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 global $wgAllowImageTag;
393 if ( $wgAllowImageTag ) {
394 $htmlsingle[] = 'img';
395 $htmlsingleonly[] = 'img';
398 $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
399 $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
401 # Convert them all to hashtables for faster lookup
402 $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
403 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
404 foreach ( $vars as $var ) {
405 $
$var = array_flip( $
$var );
407 $staticInitialised = true;
409 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
410 $extratags = array_flip( $extratags );
411 $removetags = array_flip( $removetags );
412 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
413 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ) , $removetags );
415 # Remove HTML comments
416 $text = Sanitizer
::removeHTMLcomments( $text );
417 $bits = explode( '<', $text );
418 $text = str_replace( '>', '>', array_shift( $bits ) );
420 $tagstack = $tablestack = array();
421 foreach ( $bits as $x ) {
423 # $slash: Does the current element start with a '/'?
424 # $t: Current element name
425 # $params: String between element name and >
426 # $brace: Ending '>' or '/>'
427 # $rest: Everything until the next element of $bits
428 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
429 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
431 $slash = $t = $params = $brace = $rest = null;
435 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
437 if ( $slash && isset( $htmlsingleonly[$t] ) ) {
439 } elseif ( $slash ) {
440 # Closing a tag... is it the one we just opened?
441 $ot = @array_pop
( $tagstack );
443 if ( isset( $htmlsingleallowed[$ot] ) ) {
444 # Pop all elements with an optional close tag
445 # and see if we find a match below them
447 array_push( $optstack, $ot );
448 $ot = @array_pop
( $tagstack );
449 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
450 array_push( $optstack, $ot );
451 $ot = @array_pop
( $tagstack );
454 # No match. Push the optional elements back again
456 while ( $ot = @array_pop
( $optstack ) ) {
457 array_push( $tagstack, $ot );
461 @array_push
( $tagstack, $ot );
462 # <li> can be nested in <ul> or <ol>, skip those cases:
463 if ( !isset( $htmllist[$ot] ) ||
!isset( $listtags[$t] ) ) {
468 if ( $t == 'table' ) {
469 $tagstack = array_pop( $tablestack );
474 # Keep track for later
475 if ( isset( $tabletags[$t] ) &&
476 !in_array( 'table', $tagstack ) ) {
478 } elseif ( in_array( $t, $tagstack ) &&
479 !isset( $htmlnest [$t ] ) ) {
481 #Â Is it a self closed htmlpair ? (bug 5487)
482 } elseif ( $brace == '/>' &&
483 isset( $htmlpairs[$t] ) ) {
485 } elseif ( isset( $htmlsingleonly[$t] ) ) {
486 # Hack to force empty tag for uncloseable elements
488 } elseif ( isset( $htmlsingle[$t] ) ) {
489 # Hack to not close $htmlsingle tags
491 } elseif ( isset( $tabletags[$t] )
492 && in_array( $t, $tagstack ) ) {
493 // New table tag but forgot to close the previous one
496 if ( $t == 'table' ) {
497 array_push( $tablestack, $tagstack );
500 array_push( $tagstack, $t );
503 # Replace any variables or template parameters with
505 if( is_callable( $processCallback ) ) {
506 call_user_func_array( $processCallback, array( &$params, $args ) );
509 # Strip non-approved attributes from the tag
510 $newparams = Sanitizer
::fixTagAttributes( $params, $t );
513 $rest = str_replace( '>', '>', $rest );
514 $close = ( $brace == '/>' && !$slash ) ?
' /' : '';
515 $text .= "<$slash$t$newparams$close>$rest";
519 $text .= '<' . str_replace( '>', '>', $x);
521 # Close off any remaining tags
522 while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
524 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
527 # this might be possible using tidy itself
528 foreach ( $bits as $x ) {
529 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
531 @list
( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
532 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
533 if( is_callable( $processCallback ) ) {
534 call_user_func_array( $processCallback, array( &$params, $args ) );
536 $newparams = Sanitizer
::fixTagAttributes( $params, $t );
537 $rest = str_replace( '>', '>', $rest );
538 $text .= "<$slash$t$newparams$brace$rest";
540 $text .= '<' . str_replace( '>', '>', $x);
544 wfProfileOut( __METHOD__
);
549 * Remove '<!--', '-->', and everything between.
550 * To avoid leaving blank lines, when a comment is both preceded
551 * and followed by a newline (ignoring spaces), trim leading and
552 * trailing spaces and one of the newlines.
555 * @param $text String
558 static function removeHTMLcomments( $text ) {
559 wfProfileIn( __METHOD__
);
560 while (($start = strpos($text, '<!--')) !== false) {
561 $end = strpos($text, '-->', $start +
4);
562 if ($end === false) {
563 # Unterminated comment; bail out
569 # Trim space and newline if the comment is both
570 # preceded and followed by a newline
571 $spaceStart = max($start - 1, 0);
572 $spaceLen = $end - $spaceStart;
573 while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
577 while (substr($text, $spaceStart +
$spaceLen, 1) === ' ')
579 if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart +
$spaceLen, 1) === "\n") {
580 # Remove the comment, leading and trailing
581 # spaces, and leave only one newline.
582 $text = substr_replace($text, "\n", $spaceStart, $spaceLen +
1);
585 # Remove just the comment.
586 $text = substr_replace($text, '', $start, $end - $start);
589 wfProfileOut( __METHOD__
);
594 * Take an array of attribute names and values and normalize or discard
595 * illegal values for the given element type.
597 * - Discards attributes not on a whitelist for the given element
598 * - Unsafe style attributes are discarded
599 * - Invalid id attributes are reencoded
601 * @param $attribs Array
602 * @param $element String
605 * @todo Check for legal values where the DTD limits things.
606 * @todo Check for unique id attribute :P
608 static function validateTagAttributes( $attribs, $element ) {
609 return Sanitizer
::validateAttributes( $attribs,
610 Sanitizer
::attributeWhitelist( $element ) );
614 * Take an array of attribute names and values and normalize or discard
615 * illegal values for the given whitelist.
617 * - Discards attributes not the given whitelist
618 * - Unsafe style attributes are discarded
619 * - Invalid id attributes are reencoded
621 * @param $attribs Array
622 * @param $whitelist Array: list of allowed attribute names
625 * @todo Check for legal values where the DTD limits things.
626 * @todo Check for unique id attribute :P
628 static function validateAttributes( $attribs, $whitelist ) {
629 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
631 $whitelist = array_flip( $whitelist );
632 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
635 foreach( $attribs as $attribute => $value ) {
636 #allow XML namespace declaration if RDFa is enabled
637 if ( $wgAllowRdfaAttributes && preg_match( MW_XMLNS_ATTRIBUTE_PATTRN
, $attribute ) ) {
638 if ( !preg_match( MW_EVIL_URI_PATTERN
, $value ) ) {
639 $out[$attribute] = $value;
645 if( !isset( $whitelist[$attribute] ) ) {
649 # Strip javascript "expression" from stylesheets.
650 # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
651 if( $attribute == 'style' ) {
652 $value = Sanitizer
::checkCss( $value );
653 if( $value === false ) {
659 if ( $attribute === 'id' ) {
660 $value = Sanitizer
::escapeId( $value, 'noninitial' );
663 //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
664 if ( $attribute === 'rel' ||
$attribute === 'rev' ||
665 $attribute === 'about' ||
$attribute === 'property' ||
$attribute === 'resource' ||
#RDFa
666 $attribute === 'datatype' ||
$attribute === 'typeof' ||
#RDFa
667 $attribute === 'itemid' ||
$attribute === 'itemprop' ||
$attribute === 'itemref' ||
#HTML5 microdata
668 $attribute === 'itemscope' ||
$attribute === 'itemtype' ) { #HTML5 microdata
670 //Paranoia. Allow "simple" values but suppress javascript
671 if ( preg_match( MW_EVIL_URI_PATTERN
, $value ) ) {
676 # NOTE: even though elements using href/src are not allowed directly, supply
677 # validation code that can be used by tag hook handlers, etc
678 if ( $attribute === 'href' ||
$attribute === 'src' ) {
679 if ( !preg_match( $hrefExp, $value ) ) {
680 continue; //drop any href or src attributes not using an allowed protocol.
681 //NOTE: this also drops all relative URLs
685 // If this attribute was previously set, override it.
686 // Output should only have one attribute of each name.
687 $out[$attribute] = $value;
690 if ( $wgAllowMicrodataAttributes ) {
691 # There are some complicated validity constraints we need to
692 # enforce here. First of all, we don't want to allow non-standard
694 $allowedTypes = array(
695 'http://microformats.org/profile/hcard',
696 'http://microformats.org/profile/hcalendar#vevent',
697 'http://n.whatwg.org/work',
699 if ( isset( $out['itemtype'] ) && !in_array( $out['itemtype'],
702 unset( $out['itemscope'] );
704 # itemtype, itemid, itemref don't make sense without itemscope
705 if ( !array_key_exists( 'itemscope', $out ) ) {
706 unset( $out['itemtype'] );
707 unset( $out['itemid'] );
708 unset( $out['itemref'] );
710 # TODO: Strip itemprop if we aren't descendants of an itemscope.
716 * Merge two sets of HTML attributes. Conflicting items in the second set
717 * will override those in the first, except for 'class' attributes which
718 * will be combined (if they're both strings).
720 * @todo implement merging for other attributes such as style
725 static function mergeAttributes( $a, $b ) {
726 $out = array_merge( $a, $b );
727 if( isset( $a['class'] ) && isset( $b['class'] )
728 && is_string( $a['class'] ) && is_string( $b['class'] )
729 && $a['class'] !== $b['class'] ) {
730 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
731 -1, PREG_SPLIT_NO_EMPTY
);
732 $out['class'] = implode( ' ', array_unique( $classes ) );
738 * Pick apart some CSS and check it for forbidden or unsafe structures.
739 * Returns a sanitized string, or false if it was just too evil.
741 * Currently URL references, 'expression', 'tps' are forbidden.
743 * @param $value String
746 static function checkCss( $value ) {
747 $value = Sanitizer
::decodeCharReferences( $value );
749 // Remove any comments; IE gets token splitting wrong
750 $value = StringUtils
::delimiterReplace( '/*', '*/', ' ', $value );
752 // Decode escape sequences and line continuation
753 // See the grammar in the CSS 2 spec, appendix D, Mozilla implements it accurately.
754 // IE 8 doesn't implement it at all, but there's no way to introduce url() into
755 // IE that doesn't hit Mozilla also.
757 if ( !$decodeRegex ) {
758 $space = '[\\x20\\t\\r\\n\\f]';
759 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
761 $decodeRegex = "/ $backslash
763 ($nl) | # 1. Line continuation
764 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
765 (.) # 3. backslash cancelling special meaning
768 $decoded = preg_replace_callback( $decodeRegex,
769 array( __CLASS__
, 'cssDecodeCallback' ), $value );
770 if ( preg_match( '!expression|https?://|url\s*\(!i', $decoded ) ) {
774 // Allowed, return CSS with comments stripped
779 static function cssDecodeCallback( $matches ) {
780 if ( $matches[1] !== '' ) {
782 } elseif ( $matches[2] !== '' ) {
783 return codepointToUtf8( hexdec( $matches[2] ) );
784 } elseif ( $matches[3] !== '' ) {
787 throw new MWException( __METHOD__
.': invalid match' );
792 * Take a tag soup fragment listing an HTML element's attributes
793 * and normalize it to well-formed XML, discarding unwanted attributes.
794 * Output is safe for further wikitext processing, with escaping of
795 * values that could trigger problems.
797 * - Normalizes attribute names to lowercase
798 * - Discards attributes not on a whitelist for the given element
799 * - Turns broken or invalid entities into plaintext
800 * - Double-quotes all attribute values
801 * - Attributes without values are given the name as attribute
802 * - Double attributes are discarded
803 * - Unsafe style attributes are discarded
804 * - Prepends space if there are attributes.
806 * @param $text String
807 * @param $element String
810 static function fixTagAttributes( $text, $element ) {
811 if( trim( $text ) == '' ) {
815 $stripped = Sanitizer
::validateTagAttributes(
816 Sanitizer
::decodeTagAttributes( $text ), $element );
819 foreach( $stripped as $attribute => $value ) {
820 $encAttribute = htmlspecialchars( $attribute );
821 $encValue = Sanitizer
::safeEncodeAttribute( $value );
823 $attribs[] = "$encAttribute=\"$encValue\"";
825 return count( $attribs ) ?
' ' . implode( ' ', $attribs ) : '';
829 * Encode an attribute value for HTML output.
830 * @param $text String
831 * @return HTML-encoded text fragment
833 static function encodeAttribute( $text ) {
834 $encValue = htmlspecialchars( $text, ENT_QUOTES
);
836 // Whitespace is normalized during attribute decoding,
837 // so if we've been passed non-spaces we must encode them
838 // ahead of time or they won't be preserved.
839 $encValue = strtr( $encValue, array(
849 * Encode an attribute value for HTML tags, with extra armoring
850 * against further wiki processing.
851 * @param $text String
852 * @return HTML-encoded text fragment
854 static function safeEncodeAttribute( $text ) {
855 $encValue = Sanitizer
::encodeAttribute( $text );
857 # Templates and links may be expanded in later parsing,
858 # creating invalid or dangerous output. Suppress this.
859 $encValue = strtr( $encValue, array(
860 '<' => '<', // This should never happen,
861 '>' => '>', // we've received invalid input
862 '"' => '"', // which should have been escaped.
865 "''" => '''',
866 'ISBN' => 'ISBN',
868 'PMID' => 'PMID',
874 $encValue = preg_replace_callback(
875 '/(' . wfUrlProtocols() . ')/',
876 array( 'Sanitizer', 'armorLinksCallback' ),
882 * Given a value, escape it so that it can be used in an id attribute and
883 * return it. This will use HTML5 validation if $wgExperimentalHtmlIds is
884 * true, allowing anything but ASCII whitespace. Otherwise it will use
885 * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
886 * escaped with lots of dots.
888 * To ensure we don't have to bother escaping anything, we also strip ', ",
889 * & even if $wgExperimentalIds is true. TODO: Is this the best tactic?
890 * We also strip # because it upsets IE6.
892 * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
895 * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
896 * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
897 * HTML5 definition of id attribute
899 * @param $id String: id to escape
900 * @param $options Mixed: string or array of strings (default is array()):
901 * 'noninitial': This is a non-initial fragment of an id, not a full id,
902 * so don't pay attention if the first character isn't valid at the
903 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
905 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
906 * if $wgExperimentalHtmlIds is used, so we can generate extra
907 * anchors and links won't break.
910 static function escapeId( $id, $options = array() ) {
911 global $wgHtml5, $wgExperimentalHtmlIds;
912 $options = (array)$options;
914 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
915 $id = Sanitizer
::decodeCharReferences( $id );
916 $id = preg_replace( '/[ \t\n\r\f_\'"&#]+/', '_', $id );
917 $id = trim( $id, '_' );
919 # Must have been all whitespace to start with.
926 # HTML4-style escaping
927 static $replace = array(
932 $id = urlencode( Sanitizer
::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
933 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
935 if ( !preg_match( '/^[a-zA-Z]/', $id )
936 && !in_array( 'noninitial', $options ) ) {
937 // Initial character must be a letter!
944 * Given a value, escape it so that it can be used as a CSS class and
947 * @todo For extra validity, input should be validated UTF-8.
949 * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
951 * @param $class String
954 static function escapeClass( $class ) {
955 // Convert ugly stuff to underscores and kill underscores in ugly places
956 return rtrim(preg_replace(
957 array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
963 * Given HTML input, escape with htmlspecialchars but un-escape entites.
964 * This allows (generally harmless) entities like to survive.
966 * @param $html String to escape
967 * @return String: escaped input
969 static function escapeHtmlAllowEntities( $html ) {
970 # It seems wise to escape ' as well as ", as a matter of course. Can't
972 $html = htmlspecialchars( $html, ENT_QUOTES
);
973 $html = str_replace( '&', '&', $html );
974 $html = Sanitizer
::normalizeCharReferences( $html );
979 * Regex replace callback for armoring links against further processing.
980 * @param $matches Array
983 private static function armorLinksCallback( $matches ) {
984 return str_replace( ':', ':', $matches[1] );
988 * Return an associative array of attribute names and values from
989 * a partial tag string. Attribute names are forces to lowercase,
990 * character references are decoded to UTF-8 text.
992 * @param $text String
995 public static function decodeTagAttributes( $text ) {
996 if( trim( $text ) == '' ) {
1002 if( !preg_match_all(
1006 PREG_SET_ORDER
) ) {
1010 foreach( $pairs as $set ) {
1011 $attribute = strtolower( $set[1] );
1012 $value = Sanitizer
::getTagAttributeCallback( $set );
1014 // Normalize whitespace
1015 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1016 $value = trim( $value );
1018 // Decode character references
1019 $attribs[$attribute] = Sanitizer
::decodeCharReferences( $value );
1025 * Pick the appropriate attribute value from a match set from the
1026 * MW_ATTRIBS_REGEX matches.
1031 private static function getTagAttributeCallback( $set ) {
1032 if( isset( $set[6] ) ) {
1033 # Illegal #XXXXXX color with no quotes.
1035 } elseif( isset( $set[5] ) ) {
1038 } elseif( isset( $set[4] ) ) {
1041 } elseif( isset( $set[3] ) ) {
1044 } elseif( !isset( $set[2] ) ) {
1045 # In XHTML, attributes must have a value.
1046 # For 'reduced' form, return explicitly the attribute name here.
1049 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1054 * Normalize whitespace and character references in an XML source-
1055 * encoded text for an attribute value.
1057 * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1058 * but note that we're not returning the value, but are returning
1059 * XML source fragments that will be slapped into output.
1061 * @param $text String
1064 private static function normalizeAttributeValue( $text ) {
1065 return str_replace( '"', '"',
1066 self
::normalizeWhitespace(
1067 Sanitizer
::normalizeCharReferences( $text ) ) );
1070 private static function normalizeWhitespace( $text ) {
1071 return preg_replace(
1072 '/\r\n|[\x20\x0d\x0a\x09]/',
1078 * Ensure that any entities and character references are legal
1079 * for XML and XHTML specifically. Any stray bits will be
1080 * &-escaped to result in a valid text fragment.
1082 * a. any named char refs must be known in XHTML
1083 * b. any numeric char refs must be legal chars, not invalid or forbidden
1084 * c. use &#x, not &#X
1085 * d. fix or reject non-valid attributes
1087 * @param $text String
1091 static function normalizeCharReferences( $text ) {
1092 return preg_replace_callback(
1094 array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1098 * @param $matches String
1101 static function normalizeCharReferencesCallback( $matches ) {
1103 if( $matches[1] != '' ) {
1104 $ret = Sanitizer
::normalizeEntity( $matches[1] );
1105 } elseif( $matches[2] != '' ) {
1106 $ret = Sanitizer
::decCharReference( $matches[2] );
1107 } elseif( $matches[3] != '' ) {
1108 $ret = Sanitizer
::hexCharReference( $matches[3] );
1109 } elseif( $matches[4] != '' ) {
1110 $ret = Sanitizer
::hexCharReference( $matches[4] );
1112 if( is_null( $ret ) ) {
1113 return htmlspecialchars( $matches[0] );
1120 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1121 * return the named entity reference as is. If the entity is a
1122 * MediaWiki-specific alias, returns the HTML equivalent. Otherwise,
1123 * returns HTML-escaped text of pseudo-entity source (eg &foo;)
1125 * @param $name String
1128 static function normalizeEntity( $name ) {
1129 global $wgHtmlEntities, $wgHtmlEntityAliases;
1130 if ( isset( $wgHtmlEntityAliases[$name] ) ) {
1131 return "&{$wgHtmlEntityAliases[$name]};";
1132 } elseif( isset( $wgHtmlEntities[$name] ) ) {
1135 return "&$name;";
1139 static function decCharReference( $codepoint ) {
1140 $point = intval( $codepoint );
1141 if( Sanitizer
::validateCodepoint( $point ) ) {
1142 return sprintf( '&#%d;', $point );
1148 static function hexCharReference( $codepoint ) {
1149 $point = hexdec( $codepoint );
1150 if( Sanitizer
::validateCodepoint( $point ) ) {
1151 return sprintf( '&#x%x;', $point );
1158 * Returns true if a given Unicode codepoint is a valid character in XML.
1159 * @param $codepoint Integer
1162 private static function validateCodepoint( $codepoint ) {
1163 return ($codepoint == 0x09)
1164 ||
($codepoint == 0x0a)
1165 ||
($codepoint == 0x0d)
1166 ||
($codepoint >= 0x20 && $codepoint <= 0xd7ff)
1167 ||
($codepoint >= 0xe000 && $codepoint <= 0xfffd)
1168 ||
($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1172 * Decode any character references, numeric or named entities,
1173 * in the text and return a UTF-8 string.
1175 * @param $text String
1178 public static function decodeCharReferences( $text ) {
1179 return preg_replace_callback(
1181 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1186 * Decode any character references, numeric or named entities,
1187 * in the next and normalize the resulting string. (bug 14952)
1189 * This is useful for page titles, not for text to be displayed,
1190 * MediaWiki allows HTML entities to escape normalization as a feature.
1192 * @param $text String (already normalized, containing entities)
1193 * @return String (still normalized, without entities)
1195 public static function decodeCharReferencesAndNormalize( $text ) {
1197 $text = preg_replace_callback(
1199 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1200 $text, /* limit */ -1, $count );
1203 return $wgContLang->normalize( $text );
1210 * @param $matches String
1213 static function decodeCharReferencesCallback( $matches ) {
1214 if( $matches[1] != '' ) {
1215 return Sanitizer
::decodeEntity( $matches[1] );
1216 } elseif( $matches[2] != '' ) {
1217 return Sanitizer
::decodeChar( intval( $matches[2] ) );
1218 } elseif( $matches[3] != '' ) {
1219 return Sanitizer
::decodeChar( hexdec( $matches[3] ) );
1220 } elseif( $matches[4] != '' ) {
1221 return Sanitizer
::decodeChar( hexdec( $matches[4] ) );
1223 # Last case should be an ampersand by itself
1228 * Return UTF-8 string for a codepoint if that is a valid
1229 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1230 * @param $codepoint Integer
1234 static function decodeChar( $codepoint ) {
1235 if( Sanitizer
::validateCodepoint( $codepoint ) ) {
1236 return codepointToUtf8( $codepoint );
1238 return UTF8_REPLACEMENT
;
1243 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1244 * return the UTF-8 encoding of that character. Otherwise, returns
1245 * pseudo-entity source (eg &foo;)
1247 * @param $name Strings
1250 static function decodeEntity( $name ) {
1251 global $wgHtmlEntities, $wgHtmlEntityAliases;
1252 if ( isset( $wgHtmlEntityAliases[$name] ) ) {
1253 $name = $wgHtmlEntityAliases[$name];
1255 if( isset( $wgHtmlEntities[$name] ) ) {
1256 return codepointToUtf8( $wgHtmlEntities[$name] );
1263 * Fetch the whitelist of acceptable attributes for a given element name.
1265 * @param $element String
1268 static function attributeWhitelist( $element ) {
1270 if( !isset( $list ) ) {
1271 $list = Sanitizer
::setupAttributeWhitelist();
1273 return isset( $list[$element] )
1279 * Foreach array key (an allowed HTML element), return an array
1280 * of allowed attributes
1283 static function setupAttributeWhitelist() {
1284 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1286 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1288 if ( $wgAllowRdfaAttributes ) {
1289 #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1290 $common = array_merge( $common, array(
1291 'about', 'property', 'resource', 'datatype', 'typeof',
1295 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1296 # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1297 $common = array_merge( $common, array(
1298 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1302 $block = array_merge( $common, array( 'align' ) );
1303 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1304 $tablecell = array( 'abbr',
1310 'nowrap', # deprecated
1311 'width', # deprecated
1312 'height', # deprecated
1313 'bgcolor' # deprecated
1316 # Numbers refer to sections in HTML 4.01 standard describing the element.
1317 # See: http://www.w3.org/TR/html4/
1318 $whitelist = array (
1321 'center' => $common, # deprecated
1322 'span' => $block, # ??
1340 'strong' => $common,
1351 'blockquote' => array_merge( $common, array( 'cite' ) ),
1362 'br' => array( 'id', 'class', 'title', 'style', 'clear' ),
1365 'pre' => array_merge( $common, array( 'width' ) ),
1368 'ins' => array_merge( $common, array( 'cite', 'datetime' ) ),
1369 'del' => array_merge( $common, array( 'cite', 'datetime' ) ),
1372 'ul' => array_merge( $common, array( 'type' ) ),
1373 'ol' => array_merge( $common, array( 'type', 'start' ) ),
1374 'li' => array_merge( $common, array( 'type', 'value' ) ),
1382 'table' => array_merge( $common,
1383 array( 'summary', 'width', 'border', 'frame',
1384 'rules', 'cellspacing', 'cellpadding',
1389 'caption' => array_merge( $common, array( 'align' ) ),
1392 'thead' => array_merge( $common, $tablealign ),
1393 'tfoot' => array_merge( $common, $tablealign ),
1394 'tbody' => array_merge( $common, $tablealign ),
1397 'colgroup' => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1398 'col' => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1401 'tr' => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1404 'td' => array_merge( $common, $tablecell, $tablealign ),
1405 'th' => array_merge( $common, $tablecell, $tablealign ),
1407 # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1408 'a' => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1411 # Not usually allowed, but may be used for extension-style hooks
1412 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1414 'img' => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1422 'strike' => $common,
1427 'font' => array_merge( $common, array( 'size', 'color', 'face' ) ),
1431 'hr' => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1433 # XHTML Ruby annotation text module, simple ruby only.
1434 # http://www.w3c.org/TR/ruby/
1439 'rt' => $common, #array_merge( $common, array( 'rbspan' ) ),
1442 # MathML root element, where used for extensions
1443 # 'title' may not be 100% valid here; it's XHTML
1444 # http://www.w3.org/TR/REC-MathML/
1445 'math' => array( 'class', 'style', 'id', 'title' ),
1451 * Take a fragment of (potentially invalid) HTML and return
1452 * a version with any tags removed, encoded as plain text.
1454 * Warning: this return value must be further escaped for literal
1455 * inclusion in HTML output as of 1.10!
1457 * @param $text String: HTML fragment
1460 static function stripAllTags( $text ) {
1462 $text = StringUtils
::delimiterReplace( '<', '>', '', $text );
1464 # Normalize &entities and whitespace
1465 $text = self
::decodeCharReferences( $text );
1466 $text = self
::normalizeWhitespace( $text );
1472 * Hack up a private DOCTYPE with HTML's standard entity declarations.
1473 * PHP 4 seemed to know these if you gave it an HTML doctype, but
1476 * Use for passing XHTML fragments to PHP's XML parsing functions
1480 static function hackDocType() {
1481 global $wgHtmlEntities;
1482 $out = "<!DOCTYPE html [\n";
1483 foreach( $wgHtmlEntities as $entity => $codepoint ) {
1484 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1490 static function cleanUrl( $url ) {
1491 # Normalize any HTML entities in input. They will be
1492 # re-escaped by makeExternalLink().
1493 $url = Sanitizer
::decodeCharReferences( $url );
1495 # Escape any control characters introduced by the above step
1496 $url = preg_replace( '/[\][<>"\\x00-\\x20\\x7F]/e', "urlencode('\\0')", $url );
1498 # Validate hostname portion
1500 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1501 list( /* $whole */, $protocol, $host, $rest ) = $matches;
1503 // Characters that will be ignored in IDNs.
1504 // http://tools.ietf.org/html/3454#section-3.1
1505 // Strip them before further processing so blacklists and such work.
1507 \\s| # general whitespace
1508 \xc2\xad| # 00ad SOFT HYPHEN
1509 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1510 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1511 \xe2\x81\xa0| # 2060 WORD JOINER
1512 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1513 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
1514 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1515 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1516 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1517 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1518 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1519 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe00f VARIATION SELECTOR-1-16
1522 $host = preg_replace( $strip, '', $host );
1524 // @todo Fixme: validate hostnames here
1526 return $protocol . $host . $rest;