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;
567 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
568 if( is_callable( $processCallback ) ) {
569 call_user_func_array( $processCallback, array( &$params, $args ) );
571 $newparams = Sanitizer
::fixTagAttributes( $params, $t );
572 $rest = str_replace( '>', '>', $rest );
573 $text .= "<$slash$t$newparams$brace$rest";
575 $text .= '<' . str_replace( '>', '>', $x);
579 wfProfileOut( __METHOD__
);
584 * Remove '<!--', '-->', and everything between.
585 * To avoid leaving blank lines, when a comment is both preceded
586 * and followed by a newline (ignoring spaces), trim leading and
587 * trailing spaces and one of the newlines.
590 * @param $text String
593 static function removeHTMLcomments( $text ) {
594 wfProfileIn( __METHOD__
);
595 while (($start = strpos($text, '<!--')) !== false) {
596 $end = strpos($text, '-->', $start +
4);
597 if ($end === false) {
598 # Unterminated comment; bail out
604 # Trim space and newline if the comment is both
605 # preceded and followed by a newline
606 $spaceStart = max($start - 1, 0);
607 $spaceLen = $end - $spaceStart;
608 while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
612 while (substr($text, $spaceStart +
$spaceLen, 1) === ' ')
614 if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart +
$spaceLen, 1) === "\n") {
615 # Remove the comment, leading and trailing
616 # spaces, and leave only one newline.
617 $text = substr_replace($text, "\n", $spaceStart, $spaceLen +
1);
620 # Remove just the comment.
621 $text = substr_replace($text, '', $start, $end - $start);
624 wfProfileOut( __METHOD__
);
629 * Take an array of attribute names and values and fix some deprecated values
630 * for the given element type.
631 * This does not validate properties, so you should ensure that you call
632 * validateTagAttributes AFTER this to ensure that the resulting style rule
633 * this may add is safe.
635 * - Converts most presentational attributes like align into inline css
637 * @param $attribs Array
638 * @param $element String
641 static function fixDeprecatedAttributes( $attribs, $element ) {
642 global $wgHtml5, $wgCleanupPresentationalAttributes;
644 // presentational attributes were removed from html5, we can leave them
645 // in when html5 is turned off
646 if ( !$wgHtml5 ||
!$wgCleanupPresentationalAttributes ) {
650 $table = array( 'table' );
651 $cells = array( 'td', 'th' );
652 $colls = array( 'col', 'colgroup' );
653 $tblocks = array( 'tbody', 'tfoot', 'thead' );
654 $h = array( 'h1', 'h2', 'h3', 'h4', 'h5', 'h6' );
656 $presentationalAttribs = array(
657 'align' => array( 'text-align', array_merge( array( 'caption', 'hr', 'div', 'p', 'tr' ), $table, $cells, $colls, $tblocks, $h ) ),
658 'clear' => array( 'clear', array( 'br' ) ),
659 'height' => array( 'height', $cells ),
660 'nowrap' => array( 'white-space', $cells ),
661 'size' => array( 'height', array( 'hr' ) ),
662 'type' => array( 'list-style-type', array( 'li', 'ol', 'ul' ) ),
663 'valign' => array( 'vertical-align', array_merge( $cells, $colls, $tblocks ) ),
664 'width' => array( 'width', array_merge( array( 'hr', 'pre' ), $table, $cells, $colls ) ),
667 // Ensure that any upper case or mixed case attributes are converted to lowercase
668 foreach ( $attribs as $attribute => $value ) {
669 if ( $attribute !== strtolower( $attribute ) && array_key_exists( strtolower( $attribute ), $presentationalAttribs ) ) {
670 $attribs[strtolower( $attribute )] = $value;
671 unset( $attribs[$attribute] );
676 foreach ( $presentationalAttribs as $attribute => $info ) {
677 list( $property, $elements ) = $info;
679 // Skip if this attribute is not relevant to this element
680 if ( !in_array( $element, $elements ) ) {
684 // Skip if the attribute is not used
685 if ( !array_key_exists( $attribute, $attribs ) ) {
689 $value = $attribs[$attribute];
691 // For nowrap the value should be nowrap instead of whatever text is in the value
692 if ( $attribute === 'nowrap' ) {
696 // clear="all" is clear: both; in css
697 if ( $attribute === 'clear' && strtolower( $value ) === 'all' ) {
701 // Size based properties should have px applied to them if they have no unit
702 if ( in_array( $attribute, array( 'height', 'width', 'size' ) ) ) {
703 if ( preg_match( '/^[\d.]+$/', $value ) ) {
704 $value = "{$value}px";
708 $style .= " $property: $value;";
710 unset( $attribs[$attribute] );
714 // Prepend our style rules so that they can be overridden by user css
715 if ( isset($attribs['style']) ) {
716 $style .= " " . $attribs['style'];
718 $attribs['style'] = trim($style);
725 * Takes attribute names and values for a tag and the tah name and
726 * validates that the tag is allowed to be present.
727 * This DOES NOT validate the attributes, nor does it validate the
728 * tags themselves. This method only handles the special circumstances
729 * where we may want to allow a tag within content but ONLY when it has
730 * specific attributes set.
734 static function validateTag( $params, $element ) {
735 $params = Sanitizer
::decodeTagAttributes( $params );
737 if ( $element == 'meta' ||
$element == 'link' ) {
738 if ( !isset( $params['itemprop'] ) ) {
739 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
742 if ( $element == 'meta' && !isset( $params['content'] ) ) {
743 // <meta> must have a content="" for the itemprop
746 if ( $element == 'link' && !isset( $params['href'] ) ) {
747 // <link> must have an associated href=""
756 * Take an array of attribute names and values and normalize or discard
757 * illegal values for the given element type.
759 * - Discards attributes not on a whitelist for the given element
760 * - Unsafe style attributes are discarded
761 * - Invalid id attributes are reencoded
763 * @param $attribs Array
764 * @param $element String
767 * @todo Check for legal values where the DTD limits things.
768 * @todo Check for unique id attribute :P
770 static function validateTagAttributes( $attribs, $element ) {
771 return Sanitizer
::validateAttributes( $attribs,
772 Sanitizer
::attributeWhitelist( $element ) );
776 * Take an array of attribute names and values and normalize or discard
777 * illegal values for the given whitelist.
779 * - Discards attributes not the given whitelist
780 * - Unsafe style attributes are discarded
781 * - Invalid id attributes are reencoded
783 * @param $attribs Array
784 * @param $whitelist Array: list of allowed attribute names
787 * @todo Check for legal values where the DTD limits things.
788 * @todo Check for unique id attribute :P
790 static function validateAttributes( $attribs, $whitelist ) {
791 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
793 $whitelist = array_flip( $whitelist );
794 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
797 foreach( $attribs as $attribute => $value ) {
798 #allow XML namespace declaration if RDFa is enabled
799 if ( $wgAllowRdfaAttributes && preg_match( self
::XMLNS_ATTRIBUTE_PATTERN
, $attribute ) ) {
800 if ( !preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
801 $out[$attribute] = $value;
807 # Allow any attribute beginning with "data-", if in HTML5 mode
808 if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
812 # Strip javascript "expression" from stylesheets.
813 # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
814 if( $attribute == 'style' ) {
815 $value = Sanitizer
::checkCss( $value );
818 if ( $attribute === 'id' ) {
819 $value = Sanitizer
::escapeId( $value, 'noninitial' );
822 //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
823 if ( $attribute === 'rel' ||
$attribute === 'rev' ||
824 $attribute === 'about' ||
$attribute === 'property' ||
$attribute === 'resource' ||
#RDFa
825 $attribute === 'datatype' ||
$attribute === 'typeof' ||
#RDFa
826 $attribute === 'itemid' ||
$attribute === 'itemprop' ||
$attribute === 'itemref' ||
#HTML5 microdata
827 $attribute === 'itemscope' ||
$attribute === 'itemtype' ) { #HTML5 microdata
829 //Paranoia. Allow "simple" values but suppress javascript
830 if ( preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
835 # NOTE: even though elements using href/src are not allowed directly, supply
836 # validation code that can be used by tag hook handlers, etc
837 if ( $attribute === 'href' ||
$attribute === 'src' ) {
838 if ( !preg_match( $hrefExp, $value ) ) {
839 continue; //drop any href or src attributes not using an allowed protocol.
840 //NOTE: this also drops all relative URLs
844 // If this attribute was previously set, override it.
845 // Output should only have one attribute of each name.
846 $out[$attribute] = $value;
849 if ( $wgAllowMicrodataAttributes ) {
850 # itemtype, itemid, itemref don't make sense without itemscope
851 if ( !array_key_exists( 'itemscope', $out ) ) {
852 unset( $out['itemtype'] );
853 unset( $out['itemid'] );
854 unset( $out['itemref'] );
856 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
862 * Merge two sets of HTML attributes. Conflicting items in the second set
863 * will override those in the first, except for 'class' attributes which
864 * will be combined (if they're both strings).
866 * @todo implement merging for other attributes such as style
871 static function mergeAttributes( $a, $b ) {
872 $out = array_merge( $a, $b );
873 if( isset( $a['class'] ) && isset( $b['class'] )
874 && is_string( $a['class'] ) && is_string( $b['class'] )
875 && $a['class'] !== $b['class'] ) {
876 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
877 -1, PREG_SPLIT_NO_EMPTY
);
878 $out['class'] = implode( ' ', array_unique( $classes ) );
884 * Pick apart some CSS and check it for forbidden or unsafe structures.
885 * Returns a sanitized string. This sanitized string will have
886 * character references and escape sequences decoded, and comments
887 * stripped. If the input is just too evil, only a comment complaining
888 * about evilness will be returned.
890 * Currently URL references, 'expression', 'tps' are forbidden.
892 * NOTE: Despite the fact that character references are decoded, the
893 * returned string may contain character references given certain
894 * clever input strings. These character references must
895 * be escaped before the return value is embedded in HTML.
897 * @param $value String
900 static function checkCss( $value ) {
901 // Decode character references like {
902 $value = Sanitizer
::decodeCharReferences( $value );
904 // Decode escape sequences and line continuation
905 // See the grammar in the CSS 2 spec, appendix D.
906 // This has to be done AFTER decoding character references.
907 // This means it isn't possible for this function to return
908 // unsanitized escape sequences. It is possible to manufacture
909 // input that contains character references that decode to
910 // escape sequences that decode to character references, but
911 // it's OK for the return value to contain character references
912 // because the caller is supposed to escape those anyway.
914 if ( !$decodeRegex ) {
915 $space = '[\\x20\\t\\r\\n\\f]';
916 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
918 $decodeRegex = "/ $backslash
920 ($nl) | # 1. Line continuation
921 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
922 (.) | # 3. backslash cancelling special meaning
923 () | # 4. backslash at end of string
926 $value = preg_replace_callback( $decodeRegex,
927 array( __CLASS__
, 'cssDecodeCallback' ), $value );
929 // Remove any comments; IE gets token splitting wrong
930 // This must be done AFTER decoding character references and
931 // escape sequences, because those steps can introduce comments
932 // This step cannot introduce character references or escape
933 // sequences, because it replaces comments with spaces rather
934 // than removing them completely.
935 $value = StringUtils
::delimiterReplace( '/*', '*/', ' ', $value );
937 // Remove anything after a comment-start token, to guard against
938 // incorrect client implementations.
939 $commentPos = strpos( $value, '/*' );
940 if ( $commentPos !== false ) {
941 $value = substr( $value, 0, $commentPos );
944 // Reject problematic keywords and control characters
945 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
946 return '/* invalid control char */';
947 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
948 return '/* insecure input */';
954 * @param $matches array
957 static function cssDecodeCallback( $matches ) {
958 if ( $matches[1] !== '' ) {
961 } elseif ( $matches[2] !== '' ) {
962 $char = codepointToUtf8( hexdec( $matches[2] ) );
963 } elseif ( $matches[3] !== '' ) {
968 if ( $char == "\n" ||
$char == '"' ||
$char == "'" ||
$char == '\\' ) {
969 // These characters need to be escaped in strings
970 // Clean up the escape sequence to avoid parsing errors by clients
971 return '\\' . dechex( ord( $char ) ) . ' ';
973 // Decode unnecessary escape
979 * Take a tag soup fragment listing an HTML element's attributes
980 * and normalize it to well-formed XML, discarding unwanted attributes.
981 * Output is safe for further wikitext processing, with escaping of
982 * values that could trigger problems.
984 * - Normalizes attribute names to lowercase
985 * - Discards attributes not on a whitelist for the given element
986 * - Turns broken or invalid entities into plaintext
987 * - Double-quotes all attribute values
988 * - Attributes without values are given the name as attribute
989 * - Double attributes are discarded
990 * - Unsafe style attributes are discarded
991 * - Prepends space if there are attributes.
993 * @param $text String
994 * @param $element String
997 static function fixTagAttributes( $text, $element ) {
998 if( trim( $text ) == '' ) {
1002 $decoded = Sanitizer
::decodeTagAttributes( $text );
1003 $decoded = Sanitizer
::fixDeprecatedAttributes( $decoded, $element );
1004 $stripped = Sanitizer
::validateTagAttributes( $decoded, $element );
1007 foreach( $stripped as $attribute => $value ) {
1008 $encAttribute = htmlspecialchars( $attribute );
1009 $encValue = Sanitizer
::safeEncodeAttribute( $value );
1011 $attribs[] = "$encAttribute=\"$encValue\"";
1013 return count( $attribs ) ?
' ' . implode( ' ', $attribs ) : '';
1017 * Encode an attribute value for HTML output.
1018 * @param $text String
1019 * @return HTML-encoded text fragment
1021 static function encodeAttribute( $text ) {
1022 $encValue = htmlspecialchars( $text, ENT_QUOTES
);
1024 // Whitespace is normalized during attribute decoding,
1025 // so if we've been passed non-spaces we must encode them
1026 // ahead of time or they won't be preserved.
1027 $encValue = strtr( $encValue, array(
1037 * Encode an attribute value for HTML tags, with extra armoring
1038 * against further wiki processing.
1039 * @param $text String
1040 * @return HTML-encoded text fragment
1042 static function safeEncodeAttribute( $text ) {
1043 $encValue = Sanitizer
::encodeAttribute( $text );
1045 # Templates and links may be expanded in later parsing,
1046 # creating invalid or dangerous output. Suppress this.
1047 $encValue = strtr( $encValue, array(
1048 '<' => '<', // This should never happen,
1049 '>' => '>', // we've received invalid input
1050 '"' => '"', // which should have been escaped.
1053 "''" => '''',
1054 'ISBN' => 'ISBN',
1056 'PMID' => 'PMID',
1062 $encValue = preg_replace_callback(
1063 '/(' . wfUrlProtocols() . ')/',
1064 array( 'Sanitizer', 'armorLinksCallback' ),
1070 * Given a value, escape it so that it can be used in an id attribute and
1071 * return it. This will use HTML5 validation if $wgExperimentalHtmlIds is
1072 * true, allowing anything but ASCII whitespace. Otherwise it will use
1073 * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1074 * escaped with lots of dots.
1076 * To ensure we don't have to bother escaping anything, we also strip ', ",
1077 * & even if $wgExperimentalIds is true. TODO: Is this the best tactic?
1078 * We also strip # because it upsets IE, and % because it could be
1079 * ambiguous if it's part of something that looks like a percent escape
1080 * (which don't work reliably in fragments cross-browser).
1082 * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1085 * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1086 * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1087 * HTML5 definition of id attribute
1089 * @param $id String: id to escape
1090 * @param $options Mixed: string or array of strings (default is array()):
1091 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1092 * so don't pay attention if the first character isn't valid at the
1093 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
1095 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
1096 * if $wgExperimentalHtmlIds is used, so we can generate extra
1097 * anchors and links won't break.
1100 static function escapeId( $id, $options = array() ) {
1101 global $wgHtml5, $wgExperimentalHtmlIds;
1102 $options = (array)$options;
1104 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1105 $id = Sanitizer
::decodeCharReferences( $id );
1106 $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1107 $id = trim( $id, '_' );
1109 # Must have been all whitespace to start with.
1116 # HTML4-style escaping
1117 static $replace = array(
1122 $id = urlencode( Sanitizer
::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1123 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1125 if ( !preg_match( '/^[a-zA-Z]/', $id )
1126 && !in_array( 'noninitial', $options ) ) {
1127 // Initial character must be a letter!
1134 * Given a value, escape it so that it can be used as a CSS class and
1137 * @todo For extra validity, input should be validated UTF-8.
1139 * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1141 * @param $class String
1144 static function escapeClass( $class ) {
1145 // Convert ugly stuff to underscores and kill underscores in ugly places
1146 return rtrim(preg_replace(
1147 array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
1153 * Given HTML input, escape with htmlspecialchars but un-escape entites.
1154 * This allows (generally harmless) entities like   to survive.
1156 * @param $html String to escape
1157 * @return String: escaped input
1159 static function escapeHtmlAllowEntities( $html ) {
1160 $html = Sanitizer
::decodeCharReferences( $html );
1161 # It seems wise to escape ' as well as ", as a matter of course. Can't
1163 $html = htmlspecialchars( $html, ENT_QUOTES
);
1168 * Regex replace callback for armoring links against further processing.
1169 * @param $matches Array
1172 private static function armorLinksCallback( $matches ) {
1173 return str_replace( ':', ':', $matches[1] );
1177 * Return an associative array of attribute names and values from
1178 * a partial tag string. Attribute names are forces to lowercase,
1179 * character references are decoded to UTF-8 text.
1181 * @param $text String
1184 public static function decodeTagAttributes( $text ) {
1185 if( trim( $text ) == '' ) {
1191 if( !preg_match_all(
1192 self
::getAttribsRegex(),
1195 PREG_SET_ORDER
) ) {
1199 foreach( $pairs as $set ) {
1200 $attribute = strtolower( $set[1] );
1201 $value = Sanitizer
::getTagAttributeCallback( $set );
1203 // Normalize whitespace
1204 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1205 $value = trim( $value );
1207 // Decode character references
1208 $attribs[$attribute] = Sanitizer
::decodeCharReferences( $value );
1214 * Pick the appropriate attribute value from a match set from the
1215 * attribs regex matches.
1220 private static function getTagAttributeCallback( $set ) {
1221 if( isset( $set[6] ) ) {
1222 # Illegal #XXXXXX color with no quotes.
1224 } elseif( isset( $set[5] ) ) {
1227 } elseif( isset( $set[4] ) ) {
1230 } elseif( isset( $set[3] ) ) {
1233 } elseif( !isset( $set[2] ) ) {
1234 # In XHTML, attributes must have a value.
1235 # For 'reduced' form, return explicitly the attribute name here.
1238 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1243 * Normalize whitespace and character references in an XML source-
1244 * encoded text for an attribute value.
1246 * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1247 * but note that we're not returning the value, but are returning
1248 * XML source fragments that will be slapped into output.
1250 * @param $text String
1253 private static function normalizeAttributeValue( $text ) {
1254 return str_replace( '"', '"',
1255 self
::normalizeWhitespace(
1256 Sanitizer
::normalizeCharReferences( $text ) ) );
1260 * @param $text string
1263 private static function normalizeWhitespace( $text ) {
1264 return preg_replace(
1265 '/\r\n|[\x20\x0d\x0a\x09]/',
1271 * Normalizes whitespace in a section name, such as might be returned
1272 * by Parser::stripSectionName(), for use in the id's that are used for
1275 * @param $section String
1278 static function normalizeSectionNameWhitespace( $section ) {
1279 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1283 * Ensure that any entities and character references are legal
1284 * for XML and XHTML specifically. Any stray bits will be
1285 * &-escaped to result in a valid text fragment.
1287 * a. named char refs can only be < > & ", others are
1288 * numericized (this way we're well-formed even without a DTD)
1289 * b. any numeric char refs must be legal chars, not invalid or forbidden
1290 * c. use &#x, not &#X
1291 * d. fix or reject non-valid attributes
1293 * @param $text String
1297 static function normalizeCharReferences( $text ) {
1298 return preg_replace_callback(
1299 self
::CHAR_REFS_REGEX
,
1300 array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1304 * @param $matches String
1307 static function normalizeCharReferencesCallback( $matches ) {
1309 if( $matches[1] != '' ) {
1310 $ret = Sanitizer
::normalizeEntity( $matches[1] );
1311 } elseif( $matches[2] != '' ) {
1312 $ret = Sanitizer
::decCharReference( $matches[2] );
1313 } elseif( $matches[3] != '' ) {
1314 $ret = Sanitizer
::hexCharReference( $matches[3] );
1316 if( is_null( $ret ) ) {
1317 return htmlspecialchars( $matches[0] );
1324 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1325 * return the equivalent numeric entity reference (except for the core <
1326 * > & "). If the entity is a MediaWiki-specific alias, returns
1327 * the HTML equivalent. Otherwise, returns HTML-escaped text of
1328 * pseudo-entity source (eg &foo;)
1330 * @param $name String
1333 static function normalizeEntity( $name ) {
1334 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1335 return '&' . self
::$htmlEntityAliases[$name] . ';';
1336 } elseif ( in_array( $name,
1337 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1339 } elseif ( isset( self
::$htmlEntities[$name] ) ) {
1340 return '&#' . self
::$htmlEntities[$name] . ';';
1342 return "&$name;";
1348 * @return null|string
1350 static function decCharReference( $codepoint ) {
1351 $point = intval( $codepoint );
1352 if( Sanitizer
::validateCodepoint( $point ) ) {
1353 return sprintf( '&#%d;', $point );
1361 * @return null|string
1363 static function hexCharReference( $codepoint ) {
1364 $point = hexdec( $codepoint );
1365 if( Sanitizer
::validateCodepoint( $point ) ) {
1366 return sprintf( '&#x%x;', $point );
1373 * Returns true if a given Unicode codepoint is a valid character in XML.
1374 * @param $codepoint Integer
1377 private static function validateCodepoint( $codepoint ) {
1378 return ($codepoint == 0x09)
1379 ||
($codepoint == 0x0a)
1380 ||
($codepoint == 0x0d)
1381 ||
($codepoint >= 0x20 && $codepoint <= 0xd7ff)
1382 ||
($codepoint >= 0xe000 && $codepoint <= 0xfffd)
1383 ||
($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1387 * Decode any character references, numeric or named entities,
1388 * in the text and return a UTF-8 string.
1390 * @param $text String
1393 public static function decodeCharReferences( $text ) {
1394 return preg_replace_callback(
1395 self
::CHAR_REFS_REGEX
,
1396 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1401 * Decode any character references, numeric or named entities,
1402 * in the next and normalize the resulting string. (bug 14952)
1404 * This is useful for page titles, not for text to be displayed,
1405 * MediaWiki allows HTML entities to escape normalization as a feature.
1407 * @param $text String (already normalized, containing entities)
1408 * @return String (still normalized, without entities)
1410 public static function decodeCharReferencesAndNormalize( $text ) {
1412 $text = preg_replace_callback(
1413 self
::CHAR_REFS_REGEX
,
1414 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1415 $text, /* limit */ -1, $count );
1418 return $wgContLang->normalize( $text );
1425 * @param $matches String
1428 static function decodeCharReferencesCallback( $matches ) {
1429 if( $matches[1] != '' ) {
1430 return Sanitizer
::decodeEntity( $matches[1] );
1431 } elseif( $matches[2] != '' ) {
1432 return Sanitizer
::decodeChar( intval( $matches[2] ) );
1433 } elseif( $matches[3] != '' ) {
1434 return Sanitizer
::decodeChar( hexdec( $matches[3] ) );
1436 # Last case should be an ampersand by itself
1441 * Return UTF-8 string for a codepoint if that is a valid
1442 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1443 * @param $codepoint Integer
1447 static function decodeChar( $codepoint ) {
1448 if( Sanitizer
::validateCodepoint( $codepoint ) ) {
1449 return codepointToUtf8( $codepoint );
1451 return UTF8_REPLACEMENT
;
1456 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1457 * return the UTF-8 encoding of that character. Otherwise, returns
1458 * pseudo-entity source (eg &foo;)
1460 * @param $name String
1463 static function decodeEntity( $name ) {
1464 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1465 $name = self
::$htmlEntityAliases[$name];
1467 if( isset( self
::$htmlEntities[$name] ) ) {
1468 return codepointToUtf8( self
::$htmlEntities[$name] );
1475 * Fetch the whitelist of acceptable attributes for a given element name.
1477 * @param $element String
1480 static function attributeWhitelist( $element ) {
1481 $list = Sanitizer
::setupAttributeWhitelist();
1482 return isset( $list[$element] )
1488 * Foreach array key (an allowed HTML element), return an array
1489 * of allowed attributes
1492 static function setupAttributeWhitelist() {
1493 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1495 static $whitelist, $staticInitialised;
1496 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgHtml5', 'wgAllowMicrodataAttributes' ) );
1498 if ( isset( $whitelist ) && $staticInitialised == $globalContext ) {
1502 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1504 if ( $wgAllowRdfaAttributes ) {
1505 #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1506 $common = array_merge( $common, array(
1507 'about', 'property', 'resource', 'datatype', 'typeof',
1511 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1512 # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1513 $common = array_merge( $common, array(
1514 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1518 $block = array_merge( $common, array( 'align' ) );
1519 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1520 $tablecell = array( 'abbr',
1526 'nowrap', # deprecated
1527 'width', # deprecated
1528 'height', # deprecated
1529 'bgcolor' # deprecated
1532 # Numbers refer to sections in HTML 4.01 standard describing the element.
1533 # See: http://www.w3.org/TR/html4/
1537 'center' => $common, # deprecated
1538 'span' => $block, # ??
1556 'strong' => $common,
1567 'blockquote' => array_merge( $common, array( 'cite' ) ),
1578 'br' => array( 'id', 'class', 'title', 'style', 'clear' ),
1581 'pre' => array_merge( $common, array( 'width' ) ),
1584 'ins' => array_merge( $common, array( 'cite', 'datetime' ) ),
1585 'del' => array_merge( $common, array( 'cite', 'datetime' ) ),
1588 'ul' => array_merge( $common, array( 'type' ) ),
1589 'ol' => array_merge( $common, array( 'type', 'start' ) ),
1590 'li' => array_merge( $common, array( 'type', 'value' ) ),
1598 'table' => array_merge( $common,
1599 array( 'summary', 'width', 'border', 'frame',
1600 'rules', 'cellspacing', 'cellpadding',
1605 'caption' => array_merge( $common, array( 'align' ) ),
1608 'thead' => array_merge( $common, $tablealign ),
1609 'tfoot' => array_merge( $common, $tablealign ),
1610 'tbody' => array_merge( $common, $tablealign ),
1613 'colgroup' => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1614 'col' => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1617 'tr' => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1620 'td' => array_merge( $common, $tablecell, $tablealign ),
1621 'th' => array_merge( $common, $tablecell, $tablealign ),
1623 # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1624 'a' => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1627 # Not usually allowed, but may be used for extension-style hooks
1628 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1630 'img' => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1638 'strike' => $common,
1643 'font' => array_merge( $common, array( 'size', 'color', 'face' ) ),
1647 'hr' => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1649 # XHTML Ruby annotation text module, simple ruby only.
1650 # http://www.w3c.org/TR/ruby/
1655 'rt' => $common, #array_merge( $common, array( 'rbspan' ) ),
1658 # MathML root element, where used for extensions
1659 # 'title' may not be 100% valid here; it's XHTML
1660 # http://www.w3.org/TR/REC-MathML/
1661 'math' => array( 'class', 'style', 'id', 'title' ),
1665 # HTML5 elements, defined by:
1666 # http://www.whatwg.org/specs/web-apps/current-work/multipage/
1667 $whitelist +
= array(
1668 'data' => array_merge( $common, array( 'value' ) ),
1669 'time' => array_merge( $common, array( 'datetime' ) ),
1671 // meta and link are only present when Microdata is allowed anyways
1672 // so we don't bother adding another condition here
1673 // meta and link are only valid for use as Microdata so we do not
1674 // allow the common attributes here.
1675 'meta' => array( 'itemprop', 'content' ),
1676 'link' => array( 'itemprop', 'href' ),
1680 $staticInitialised = $globalContext;
1686 * Take a fragment of (potentially invalid) HTML and return
1687 * a version with any tags removed, encoded as plain text.
1689 * Warning: this return value must be further escaped for literal
1690 * inclusion in HTML output as of 1.10!
1692 * @param $text String: HTML fragment
1695 static function stripAllTags( $text ) {
1697 $text = StringUtils
::delimiterReplace( '<', '>', '', $text );
1699 # Normalize &entities and whitespace
1700 $text = self
::decodeCharReferences( $text );
1701 $text = self
::normalizeWhitespace( $text );
1707 * Hack up a private DOCTYPE with HTML's standard entity declarations.
1708 * PHP 4 seemed to know these if you gave it an HTML doctype, but
1711 * Use for passing XHTML fragments to PHP's XML parsing functions
1715 static function hackDocType() {
1716 $out = "<!DOCTYPE html [\n";
1717 foreach( self
::$htmlEntities as $entity => $codepoint ) {
1718 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1725 * @param $url string
1726 * @return mixed|string
1728 static function cleanUrl( $url ) {
1729 # Normalize any HTML entities in input. They will be
1730 # re-escaped by makeExternalLink().
1731 $url = Sanitizer
::decodeCharReferences( $url );
1733 # Escape any control characters introduced by the above step
1734 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1735 array( __CLASS__
, 'cleanUrlCallback' ), $url );
1737 # Validate hostname portion
1739 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1740 list( /* $whole */, $protocol, $host, $rest ) = $matches;
1742 // Characters that will be ignored in IDNs.
1743 // http://tools.ietf.org/html/3454#section-3.1
1744 // Strip them before further processing so blacklists and such work.
1746 \\s| # general whitespace
1747 \xc2\xad| # 00ad SOFT HYPHEN
1748 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1749 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1750 \xe2\x81\xa0| # 2060 WORD JOINER
1751 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1752 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
1753 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1754 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1755 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1756 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1757 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1758 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1761 $host = preg_replace( $strip, '', $host );
1763 // @todo FIXME: Validate hostnames here
1765 return $protocol . $host . $rest;
1772 * @param $matches array
1775 static function cleanUrlCallback( $matches ) {
1776 return urlencode( $matches[0] );
1780 * Does a string look like an e-mail address?
1782 * This validates an email address using an HTML5 specification found at:
1783 * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1784 * Which as of 2011-01-24 says:
1786 * A valid e-mail address is a string that matches the ABNF production
1787 * 1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1788 * in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1791 * This function is an implementation of the specification as requested in
1794 * Client-side forms will use the same standard validation rules via JS or
1795 * HTML 5 validation; additional restrictions can be enforced server-side
1796 * by extensions via the 'isValidEmailAddr' hook.
1798 * Note that this validation doesn't 100% match RFC 2822, but is believed
1799 * to be liberal enough for wide use. Some invalid addresses will still
1800 * pass validation here.
1804 * @param $addr String E-mail address
1807 public static function validateEmail( $addr ) {
1809 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1813 // Please note strings below are enclosed in brackets [], this make the
1814 // hyphen "-" a range indicator. Hence it is double backslashed below.
1816 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1817 $rfc1034_ldh_str = "a-z0-9\\-" ;
1819 $HTML5_email_regexp = "/
1821 [$rfc5322_atext\\.]+ # user part which is liberal :p
1823 [$rfc1034_ldh_str]+ # First domain part
1824 (\\.[$rfc1034_ldh_str]+)* # Following part prefixed with a dot
1826 /ix" ; // case Insensitive, eXtended
1828 return (bool) preg_match( $HTML5_email_regexp, $addr );