3 * HTML sanitizer for %MediaWiki.
5 * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
6 * https://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 * HTML 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.
58 private static $htmlEntities = array(
74 'apos' => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
315 * Character entity aliases accepted by MediaWiki
317 private static $htmlEntityAliases = array(
323 * Lazy-initialised attributes regex, see getAttribsRegex()
325 private static $attribsRegex;
328 * Regular expression to match HTML/XML attribute pairs within a tag.
329 * Allows some... latitude.
330 * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
333 static function getAttribsRegex() {
334 if ( self
::$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 string $text
360 * @param callable $processCallback Callback to do any variable or parameter
361 * replacements in HTML attribute values
362 * @param array|bool $args Arguments for the processing callback
363 * @param array $extratags For any extra tags to include
364 * @param array $removetags For any tags (default or extra) to exclude
367 static function removeHTMLtags( $text, $processCallback = null,
368 $args = array(), $extratags = array(), $removetags = array()
370 global $wgUseTidy, $wgAllowMicrodataAttributes, $wgAllowImageTag;
372 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
373 $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
375 wfProfileIn( __METHOD__
);
377 // Base our staticInitialised variable off of the global config state so that if the globals
378 // are changed (like in the screwed up test system) we will re-initialise the settings.
379 $globalContext = implode( '-', compact( 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
380 if ( !$staticInitialised ||
$staticInitialised != $globalContext ) {
382 $htmlpairsStatic = array( # Tags that must be closed
383 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
384 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
385 'strike', 'strong', 'tt', 'var', 'div', 'center',
386 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
387 'ruby', 'rb', 'rp', 'rt', 'rtc', 'p', 'span', 'abbr', 'dfn',
388 'kbd', 'samp', 'data', 'time', 'mark'
391 'br', 'wbr', 'hr', 'li', 'dt', 'dd'
393 $htmlsingleonly = array( # Elements that cannot have close tags
396 if ( $wgAllowMicrodataAttributes ) {
397 $htmlsingle[] = $htmlsingleonly[] = 'meta';
398 $htmlsingle[] = $htmlsingleonly[] = 'link';
400 $htmlnest = array( # Tags that can be nested--??
401 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
402 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
403 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
405 $tabletags = array( # Can only appear inside table, we will close them
408 $htmllist = array( # Tags used by list
411 $listtags = array( # Tags that can appear in a list
415 if ( $wgAllowImageTag ) {
416 $htmlsingle[] = 'img';
417 $htmlsingleonly[] = 'img';
420 $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
421 $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
423 # Convert them all to hashtables for faster lookup
424 $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
425 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
426 foreach ( $vars as $var ) {
427 $
$var = array_flip( $
$var );
429 $staticInitialised = $globalContext;
431 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
432 $extratags = array_flip( $extratags );
433 $removetags = array_flip( $removetags );
434 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
435 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
437 # Remove HTML comments
438 $text = Sanitizer
::removeHTMLcomments( $text );
439 $bits = explode( '<', $text );
440 $text = str_replace( '>', '>', array_shift( $bits ) );
442 $tagstack = $tablestack = array();
443 foreach ( $bits as $x ) {
445 # $slash: Does the current element start with a '/'?
446 # $t: Current element name
447 # $params: String between element name and >
448 # $brace: Ending '>' or '/>'
449 # $rest: Everything until the next element of $bits
450 if ( preg_match( '!^(/?)([^\\s/>]+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
451 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
453 $slash = $t = $params = $brace = $rest = null;
457 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
459 if ( $slash && isset( $htmlsingleonly[$t] ) ) {
461 } elseif ( $slash ) {
462 # Closing a tag... is it the one we just opened?
463 wfSuppressWarnings();
464 $ot = array_pop( $tagstack );
468 if ( isset( $htmlsingleallowed[$ot] ) ) {
469 # Pop all elements with an optional close tag
470 # and see if we find a match below them
472 array_push( $optstack, $ot );
473 wfSuppressWarnings();
474 $ot = array_pop( $tagstack );
476 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
477 array_push( $optstack, $ot );
478 wfSuppressWarnings();
479 $ot = array_pop( $tagstack );
483 # No match. Push the optional elements back again
485 wfSuppressWarnings();
486 $ot = array_pop( $optstack );
489 array_push( $tagstack, $ot );
490 wfSuppressWarnings();
491 $ot = array_pop( $optstack );
496 wfSuppressWarnings();
497 array_push( $tagstack, $ot );
500 # <li> can be nested in <ul> or <ol>, skip those cases:
501 if ( !isset( $htmllist[$ot] ) ||
!isset( $listtags[$t] ) ) {
506 if ( $t == 'table' ) {
507 $tagstack = array_pop( $tablestack );
512 # Keep track for later
513 if ( isset( $tabletags[$t] ) &&
514 !in_array( 'table', $tagstack ) ) {
516 } elseif ( in_array( $t, $tagstack ) &&
517 !isset( $htmlnest[$t] ) ) {
519 # Is it a self closed htmlpair ? (bug 5487)
520 } elseif ( $brace == '/>' &&
521 isset( $htmlpairs[$t] ) ) {
523 } elseif ( isset( $htmlsingleonly[$t] ) ) {
524 # Hack to force empty tag for unclosable elements
526 } elseif ( isset( $htmlsingle[$t] ) ) {
527 # Hack to not close $htmlsingle tags
529 # Still need to push this optionally-closed tag to
530 # the tag stack so that we can match end tags
531 # instead of marking them as bad.
532 array_push( $tagstack, $t );
533 } elseif ( isset( $tabletags[$t] )
534 && in_array( $t, $tagstack ) ) {
535 // New table tag but forgot to close the previous one
538 if ( $t == 'table' ) {
539 array_push( $tablestack, $tagstack );
542 array_push( $tagstack, $t );
545 # Replace any variables or template parameters with
547 if ( is_callable( $processCallback ) ) {
548 call_user_func_array( $processCallback, array( &$params, $args ) );
551 if ( !Sanitizer
::validateTag( $params, $t ) ) {
555 # Strip non-approved attributes from the tag
556 $newparams = Sanitizer
::fixTagAttributes( $params, $t );
559 $rest = str_replace( '>', '>', $rest );
560 $close = ( $brace == '/>' && !$slash ) ?
' /' : '';
561 $text .= "<$slash$t$newparams$close>$rest";
565 $text .= '<' . str_replace( '>', '>', $x );
567 # Close off any remaining tags
568 while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
570 if ( $t == 'table' ) {
571 $tagstack = array_pop( $tablestack );
575 # this might be possible using tidy itself
576 foreach ( $bits as $x ) {
578 '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
583 wfSuppressWarnings();
584 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
588 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
589 if ( is_callable( $processCallback ) ) {
590 call_user_func_array( $processCallback, array( &$params, $args ) );
593 if ( !Sanitizer
::validateTag( $params, $t ) ) {
597 $newparams = Sanitizer
::fixTagAttributes( $params, $t );
599 $rest = str_replace( '>', '>', $rest );
600 $text .= "<$slash$t$newparams$brace$rest";
604 $text .= '<' . str_replace( '>', '>', $x );
607 wfProfileOut( __METHOD__
);
612 * Remove '<!--', '-->', and everything between.
613 * To avoid leaving blank lines, when a comment is both preceded
614 * and followed by a newline (ignoring spaces), trim leading and
615 * trailing spaces and one of the newlines.
618 * @param string $text
621 static function removeHTMLcomments( $text ) {
622 wfProfileIn( __METHOD__
);
623 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
624 $end = strpos( $text, '-->', $start +
4 );
625 if ( $end === false ) {
626 # Unterminated comment; bail out
632 # Trim space and newline if the comment is both
633 # preceded and followed by a newline
634 $spaceStart = max( $start - 1, 0 );
635 $spaceLen = $end - $spaceStart;
636 while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
640 while ( substr( $text, $spaceStart +
$spaceLen, 1 ) === ' ' ) {
643 if ( substr( $text, $spaceStart, 1 ) === "\n"
644 && substr( $text, $spaceStart +
$spaceLen, 1 ) === "\n" ) {
645 # Remove the comment, leading and trailing
646 # spaces, and leave only one newline.
647 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen +
1 );
649 # Remove just the comment.
650 $text = substr_replace( $text, '', $start, $end - $start );
653 wfProfileOut( __METHOD__
);
658 * Takes attribute names and values for a tag and the tag name and
659 * validates that the tag is allowed to be present.
660 * This DOES NOT validate the attributes, nor does it validate the
661 * tags themselves. This method only handles the special circumstances
662 * where we may want to allow a tag within content but ONLY when it has
663 * specific attributes set.
665 * @param string $params
666 * @param string $element
669 static function validateTag( $params, $element ) {
670 $params = Sanitizer
::decodeTagAttributes( $params );
672 if ( $element == 'meta' ||
$element == 'link' ) {
673 if ( !isset( $params['itemprop'] ) ) {
674 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
677 if ( $element == 'meta' && !isset( $params['content'] ) ) {
678 // <meta> must have a content="" for the itemprop
681 if ( $element == 'link' && !isset( $params['href'] ) ) {
682 // <link> must have an associated href=""
691 * Take an array of attribute names and values and normalize or discard
692 * illegal values for the given element type.
694 * - Discards attributes not on a whitelist for the given element
695 * - Unsafe style attributes are discarded
696 * - Invalid id attributes are re-encoded
698 * @param array $attribs
699 * @param string $element
702 * @todo Check for legal values where the DTD limits things.
703 * @todo Check for unique id attribute :P
705 static function validateTagAttributes( $attribs, $element ) {
706 return Sanitizer
::validateAttributes( $attribs,
707 Sanitizer
::attributeWhitelist( $element ) );
711 * Take an array of attribute names and values and normalize or discard
712 * illegal values for the given whitelist.
714 * - Discards attributes not the given whitelist
715 * - Unsafe style attributes are discarded
716 * - Invalid id attributes are re-encoded
718 * @param array $attribs
719 * @param array $whitelist List of allowed attribute names
722 * @todo Check for legal values where the DTD limits things.
723 * @todo Check for unique id attribute :P
725 static function validateAttributes( $attribs, $whitelist ) {
726 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
728 $whitelist = array_flip( $whitelist );
729 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
732 foreach ( $attribs as $attribute => $value ) {
733 #allow XML namespace declaration if RDFa is enabled
734 if ( $wgAllowRdfaAttributes && preg_match( self
::XMLNS_ATTRIBUTE_PATTERN
, $attribute ) ) {
735 if ( !preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
736 $out[$attribute] = $value;
742 # Allow any attribute beginning with "data-"
743 if ( !preg_match( '/^data-/i', $attribute ) && !isset( $whitelist[$attribute] ) ) {
747 # Strip javascript "expression" from stylesheets.
748 # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
749 if ( $attribute == 'style' ) {
750 $value = Sanitizer
::checkCss( $value );
753 if ( $attribute === 'id' ) {
754 $value = Sanitizer
::escapeId( $value, 'noninitial' );
758 # http://www.w3.org/TR/wai-aria/
759 # http://www.whatwg.org/html/elements.html#wai-aria
760 # For now we only support role="presentation" until we work out what roles should be
761 # usable by content and we ensure that our code explicitly rejects patterns that
762 # violate HTML5's ARIA restrictions.
763 if ( $attribute === 'role' && $value !== 'presentation' ) {
767 // RDFa and microdata properties allow URLs, URIs and/or CURIs.
768 // Check them for sanity.
769 if ( $attribute === 'rel' ||
$attribute === 'rev'
771 ||
$attribute === 'about' ||
$attribute === 'property'
772 ||
$attribute === 'resource' ||
$attribute === 'datatype'
773 ||
$attribute === 'typeof'
775 ||
$attribute === 'itemid' ||
$attribute === 'itemprop'
776 ||
$attribute === 'itemref' ||
$attribute === 'itemscope'
777 ||
$attribute === 'itemtype'
779 //Paranoia. Allow "simple" values but suppress javascript
780 if ( preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
785 # NOTE: even though elements using href/src are not allowed directly, supply
786 # validation code that can be used by tag hook handlers, etc
787 if ( $attribute === 'href' ||
$attribute === 'src' ) {
788 if ( !preg_match( $hrefExp, $value ) ) {
789 continue; //drop any href or src attributes not using an allowed protocol.
790 // NOTE: this also drops all relative URLs
794 // If this attribute was previously set, override it.
795 // Output should only have one attribute of each name.
796 $out[$attribute] = $value;
799 if ( $wgAllowMicrodataAttributes ) {
800 # itemtype, itemid, itemref don't make sense without itemscope
801 if ( !array_key_exists( 'itemscope', $out ) ) {
802 unset( $out['itemtype'] );
803 unset( $out['itemid'] );
804 unset( $out['itemref'] );
806 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
812 * Merge two sets of HTML attributes. Conflicting items in the second set
813 * will override those in the first, except for 'class' attributes which
814 * will be combined (if they're both strings).
816 * @todo implement merging for other attributes such as style
821 static function mergeAttributes( $a, $b ) {
822 $out = array_merge( $a, $b );
823 if ( isset( $a['class'] ) && isset( $b['class'] )
824 && is_string( $a['class'] ) && is_string( $b['class'] )
825 && $a['class'] !== $b['class']
827 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
828 -1, PREG_SPLIT_NO_EMPTY
);
829 $out['class'] = implode( ' ', array_unique( $classes ) );
835 * Pick apart some CSS and check it for forbidden or unsafe structures.
836 * Returns a sanitized string. This sanitized string will have
837 * character references and escape sequences decoded and comments
838 * stripped (unless it is itself one valid comment, in which case the value
839 * will be passed through). If the input is just too evil, only a comment
840 * complaining about evilness will be returned.
842 * Currently URL references, 'expression', 'tps' are forbidden.
844 * NOTE: Despite the fact that character references are decoded, the
845 * returned string may contain character references given certain
846 * clever input strings. These character references must
847 * be escaped before the return value is embedded in HTML.
849 * @param string $value
852 static function checkCss( $value ) {
853 // Decode character references like {
854 $value = Sanitizer
::decodeCharReferences( $value );
856 // Decode escape sequences and line continuation
857 // See the grammar in the CSS 2 spec, appendix D.
858 // This has to be done AFTER decoding character references.
859 // This means it isn't possible for this function to return
860 // unsanitized escape sequences. It is possible to manufacture
861 // input that contains character references that decode to
862 // escape sequences that decode to character references, but
863 // it's OK for the return value to contain character references
864 // because the caller is supposed to escape those anyway.
866 if ( !$decodeRegex ) {
867 $space = '[\\x20\\t\\r\\n\\f]';
868 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
870 $decodeRegex = "/ $backslash
872 ($nl) | # 1. Line continuation
873 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
874 (.) | # 3. backslash cancelling special meaning
875 () | # 4. backslash at end of string
878 $value = preg_replace_callback( $decodeRegex,
879 array( __CLASS__
, 'cssDecodeCallback' ), $value );
881 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
882 $value = preg_replace_callback(
883 '/[!-[]-z]/u', // U+FF01 to U+FF5A, excluding U+FF3C (bug 58088)
884 function ( $matches ) {
885 $cp = utf8ToCodepoint( $matches[0] );
886 if ( $cp === false ) {
889 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
894 // Convert more characters IE6 might treat as ascii
895 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
896 $value = str_replace(
897 array( 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ),
898 array( 'r', 'n', 'n', 'l', 'i', '(', '(' ),
902 // Let the value through if it's nothing but a single comment, to
903 // allow other functions which may reject it to pass some error
905 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
906 // Remove any comments; IE gets token splitting wrong
907 // This must be done AFTER decoding character references and
908 // escape sequences, because those steps can introduce comments
909 // This step cannot introduce character references or escape
910 // sequences, because it replaces comments with spaces rather
911 // than removing them completely.
912 $value = StringUtils
::delimiterReplace( '/*', '*/', ' ', $value );
914 // Remove anything after a comment-start token, to guard against
915 // incorrect client implementations.
916 $commentPos = strpos( $value, '/*' );
917 if ( $commentPos !== false ) {
918 $value = substr( $value, 0, $commentPos );
922 // S followed by repeat, iteration, or prolonged sound marks,
923 // which IE will treat as "ss"
924 $value = preg_replace(
926 \xE3\x80\xB1 | # U+3031
927 \xE3\x82\x9D | # U+309D
928 \xE3\x83\xBC | # U+30FC
929 \xE3\x83\xBD | # U+30FD
930 \xEF\xB9\xBC | # U+FE7C
931 \xEF\xB9\xBD | # U+FE7D
932 \xEF\xBD\xB0 # U+FF70
938 // Reject problematic keywords and control characters
939 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ) {
940 return '/* invalid control char */';
941 } elseif ( preg_match(
952 return '/* insecure input */';
958 * @param array $matches
961 static function cssDecodeCallback( $matches ) {
962 if ( $matches[1] !== '' ) {
965 } elseif ( $matches[2] !== '' ) {
966 $char = codepointToUtf8( hexdec( $matches[2] ) );
967 } elseif ( $matches[3] !== '' ) {
972 if ( $char == "\n" ||
$char == '"' ||
$char == "'" ||
$char == '\\' ) {
973 // These characters need to be escaped in strings
974 // Clean up the escape sequence to avoid parsing errors by clients
975 return '\\' . dechex( ord( $char ) ) . ' ';
977 // Decode unnecessary escape
983 * Take a tag soup fragment listing an HTML element's attributes
984 * and normalize it to well-formed XML, discarding unwanted attributes.
985 * Output is safe for further wikitext processing, with escaping of
986 * values that could trigger problems.
988 * - Normalizes attribute names to lowercase
989 * - Discards attributes not on a whitelist for the given element
990 * - Turns broken or invalid entities into plaintext
991 * - Double-quotes all attribute values
992 * - Attributes without values are given the name as attribute
993 * - Double attributes are discarded
994 * - Unsafe style attributes are discarded
995 * - Prepends space if there are attributes.
997 * @param string $text
998 * @param string $element
1001 static function fixTagAttributes( $text, $element ) {
1002 if ( trim( $text ) == '' ) {
1006 $decoded = Sanitizer
::decodeTagAttributes( $text );
1007 $stripped = Sanitizer
::validateTagAttributes( $decoded, $element );
1009 return Sanitizer
::safeEncodeTagAttributes( $stripped );
1013 * Encode an attribute value for HTML output.
1014 * @param string $text
1015 * @return string HTML-encoded text fragment
1017 static function encodeAttribute( $text ) {
1018 $encValue = htmlspecialchars( $text, ENT_QUOTES
);
1020 // Whitespace is normalized during attribute decoding,
1021 // so if we've been passed non-spaces we must encode them
1022 // ahead of time or they won't be preserved.
1023 $encValue = strtr( $encValue, array(
1033 * Encode an attribute value for HTML tags, with extra armoring
1034 * against further wiki processing.
1035 * @param string $text
1036 * @return string HTML-encoded text fragment
1038 static function safeEncodeAttribute( $text ) {
1039 $encValue = Sanitizer
::encodeAttribute( $text );
1041 # Templates and links may be expanded in later parsing,
1042 # creating invalid or dangerous output. Suppress this.
1043 $encValue = strtr( $encValue, array(
1044 '<' => '<', // This should never happen,
1045 '>' => '>', // we've received invalid input
1046 '"' => '"', // which should have been escaped.
1049 "''" => '''',
1050 'ISBN' => 'ISBN',
1052 'PMID' => 'PMID',
1058 $encValue = preg_replace_callback(
1059 '/((?i)' . wfUrlProtocols() . ')/',
1060 array( 'Sanitizer', 'armorLinksCallback' ),
1066 * Given a value, escape it so that it can be used in an id attribute and
1067 * return it. This will use HTML5 validation if $wgExperimentalHtmlIds is
1068 * true, allowing anything but ASCII whitespace. Otherwise it will use
1069 * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1070 * escaped with lots of dots.
1072 * To ensure we don't have to bother escaping anything, we also strip ', ",
1073 * & even if $wgExperimentalIds is true. TODO: Is this the best tactic?
1074 * We also strip # because it upsets IE, and % because it could be
1075 * ambiguous if it's part of something that looks like a percent escape
1076 * (which don't work reliably in fragments cross-browser).
1078 * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1079 * in the id and name attributes
1080 * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1082 * @see http://www.whatwg.org/html/elements.html#the-id-attribute
1083 * HTML5 definition of id attribute
1085 * @param string $id Id to escape
1086 * @param string|array $options String or array of strings (default is array()):
1087 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1088 * so don't pay attention if the first character isn't valid at the
1089 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
1091 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
1092 * if $wgExperimentalHtmlIds is used, so we can generate extra
1093 * anchors and links won't break.
1096 static function escapeId( $id, $options = array() ) {
1097 global $wgExperimentalHtmlIds;
1098 $options = (array)$options;
1100 $id = Sanitizer
::decodeCharReferences( $id );
1102 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1103 $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1104 $id = trim( $id, '_' );
1106 # Must have been all whitespace to start with.
1113 # HTML4-style escaping
1114 static $replace = array(
1119 $id = urlencode( strtr( $id, ' ', '_' ) );
1120 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1122 if ( !preg_match( '/^[a-zA-Z]/', $id )
1123 && !in_array( 'noninitial', $options ) ) {
1124 // Initial character must be a letter!
1131 * Given a value, escape it so that it can be used as a CSS class and
1134 * @todo For extra validity, input should be validated UTF-8.
1136 * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1138 * @param string $class
1141 static function escapeClass( $class ) {
1142 // Convert ugly stuff to underscores and kill underscores in ugly places
1143 return rtrim( preg_replace(
1144 array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1150 * Given HTML input, escape with htmlspecialchars but un-escape entities.
1151 * This allows (generally harmless) entities like   to survive.
1153 * @param string $html HTML to escape
1154 * @return string Escaped input
1156 static function escapeHtmlAllowEntities( $html ) {
1157 $html = Sanitizer
::decodeCharReferences( $html );
1158 # It seems wise to escape ' as well as ", as a matter of course. Can't
1160 $html = htmlspecialchars( $html, ENT_QUOTES
);
1165 * Regex replace callback for armoring links against further processing.
1166 * @param array $matches
1169 private static function armorLinksCallback( $matches ) {
1170 return str_replace( ':', ':', $matches[1] );
1174 * Return an associative array of attribute names and values from
1175 * a partial tag string. Attribute names are forces to lowercase,
1176 * character references are decoded to UTF-8 text.
1178 * @param string $text
1181 public static function decodeTagAttributes( $text ) {
1182 if ( trim( $text ) == '' ) {
1188 if ( !preg_match_all(
1189 self
::getAttribsRegex(),
1192 PREG_SET_ORDER
) ) {
1196 foreach ( $pairs as $set ) {
1197 $attribute = strtolower( $set[1] );
1198 $value = Sanitizer
::getTagAttributeCallback( $set );
1200 // Normalize whitespace
1201 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1202 $value = trim( $value );
1204 // Decode character references
1205 $attribs[$attribute] = Sanitizer
::decodeCharReferences( $value );
1211 * Build a partial tag string from an associative array of attribute
1212 * names and values as returned by decodeTagAttributes.
1214 * @param array $assoc_array
1217 public static function safeEncodeTagAttributes( $assoc_array ) {
1219 foreach ( $assoc_array as $attribute => $value ) {
1220 $encAttribute = htmlspecialchars( $attribute );
1221 $encValue = Sanitizer
::safeEncodeAttribute( $value );
1223 $attribs[] = "$encAttribute=\"$encValue\"";
1225 return count( $attribs ) ?
' ' . implode( ' ', $attribs ) : '';
1229 * Pick the appropriate attribute value from a match set from the
1230 * attribs regex matches.
1233 * @throws MWException When tag conditions are not met.
1236 private static function getTagAttributeCallback( $set ) {
1237 if ( isset( $set[6] ) ) {
1238 # Illegal #XXXXXX color with no quotes.
1240 } elseif ( isset( $set[5] ) ) {
1243 } elseif ( isset( $set[4] ) ) {
1246 } elseif ( isset( $set[3] ) ) {
1249 } elseif ( !isset( $set[2] ) ) {
1250 # In XHTML, attributes must have a value.
1251 # For 'reduced' form, return explicitly the attribute name here.
1254 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1259 * Normalize whitespace and character references in an XML source-
1260 * encoded text for an attribute value.
1262 * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1263 * but note that we're not returning the value, but are returning
1264 * XML source fragments that will be slapped into output.
1266 * @param string $text
1268 * @todo Remove, unused?
1270 private static function normalizeAttributeValue( $text ) {
1271 return str_replace( '"', '"',
1272 self
::normalizeWhitespace(
1273 Sanitizer
::normalizeCharReferences( $text ) ) );
1277 * @param string $text
1280 private static function normalizeWhitespace( $text ) {
1281 return preg_replace(
1282 '/\r\n|[\x20\x0d\x0a\x09]/',
1288 * Normalizes whitespace in a section name, such as might be returned
1289 * by Parser::stripSectionName(), for use in the id's that are used for
1292 * @param string $section
1295 static function normalizeSectionNameWhitespace( $section ) {
1296 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1300 * Ensure that any entities and character references are legal
1301 * for XML and XHTML specifically. Any stray bits will be
1302 * &-escaped to result in a valid text fragment.
1304 * a. named char refs can only be < > & ", others are
1305 * numericized (this way we're well-formed even without a DTD)
1306 * b. any numeric char refs must be legal chars, not invalid or forbidden
1307 * c. use lower cased "&#x", not "&#X"
1308 * d. fix or reject non-valid attributes
1310 * @param string $text
1314 static function normalizeCharReferences( $text ) {
1315 return preg_replace_callback(
1316 self
::CHAR_REFS_REGEX
,
1317 array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1322 * @param string $matches
1325 static function normalizeCharReferencesCallback( $matches ) {
1327 if ( $matches[1] != '' ) {
1328 $ret = Sanitizer
::normalizeEntity( $matches[1] );
1329 } elseif ( $matches[2] != '' ) {
1330 $ret = Sanitizer
::decCharReference( $matches[2] );
1331 } elseif ( $matches[3] != '' ) {
1332 $ret = Sanitizer
::hexCharReference( $matches[3] );
1334 if ( is_null( $ret ) ) {
1335 return htmlspecialchars( $matches[0] );
1342 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1343 * return the equivalent numeric entity reference (except for the core <
1344 * > & "). If the entity is a MediaWiki-specific alias, returns
1345 * the HTML equivalent. Otherwise, returns HTML-escaped text of
1346 * pseudo-entity source (eg &foo;)
1348 * @param string $name
1351 static function normalizeEntity( $name ) {
1352 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1353 return '&' . self
::$htmlEntityAliases[$name] . ';';
1354 } elseif ( in_array( $name,
1355 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1357 } elseif ( isset( self
::$htmlEntities[$name] ) ) {
1358 return '&#' . self
::$htmlEntities[$name] . ';';
1360 return "&$name;";
1365 * @param int $codepoint
1366 * @return null|string
1368 static function decCharReference( $codepoint ) {
1369 $point = intval( $codepoint );
1370 if ( Sanitizer
::validateCodepoint( $point ) ) {
1371 return sprintf( '&#%d;', $point );
1378 * @param int $codepoint
1379 * @return null|string
1381 static function hexCharReference( $codepoint ) {
1382 $point = hexdec( $codepoint );
1383 if ( Sanitizer
::validateCodepoint( $point ) ) {
1384 return sprintf( '&#x%x;', $point );
1391 * Returns true if a given Unicode codepoint is a valid character in XML.
1392 * @param int $codepoint
1395 private static function validateCodepoint( $codepoint ) {
1396 return $codepoint == 0x09
1397 ||
$codepoint == 0x0a
1398 ||
$codepoint == 0x0d
1399 ||
( $codepoint >= 0x20 && $codepoint <= 0xd7ff )
1400 ||
( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1401 ||
( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1405 * Decode any character references, numeric or named entities,
1406 * in the text and return a UTF-8 string.
1408 * @param string $text
1411 public static function decodeCharReferences( $text ) {
1412 return preg_replace_callback(
1413 self
::CHAR_REFS_REGEX
,
1414 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1419 * Decode any character references, numeric or named entities,
1420 * in the next and normalize the resulting string. (bug 14952)
1422 * This is useful for page titles, not for text to be displayed,
1423 * MediaWiki allows HTML entities to escape normalization as a feature.
1425 * @param string $text Already normalized, containing entities
1426 * @return string Still normalized, without entities
1428 public static function decodeCharReferencesAndNormalize( $text ) {
1430 $text = preg_replace_callback(
1431 self
::CHAR_REFS_REGEX
,
1432 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1433 $text, /* limit */ -1, $count );
1436 return $wgContLang->normalize( $text );
1443 * @param string $matches
1446 static function decodeCharReferencesCallback( $matches ) {
1447 if ( $matches[1] != '' ) {
1448 return Sanitizer
::decodeEntity( $matches[1] );
1449 } elseif ( $matches[2] != '' ) {
1450 return Sanitizer
::decodeChar( intval( $matches[2] ) );
1451 } elseif ( $matches[3] != '' ) {
1452 return Sanitizer
::decodeChar( hexdec( $matches[3] ) );
1454 # Last case should be an ampersand by itself
1459 * Return UTF-8 string for a codepoint if that is a valid
1460 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1461 * @param int $codepoint
1465 static function decodeChar( $codepoint ) {
1466 if ( Sanitizer
::validateCodepoint( $codepoint ) ) {
1467 return codepointToUtf8( $codepoint );
1469 return UTF8_REPLACEMENT
;
1474 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1475 * return the UTF-8 encoding of that character. Otherwise, returns
1476 * pseudo-entity source (eg "&foo;")
1478 * @param string $name
1481 static function decodeEntity( $name ) {
1482 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1483 $name = self
::$htmlEntityAliases[$name];
1485 if ( isset( self
::$htmlEntities[$name] ) ) {
1486 return codepointToUtf8( self
::$htmlEntities[$name] );
1493 * Fetch the whitelist of acceptable attributes for a given element name.
1495 * @param string $element
1498 static function attributeWhitelist( $element ) {
1499 $list = Sanitizer
::setupAttributeWhitelist();
1500 return isset( $list[$element] )
1506 * Foreach array key (an allowed HTML element), return an array
1507 * of allowed attributes
1510 static function setupAttributeWhitelist() {
1511 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
1512 static $whitelist, $staticInitialised;
1514 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgAllowMicrodataAttributes' ) );
1516 if ( $whitelist !== null && $staticInitialised == $globalContext ) {
1533 if ( $wgAllowRdfaAttributes ) {
1534 # RDFa attributes as specified in section 9 of
1535 # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1536 $common = array_merge( $common, array(
1537 'about', 'property', 'resource', 'datatype', 'typeof',
1541 if ( $wgAllowMicrodataAttributes ) {
1542 # add HTML5 microdata tags as specified by
1543 # http://www.whatwg.org/html/microdata.html#the-microdata-model
1544 $common = array_merge( $common, array(
1545 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1549 $block = array_merge( $common, array( 'align' ) );
1550 $tablealign = array( 'align', 'valign' );
1558 'nowrap', # deprecated
1559 'width', # deprecated
1560 'height', # deprecated
1561 'bgcolor', # deprecated
1564 # Numbers refer to sections in HTML 4.01 standard describing the element.
1565 # See: http://www.w3.org/TR/html4/
1569 'center' => $common, # deprecated
1588 'strong' => $common,
1599 'blockquote' => array_merge( $common, array( 'cite' ) ),
1600 'q' => array_merge( $common, array( 'cite' ) ),
1610 'br' => array_merge( $common, array( 'clear' ) ),
1612 # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1616 'pre' => array_merge( $common, array( 'width' ) ),
1619 'ins' => array_merge( $common, array( 'cite', 'datetime' ) ),
1620 'del' => array_merge( $common, array( 'cite', 'datetime' ) ),
1623 'ul' => array_merge( $common, array( 'type' ) ),
1624 'ol' => array_merge( $common, array( 'type', 'start' ) ),
1625 'li' => array_merge( $common, array( 'type', 'value' ) ),
1633 'table' => array_merge( $common,
1634 array( 'summary', 'width', 'border', 'frame',
1635 'rules', 'cellspacing', 'cellpadding',
1640 'caption' => $block,
1648 'colgroup' => array_merge( $common, array( 'span' ) ),
1649 'col' => array_merge( $common, array( 'span' ) ),
1652 'tr' => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1655 'td' => array_merge( $common, $tablecell, $tablealign ),
1656 'th' => array_merge( $common, $tablecell, $tablealign ),
1659 # NOTE: <a> is not allowed directly, but the attrib
1660 # whitelist is used from the Parser object
1661 'a' => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1664 # Not usually allowed, but may be used for extension-style hooks
1665 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1667 'img' => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1675 'strike' => $common,
1680 'font' => array_merge( $common, array( 'size', 'color', 'face' ) ),
1684 'hr' => array_merge( $common, array( 'width' ) ),
1686 # HTML Ruby annotation text module, simple ruby only.
1687 # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
1692 'rt' => $common, #array_merge( $common, array( 'rbspan' ) ),
1695 # MathML root element, where used for extensions
1696 # 'title' may not be 100% valid here; it's XHTML
1697 # http://www.w3.org/TR/REC-MathML/
1698 'math' => array( 'class', 'style', 'id', 'title' ),
1700 # HTML 5 section 4.6
1703 # HTML5 elements, defined by:
1704 # http://www.whatwg.org/html/
1705 'data' => array_merge( $common, array( 'value' ) ),
1706 'time' => array_merge( $common, array( 'datetime' ) ),
1709 // meta and link are only permitted by removeHTMLtags when Microdata
1710 // is enabled so we don't bother adding a conditional to hide these
1711 // Also meta and link are only valid in WikiText as Microdata elements
1712 // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1713 // So we don't bother including $common attributes that have no purpose.
1714 'meta' => array( 'itemprop', 'content' ),
1715 'link' => array( 'itemprop', 'href' ),
1718 $staticInitialised = $globalContext;
1724 * Take a fragment of (potentially invalid) HTML and return
1725 * a version with any tags removed, encoded as plain text.
1727 * Warning: this return value must be further escaped for literal
1728 * inclusion in HTML output as of 1.10!
1730 * @param string $text HTML fragment
1733 static function stripAllTags( $text ) {
1735 $text = StringUtils
::delimiterReplace( '<', '>', '', $text );
1737 # Normalize &entities and whitespace
1738 $text = self
::decodeCharReferences( $text );
1739 $text = self
::normalizeWhitespace( $text );
1745 * Hack up a private DOCTYPE with HTML's standard entity declarations.
1746 * PHP 4 seemed to know these if you gave it an HTML doctype, but
1749 * Use for passing XHTML fragments to PHP's XML parsing functions
1753 static function hackDocType() {
1754 $out = "<!DOCTYPE html [\n";
1755 foreach ( self
::$htmlEntities as $entity => $codepoint ) {
1756 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1763 * @param string $url
1764 * @return mixed|string
1766 static function cleanUrl( $url ) {
1767 # Normalize any HTML entities in input. They will be
1768 # re-escaped by makeExternalLink().
1769 $url = Sanitizer
::decodeCharReferences( $url );
1771 # Escape any control characters introduced by the above step
1772 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1773 array( __CLASS__
, 'cleanUrlCallback' ), $url );
1775 # Validate hostname portion
1777 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1778 list( /* $whole */, $protocol, $host, $rest ) = $matches;
1780 // Characters that will be ignored in IDNs.
1781 // http://tools.ietf.org/html/3454#section-3.1
1782 // Strip them before further processing so blacklists and such work.
1784 \\s| # general whitespace
1785 \xc2\xad| # 00ad SOFT HYPHEN
1786 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1787 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1788 \xe2\x81\xa0| # 2060 WORD JOINER
1789 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1790 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
1791 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1792 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1793 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1794 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1795 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1796 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1799 $host = preg_replace( $strip, '', $host );
1801 // @todo FIXME: Validate hostnames here
1803 return $protocol . $host . $rest;
1810 * @param array $matches
1813 static function cleanUrlCallback( $matches ) {
1814 return urlencode( $matches[0] );
1818 * Does a string look like an e-mail address?
1820 * This validates an email address using an HTML5 specification found at:
1821 * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1822 * Which as of 2011-01-24 says:
1824 * A valid e-mail address is a string that matches the ABNF production
1825 * 1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1826 * in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1829 * This function is an implementation of the specification as requested in
1832 * Client-side forms will use the same standard validation rules via JS or
1833 * HTML 5 validation; additional restrictions can be enforced server-side
1834 * by extensions via the 'isValidEmailAddr' hook.
1836 * Note that this validation doesn't 100% match RFC 2822, but is believed
1837 * to be liberal enough for wide use. Some invalid addresses will still
1838 * pass validation here.
1842 * @param string $addr E-mail address
1845 public static function validateEmail( $addr ) {
1847 if ( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1851 // Please note strings below are enclosed in brackets [], this make the
1852 // hyphen "-" a range indicator. Hence it is double backslashed below.
1854 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1855 $rfc1034_ldh_str = "a-z0-9\\-";
1857 $html5_email_regexp = "/
1859 [$rfc5322_atext\\.]+ # user part which is liberal :p
1861 [$rfc1034_ldh_str]+ # First domain part
1862 (\\.[$rfc1034_ldh_str]+)* # Following part prefixed with a dot
1864 /ix"; // case Insensitive, eXtended
1866 return (bool)preg_match( $html5_email_regexp, $addr );