1 # cython: language_level=2
3 """A cleanup tool for HTML.
5 Removes unwanted tags and content. See the `Cleaner` class for
9 from __future__ import absolute_import
14 from urlparse import urlsplit
15 from urllib import unquote_plus
18 from urllib.parse import urlsplit, unquote_plus
19 from lxml import etree
20 from lxml.html import defs
21 from lxml.html import fromstring, XHTML_NAMESPACE
22 from lxml.html import xhtml_to_html, _transform_result
37 basestring = (str, bytes)
40 __all__ = ['clean_html', 'clean', 'Cleaner', 'autolink', 'autolink_html',
41 'word_break', 'word_break_html']
43 # Look at http://code.sixapart.com/trac/livejournal/browser/trunk/cgi-bin/cleanhtml.pl
44 # Particularly the CSS cleaning; most of the tag cleaning is integrated now
45 # I have multiple kinds of schemes searched; but should schemes be
46 # whitelisted instead?
48 # remove images? Also in CSS? background attribute?
49 # Some way to whitelist object, iframe, etc (e.g., if you want to
50 # allow *just* embedded YouTube movies)
51 # Log what was deleted and why?
52 # style="behavior: ..." might be bad in IE?
53 # Should we have something for just <meta http-equiv>? That's the worst of the
55 # UTF-7 detections? Example:
56 # <HEAD><META HTTP-EQUIV="CONTENT-TYPE" CONTENT="text/html; charset=UTF-7"> </HEAD>+ADw-SCRIPT+AD4-alert('XSS');+ADw-/SCRIPT+AD4-
57 # you don't always have to have the charset set, if the page has no charset
58 # and there's UTF7-like code in it.
59 # Look at these tests: http://htmlpurifier.org/live/smoketests/xssAttacks.php
62 # This is an IE-specific construct you can have in a stylesheet to
63 # run some Javascript:
64 _css_javascript_re = re.compile(
65 r'expression\s*\(.*?\)', re.S|re.I)
67 # Do I have to worry about @\nimport?
68 _css_import_re = re.compile(
71 # All kinds of schemes besides just javascript: that can cause
73 _is_image_dataurl = re.compile(
74 r'^data:image/.+;base64', re.I).search
75 _is_possibly_malicious_scheme = re.compile(
76 r'(?:javascript|jscript|livescript|vbscript|data|about|mocha):',
78 def _is_javascript_scheme(s):
79 if _is_image_dataurl(s):
81 return _is_possibly_malicious_scheme(s)
83 _substitute_whitespace = re.compile(r'[\s\x00-\x08\x0B\x0C\x0E-\x19]+').sub
84 # FIXME: should data: be blocked?
86 # FIXME: check against: http://msdn2.microsoft.com/en-us/library/ms537512.aspx
87 _conditional_comment_re = re.compile(
88 r'\[if[\s\n\r]+.*?][\s\n\r]*>', re.I|re.S)
90 _find_styled_elements = etree.XPath(
91 "descendant-or-self::*[@style]")
93 _find_external_links = etree.XPath(
94 ("descendant-or-self::a [normalize-space(@href) and substring(normalize-space(@href),1,1) != '#'] |"
95 "descendant-or-self::x:a[normalize-space(@href) and substring(normalize-space(@href),1,1) != '#']"),
96 namespaces={'x':XHTML_NAMESPACE})
99 class Cleaner(object):
101 Instances cleans the document of each of the possible offending
102 elements. The cleaning is controlled by attributes; you can
103 override attributes in a subclass, or set them in the constructor.
106 Removes any ``<script>`` tags.
109 Removes any Javascript, like an ``onclick`` attribute. Also removes stylesheets
110 as they could contain Javascript.
113 Removes any comments.
116 Removes any style tags.
119 Removes any style attributes. Defaults to the value of the ``style`` option.
122 Removes any ``<link>`` tags
125 Removes any ``<meta>`` tags
128 Structural parts of a page: ``<head>``, ``<html>``, ``<title>``.
130 ``processing_instructions``:
131 Removes any processing instructions.
134 Removes any embedded objects (flash, iframes)
137 Removes any frame-related tags
140 Removes any form tags
143 Tags that aren't *wrong*, but are annoying. ``<blink>`` and ``<marquee>``
146 A list of tags to remove. Only the tags will be removed,
147 their content will get pulled up into the parent tag.
150 A list of tags to kill. Killing also removes the tag's content,
151 i.e. the whole subtree, not just the tag itself.
154 A list of tags to include (default include all).
156 ``remove_unknown_tags``:
157 Remove any tags that aren't standard parts of HTML.
160 If true, only include 'safe' attributes (specifically the list
161 from the feedparser HTML sanitisation web site).
164 A set of attribute names to override the default list of attributes
165 considered 'safe' (when safe_attrs_only=True).
168 If true, then any <a> tags will have ``rel="nofollow"`` added to them.
171 A list or set of hosts that you can use for embedded content
172 (for content like ``<object>``, ``<link rel="stylesheet">``, etc).
173 You can also implement/override the method
174 ``allow_embedded_url(el, url)`` or ``allow_element(el)`` to
175 implement more complex rules for what can be embedded.
176 Anything that passes this test will be shown, regardless of
177 the value of (for instance) ``embedded``.
179 Note that this parameter might not work as intended if you do not
180 make the links absolute before doing the cleaning.
182 Note that you may also need to set ``whitelist_tags``.
185 A set of tags that can be included with ``host_whitelist``.
186 The default is ``iframe`` and ``embed``; you may wish to
187 include other tags like ``script``, or you may want to
188 implement ``allow_embedded_url`` for more control. Set to None to
191 This modifies the document *in place*.
201 page_structure = True
202 processing_instructions = True
210 remove_unknown_tags = True
211 safe_attrs_only = True
212 safe_attrs = defs.safe_attrs
215 whitelist_tags = {'iframe', 'embed'}
217 def __init__(self, **kw):
218 for name, value in kw.items():
219 if not hasattr(self, name):
221 "Unknown parameter: %s=%r" % (name, value))
222 setattr(self, name, value)
223 if self.inline_style is None and 'inline_style' not in kw:
224 self.inline_style = self.style
226 # Used to lookup the primary URL for a given tag that is up for
228 _tag_link_attrs = dict(
231 # From: http://java.sun.com/j2se/1.4.2/docs/guide/misc/applet.html
232 # From what I can tell, both attributes can contain a link:
233 applet=['code', 'object'],
237 # FIXME: there doesn't really seem like a general way to figure out what
238 # links an <object> tag uses; links often go in <param> tags with values
239 # that we don't really know. You'd have to have knowledge about specific
240 # kinds of plugins (probably keyed off classid), and match against those.
242 # FIXME: not looking at the action currently, because it is more complex
243 # than than -- if you keep the form, you should keep the form controls.
248 def __call__(self, doc):
252 if hasattr(doc, 'getroot'):
253 # ElementTree instance, instead of an element
255 # convert XHTML to HTML
257 # Normalize a case that IE treats <image> like <img>, and that
258 # can confuse either this step or later steps.
259 for el in doc.iter('image'):
261 if not self.comments:
262 # Of course, if we were going to kill comments anyway, we don't
263 # need to worry about this
264 self.kill_conditional_comments(doc)
266 kill_tags = set(self.kill_tags or ())
267 remove_tags = set(self.remove_tags or ())
268 allow_tags = set(self.allow_tags or ())
271 kill_tags.add('script')
272 if self.safe_attrs_only:
273 safe_attrs = set(self.safe_attrs)
274 for el in doc.iter(etree.Element):
276 for aname in attrib.keys():
277 if aname not in safe_attrs:
280 if not (self.safe_attrs_only and
281 self.safe_attrs == defs.safe_attrs):
282 # safe_attrs handles events attributes itself
283 for el in doc.iter(etree.Element):
285 for aname in attrib.keys():
286 if aname.startswith('on'):
288 doc.rewrite_links(self._remove_javascript_link,
289 resolve_base_href=False)
290 # If we're deleting style then we don't have to remove JS links
291 # from styles, otherwise...
292 if not self.inline_style:
293 for el in _find_styled_elements(doc):
294 old = el.get('style')
295 new = _css_javascript_re.sub('', old)
296 new = _css_import_re.sub('', new)
297 if self._has_sneaky_javascript(new):
298 # Something tricky is going on...
299 del el.attrib['style']
303 for el in list(doc.iter('style')):
304 if el.get('type', '').lower().strip() == 'text/javascript':
308 new = _css_javascript_re.sub('', old)
309 # The imported CSS can do anything; we just can't allow:
310 new = _css_import_re.sub('', old)
311 if self._has_sneaky_javascript(new):
312 # Something tricky is going on...
313 el.text = '/* deleted */'
316 if self.comments or self.processing_instructions:
317 # FIXME: why either? I feel like there's some obscure reason
318 # because you can put PIs in comments...? But I've already
320 kill_tags.add(etree.Comment)
321 if self.processing_instructions:
322 kill_tags.add(etree.ProcessingInstruction)
324 kill_tags.add('style')
325 if self.inline_style:
326 etree.strip_attributes(doc, 'style')
328 kill_tags.add('link')
329 elif self.style or self.javascript:
330 # We must get rid of included stylesheets if Javascript is not
331 # allowed, as you can put Javascript in them
332 for el in list(doc.iter('link')):
333 if 'stylesheet' in el.get('rel', '').lower():
334 # Note this kills alternate stylesheets as well
335 if not self.allow_element(el):
338 kill_tags.add('meta')
339 if self.page_structure:
340 remove_tags.update(('head', 'html', 'title'))
342 # FIXME: is <layer> really embedded?
343 # We should get rid of any <param> tags not inside <applet>;
344 # These are not really valid anyway.
345 for el in list(doc.iter('param')):
347 parent = el.getparent()
348 while parent is not None and parent.tag not in ('applet', 'object'):
349 parent = parent.getparent()
352 kill_tags.update(('applet',))
353 # The alternate contents that are in an iframe are a good fallback:
354 remove_tags.update(('iframe', 'embed', 'layer', 'object', 'param'))
356 # FIXME: ideally we should look at the frame links, but
357 # generally frames don't mix properly with an HTML
359 kill_tags.update(defs.frame_tags)
361 remove_tags.add('form')
362 kill_tags.update(('button', 'input', 'select', 'textarea'))
363 if self.annoying_tags:
364 remove_tags.update(('blink', 'marquee'))
368 for el in doc.iter():
369 if el.tag in kill_tags:
370 if self.allow_element(el):
373 elif el.tag in remove_tags:
374 if self.allow_element(el):
378 if _remove and _remove[0] == doc:
379 # We have to drop the parent-most tag, which we can't
380 # do. Instead we'll rewrite it:
384 elif _kill and _kill[0] == doc:
385 # We have to drop the parent-most element, which we can't
386 # do. Instead we'll clear it:
392 _kill.reverse() # start with innermost tags
398 if self.remove_unknown_tags:
401 "It does not make sense to pass in both allow_tags and remove_unknown_tags")
402 allow_tags = set(defs.tags)
405 for el in doc.iter():
406 if el.tag not in allow_tags:
415 if self.add_nofollow:
416 for el in _find_external_links(doc):
417 if not self.allow_follow(el):
420 if ('nofollow' in rel
421 and ' nofollow ' in (' %s ' % rel)):
423 rel = '%s nofollow' % rel
428 def allow_follow(self, anchor):
430 Override to suppress rel="nofollow" on some anchors.
434 def allow_element(self, el):
436 Decide whether an element is configured to be accepted or rejected.
438 :param el: an element.
439 :return: true to accept the element or false to reject/discard it.
441 if el.tag not in self._tag_link_attrs:
443 attr = self._tag_link_attrs[el.tag]
444 if isinstance(attr, (list, tuple)):
445 for one_attr in attr:
446 url = el.get(one_attr)
449 if not self.allow_embedded_url(el, url):
456 return self.allow_embedded_url(el, url)
458 def allow_embedded_url(self, el, url):
460 Decide whether a URL that was found in an element's attributes or text
461 if configured to be accepted or rejected.
463 :param el: an element.
464 :param url: a URL found on the element.
465 :return: true to accept the URL and false to reject it.
467 if self.whitelist_tags is not None and el.tag not in self.whitelist_tags:
469 scheme, netloc, path, query, fragment = urlsplit(url)
470 netloc = netloc.lower().split(':', 1)[0]
471 if scheme not in ('http', 'https'):
473 if netloc in self.host_whitelist:
477 def kill_conditional_comments(self, doc):
479 IE conditional comments basically embed HTML that the parser
480 doesn't normally see. We can't allow anything like that, so
481 we'll kill any comments that could be conditional.
485 doc, lambda el: _conditional_comment_re.search(el.text),
488 def _kill_elements(self, doc, condition, iterate=None):
490 for el in doc.iter(iterate):
496 def _remove_javascript_link(self, link):
497 # links like "j a v a s c r i p t:" might be interpreted in IE
498 new = _substitute_whitespace('', unquote_plus(link))
499 if _is_javascript_scheme(new):
500 # FIXME: should this be None to delete?
504 _substitute_comments = re.compile(r'/\*.*?\*/', re.S).sub
506 def _has_sneaky_javascript(self, style):
508 Depending on the browser, stuff like ``e x p r e s s i o n(...)``
509 can get interpreted, or ``expre/* stuff */ssion(...)``. This
510 checks for attempt to do stuff like this.
512 Typically the response will be to kill the entire style; if you
513 have just a bit of Javascript in the style another rule will catch
514 that and remove only the Javascript from the style; this catches
515 more sneaky attempts.
517 style = self._substitute_comments('', style)
518 style = style.replace('\\', '')
519 style = _substitute_whitespace('', style)
520 style = style.lower()
521 if 'javascript:' in style:
523 if 'expression(' in style:
527 def clean_html(self, html):
528 result_type = type(html)
529 if isinstance(html, basestring):
530 doc = fromstring(html)
532 doc = copy.deepcopy(html)
534 return _transform_result(result_type, doc)
537 clean_html = clean.clean_html
539 ############################################################
541 ############################################################
544 re.compile(r'(?P<body>https?://(?P<host>[a-z0-9._-]+)(?:/[/\-_.,a-z0-9%&?;=~]*)?(?:\([/\-_.,a-z0-9%&?;=~]*\))?)', re.I),
545 # This is conservative, but autolinking can be a bit conservative:
546 re.compile(r'mailto:(?P<body>[a-z0-9._-]+@(?P<host>[a-z0-9_.-]+[a-z]))', re.I),
549 _avoid_elements = ['textarea', 'pre', 'code', 'head', 'select', 'a']
552 re.compile(r'^localhost', re.I),
553 re.compile(r'\bexample\.(?:com|org|net)$', re.I),
554 re.compile(r'^127\.0\.0\.1$'),
557 _avoid_classes = ['nolink']
559 def autolink(el, link_regexes=_link_regexes,
560 avoid_elements=_avoid_elements,
561 avoid_hosts=_avoid_hosts,
562 avoid_classes=_avoid_classes):
564 Turn any URLs into links.
566 It will search for links identified by the given regular
567 expressions (by default mailto and http(s) links).
569 It won't link text in an element in avoid_elements, or an element
570 with a class in avoid_classes. It won't link to anything with a
571 host that matches one of the regular expressions in avoid_hosts
572 (default localhost and 127.0.0.1).
574 If you pass in an element, the element's tail will not be
575 substituted, only the contents of the element.
577 if el.tag in avoid_elements:
579 class_name = el.get('class')
581 class_name = class_name.split()
582 for match_class in avoid_classes:
583 if match_class in class_name:
585 for child in list(el):
586 autolink(child, link_regexes=link_regexes,
587 avoid_elements=avoid_elements,
588 avoid_hosts=avoid_hosts,
589 avoid_classes=avoid_classes)
591 text, tail_children = _link_text(
592 child.tail, link_regexes, avoid_hosts, factory=el.makeelement)
595 index = el.index(child)
596 el[index+1:index+1] = tail_children
598 text, pre_children = _link_text(
599 el.text, link_regexes, avoid_hosts, factory=el.makeelement)
602 el[:0] = pre_children
604 def _link_text(text, link_regexes, avoid_hosts, factory):
609 best_match, best_pos = None, None
610 for regex in link_regexes:
613 match = regex.search(text, pos=regex_pos)
616 host = match.group('host')
617 for host_regex in avoid_hosts:
618 if host_regex.search(host):
619 regex_pos = match.end()
625 if best_pos is None or match.start() < best_pos:
627 best_pos = match.start()
628 if best_match is None:
631 assert not links[-1].tail
632 links[-1].tail = text
634 assert not leading_text
637 link = best_match.group(0)
638 end = best_match.end()
639 if link.endswith('.') or link.endswith(','):
640 # These punctuation marks shouldn't end a link
643 prev_text = text[:best_match.start()]
645 assert not links[-1].tail
646 links[-1].tail = prev_text
648 assert not leading_text
649 leading_text = prev_text
650 anchor = factory('a')
651 anchor.set('href', link)
652 body = best_match.group('body')
655 if body.endswith('.') or body.endswith(','):
660 return leading_text, links
662 def autolink_html(html, *args, **kw):
663 result_type = type(html)
664 if isinstance(html, basestring):
665 doc = fromstring(html)
667 doc = copy.deepcopy(html)
668 autolink(doc, *args, **kw)
669 return _transform_result(result_type, doc)
671 autolink_html.__doc__ = autolink.__doc__
673 ############################################################
675 ############################################################
677 _avoid_word_break_elements = ['pre', 'textarea', 'code']
678 _avoid_word_break_classes = ['nobreak']
680 def word_break(el, max_width=40,
681 avoid_elements=_avoid_word_break_elements,
682 avoid_classes=_avoid_word_break_classes,
683 break_character=unichr(0x200b)):
685 Breaks any long words found in the body of the text (not attributes).
687 Doesn't effect any of the tags in avoid_elements, by default
688 ``<textarea>`` and ``<pre>``
690 Breaks words by inserting ​, which is a unicode character
691 for Zero Width Space character. This generally takes up no space
692 in rendering, but does copy as a space, and in monospace contexts
693 usually takes up space.
695 See http://www.cs.tut.fi/~jkorpela/html/nobr.html for a discussion
697 # Character suggestion of ​ comes from:
698 # http://www.cs.tut.fi/~jkorpela/html/nobr.html
699 if el.tag in _avoid_word_break_elements:
701 class_name = el.get('class')
704 class_name = class_name.split()
705 for avoid in avoid_classes:
706 if avoid in class_name:
712 el.text = _break_text(el.text, max_width, break_character)
714 word_break(child, max_width=max_width,
715 avoid_elements=avoid_elements,
716 avoid_classes=avoid_classes,
717 break_character=break_character)
719 child.tail = _break_text(child.tail, max_width, break_character)
721 def word_break_html(html, *args, **kw):
722 result_type = type(html)
723 doc = fromstring(html)
724 word_break(doc, *args, **kw)
725 return _transform_result(result_type, doc)
727 def _break_text(text, max_width, break_character):
730 if len(word) > max_width:
731 replacement = _insert_break(word, max_width, break_character)
732 text = text.replace(word, replacement)
735 _break_prefer_re = re.compile(r'[^a-z]', re.I)
737 def _insert_break(word, width, break_character):
740 while len(word) > width:
742 breaks = list(_break_prefer_re.finditer(start))
744 last_break = breaks[-1]
745 # Only walk back up to 10 characters to find a nice break:
746 if last_break.end() > width-10:
747 # FIXME: should the break character be at the end of the
748 # chunk, or the beginning of the next chunk?
749 start = word[:last_break.end()]
750 result += start + break_character
751 word = word[len(start):]