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() ) {
369 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
370 $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
372 wfProfileIn( __METHOD__
);
374 if ( !$staticInitialised ) {
376 $htmlpairsStatic = array( # Tags that must be closed
377 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
378 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
379 'strike', 'strong', 'tt', 'var', 'div', 'center',
380 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
381 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'abbr', 'dfn',
385 'br', 'hr', 'li', 'dt', 'dd'
387 $htmlsingleonly = array( # Elements that cannot have close tags
390 $htmlnest = array( # Tags that can be nested--??
391 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
392 'dl', 'font', 'big', 'small', 'sub', 'sup', 'span'
394 $tabletags = array( # Can only appear inside table, we will close them
397 $htmllist = array( # Tags used by list
400 $listtags = array( # Tags that can appear in a list
404 global $wgAllowImageTag;
405 if ( $wgAllowImageTag ) {
406 $htmlsingle[] = 'img';
407 $htmlsingleonly[] = 'img';
410 $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
411 $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
413 # Convert them all to hashtables for faster lookup
414 $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
415 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
416 foreach ( $vars as $var ) {
417 $
$var = array_flip( $
$var );
419 $staticInitialised = true;
421 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
422 $extratags = array_flip( $extratags );
423 $removetags = array_flip( $removetags );
424 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
425 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ) , $removetags );
427 # Remove HTML comments
428 $text = Sanitizer
::removeHTMLcomments( $text );
429 $bits = explode( '<', $text );
430 $text = str_replace( '>', '>', array_shift( $bits ) );
432 $tagstack = $tablestack = array();
433 foreach ( $bits as $x ) {
435 # $slash: Does the current element start with a '/'?
436 # $t: Current element name
437 # $params: String between element name and >
438 # $brace: Ending '>' or '/>'
439 # $rest: Everything until the next element of $bits
440 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
441 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
443 $slash = $t = $params = $brace = $rest = null;
447 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
449 if ( $slash && isset( $htmlsingleonly[$t] ) ) {
451 } elseif ( $slash ) {
452 # Closing a tag... is it the one we just opened?
453 $ot = @array_pop
( $tagstack );
455 if ( isset( $htmlsingleallowed[$ot] ) ) {
456 # Pop all elements with an optional close tag
457 # and see if we find a match below them
459 array_push( $optstack, $ot );
460 wfSuppressWarnings();
461 $ot = array_pop( $tagstack );
463 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
464 array_push( $optstack, $ot );
465 wfSuppressWarnings();
466 $ot = array_pop( $tagstack );
470 # No match. Push the optional elements back again
472 wfSuppressWarnings();
473 $ot = array_pop( $optstack );
476 array_push( $tagstack, $ot );
477 wfSuppressWarnings();
478 $ot = array_pop( $optstack );
483 @array_push
( $tagstack, $ot );
484 # <li> can be nested in <ul> or <ol>, skip those cases:
485 if ( !isset( $htmllist[$ot] ) ||
!isset( $listtags[$t] ) ) {
490 if ( $t == 'table' ) {
491 $tagstack = array_pop( $tablestack );
496 # Keep track for later
497 if ( isset( $tabletags[$t] ) &&
498 !in_array( 'table', $tagstack ) ) {
500 } elseif ( in_array( $t, $tagstack ) &&
501 !isset( $htmlnest [$t ] ) ) {
503 # Is it a self closed htmlpair ? (bug 5487)
504 } elseif ( $brace == '/>' &&
505 isset( $htmlpairs[$t] ) ) {
507 } elseif ( isset( $htmlsingleonly[$t] ) ) {
508 # Hack to force empty tag for uncloseable elements
510 } elseif ( isset( $htmlsingle[$t] ) ) {
511 # Hack to not close $htmlsingle tags
513 } elseif ( isset( $tabletags[$t] )
514 && in_array( $t, $tagstack ) ) {
515 // New table tag but forgot to close the previous one
518 if ( $t == 'table' ) {
519 array_push( $tablestack, $tagstack );
522 array_push( $tagstack, $t );
525 # Replace any variables or template parameters with
527 if( is_callable( $processCallback ) ) {
528 call_user_func_array( $processCallback, array( &$params, $args ) );
531 # Strip non-approved attributes from the tag
532 $newparams = Sanitizer
::fixTagAttributes( $params, $t );
535 $rest = str_replace( '>', '>', $rest );
536 $close = ( $brace == '/>' && !$slash ) ?
' /' : '';
537 $text .= "<$slash$t$newparams$close>$rest";
541 $text .= '<' . str_replace( '>', '>', $x);
543 # Close off any remaining tags
544 while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
546 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
549 # this might be possible using tidy itself
550 foreach ( $bits as $x ) {
551 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
553 @list
( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
554 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
555 if( is_callable( $processCallback ) ) {
556 call_user_func_array( $processCallback, array( &$params, $args ) );
558 $newparams = Sanitizer
::fixTagAttributes( $params, $t );
559 $rest = str_replace( '>', '>', $rest );
560 $text .= "<$slash$t$newparams$brace$rest";
562 $text .= '<' . str_replace( '>', '>', $x);
566 wfProfileOut( __METHOD__
);
571 * Remove '<!--', '-->', and everything between.
572 * To avoid leaving blank lines, when a comment is both preceded
573 * and followed by a newline (ignoring spaces), trim leading and
574 * trailing spaces and one of the newlines.
577 * @param $text String
580 static function removeHTMLcomments( $text ) {
581 wfProfileIn( __METHOD__
);
582 while (($start = strpos($text, '<!--')) !== false) {
583 $end = strpos($text, '-->', $start +
4);
584 if ($end === false) {
585 # Unterminated comment; bail out
591 # Trim space and newline if the comment is both
592 # preceded and followed by a newline
593 $spaceStart = max($start - 1, 0);
594 $spaceLen = $end - $spaceStart;
595 while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
599 while (substr($text, $spaceStart +
$spaceLen, 1) === ' ')
601 if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart +
$spaceLen, 1) === "\n") {
602 # Remove the comment, leading and trailing
603 # spaces, and leave only one newline.
604 $text = substr_replace($text, "\n", $spaceStart, $spaceLen +
1);
607 # Remove just the comment.
608 $text = substr_replace($text, '', $start, $end - $start);
611 wfProfileOut( __METHOD__
);
616 * Take an array of attribute names and values and fix some deprecated values
617 * for the given element type.
618 * This does not validate properties, so you should ensure that you call
619 * validateTagAttributes AFTER this to ensure that the resulting style rule
620 * this may add is safe.
622 * - Converts most presentational attributes like align into inline css
624 * @param $attribs Array
625 * @param $element String
628 static function fixDeprecatedAttributes( $attribs, $element ) {
629 global $wgHtml5, $wgCleanupPresentationalAttributes;
631 // presentational attributes were removed from html5, we can leave them
632 // in when html5 is turned off
633 if ( !$wgHtml5 ||
!$wgCleanupPresentationalAttributes ) {
637 $table = array( 'table' );
638 $cells = array( 'td', 'th' );
639 $colls = array( 'col', 'colgroup' );
640 $tblocks = array( 'tbody', 'tfoot', 'thead' );
641 $h = array( 'h1', 'h2', 'h3', 'h4', 'h5', 'h6' );
643 $presentationalAttribs = array(
644 'align' => array( 'text-align', array_merge( array( 'caption', 'hr', 'div', 'p', 'tr' ), $table, $cells, $colls, $tblocks, $h ) ),
645 'clear' => array( 'clear', array( 'br' ) ),
646 'height' => array( 'height', $cells ),
647 'nowrap' => array( 'white-space', $cells ),
648 'size' => array( 'height', array( 'hr' ) ),
649 'type' => array( 'list-style-type', array( 'li', 'ol', 'ul' ) ),
650 'valign' => array( 'vertical-align', array_merge( $cells, $colls, $tblocks ) ),
651 'width' => array( 'width', array_merge( array( 'hr', 'pre' ), $table, $cells, $colls ) ),
654 // Ensure that any upper case or mixed case attributes are converted to lowercase
655 foreach ( $attribs as $attribute => $value ) {
656 if ( $attribute !== strtolower( $attribute ) && array_key_exists( strtolower( $attribute ), $presentationalAttribs ) ) {
657 $attribs[strtolower( $attribute )] = $value;
658 unset( $attribs[$attribute] );
663 foreach ( $presentationalAttribs as $attribute => $info ) {
664 list( $property, $elements ) = $info;
666 // Skip if this attribute is not relevant to this element
667 if ( !in_array( $element, $elements ) ) {
671 // Skip if the attribute is not used
672 if ( !array_key_exists( $attribute, $attribs ) ) {
676 $value = $attribs[$attribute];
678 // For nowrap the value should be nowrap instead of whatever text is in the value
679 if ( $attribute === 'nowrap' ) {
683 // clear="all" is clear: both; in css
684 if ( $attribute === 'clear' && strtolower( $value ) === 'all' ) {
688 // Size based properties should have px applied to them if they have no unit
689 if ( in_array( $attribute, array( 'height', 'width', 'size' ) ) ) {
690 if ( preg_match( '/^[\d.]+$/', $value ) ) {
691 $value = "{$value}px";
695 if ( $attribute === 'align' && !in_array( $element, $cells ) ) {
696 if ( $value === 'center' ) {
697 $style .= ' margin-left: auto;';
698 $property = 'margin-right';
705 $style .= " $property: $value;";
707 unset( $attribs[$attribute] );
711 // Prepend our style rules so that they can be overridden by user css
712 if ( isset($attribs['style']) ) {
713 $style .= " " . $attribs['style'];
715 $attribs['style'] = trim($style);
722 * Take an array of attribute names and values and normalize or discard
723 * illegal values for the given element type.
725 * - Discards attributes not on a whitelist for the given element
726 * - Unsafe style attributes are discarded
727 * - Invalid id attributes are reencoded
729 * @param $attribs Array
730 * @param $element String
733 * @todo Check for legal values where the DTD limits things.
734 * @todo Check for unique id attribute :P
736 static function validateTagAttributes( $attribs, $element ) {
737 return Sanitizer
::validateAttributes( $attribs,
738 Sanitizer
::attributeWhitelist( $element ) );
742 * Take an array of attribute names and values and normalize or discard
743 * illegal values for the given whitelist.
745 * - Discards attributes not the given whitelist
746 * - Unsafe style attributes are discarded
747 * - Invalid id attributes are reencoded
749 * @param $attribs Array
750 * @param $whitelist Array: list of allowed attribute names
753 * @todo Check for legal values where the DTD limits things.
754 * @todo Check for unique id attribute :P
756 static function validateAttributes( $attribs, $whitelist ) {
757 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
759 $whitelist = array_flip( $whitelist );
760 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
763 foreach( $attribs as $attribute => $value ) {
764 #allow XML namespace declaration if RDFa is enabled
765 if ( $wgAllowRdfaAttributes && preg_match( self
::XMLNS_ATTRIBUTE_PATTERN
, $attribute ) ) {
766 if ( !preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
767 $out[$attribute] = $value;
773 # Allow any attribute beginning with "data-", if in HTML5 mode
774 if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
778 # Strip javascript "expression" from stylesheets.
779 # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
780 if( $attribute == 'style' ) {
781 $value = Sanitizer
::checkCss( $value );
784 if ( $attribute === 'id' ) {
785 $value = Sanitizer
::escapeId( $value, 'noninitial' );
788 //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
789 if ( $attribute === 'rel' ||
$attribute === 'rev' ||
790 $attribute === 'about' ||
$attribute === 'property' ||
$attribute === 'resource' ||
#RDFa
791 $attribute === 'datatype' ||
$attribute === 'typeof' ||
#RDFa
792 $attribute === 'itemid' ||
$attribute === 'itemprop' ||
$attribute === 'itemref' ||
#HTML5 microdata
793 $attribute === 'itemscope' ||
$attribute === 'itemtype' ) { #HTML5 microdata
795 //Paranoia. Allow "simple" values but suppress javascript
796 if ( preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
801 # NOTE: even though elements using href/src are not allowed directly, supply
802 # validation code that can be used by tag hook handlers, etc
803 if ( $attribute === 'href' ||
$attribute === 'src' ) {
804 if ( !preg_match( $hrefExp, $value ) ) {
805 continue; //drop any href or src attributes not using an allowed protocol.
806 //NOTE: this also drops all relative URLs
810 // If this attribute was previously set, override it.
811 // Output should only have one attribute of each name.
812 $out[$attribute] = $value;
815 if ( $wgAllowMicrodataAttributes ) {
816 # itemtype, itemid, itemref don't make sense without itemscope
817 if ( !array_key_exists( 'itemscope', $out ) ) {
818 unset( $out['itemtype'] );
819 unset( $out['itemid'] );
820 unset( $out['itemref'] );
822 # TODO: Strip itemprop if we aren't descendants of an itemscope.
828 * Merge two sets of HTML attributes. Conflicting items in the second set
829 * will override those in the first, except for 'class' attributes which
830 * will be combined (if they're both strings).
832 * @todo implement merging for other attributes such as style
837 static function mergeAttributes( $a, $b ) {
838 $out = array_merge( $a, $b );
839 if( isset( $a['class'] ) && isset( $b['class'] )
840 && is_string( $a['class'] ) && is_string( $b['class'] )
841 && $a['class'] !== $b['class'] ) {
842 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
843 -1, PREG_SPLIT_NO_EMPTY
);
844 $out['class'] = implode( ' ', array_unique( $classes ) );
850 * Pick apart some CSS and check it for forbidden or unsafe structures.
851 * Returns a sanitized string. This sanitized string will have
852 * character references and escape sequences decoded, and comments
853 * stripped. If the input is just too evil, only a comment complaining
854 * about evilness will be returned.
856 * Currently URL references, 'expression', 'tps' are forbidden.
858 * NOTE: Despite the fact that character references are decoded, the
859 * returned string may contain character references given certain
860 * clever input strings. These character references must
861 * be escaped before the return value is embedded in HTML.
863 * @param $value String
866 static function checkCss( $value ) {
867 // Decode character references like {
868 $value = Sanitizer
::decodeCharReferences( $value );
870 // Decode escape sequences and line continuation
871 // See the grammar in the CSS 2 spec, appendix D.
872 // This has to be done AFTER decoding character references.
873 // This means it isn't possible for this function to return
874 // unsanitized escape sequences. It is possible to manufacture
875 // input that contains character references that decode to
876 // escape sequences that decode to character references, but
877 // it's OK for the return value to contain character references
878 // because the caller is supposed to escape those anyway.
880 if ( !$decodeRegex ) {
881 $space = '[\\x20\\t\\r\\n\\f]';
882 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
884 $decodeRegex = "/ $backslash
886 ($nl) | # 1. Line continuation
887 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
888 (.) | # 3. backslash cancelling special meaning
889 () | # 4. backslash at end of string
892 $value = preg_replace_callback( $decodeRegex,
893 array( __CLASS__
, 'cssDecodeCallback' ), $value );
895 // Remove any comments; IE gets token splitting wrong
896 // This must be done AFTER decoding character references and
897 // escape sequences, because those steps can introduce comments
898 // This step cannot introduce character references or escape
899 // sequences, because it replaces comments with spaces rather
900 // than removing them completely.
901 $value = StringUtils
::delimiterReplace( '/*', '*/', ' ', $value );
903 // Remove anything after a comment-start token, to guard against
904 // incorrect client implementations.
905 $commentPos = strpos( $value, '/*' );
906 if ( $commentPos !== false ) {
907 $value = substr( $value, 0, $commentPos );
910 // Reject problematic keywords and control characters
911 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
912 return '/* invalid control char */';
913 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
914 return '/* insecure input */';
920 * @param $matches array
923 static function cssDecodeCallback( $matches ) {
924 if ( $matches[1] !== '' ) {
927 } elseif ( $matches[2] !== '' ) {
928 $char = codepointToUtf8( hexdec( $matches[2] ) );
929 } elseif ( $matches[3] !== '' ) {
934 if ( $char == "\n" ||
$char == '"' ||
$char == "'" ||
$char == '\\' ) {
935 // These characters need to be escaped in strings
936 // Clean up the escape sequence to avoid parsing errors by clients
937 return '\\' . dechex( ord( $char ) ) . ' ';
939 // Decode unnecessary escape
945 * Take a tag soup fragment listing an HTML element's attributes
946 * and normalize it to well-formed XML, discarding unwanted attributes.
947 * Output is safe for further wikitext processing, with escaping of
948 * values that could trigger problems.
950 * - Normalizes attribute names to lowercase
951 * - Discards attributes not on a whitelist for the given element
952 * - Turns broken or invalid entities into plaintext
953 * - Double-quotes all attribute values
954 * - Attributes without values are given the name as attribute
955 * - Double attributes are discarded
956 * - Unsafe style attributes are discarded
957 * - Prepends space if there are attributes.
959 * @param $text String
960 * @param $element String
963 static function fixTagAttributes( $text, $element ) {
964 if( trim( $text ) == '' ) {
968 $decoded = Sanitizer
::decodeTagAttributes( $text );
969 $decoded = Sanitizer
::fixDeprecatedAttributes( $decoded, $element );
970 $stripped = Sanitizer
::validateTagAttributes( $decoded, $element );
973 foreach( $stripped as $attribute => $value ) {
974 $encAttribute = htmlspecialchars( $attribute );
975 $encValue = Sanitizer
::safeEncodeAttribute( $value );
977 $attribs[] = "$encAttribute=\"$encValue\"";
979 return count( $attribs ) ?
' ' . implode( ' ', $attribs ) : '';
983 * Encode an attribute value for HTML output.
984 * @param $text String
985 * @return HTML-encoded text fragment
987 static function encodeAttribute( $text ) {
988 $encValue = htmlspecialchars( $text, ENT_QUOTES
);
990 // Whitespace is normalized during attribute decoding,
991 // so if we've been passed non-spaces we must encode them
992 // ahead of time or they won't be preserved.
993 $encValue = strtr( $encValue, array(
1003 * Encode an attribute value for HTML tags, with extra armoring
1004 * against further wiki processing.
1005 * @param $text String
1006 * @return HTML-encoded text fragment
1008 static function safeEncodeAttribute( $text ) {
1009 $encValue = Sanitizer
::encodeAttribute( $text );
1011 # Templates and links may be expanded in later parsing,
1012 # creating invalid or dangerous output. Suppress this.
1013 $encValue = strtr( $encValue, array(
1014 '<' => '<', // This should never happen,
1015 '>' => '>', // we've received invalid input
1016 '"' => '"', // which should have been escaped.
1019 "''" => '''',
1020 'ISBN' => 'ISBN',
1022 'PMID' => 'PMID',
1028 $encValue = preg_replace_callback(
1029 '/(' . wfUrlProtocols() . ')/',
1030 array( 'Sanitizer', 'armorLinksCallback' ),
1036 * Given a value, escape it so that it can be used in an id attribute and
1037 * return it. This will use HTML5 validation if $wgExperimentalHtmlIds is
1038 * true, allowing anything but ASCII whitespace. Otherwise it will use
1039 * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1040 * escaped with lots of dots.
1042 * To ensure we don't have to bother escaping anything, we also strip ', ",
1043 * & even if $wgExperimentalIds is true. TODO: Is this the best tactic?
1044 * We also strip # because it upsets IE, and % because it could be
1045 * ambiguous if it's part of something that looks like a percent escape
1046 * (which don't work reliably in fragments cross-browser).
1048 * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1051 * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1052 * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1053 * HTML5 definition of id attribute
1055 * @param $id String: id to escape
1056 * @param $options Mixed: string or array of strings (default is array()):
1057 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1058 * so don't pay attention if the first character isn't valid at the
1059 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
1061 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
1062 * if $wgExperimentalHtmlIds is used, so we can generate extra
1063 * anchors and links won't break.
1066 static function escapeId( $id, $options = array() ) {
1067 global $wgHtml5, $wgExperimentalHtmlIds;
1068 $options = (array)$options;
1070 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1071 $id = Sanitizer
::decodeCharReferences( $id );
1072 $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1073 $id = trim( $id, '_' );
1075 # Must have been all whitespace to start with.
1082 # HTML4-style escaping
1083 static $replace = array(
1088 $id = urlencode( Sanitizer
::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1089 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1091 if ( !preg_match( '/^[a-zA-Z]/', $id )
1092 && !in_array( 'noninitial', $options ) ) {
1093 // Initial character must be a letter!
1100 * Given a value, escape it so that it can be used as a CSS class and
1103 * @todo For extra validity, input should be validated UTF-8.
1105 * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1107 * @param $class String
1110 static function escapeClass( $class ) {
1111 // Convert ugly stuff to underscores and kill underscores in ugly places
1112 return rtrim(preg_replace(
1113 array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
1119 * Given HTML input, escape with htmlspecialchars but un-escape entites.
1120 * This allows (generally harmless) entities like   to survive.
1122 * @param $html String to escape
1123 * @return String: escaped input
1125 static function escapeHtmlAllowEntities( $html ) {
1126 $html = Sanitizer
::decodeCharReferences( $html );
1127 # It seems wise to escape ' as well as ", as a matter of course. Can't
1129 $html = htmlspecialchars( $html, ENT_QUOTES
);
1134 * Regex replace callback for armoring links against further processing.
1135 * @param $matches Array
1138 private static function armorLinksCallback( $matches ) {
1139 return str_replace( ':', ':', $matches[1] );
1143 * Return an associative array of attribute names and values from
1144 * a partial tag string. Attribute names are forces to lowercase,
1145 * character references are decoded to UTF-8 text.
1147 * @param $text String
1150 public static function decodeTagAttributes( $text ) {
1151 if( trim( $text ) == '' ) {
1157 if( !preg_match_all(
1158 self
::getAttribsRegex(),
1161 PREG_SET_ORDER
) ) {
1165 foreach( $pairs as $set ) {
1166 $attribute = strtolower( $set[1] );
1167 $value = Sanitizer
::getTagAttributeCallback( $set );
1169 // Normalize whitespace
1170 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1171 $value = trim( $value );
1173 // Decode character references
1174 $attribs[$attribute] = Sanitizer
::decodeCharReferences( $value );
1180 * Pick the appropriate attribute value from a match set from the
1181 * attribs regex matches.
1186 private static function getTagAttributeCallback( $set ) {
1187 if( isset( $set[6] ) ) {
1188 # Illegal #XXXXXX color with no quotes.
1190 } elseif( isset( $set[5] ) ) {
1193 } elseif( isset( $set[4] ) ) {
1196 } elseif( isset( $set[3] ) ) {
1199 } elseif( !isset( $set[2] ) ) {
1200 # In XHTML, attributes must have a value.
1201 # For 'reduced' form, return explicitly the attribute name here.
1204 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1209 * Normalize whitespace and character references in an XML source-
1210 * encoded text for an attribute value.
1212 * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1213 * but note that we're not returning the value, but are returning
1214 * XML source fragments that will be slapped into output.
1216 * @param $text String
1219 private static function normalizeAttributeValue( $text ) {
1220 return str_replace( '"', '"',
1221 self
::normalizeWhitespace(
1222 Sanitizer
::normalizeCharReferences( $text ) ) );
1226 * @param $text string
1229 private static function normalizeWhitespace( $text ) {
1230 return preg_replace(
1231 '/\r\n|[\x20\x0d\x0a\x09]/',
1237 * Normalizes whitespace in a section name, such as might be returned
1238 * by Parser::stripSectionName(), for use in the id's that are used for
1241 * @param $section String
1244 static function normalizeSectionNameWhitespace( $section ) {
1245 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1249 * Ensure that any entities and character references are legal
1250 * for XML and XHTML specifically. Any stray bits will be
1251 * &-escaped to result in a valid text fragment.
1253 * a. named char refs can only be < > & ", others are
1254 * numericized (this way we're well-formed even without a DTD)
1255 * b. any numeric char refs must be legal chars, not invalid or forbidden
1256 * c. use lower cased "&#x", not "&#X"
1257 * d. fix or reject non-valid attributes
1259 * @param $text String
1263 static function normalizeCharReferences( $text ) {
1264 return preg_replace_callback(
1265 self
::CHAR_REFS_REGEX
,
1266 array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1270 * @param $matches String
1273 static function normalizeCharReferencesCallback( $matches ) {
1275 if( $matches[1] != '' ) {
1276 $ret = Sanitizer
::normalizeEntity( $matches[1] );
1277 } elseif( $matches[2] != '' ) {
1278 $ret = Sanitizer
::decCharReference( $matches[2] );
1279 } elseif( $matches[3] != '' ) {
1280 $ret = Sanitizer
::hexCharReference( $matches[3] );
1282 if( is_null( $ret ) ) {
1283 return htmlspecialchars( $matches[0] );
1290 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1291 * return the equivalent numeric entity reference (except for the core <
1292 * > & "). If the entity is a MediaWiki-specific alias, returns
1293 * the HTML equivalent. Otherwise, returns HTML-escaped text of
1294 * pseudo-entity source (eg &foo;)
1296 * @param $name String
1299 static function normalizeEntity( $name ) {
1300 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1301 return '&' . self
::$htmlEntityAliases[$name] . ';';
1302 } elseif ( in_array( $name,
1303 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1305 } elseif ( isset( self
::$htmlEntities[$name] ) ) {
1306 return '&#' . self
::$htmlEntities[$name] . ';';
1308 return "&$name;";
1314 * @return null|string
1316 static function decCharReference( $codepoint ) {
1317 $point = intval( $codepoint );
1318 if( Sanitizer
::validateCodepoint( $point ) ) {
1319 return sprintf( '&#%d;', $point );
1327 * @return null|string
1329 static function hexCharReference( $codepoint ) {
1330 $point = hexdec( $codepoint );
1331 if( Sanitizer
::validateCodepoint( $point ) ) {
1332 return sprintf( '&#x%x;', $point );
1339 * Returns true if a given Unicode codepoint is a valid character in XML.
1340 * @param $codepoint Integer
1343 private static function validateCodepoint( $codepoint ) {
1344 return ($codepoint == 0x09)
1345 ||
($codepoint == 0x0a)
1346 ||
($codepoint == 0x0d)
1347 ||
($codepoint >= 0x20 && $codepoint <= 0xd7ff)
1348 ||
($codepoint >= 0xe000 && $codepoint <= 0xfffd)
1349 ||
($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1353 * Decode any character references, numeric or named entities,
1354 * in the text and return a UTF-8 string.
1356 * @param $text String
1359 public static function decodeCharReferences( $text ) {
1360 return preg_replace_callback(
1361 self
::CHAR_REFS_REGEX
,
1362 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1367 * Decode any character references, numeric or named entities,
1368 * in the next and normalize the resulting string. (bug 14952)
1370 * This is useful for page titles, not for text to be displayed,
1371 * MediaWiki allows HTML entities to escape normalization as a feature.
1373 * @param $text String (already normalized, containing entities)
1374 * @return String (still normalized, without entities)
1376 public static function decodeCharReferencesAndNormalize( $text ) {
1378 $text = preg_replace_callback(
1379 self
::CHAR_REFS_REGEX
,
1380 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1381 $text, /* limit */ -1, $count );
1384 return $wgContLang->normalize( $text );
1391 * @param $matches String
1394 static function decodeCharReferencesCallback( $matches ) {
1395 if( $matches[1] != '' ) {
1396 return Sanitizer
::decodeEntity( $matches[1] );
1397 } elseif( $matches[2] != '' ) {
1398 return Sanitizer
::decodeChar( intval( $matches[2] ) );
1399 } elseif( $matches[3] != '' ) {
1400 return Sanitizer
::decodeChar( hexdec( $matches[3] ) );
1402 # Last case should be an ampersand by itself
1407 * Return UTF-8 string for a codepoint if that is a valid
1408 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1409 * @param $codepoint Integer
1413 static function decodeChar( $codepoint ) {
1414 if( Sanitizer
::validateCodepoint( $codepoint ) ) {
1415 return codepointToUtf8( $codepoint );
1417 return UTF8_REPLACEMENT
;
1422 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1423 * return the UTF-8 encoding of that character. Otherwise, returns
1424 * pseudo-entity source (eg "&foo;")
1426 * @param $name String
1429 static function decodeEntity( $name ) {
1430 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1431 $name = self
::$htmlEntityAliases[$name];
1433 if( isset( self
::$htmlEntities[$name] ) ) {
1434 return codepointToUtf8( self
::$htmlEntities[$name] );
1441 * Fetch the whitelist of acceptable attributes for a given element name.
1443 * @param $element String
1446 static function attributeWhitelist( $element ) {
1448 if( !isset( $list ) ) {
1449 $list = Sanitizer
::setupAttributeWhitelist();
1451 return isset( $list[$element] )
1457 * Foreach array key (an allowed HTML element), return an array
1458 * of allowed attributes
1461 static function setupAttributeWhitelist() {
1462 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1464 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1466 if ( $wgAllowRdfaAttributes ) {
1467 #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1468 $common = array_merge( $common, array(
1469 'about', 'property', 'resource', 'datatype', 'typeof',
1473 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1474 # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1475 $common = array_merge( $common, array(
1476 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1480 $block = array_merge( $common, array( 'align' ) );
1481 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1482 $tablecell = array( 'abbr',
1488 'nowrap', # deprecated
1489 'width', # deprecated
1490 'height', # deprecated
1491 'bgcolor' # deprecated
1494 # Numbers refer to sections in HTML 4.01 standard describing the element.
1495 # See: http://www.w3.org/TR/html4/
1496 $whitelist = array (
1499 'center' => $common, # deprecated
1500 'span' => $block, # ??
1518 'strong' => $common,
1529 'blockquote' => array_merge( $common, array( 'cite' ) ),
1540 'br' => array( 'id', 'class', 'title', 'style', 'clear' ),
1543 'pre' => array_merge( $common, array( 'width' ) ),
1546 'ins' => array_merge( $common, array( 'cite', 'datetime' ) ),
1547 'del' => array_merge( $common, array( 'cite', 'datetime' ) ),
1550 'ul' => array_merge( $common, array( 'type' ) ),
1551 'ol' => array_merge( $common, array( 'type', 'start' ) ),
1552 'li' => array_merge( $common, array( 'type', 'value' ) ),
1560 'table' => array_merge( $common,
1561 array( 'summary', 'width', 'border', 'frame',
1562 'rules', 'cellspacing', 'cellpadding',
1567 'caption' => array_merge( $common, array( 'align' ) ),
1570 'thead' => array_merge( $common, $tablealign ),
1571 'tfoot' => array_merge( $common, $tablealign ),
1572 'tbody' => array_merge( $common, $tablealign ),
1575 'colgroup' => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1576 'col' => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1579 'tr' => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1582 'td' => array_merge( $common, $tablecell, $tablealign ),
1583 'th' => array_merge( $common, $tablecell, $tablealign ),
1585 # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1586 'a' => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1589 # Not usually allowed, but may be used for extension-style hooks
1590 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1592 'img' => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1600 'strike' => $common,
1605 'font' => array_merge( $common, array( 'size', 'color', 'face' ) ),
1609 'hr' => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1611 # XHTML Ruby annotation text module, simple ruby only.
1612 # http://www.w3c.org/TR/ruby/
1617 'rt' => $common, #array_merge( $common, array( 'rbspan' ) ),
1620 # MathML root element, where used for extensions
1621 # 'title' may not be 100% valid here; it's XHTML
1622 # http://www.w3.org/TR/REC-MathML/
1623 'math' => array( 'class', 'style', 'id', 'title' ),
1625 # HTML 5 section 4.6
1633 * Take a fragment of (potentially invalid) HTML and return
1634 * a version with any tags removed, encoded as plain text.
1636 * Warning: this return value must be further escaped for literal
1637 * inclusion in HTML output as of 1.10!
1639 * @param $text String: HTML fragment
1642 static function stripAllTags( $text ) {
1644 $text = StringUtils
::delimiterReplace( '<', '>', '', $text );
1646 # Normalize &entities and whitespace
1647 $text = self
::decodeCharReferences( $text );
1648 $text = self
::normalizeWhitespace( $text );
1654 * Hack up a private DOCTYPE with HTML's standard entity declarations.
1655 * PHP 4 seemed to know these if you gave it an HTML doctype, but
1658 * Use for passing XHTML fragments to PHP's XML parsing functions
1662 static function hackDocType() {
1663 $out = "<!DOCTYPE html [\n";
1664 foreach( self
::$htmlEntities as $entity => $codepoint ) {
1665 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1672 * @param $url string
1673 * @return mixed|string
1675 static function cleanUrl( $url ) {
1676 # Normalize any HTML entities in input. They will be
1677 # re-escaped by makeExternalLink().
1678 $url = Sanitizer
::decodeCharReferences( $url );
1680 # Escape any control characters introduced by the above step
1681 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1682 array( __CLASS__
, 'cleanUrlCallback' ), $url );
1684 # Validate hostname portion
1686 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1687 list( /* $whole */, $protocol, $host, $rest ) = $matches;
1689 // Characters that will be ignored in IDNs.
1690 // http://tools.ietf.org/html/3454#section-3.1
1691 // Strip them before further processing so blacklists and such work.
1693 \\s| # general whitespace
1694 \xc2\xad| # 00ad SOFT HYPHEN
1695 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1696 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1697 \xe2\x81\xa0| # 2060 WORD JOINER
1698 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1699 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
1700 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1701 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1702 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1703 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1704 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1705 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1708 $host = preg_replace( $strip, '', $host );
1710 // @todo FIXME: Validate hostnames here
1712 return $protocol . $host . $rest;
1719 * @param $matches array
1722 static function cleanUrlCallback( $matches ) {
1723 return urlencode( $matches[0] );
1727 * Does a string look like an e-mail address?
1729 * This validates an email address using an HTML5 specification found at:
1730 * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1731 * Which as of 2011-01-24 says:
1733 * A valid e-mail address is a string that matches the ABNF production
1734 * 1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1735 * in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1738 * This function is an implementation of the specification as requested in
1741 * Client-side forms will use the same standard validation rules via JS or
1742 * HTML 5 validation; additional restrictions can be enforced server-side
1743 * by extensions via the 'isValidEmailAddr' hook.
1745 * Note that this validation doesn't 100% match RFC 2822, but is believed
1746 * to be liberal enough for wide use. Some invalid addresses will still
1747 * pass validation here.
1751 * @param $addr String E-mail address
1754 public static function validateEmail( $addr ) {
1756 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1760 // Please note strings below are enclosed in brackets [], this make the
1761 // hyphen "-" a range indicator. Hence it is double backslashed below.
1763 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1764 $rfc1034_ldh_str = "a-z0-9\\-" ;
1766 $HTML5_email_regexp = "/
1768 [$rfc5322_atext\\.]+ # user part which is liberal :p
1770 [$rfc1034_ldh_str]+ # First domain part
1771 (\\.[$rfc1034_ldh_str]+)* # Following part prefixed with a dot
1773 /ix" ; // case Insensitive, eXtended
1775 return (bool) preg_match( $HTML5_email_regexp, $addr );