3 * HTML 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 * HTML sanitizer for MediaWiki
33 * Regular expression to match various types of character references in
34 * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
36 const CHAR_REFS_REGEX
=
37 '/&([A-Za-z0-9\x80-\xff]+);
39 |&\#[xX]([0-9A-Fa-f]+);
43 * Blacklist for evil uris like javascript:
44 * WARNING: DO NOT use this in any place that actually requires blacklisting
45 * for security reasons. There are NUMEROUS[1] ways to bypass blacklisting, the
46 * only way to be secure from javascript: uri based xss vectors is to whitelist
47 * things that you know are safe and deny everything else.
48 * [1]: http://ha.ckers.org/xss.html
50 const EVIL_URI_PATTERN
= '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
51 const XMLNS_ATTRIBUTE_PATTERN
= "/^xmlns:[:A-Z_a-z-.0-9]+$/";
54 * List of all named character entities defined in HTML 4.01
55 * http://www.w3.org/TR/html4/sgml/entities.html
56 * As well as ' which is only defined starting in XHTML1.
58 private static $htmlEntities = array(
74 'apos' => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
315 * Character entity aliases accepted by MediaWiki
317 private static $htmlEntityAliases = array(
323 * Lazy-initialised attributes regex, see getAttribsRegex()
325 private static $attribsRegex;
328 * Regular expression to match HTML/XML attribute pairs within a tag.
329 * Allows some... latitude.
330 * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
332 static function getAttribsRegex() {
333 if ( self
::$attribsRegex === null ) {
334 $attribFirst = '[:A-Z_a-z0-9]';
335 $attrib = '[:A-Z_a-z-.0-9]';
336 $space = '[\x09\x0a\x0d\x20]';
337 self
::$attribsRegex =
338 "/(?:^|$space)({$attribFirst}{$attrib}*)
341 # The attribute value: quoted or alone
344 | ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
345 | (\#[0-9a-fA-F]+) # Technically wrong, but lots of
346 # colors are specified like this.
347 # We'll be normalizing it.
351 return self
::$attribsRegex;
355 * Cleans up HTML, removes dangerous tags and attributes, and
356 * removes HTML comments
358 * @param string $text
359 * @param callable $processCallback Callback to do any variable or parameter
360 * replacements in HTML attribute values
361 * @param array $args Arguments for the processing callback
362 * @param array $extratags For any extra tags to include
363 * @param array $removetags For any tags (default or extra) to exclude
366 static function removeHTMLtags( $text, $processCallback = null,
367 $args = array(), $extratags = array(), $removetags = array()
369 global $wgUseTidy, $wgAllowMicrodataAttributes, $wgAllowImageTag;
371 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
372 $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
374 wfProfileIn( __METHOD__
);
376 // Base our staticInitialised variable off of the global config state so that if the globals
377 // are changed (like in the screwed up test system) we will re-initialise the settings.
378 $globalContext = implode( '-', compact( 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
379 if ( !$staticInitialised ||
$staticInitialised != $globalContext ) {
381 $htmlpairsStatic = array( # Tags that must be closed
382 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
383 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
384 'strike', 'strong', 'tt', 'var', 'div', 'center',
385 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
386 'ruby', 'rt', 'rb', 'rp', 'p', 'span', 'abbr', 'dfn',
387 'kbd', 'samp', 'data', 'time', 'mark'
390 'br', 'wbr', 'hr', 'li', 'dt', 'dd'
392 $htmlsingleonly = array( # Elements that cannot have close tags
395 if ( $wgAllowMicrodataAttributes ) {
396 $htmlsingle[] = $htmlsingleonly[] = 'meta';
397 $htmlsingle[] = $htmlsingleonly[] = 'link';
399 $htmlnest = array( # Tags that can be nested--??
400 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
401 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
402 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
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 $bits
449 if ( preg_match( '!^(/?)([^\\s/>]+)([^>]*?)(/{0,1}>)([^<]*)$!', $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 unclosable 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' ) {
564 $tagstack = array_pop( $tablestack );
568 # this might be possible using tidy itself
569 foreach ( $bits as $x ) {
570 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
572 @list
( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
574 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
575 if ( is_callable( $processCallback ) ) {
576 call_user_func_array( $processCallback, array( &$params, $args ) );
579 if ( !Sanitizer
::validateTag( $params, $t ) ) {
583 $newparams = Sanitizer
::fixTagAttributes( $params, $t );
585 $rest = str_replace( '>', '>', $rest );
586 $text .= "<$slash$t$newparams$brace$rest";
590 $text .= '<' . str_replace( '>', '>', $x );
593 wfProfileOut( __METHOD__
);
598 * Remove '<!--', '-->', and everything between.
599 * To avoid leaving blank lines, when a comment is both preceded
600 * and followed by a newline (ignoring spaces), trim leading and
601 * trailing spaces and one of the newlines.
604 * @param string $text
607 static function removeHTMLcomments( $text ) {
608 wfProfileIn( __METHOD__
);
609 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
610 $end = strpos( $text, '-->', $start +
4 );
611 if ( $end === false ) {
612 # Unterminated comment; bail out
618 # Trim space and newline if the comment is both
619 # preceded and followed by a newline
620 $spaceStart = max( $start - 1, 0 );
621 $spaceLen = $end - $spaceStart;
622 while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
626 while ( substr( $text, $spaceStart +
$spaceLen, 1 ) === ' ' ) {
629 if ( substr( $text, $spaceStart, 1 ) === "\n"
630 && substr( $text, $spaceStart +
$spaceLen, 1 ) === "\n" ) {
631 # Remove the comment, leading and trailing
632 # spaces, and leave only one newline.
633 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen +
1 );
636 # Remove just the comment.
637 $text = substr_replace( $text, '', $start, $end - $start );
640 wfProfileOut( __METHOD__
);
645 * Takes attribute names and values for a tag and the tag name and
646 * validates that the tag is allowed to be present.
647 * This DOES NOT validate the attributes, nor does it validate the
648 * tags themselves. This method only handles the special circumstances
649 * where we may want to allow a tag within content but ONLY when it has
650 * specific attributes set.
652 * @param string $params
653 * @param string $element
656 static function validateTag( $params, $element ) {
657 $params = Sanitizer
::decodeTagAttributes( $params );
659 if ( $element == 'meta' ||
$element == 'link' ) {
660 if ( !isset( $params['itemprop'] ) ) {
661 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
664 if ( $element == 'meta' && !isset( $params['content'] ) ) {
665 // <meta> must have a content="" for the itemprop
668 if ( $element == 'link' && !isset( $params['href'] ) ) {
669 // <link> must have an associated href=""
678 * Take an array of attribute names and values and normalize or discard
679 * illegal values for the given element type.
681 * - Discards attributes not on a whitelist for the given element
682 * - Unsafe style attributes are discarded
683 * - Invalid id attributes are re-encoded
685 * @param array $attribs
686 * @param string $element
689 * @todo Check for legal values where the DTD limits things.
690 * @todo Check for unique id attribute :P
692 static function validateTagAttributes( $attribs, $element ) {
693 return Sanitizer
::validateAttributes( $attribs,
694 Sanitizer
::attributeWhitelist( $element ) );
698 * Take an array of attribute names and values and normalize or discard
699 * illegal values for the given whitelist.
701 * - Discards attributes not the given whitelist
702 * - Unsafe style attributes are discarded
703 * - Invalid id attributes are re-encoded
705 * @param array $attribs
706 * @param array $whitelist list of allowed attribute names
709 * @todo Check for legal values where the DTD limits things.
710 * @todo Check for unique id attribute :P
712 static function validateAttributes( $attribs, $whitelist ) {
713 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
715 $whitelist = array_flip( $whitelist );
716 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
719 foreach ( $attribs as $attribute => $value ) {
720 #allow XML namespace declaration if RDFa is enabled
721 if ( $wgAllowRdfaAttributes && preg_match( self
::XMLNS_ATTRIBUTE_PATTERN
, $attribute ) ) {
722 if ( !preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
723 $out[$attribute] = $value;
729 # Allow any attribute beginning with "data-"
730 if ( !preg_match( '/^data-/i', $attribute ) && !isset( $whitelist[$attribute] ) ) {
734 # Strip javascript "expression" from stylesheets.
735 # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
736 if ( $attribute == 'style' ) {
737 $value = Sanitizer
::checkCss( $value );
740 if ( $attribute === 'id' ) {
741 $value = Sanitizer
::escapeId( $value, 'noninitial' );
745 # http://www.w3.org/TR/wai-aria/
746 # http://www.whatwg.org/html/elements.html#wai-aria
747 # For now we only support role="presentation" until we work out what roles should be
748 # usable by content and we ensure that our code explicitly rejects patterns that
749 # violate HTML5's ARIA restrictions.
750 if ( $attribute === 'role' && $value !== 'presentation' ) {
754 // RDFa and microdata properties allow URLs, URIs and/or CURIs.
755 // Check them for sanity.
756 if ( $attribute === 'rel' ||
$attribute === 'rev'
758 ||
$attribute === 'about' ||
$attribute === 'property'
759 ||
$attribute === 'resource' ||
$attribute === 'datatype'
760 ||
$attribute === 'typeof'
762 ||
$attribute === 'itemid' ||
$attribute === 'itemprop'
763 ||
$attribute === 'itemref' ||
$attribute === 'itemscope'
764 ||
$attribute === 'itemtype'
766 //Paranoia. Allow "simple" values but suppress javascript
767 if ( preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
772 # NOTE: even though elements using href/src are not allowed directly, supply
773 # validation code that can be used by tag hook handlers, etc
774 if ( $attribute === 'href' ||
$attribute === 'src' ) {
775 if ( !preg_match( $hrefExp, $value ) ) {
776 continue; //drop any href or src attributes not using an allowed protocol.
777 // NOTE: this also drops all relative URLs
781 // If this attribute was previously set, override it.
782 // Output should only have one attribute of each name.
783 $out[$attribute] = $value;
786 if ( $wgAllowMicrodataAttributes ) {
787 # itemtype, itemid, itemref don't make sense without itemscope
788 if ( !array_key_exists( 'itemscope', $out ) ) {
789 unset( $out['itemtype'] );
790 unset( $out['itemid'] );
791 unset( $out['itemref'] );
793 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
799 * Merge two sets of HTML attributes. Conflicting items in the second set
800 * will override those in the first, except for 'class' attributes which
801 * will be combined (if they're both strings).
803 * @todo implement merging for other attributes such as style
808 static function mergeAttributes( $a, $b ) {
809 $out = array_merge( $a, $b );
810 if ( isset( $a['class'] ) && isset( $b['class'] )
811 && is_string( $a['class'] ) && is_string( $b['class'] )
812 && $a['class'] !== $b['class']
814 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
815 -1, PREG_SPLIT_NO_EMPTY
);
816 $out['class'] = implode( ' ', array_unique( $classes ) );
822 * Pick apart some CSS and check it for forbidden or unsafe structures.
823 * Returns a sanitized string. This sanitized string will have
824 * character references and escape sequences decoded and comments
825 * stripped (unless it is itself one valid comment, in which case the value
826 * will be passed through). If the input is just too evil, only a comment
827 * complaining about evilness will be returned.
829 * Currently URL references, 'expression', 'tps' are forbidden.
831 * NOTE: Despite the fact that character references are decoded, the
832 * returned string may contain character references given certain
833 * clever input strings. These character references must
834 * be escaped before the return value is embedded in HTML.
836 * @param string $value
839 static function checkCss( $value ) {
840 // Decode character references like {
841 $value = Sanitizer
::decodeCharReferences( $value );
843 // Decode escape sequences and line continuation
844 // See the grammar in the CSS 2 spec, appendix D.
845 // This has to be done AFTER decoding character references.
846 // This means it isn't possible for this function to return
847 // unsanitized escape sequences. It is possible to manufacture
848 // input that contains character references that decode to
849 // escape sequences that decode to character references, but
850 // it's OK for the return value to contain character references
851 // because the caller is supposed to escape those anyway.
853 if ( !$decodeRegex ) {
854 $space = '[\\x20\\t\\r\\n\\f]';
855 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
857 $decodeRegex = "/ $backslash
859 ($nl) | # 1. Line continuation
860 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
861 (.) | # 3. backslash cancelling special meaning
862 () | # 4. backslash at end of string
865 $value = preg_replace_callback( $decodeRegex,
866 array( __CLASS__
, 'cssDecodeCallback' ), $value );
868 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
869 $value = preg_replace_callback(
870 '/[!-z]/u', // U+FF01 to U+FF5A
871 function ( $matches ) {
872 $cp = utf8ToCodepoint( $matches[0] );
873 if ( $cp === false ) {
876 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
881 // Convert more characters IE6 might treat as ascii
882 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
883 $value = str_replace(
884 array( 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ),
885 array( 'r', 'n', 'n', 'l', 'i', '(', '(' ),
889 // Let the value through if it's nothing but a single comment, to
890 // allow other functions which may reject it to pass some error
892 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
893 // Remove any comments; IE gets token splitting wrong
894 // This must be done AFTER decoding character references and
895 // escape sequences, because those steps can introduce comments
896 // This step cannot introduce character references or escape
897 // sequences, because it replaces comments with spaces rather
898 // than removing them completely.
899 $value = StringUtils
::delimiterReplace( '/*', '*/', ' ', $value );
901 // Remove anything after a comment-start token, to guard against
902 // incorrect client implementations.
903 $commentPos = strpos( $value, '/*' );
904 if ( $commentPos !== false ) {
905 $value = substr( $value, 0, $commentPos );
909 // S followed by repeat, iteration, or prolonged sound marks,
910 // which IE will treat as "ss"
911 $value = preg_replace(
913 \xE3\x80\xB1 | # U+3031
914 \xE3\x82\x9D | # U+309D
915 \xE3\x83\xBC | # U+30FC
916 \xE3\x83\xBD | # U+30FD
917 \xEF\xB9\xBC | # U+FE7C
918 \xEF\xB9\xBD | # U+FE7D
919 \xEF\xBD\xB0 # U+FF70
925 // Reject problematic keywords and control characters
926 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ) {
927 return '/* invalid control char */';
928 } elseif ( preg_match(
929 '! expression | filter\s*: | accelerator\s*: | url\s*\( | image\s*\( | image-set\s*\( !ix',
932 return '/* insecure input */';
938 * @param array $matches
941 static function cssDecodeCallback( $matches ) {
942 if ( $matches[1] !== '' ) {
945 } elseif ( $matches[2] !== '' ) {
946 $char = codepointToUtf8( hexdec( $matches[2] ) );
947 } elseif ( $matches[3] !== '' ) {
952 if ( $char == "\n" ||
$char == '"' ||
$char == "'" ||
$char == '\\' ) {
953 // These characters need to be escaped in strings
954 // Clean up the escape sequence to avoid parsing errors by clients
955 return '\\' . dechex( ord( $char ) ) . ' ';
957 // Decode unnecessary escape
963 * Take a tag soup fragment listing an HTML element's attributes
964 * and normalize it to well-formed XML, discarding unwanted attributes.
965 * Output is safe for further wikitext processing, with escaping of
966 * values that could trigger problems.
968 * - Normalizes attribute names to lowercase
969 * - Discards attributes not on a whitelist for the given element
970 * - Turns broken or invalid entities into plaintext
971 * - Double-quotes all attribute values
972 * - Attributes without values are given the name as attribute
973 * - Double attributes are discarded
974 * - Unsafe style attributes are discarded
975 * - Prepends space if there are attributes.
977 * @param string $text
978 * @param string $element
981 static function fixTagAttributes( $text, $element ) {
982 if ( trim( $text ) == '' ) {
986 $decoded = Sanitizer
::decodeTagAttributes( $text );
987 $stripped = Sanitizer
::validateTagAttributes( $decoded, $element );
989 return Sanitizer
::safeEncodeTagAttributes( $stripped );
993 * Encode an attribute value for HTML output.
994 * @param string $text
995 * @return string HTML-encoded text fragment
997 static function encodeAttribute( $text ) {
998 $encValue = htmlspecialchars( $text, ENT_QUOTES
);
1000 // Whitespace is normalized during attribute decoding,
1001 // so if we've been passed non-spaces we must encode them
1002 // ahead of time or they won't be preserved.
1003 $encValue = strtr( $encValue, array(
1013 * Encode an attribute value for HTML tags, with extra armoring
1014 * against further wiki processing.
1015 * @param string $text
1016 * @return string HTML-encoded text fragment
1018 static function safeEncodeAttribute( $text ) {
1019 $encValue = Sanitizer
::encodeAttribute( $text );
1021 # Templates and links may be expanded in later parsing,
1022 # creating invalid or dangerous output. Suppress this.
1023 $encValue = strtr( $encValue, array(
1024 '<' => '<', // This should never happen,
1025 '>' => '>', // we've received invalid input
1026 '"' => '"', // which should have been escaped.
1029 "''" => '''',
1030 'ISBN' => 'ISBN',
1032 'PMID' => 'PMID',
1038 $encValue = preg_replace_callback(
1039 '/((?i)' . wfUrlProtocols() . ')/',
1040 array( 'Sanitizer', 'armorLinksCallback' ),
1046 * Given a value, escape it so that it can be used in an id attribute and
1047 * return it. This will use HTML5 validation if $wgExperimentalHtmlIds is
1048 * true, allowing anything but ASCII whitespace. Otherwise it will use
1049 * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1050 * escaped with lots of dots.
1052 * To ensure we don't have to bother escaping anything, we also strip ', ",
1053 * & even if $wgExperimentalIds is true. TODO: Is this the best tactic?
1054 * We also strip # because it upsets IE, and % because it could be
1055 * ambiguous if it's part of something that looks like a percent escape
1056 * (which don't work reliably in fragments cross-browser).
1058 * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1059 * in the id and name attributes
1060 * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1062 * @see http://www.whatwg.org/html/elements.html#the-id-attribute
1063 * HTML5 definition of id attribute
1065 * @param string $id id to escape
1066 * @param $options Mixed: string or array of strings (default is array()):
1067 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1068 * so don't pay attention if the first character isn't valid at the
1069 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
1071 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
1072 * if $wgExperimentalHtmlIds is used, so we can generate extra
1073 * anchors and links won't break.
1076 static function escapeId( $id, $options = array() ) {
1077 global $wgExperimentalHtmlIds;
1078 $options = (array)$options;
1080 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1081 $id = Sanitizer
::decodeCharReferences( $id );
1082 $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1083 $id = trim( $id, '_' );
1085 # Must have been all whitespace to start with.
1092 # HTML4-style escaping
1093 static $replace = array(
1098 $id = urlencode( Sanitizer
::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1099 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1101 if ( !preg_match( '/^[a-zA-Z]/', $id )
1102 && !in_array( 'noninitial', $options ) ) {
1103 // Initial character must be a letter!
1110 * Given a value, escape it so that it can be used as a CSS class and
1113 * @todo For extra validity, input should be validated UTF-8.
1115 * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1117 * @param string $class
1120 static function escapeClass( $class ) {
1121 // Convert ugly stuff to underscores and kill underscores in ugly places
1122 return rtrim( preg_replace(
1123 array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1129 * Given HTML input, escape with htmlspecialchars but un-escape entities.
1130 * This allows (generally harmless) entities like   to survive.
1132 * @param string $html HTML to escape
1133 * @return string: escaped input
1135 static function escapeHtmlAllowEntities( $html ) {
1136 $html = Sanitizer
::decodeCharReferences( $html );
1137 # It seems wise to escape ' as well as ", as a matter of course. Can't
1139 $html = htmlspecialchars( $html, ENT_QUOTES
);
1144 * Regex replace callback for armoring links against further processing.
1145 * @param array $matches
1148 private static function armorLinksCallback( $matches ) {
1149 return str_replace( ':', ':', $matches[1] );
1153 * Return an associative array of attribute names and values from
1154 * a partial tag string. Attribute names are forces to lowercase,
1155 * character references are decoded to UTF-8 text.
1157 * @param string $text
1160 public static function decodeTagAttributes( $text ) {
1161 if ( trim( $text ) == '' ) {
1167 if ( !preg_match_all(
1168 self
::getAttribsRegex(),
1171 PREG_SET_ORDER
) ) {
1175 foreach ( $pairs as $set ) {
1176 $attribute = strtolower( $set[1] );
1177 $value = Sanitizer
::getTagAttributeCallback( $set );
1179 // Normalize whitespace
1180 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1181 $value = trim( $value );
1183 // Decode character references
1184 $attribs[$attribute] = Sanitizer
::decodeCharReferences( $value );
1190 * Build a partial tag string from an associative array of attribute
1191 * names and values as returned by decodeTagAttributes.
1193 * @param array $assoc_array
1196 public static function safeEncodeTagAttributes( $assoc_array ) {
1198 foreach ( $assoc_array as $attribute => $value ) {
1199 $encAttribute = htmlspecialchars( $attribute );
1200 $encValue = Sanitizer
::safeEncodeAttribute( $value );
1202 $attribs[] = "$encAttribute=\"$encValue\"";
1204 return count( $attribs ) ?
' ' . implode( ' ', $attribs ) : '';
1208 * Pick the appropriate attribute value from a match set from the
1209 * attribs regex matches.
1212 * @throws MWException when tag conditions are not met.
1215 private static function getTagAttributeCallback( $set ) {
1216 if ( isset( $set[6] ) ) {
1217 # Illegal #XXXXXX color with no quotes.
1219 } elseif ( isset( $set[5] ) ) {
1222 } elseif ( isset( $set[4] ) ) {
1225 } elseif ( isset( $set[3] ) ) {
1228 } elseif ( !isset( $set[2] ) ) {
1229 # In XHTML, attributes must have a value.
1230 # For 'reduced' form, return explicitly the attribute name here.
1233 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1238 * Normalize whitespace and character references in an XML source-
1239 * encoded text for an attribute value.
1241 * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1242 * but note that we're not returning the value, but are returning
1243 * XML source fragments that will be slapped into output.
1245 * @param string $text
1247 * @todo Remove, unused?
1249 private static function normalizeAttributeValue( $text ) {
1250 return str_replace( '"', '"',
1251 self
::normalizeWhitespace(
1252 Sanitizer
::normalizeCharReferences( $text ) ) );
1256 * @param string $text
1259 private static function normalizeWhitespace( $text ) {
1260 return preg_replace(
1261 '/\r\n|[\x20\x0d\x0a\x09]/',
1267 * Normalizes whitespace in a section name, such as might be returned
1268 * by Parser::stripSectionName(), for use in the id's that are used for
1271 * @param string $section
1274 static function normalizeSectionNameWhitespace( $section ) {
1275 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1279 * Ensure that any entities and character references are legal
1280 * for XML and XHTML specifically. Any stray bits will be
1281 * &-escaped to result in a valid text fragment.
1283 * a. named char refs can only be < > & ", others are
1284 * numericized (this way we're well-formed even without a DTD)
1285 * b. any numeric char refs must be legal chars, not invalid or forbidden
1286 * c. use lower cased "&#x", not "&#X"
1287 * d. fix or reject non-valid attributes
1289 * @param string $text
1293 static function normalizeCharReferences( $text ) {
1294 return preg_replace_callback(
1295 self
::CHAR_REFS_REGEX
,
1296 array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1300 * @param string $matches
1303 static function normalizeCharReferencesCallback( $matches ) {
1305 if ( $matches[1] != '' ) {
1306 $ret = Sanitizer
::normalizeEntity( $matches[1] );
1307 } elseif ( $matches[2] != '' ) {
1308 $ret = Sanitizer
::decCharReference( $matches[2] );
1309 } elseif ( $matches[3] != '' ) {
1310 $ret = Sanitizer
::hexCharReference( $matches[3] );
1312 if ( is_null( $ret ) ) {
1313 return htmlspecialchars( $matches[0] );
1320 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1321 * return the equivalent numeric entity reference (except for the core <
1322 * > & "). If the entity is a MediaWiki-specific alias, returns
1323 * the HTML equivalent. Otherwise, returns HTML-escaped text of
1324 * pseudo-entity source (eg &foo;)
1326 * @param string $name
1329 static function normalizeEntity( $name ) {
1330 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1331 return '&' . self
::$htmlEntityAliases[$name] . ';';
1332 } elseif ( in_array( $name,
1333 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1335 } elseif ( isset( self
::$htmlEntities[$name] ) ) {
1336 return '&#' . self
::$htmlEntities[$name] . ';';
1338 return "&$name;";
1343 * @param int $codepoint
1344 * @return null|string
1346 static function decCharReference( $codepoint ) {
1347 $point = intval( $codepoint );
1348 if ( Sanitizer
::validateCodepoint( $point ) ) {
1349 return sprintf( '&#%d;', $point );
1356 * @param int $codepoint
1357 * @return null|string
1359 static function hexCharReference( $codepoint ) {
1360 $point = hexdec( $codepoint );
1361 if ( Sanitizer
::validateCodepoint( $point ) ) {
1362 return sprintf( '&#x%x;', $point );
1369 * Returns true if a given Unicode codepoint is a valid character in XML.
1370 * @param int $codepoint
1373 private static function validateCodepoint( $codepoint ) {
1374 return $codepoint == 0x09
1375 ||
$codepoint == 0x0a
1376 ||
$codepoint == 0x0d
1377 ||
( $codepoint >= 0x20 && $codepoint <= 0xd7ff )
1378 ||
( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1379 ||
( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1383 * Decode any character references, numeric or named entities,
1384 * in the text and return a UTF-8 string.
1386 * @param string $text
1389 public static function decodeCharReferences( $text ) {
1390 return preg_replace_callback(
1391 self
::CHAR_REFS_REGEX
,
1392 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1397 * Decode any character references, numeric or named entities,
1398 * in the next and normalize the resulting string. (bug 14952)
1400 * This is useful for page titles, not for text to be displayed,
1401 * MediaWiki allows HTML entities to escape normalization as a feature.
1403 * @param string $text Already normalized, containing entities
1404 * @return string Still normalized, without entities
1406 public static function decodeCharReferencesAndNormalize( $text ) {
1408 $text = preg_replace_callback(
1409 self
::CHAR_REFS_REGEX
,
1410 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1411 $text, /* limit */ -1, $count );
1414 return $wgContLang->normalize( $text );
1421 * @param string $matches
1424 static function decodeCharReferencesCallback( $matches ) {
1425 if ( $matches[1] != '' ) {
1426 return Sanitizer
::decodeEntity( $matches[1] );
1427 } elseif ( $matches[2] != '' ) {
1428 return Sanitizer
::decodeChar( intval( $matches[2] ) );
1429 } elseif ( $matches[3] != '' ) {
1430 return Sanitizer
::decodeChar( hexdec( $matches[3] ) );
1432 # Last case should be an ampersand by itself
1437 * Return UTF-8 string for a codepoint if that is a valid
1438 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1439 * @param int $codepoint
1443 static function decodeChar( $codepoint ) {
1444 if ( Sanitizer
::validateCodepoint( $codepoint ) ) {
1445 return codepointToUtf8( $codepoint );
1447 return UTF8_REPLACEMENT
;
1452 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1453 * return the UTF-8 encoding of that character. Otherwise, returns
1454 * pseudo-entity source (eg "&foo;")
1456 * @param string $name
1459 static function decodeEntity( $name ) {
1460 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1461 $name = self
::$htmlEntityAliases[$name];
1463 if ( isset( self
::$htmlEntities[$name] ) ) {
1464 return codepointToUtf8( self
::$htmlEntities[$name] );
1471 * Fetch the whitelist of acceptable attributes for a given element name.
1473 * @param string $element
1476 static function attributeWhitelist( $element ) {
1477 $list = Sanitizer
::setupAttributeWhitelist();
1478 return isset( $list[$element] )
1484 * Foreach array key (an allowed HTML element), return an array
1485 * of allowed attributes
1488 static function setupAttributeWhitelist() {
1489 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
1491 static $whitelist, $staticInitialised;
1492 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgAllowMicrodataAttributes' ) );
1494 if ( isset( $whitelist ) && $staticInitialised == $globalContext ) {
1511 if ( $wgAllowRdfaAttributes ) {
1512 # RDFa attributes as specified in section 9 of
1513 # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1514 $common = array_merge( $common, array(
1515 'about', 'property', 'resource', 'datatype', 'typeof',
1519 if ( $wgAllowMicrodataAttributes ) {
1520 # add HTML5 microdata tags as specified by
1521 # http://www.whatwg.org/html/microdata.html#the-microdata-model
1522 $common = array_merge( $common, array(
1523 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1527 $block = array_merge( $common, array( 'align' ) );
1528 $tablealign = array( 'align', 'valign' );
1536 'nowrap', # deprecated
1537 'width', # deprecated
1538 'height', # deprecated
1539 'bgcolor', # deprecated
1542 # Numbers refer to sections in HTML 4.01 standard describing the element.
1543 # See: http://www.w3.org/TR/html4/
1547 'center' => $common, # deprecated
1566 'strong' => $common,
1577 'blockquote' => array_merge( $common, array( 'cite' ) ),
1578 'q' => array_merge( $common, array( 'cite' ) ),
1588 'br' => array_merge( $common, array( 'clear' ) ),
1590 # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1594 'pre' => array_merge( $common, array( 'width' ) ),
1597 'ins' => array_merge( $common, array( 'cite', 'datetime' ) ),
1598 'del' => array_merge( $common, array( 'cite', 'datetime' ) ),
1601 'ul' => array_merge( $common, array( 'type' ) ),
1602 'ol' => array_merge( $common, array( 'type', 'start' ) ),
1603 'li' => array_merge( $common, array( 'type', 'value' ) ),
1611 'table' => array_merge( $common,
1612 array( 'summary', 'width', 'border', 'frame',
1613 'rules', 'cellspacing', 'cellpadding',
1618 'caption' => $block,
1626 'colgroup' => array_merge( $common, array( 'span' ) ),
1627 'col' => array_merge( $common, array( 'span' ) ),
1630 'tr' => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1633 'td' => array_merge( $common, $tablecell, $tablealign ),
1634 'th' => array_merge( $common, $tablecell, $tablealign ),
1637 # NOTE: <a> is not allowed directly, but the attrib
1638 # whitelist is used from the Parser object
1639 'a' => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1642 # Not usually allowed, but may be used for extension-style hooks
1643 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1645 'img' => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1653 'strike' => $common,
1658 'font' => array_merge( $common, array( 'size', 'color', 'face' ) ),
1662 'hr' => array_merge( $common, array( 'width' ) ),
1664 # HTML Ruby annotation text module, simple ruby only.
1665 # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
1670 'rt' => $common, #array_merge( $common, array( 'rbspan' ) ),
1673 # MathML root element, where used for extensions
1674 # 'title' may not be 100% valid here; it's XHTML
1675 # http://www.w3.org/TR/REC-MathML/
1676 'math' => array( 'class', 'style', 'id', 'title' ),
1678 # HTML 5 section 4.6
1681 # HTML5 elements, defined by:
1682 # http://www.whatwg.org/html/
1683 'data' => array_merge( $common, array( 'value' ) ),
1684 'time' => array_merge( $common, array( 'datetime' ) ),
1687 // meta and link are only permitted by removeHTMLtags when Microdata
1688 // is enabled so we don't bother adding a conditional to hide these
1689 // Also meta and link are only valid in WikiText as Microdata elements
1690 // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1691 // So we don't bother including $common attributes that have no purpose.
1692 'meta' => array( 'itemprop', 'content' ),
1693 'link' => array( 'itemprop', 'href' ),
1696 $staticInitialised = $globalContext;
1702 * Take a fragment of (potentially invalid) HTML and return
1703 * a version with any tags removed, encoded as plain text.
1705 * Warning: this return value must be further escaped for literal
1706 * inclusion in HTML output as of 1.10!
1708 * @param string $text HTML fragment
1711 static function stripAllTags( $text ) {
1713 $text = StringUtils
::delimiterReplace( '<', '>', '', $text );
1715 # Normalize &entities and whitespace
1716 $text = self
::decodeCharReferences( $text );
1717 $text = self
::normalizeWhitespace( $text );
1723 * Hack up a private DOCTYPE with HTML's standard entity declarations.
1724 * PHP 4 seemed to know these if you gave it an HTML doctype, but
1727 * Use for passing XHTML fragments to PHP's XML parsing functions
1731 static function hackDocType() {
1732 $out = "<!DOCTYPE html [\n";
1733 foreach ( self
::$htmlEntities as $entity => $codepoint ) {
1734 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1741 * @param string $url
1742 * @return mixed|string
1744 static function cleanUrl( $url ) {
1745 # Normalize any HTML entities in input. They will be
1746 # re-escaped by makeExternalLink().
1747 $url = Sanitizer
::decodeCharReferences( $url );
1749 # Escape any control characters introduced by the above step
1750 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1751 array( __CLASS__
, 'cleanUrlCallback' ), $url );
1753 # Validate hostname portion
1755 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1756 list( /* $whole */, $protocol, $host, $rest ) = $matches;
1758 // Characters that will be ignored in IDNs.
1759 // http://tools.ietf.org/html/3454#section-3.1
1760 // Strip them before further processing so blacklists and such work.
1762 \\s| # general whitespace
1763 \xc2\xad| # 00ad SOFT HYPHEN
1764 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1765 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1766 \xe2\x81\xa0| # 2060 WORD JOINER
1767 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1768 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
1769 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1770 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1771 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1772 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1773 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1774 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1777 $host = preg_replace( $strip, '', $host );
1779 // @todo FIXME: Validate hostnames here
1781 return $protocol . $host . $rest;
1788 * @param array $matches
1791 static function cleanUrlCallback( $matches ) {
1792 return urlencode( $matches[0] );
1796 * Does a string look like an e-mail address?
1798 * This validates an email address using an HTML5 specification found at:
1799 * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1800 * Which as of 2011-01-24 says:
1802 * A valid e-mail address is a string that matches the ABNF production
1803 * 1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1804 * in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1807 * This function is an implementation of the specification as requested in
1810 * Client-side forms will use the same standard validation rules via JS or
1811 * HTML 5 validation; additional restrictions can be enforced server-side
1812 * by extensions via the 'isValidEmailAddr' hook.
1814 * Note that this validation doesn't 100% match RFC 2822, but is believed
1815 * to be liberal enough for wide use. Some invalid addresses will still
1816 * pass validation here.
1820 * @param string $addr E-mail address
1823 public static function validateEmail( $addr ) {
1825 if ( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1829 // Please note strings below are enclosed in brackets [], this make the
1830 // hyphen "-" a range indicator. Hence it is double backslashed below.
1832 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1833 $rfc1034_ldh_str = "a-z0-9\\-";
1835 $HTML5_email_regexp = "/
1837 [$rfc5322_atext\\.]+ # user part which is liberal :p
1839 [$rfc1034_ldh_str]+ # First domain part
1840 (\\.[$rfc1034_ldh_str]+)* # Following part prefixed with a dot
1842 /ix"; // case Insensitive, eXtended
1844 return (bool)preg_match( $HTML5_email_regexp, $addr );