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', 'bdi', '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', 'mark' ) );
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 'li', 'dl', 'dt', 'dd', '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 # Still need to push this optionally-closed tag to
523 # the tag stack so that we can match end tags
524 # instead of marking them as bad.
525 array_push( $tagstack, $t );
526 } elseif ( isset( $tabletags[$t] )
527 && in_array( $t, $tagstack ) ) {
528 // New table tag but forgot to close the previous one
531 if ( $t == 'table' ) {
532 array_push( $tablestack, $tagstack );
535 array_push( $tagstack, $t );
538 # Replace any variables or template parameters with
540 if( is_callable( $processCallback ) ) {
541 call_user_func_array( $processCallback, array( &$params, $args ) );
544 if ( !Sanitizer
::validateTag( $params, $t ) ) {
548 # Strip non-approved attributes from the tag
549 $newparams = Sanitizer
::fixTagAttributes( $params, $t );
552 $rest = str_replace( '>', '>', $rest );
553 $close = ( $brace == '/>' && !$slash ) ?
' /' : '';
554 $text .= "<$slash$t$newparams$close>$rest";
558 $text .= '<' . str_replace( '>', '>', $x);
560 # Close off any remaining tags
561 while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
563 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
566 # this might be possible using tidy itself
567 foreach ( $bits as $x ) {
568 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
570 @list
( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
572 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
573 if( is_callable( $processCallback ) ) {
574 call_user_func_array( $processCallback, array( &$params, $args ) );
577 if ( !Sanitizer
::validateTag( $params, $t ) ) {
581 $newparams = Sanitizer
::fixTagAttributes( $params, $t );
583 $rest = str_replace( '>', '>', $rest );
584 $text .= "<$slash$t$newparams$brace$rest";
588 $text .= '<' . str_replace( '>', '>', $x);
591 wfProfileOut( __METHOD__
);
596 * Remove '<!--', '-->', and everything between.
597 * To avoid leaving blank lines, when a comment is both preceded
598 * and followed by a newline (ignoring spaces), trim leading and
599 * trailing spaces and one of the newlines.
602 * @param $text String
605 static function removeHTMLcomments( $text ) {
606 wfProfileIn( __METHOD__
);
607 while ( ($start = strpos( $text, '<!--' ) ) !== false ) {
608 $end = strpos( $text, '-->', $start +
4 );
609 if ( $end === false ) {
610 # Unterminated comment; bail out
616 # Trim space and newline if the comment is both
617 # preceded and followed by a newline
618 $spaceStart = max( $start - 1, 0 );
619 $spaceLen = $end - $spaceStart;
620 while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
624 while ( substr( $text, $spaceStart +
$spaceLen, 1 ) === ' ' )
626 if ( substr( $text, $spaceStart, 1 ) === "\n" and substr( $text, $spaceStart +
$spaceLen, 1 ) === "\n" ) {
627 # Remove the comment, leading and trailing
628 # spaces, and leave only one newline.
629 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen +
1 );
632 # Remove just the comment.
633 $text = substr_replace( $text, '', $start, $end - $start );
636 wfProfileOut( __METHOD__
);
641 * Takes attribute names and values for a tag and the tag name and
642 * validates that the tag is allowed to be present.
643 * This DOES NOT validate the attributes, nor does it validate the
644 * tags themselves. This method only handles the special circumstances
645 * where we may want to allow a tag within content but ONLY when it has
646 * specific attributes set.
651 static function validateTag( $params, $element ) {
652 $params = Sanitizer
::decodeTagAttributes( $params );
654 if ( $element == 'meta' ||
$element == 'link' ) {
655 if ( !isset( $params['itemprop'] ) ) {
656 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
659 if ( $element == 'meta' && !isset( $params['content'] ) ) {
660 // <meta> must have a content="" for the itemprop
663 if ( $element == 'link' && !isset( $params['href'] ) ) {
664 // <link> must have an associated href=""
673 * Take an array of attribute names and values and normalize or discard
674 * illegal values for the given element type.
676 * - Discards attributes not on a whitelist for the given element
677 * - Unsafe style attributes are discarded
678 * - Invalid id attributes are reencoded
680 * @param $attribs Array
681 * @param $element String
684 * @todo Check for legal values where the DTD limits things.
685 * @todo Check for unique id attribute :P
687 static function validateTagAttributes( $attribs, $element ) {
688 return Sanitizer
::validateAttributes( $attribs,
689 Sanitizer
::attributeWhitelist( $element ) );
693 * Take an array of attribute names and values and normalize or discard
694 * illegal values for the given whitelist.
696 * - Discards attributes not the given whitelist
697 * - Unsafe style attributes are discarded
698 * - Invalid id attributes are reencoded
700 * @param $attribs Array
701 * @param $whitelist Array: list of allowed attribute names
704 * @todo Check for legal values where the DTD limits things.
705 * @todo Check for unique id attribute :P
707 static function validateAttributes( $attribs, $whitelist ) {
708 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
710 $whitelist = array_flip( $whitelist );
711 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
714 foreach( $attribs as $attribute => $value ) {
715 #allow XML namespace declaration if RDFa is enabled
716 if ( $wgAllowRdfaAttributes && preg_match( self
::XMLNS_ATTRIBUTE_PATTERN
, $attribute ) ) {
717 if ( !preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
718 $out[$attribute] = $value;
724 # Allow any attribute beginning with "data-", if in HTML5 mode
725 if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
729 # Strip javascript "expression" from stylesheets.
730 # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
731 if( $attribute == 'style' ) {
732 $value = Sanitizer
::checkCss( $value );
735 if ( $attribute === 'id' ) {
736 $value = Sanitizer
::escapeId( $value, 'noninitial' );
740 # http://www.w3.org/TR/wai-aria/
741 # http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#wai-aria
742 # For now we only support role="presentation" until we work out what roles should be
743 # usable by content and we ensure that our code explicitly rejects patterns that
744 # violate HTML5's ARIA restrictions.
745 if ( $attribute === 'role' && $value !== 'presentation' ) {
749 //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
750 if ( $attribute === 'rel' ||
$attribute === 'rev' ||
751 $attribute === 'about' ||
$attribute === 'property' ||
$attribute === 'resource' ||
#RDFa
752 $attribute === 'datatype' ||
$attribute === 'typeof' ||
#RDFa
753 $attribute === 'itemid' ||
$attribute === 'itemprop' ||
$attribute === 'itemref' ||
#HTML5 microdata
754 $attribute === 'itemscope' ||
$attribute === 'itemtype' ) { #HTML5 microdata
756 //Paranoia. Allow "simple" values but suppress javascript
757 if ( preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
762 # NOTE: even though elements using href/src are not allowed directly, supply
763 # validation code that can be used by tag hook handlers, etc
764 if ( $attribute === 'href' ||
$attribute === 'src' ) {
765 if ( !preg_match( $hrefExp, $value ) ) {
766 continue; //drop any href or src attributes not using an allowed protocol.
767 //NOTE: this also drops all relative URLs
771 // If this attribute was previously set, override it.
772 // Output should only have one attribute of each name.
773 $out[$attribute] = $value;
776 if ( $wgAllowMicrodataAttributes ) {
777 # itemtype, itemid, itemref don't make sense without itemscope
778 if ( !array_key_exists( 'itemscope', $out ) ) {
779 unset( $out['itemtype'] );
780 unset( $out['itemid'] );
781 unset( $out['itemref'] );
783 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
789 * Merge two sets of HTML attributes. Conflicting items in the second set
790 * will override those in the first, except for 'class' attributes which
791 * will be combined (if they're both strings).
793 * @todo implement merging for other attributes such as style
798 static function mergeAttributes( $a, $b ) {
799 $out = array_merge( $a, $b );
800 if( isset( $a['class'] ) && isset( $b['class'] )
801 && is_string( $a['class'] ) && is_string( $b['class'] )
802 && $a['class'] !== $b['class'] ) {
803 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
804 -1, PREG_SPLIT_NO_EMPTY
);
805 $out['class'] = implode( ' ', array_unique( $classes ) );
811 * Pick apart some CSS and check it for forbidden or unsafe structures.
812 * Returns a sanitized string. This sanitized string will have
813 * character references and escape sequences decoded, and comments
814 * stripped. If the input is just too evil, only a comment complaining
815 * about evilness will be returned.
817 * Currently URL references, 'expression', 'tps' are forbidden.
819 * NOTE: Despite the fact that character references are decoded, the
820 * returned string may contain character references given certain
821 * clever input strings. These character references must
822 * be escaped before the return value is embedded in HTML.
824 * @param $value String
827 static function checkCss( $value ) {
828 // Decode character references like {
829 $value = Sanitizer
::decodeCharReferences( $value );
831 // Decode escape sequences and line continuation
832 // See the grammar in the CSS 2 spec, appendix D.
833 // This has to be done AFTER decoding character references.
834 // This means it isn't possible for this function to return
835 // unsanitized escape sequences. It is possible to manufacture
836 // input that contains character references that decode to
837 // escape sequences that decode to character references, but
838 // it's OK for the return value to contain character references
839 // because the caller is supposed to escape those anyway.
841 if ( !$decodeRegex ) {
842 $space = '[\\x20\\t\\r\\n\\f]';
843 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
845 $decodeRegex = "/ $backslash
847 ($nl) | # 1. Line continuation
848 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
849 (.) | # 3. backslash cancelling special meaning
850 () | # 4. backslash at end of string
853 $value = preg_replace_callback( $decodeRegex,
854 array( __CLASS__
, 'cssDecodeCallback' ), $value );
856 // Remove any comments; IE gets token splitting wrong
857 // This must be done AFTER decoding character references and
858 // escape sequences, because those steps can introduce comments
859 // This step cannot introduce character references or escape
860 // sequences, because it replaces comments with spaces rather
861 // than removing them completely.
862 $value = StringUtils
::delimiterReplace( '/*', '*/', ' ', $value );
864 // Remove anything after a comment-start token, to guard against
865 // incorrect client implementations.
866 $commentPos = strpos( $value, '/*' );
867 if ( $commentPos !== false ) {
868 $value = substr( $value, 0, $commentPos );
871 // Reject problematic keywords and control characters
872 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
873 return '/* invalid control char */';
874 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( | image\s*\( | image-set\s*\( !ix', $value ) ) {
875 return '/* insecure input */';
881 * @param $matches array
884 static function cssDecodeCallback( $matches ) {
885 if ( $matches[1] !== '' ) {
888 } elseif ( $matches[2] !== '' ) {
889 $char = codepointToUtf8( hexdec( $matches[2] ) );
890 } elseif ( $matches[3] !== '' ) {
895 if ( $char == "\n" ||
$char == '"' ||
$char == "'" ||
$char == '\\' ) {
896 // These characters need to be escaped in strings
897 // Clean up the escape sequence to avoid parsing errors by clients
898 return '\\' . dechex( ord( $char ) ) . ' ';
900 // Decode unnecessary escape
906 * Take a tag soup fragment listing an HTML element's attributes
907 * and normalize it to well-formed XML, discarding unwanted attributes.
908 * Output is safe for further wikitext processing, with escaping of
909 * values that could trigger problems.
911 * - Normalizes attribute names to lowercase
912 * - Discards attributes not on a whitelist for the given element
913 * - Turns broken or invalid entities into plaintext
914 * - Double-quotes all attribute values
915 * - Attributes without values are given the name as attribute
916 * - Double attributes are discarded
917 * - Unsafe style attributes are discarded
918 * - Prepends space if there are attributes.
920 * @param $text String
921 * @param $element String
924 static function fixTagAttributes( $text, $element ) {
925 if( trim( $text ) == '' ) {
929 $decoded = Sanitizer
::decodeTagAttributes( $text );
930 $stripped = Sanitizer
::validateTagAttributes( $decoded, $element );
933 foreach( $stripped as $attribute => $value ) {
934 $encAttribute = htmlspecialchars( $attribute );
935 $encValue = Sanitizer
::safeEncodeAttribute( $value );
937 $attribs[] = "$encAttribute=\"$encValue\"";
939 return count( $attribs ) ?
' ' . implode( ' ', $attribs ) : '';
943 * Encode an attribute value for HTML output.
944 * @param $text String
945 * @return HTML-encoded text fragment
947 static function encodeAttribute( $text ) {
948 $encValue = htmlspecialchars( $text, ENT_QUOTES
);
950 // Whitespace is normalized during attribute decoding,
951 // so if we've been passed non-spaces we must encode them
952 // ahead of time or they won't be preserved.
953 $encValue = strtr( $encValue, array(
963 * Encode an attribute value for HTML tags, with extra armoring
964 * against further wiki processing.
965 * @param $text String
966 * @return HTML-encoded text fragment
968 static function safeEncodeAttribute( $text ) {
969 $encValue = Sanitizer
::encodeAttribute( $text );
971 # Templates and links may be expanded in later parsing,
972 # creating invalid or dangerous output. Suppress this.
973 $encValue = strtr( $encValue, array(
974 '<' => '<', // This should never happen,
975 '>' => '>', // we've received invalid input
976 '"' => '"', // which should have been escaped.
979 "''" => '''',
980 'ISBN' => 'ISBN',
982 'PMID' => 'PMID',
988 $encValue = preg_replace_callback(
989 '/((?i)' . wfUrlProtocols() . ')/',
990 array( 'Sanitizer', 'armorLinksCallback' ),
996 * Given a value, escape it so that it can be used in an id attribute and
997 * return it. This will use HTML5 validation if $wgExperimentalHtmlIds is
998 * true, allowing anything but ASCII whitespace. Otherwise it will use
999 * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1000 * escaped with lots of dots.
1002 * To ensure we don't have to bother escaping anything, we also strip ', ",
1003 * & even if $wgExperimentalIds is true. TODO: Is this the best tactic?
1004 * We also strip # because it upsets IE, and % because it could be
1005 * ambiguous if it's part of something that looks like a percent escape
1006 * (which don't work reliably in fragments cross-browser).
1008 * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1011 * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1012 * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1013 * HTML5 definition of id attribute
1015 * @param $id String: id to escape
1016 * @param $options Mixed: string or array of strings (default is array()):
1017 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1018 * so don't pay attention if the first character isn't valid at the
1019 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
1021 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
1022 * if $wgExperimentalHtmlIds is used, so we can generate extra
1023 * anchors and links won't break.
1026 static function escapeId( $id, $options = array() ) {
1027 global $wgHtml5, $wgExperimentalHtmlIds;
1028 $options = (array)$options;
1030 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1031 $id = Sanitizer
::decodeCharReferences( $id );
1032 $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1033 $id = trim( $id, '_' );
1035 # Must have been all whitespace to start with.
1042 # HTML4-style escaping
1043 static $replace = array(
1048 $id = urlencode( Sanitizer
::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1049 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1051 if ( !preg_match( '/^[a-zA-Z]/', $id )
1052 && !in_array( 'noninitial', $options ) ) {
1053 // Initial character must be a letter!
1060 * Given a value, escape it so that it can be used as a CSS class and
1063 * @todo For extra validity, input should be validated UTF-8.
1065 * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1067 * @param $class String
1070 static function escapeClass( $class ) {
1071 // Convert ugly stuff to underscores and kill underscores in ugly places
1072 return rtrim( preg_replace(
1073 array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1079 * Given HTML input, escape with htmlspecialchars but un-escape entites.
1080 * This allows (generally harmless) entities like   to survive.
1082 * @param $html String to escape
1083 * @return String: escaped input
1085 static function escapeHtmlAllowEntities( $html ) {
1086 $html = Sanitizer
::decodeCharReferences( $html );
1087 # It seems wise to escape ' as well as ", as a matter of course. Can't
1089 $html = htmlspecialchars( $html, ENT_QUOTES
);
1094 * Regex replace callback for armoring links against further processing.
1095 * @param $matches Array
1098 private static function armorLinksCallback( $matches ) {
1099 return str_replace( ':', ':', $matches[1] );
1103 * Return an associative array of attribute names and values from
1104 * a partial tag string. Attribute names are forces to lowercase,
1105 * character references are decoded to UTF-8 text.
1107 * @param $text String
1110 public static function decodeTagAttributes( $text ) {
1111 if( trim( $text ) == '' ) {
1117 if( !preg_match_all(
1118 self
::getAttribsRegex(),
1121 PREG_SET_ORDER
) ) {
1125 foreach( $pairs as $set ) {
1126 $attribute = strtolower( $set[1] );
1127 $value = Sanitizer
::getTagAttributeCallback( $set );
1129 // Normalize whitespace
1130 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1131 $value = trim( $value );
1133 // Decode character references
1134 $attribs[$attribute] = Sanitizer
::decodeCharReferences( $value );
1140 * Pick the appropriate attribute value from a match set from the
1141 * attribs regex matches.
1144 * @throws MWException
1147 private static function getTagAttributeCallback( $set ) {
1148 if( isset( $set[6] ) ) {
1149 # Illegal #XXXXXX color with no quotes.
1151 } elseif( isset( $set[5] ) ) {
1154 } elseif( isset( $set[4] ) ) {
1157 } elseif( isset( $set[3] ) ) {
1160 } elseif( !isset( $set[2] ) ) {
1161 # In XHTML, attributes must have a value.
1162 # For 'reduced' form, return explicitly the attribute name here.
1165 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1170 * Normalize whitespace and character references in an XML source-
1171 * encoded text for an attribute value.
1173 * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1174 * but note that we're not returning the value, but are returning
1175 * XML source fragments that will be slapped into output.
1177 * @param $text String
1180 private static function normalizeAttributeValue( $text ) {
1181 return str_replace( '"', '"',
1182 self
::normalizeWhitespace(
1183 Sanitizer
::normalizeCharReferences( $text ) ) );
1187 * @param $text string
1190 private static function normalizeWhitespace( $text ) {
1191 return preg_replace(
1192 '/\r\n|[\x20\x0d\x0a\x09]/',
1198 * Normalizes whitespace in a section name, such as might be returned
1199 * by Parser::stripSectionName(), for use in the id's that are used for
1202 * @param $section String
1205 static function normalizeSectionNameWhitespace( $section ) {
1206 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1210 * Ensure that any entities and character references are legal
1211 * for XML and XHTML specifically. Any stray bits will be
1212 * &-escaped to result in a valid text fragment.
1214 * a. named char refs can only be < > & ", others are
1215 * numericized (this way we're well-formed even without a DTD)
1216 * b. any numeric char refs must be legal chars, not invalid or forbidden
1217 * c. use lower cased "&#x", not "&#X"
1218 * d. fix or reject non-valid attributes
1220 * @param $text String
1224 static function normalizeCharReferences( $text ) {
1225 return preg_replace_callback(
1226 self
::CHAR_REFS_REGEX
,
1227 array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1231 * @param $matches String
1234 static function normalizeCharReferencesCallback( $matches ) {
1236 if( $matches[1] != '' ) {
1237 $ret = Sanitizer
::normalizeEntity( $matches[1] );
1238 } elseif( $matches[2] != '' ) {
1239 $ret = Sanitizer
::decCharReference( $matches[2] );
1240 } elseif( $matches[3] != '' ) {
1241 $ret = Sanitizer
::hexCharReference( $matches[3] );
1243 if( is_null( $ret ) ) {
1244 return htmlspecialchars( $matches[0] );
1251 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1252 * return the equivalent numeric entity reference (except for the core <
1253 * > & "). If the entity is a MediaWiki-specific alias, returns
1254 * the HTML equivalent. Otherwise, returns HTML-escaped text of
1255 * pseudo-entity source (eg &foo;)
1257 * @param $name String
1260 static function normalizeEntity( $name ) {
1261 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1262 return '&' . self
::$htmlEntityAliases[$name] . ';';
1263 } elseif ( in_array( $name,
1264 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1266 } elseif ( isset( self
::$htmlEntities[$name] ) ) {
1267 return '&#' . self
::$htmlEntities[$name] . ';';
1269 return "&$name;";
1275 * @return null|string
1277 static function decCharReference( $codepoint ) {
1278 $point = intval( $codepoint );
1279 if( Sanitizer
::validateCodepoint( $point ) ) {
1280 return sprintf( '&#%d;', $point );
1288 * @return null|string
1290 static function hexCharReference( $codepoint ) {
1291 $point = hexdec( $codepoint );
1292 if( Sanitizer
::validateCodepoint( $point ) ) {
1293 return sprintf( '&#x%x;', $point );
1300 * Returns true if a given Unicode codepoint is a valid character in XML.
1301 * @param $codepoint Integer
1304 private static function validateCodepoint( $codepoint ) {
1305 return ($codepoint == 0x09)
1306 ||
($codepoint == 0x0a)
1307 ||
($codepoint == 0x0d)
1308 ||
($codepoint >= 0x20 && $codepoint <= 0xd7ff)
1309 ||
($codepoint >= 0xe000 && $codepoint <= 0xfffd)
1310 ||
($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1314 * Decode any character references, numeric or named entities,
1315 * in the text and return a UTF-8 string.
1317 * @param $text String
1320 public static function decodeCharReferences( $text ) {
1321 return preg_replace_callback(
1322 self
::CHAR_REFS_REGEX
,
1323 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1328 * Decode any character references, numeric or named entities,
1329 * in the next and normalize the resulting string. (bug 14952)
1331 * This is useful for page titles, not for text to be displayed,
1332 * MediaWiki allows HTML entities to escape normalization as a feature.
1334 * @param $text String (already normalized, containing entities)
1335 * @return String (still normalized, without entities)
1337 public static function decodeCharReferencesAndNormalize( $text ) {
1339 $text = preg_replace_callback(
1340 self
::CHAR_REFS_REGEX
,
1341 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1342 $text, /* limit */ -1, $count );
1345 return $wgContLang->normalize( $text );
1352 * @param $matches String
1355 static function decodeCharReferencesCallback( $matches ) {
1356 if( $matches[1] != '' ) {
1357 return Sanitizer
::decodeEntity( $matches[1] );
1358 } elseif( $matches[2] != '' ) {
1359 return Sanitizer
::decodeChar( intval( $matches[2] ) );
1360 } elseif( $matches[3] != '' ) {
1361 return Sanitizer
::decodeChar( hexdec( $matches[3] ) );
1363 # Last case should be an ampersand by itself
1368 * Return UTF-8 string for a codepoint if that is a valid
1369 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1370 * @param $codepoint Integer
1374 static function decodeChar( $codepoint ) {
1375 if( Sanitizer
::validateCodepoint( $codepoint ) ) {
1376 return codepointToUtf8( $codepoint );
1378 return UTF8_REPLACEMENT
;
1383 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1384 * return the UTF-8 encoding of that character. Otherwise, returns
1385 * pseudo-entity source (eg "&foo;")
1387 * @param $name String
1390 static function decodeEntity( $name ) {
1391 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1392 $name = self
::$htmlEntityAliases[$name];
1394 if( isset( self
::$htmlEntities[$name] ) ) {
1395 return codepointToUtf8( self
::$htmlEntities[$name] );
1402 * Fetch the whitelist of acceptable attributes for a given element name.
1404 * @param $element String
1407 static function attributeWhitelist( $element ) {
1408 $list = Sanitizer
::setupAttributeWhitelist();
1409 return isset( $list[$element] )
1415 * Foreach array key (an allowed HTML element), return an array
1416 * of allowed attributes
1419 static function setupAttributeWhitelist() {
1420 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1422 static $whitelist, $staticInitialised;
1423 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgHtml5', 'wgAllowMicrodataAttributes' ) );
1425 if ( isset( $whitelist ) && $staticInitialised == $globalContext ) {
1442 if ( $wgAllowRdfaAttributes ) {
1443 #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1444 $common = array_merge( $common, array(
1445 'about', 'property', 'resource', 'datatype', 'typeof',
1449 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1450 # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1451 $common = array_merge( $common, array(
1452 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1456 $block = array_merge( $common, array( 'align' ) );
1457 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1458 $tablecell = array( 'abbr',
1464 'nowrap', # deprecated
1465 'width', # deprecated
1466 'height', # deprecated
1467 'bgcolor' # deprecated
1470 # Numbers refer to sections in HTML 4.01 standard describing the element.
1471 # See: http://www.w3.org/TR/html4/
1475 'center' => $common, # deprecated
1476 'span' => $block, # ??
1494 'strong' => $common,
1505 'blockquote' => array_merge( $common, array( 'cite' ) ),
1516 'br' => array( 'id', 'class', 'title', 'style', 'clear' ),
1519 'pre' => array_merge( $common, array( 'width' ) ),
1522 'ins' => array_merge( $common, array( 'cite', 'datetime' ) ),
1523 'del' => array_merge( $common, array( 'cite', 'datetime' ) ),
1526 'ul' => array_merge( $common, array( 'type' ) ),
1527 'ol' => array_merge( $common, array( 'type', 'start' ) ),
1528 'li' => array_merge( $common, array( 'type', 'value' ) ),
1536 'table' => array_merge( $common,
1537 array( 'summary', 'width', 'border', 'frame',
1538 'rules', 'cellspacing', 'cellpadding',
1543 'caption' => array_merge( $common, array( 'align' ) ),
1546 'thead' => array_merge( $common, $tablealign ),
1547 'tfoot' => array_merge( $common, $tablealign ),
1548 'tbody' => array_merge( $common, $tablealign ),
1551 'colgroup' => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1552 'col' => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1555 'tr' => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1558 'td' => array_merge( $common, $tablecell, $tablealign ),
1559 'th' => array_merge( $common, $tablecell, $tablealign ),
1561 # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1562 'a' => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1565 # Not usually allowed, but may be used for extension-style hooks
1566 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1568 'img' => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1576 'strike' => $common,
1581 'font' => array_merge( $common, array( 'size', 'color', 'face' ) ),
1585 'hr' => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1587 # XHTML Ruby annotation text module, simple ruby only.
1588 # http://www.w3c.org/TR/ruby/
1593 'rt' => $common, #array_merge( $common, array( 'rbspan' ) ),
1596 # MathML root element, where used for extensions
1597 # 'title' may not be 100% valid here; it's XHTML
1598 # http://www.w3.org/TR/REC-MathML/
1599 'math' => array( 'class', 'style', 'id', 'title' ),
1601 # HTML 5 section 4.6
1607 # HTML5 elements, defined by:
1608 # http://www.whatwg.org/specs/web-apps/current-work/multipage/
1609 $whitelist +
= array(
1610 'data' => array_merge( $common, array( 'value' ) ),
1611 'time' => array_merge( $common, array( 'datetime' ) ),
1614 // meta and link are only permitted by removeHTMLtags when Microdata
1615 // is enabled so we don't bother adding a conditional to hide these
1616 // Also meta and link are only valid in WikiText as Microdata elements
1617 // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1618 // So we don't bother including $common attributes that have no purpose.
1619 'meta' => array( 'itemprop', 'content' ),
1620 'link' => array( 'itemprop', 'href' ),
1624 $staticInitialised = $globalContext;
1630 * Take a fragment of (potentially invalid) HTML and return
1631 * a version with any tags removed, encoded as plain text.
1633 * Warning: this return value must be further escaped for literal
1634 * inclusion in HTML output as of 1.10!
1636 * @param $text String: HTML fragment
1639 static function stripAllTags( $text ) {
1641 $text = StringUtils
::delimiterReplace( '<', '>', '', $text );
1643 # Normalize &entities and whitespace
1644 $text = self
::decodeCharReferences( $text );
1645 $text = self
::normalizeWhitespace( $text );
1651 * Hack up a private DOCTYPE with HTML's standard entity declarations.
1652 * PHP 4 seemed to know these if you gave it an HTML doctype, but
1655 * Use for passing XHTML fragments to PHP's XML parsing functions
1659 static function hackDocType() {
1660 $out = "<!DOCTYPE html [\n";
1661 foreach( self
::$htmlEntities as $entity => $codepoint ) {
1662 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1669 * @param $url string
1670 * @return mixed|string
1672 static function cleanUrl( $url ) {
1673 # Normalize any HTML entities in input. They will be
1674 # re-escaped by makeExternalLink().
1675 $url = Sanitizer
::decodeCharReferences( $url );
1677 # Escape any control characters introduced by the above step
1678 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1679 array( __CLASS__
, 'cleanUrlCallback' ), $url );
1681 # Validate hostname portion
1683 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1684 list( /* $whole */, $protocol, $host, $rest ) = $matches;
1686 // Characters that will be ignored in IDNs.
1687 // http://tools.ietf.org/html/3454#section-3.1
1688 // Strip them before further processing so blacklists and such work.
1690 \\s| # general whitespace
1691 \xc2\xad| # 00ad SOFT HYPHEN
1692 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1693 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1694 \xe2\x81\xa0| # 2060 WORD JOINER
1695 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1696 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
1697 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1698 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1699 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1700 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1701 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1702 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1705 $host = preg_replace( $strip, '', $host );
1707 // @todo FIXME: Validate hostnames here
1709 return $protocol . $host . $rest;
1716 * @param $matches array
1719 static function cleanUrlCallback( $matches ) {
1720 return urlencode( $matches[0] );
1724 * Does a string look like an e-mail address?
1726 * This validates an email address using an HTML5 specification found at:
1727 * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1728 * Which as of 2011-01-24 says:
1730 * A valid e-mail address is a string that matches the ABNF production
1731 * 1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1732 * in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1735 * This function is an implementation of the specification as requested in
1738 * Client-side forms will use the same standard validation rules via JS or
1739 * HTML 5 validation; additional restrictions can be enforced server-side
1740 * by extensions via the 'isValidEmailAddr' hook.
1742 * Note that this validation doesn't 100% match RFC 2822, but is believed
1743 * to be liberal enough for wide use. Some invalid addresses will still
1744 * pass validation here.
1748 * @param $addr String E-mail address
1751 public static function validateEmail( $addr ) {
1753 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1757 // Please note strings below are enclosed in brackets [], this make the
1758 // hyphen "-" a range indicator. Hence it is double backslashed below.
1760 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1761 $rfc1034_ldh_str = "a-z0-9\\-";
1763 $HTML5_email_regexp = "/
1765 [$rfc5322_atext\\.]+ # user part which is liberal :p
1767 [$rfc1034_ldh_str]+ # First domain part
1768 (\\.[$rfc1034_ldh_str]+)* # Following part prefixed with a dot
1770 /ix"; // case Insensitive, eXtended
1772 return (bool) preg_match( $HTML5_email_regexp, $addr );