86 to filter tokens returned. This method has to return an iterable of
87 :class:`~jinja2.lexer.Token`\s, but it doesn't have to return a
88 :class:`~jinja2.lexer.TokenStream`.
90 In the `ext` folder of the Jinja2 source distribution there is a file
91 called `inlinegettext.py` which implements a filter that utilizes this
96 def parse(self, parser):
97 """If any of the :attr:`tags` matched this method is called with the
98 parser as first argument. The token the parser stream is pointing at
99 is the name token that matched. This method has to return one or a
100 list of multiple nodes.
102 raise NotImplementedError()
104 def attr(self, name, lineno=None):
105 """Return an attribute node for the current extension. This is useful
106 to pass constants on extensions to generated template code::
108 self.attr('_my_attribute', lineno=lineno)
110 return nodes.ExtensionAttribute(self.identifier, name, lineno=lineno)
112 def call_method(self, name, args=None, kwargs=None, dyn_args=None,
113 dyn_kwargs=None, lineno=None):
114 """Call a method of the extension. This is a shortcut for
115 :meth:`attr` + :class:`jinja2.nodes.Call`.
121 return nodes.Call(self.attr(name, lineno=lineno), args, kwargs,
122 dyn_args, dyn_kwargs, lineno=lineno)
126 def _gettext_alias(__context, *args, **kwargs):
127 return __context.resolve('gettext')(*args, **kwargs)
130 def _make_new_gettext(func):
132 def gettext(__context, __string, **variables):
134 if __context.eval_ctx.autoescape:
136 return rv % variables
140 def _make_new_ngettext(func):
142 def ngettext(__context, __singular, __plural, num, **variables):
143 variables.setdefault('num', num)
144 rv = func(__singular, __plural, num)
145 if __context.eval_ctx.autoescape:
147 return rv % variables
151 class InternationalizationExtension(Extension):
152 """This extension adds gettext support to Jinja2."""
153 tags = set(['trans'])
155 # TODO: the i18n extension is currently reevaluating values in a few
156 # situations. Take this example:
157 # {% trans count=something() %}{{ count }} foo{% pluralize
158 # %}{{ count }} fooss{% endtrans %}
159 # something is called twice here. One time for the gettext value and
160 # the other time for the n-parameter of the ngettext function.
162 def __init__(self, environment):
163 Extension.__init__(self, environment)
164 environment.globals['_'] = _gettext_alias
166 install_gettext_translations=self._install,
167 install_null_translations=self._install_null,
168 install_gettext_callables=self._install_callables,
169 uninstall_gettext_translations=self._uninstall,
170 extract_translations=self._extract,
171 newstyle_gettext=False
174 def _install(self, translations, newstyle=None):
175 gettext = getattr(translations, 'ugettext', None)
177 gettext = translations.gettext
178 ngettext = getattr(translations, 'ungettext', None)
180 ngettext = translations.ngettext
181 self._install_callables(gettext, ngettext, newstyle)
183 def _install_null(self, newstyle=None):
184 self._install_callables(
186 lambda s, p, n: (n != 1 and (p,) or (s,))[0],
190 def _install_callables(self, gettext, ngettext, newstyle=None):
191 if newstyle is not None:
192 self.environment.newstyle_gettext = newstyle
193 if self.environment.newstyle_gettext:
194 gettext = _make_new_gettext(gettext)
195 ngettext = _make_new_ngettext(ngettext)
196 self.environment.globals.update(
201 def _uninstall(self, translations):
202 for key in 'gettext', 'ngettext':
203 self.environment.globals.pop(key, None)
205 def _extract(self, source, gettext_functions=GETTEXT_FUNCTIONS):
206 if isinstance(source, basestring):
207 source = self.environment.parse(source)
208 return extract_from_ast(source, gettext_functions)
210 def parse(self, parser):
211 """Parse a translatable tag."""
212 lineno = next(parser.stream).lineno
214 # find all the variables referenced. Additionally a variable can be
215 # defined in the body of the trans block too, but this is checked at
219 while parser.stream.current.type != 'block_end':
221 parser.stream.expect('comma')
223 # skip colon for python compatibility
224 if parser.stream.skip_if('colon'):
227 name = parser.stream.expect('name')
228 if name.value in variables:
229 parser.fail('translatable variable %r defined twice.' %
230 name.value, name.lineno,
231 exc=TemplateAssertionError)
234 if parser.stream.current.type == 'assign':
236 variables[name.value] = var = parser.parse_expression()
238 variables[name.value] = var = nodes.Name(name.value, 'load')
239 if plural_expr is None:
242 parser.stream.expect('block_end')
244 plural = plural_names = None
248 # now parse until endtrans or pluralize
249 singular_names, singular = self._parse_block(parser, True)
251 referenced.update(singular_names)
252 if plural_expr is None:
253 plural_expr = nodes.Name(singular_names[0], 'load')
255 # if we have a pluralize block, we parse that too
256 if parser.stream.current.test('name:pluralize'):
259 if parser.stream.current.type != 'block_end':
260 name = parser.stream.expect('name')
261 if name.value not in variables:
262 parser.fail('unknown variable %r for pluralization' %
263 name.value, name.lineno,
264 exc=TemplateAssertionError)
265 plural_expr = variables[name.value]
266 parser.stream.expect('block_end')
267 plural_names, plural = self._parse_block(parser, False)
269 referenced.update(plural_names)
273 # register free names as simple name expressions
274 for var in referenced:
275 if var not in variables:
276 variables[var] = nodes.Name(var, 'load')
278 # no variables referenced? no need to escape
280 singular = singular.replace('%%', '%')
282 plural = plural.replace('%%', '%')
286 elif plural_expr is None:
287 parser.fail('pluralize without variables', lineno)
290 variables = nodes.Dict([nodes.Pair(nodes.Const(x, lineno=lineno), y)
291 for x, y in variables.items()])
295 node = self._make_node(singular, plural, variables, plural_expr)
296 node.set_lineno(lineno)
299 def _parse_block(self, parser, allow_pluralize):
300 """Parse until the next block tag with a given name."""
304 if parser.stream.current.type == 'data':
305 buf.append(parser.stream.current.value.replace('%', '%%'))
307 elif parser.stream.current.type == 'variable_begin':
309 name = parser.stream.expect('name').value
310 referenced.append(name)
311 buf.append('%%(%s)s' % name)
312 parser.stream.expect('variable_end')
313 elif parser.stream.current.type == 'block_begin':
315 if parser.stream.current.test('name:endtrans'):
317 elif parser.stream.current.test('name:pluralize'):
320 parser.fail('a translatable section can have only one '
322 parser.fail('control structures in translatable sections are '
324 elif parser.stream.eos:
325 parser.fail('unclosed translation block')
327 assert False, 'internal parser error'
329 return referenced, concat(buf)
331 def _make_node(self, singular, plural, variables, plural_expr):
332 """Generates a useful node from the data provided."""
334 if plural_expr is None:
335 gettext = nodes.Name('gettext', 'load')
336 node = nodes.Call(gettext, [nodes.Const(singular)],
339 # singular and plural
341 ngettext = nodes.Name('ngettext', 'load')
342 node = nodes.Call(ngettext, [
343 nodes.Const(singular),
348 # in case newstyle gettext is used, the method is powerful
349 # enough to handle the variable expansion and autoescape
351 if self.environment.newstyle_gettext:
352 if variables is None:
353 variables = nodes.Dict([])
354 node.kwargs = variables
356 # otherwise do that here
358 # mark the return value as safe if we are in an
359 # environment with autoescaping turned on
360 node = nodes.MarkSafeIfAutoescape(node)
362 node = nodes.Mod(node, variables)
363 return nodes.Output([node])
366 class ExprStmtExtension(Extension):
367 """Adds a `do` tag to Jinja2 that works like the print statement just
368 that it doesn't print the return value.
372 def parse(self, parser):
373 node = nodes.ExprStmt(lineno=next(parser.stream).lineno)
374 node.node = parser.parse_tuple()
378 class LoopControlExtension(Extension):
379 """Adds break and continue to the template engine."""
380 tags = set(['break', 'continue'])
382 def parse(self, parser):
383 token = next(parser.stream)
384 if token.value == 'break':
385 return nodes.Break(lineno=token.lineno)
386 return nodes.Continue(lineno=token.lineno)
389 class WithExtension(Extension):
390 """Adds support for a django-like with block."""
393 def parse(self, parser):
394 node = nodes.Scope(lineno=next(parser.stream).lineno)
396 while parser.stream.current.type != 'block_end':
397 lineno = parser.stream.current.lineno
399 parser.stream.expect('comma')
400 target = parser.parse_assign_target()
401 parser.stream.expect('assign')
402 expr = parser.parse_expression()
403 assignments.append(nodes.Assign(target, expr, lineno=lineno))
404 node.body = assignments + \
405 list(parser.parse_statements(('name:endwith',),
410 class AutoEscapeExtension(Extension):
411 """Changes auto escape rules for a scope."""
412 tags = set(['autoescape'])
414 def parse(self, parser):
415 node = nodes.ScopedEvalContextModifier(lineno=next(parser.stream).lineno)
417 nodes.Keyword('autoescape', parser.parse_expression())
419 node.body = parser.parse_statements(('name:endautoescape',),
421 return nodes.Scope([node])
424 def extract_from_ast(node, gettext_functions=GETTEXT_FUNCTIONS,
426 """Extract localizable strings from the given template node. Per
427 default this function returns matches in babel style that means non string
428 parameters as well as keyword arguments are returned as `None`. This
429 allows Babel to figure out what you really meant if you are using
430 gettext functions that allow keyword arguments for placeholder expansion.
431 If you don't want that behavior set the `babel_style` parameter to `False`
432 which causes only strings to be returned and parameters are always stored
433 in tuples. As a consequence invalid gettext calls (calls without a single
434 string parameter or string parameters after non-string parameters) are
437 This example explains the behavior:
439 >>> from jinja2 import Environment
440 >>> env = Environment()
441 >>> node = env.parse('{{ (_("foo"), _(), ngettext("foo", "bar", 42)) }}')
442 >>> list(extract_from_ast(node))
443 [(1, '_', 'foo'), (1, '_', ()), (1, 'ngettext', ('foo', 'bar', None))]
444 >>> list(extract_from_ast(node, babel_style=False))
445 [(1, '_', ('foo',)), (1, 'ngettext', ('foo', 'bar'))]
447 For every string found this function yields a ``(lineno, function,
448 message)`` tuple, where:
450 * ``lineno`` is the number of the line on which the string was found,
451 * ``function`` is the name of the ``gettext`` function used (if the
452 string was extracted from embedded Python code), and
453 * ``message`` is the string itself (a ``unicode`` object, or a tuple
454 of ``unicode`` objects for functions with multiple string arguments).
456 This extraction function operates on the AST and is because of that unable
457 to extract any comments. For comment support you have to use the babel
458 extraction interface or extract comments yourself.
460 for node in node.find_all(nodes.Call):
461 if not isinstance(node.node, nodes.Name) or \
462 node.node.name not in gettext_functions:
466 for arg in node.args:
467 if isinstance(arg, nodes.Const) and \
468 isinstance(arg.value, basestring):
469 strings.append(arg.value)
473 for arg in node.kwargs:
475 if node.dyn_args is not None:
477 if node.dyn_kwargs is not None:
481 strings = tuple(x for x in strings if x is not None)
485 if len(strings) == 1:
488 strings = tuple(strings)
489 yield node.lineno, node.node.name, strings
492 class _CommentFinder(object):
493 """Helper class to find comments in a token stream. Can only
494 find comments for gettext calls forwards. Once the comment
495 from line 4 is found, a comment for line 1 will not return a
499 def __init__(self, tokens, comment_tags):
501 self.comment_tags = comment_tags
505 def find_backwards(self, offset):
507 for _, token_type, token_value in \
508 reversed(self.tokens[self.offset:offset]):
509 if token_type in ('comment', 'linecomment'):
511 prefix, comment = token_value.split(None, 1)
514 if prefix in self.comment_tags:
515 return [comment.rstrip()]
520 def find_comments(self, lineno):
521 if not self.comment_tags or self.last_lineno > lineno:
523 for idx, (token_lineno, _, _) in enumerate(self.tokens[self.offset:]):
524 if token_lineno > lineno:
525 return self.find_backwards(self.offset + idx)
526 return self.find_backwards(len(self.tokens))
529 def babel_extract(fileobj, keywords, comment_tags, options):
530 """Babel extraction method for Jinja templates.
532 .. versionchanged:: 2.3
533 Basic support for translation comments was added. If `comment_tags`
534 is now set to a list of keywords for extraction, the extractor will
535 try to find the best preceeding comment that begins with one of the
536 keywords. For best results, make sure to not have more than one
537 gettext call in one line of code and the matching comment in the
538 same line or the line before.
540 :param fileobj: the file-like object the messages should be extracted from
541 :param keywords: a list of keywords (i.e. function names) that should be
542 recognized as translation functions
543 :param comment_tags: a list of translator tags to search for and include
545 :param options: a dictionary of additional options (optional)
546 :return: an iterator over ``(lineno, funcname, message, comments)`` tuples.
547 (comments will be empty currently)
550 for extension in options.get('extensions', '').split(','):
551 extension = extension.strip()
554 extensions.add(import_string(extension))
555 if InternationalizationExtension not in extensions:
556 extensions.add(InternationalizationExtension)
558 environment = get_spontaneous_environment(
559 options.get('block_start_string', BLOCK_START_STRING),
560 options.get('block_end_string', BLOCK_END_STRING),
561 options.get('variable_start_string', VARIABLE_START_STRING),
562 options.get('variable_end_string', VARIABLE_END_STRING),
563 options.get('comment_start_string', COMMENT_START_STRING),
564 options.get('comment_end_string', COMMENT_END_STRING),
565 options.get('line_statement_prefix') or LINE_STATEMENT_PREFIX,
566 options.get('line_comment_prefix') or LINE_COMMENT_PREFIX,
567 str(options.get('trim_blocks', TRIM_BLOCKS)).lower() in \
568 ('1', 'on', 'yes', 'true'),
569 NEWLINE_SEQUENCE, frozenset(extensions),
570 # fill with defaults so that environments are shared
571 # with other spontaneus environments. The rest of the
572 # arguments are optimizer, undefined, finalize, autoescape,
573 # loader, cache size, auto reloading setting and the
575 True, Undefined, None, False, None, 0, False, None
578 source = fileobj.read().decode(options.get('encoding', 'utf-8'))
580 node = environment.parse(source)
581 tokens = list(environment.lex(environment.preprocess(source)))
582 except TemplateSyntaxError, e:
583 # skip templates with syntax errors
586 finder = _CommentFinder(tokens, comment_tags)
587 for lineno, func, message in extract_from_ast(node, keywords):
588 yield lineno, func, message, finder.find_comments(lineno)
591 #: nicer import names
592 i18n = InternationalizationExtension
593 do = ExprStmtExtension
594 loopcontrols = LoopControlExtension
595 with_ = WithExtension
596 autoescape = AutoEscapeExtension