3 * XHTML sanitizer for MediaWiki
5 * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
6 * http://www.mediawiki.org/
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 2 of the License, or
11 * (at your option) any later version.
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License along
19 * with this program; if not, write to the Free Software Foundation, Inc.,
20 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
21 * http://www.gnu.org/copyleft/gpl.html
28 * XHTML sanitizer for MediaWiki
33 * Regular expression to match various types of character references in
34 * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
36 const CHAR_REFS_REGEX
=
37 '/&([A-Za-z0-9\x80-\xff]+);
39 |&\#[xX]([0-9A-Fa-f]+);
43 * Blacklist for evil uris like javascript:
44 * WARNING: DO NOT use this in any place that actually requires blacklisting
45 * for security reasons. There are NUMEROUS[1] ways to bypass blacklisting, the
46 * only way to be secure from javascript: uri based xss vectors is to whitelist
47 * things that you know are safe and deny everything else.
48 * [1]: http://ha.ckers.org/xss.html
50 const EVIL_URI_PATTERN
= '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
51 const XMLNS_ATTRIBUTE_PATTERN
= "/^xmlns:[:A-Z_a-z-.0-9]+$/";
54 * List of all named character entities defined in HTML 4.01
55 * http://www.w3.org/TR/html4/sgml/entities.html
56 * As well as ' which is only defined starting in XHTML1.
59 static $htmlEntities = array(
75 'apos' => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
316 * Character entity aliases accepted by MediaWiki
318 static $htmlEntityAliases = array(
324 * Lazy-initialised attributes regex, see getAttribsRegex()
326 static $attribsRegex;
329 * Regular expression to match HTML/XML attribute pairs within a tag.
330 * Allows some... latitude.
331 * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
333 static function getAttribsRegex() {
334 if ( self
::$attribsRegex === null ) {
335 $attribFirst = '[:A-Z_a-z0-9]';
336 $attrib = '[:A-Z_a-z-.0-9]';
337 $space = '[\x09\x0a\x0d\x20]';
338 self
::$attribsRegex =
339 "/(?:^|$space)({$attribFirst}{$attrib}*)
342 # The attribute value: quoted or alone
345 | ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
346 | (\#[0-9a-fA-F]+) # Technically wrong, but lots of
347 # colors are specified like this.
348 # We'll be normalizing it.
352 return self
::$attribsRegex;
356 * Cleans up HTML, removes dangerous tags and attributes, and
357 * removes HTML comments
359 * @param $text String
360 * @param $processCallback Callback to do any variable or parameter replacements in HTML attribute values
361 * @param $args Array for the processing callback
362 * @param $extratags Array for any extra tags to include
363 * @param $removetags Array for any tags (default or extra) to exclude
366 static function removeHTMLtags( $text, $processCallback = null, $args = array(), $extratags = array(), $removetags = array() ) {
369 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
370 $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
372 wfProfileIn( __METHOD__
);
374 if ( !$staticInitialised ) {
376 $htmlpairsStatic = array( # Tags that must be closed
377 'b', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
378 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
379 'strike', 'strong', 'tt', 'var', 'div', 'center',
380 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
381 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'abbr', 'dfn',
382 'kbd', 'samp', 'thead', 'tbody', 'tfoot'
385 'br', 'hr', 'li', 'dt', 'dd'
387 $htmlsingleonly = array( # Elements that cannot have close tags
390 $htmlnest = array( # Tags that can be nested--??
391 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
392 'dl', 'font', 'big', 'small', 'sub', 'sup', 'span'
394 $tabletags = array( # Can only appear inside table, we will close them
397 $htmllist = array( # Tags used by list
400 $listtags = array( # Tags that can appear in a list
404 global $wgAllowImageTag;
405 if ( $wgAllowImageTag ) {
406 $htmlsingle[] = 'img';
407 $htmlsingleonly[] = 'img';
410 $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
411 $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
413 # Convert them all to hashtables for faster lookup
414 $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
415 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
416 foreach ( $vars as $var ) {
417 $
$var = array_flip( $
$var );
419 $staticInitialised = true;
421 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
422 $extratags = array_flip( $extratags );
423 $removetags = array_flip( $removetags );
424 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
425 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ) , $removetags );
427 # Remove HTML comments
428 $text = Sanitizer
::removeHTMLcomments( $text );
429 $bits = explode( '<', $text );
430 $text = str_replace( '>', '>', array_shift( $bits ) );
432 $tagstack = $tablestack = array();
433 foreach ( $bits as $x ) {
435 # $slash: Does the current element start with a '/'?
436 # $t: Current element name
437 # $params: String between element name and >
438 # $brace: Ending '>' or '/>'
439 # $rest: Everything until the next element of $bits
440 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
441 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
443 $slash = $t = $params = $brace = $rest = null;
447 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
449 if ( $slash && isset( $htmlsingleonly[$t] ) ) {
451 } elseif ( $slash ) {
452 # Closing a tag... is it the one we just opened?
453 $ot = @array_pop
( $tagstack );
455 if ( isset( $htmlsingleallowed[$ot] ) ) {
456 # Pop all elements with an optional close tag
457 # and see if we find a match below them
459 array_push( $optstack, $ot );
460 $ot = @array_pop
( $tagstack );
461 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
462 array_push( $optstack, $ot );
463 $ot = @array_pop
( $tagstack );
466 # No match. Push the optional elements back again
468 while ( $ot = @array_pop
( $optstack ) ) {
469 array_push( $tagstack, $ot );
473 @array_push
( $tagstack, $ot );
474 # <li> can be nested in <ul> or <ol>, skip those cases:
475 if ( !isset( $htmllist[$ot] ) ||
!isset( $listtags[$t] ) ) {
480 if ( $t == 'table' ) {
481 $tagstack = array_pop( $tablestack );
486 # Keep track for later
487 if ( isset( $tabletags[$t] ) &&
488 !in_array( 'table', $tagstack ) ) {
490 } elseif ( in_array( $t, $tagstack ) &&
491 !isset( $htmlnest [$t ] ) ) {
493 # Is it a self closed htmlpair ? (bug 5487)
494 } elseif ( $brace == '/>' &&
495 isset( $htmlpairs[$t] ) ) {
497 } elseif ( isset( $htmlsingleonly[$t] ) ) {
498 # Hack to force empty tag for uncloseable elements
500 } elseif ( isset( $htmlsingle[$t] ) ) {
501 # Hack to not close $htmlsingle tags
503 } elseif ( isset( $tabletags[$t] )
504 && in_array( $t, $tagstack ) ) {
505 // New table tag but forgot to close the previous one
508 if ( $t == 'table' ) {
509 array_push( $tablestack, $tagstack );
512 array_push( $tagstack, $t );
515 # Replace any variables or template parameters with
517 if( is_callable( $processCallback ) ) {
518 call_user_func_array( $processCallback, array( &$params, $args ) );
521 # Strip non-approved attributes from the tag
522 $newparams = Sanitizer
::fixTagAttributes( $params, $t );
525 $rest = str_replace( '>', '>', $rest );
526 $close = ( $brace == '/>' && !$slash ) ?
' /' : '';
527 $text .= "<$slash$t$newparams$close>$rest";
531 $text .= '<' . str_replace( '>', '>', $x);
533 # Close off any remaining tags
534 while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
536 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
539 # this might be possible using tidy itself
540 foreach ( $bits as $x ) {
541 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
543 @list
( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
544 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
545 if( is_callable( $processCallback ) ) {
546 call_user_func_array( $processCallback, array( &$params, $args ) );
548 $newparams = Sanitizer
::fixTagAttributes( $params, $t );
549 $rest = str_replace( '>', '>', $rest );
550 $text .= "<$slash$t$newparams$brace$rest";
552 $text .= '<' . str_replace( '>', '>', $x);
556 wfProfileOut( __METHOD__
);
561 * Remove '<!--', '-->', and everything between.
562 * To avoid leaving blank lines, when a comment is both preceded
563 * and followed by a newline (ignoring spaces), trim leading and
564 * trailing spaces and one of the newlines.
567 * @param $text String
570 static function removeHTMLcomments( $text ) {
571 wfProfileIn( __METHOD__
);
572 while (($start = strpos($text, '<!--')) !== false) {
573 $end = strpos($text, '-->', $start +
4);
574 if ($end === false) {
575 # Unterminated comment; bail out
581 # Trim space and newline if the comment is both
582 # preceded and followed by a newline
583 $spaceStart = max($start - 1, 0);
584 $spaceLen = $end - $spaceStart;
585 while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
589 while (substr($text, $spaceStart +
$spaceLen, 1) === ' ')
591 if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart +
$spaceLen, 1) === "\n") {
592 # Remove the comment, leading and trailing
593 # spaces, and leave only one newline.
594 $text = substr_replace($text, "\n", $spaceStart, $spaceLen +
1);
597 # Remove just the comment.
598 $text = substr_replace($text, '', $start, $end - $start);
601 wfProfileOut( __METHOD__
);
606 * Take an array of attribute names and values and normalize or discard
607 * illegal values for the given element type.
609 * - Discards attributes not on a whitelist for the given element
610 * - Unsafe style attributes are discarded
611 * - Invalid id attributes are reencoded
613 * @param $attribs Array
614 * @param $element String
617 * @todo Check for legal values where the DTD limits things.
618 * @todo Check for unique id attribute :P
620 static function validateTagAttributes( $attribs, $element ) {
621 return Sanitizer
::validateAttributes( $attribs,
622 Sanitizer
::attributeWhitelist( $element ) );
626 * Take an array of attribute names and values and normalize or discard
627 * illegal values for the given whitelist.
629 * - Discards attributes not the given whitelist
630 * - Unsafe style attributes are discarded
631 * - Invalid id attributes are reencoded
633 * @param $attribs Array
634 * @param $whitelist Array: list of allowed attribute names
637 * @todo Check for legal values where the DTD limits things.
638 * @todo Check for unique id attribute :P
640 static function validateAttributes( $attribs, $whitelist ) {
641 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
643 $whitelist = array_flip( $whitelist );
644 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
647 foreach( $attribs as $attribute => $value ) {
648 #allow XML namespace declaration if RDFa is enabled
649 if ( $wgAllowRdfaAttributes && preg_match( self
::XMLNS_ATTRIBUTE_PATTERN
, $attribute ) ) {
650 if ( !preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
651 $out[$attribute] = $value;
657 # Allow any attribute beginning with "data-", if in HTML5 mode
658 if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
662 # Strip javascript "expression" from stylesheets.
663 # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
664 if( $attribute == 'style' ) {
665 $value = Sanitizer
::checkCss( $value );
668 if ( $attribute === 'id' ) {
669 $value = Sanitizer
::escapeId( $value, 'noninitial' );
672 //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
673 if ( $attribute === 'rel' ||
$attribute === 'rev' ||
674 $attribute === 'about' ||
$attribute === 'property' ||
$attribute === 'resource' ||
#RDFa
675 $attribute === 'datatype' ||
$attribute === 'typeof' ||
#RDFa
676 $attribute === 'itemid' ||
$attribute === 'itemprop' ||
$attribute === 'itemref' ||
#HTML5 microdata
677 $attribute === 'itemscope' ||
$attribute === 'itemtype' ) { #HTML5 microdata
679 //Paranoia. Allow "simple" values but suppress javascript
680 if ( preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
685 # NOTE: even though elements using href/src are not allowed directly, supply
686 # validation code that can be used by tag hook handlers, etc
687 if ( $attribute === 'href' ||
$attribute === 'src' ) {
688 if ( !preg_match( $hrefExp, $value ) ) {
689 continue; //drop any href or src attributes not using an allowed protocol.
690 //NOTE: this also drops all relative URLs
694 // If this attribute was previously set, override it.
695 // Output should only have one attribute of each name.
696 $out[$attribute] = $value;
699 if ( $wgAllowMicrodataAttributes ) {
700 # itemtype, itemid, itemref don't make sense without itemscope
701 if ( !array_key_exists( 'itemscope', $out ) ) {
702 unset( $out['itemtype'] );
703 unset( $out['itemid'] );
704 unset( $out['itemref'] );
706 # TODO: Strip itemprop if we aren't descendants of an itemscope.
712 * Merge two sets of HTML attributes. Conflicting items in the second set
713 * will override those in the first, except for 'class' attributes which
714 * will be combined (if they're both strings).
716 * @todo implement merging for other attributes such as style
721 static function mergeAttributes( $a, $b ) {
722 $out = array_merge( $a, $b );
723 if( isset( $a['class'] ) && isset( $b['class'] )
724 && is_string( $a['class'] ) && is_string( $b['class'] )
725 && $a['class'] !== $b['class'] ) {
726 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
727 -1, PREG_SPLIT_NO_EMPTY
);
728 $out['class'] = implode( ' ', array_unique( $classes ) );
734 * Pick apart some CSS and check it for forbidden or unsafe structures.
735 * Returns a sanitized string. This sanitized string will have
736 * character references and escape sequences decoded, and comments
737 * stripped. If the input is just too evil, only a comment complaining
738 * about evilness will be returned.
740 * Currently URL references, 'expression', 'tps' are forbidden.
742 * NOTE: Despite the fact that character references are decoded, the
743 * returned string may contain character references given certain
744 * clever input strings. These character references must
745 * be escaped before the return value is embedded in HTML.
747 * @param $value String
750 static function checkCss( $value ) {
751 // Decode character references like {
752 $value = Sanitizer
::decodeCharReferences( $value );
754 // Decode escape sequences and line continuation
755 // See the grammar in the CSS 2 spec, appendix D.
756 // This has to be done AFTER decoding character references.
757 // This means it isn't possible for this function to return
758 // unsanitized escape sequences. It is possible to manufacture
759 // input that contains character references that decode to
760 // escape sequences that decode to character references, but
761 // it's OK for the return value to contain character references
762 // because the caller is supposed to escape those anyway.
764 if ( !$decodeRegex ) {
765 $space = '[\\x20\\t\\r\\n\\f]';
766 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
768 $decodeRegex = "/ $backslash
770 ($nl) | # 1. Line continuation
771 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
772 (.) | # 3. backslash cancelling special meaning
773 () | # 4. backslash at end of string
776 $value = preg_replace_callback( $decodeRegex,
777 array( __CLASS__
, 'cssDecodeCallback' ), $value );
779 // Remove any comments; IE gets token splitting wrong
780 // This must be done AFTER decoding character references and
781 // escape sequences, because those steps can introduce comments
782 // This step cannot introduce character references or escape
783 // sequences, because it replaces comments with spaces rather
784 // than removing them completely.
785 $value = StringUtils
::delimiterReplace( '/*', '*/', ' ', $value );
787 // Remove anything after a comment-start token, to guard against
788 // incorrect client implementations.
789 $commentPos = strpos( $value, '/*' );
790 if ( $commentPos !== false ) {
791 $value = substr( $value, 0, $commentPos );
794 // Reject problematic keywords and control characters
795 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
796 return '/* invalid control char */';
797 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
798 return '/* insecure input */';
804 * @param $matches array
807 static function cssDecodeCallback( $matches ) {
808 if ( $matches[1] !== '' ) {
811 } elseif ( $matches[2] !== '' ) {
812 $char = codepointToUtf8( hexdec( $matches[2] ) );
813 } elseif ( $matches[3] !== '' ) {
818 if ( $char == "\n" ||
$char == '"' ||
$char == "'" ||
$char == '\\' ) {
819 // These characters need to be escaped in strings
820 // Clean up the escape sequence to avoid parsing errors by clients
821 return '\\' . dechex( ord( $char ) ) . ' ';
823 // Decode unnecessary escape
829 * Take a tag soup fragment listing an HTML element's attributes
830 * and normalize it to well-formed XML, discarding unwanted attributes.
831 * Output is safe for further wikitext processing, with escaping of
832 * values that could trigger problems.
834 * - Normalizes attribute names to lowercase
835 * - Discards attributes not on a whitelist for the given element
836 * - Turns broken or invalid entities into plaintext
837 * - Double-quotes all attribute values
838 * - Attributes without values are given the name as attribute
839 * - Double attributes are discarded
840 * - Unsafe style attributes are discarded
841 * - Prepends space if there are attributes.
843 * @param $text String
844 * @param $element String
847 static function fixTagAttributes( $text, $element ) {
848 if( trim( $text ) == '' ) {
852 $stripped = Sanitizer
::validateTagAttributes(
853 Sanitizer
::decodeTagAttributes( $text ), $element );
856 foreach( $stripped as $attribute => $value ) {
857 $encAttribute = htmlspecialchars( $attribute );
858 $encValue = Sanitizer
::safeEncodeAttribute( $value );
860 $attribs[] = "$encAttribute=\"$encValue\"";
862 return count( $attribs ) ?
' ' . implode( ' ', $attribs ) : '';
866 * Encode an attribute value for HTML output.
867 * @param $text String
868 * @return HTML-encoded text fragment
870 static function encodeAttribute( $text ) {
871 $encValue = htmlspecialchars( $text, ENT_QUOTES
);
873 // Whitespace is normalized during attribute decoding,
874 // so if we've been passed non-spaces we must encode them
875 // ahead of time or they won't be preserved.
876 $encValue = strtr( $encValue, array(
886 * Encode an attribute value for HTML tags, with extra armoring
887 * against further wiki processing.
888 * @param $text String
889 * @return HTML-encoded text fragment
891 static function safeEncodeAttribute( $text ) {
892 $encValue = Sanitizer
::encodeAttribute( $text );
894 # Templates and links may be expanded in later parsing,
895 # creating invalid or dangerous output. Suppress this.
896 $encValue = strtr( $encValue, array(
897 '<' => '<', // This should never happen,
898 '>' => '>', // we've received invalid input
899 '"' => '"', // which should have been escaped.
902 "''" => '''',
903 'ISBN' => 'ISBN',
905 'PMID' => 'PMID',
911 $encValue = preg_replace_callback(
912 '/(' . wfUrlProtocols() . ')/',
913 array( 'Sanitizer', 'armorLinksCallback' ),
919 * Given a value, escape it so that it can be used in an id attribute and
920 * return it. This will use HTML5 validation if $wgExperimentalHtmlIds is
921 * true, allowing anything but ASCII whitespace. Otherwise it will use
922 * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
923 * escaped with lots of dots.
925 * To ensure we don't have to bother escaping anything, we also strip ', ",
926 * & even if $wgExperimentalIds is true. TODO: Is this the best tactic?
927 * We also strip # because it upsets IE, and % because it could be
928 * ambiguous if it's part of something that looks like a percent escape
929 * (which don't work reliably in fragments cross-browser).
931 * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
934 * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
935 * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
936 * HTML5 definition of id attribute
938 * @param $id String: id to escape
939 * @param $options Mixed: string or array of strings (default is array()):
940 * 'noninitial': This is a non-initial fragment of an id, not a full id,
941 * so don't pay attention if the first character isn't valid at the
942 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
944 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
945 * if $wgExperimentalHtmlIds is used, so we can generate extra
946 * anchors and links won't break.
949 static function escapeId( $id, $options = array() ) {
950 global $wgHtml5, $wgExperimentalHtmlIds;
951 $options = (array)$options;
953 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
954 $id = Sanitizer
::decodeCharReferences( $id );
955 $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
956 $id = trim( $id, '_' );
958 # Must have been all whitespace to start with.
965 # HTML4-style escaping
966 static $replace = array(
971 $id = urlencode( Sanitizer
::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
972 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
974 if ( !preg_match( '/^[a-zA-Z]/', $id )
975 && !in_array( 'noninitial', $options ) ) {
976 // Initial character must be a letter!
983 * Given a value, escape it so that it can be used as a CSS class and
986 * @todo For extra validity, input should be validated UTF-8.
988 * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
990 * @param $class String
993 static function escapeClass( $class ) {
994 // Convert ugly stuff to underscores and kill underscores in ugly places
995 return rtrim(preg_replace(
996 array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
1002 * Given HTML input, escape with htmlspecialchars but un-escape entites.
1003 * This allows (generally harmless) entities like   to survive.
1005 * @param $html String to escape
1006 * @return String: escaped input
1008 static function escapeHtmlAllowEntities( $html ) {
1009 $html = Sanitizer
::decodeCharReferences( $html );
1010 # It seems wise to escape ' as well as ", as a matter of course. Can't
1012 $html = htmlspecialchars( $html, ENT_QUOTES
);
1017 * Regex replace callback for armoring links against further processing.
1018 * @param $matches Array
1021 private static function armorLinksCallback( $matches ) {
1022 return str_replace( ':', ':', $matches[1] );
1026 * Return an associative array of attribute names and values from
1027 * a partial tag string. Attribute names are forces to lowercase,
1028 * character references are decoded to UTF-8 text.
1030 * @param $text String
1033 public static function decodeTagAttributes( $text ) {
1034 if( trim( $text ) == '' ) {
1040 if( !preg_match_all(
1041 self
::getAttribsRegex(),
1044 PREG_SET_ORDER
) ) {
1048 foreach( $pairs as $set ) {
1049 $attribute = strtolower( $set[1] );
1050 $value = Sanitizer
::getTagAttributeCallback( $set );
1052 // Normalize whitespace
1053 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1054 $value = trim( $value );
1056 // Decode character references
1057 $attribs[$attribute] = Sanitizer
::decodeCharReferences( $value );
1063 * Pick the appropriate attribute value from a match set from the
1064 * attribs regex matches.
1069 private static function getTagAttributeCallback( $set ) {
1070 if( isset( $set[6] ) ) {
1071 # Illegal #XXXXXX color with no quotes.
1073 } elseif( isset( $set[5] ) ) {
1076 } elseif( isset( $set[4] ) ) {
1079 } elseif( isset( $set[3] ) ) {
1082 } elseif( !isset( $set[2] ) ) {
1083 # In XHTML, attributes must have a value.
1084 # For 'reduced' form, return explicitly the attribute name here.
1087 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1092 * Normalize whitespace and character references in an XML source-
1093 * encoded text for an attribute value.
1095 * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1096 * but note that we're not returning the value, but are returning
1097 * XML source fragments that will be slapped into output.
1099 * @param $text String
1102 private static function normalizeAttributeValue( $text ) {
1103 return str_replace( '"', '"',
1104 self
::normalizeWhitespace(
1105 Sanitizer
::normalizeCharReferences( $text ) ) );
1109 * @param $text string
1112 private static function normalizeWhitespace( $text ) {
1113 return preg_replace(
1114 '/\r\n|[\x20\x0d\x0a\x09]/',
1120 * Normalizes whitespace in a section name, such as might be returned
1121 * by Parser::stripSectionName(), for use in the id's that are used for
1124 * @param $section String
1127 static function normalizeSectionNameWhitespace( $section ) {
1128 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1132 * Ensure that any entities and character references are legal
1133 * for XML and XHTML specifically. Any stray bits will be
1134 * &-escaped to result in a valid text fragment.
1136 * a. named char refs can only be < > & ", others are
1137 * numericized (this way we're well-formed even without a DTD)
1138 * b. any numeric char refs must be legal chars, not invalid or forbidden
1139 * c. use &#x, not &#X
1140 * d. fix or reject non-valid attributes
1142 * @param $text String
1146 static function normalizeCharReferences( $text ) {
1147 return preg_replace_callback(
1148 self
::CHAR_REFS_REGEX
,
1149 array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1153 * @param $matches String
1156 static function normalizeCharReferencesCallback( $matches ) {
1158 if( $matches[1] != '' ) {
1159 $ret = Sanitizer
::normalizeEntity( $matches[1] );
1160 } elseif( $matches[2] != '' ) {
1161 $ret = Sanitizer
::decCharReference( $matches[2] );
1162 } elseif( $matches[3] != '' ) {
1163 $ret = Sanitizer
::hexCharReference( $matches[3] );
1165 if( is_null( $ret ) ) {
1166 return htmlspecialchars( $matches[0] );
1173 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1174 * return the equivalent numeric entity reference (except for the core <
1175 * > & "). If the entity is a MediaWiki-specific alias, returns
1176 * the HTML equivalent. Otherwise, returns HTML-escaped text of
1177 * pseudo-entity source (eg &foo;)
1179 * @param $name String
1182 static function normalizeEntity( $name ) {
1183 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1184 return '&' . self
::$htmlEntityAliases[$name] . ';';
1185 } elseif ( in_array( $name,
1186 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1188 } elseif ( isset( self
::$htmlEntities[$name] ) ) {
1189 return '&#' . self
::$htmlEntities[$name] . ';';
1191 return "&$name;";
1197 * @return null|string
1199 static function decCharReference( $codepoint ) {
1200 $point = intval( $codepoint );
1201 if( Sanitizer
::validateCodepoint( $point ) ) {
1202 return sprintf( '&#%d;', $point );
1210 * @return null|string
1212 static function hexCharReference( $codepoint ) {
1213 $point = hexdec( $codepoint );
1214 if( Sanitizer
::validateCodepoint( $point ) ) {
1215 return sprintf( '&#x%x;', $point );
1222 * Returns true if a given Unicode codepoint is a valid character in XML.
1223 * @param $codepoint Integer
1226 private static function validateCodepoint( $codepoint ) {
1227 return ($codepoint == 0x09)
1228 ||
($codepoint == 0x0a)
1229 ||
($codepoint == 0x0d)
1230 ||
($codepoint >= 0x20 && $codepoint <= 0xd7ff)
1231 ||
($codepoint >= 0xe000 && $codepoint <= 0xfffd)
1232 ||
($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1236 * Decode any character references, numeric or named entities,
1237 * in the text and return a UTF-8 string.
1239 * @param $text String
1242 public static function decodeCharReferences( $text ) {
1243 return preg_replace_callback(
1244 self
::CHAR_REFS_REGEX
,
1245 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1250 * Decode any character references, numeric or named entities,
1251 * in the next and normalize the resulting string. (bug 14952)
1253 * This is useful for page titles, not for text to be displayed,
1254 * MediaWiki allows HTML entities to escape normalization as a feature.
1256 * @param $text String (already normalized, containing entities)
1257 * @return String (still normalized, without entities)
1259 public static function decodeCharReferencesAndNormalize( $text ) {
1261 $text = preg_replace_callback(
1262 self
::CHAR_REFS_REGEX
,
1263 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1264 $text, /* limit */ -1, $count );
1267 return $wgContLang->normalize( $text );
1274 * @param $matches String
1277 static function decodeCharReferencesCallback( $matches ) {
1278 if( $matches[1] != '' ) {
1279 return Sanitizer
::decodeEntity( $matches[1] );
1280 } elseif( $matches[2] != '' ) {
1281 return Sanitizer
::decodeChar( intval( $matches[2] ) );
1282 } elseif( $matches[3] != '' ) {
1283 return Sanitizer
::decodeChar( hexdec( $matches[3] ) );
1285 # Last case should be an ampersand by itself
1290 * Return UTF-8 string for a codepoint if that is a valid
1291 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1292 * @param $codepoint Integer
1296 static function decodeChar( $codepoint ) {
1297 if( Sanitizer
::validateCodepoint( $codepoint ) ) {
1298 return codepointToUtf8( $codepoint );
1300 return UTF8_REPLACEMENT
;
1305 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1306 * return the UTF-8 encoding of that character. Otherwise, returns
1307 * pseudo-entity source (eg &foo;)
1309 * @param $name String
1312 static function decodeEntity( $name ) {
1313 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1314 $name = self
::$htmlEntityAliases[$name];
1316 if( isset( self
::$htmlEntities[$name] ) ) {
1317 return codepointToUtf8( self
::$htmlEntities[$name] );
1324 * Fetch the whitelist of acceptable attributes for a given element name.
1326 * @param $element String
1329 static function attributeWhitelist( $element ) {
1331 if( !isset( $list ) ) {
1332 $list = Sanitizer
::setupAttributeWhitelist();
1334 return isset( $list[$element] )
1340 * Foreach array key (an allowed HTML element), return an array
1341 * of allowed attributes
1344 static function setupAttributeWhitelist() {
1345 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1347 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1349 if ( $wgAllowRdfaAttributes ) {
1350 #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1351 $common = array_merge( $common, array(
1352 'about', 'property', 'resource', 'datatype', 'typeof',
1356 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1357 # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1358 $common = array_merge( $common, array(
1359 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1363 $block = array_merge( $common, array( 'align' ) );
1364 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1365 $tablecell = array( 'abbr',
1371 'nowrap', # deprecated
1372 'width', # deprecated
1373 'height', # deprecated
1374 'bgcolor' # deprecated
1377 # Numbers refer to sections in HTML 4.01 standard describing the element.
1378 # See: http://www.w3.org/TR/html4/
1379 $whitelist = array (
1382 'center' => $common, # deprecated
1383 'span' => $block, # ??
1401 'strong' => $common,
1412 'blockquote' => array_merge( $common, array( 'cite' ) ),
1423 'br' => array( 'id', 'class', 'title', 'style', 'clear' ),
1426 'pre' => array_merge( $common, array( 'width' ) ),
1429 'ins' => array_merge( $common, array( 'cite', 'datetime' ) ),
1430 'del' => array_merge( $common, array( 'cite', 'datetime' ) ),
1433 'ul' => array_merge( $common, array( 'type' ) ),
1434 'ol' => array_merge( $common, array( 'type', 'start' ) ),
1435 'li' => array_merge( $common, array( 'type', 'value' ) ),
1443 'table' => array_merge( $common,
1444 array( 'summary', 'width', 'border', 'frame',
1445 'rules', 'cellspacing', 'cellpadding',
1450 'caption' => array_merge( $common, array( 'align' ) ),
1453 'thead' => array_merge( $common, $tablealign ),
1454 'tfoot' => array_merge( $common, $tablealign ),
1455 'tbody' => array_merge( $common, $tablealign ),
1458 'colgroup' => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1459 'col' => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1462 'tr' => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1465 'td' => array_merge( $common, $tablecell, $tablealign ),
1466 'th' => array_merge( $common, $tablecell, $tablealign ),
1468 # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1469 'a' => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1472 # Not usually allowed, but may be used for extension-style hooks
1473 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1475 'img' => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1483 'strike' => $common,
1488 'font' => array_merge( $common, array( 'size', 'color', 'face' ) ),
1492 'hr' => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1494 # XHTML Ruby annotation text module, simple ruby only.
1495 # http://www.w3c.org/TR/ruby/
1500 'rt' => $common, #array_merge( $common, array( 'rbspan' ) ),
1503 # MathML root element, where used for extensions
1504 # 'title' may not be 100% valid here; it's XHTML
1505 # http://www.w3.org/TR/REC-MathML/
1506 'math' => array( 'class', 'style', 'id', 'title' ),
1512 * Take a fragment of (potentially invalid) HTML and return
1513 * a version with any tags removed, encoded as plain text.
1515 * Warning: this return value must be further escaped for literal
1516 * inclusion in HTML output as of 1.10!
1518 * @param $text String: HTML fragment
1521 static function stripAllTags( $text ) {
1523 $text = StringUtils
::delimiterReplace( '<', '>', '', $text );
1525 # Normalize &entities and whitespace
1526 $text = self
::decodeCharReferences( $text );
1527 $text = self
::normalizeWhitespace( $text );
1533 * Hack up a private DOCTYPE with HTML's standard entity declarations.
1534 * PHP 4 seemed to know these if you gave it an HTML doctype, but
1537 * Use for passing XHTML fragments to PHP's XML parsing functions
1541 static function hackDocType() {
1542 $out = "<!DOCTYPE html [\n";
1543 foreach( self
::$htmlEntities as $entity => $codepoint ) {
1544 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1551 * @param $url string
1552 * @return mixed|string
1554 static function cleanUrl( $url ) {
1555 # Normalize any HTML entities in input. They will be
1556 # re-escaped by makeExternalLink().
1557 $url = Sanitizer
::decodeCharReferences( $url );
1559 # Escape any control characters introduced by the above step
1560 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1561 array( __CLASS__
, 'cleanUrlCallback' ), $url );
1563 # Validate hostname portion
1565 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1566 list( /* $whole */, $protocol, $host, $rest ) = $matches;
1568 // Characters that will be ignored in IDNs.
1569 // http://tools.ietf.org/html/3454#section-3.1
1570 // Strip them before further processing so blacklists and such work.
1572 \\s| # general whitespace
1573 \xc2\xad| # 00ad SOFT HYPHEN
1574 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1575 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1576 \xe2\x81\xa0| # 2060 WORD JOINER
1577 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1578 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
1579 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1580 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1581 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1582 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1583 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1584 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe00f VARIATION SELECTOR-1-16
1587 $host = preg_replace( $strip, '', $host );
1589 // @todo FIXME: Validate hostnames here
1591 return $protocol . $host . $rest;
1598 * @param $matches array
1601 static function cleanUrlCallback( $matches ) {
1602 return urlencode( $matches[0] );
1606 * Does a string look like an e-mail address?
1608 * This validates an email address using an HTML5 specification found at:
1609 * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1610 * Which as of 2011-01-24 says:
1612 * A valid e-mail address is a string that matches the ABNF production
1613 * 1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1614 * in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1617 * This function is an implementation of the specification as requested in
1620 * Client-side forms will use the same standard validation rules via JS or
1621 * HTML 5 validation; additional restrictions can be enforced server-side
1622 * by extensions via the 'isValidEmailAddr' hook.
1624 * Note that this validation doesn't 100% match RFC 2822, but is believed
1625 * to be liberal enough for wide use. Some invalid addresses will still
1626 * pass validation here.
1630 * @param $addr String E-mail address
1633 public static function validateEmail( $addr ) {
1635 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1639 // Please note strings below are enclosed in brackets [], this make the
1640 // hyphen "-" a range indicator. Hence it is double backslashed below.
1642 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1643 $rfc1034_ldh_str = "a-z0-9\\-" ;
1645 $HTML5_email_regexp = "/
1647 [$rfc5322_atext\\.]+ # user part which is liberal :p
1649 [$rfc1034_ldh_str]+ # First domain part
1650 (\\.[$rfc1034_ldh_str]+)* # Following part prefixed with a dot
1652 /ix" ; // case Insensitive, eXtended
1654 return (bool) preg_match( $HTML5_email_regexp, $addr );