3 * XHTML sanitizer for MediaWiki
5 * Copyright (C) 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 * Regular expression to match various types of character references in
29 * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
31 define( 'MW_CHAR_REFS_REGEX',
32 '/&([A-Za-z0-9\x80-\xff]+);
39 * Regular expression to match HTML/XML attribute pairs within a tag.
40 * Allows some... latitude.
41 * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
43 $attrib = '[A-Za-z0-9]';
44 $space = '[\x09\x0a\x0d\x20]';
45 define( 'MW_ATTRIBS_REGEX',
46 "/(?:^|$space)((?:xml:|xmlns:)?$attrib+)
49 # The attribute value: quoted or alone
52 | ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
53 | (\#[0-9a-fA-F]+) # Technically wrong, but lots of
54 # colors are specified like this.
55 # We'll be normalizing it.
57 )?(?=$space|\$)/sx" );
60 * Regular expression to match URIs that could trigger script execution
62 define( 'MW_EVIL_URI_PATTERN', '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i' );
65 * Regular expression to match namespace attributes
67 define( 'MW_XMLNS_ATTRIBUTE_PATTRN', "/^xmlns:$attrib+$/" );
70 * List of all named character entities defined in HTML 4.01
71 * http://www.w3.org/TR/html4/sgml/entities.html
74 global $wgHtmlEntities;
75 $wgHtmlEntities = array(
330 * Character entity aliases accepted by MediaWiki
332 global $wgHtmlEntityAliases;
333 $wgHtmlEntityAliases = array(
340 * XHTML sanitizer for MediaWiki
345 * Cleans up HTML, removes dangerous tags and attributes, and
346 * removes HTML comments
348 * @param $text String
349 * @param $processCallback Callback to do any variable or parameter replacements in HTML attribute values
350 * @param $args Array for the processing callback
351 * @param $extratags Array for any extra tags to include
352 * @param $removetags Array for any tags (default or extra) to exclude
355 static function removeHTMLtags( $text, $processCallback = null, $args = array(), $extratags = array(), $removetags = array() ) {
358 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
359 $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
361 wfProfileIn( __METHOD__ );
363 if ( !$staticInitialised ) {
365 $htmlpairsStatic = array( # Tags that must be closed
366 'b', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
367 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
368 'strike', 'strong', 'tt', 'var', 'div', 'center',
369 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
370 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'u', 'abbr'
373 'br', 'hr', 'li', 'dt', 'dd'
375 $htmlsingleonly = array( # Elements that cannot have close tags
378 $htmlnest = array( # Tags that can be nested--??
379 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
380 'dl', 'font', 'big', 'small', 'sub', 'sup', 'span'
382 $tabletags = array( # Can only appear inside table, we will close them
385 $htmllist = array( # Tags used by list
388 $listtags = array( # Tags that can appear in a list
392 $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
393 $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
395 # Convert them all to hashtables for faster lookup
396 $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
397 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
398 foreach ( $vars as $var ) {
399 $$var = array_flip( $$var );
401 $staticInitialised = true;
403 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
404 $extratags = array_flip( $extratags );
405 $removetags = array_flip( $removetags );
406 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
407 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ) , $removetags );
409 # Remove HTML comments
410 $text = Sanitizer::removeHTMLcomments( $text );
411 $bits = explode( '<', $text );
412 $text = str_replace( '>', '>', array_shift( $bits ) );
414 $tagstack = $tablestack = array();
415 foreach ( $bits as $x ) {
417 # $slash: Does the current element start with a '/'?
418 # $t: Current element name
419 # $params: String between element name and >
420 # $brace: Ending '>' or '/>'
421 # $rest: Everything until the next element of $bits
422 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
423 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
425 $slash = $t = $params = $brace = $rest = null;
429 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
431 if ( $slash && isset( $htmlsingleonly[$t] ) ) {
433 } elseif ( $slash ) {
434 # Closing a tag... is it the one we just opened?
435 $ot = @array_pop( $tagstack );
437 if ( isset( $htmlsingleallowed[$ot] ) ) {
438 # Pop all elements with an optional close tag
439 # and see if we find a match below them
441 array_push( $optstack, $ot );
442 $ot = @array_pop( $tagstack );
443 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
444 array_push( $optstack, $ot );
445 $ot = @array_pop( $tagstack );
448 # No match. Push the optional elements back again
450 while ( $ot = @array_pop( $optstack ) ) {
451 array_push( $tagstack, $ot );
455 @array_push( $tagstack, $ot );
456 # <li> can be nested in <ul> or <ol>, skip those cases:
457 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
462 if ( $t == 'table' ) {
463 $tagstack = array_pop( $tablestack );
468 # Keep track for later
469 if ( isset( $tabletags[$t] ) &&
470 !in_array( 'table', $tagstack ) ) {
472 } elseif ( in_array( $t, $tagstack ) &&
473 !isset( $htmlnest [$t ] ) ) {
475 #Â Is it a self closed htmlpair ? (bug 5487)
476 } elseif ( $brace == '/>' &&
477 isset( $htmlpairs[$t] ) ) {
479 } elseif ( isset( $htmlsingleonly[$t] ) ) {
480 # Hack to force empty tag for uncloseable elements
482 } elseif ( isset( $htmlsingle[$t] ) ) {
483 # Hack to not close $htmlsingle tags
485 } elseif ( isset( $tabletags[$t] )
486 && in_array( $t, $tagstack ) ) {
487 // New table tag but forgot to close the previous one
490 if ( $t == 'table' ) {
491 array_push( $tablestack, $tagstack );
494 array_push( $tagstack, $t );
497 # Replace any variables or template parameters with
499 if( is_callable( $processCallback ) ) {
500 call_user_func_array( $processCallback, array( &$params, $args ) );
503 # Strip non-approved attributes from the tag
504 $newparams = Sanitizer::fixTagAttributes( $params, $t );
507 $rest = str_replace( '>', '>', $rest );
508 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
509 $text .= "<$slash$t$newparams$close>$rest";
513 $text .= '<' . str_replace( '>', '>', $x);
515 # Close off any remaining tags
516 while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
518 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
521 # this might be possible using tidy itself
522 foreach ( $bits as $x ) {
523 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
525 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
526 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
527 if( is_callable( $processCallback ) ) {
528 call_user_func_array( $processCallback, array( &$params, $args ) );
530 $newparams = Sanitizer::fixTagAttributes( $params, $t );
531 $rest = str_replace( '>', '>', $rest );
532 $text .= "<$slash$t$newparams$brace$rest";
534 $text .= '<' . str_replace( '>', '>', $x);
538 wfProfileOut( __METHOD__ );
543 * Remove '<!--', '-->', and everything between.
544 * To avoid leaving blank lines, when a comment is both preceded
545 * and followed by a newline (ignoring spaces), trim leading and
546 * trailing spaces and one of the newlines.
549 * @param $text String
552 static function removeHTMLcomments( $text ) {
553 wfProfileIn( __METHOD__ );
554 while (($start = strpos($text, '<!--')) !== false) {
555 $end = strpos($text, '-->', $start + 4);
556 if ($end === false) {
557 # Unterminated comment; bail out
563 # Trim space and newline if the comment is both
564 # preceded and followed by a newline
565 $spaceStart = max($start - 1, 0);
566 $spaceLen = $end - $spaceStart;
567 while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
571 while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
573 if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
574 # Remove the comment, leading and trailing
575 # spaces, and leave only one newline.
576 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
579 # Remove just the comment.
580 $text = substr_replace($text, '', $start, $end - $start);
583 wfProfileOut( __METHOD__ );
588 * Take an array of attribute names and values and normalize or discard
589 * illegal values for the given element type.
591 * - Discards attributes not on a whitelist for the given element
592 * - Unsafe style attributes are discarded
593 * - Invalid id attributes are reencoded
595 * @param $attribs Array
596 * @param $element String
599 * @todo Check for legal values where the DTD limits things.
600 * @todo Check for unique id attribute :P
602 static function validateTagAttributes( $attribs, $element ) {
603 return Sanitizer::validateAttributes( $attribs,
604 Sanitizer::attributeWhitelist( $element ) );
608 * Take an array of attribute names and values and normalize or discard
609 * illegal values for the given whitelist.
611 * - Discards attributes not the given whitelist
612 * - Unsafe style attributes are discarded
613 * - Invalid id attributes are reencoded
615 * @param $attribs Array
616 * @param $whitelist Array: list of allowed attribute names
619 * @todo Check for legal values where the DTD limits things.
620 * @todo Check for unique id attribute :P
622 static function validateAttributes( $attribs, $whitelist ) {
623 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
625 $whitelist = array_flip( $whitelist );
626 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
629 foreach( $attribs as $attribute => $value ) {
630 #allow XML namespace declaration if RDFa is enabled
631 if ( $wgAllowRdfaAttributes && preg_match( MW_XMLNS_ATTRIBUTE_PATTRN, $attribute ) ) {
632 if ( !preg_match( MW_EVIL_URI_PATTERN, $value ) ) {
633 $out[$attribute] = $value;
639 if( !isset( $whitelist[$attribute] ) ) {
643 # Strip javascript "expression" from stylesheets.
644 # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
645 if( $attribute == 'style' ) {
646 $value = Sanitizer::checkCss( $value );
649 if ( $attribute === 'id' ) {
650 $value = Sanitizer::escapeId( $value, 'noninitial' );
653 //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
654 if ( $attribute === 'rel' || $attribute === 'rev' ||
655 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
656 $attribute === 'datatype' || $attribute === 'typeof' || #RDFa
657 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
658 $attribute === 'itemscope' || $attribute === 'itemtype' ) { #HTML5 microdata
660 //Paranoia. Allow "simple" values but suppress javascript
661 if ( preg_match( MW_EVIL_URI_PATTERN, $value ) ) {
666 # NOTE: even though elements using href/src are not allowed directly, supply
667 # validation code that can be used by tag hook handlers, etc
668 if ( $attribute === 'href' || $attribute === 'src' ) {
669 if ( !preg_match( $hrefExp, $value ) ) {
670 continue; //drop any href or src attributes not using an allowed protocol.
671 //NOTE: this also drops all relative URLs
675 // If this attribute was previously set, override it.
676 // Output should only have one attribute of each name.
677 $out[$attribute] = $value;
680 if ( $wgAllowMicrodataAttributes ) {
681 # There are some complicated validity constraints we need to
682 # enforce here. First of all, we don't want to allow non-standard
684 $allowedTypes = array(
685 'http://microformats.org/profile/hcard',
686 'http://microformats.org/profile/hcalendar#vevent',
687 'http://n.whatwg.org/work',
689 if ( isset( $out['itemtype'] ) && !in_array( $out['itemtype'],
692 unset( $out['itemscope'] );
694 # itemtype, itemid, itemref don't make sense without itemscope
695 if ( !array_key_exists( 'itemscope', $out ) ) {
696 unset( $out['itemtype'] );
697 unset( $out['itemid'] );
698 unset( $out['itemref'] );
700 # TODO: Strip itemprop if we aren't descendants of an itemscope.
706 * Merge two sets of HTML attributes. Conflicting items in the second set
707 * will override those in the first, except for 'class' attributes which
708 * will be combined (if they're both strings).
710 * @todo implement merging for other attributes such as style
715 static function mergeAttributes( $a, $b ) {
716 $out = array_merge( $a, $b );
717 if( isset( $a['class'] ) && isset( $b['class'] )
718 && is_string( $a['class'] ) && is_string( $b['class'] )
719 && $a['class'] !== $b['class'] ) {
720 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
721 -1, PREG_SPLIT_NO_EMPTY );
722 $out['class'] = implode( ' ', array_unique( $classes ) );
728 * Pick apart some CSS and check it for forbidden or unsafe structures.
729 * Returns a sanitized string, or false if it was just too evil.
731 * Currently URL references, 'expression', 'tps' are forbidden.
733 * @param $value String
736 static function checkCss( $value ) {
737 $value = Sanitizer::decodeCharReferences( $value );
739 // Remove any comments; IE gets token splitting wrong
740 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
742 // Decode escape sequences and line continuation
743 // See the grammar in the CSS 2 spec, appendix D.
744 static $decodeRegex, $reencodeTable;
745 if ( !$decodeRegex ) {
746 $space = '[\\x20\\t\\r\\n\\f]';
747 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
749 $decodeRegex = "/ $backslash
751 ($nl) | # 1. Line continuation
752 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
753 (.) | # 3. backslash cancelling special meaning
754 () | # 4. backslash at end of string
757 $value = preg_replace_callback( $decodeRegex,
758 array( __CLASS__, 'cssDecodeCallback' ), $value );
760 // Reject problematic keywords and control characters
761 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
762 return '/* invalid control char */';
763 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
764 return '/* insecure input */';
769 static function cssDecodeCallback( $matches ) {
770 if ( $matches[1] !== '' ) {
773 } elseif ( $matches[2] !== '' ) {
774 $char = codepointToUtf8( hexdec( $matches[2] ) );
775 } elseif ( $matches[3] !== '' ) {
780 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
781 // These characters need to be escaped in strings
782 // Clean up the escape sequence to avoid parsing errors by clients
783 return '\\' . dechex( ord( $char ) ) . ' ';
785 // Decode unnecessary escape
791 * Take a tag soup fragment listing an HTML element's attributes
792 * and normalize it to well-formed XML, discarding unwanted attributes.
793 * Output is safe for further wikitext processing, with escaping of
794 * values that could trigger problems.
796 * - Normalizes attribute names to lowercase
797 * - Discards attributes not on a whitelist for the given element
798 * - Turns broken or invalid entities into plaintext
799 * - Double-quotes all attribute values
800 * - Attributes without values are given the name as attribute
801 * - Double attributes are discarded
802 * - Unsafe style attributes are discarded
803 * - Prepends space if there are attributes.
805 * @param $text String
806 * @param $element String
809 static function fixTagAttributes( $text, $element ) {
810 if( trim( $text ) == '' ) {
814 $stripped = Sanitizer::validateTagAttributes(
815 Sanitizer::decodeTagAttributes( $text ), $element );
818 foreach( $stripped as $attribute => $value ) {
819 $encAttribute = htmlspecialchars( $attribute );
820 $encValue = Sanitizer::safeEncodeAttribute( $value );
822 $attribs[] = "$encAttribute=\"$encValue\"";
824 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
828 * Encode an attribute value for HTML output.
829 * @param $text String
830 * @return HTML-encoded text fragment
832 static function encodeAttribute( $text ) {
833 $encValue = htmlspecialchars( $text, ENT_QUOTES );
835 // Whitespace is normalized during attribute decoding,
836 // so if we've been passed non-spaces we must encode them
837 // ahead of time or they won't be preserved.
838 $encValue = strtr( $encValue, array(
848 * Encode an attribute value for HTML tags, with extra armoring
849 * against further wiki processing.
850 * @param $text String
851 * @return HTML-encoded text fragment
853 static function safeEncodeAttribute( $text ) {
854 $encValue = Sanitizer::encodeAttribute( $text );
856 # Templates and links may be expanded in later parsing,
857 # creating invalid or dangerous output. Suppress this.
858 $encValue = strtr( $encValue, array(
859 '<' => '<', // This should never happen,
860 '>' => '>', // we've received invalid input
861 '"' => '"', // which should have been escaped.
864 "''" => '''',
865 'ISBN' => 'ISBN',
867 'PMID' => 'PMID',
873 $encValue = preg_replace_callback(
874 '/(' . wfUrlProtocols() . ')/',
875 array( 'Sanitizer', 'armorLinksCallback' ),
881 * Given a value, escape it so that it can be used in an id attribute and
882 * return it. This will use HTML5 validation if $wgExperimentalHtmlIds is
883 * true, allowing anything but ASCII whitespace. Otherwise it will use
884 * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
885 * escaped with lots of dots.
887 * To ensure we don't have to bother escaping anything, we also strip ', ",
888 * & even if $wgExperimentalIds is true. TODO: Is this the best tactic?
889 * We also strip # because it upsets IE6.
891 * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
894 * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
895 * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
896 * HTML5 definition of id attribute
898 * @param $id String: id to escape
899 * @param $options Mixed: string or array of strings (default is array()):
900 * 'noninitial': This is a non-initial fragment of an id, not a full id,
901 * so don't pay attention if the first character isn't valid at the
902 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
904 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
905 * if $wgExperimentalHtmlIds is used, so we can generate extra
906 * anchors and links won't break.
909 static function escapeId( $id, $options = array() ) {
910 global $wgHtml5, $wgExperimentalHtmlIds;
911 $options = (array)$options;
913 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
914 $id = Sanitizer::decodeCharReferences( $id );
915 $id = preg_replace( '/[ \t\n\r\f_\'"&#]+/', '_', $id );
916 $id = trim( $id, '_' );
918 # Must have been all whitespace to start with.
925 # HTML4-style escaping
926 static $replace = array(
931 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
932 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
934 if ( !preg_match( '/^[a-zA-Z]/', $id )
935 && !in_array( 'noninitial', $options ) ) {
936 // Initial character must be a letter!
943 * Given a value, escape it so that it can be used as a CSS class and
946 * @todo For extra validity, input should be validated UTF-8.
948 * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
950 * @param $class String
953 static function escapeClass( $class ) {
954 // Convert ugly stuff to underscores and kill underscores in ugly places
955 return rtrim(preg_replace(
956 array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
962 * Given HTML input, escape with htmlspecialchars but un-escape entites.
963 * This allows (generally harmless) entities like to survive.
965 * @param $html String to escape
966 * @return String: escaped input
968 static function escapeHtmlAllowEntities( $html ) {
969 # It seems wise to escape ' as well as ", as a matter of course. Can't
971 $html = htmlspecialchars( $html, ENT_QUOTES );
972 $html = str_replace( '&', '&', $html );
973 $html = Sanitizer::normalizeCharReferences( $html );
978 * Regex replace callback for armoring links against further processing.
979 * @param $matches Array
982 private static function armorLinksCallback( $matches ) {
983 return str_replace( ':', ':', $matches[1] );
987 * Return an associative array of attribute names and values from
988 * a partial tag string. Attribute names are forces to lowercase,
989 * character references are decoded to UTF-8 text.
991 * @param $text String
994 public static function decodeTagAttributes( $text ) {
995 if( trim( $text ) == '' ) {
1001 if( !preg_match_all(
1005 PREG_SET_ORDER ) ) {
1009 foreach( $pairs as $set ) {
1010 $attribute = strtolower( $set[1] );
1011 $value = Sanitizer::getTagAttributeCallback( $set );
1013 // Normalize whitespace
1014 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1015 $value = trim( $value );
1017 // Decode character references
1018 $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1024 * Pick the appropriate attribute value from a match set from the
1025 * MW_ATTRIBS_REGEX matches.
1030 private static function getTagAttributeCallback( $set ) {
1031 if( isset( $set[6] ) ) {
1032 # Illegal #XXXXXX color with no quotes.
1034 } elseif( isset( $set[5] ) ) {
1037 } elseif( isset( $set[4] ) ) {
1040 } elseif( isset( $set[3] ) ) {
1043 } elseif( !isset( $set[2] ) ) {
1044 # In XHTML, attributes must have a value.
1045 # For 'reduced' form, return explicitly the attribute name here.
1048 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1053 * Normalize whitespace and character references in an XML source-
1054 * encoded text for an attribute value.
1056 * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1057 * but note that we're not returning the value, but are returning
1058 * XML source fragments that will be slapped into output.
1060 * @param $text String
1063 private static function normalizeAttributeValue( $text ) {
1064 return str_replace( '"', '"',
1065 self::normalizeWhitespace(
1066 Sanitizer::normalizeCharReferences( $text ) ) );
1069 private static function normalizeWhitespace( $text ) {
1070 return preg_replace(
1071 '/\r\n|[\x20\x0d\x0a\x09]/',
1077 * Ensure that any entities and character references are legal
1078 * for XML and XHTML specifically. Any stray bits will be
1079 * &-escaped to result in a valid text fragment.
1081 * a. any named char refs must be known in XHTML
1082 * b. any numeric char refs must be legal chars, not invalid or forbidden
1083 * c. use &#x, not &#X
1084 * d. fix or reject non-valid attributes
1086 * @param $text String
1090 static function normalizeCharReferences( $text ) {
1091 return preg_replace_callback(
1093 array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1097 * @param $matches String
1100 static function normalizeCharReferencesCallback( $matches ) {
1102 if( $matches[1] != '' ) {
1103 $ret = Sanitizer::normalizeEntity( $matches[1] );
1104 } elseif( $matches[2] != '' ) {
1105 $ret = Sanitizer::decCharReference( $matches[2] );
1106 } elseif( $matches[3] != '' ) {
1107 $ret = Sanitizer::hexCharReference( $matches[3] );
1108 } elseif( $matches[4] != '' ) {
1109 $ret = Sanitizer::hexCharReference( $matches[4] );
1111 if( is_null( $ret ) ) {
1112 return htmlspecialchars( $matches[0] );
1119 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1120 * return the named entity reference as is. If the entity is a
1121 * MediaWiki-specific alias, returns the HTML equivalent. Otherwise,
1122 * returns HTML-escaped text of pseudo-entity source (eg &foo;)
1124 * @param $name String
1127 static function normalizeEntity( $name ) {
1128 global $wgHtmlEntities, $wgHtmlEntityAliases;
1129 if ( isset( $wgHtmlEntityAliases[$name] ) ) {
1130 return "&{$wgHtmlEntityAliases[$name]};";
1131 } elseif( isset( $wgHtmlEntities[$name] ) ) {
1134 return "&$name;";
1138 static function decCharReference( $codepoint ) {
1139 $point = intval( $codepoint );
1140 if( Sanitizer::validateCodepoint( $point ) ) {
1141 return sprintf( '&#%d;', $point );
1147 static function hexCharReference( $codepoint ) {
1148 $point = hexdec( $codepoint );
1149 if( Sanitizer::validateCodepoint( $point ) ) {
1150 return sprintf( '&#x%x;', $point );
1157 * Returns true if a given Unicode codepoint is a valid character in XML.
1158 * @param $codepoint Integer
1161 private static function validateCodepoint( $codepoint ) {
1162 return ($codepoint == 0x09)
1163 || ($codepoint == 0x0a)
1164 || ($codepoint == 0x0d)
1165 || ($codepoint >= 0x20 && $codepoint <= 0xd7ff)
1166 || ($codepoint >= 0xe000 && $codepoint <= 0xfffd)
1167 || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1171 * Decode any character references, numeric or named entities,
1172 * in the text and return a UTF-8 string.
1174 * @param $text String
1177 public static function decodeCharReferences( $text ) {
1178 return preg_replace_callback(
1180 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1185 * @param $matches String
1188 static function decodeCharReferencesCallback( $matches ) {
1189 if( $matches[1] != '' ) {
1190 return Sanitizer::decodeEntity( $matches[1] );
1191 } elseif( $matches[2] != '' ) {
1192 return Sanitizer::decodeChar( intval( $matches[2] ) );
1193 } elseif( $matches[3] != '' ) {
1194 return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1195 } elseif( $matches[4] != '' ) {
1196 return Sanitizer::decodeChar( hexdec( $matches[4] ) );
1198 # Last case should be an ampersand by itself
1203 * Return UTF-8 string for a codepoint if that is a valid
1204 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1205 * @param $codepoint Integer
1209 static function decodeChar( $codepoint ) {
1210 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1211 return codepointToUtf8( $codepoint );
1213 return UTF8_REPLACEMENT;
1218 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1219 * return the UTF-8 encoding of that character. Otherwise, returns
1220 * pseudo-entity source (eg &foo;)
1222 * @param $name Strings
1225 static function decodeEntity( $name ) {
1226 global $wgHtmlEntities, $wgHtmlEntityAliases;
1227 if ( isset( $wgHtmlEntityAliases[$name] ) ) {
1228 $name = $wgHtmlEntityAliases[$name];
1230 if( isset( $wgHtmlEntities[$name] ) ) {
1231 return codepointToUtf8( $wgHtmlEntities[$name] );
1238 * Fetch the whitelist of acceptable attributes for a given element name.
1240 * @param $element String
1243 static function attributeWhitelist( $element ) {
1245 if( !isset( $list ) ) {
1246 $list = Sanitizer::setupAttributeWhitelist();
1248 return isset( $list[$element] )
1254 * Foreach array key (an allowed HTML element), return an array
1255 * of allowed attributes
1258 static function setupAttributeWhitelist() {
1259 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1261 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1263 if ( $wgAllowRdfaAttributes ) {
1264 #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1265 $common = array_merge( $common, array(
1266 'about', 'property', 'resource', 'datatype', 'typeof',
1270 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1271 # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1272 $common = array_merge( $common, array(
1273 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1277 $block = array_merge( $common, array( 'align' ) );
1278 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1279 $tablecell = array( 'abbr',
1285 'nowrap', # deprecated
1286 'width', # deprecated
1287 'height', # deprecated
1288 'bgcolor' # deprecated
1291 # Numbers refer to sections in HTML 4.01 standard describing the element.
1292 # See: http://www.w3.org/TR/html4/
1293 $whitelist = array (
1296 'center' => $common, # deprecated
1297 'span' => $block, # ??
1315 'strong' => $common,
1326 'blockquote' => array_merge( $common, array( 'cite' ) ),
1337 'br' => array( 'id', 'class', 'title', 'style', 'clear' ),
1340 'pre' => array_merge( $common, array( 'width' ) ),
1343 'ins' => array_merge( $common, array( 'cite', 'datetime' ) ),
1344 'del' => array_merge( $common, array( 'cite', 'datetime' ) ),
1347 'ul' => array_merge( $common, array( 'type' ) ),
1348 'ol' => array_merge( $common, array( 'type', 'start' ) ),
1349 'li' => array_merge( $common, array( 'type', 'value' ) ),
1357 'table' => array_merge( $common,
1358 array( 'summary', 'width', 'border', 'frame',
1359 'rules', 'cellspacing', 'cellpadding',
1364 'caption' => array_merge( $common, array( 'align' ) ),
1367 'thead' => array_merge( $common, $tablealign ),
1368 'tfoot' => array_merge( $common, $tablealign ),
1369 'tbody' => array_merge( $common, $tablealign ),
1372 'colgroup' => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1373 'col' => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1376 'tr' => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1379 'td' => array_merge( $common, $tablecell, $tablealign ),
1380 'th' => array_merge( $common, $tablecell, $tablealign ),
1382 # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1383 'a' => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1386 # Not usually allowed, but may be used for extension-style hooks
1387 # such as <math> when it is rasterized
1388 'img' => array_merge( $common, array( 'alt' ) ),
1396 'strike' => $common,
1401 'font' => array_merge( $common, array( 'size', 'color', 'face' ) ),
1405 'hr' => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1407 # XHTML Ruby annotation text module, simple ruby only.
1408 # http://www.w3c.org/TR/ruby/
1413 'rt' => $common, #array_merge( $common, array( 'rbspan' ) ),
1416 # MathML root element, where used for extensions
1417 # 'title' may not be 100% valid here; it's XHTML
1418 # http://www.w3.org/TR/REC-MathML/
1419 'math' => array( 'class', 'style', 'id', 'title' ),
1425 * Take a fragment of (potentially invalid) HTML and return
1426 * a version with any tags removed, encoded as plain text.
1428 * Warning: this return value must be further escaped for literal
1429 * inclusion in HTML output as of 1.10!
1431 * @param $text String: HTML fragment
1434 static function stripAllTags( $text ) {
1436 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1438 # Normalize &entities and whitespace
1439 $text = self::decodeCharReferences( $text );
1440 $text = self::normalizeWhitespace( $text );
1446 * Hack up a private DOCTYPE with HTML's standard entity declarations.
1447 * PHP 4 seemed to know these if you gave it an HTML doctype, but
1450 * Use for passing XHTML fragments to PHP's XML parsing functions
1454 static function hackDocType() {
1455 global $wgHtmlEntities;
1456 $out = "<!DOCTYPE html [\n";
1457 foreach( $wgHtmlEntities as $entity => $codepoint ) {
1458 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1464 static function cleanUrl( $url ) {
1465 # Normalize any HTML entities in input. They will be
1466 # re-escaped by makeExternalLink().
1467 $url = Sanitizer::decodeCharReferences( $url );
1469 # Escape any control characters introduced by the above step
1470 $url = preg_replace( '/[\][<>"\\x00-\\x20\\x7F]/e', "urlencode('\\0')", $url );
1472 # Validate hostname portion
1474 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1475 list( /* $whole */, $protocol, $host, $rest ) = $matches;
1477 // Characters that will be ignored in IDNs.
1478 // http://tools.ietf.org/html/3454#section-3.1
1479 // Strip them before further processing so blacklists and such work.
1481 \\s| # general whitespace
1482 \xc2\xad| # 00ad SOFT HYPHEN
1483 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1484 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1485 \xe2\x81\xa0| # 2060 WORD JOINER
1486 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1487 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
1488 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1489 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1490 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1491 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1492 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1493 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe00f VARIATION SELECTOR-1-16
1496 $host = preg_replace( $strip, '', $host );
1498 // @todo Fixme: validate hostnames here
1500 return $protocol . $host . $rest;