3 * XHTML sanitizer for MediaWiki
5 * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
6 * http://www.mediawiki.org/
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 2 of the License, or
11 * (at your option) any later version.
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License along
19 * with this program; if not, write to the Free Software Foundation, Inc.,
20 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
21 * http://www.gnu.org/copyleft/gpl.html
28 * XHTML sanitizer for MediaWiki
33 * Regular expression to match various types of character references in
34 * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
36 const CHAR_REFS_REGEX
=
37 '/&([A-Za-z0-9\x80-\xff]+);
39 |&\#[xX]([0-9A-Fa-f]+);
43 * Blacklist for evil uris like javascript:
44 * WARNING: DO NOT use this in any place that actually requires blacklisting
45 * for security reasons. There are NUMEROUS[1] ways to bypass blacklisting, the
46 * only way to be secure from javascript: uri based xss vectors is to whitelist
47 * things that you know are safe and deny everything else.
48 * [1]: http://ha.ckers.org/xss.html
50 const EVIL_URI_PATTERN
= '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
51 const XMLNS_ATTRIBUTE_PATTERN
= "/^xmlns:[:A-Z_a-z-.0-9]+$/";
54 * List of all named character entities defined in HTML 4.01
55 * http://www.w3.org/TR/html4/sgml/entities.html
56 * As well as ' which is only defined starting in XHTML1.
59 static $htmlEntities = array(
75 'apos' => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
316 * Character entity aliases accepted by MediaWiki
318 static $htmlEntityAliases = array(
324 * Lazy-initialised attributes regex, see getAttribsRegex()
326 static $attribsRegex;
329 * Regular expression to match HTML/XML attribute pairs within a tag.
330 * Allows some... latitude.
331 * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
333 static function getAttribsRegex() {
334 if ( self
::$attribsRegex === null ) {
335 $attribFirst = '[:A-Z_a-z0-9]';
336 $attrib = '[:A-Z_a-z-.0-9]';
337 $space = '[\x09\x0a\x0d\x20]';
338 self
::$attribsRegex =
339 "/(?:^|$space)({$attribFirst}{$attrib}*)
342 # The attribute value: quoted or alone
345 | ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
346 | (\#[0-9a-fA-F]+) # Technically wrong, but lots of
347 # colors are specified like this.
348 # We'll be normalizing it.
352 return self
::$attribsRegex;
356 * Cleans up HTML, removes dangerous tags and attributes, and
357 * removes HTML comments
359 * @param $text String
360 * @param $processCallback Callback to do any variable or parameter replacements in HTML attribute values
361 * @param $args Array for the processing callback
362 * @param $extratags Array for any extra tags to include
363 * @param $removetags Array for any tags (default or extra) to exclude
366 static function removeHTMLtags( $text, $processCallback = null, $args = array(), $extratags = array(), $removetags = array() ) {
367 global $wgUseTidy, $wgHtml5, $wgAllowMicrodataAttributes, $wgAllowImageTag;
369 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
370 $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
372 wfProfileIn( __METHOD__
);
374 // Base our staticInitialised variable off of the global config state so that if the globals
375 // are changed (like in the secrewed up test system) we will re-initialise the settings.
376 $globalContext = implode( '-', compact( 'wgHtml5', 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
377 if ( !$staticInitialised ||
$staticInitialised != $globalContext ) {
379 $htmlpairsStatic = array( # Tags that must be closed
380 'b', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
381 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
382 'strike', 'strong', 'tt', 'var', 'div', 'center',
383 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
384 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'abbr', 'dfn',
388 $htmlpairsStatic = array_merge( $htmlpairsStatic, array( 'data', 'time' ) );
391 'br', 'hr', 'li', 'dt', 'dd'
393 $htmlsingleonly = array( # Elements that cannot have close tags
396 if ( $wgHtml5 && $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 'dl', 'font', 'big', 'small', 'sub', 'sup', 'span'
404 $tabletags = array( # Can only appear inside table, we will close them
407 $htmllist = array( # Tags used by list
410 $listtags = array( # Tags that can appear in a list
414 if ( $wgAllowImageTag ) {
415 $htmlsingle[] = 'img';
416 $htmlsingleonly[] = 'img';
419 $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
420 $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
422 # Convert them all to hashtables for faster lookup
423 $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
424 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
425 foreach ( $vars as $var ) {
426 $
$var = array_flip( $
$var );
428 $staticInitialised = $globalContext;
430 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
431 $extratags = array_flip( $extratags );
432 $removetags = array_flip( $removetags );
433 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
434 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ) , $removetags );
436 # Remove HTML comments
437 $text = Sanitizer
::removeHTMLcomments( $text );
438 $bits = explode( '<', $text );
439 $text = str_replace( '>', '>', array_shift( $bits ) );
441 $tagstack = $tablestack = array();
442 foreach ( $bits as $x ) {
444 # $slash: Does the current element start with a '/'?
445 # $t: Current element name
446 # $params: String between element name and >
447 # $brace: Ending '>' or '/>'
448 # $rest: Everything until the next element of $bits
449 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
450 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
452 $slash = $t = $params = $brace = $rest = null;
456 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
458 if ( $slash && isset( $htmlsingleonly[$t] ) ) {
460 } elseif ( $slash ) {
461 # Closing a tag... is it the one we just opened?
462 $ot = @array_pop
( $tagstack );
464 if ( isset( $htmlsingleallowed[$ot] ) ) {
465 # Pop all elements with an optional close tag
466 # and see if we find a match below them
468 array_push( $optstack, $ot );
469 wfSuppressWarnings();
470 $ot = array_pop( $tagstack );
472 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
473 array_push( $optstack, $ot );
474 wfSuppressWarnings();
475 $ot = array_pop( $tagstack );
479 # No match. Push the optional elements back again
481 wfSuppressWarnings();
482 $ot = array_pop( $optstack );
485 array_push( $tagstack, $ot );
486 wfSuppressWarnings();
487 $ot = array_pop( $optstack );
492 @array_push
( $tagstack, $ot );
493 # <li> can be nested in <ul> or <ol>, skip those cases:
494 if ( !isset( $htmllist[$ot] ) ||
!isset( $listtags[$t] ) ) {
499 if ( $t == 'table' ) {
500 $tagstack = array_pop( $tablestack );
505 # Keep track for later
506 if ( isset( $tabletags[$t] ) &&
507 !in_array( 'table', $tagstack ) ) {
509 } elseif ( in_array( $t, $tagstack ) &&
510 !isset( $htmlnest [$t ] ) ) {
512 # Is it a self closed htmlpair ? (bug 5487)
513 } elseif ( $brace == '/>' &&
514 isset( $htmlpairs[$t] ) ) {
516 } elseif ( isset( $htmlsingleonly[$t] ) ) {
517 # Hack to force empty tag for uncloseable elements
519 } elseif ( isset( $htmlsingle[$t] ) ) {
520 # Hack to not close $htmlsingle tags
522 } elseif ( isset( $tabletags[$t] )
523 && in_array( $t, $tagstack ) ) {
524 // New table tag but forgot to close the previous one
527 if ( $t == 'table' ) {
528 array_push( $tablestack, $tagstack );
531 array_push( $tagstack, $t );
534 # Replace any variables or template parameters with
536 if( is_callable( $processCallback ) ) {
537 call_user_func_array( $processCallback, array( &$params, $args ) );
540 if ( !Sanitizer
::validateTag( $params, $t ) ) {
544 # Strip non-approved attributes from the tag
545 $newparams = Sanitizer
::fixTagAttributes( $params, $t );
548 $rest = str_replace( '>', '>', $rest );
549 $close = ( $brace == '/>' && !$slash ) ?
' /' : '';
550 $text .= "<$slash$t$newparams$close>$rest";
554 $text .= '<' . str_replace( '>', '>', $x);
556 # Close off any remaining tags
557 while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
559 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
562 # this might be possible using tidy itself
563 foreach ( $bits as $x ) {
564 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
566 @list
( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
568 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
569 if( is_callable( $processCallback ) ) {
570 call_user_func_array( $processCallback, array( &$params, $args ) );
573 if ( !Sanitizer
::validateTag( $params, $t ) ) {
577 $newparams = Sanitizer
::fixTagAttributes( $params, $t );
579 $rest = str_replace( '>', '>', $rest );
580 $text .= "<$slash$t$newparams$brace$rest";
584 $text .= '<' . str_replace( '>', '>', $x);
587 wfProfileOut( __METHOD__
);
592 * Remove '<!--', '-->', and everything between.
593 * To avoid leaving blank lines, when a comment is both preceded
594 * and followed by a newline (ignoring spaces), trim leading and
595 * trailing spaces and one of the newlines.
598 * @param $text String
601 static function removeHTMLcomments( $text ) {
602 wfProfileIn( __METHOD__
);
603 while (($start = strpos($text, '<!--')) !== false) {
604 $end = strpos($text, '-->', $start +
4);
605 if ($end === false) {
606 # Unterminated comment; bail out
612 # Trim space and newline if the comment is both
613 # preceded and followed by a newline
614 $spaceStart = max($start - 1, 0);
615 $spaceLen = $end - $spaceStart;
616 while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
620 while (substr($text, $spaceStart +
$spaceLen, 1) === ' ')
622 if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart +
$spaceLen, 1) === "\n") {
623 # Remove the comment, leading and trailing
624 # spaces, and leave only one newline.
625 $text = substr_replace($text, "\n", $spaceStart, $spaceLen +
1);
628 # Remove just the comment.
629 $text = substr_replace($text, '', $start, $end - $start);
632 wfProfileOut( __METHOD__
);
637 * Take an array of attribute names and values and fix some deprecated values
638 * for the given element type.
639 * This does not validate properties, so you should ensure that you call
640 * validateTagAttributes AFTER this to ensure that the resulting style rule
641 * this may add is safe.
643 * - Converts most presentational attributes like align into inline css
645 * @param $attribs Array
646 * @param $element String
649 static function fixDeprecatedAttributes( $attribs, $element ) {
650 global $wgHtml5, $wgCleanupPresentationalAttributes;
652 // presentational attributes were removed from html5, we can leave them
653 // in when html5 is turned off
654 if ( !$wgHtml5 ||
!$wgCleanupPresentationalAttributes ) {
658 $table = array( 'table' );
659 $cells = array( 'td', 'th' );
660 $colls = array( 'col', 'colgroup' );
661 $tblocks = array( 'tbody', 'tfoot', 'thead' );
662 $h = array( 'h1', 'h2', 'h3', 'h4', 'h5', 'h6' );
664 $presentationalAttribs = array(
665 'align' => array( 'text-align', array_merge( array( 'caption', 'hr', 'div', 'p', 'tr' ), $table, $cells, $colls, $tblocks, $h ) ),
666 'clear' => array( 'clear', array( 'br' ) ),
667 'height' => array( 'height', $cells ),
668 'nowrap' => array( 'white-space', $cells ),
669 'size' => array( 'height', array( 'hr' ) ),
670 'type' => array( 'list-style-type', array( 'li', 'ol', 'ul' ) ),
671 'valign' => array( 'vertical-align', array_merge( $cells, $colls, $tblocks ) ),
672 'width' => array( 'width', array_merge( array( 'hr', 'pre' ), $table, $cells, $colls ) ),
675 // Ensure that any upper case or mixed case attributes are converted to lowercase
676 foreach ( $attribs as $attribute => $value ) {
677 if ( $attribute !== strtolower( $attribute ) && array_key_exists( strtolower( $attribute ), $presentationalAttribs ) ) {
678 $attribs[strtolower( $attribute )] = $value;
679 unset( $attribs[$attribute] );
684 foreach ( $presentationalAttribs as $attribute => $info ) {
685 list( $property, $elements ) = $info;
687 // Skip if this attribute is not relevant to this element
688 if ( !in_array( $element, $elements ) ) {
692 // Skip if the attribute is not used
693 if ( !array_key_exists( $attribute, $attribs ) ) {
697 $value = $attribs[$attribute];
699 // For nowrap the value should be nowrap instead of whatever text is in the value
700 if ( $attribute === 'nowrap' ) {
704 // clear="all" is clear: both; in css
705 if ( $attribute === 'clear' && strtolower( $value ) === 'all' ) {
709 // Size based properties should have px applied to them if they have no unit
710 if ( in_array( $attribute, array( 'height', 'width', 'size' ) ) ) {
711 if ( preg_match( '/^[\d.]+$/', $value ) ) {
712 $value = "{$value}px";
716 $style .= " $property: $value;";
718 unset( $attribs[$attribute] );
722 // Prepend our style rules so that they can be overridden by user css
723 if ( isset($attribs['style']) ) {
724 $style .= " " . $attribs['style'];
726 $attribs['style'] = trim($style);
733 * Takes attribute names and values for a tag and the tah name and
734 * validates that the tag is allowed to be present.
735 * This DOES NOT validate the attributes, nor does it validate the
736 * tags themselves. This method only handles the special circumstances
737 * where we may want to allow a tag within content but ONLY when it has
738 * specific attributes set.
742 static function validateTag( $params, $element ) {
743 $params = Sanitizer
::decodeTagAttributes( $params );
745 if ( $element == 'meta' ||
$element == 'link' ) {
746 if ( !isset( $params['itemprop'] ) ) {
747 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
750 if ( $element == 'meta' && !isset( $params['content'] ) ) {
751 // <meta> must have a content="" for the itemprop
754 if ( $element == 'link' && !isset( $params['href'] ) ) {
755 // <link> must have an associated href=""
764 * Take an array of attribute names and values and normalize or discard
765 * illegal values for the given element type.
767 * - Discards attributes not on a whitelist for the given element
768 * - Unsafe style attributes are discarded
769 * - Invalid id attributes are reencoded
771 * @param $attribs Array
772 * @param $element String
775 * @todo Check for legal values where the DTD limits things.
776 * @todo Check for unique id attribute :P
778 static function validateTagAttributes( $attribs, $element ) {
779 return Sanitizer
::validateAttributes( $attribs,
780 Sanitizer
::attributeWhitelist( $element ) );
784 * Take an array of attribute names and values and normalize or discard
785 * illegal values for the given whitelist.
787 * - Discards attributes not the given whitelist
788 * - Unsafe style attributes are discarded
789 * - Invalid id attributes are reencoded
791 * @param $attribs Array
792 * @param $whitelist Array: list of allowed attribute names
795 * @todo Check for legal values where the DTD limits things.
796 * @todo Check for unique id attribute :P
798 static function validateAttributes( $attribs, $whitelist ) {
799 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
801 $whitelist = array_flip( $whitelist );
802 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
805 foreach( $attribs as $attribute => $value ) {
806 #allow XML namespace declaration if RDFa is enabled
807 if ( $wgAllowRdfaAttributes && preg_match( self
::XMLNS_ATTRIBUTE_PATTERN
, $attribute ) ) {
808 if ( !preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
809 $out[$attribute] = $value;
815 # Allow any attribute beginning with "data-", if in HTML5 mode
816 if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
820 # Strip javascript "expression" from stylesheets.
821 # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
822 if( $attribute == 'style' ) {
823 $value = Sanitizer
::checkCss( $value );
826 if ( $attribute === 'id' ) {
827 $value = Sanitizer
::escapeId( $value, 'noninitial' );
830 //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
831 if ( $attribute === 'rel' ||
$attribute === 'rev' ||
832 $attribute === 'about' ||
$attribute === 'property' ||
$attribute === 'resource' ||
#RDFa
833 $attribute === 'datatype' ||
$attribute === 'typeof' ||
#RDFa
834 $attribute === 'itemid' ||
$attribute === 'itemprop' ||
$attribute === 'itemref' ||
#HTML5 microdata
835 $attribute === 'itemscope' ||
$attribute === 'itemtype' ) { #HTML5 microdata
837 //Paranoia. Allow "simple" values but suppress javascript
838 if ( preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
843 # NOTE: even though elements using href/src are not allowed directly, supply
844 # validation code that can be used by tag hook handlers, etc
845 if ( $attribute === 'href' ||
$attribute === 'src' ) {
846 if ( !preg_match( $hrefExp, $value ) ) {
847 continue; //drop any href or src attributes not using an allowed protocol.
848 //NOTE: this also drops all relative URLs
852 // If this attribute was previously set, override it.
853 // Output should only have one attribute of each name.
854 $out[$attribute] = $value;
857 if ( $wgAllowMicrodataAttributes ) {
858 # itemtype, itemid, itemref don't make sense without itemscope
859 if ( !array_key_exists( 'itemscope', $out ) ) {
860 unset( $out['itemtype'] );
861 unset( $out['itemid'] );
862 unset( $out['itemref'] );
864 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
870 * Merge two sets of HTML attributes. Conflicting items in the second set
871 * will override those in the first, except for 'class' attributes which
872 * will be combined (if they're both strings).
874 * @todo implement merging for other attributes such as style
879 static function mergeAttributes( $a, $b ) {
880 $out = array_merge( $a, $b );
881 if( isset( $a['class'] ) && isset( $b['class'] )
882 && is_string( $a['class'] ) && is_string( $b['class'] )
883 && $a['class'] !== $b['class'] ) {
884 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
885 -1, PREG_SPLIT_NO_EMPTY
);
886 $out['class'] = implode( ' ', array_unique( $classes ) );
892 * Pick apart some CSS and check it for forbidden or unsafe structures.
893 * Returns a sanitized string. This sanitized string will have
894 * character references and escape sequences decoded, and comments
895 * stripped. If the input is just too evil, only a comment complaining
896 * about evilness will be returned.
898 * Currently URL references, 'expression', 'tps' are forbidden.
900 * NOTE: Despite the fact that character references are decoded, the
901 * returned string may contain character references given certain
902 * clever input strings. These character references must
903 * be escaped before the return value is embedded in HTML.
905 * @param $value String
908 static function checkCss( $value ) {
909 // Decode character references like {
910 $value = Sanitizer
::decodeCharReferences( $value );
912 // Decode escape sequences and line continuation
913 // See the grammar in the CSS 2 spec, appendix D.
914 // This has to be done AFTER decoding character references.
915 // This means it isn't possible for this function to return
916 // unsanitized escape sequences. It is possible to manufacture
917 // input that contains character references that decode to
918 // escape sequences that decode to character references, but
919 // it's OK for the return value to contain character references
920 // because the caller is supposed to escape those anyway.
922 if ( !$decodeRegex ) {
923 $space = '[\\x20\\t\\r\\n\\f]';
924 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
926 $decodeRegex = "/ $backslash
928 ($nl) | # 1. Line continuation
929 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
930 (.) | # 3. backslash cancelling special meaning
931 () | # 4. backslash at end of string
934 $value = preg_replace_callback( $decodeRegex,
935 array( __CLASS__
, 'cssDecodeCallback' ), $value );
937 // Remove any comments; IE gets token splitting wrong
938 // This must be done AFTER decoding character references and
939 // escape sequences, because those steps can introduce comments
940 // This step cannot introduce character references or escape
941 // sequences, because it replaces comments with spaces rather
942 // than removing them completely.
943 $value = StringUtils
::delimiterReplace( '/*', '*/', ' ', $value );
945 // Remove anything after a comment-start token, to guard against
946 // incorrect client implementations.
947 $commentPos = strpos( $value, '/*' );
948 if ( $commentPos !== false ) {
949 $value = substr( $value, 0, $commentPos );
952 // Reject problematic keywords and control characters
953 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
954 return '/* invalid control char */';
955 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
956 return '/* insecure input */';
962 * @param $matches array
965 static function cssDecodeCallback( $matches ) {
966 if ( $matches[1] !== '' ) {
969 } elseif ( $matches[2] !== '' ) {
970 $char = codepointToUtf8( hexdec( $matches[2] ) );
971 } elseif ( $matches[3] !== '' ) {
976 if ( $char == "\n" ||
$char == '"' ||
$char == "'" ||
$char == '\\' ) {
977 // These characters need to be escaped in strings
978 // Clean up the escape sequence to avoid parsing errors by clients
979 return '\\' . dechex( ord( $char ) ) . ' ';
981 // Decode unnecessary escape
987 * Take a tag soup fragment listing an HTML element's attributes
988 * and normalize it to well-formed XML, discarding unwanted attributes.
989 * Output is safe for further wikitext processing, with escaping of
990 * values that could trigger problems.
992 * - Normalizes attribute names to lowercase
993 * - Discards attributes not on a whitelist for the given element
994 * - Turns broken or invalid entities into plaintext
995 * - Double-quotes all attribute values
996 * - Attributes without values are given the name as attribute
997 * - Double attributes are discarded
998 * - Unsafe style attributes are discarded
999 * - Prepends space if there are attributes.
1001 * @param $text String
1002 * @param $element String
1005 static function fixTagAttributes( $text, $element ) {
1006 if( trim( $text ) == '' ) {
1010 $decoded = Sanitizer
::decodeTagAttributes( $text );
1011 $decoded = Sanitizer
::fixDeprecatedAttributes( $decoded, $element );
1012 $stripped = Sanitizer
::validateTagAttributes( $decoded, $element );
1015 foreach( $stripped as $attribute => $value ) {
1016 $encAttribute = htmlspecialchars( $attribute );
1017 $encValue = Sanitizer
::safeEncodeAttribute( $value );
1019 $attribs[] = "$encAttribute=\"$encValue\"";
1021 return count( $attribs ) ?
' ' . implode( ' ', $attribs ) : '';
1025 * Encode an attribute value for HTML output.
1026 * @param $text String
1027 * @return HTML-encoded text fragment
1029 static function encodeAttribute( $text ) {
1030 $encValue = htmlspecialchars( $text, ENT_QUOTES
);
1032 // Whitespace is normalized during attribute decoding,
1033 // so if we've been passed non-spaces we must encode them
1034 // ahead of time or they won't be preserved.
1035 $encValue = strtr( $encValue, array(
1045 * Encode an attribute value for HTML tags, with extra armoring
1046 * against further wiki processing.
1047 * @param $text String
1048 * @return HTML-encoded text fragment
1050 static function safeEncodeAttribute( $text ) {
1051 $encValue = Sanitizer
::encodeAttribute( $text );
1053 # Templates and links may be expanded in later parsing,
1054 # creating invalid or dangerous output. Suppress this.
1055 $encValue = strtr( $encValue, array(
1056 '<' => '<', // This should never happen,
1057 '>' => '>', // we've received invalid input
1058 '"' => '"', // which should have been escaped.
1061 "''" => '''',
1062 'ISBN' => 'ISBN',
1064 'PMID' => 'PMID',
1070 $encValue = preg_replace_callback(
1071 '/(' . wfUrlProtocols() . ')/',
1072 array( 'Sanitizer', 'armorLinksCallback' ),
1078 * Given a value, escape it so that it can be used in an id attribute and
1079 * return it. This will use HTML5 validation if $wgExperimentalHtmlIds is
1080 * true, allowing anything but ASCII whitespace. Otherwise it will use
1081 * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1082 * escaped with lots of dots.
1084 * To ensure we don't have to bother escaping anything, we also strip ', ",
1085 * & even if $wgExperimentalIds is true. TODO: Is this the best tactic?
1086 * We also strip # because it upsets IE, and % because it could be
1087 * ambiguous if it's part of something that looks like a percent escape
1088 * (which don't work reliably in fragments cross-browser).
1090 * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1093 * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1094 * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1095 * HTML5 definition of id attribute
1097 * @param $id String: id to escape
1098 * @param $options Mixed: string or array of strings (default is array()):
1099 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1100 * so don't pay attention if the first character isn't valid at the
1101 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
1103 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
1104 * if $wgExperimentalHtmlIds is used, so we can generate extra
1105 * anchors and links won't break.
1108 static function escapeId( $id, $options = array() ) {
1109 global $wgHtml5, $wgExperimentalHtmlIds;
1110 $options = (array)$options;
1112 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1113 $id = Sanitizer
::decodeCharReferences( $id );
1114 $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1115 $id = trim( $id, '_' );
1117 # Must have been all whitespace to start with.
1124 # HTML4-style escaping
1125 static $replace = array(
1130 $id = urlencode( Sanitizer
::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1131 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1133 if ( !preg_match( '/^[a-zA-Z]/', $id )
1134 && !in_array( 'noninitial', $options ) ) {
1135 // Initial character must be a letter!
1142 * Given a value, escape it so that it can be used as a CSS class and
1145 * @todo For extra validity, input should be validated UTF-8.
1147 * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1149 * @param $class String
1152 static function escapeClass( $class ) {
1153 // Convert ugly stuff to underscores and kill underscores in ugly places
1154 return rtrim(preg_replace(
1155 array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
1161 * Given HTML input, escape with htmlspecialchars but un-escape entites.
1162 * This allows (generally harmless) entities like   to survive.
1164 * @param $html String to escape
1165 * @return String: escaped input
1167 static function escapeHtmlAllowEntities( $html ) {
1168 $html = Sanitizer
::decodeCharReferences( $html );
1169 # It seems wise to escape ' as well as ", as a matter of course. Can't
1171 $html = htmlspecialchars( $html, ENT_QUOTES
);
1176 * Regex replace callback for armoring links against further processing.
1177 * @param $matches Array
1180 private static function armorLinksCallback( $matches ) {
1181 return str_replace( ':', ':', $matches[1] );
1185 * Return an associative array of attribute names and values from
1186 * a partial tag string. Attribute names are forces to lowercase,
1187 * character references are decoded to UTF-8 text.
1189 * @param $text String
1192 public static function decodeTagAttributes( $text ) {
1193 if( trim( $text ) == '' ) {
1199 if( !preg_match_all(
1200 self
::getAttribsRegex(),
1203 PREG_SET_ORDER
) ) {
1207 foreach( $pairs as $set ) {
1208 $attribute = strtolower( $set[1] );
1209 $value = Sanitizer
::getTagAttributeCallback( $set );
1211 // Normalize whitespace
1212 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1213 $value = trim( $value );
1215 // Decode character references
1216 $attribs[$attribute] = Sanitizer
::decodeCharReferences( $value );
1222 * Pick the appropriate attribute value from a match set from the
1223 * attribs regex matches.
1228 private static function getTagAttributeCallback( $set ) {
1229 if( isset( $set[6] ) ) {
1230 # Illegal #XXXXXX color with no quotes.
1232 } elseif( isset( $set[5] ) ) {
1235 } elseif( isset( $set[4] ) ) {
1238 } elseif( isset( $set[3] ) ) {
1241 } elseif( !isset( $set[2] ) ) {
1242 # In XHTML, attributes must have a value.
1243 # For 'reduced' form, return explicitly the attribute name here.
1246 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1251 * Normalize whitespace and character references in an XML source-
1252 * encoded text for an attribute value.
1254 * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1255 * but note that we're not returning the value, but are returning
1256 * XML source fragments that will be slapped into output.
1258 * @param $text String
1261 private static function normalizeAttributeValue( $text ) {
1262 return str_replace( '"', '"',
1263 self
::normalizeWhitespace(
1264 Sanitizer
::normalizeCharReferences( $text ) ) );
1268 * @param $text string
1271 private static function normalizeWhitespace( $text ) {
1272 return preg_replace(
1273 '/\r\n|[\x20\x0d\x0a\x09]/',
1279 * Normalizes whitespace in a section name, such as might be returned
1280 * by Parser::stripSectionName(), for use in the id's that are used for
1283 * @param $section String
1286 static function normalizeSectionNameWhitespace( $section ) {
1287 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1291 * Ensure that any entities and character references are legal
1292 * for XML and XHTML specifically. Any stray bits will be
1293 * &-escaped to result in a valid text fragment.
1295 * a. named char refs can only be < > & ", others are
1296 * numericized (this way we're well-formed even without a DTD)
1297 * b. any numeric char refs must be legal chars, not invalid or forbidden
1298 * c. use &#x, not &#X
1299 * d. fix or reject non-valid attributes
1301 * @param $text String
1305 static function normalizeCharReferences( $text ) {
1306 return preg_replace_callback(
1307 self
::CHAR_REFS_REGEX
,
1308 array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1312 * @param $matches String
1315 static function normalizeCharReferencesCallback( $matches ) {
1317 if( $matches[1] != '' ) {
1318 $ret = Sanitizer
::normalizeEntity( $matches[1] );
1319 } elseif( $matches[2] != '' ) {
1320 $ret = Sanitizer
::decCharReference( $matches[2] );
1321 } elseif( $matches[3] != '' ) {
1322 $ret = Sanitizer
::hexCharReference( $matches[3] );
1324 if( is_null( $ret ) ) {
1325 return htmlspecialchars( $matches[0] );
1332 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1333 * return the equivalent numeric entity reference (except for the core <
1334 * > & "). If the entity is a MediaWiki-specific alias, returns
1335 * the HTML equivalent. Otherwise, returns HTML-escaped text of
1336 * pseudo-entity source (eg &foo;)
1338 * @param $name String
1341 static function normalizeEntity( $name ) {
1342 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1343 return '&' . self
::$htmlEntityAliases[$name] . ';';
1344 } elseif ( in_array( $name,
1345 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1347 } elseif ( isset( self
::$htmlEntities[$name] ) ) {
1348 return '&#' . self
::$htmlEntities[$name] . ';';
1350 return "&$name;";
1356 * @return null|string
1358 static function decCharReference( $codepoint ) {
1359 $point = intval( $codepoint );
1360 if( Sanitizer
::validateCodepoint( $point ) ) {
1361 return sprintf( '&#%d;', $point );
1369 * @return null|string
1371 static function hexCharReference( $codepoint ) {
1372 $point = hexdec( $codepoint );
1373 if( Sanitizer
::validateCodepoint( $point ) ) {
1374 return sprintf( '&#x%x;', $point );
1381 * Returns true if a given Unicode codepoint is a valid character in XML.
1382 * @param $codepoint Integer
1385 private static function validateCodepoint( $codepoint ) {
1386 return ($codepoint == 0x09)
1387 ||
($codepoint == 0x0a)
1388 ||
($codepoint == 0x0d)
1389 ||
($codepoint >= 0x20 && $codepoint <= 0xd7ff)
1390 ||
($codepoint >= 0xe000 && $codepoint <= 0xfffd)
1391 ||
($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1395 * Decode any character references, numeric or named entities,
1396 * in the text and return a UTF-8 string.
1398 * @param $text String
1401 public static function decodeCharReferences( $text ) {
1402 return preg_replace_callback(
1403 self
::CHAR_REFS_REGEX
,
1404 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1409 * Decode any character references, numeric or named entities,
1410 * in the next and normalize the resulting string. (bug 14952)
1412 * This is useful for page titles, not for text to be displayed,
1413 * MediaWiki allows HTML entities to escape normalization as a feature.
1415 * @param $text String (already normalized, containing entities)
1416 * @return String (still normalized, without entities)
1418 public static function decodeCharReferencesAndNormalize( $text ) {
1420 $text = preg_replace_callback(
1421 self
::CHAR_REFS_REGEX
,
1422 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1423 $text, /* limit */ -1, $count );
1426 return $wgContLang->normalize( $text );
1433 * @param $matches String
1436 static function decodeCharReferencesCallback( $matches ) {
1437 if( $matches[1] != '' ) {
1438 return Sanitizer
::decodeEntity( $matches[1] );
1439 } elseif( $matches[2] != '' ) {
1440 return Sanitizer
::decodeChar( intval( $matches[2] ) );
1441 } elseif( $matches[3] != '' ) {
1442 return Sanitizer
::decodeChar( hexdec( $matches[3] ) );
1444 # Last case should be an ampersand by itself
1449 * Return UTF-8 string for a codepoint if that is a valid
1450 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1451 * @param $codepoint Integer
1455 static function decodeChar( $codepoint ) {
1456 if( Sanitizer
::validateCodepoint( $codepoint ) ) {
1457 return codepointToUtf8( $codepoint );
1459 return UTF8_REPLACEMENT
;
1464 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1465 * return the UTF-8 encoding of that character. Otherwise, returns
1466 * pseudo-entity source (eg &foo;)
1468 * @param $name String
1471 static function decodeEntity( $name ) {
1472 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1473 $name = self
::$htmlEntityAliases[$name];
1475 if( isset( self
::$htmlEntities[$name] ) ) {
1476 return codepointToUtf8( self
::$htmlEntities[$name] );
1483 * Fetch the whitelist of acceptable attributes for a given element name.
1485 * @param $element String
1488 static function attributeWhitelist( $element ) {
1489 $list = Sanitizer
::setupAttributeWhitelist();
1490 return isset( $list[$element] )
1496 * Foreach array key (an allowed HTML element), return an array
1497 * of allowed attributes
1500 static function setupAttributeWhitelist() {
1501 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1503 static $whitelist, $staticInitialised;
1504 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgHtml5', 'wgAllowMicrodataAttributes' ) );
1506 if ( isset( $whitelist ) && $staticInitialised == $globalContext ) {
1510 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1512 if ( $wgAllowRdfaAttributes ) {
1513 #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1514 $common = array_merge( $common, array(
1515 'about', 'property', 'resource', 'datatype', 'typeof',
1519 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1520 # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1521 $common = array_merge( $common, array(
1522 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1526 $block = array_merge( $common, array( 'align' ) );
1527 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1528 $tablecell = array( 'abbr',
1534 'nowrap', # deprecated
1535 'width', # deprecated
1536 'height', # deprecated
1537 'bgcolor' # deprecated
1540 # Numbers refer to sections in HTML 4.01 standard describing the element.
1541 # See: http://www.w3.org/TR/html4/
1545 'center' => $common, # deprecated
1546 'span' => $block, # ??
1564 'strong' => $common,
1575 'blockquote' => array_merge( $common, array( 'cite' ) ),
1586 'br' => array( 'id', 'class', 'title', 'style', 'clear' ),
1589 'pre' => array_merge( $common, array( 'width' ) ),
1592 'ins' => array_merge( $common, array( 'cite', 'datetime' ) ),
1593 'del' => array_merge( $common, array( 'cite', 'datetime' ) ),
1596 'ul' => array_merge( $common, array( 'type' ) ),
1597 'ol' => array_merge( $common, array( 'type', 'start' ) ),
1598 'li' => array_merge( $common, array( 'type', 'value' ) ),
1606 'table' => array_merge( $common,
1607 array( 'summary', 'width', 'border', 'frame',
1608 'rules', 'cellspacing', 'cellpadding',
1613 'caption' => array_merge( $common, array( 'align' ) ),
1616 'thead' => array_merge( $common, $tablealign ),
1617 'tfoot' => array_merge( $common, $tablealign ),
1618 'tbody' => array_merge( $common, $tablealign ),
1621 'colgroup' => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1622 'col' => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1625 'tr' => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1628 'td' => array_merge( $common, $tablecell, $tablealign ),
1629 'th' => array_merge( $common, $tablecell, $tablealign ),
1631 # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1632 'a' => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1635 # Not usually allowed, but may be used for extension-style hooks
1636 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1638 'img' => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1646 'strike' => $common,
1651 'font' => array_merge( $common, array( 'size', 'color', 'face' ) ),
1655 'hr' => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1657 # XHTML Ruby annotation text module, simple ruby only.
1658 # http://www.w3c.org/TR/ruby/
1663 'rt' => $common, #array_merge( $common, array( 'rbspan' ) ),
1666 # MathML root element, where used for extensions
1667 # 'title' may not be 100% valid here; it's XHTML
1668 # http://www.w3.org/TR/REC-MathML/
1669 'math' => array( 'class', 'style', 'id', 'title' ),
1673 # HTML5 elements, defined by:
1674 # http://www.whatwg.org/specs/web-apps/current-work/multipage/
1675 $whitelist +
= array(
1676 'data' => array_merge( $common, array( 'value' ) ),
1677 'time' => array_merge( $common, array( 'datetime' ) ),
1679 // meta and link are only present when Microdata is allowed anyways
1680 // so we don't bother adding another condition here
1681 // meta and link are only valid for use as Microdata so we do not
1682 // allow the common attributes here.
1683 'meta' => array( 'itemprop', 'content' ),
1684 'link' => array( 'itemprop', 'href' ),
1688 $staticInitialised = $globalContext;
1694 * Take a fragment of (potentially invalid) HTML and return
1695 * a version with any tags removed, encoded as plain text.
1697 * Warning: this return value must be further escaped for literal
1698 * inclusion in HTML output as of 1.10!
1700 * @param $text String: HTML fragment
1703 static function stripAllTags( $text ) {
1705 $text = StringUtils
::delimiterReplace( '<', '>', '', $text );
1707 # Normalize &entities and whitespace
1708 $text = self
::decodeCharReferences( $text );
1709 $text = self
::normalizeWhitespace( $text );
1715 * Hack up a private DOCTYPE with HTML's standard entity declarations.
1716 * PHP 4 seemed to know these if you gave it an HTML doctype, but
1719 * Use for passing XHTML fragments to PHP's XML parsing functions
1723 static function hackDocType() {
1724 $out = "<!DOCTYPE html [\n";
1725 foreach( self
::$htmlEntities as $entity => $codepoint ) {
1726 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1733 * @param $url string
1734 * @return mixed|string
1736 static function cleanUrl( $url ) {
1737 # Normalize any HTML entities in input. They will be
1738 # re-escaped by makeExternalLink().
1739 $url = Sanitizer
::decodeCharReferences( $url );
1741 # Escape any control characters introduced by the above step
1742 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1743 array( __CLASS__
, 'cleanUrlCallback' ), $url );
1745 # Validate hostname portion
1747 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1748 list( /* $whole */, $protocol, $host, $rest ) = $matches;
1750 // Characters that will be ignored in IDNs.
1751 // http://tools.ietf.org/html/3454#section-3.1
1752 // Strip them before further processing so blacklists and such work.
1754 \\s| # general whitespace
1755 \xc2\xad| # 00ad SOFT HYPHEN
1756 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1757 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1758 \xe2\x81\xa0| # 2060 WORD JOINER
1759 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1760 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
1761 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1762 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1763 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1764 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1765 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1766 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1769 $host = preg_replace( $strip, '', $host );
1771 // @todo FIXME: Validate hostnames here
1773 return $protocol . $host . $rest;
1780 * @param $matches array
1783 static function cleanUrlCallback( $matches ) {
1784 return urlencode( $matches[0] );
1788 * Does a string look like an e-mail address?
1790 * This validates an email address using an HTML5 specification found at:
1791 * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1792 * Which as of 2011-01-24 says:
1794 * A valid e-mail address is a string that matches the ABNF production
1795 * 1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1796 * in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1799 * This function is an implementation of the specification as requested in
1802 * Client-side forms will use the same standard validation rules via JS or
1803 * HTML 5 validation; additional restrictions can be enforced server-side
1804 * by extensions via the 'isValidEmailAddr' hook.
1806 * Note that this validation doesn't 100% match RFC 2822, but is believed
1807 * to be liberal enough for wide use. Some invalid addresses will still
1808 * pass validation here.
1812 * @param $addr String E-mail address
1815 public static function validateEmail( $addr ) {
1817 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1821 // Please note strings below are enclosed in brackets [], this make the
1822 // hyphen "-" a range indicator. Hence it is double backslashed below.
1824 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1825 $rfc1034_ldh_str = "a-z0-9\\-" ;
1827 $HTML5_email_regexp = "/
1829 [$rfc5322_atext\\.]+ # user part which is liberal :p
1831 [$rfc1034_ldh_str]+ # First domain part
1832 (\\.[$rfc1034_ldh_str]+)* # Following part prefixed with a dot
1834 /ix" ; // case Insensitive, eXtended
1836 return (bool) preg_match( $HTML5_email_regexp, $addr );