3 * HTML sanitizer for %MediaWiki.
5 * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
6 * https://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
27 use MediaWiki\MediaWikiServices
;
30 * HTML sanitizer for MediaWiki
35 * Regular expression to match various types of character references in
36 * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
38 const CHAR_REFS_REGEX
=
39 '/&([A-Za-z0-9\x80-\xff]+);
41 |&\#[xX]([0-9A-Fa-f]+);
45 * Acceptable tag name charset from HTML5 parsing spec
46 * https://www.w3.org/TR/html5/syntax.html#tag-open-state
48 const ELEMENT_BITS_REGEX
= '!^(/?)([A-Za-z][^\t\n\v />\0]*+)([^>]*?)(/?>)([^<]*)$!';
51 * Blacklist for evil uris like javascript:
52 * WARNING: DO NOT use this in any place that actually requires blacklisting
53 * for security reasons. There are NUMEROUS[1] ways to bypass blacklisting, the
54 * only way to be secure from javascript: uri based xss vectors is to whitelist
55 * things that you know are safe and deny everything else.
56 * [1]: http://ha.ckers.org/xss.html
58 const EVIL_URI_PATTERN
= '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
59 const XMLNS_ATTRIBUTE_PATTERN
= "/^xmlns:[:A-Z_a-z-.0-9]+$/";
62 * Tells escapeUrlForHtml() to encode the ID using the wiki's primary encoding.
69 * Tells escapeUrlForHtml() to encode the ID using the fallback encoding, or return false
70 * if no fallback is configured.
74 const ID_FALLBACK
= 1;
77 * List of all named character entities defined in HTML 4.01
78 * https://www.w3.org/TR/html4/sgml/entities.html
79 * As well as ' which is only defined starting in XHTML1.
81 private static $htmlEntities = [
97 'apos' => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
338 * Character entity aliases accepted by MediaWiki
340 private static $htmlEntityAliases = [
346 * Lazy-initialised attributes regex, see getAttribsRegex()
348 private static $attribsRegex;
351 * Regular expression to match HTML/XML attribute pairs within a tag.
352 * Based on https://www.w3.org/TR/html5/syntax.html#before-attribute-name-state
353 * Used in Sanitizer::decodeTagAttributes
356 static function getAttribsRegex() {
357 if ( self
::$attribsRegex === null ) {
358 $spaceChars = '\x09\x0a\x0c\x0d\x20';
359 $space = "[{$spaceChars}]";
360 $attrib = "[^{$spaceChars}\/>=]";
361 $attribFirst = "(?:{$attrib}|=)";
362 self
::$attribsRegex =
363 "/({$attribFirst}{$attrib}*)
366 # The attribute value: quoted or alone
373 return self
::$attribsRegex;
377 * Lazy-initialised attribute name regex, see getAttribNameRegex()
379 private static $attribNameRegex;
382 * Used in Sanitizer::decodeTagAttributes to filter attributes.
385 static function getAttribNameRegex() {
386 if ( self
::$attribNameRegex === null ) {
387 $attribFirst = "[:_\p{L}\p{N}]";
388 $attrib = "[:_\.\-\p{L}\p{N}]";
389 self
::$attribNameRegex = "/^({$attribFirst}{$attrib}*)$/sxu";
391 return self
::$attribNameRegex;
395 * Return the various lists of recognized tags
396 * @param array $extratags For any extra tags to include
397 * @param array $removetags For any tags (default or extra) to exclude
400 public static function getRecognizedTagData( $extratags = [], $removetags = [] ) {
401 global $wgAllowImageTag;
403 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
404 $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
406 // Base our staticInitialised variable off of the global config state so that if the globals
407 // are changed (like in the screwed up test system) we will re-initialise the settings.
408 $globalContext = $wgAllowImageTag;
409 if ( !$staticInitialised ||
$staticInitialised != $globalContext ) {
410 $htmlpairsStatic = [ # Tags that must be closed
411 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
412 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
413 'strike', 'strong', 'tt', 'var', 'div', 'center',
414 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
415 'ruby', 'rb', 'rp', 'rt', 'rtc', 'p', 'span', 'abbr', 'dfn',
416 'kbd', 'samp', 'data', 'time', 'mark'
419 'br', 'wbr', 'hr', 'li', 'dt', 'dd', 'meta', 'link'
422 # Elements that cannot have close tags. This is (not coincidentally)
423 # also the list of tags for which the HTML 5 parsing algorithm
424 # requires you to "acknowledge the token's self-closing flag", i.e.
425 # a self-closing tag like <br/> is not an HTML 5 parse error only
428 'br', 'wbr', 'hr', 'meta', 'link'
431 $htmlnest = [ # Tags that can be nested--??
432 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
433 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
434 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
436 $tabletags = [ # Can only appear inside table, we will close them
439 $htmllist = [ # Tags used by list
442 $listtags = [ # Tags that can appear in a list
446 if ( $wgAllowImageTag ) {
447 $htmlsingle[] = 'img';
448 $htmlsingleonly[] = 'img';
451 $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
452 $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
454 # Convert them all to hashtables for faster lookup
455 $vars = [ 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
456 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' ];
457 foreach ( $vars as $var ) {
458 $
$var = array_flip( $
$var );
460 $staticInitialised = $globalContext;
463 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
464 $extratags = array_flip( $extratags );
465 $removetags = array_flip( $removetags );
466 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
467 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
470 'htmlpairs' => $htmlpairs,
471 'htmlsingle' => $htmlsingle,
472 'htmlsingleonly' => $htmlsingleonly,
473 'htmlnest' => $htmlnest,
474 'tabletags' => $tabletags,
475 'htmllist' => $htmllist,
476 'listtags' => $listtags,
477 'htmlsingleallowed' => $htmlsingleallowed,
478 'htmlelements' => $htmlelements,
483 * Cleans up HTML, removes dangerous tags and attributes, and
484 * removes HTML comments
485 * @param string $text
486 * @param callable|null $processCallback Callback to do any variable or parameter
487 * replacements in HTML attribute values
488 * @param array|bool $args Arguments for the processing callback
489 * @param array $extratags For any extra tags to include
490 * @param array $removetags For any tags (default or extra) to exclude
491 * @param callable|null $warnCallback (Deprecated) Callback allowing the
492 * addition of a tracking category when bad input is encountered.
493 * DO NOT ADD NEW PARAMETERS AFTER $warnCallback, since it will be
497 public static function removeHTMLtags( $text, $processCallback = null,
498 $args = [], $extratags = [], $removetags = [], $warnCallback = null
500 $tagData = self
::getRecognizedTagData( $extratags, $removetags );
501 $htmlpairs = $tagData['htmlpairs'];
502 $htmlsingle = $tagData['htmlsingle'];
503 $htmlsingleonly = $tagData['htmlsingleonly'];
504 $htmlnest = $tagData['htmlnest'];
505 $tabletags = $tagData['tabletags'];
506 $htmllist = $tagData['htmllist'];
507 $listtags = $tagData['listtags'];
508 $htmlsingleallowed = $tagData['htmlsingleallowed'];
509 $htmlelements = $tagData['htmlelements'];
511 # Remove HTML comments
512 $text = self
::removeHTMLcomments( $text );
513 $bits = explode( '<', $text );
514 $text = str_replace( '>', '>', array_shift( $bits ) );
515 if ( !MWTidy
::isEnabled() ) {
516 wfDeprecated( 'disabling tidy', '1.33' );
517 $tagstack = $tablestack = [];
518 foreach ( $bits as $x ) {
520 # $slash: Does the current element start with a '/'?
521 # $t: Current element name
522 # $params: String between element name and >
523 # $brace: Ending '>' or '/>'
524 # $rest: Everything until the next element of $bits
525 if ( preg_match( self
::ELEMENT_BITS_REGEX
, $x, $regs ) ) {
526 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
528 $slash = $t = $params = $brace = $rest = null;
532 $t = strtolower( $t );
533 if ( isset( $htmlelements[$t] ) ) {
535 if ( $slash && isset( $htmlsingleonly[$t] ) ) {
537 } elseif ( $slash ) {
538 # Closing a tag... is it the one we just opened?
539 Wikimedia\
suppressWarnings();
540 $ot = array_pop( $tagstack );
541 Wikimedia\restoreWarnings
();
544 if ( isset( $htmlsingleallowed[$ot] ) ) {
545 # Pop all elements with an optional close tag
546 # and see if we find a match below them
548 array_push( $optstack, $ot );
549 Wikimedia\
suppressWarnings();
550 $ot = array_pop( $tagstack );
551 Wikimedia\restoreWarnings
();
552 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
553 array_push( $optstack, $ot );
554 Wikimedia\
suppressWarnings();
555 $ot = array_pop( $tagstack );
556 Wikimedia\restoreWarnings
();
559 # No match. Push the optional elements back again
561 Wikimedia\
suppressWarnings();
562 $ot = array_pop( $optstack );
563 Wikimedia\restoreWarnings
();
565 array_push( $tagstack, $ot );
566 Wikimedia\
suppressWarnings();
567 $ot = array_pop( $optstack );
568 Wikimedia\restoreWarnings
();
572 Wikimedia\
suppressWarnings();
573 array_push( $tagstack, $ot );
574 Wikimedia\restoreWarnings
();
576 # <li> can be nested in <ul> or <ol>, skip those cases:
577 if ( !isset( $htmllist[$ot] ) ||
!isset( $listtags[$t] ) ) {
581 } elseif ( $t == 'table' ) {
582 $tagstack = array_pop( $tablestack );
586 # Keep track for later
587 if ( isset( $tabletags[$t] ) && !in_array( 'table', $tagstack ) ) {
589 } elseif ( in_array( $t, $tagstack ) && !isset( $htmlnest[$t] ) ) {
591 # Is it a self closed htmlpair ? (T7487)
592 } elseif ( $brace == '/>' && isset( $htmlpairs[$t] ) ) {
593 // Eventually we'll just remove the self-closing
594 // slash, in order to be consistent with HTML5
597 // For now, let's just warn authors to clean up.
598 if ( is_callable( $warnCallback ) ) {
599 call_user_func_array( $warnCallback, [ 'deprecated-self-close-category' ] );
602 } elseif ( isset( $htmlsingleonly[$t] ) ) {
603 # Hack to force empty tag for unclosable elements
605 } elseif ( isset( $htmlsingle[$t] ) ) {
606 # Hack to not close $htmlsingle tags
608 # Still need to push this optionally-closed tag to
609 # the tag stack so that we can match end tags
610 # instead of marking them as bad.
611 array_push( $tagstack, $t );
612 } elseif ( isset( $tabletags[$t] ) && in_array( $t, $tagstack ) ) {
613 // New table tag but forgot to close the previous one
616 if ( $t == 'table' ) {
617 array_push( $tablestack, $tagstack );
620 array_push( $tagstack, $t );
623 # Replace any variables or template parameters with
625 if ( is_callable( $processCallback ) ) {
626 call_user_func_array( $processCallback, [ &$params, $args ] );
629 if ( !self
::validateTag( $params, $t ) ) {
633 # Strip non-approved attributes from the tag
634 $newparams = self
::fixTagAttributes( $params, $t );
637 $rest = str_replace( '>', '>', $rest );
638 $close = ( $brace == '/>' && !$slash ) ?
' /' : '';
639 $text .= "<$slash$t$newparams$close>$rest";
643 $text .= '<' . str_replace( '>', '>', $x );
645 # Close off any remaining tags
646 while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
648 if ( $t == 'table' ) {
649 $tagstack = array_pop( $tablestack );
653 # this might be possible using tidy itself
654 foreach ( $bits as $x ) {
655 if ( preg_match( self
::ELEMENT_BITS_REGEX
, $x, $regs ) ) {
656 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
659 $t = strtolower( $t );
660 if ( isset( $htmlelements[$t] ) ) {
661 if ( is_callable( $processCallback ) ) {
662 call_user_func_array( $processCallback, [ &$params, $args ] );
665 if ( $brace == '/>' && !( isset( $htmlsingle[$t] ) ||
isset( $htmlsingleonly[$t] ) ) ) {
666 // Eventually we'll just remove the self-closing
667 // slash, in order to be consistent with HTML5
670 // For now, let's just warn authors to clean up.
671 if ( is_callable( $warnCallback ) ) {
672 call_user_func_array( $warnCallback, [ 'deprecated-self-close-category' ] );
675 if ( !self
::validateTag( $params, $t ) ) {
679 $newparams = self
::fixTagAttributes( $params, $t );
681 if ( $brace === '/>' && !isset( $htmlsingleonly[$t] ) ) {
682 # Interpret self-closing tags as empty tags even when
683 # HTML 5 would interpret them as start tags. Such input
684 # is commonly seen on Wikimedia wikis with this intention.
688 $rest = str_replace( '>', '>', $rest );
689 $text .= "<$slash$t$newparams$brace$rest";
694 $text .= '<' . str_replace( '>', '>', $x );
701 * Remove '<!--', '-->', and everything between.
702 * To avoid leaving blank lines, when a comment is both preceded
703 * and followed by a newline (ignoring spaces), trim leading and
704 * trailing spaces and one of the newlines.
706 * @param string $text
709 public static function removeHTMLcomments( $text ) {
710 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
711 $end = strpos( $text, '-->', $start +
4 );
712 if ( $end === false ) {
713 # Unterminated comment; bail out
719 # Trim space and newline if the comment is both
720 # preceded and followed by a newline
721 $spaceStart = max( $start - 1, 0 );
722 $spaceLen = $end - $spaceStart;
723 while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
727 while ( substr( $text, $spaceStart +
$spaceLen, 1 ) === ' ' ) {
730 if ( substr( $text, $spaceStart, 1 ) === "\n"
731 && substr( $text, $spaceStart +
$spaceLen, 1 ) === "\n" ) {
732 # Remove the comment, leading and trailing
733 # spaces, and leave only one newline.
734 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen +
1 );
736 # Remove just the comment.
737 $text = substr_replace( $text, '', $start, $end - $start );
744 * Takes attribute names and values for a tag and the tag name and
745 * validates that the tag is allowed to be present.
746 * This DOES NOT validate the attributes, nor does it validate the
747 * tags themselves. This method only handles the special circumstances
748 * where we may want to allow a tag within content but ONLY when it has
749 * specific attributes set.
751 * @param string $params
752 * @param string $element
755 static function validateTag( $params, $element ) {
756 $params = self
::decodeTagAttributes( $params );
758 if ( $element == 'meta' ||
$element == 'link' ) {
759 if ( !isset( $params['itemprop'] ) ) {
760 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
763 if ( $element == 'meta' && !isset( $params['content'] ) ) {
764 // <meta> must have a content="" for the itemprop
767 if ( $element == 'link' && !isset( $params['href'] ) ) {
768 // <link> must have an associated href=""
777 * Take an array of attribute names and values and normalize or discard
778 * illegal values for the given element type.
780 * - Discards attributes not on a whitelist for the given element
781 * - Unsafe style attributes are discarded
782 * - Invalid id attributes are re-encoded
784 * @param array $attribs
785 * @param string $element
788 * @todo Check for legal values where the DTD limits things.
789 * @todo Check for unique id attribute :P
791 static function validateTagAttributes( $attribs, $element ) {
792 return self
::validateAttributes( $attribs,
793 self
::attributeWhitelist( $element ) );
797 * Take an array of attribute names and values and normalize or discard
798 * illegal values for the given whitelist.
800 * - Discards attributes not on the given whitelist
801 * - Unsafe style attributes are discarded
802 * - Invalid id attributes are re-encoded
804 * @param array $attribs
805 * @param array $whitelist List of allowed attribute names
808 * @todo Check for legal values where the DTD limits things.
809 * @todo Check for unique id attribute :P
811 static function validateAttributes( $attribs, $whitelist ) {
812 $whitelist = array_flip( $whitelist );
813 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
816 foreach ( $attribs as $attribute => $value ) {
817 # Allow XML namespace declaration to allow RDFa
818 if ( preg_match( self
::XMLNS_ATTRIBUTE_PATTERN
, $attribute ) ) {
819 if ( !preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
820 $out[$attribute] = $value;
826 # Allow any attribute beginning with "data-"
828 # * Disallow data attributes used by MediaWiki code
829 # * Ensure that the attribute is not namespaced by banning
831 if ( !preg_match( '/^data-[^:]*$/i', $attribute )
832 && !isset( $whitelist[$attribute] )
833 || self
::isReservedDataAttribute( $attribute )
838 # Strip javascript "expression" from stylesheets.
839 # https://msdn.microsoft.com/en-us/library/ms537634.aspx
840 if ( $attribute == 'style' ) {
841 $value = self
::checkCss( $value );
844 # Escape HTML id attributes
845 if ( $attribute === 'id' ) {
846 $value = self
::escapeIdForAttribute( $value, self
::ID_PRIMARY
);
849 # Escape HTML id reference lists
850 if ( $attribute === 'aria-describedby'
851 ||
$attribute === 'aria-flowto'
852 ||
$attribute === 'aria-labelledby'
853 ||
$attribute === 'aria-owns'
855 $value = self
::escapeIdReferenceList( $value );
858 // RDFa and microdata properties allow URLs, URIs and/or CURIs.
859 // Check them for sanity.
860 if ( $attribute === 'rel' ||
$attribute === 'rev'
862 ||
$attribute === 'about' ||
$attribute === 'property'
863 ||
$attribute === 'resource' ||
$attribute === 'datatype'
864 ||
$attribute === 'typeof'
866 ||
$attribute === 'itemid' ||
$attribute === 'itemprop'
867 ||
$attribute === 'itemref' ||
$attribute === 'itemscope'
868 ||
$attribute === 'itemtype'
870 // Paranoia. Allow "simple" values but suppress javascript
871 if ( preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
876 # NOTE: even though elements using href/src are not allowed directly, supply
877 # validation code that can be used by tag hook handlers, etc
878 if ( $attribute === 'href' ||
$attribute === 'src' ||
$attribute === 'poster' ) {
879 if ( !preg_match( $hrefExp, $value ) ) {
880 continue; // drop any href or src attributes not using an allowed protocol.
881 // NOTE: this also drops all relative URLs
885 // If this attribute was previously set, override it.
886 // Output should only have one attribute of each name.
887 $out[$attribute] = $value;
890 # itemtype, itemid, itemref don't make sense without itemscope
891 if ( !array_key_exists( 'itemscope', $out ) ) {
892 unset( $out['itemtype'] );
893 unset( $out['itemid'] );
894 unset( $out['itemref'] );
896 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
902 * Given an attribute name, checks whether it is a reserved data attribute
903 * (such as data-mw-foo) which is unavailable to user-generated HTML so MediaWiki
904 * core and extension code can safely use it to communicate with frontend code.
905 * @param string $attr Attribute name.
908 public static function isReservedDataAttribute( $attr ) {
909 // data-ooui is reserved for ooui.
910 // data-mw and data-parsoid are reserved for parsoid.
911 // data-mw-<name here> is reserved for extensions (or core) if
912 // they need to communicate some data to the client and want to be
913 // sure that it isn't coming from an untrusted user.
914 // We ignore the possibility of namespaces since user-generated HTML
915 // can't use them anymore.
916 return (bool)preg_match( '/^data-(ooui|mw|parsoid)/i', $attr );
920 * Merge two sets of HTML attributes. Conflicting items in the second set
921 * will override those in the first, except for 'class' attributes which
922 * will be combined (if they're both strings).
924 * @todo implement merging for other attributes such as style
929 static function mergeAttributes( $a, $b ) {
930 $out = array_merge( $a, $b );
931 if ( isset( $a['class'] ) && isset( $b['class'] )
932 && is_string( $a['class'] ) && is_string( $b['class'] )
933 && $a['class'] !== $b['class']
935 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
936 -1, PREG_SPLIT_NO_EMPTY
);
937 $out['class'] = implode( ' ', array_unique( $classes ) );
943 * Normalize CSS into a format we can easily search for hostile input
944 * - decode character references
945 * - decode escape sequences
946 * - convert characters that IE6 interprets into ascii
947 * - remove comments, unless the entire value is one single comment
948 * @param string $value the css string
949 * @return string normalized css
951 public static function normalizeCss( $value ) {
952 // Decode character references like {
953 $value = self
::decodeCharReferences( $value );
955 // Decode escape sequences and line continuation
956 // See the grammar in the CSS 2 spec, appendix D.
957 // This has to be done AFTER decoding character references.
958 // This means it isn't possible for this function to return
959 // unsanitized escape sequences. It is possible to manufacture
960 // input that contains character references that decode to
961 // escape sequences that decode to character references, but
962 // it's OK for the return value to contain character references
963 // because the caller is supposed to escape those anyway.
965 if ( !$decodeRegex ) {
966 $space = '[\\x20\\t\\r\\n\\f]';
967 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
969 $decodeRegex = "/ $backslash
971 ($nl) | # 1. Line continuation
972 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
973 (.) | # 3. backslash cancelling special meaning
974 () | # 4. backslash at end of string
977 $value = preg_replace_callback( $decodeRegex,
978 [ __CLASS__
, 'cssDecodeCallback' ], $value );
980 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
981 $value = preg_replace_callback(
982 '/[!-[]-z]/u', // U+FF01 to U+FF5A, excluding U+FF3C (T60088)
983 function ( $matches ) {
984 $cp = UtfNormal\Utils
::utf8ToCodepoint( $matches[0] );
985 if ( $cp === false ) {
988 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
993 // Convert more characters IE6 might treat as ascii
994 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
995 $value = str_replace(
996 [ 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ],
997 [ 'r', 'n', 'n', 'l', 'i', '(', '(' ],
1001 // Let the value through if it's nothing but a single comment, to
1002 // allow other functions which may reject it to pass some error
1004 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
1005 // Remove any comments; IE gets token splitting wrong
1006 // This must be done AFTER decoding character references and
1007 // escape sequences, because those steps can introduce comments
1008 // This step cannot introduce character references or escape
1009 // sequences, because it replaces comments with spaces rather
1010 // than removing them completely.
1011 $value = StringUtils
::delimiterReplace( '/*', '*/', ' ', $value );
1013 // Remove anything after a comment-start token, to guard against
1014 // incorrect client implementations.
1015 $commentPos = strpos( $value, '/*' );
1016 if ( $commentPos !== false ) {
1017 $value = substr( $value, 0, $commentPos );
1021 // S followed by repeat, iteration, or prolonged sound marks,
1022 // which IE will treat as "ss"
1023 $value = preg_replace(
1025 \xE3\x80\xB1 | # U+3031
1026 \xE3\x82\x9D | # U+309D
1027 \xE3\x83\xBC | # U+30FC
1028 \xE3\x83\xBD | # U+30FD
1029 \xEF\xB9\xBC | # U+FE7C
1030 \xEF\xB9\xBD | # U+FE7D
1031 \xEF\xBD\xB0 # U+FF70
1041 * Pick apart some CSS and check it for forbidden or unsafe structures.
1042 * Returns a sanitized string. This sanitized string will have
1043 * character references and escape sequences decoded and comments
1044 * stripped (unless it is itself one valid comment, in which case the value
1045 * will be passed through). If the input is just too evil, only a comment
1046 * complaining about evilness will be returned.
1048 * Currently URL references, 'expression', 'tps' are forbidden.
1050 * NOTE: Despite the fact that character references are decoded, the
1051 * returned string may contain character references given certain
1052 * clever input strings. These character references must
1053 * be escaped before the return value is embedded in HTML.
1055 * @param string $value
1058 static function checkCss( $value ) {
1059 $value = self
::normalizeCss( $value );
1061 // Reject problematic keywords and control characters
1062 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ||
1063 strpos( $value, UtfNormal\Constants
::UTF8_REPLACEMENT
) !== false ) {
1064 return '/* invalid control char */';
1065 } elseif ( preg_match(
1070 | -o-link-source\s*:
1075 | attr\s*\([^)]+[\s,]+url
1078 return '/* insecure input */';
1084 * @param array $matches
1087 static function cssDecodeCallback( $matches ) {
1088 if ( $matches[1] !== '' ) {
1089 // Line continuation
1091 } elseif ( $matches[2] !== '' ) {
1092 $char = UtfNormal\Utils
::codepointToUtf8( hexdec( $matches[2] ) );
1093 } elseif ( $matches[3] !== '' ) {
1094 $char = $matches[3];
1098 if ( $char == "\n" ||
$char == '"' ||
$char == "'" ||
$char == '\\' ) {
1099 // These characters need to be escaped in strings
1100 // Clean up the escape sequence to avoid parsing errors by clients
1101 return '\\' . dechex( ord( $char ) ) . ' ';
1103 // Decode unnecessary escape
1109 * Take a tag soup fragment listing an HTML element's attributes
1110 * and normalize it to well-formed XML, discarding unwanted attributes.
1111 * Output is safe for further wikitext processing, with escaping of
1112 * values that could trigger problems.
1114 * - Normalizes attribute names to lowercase
1115 * - Discards attributes not on a whitelist for the given element
1116 * - Turns broken or invalid entities into plaintext
1117 * - Double-quotes all attribute values
1118 * - Attributes without values are given the name as attribute
1119 * - Double attributes are discarded
1120 * - Unsafe style attributes are discarded
1121 * - Prepends space if there are attributes.
1122 * - (Optionally) Sorts attributes by name.
1124 * @param string $text
1125 * @param string $element
1126 * @param bool $sorted Whether to sort the attributes (default: false)
1129 static function fixTagAttributes( $text, $element, $sorted = false ) {
1130 if ( trim( $text ) == '' ) {
1134 $decoded = self
::decodeTagAttributes( $text );
1135 $stripped = self
::validateTagAttributes( $decoded, $element );
1141 return self
::safeEncodeTagAttributes( $stripped );
1145 * Encode an attribute value for HTML output.
1146 * @param string $text
1147 * @return string HTML-encoded text fragment
1149 static function encodeAttribute( $text ) {
1150 $encValue = htmlspecialchars( $text, ENT_QUOTES
);
1152 // Whitespace is normalized during attribute decoding,
1153 // so if we've been passed non-spaces we must encode them
1154 // ahead of time or they won't be preserved.
1155 $encValue = strtr( $encValue, [
1165 * Armor French spaces with a replacement character
1168 * @param string $text Text to armor
1169 * @param string $space Space character for the French spaces, defaults to ' '
1170 * @return string Armored text
1172 public static function armorFrenchSpaces( $text, $space = ' ' ) {
1173 // Replace $ with \$ and \ with \\
1174 $space = preg_replace( '#(?<!\\\\)(\\$|\\\\)#', '\\\\$1', $space );
1176 # French spaces, last one Guillemet-left
1177 # only if there is something before the space
1178 # and a non-word character after the punctuation.
1179 '/(\S) (?=[?:;!%»›](?!\w))/u' => "\\1$space",
1180 # French spaces, Guillemet-right
1181 '/([«‹]) /u' => "\\1$space",
1183 return preg_replace( array_keys( $fixtags ), array_values( $fixtags ), $text );
1187 * Encode an attribute value for HTML tags, with extra armoring
1188 * against further wiki processing.
1189 * @param string $text
1190 * @return string HTML-encoded text fragment
1192 static function safeEncodeAttribute( $text ) {
1193 $encValue = self
::encodeAttribute( $text );
1195 # Templates and links may be expanded in later parsing,
1196 # creating invalid or dangerous output. Suppress this.
1197 $encValue = strtr( $encValue, [
1198 '<' => '<', // This should never happen,
1199 '>' => '>', // we've received invalid input
1200 '"' => '"', // which should have been escaped.
1202 '}' => '}', // prevent unpaired language conversion syntax
1205 "''" => '''',
1206 'ISBN' => 'ISBN',
1208 'PMID' => 'PMID',
1213 # Armor against French spaces detection (T5158)
1214 $encValue = self
::armorFrenchSpaces( $encValue, ' ' );
1217 $encValue = preg_replace_callback(
1218 '/((?i)' . wfUrlProtocols() . ')/',
1219 function ( $matches ) {
1220 return str_replace( ':', ':', $matches[1] );
1227 * Given a value, escape it so that it can be used in an id attribute and
1228 * return it. This will use HTML5 validation, allowing anything but ASCII
1231 * To ensure we don't have to bother escaping anything, we also strip ', ".
1232 * TODO: Is this the best tactic?
1234 * We also strip # because it upsets IE, and % because it could be
1235 * ambiguous if it's part of something that looks like a percent escape
1236 * (which don't work reliably in fragments cross-browser).
1238 * @deprecated since 1.30, use one of this class' escapeIdFor*() functions
1240 * @see https://www.w3.org/TR/html401/types.html#type-name Valid characters
1241 * in the id and name attributes
1242 * @see https://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1244 * @see https://www.w3.org/TR/html5/dom.html#the-id-attribute
1245 * HTML5 definition of id attribute
1247 * @param string $id Id to escape
1248 * @param string|array $options String or array of strings (default is array()):
1249 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1250 * so don't pay attention if the first character isn't valid at the
1251 * beginning of an id.
1254 static function escapeId( $id, $options = [] ) {
1255 $options = (array)$options;
1257 // HTML4-style escaping
1263 $id = urlencode( strtr( $id, ' ', '_' ) );
1264 $id = strtr( $id, $replace );
1266 if ( !preg_match( '/^[a-zA-Z]/', $id ) && !in_array( 'noninitial', $options ) ) {
1267 // Initial character must be a letter!
1274 * Given a section name or other user-generated or otherwise unsafe string, escapes it to be
1275 * a valid HTML id attribute.
1277 * WARNING: unlike escapeId(), the output of this function is not guaranteed to be HTML safe,
1278 * be sure to use proper escaping.
1280 * @param string $id String to escape
1281 * @param int $mode One of ID_* constants, specifying whether the primary or fallback encoding
1283 * @return string|bool Escaped ID or false if fallback encoding is requested but it's not
1288 public static function escapeIdForAttribute( $id, $mode = self
::ID_PRIMARY
) {
1289 global $wgFragmentMode;
1291 if ( !isset( $wgFragmentMode[$mode] ) ) {
1292 if ( $mode === self
::ID_PRIMARY
) {
1293 throw new UnexpectedValueException( '$wgFragmentMode is configured with no primary mode' );
1298 $internalMode = $wgFragmentMode[$mode];
1300 return self
::escapeIdInternal( $id, $internalMode );
1304 * Given a section name or other user-generated or otherwise unsafe string, escapes it to be
1305 * a valid URL fragment.
1307 * WARNING: unlike escapeId(), the output of this function is not guaranteed to be HTML safe,
1308 * be sure to use proper escaping.
1310 * @param string $id String to escape
1311 * @return string Escaped ID
1315 public static function escapeIdForLink( $id ) {
1316 global $wgFragmentMode;
1318 if ( !isset( $wgFragmentMode[self
::ID_PRIMARY
] ) ) {
1319 throw new UnexpectedValueException( '$wgFragmentMode is configured with no primary mode' );
1322 $mode = $wgFragmentMode[self
::ID_PRIMARY
];
1324 $id = self
::escapeIdInternal( $id, $mode );
1330 * Given a section name or other user-generated or otherwise unsafe string, escapes it to be
1331 * a valid URL fragment for external interwikis.
1333 * @param string $id String to escape
1334 * @return string Escaped ID
1338 public static function escapeIdForExternalInterwiki( $id ) {
1339 global $wgExternalInterwikiFragmentMode;
1341 $id = self
::escapeIdInternal( $id, $wgExternalInterwikiFragmentMode );
1347 * Helper for escapeIdFor*() functions. Performs most of the actual escaping.
1349 * @param string $id String to escape
1350 * @param string $mode One of modes from $wgFragmentMode
1353 private static function escapeIdInternal( $id, $mode ) {
1356 $id = str_replace( ' ', '_', $id );
1359 // This corresponds to 'noninitial' mode of the old escapeId()
1365 $id = urlencode( str_replace( ' ', '_', $id ) );
1366 $id = strtr( $id, $replace );
1369 throw new InvalidArgumentException( "Invalid mode '$mode' passed to '" . __METHOD__
);
1376 * Given a string containing a space delimited list of ids, escape each id
1377 * to match ids escaped by the escapeIdForAttribute() function.
1381 * @param string $referenceString Space delimited list of ids
1384 public static function escapeIdReferenceList( $referenceString ) {
1385 # Explode the space delimited list string into an array of tokens
1386 $references = preg_split( '/\s+/', "{$referenceString}", -1, PREG_SPLIT_NO_EMPTY
);
1388 # Escape each token as an id
1389 foreach ( $references as &$ref ) {
1390 $ref = self
::escapeIdForAttribute( $ref );
1393 # Merge the array back to a space delimited list string
1394 # If the array is empty, the result will be an empty string ('')
1395 $referenceString = implode( ' ', $references );
1397 return $referenceString;
1401 * Given a value, escape it so that it can be used as a CSS class and
1404 * @todo For extra validity, input should be validated UTF-8.
1406 * @see https://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1408 * @param string $class
1411 static function escapeClass( $class ) {
1412 // Convert ugly stuff to underscores and kill underscores in ugly places
1413 return rtrim( preg_replace(
1414 [ '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ],
1420 * Given HTML input, escape with htmlspecialchars but un-escape entities.
1421 * This allows (generally harmless) entities like   to survive.
1423 * @param string $html HTML to escape
1424 * @return string Escaped input
1426 static function escapeHtmlAllowEntities( $html ) {
1427 $html = self
::decodeCharReferences( $html );
1428 # It seems wise to escape ' as well as ", as a matter of course. Can't
1429 # hurt. Use ENT_SUBSTITUTE so that incorrectly truncated multibyte characters
1430 # don't cause the entire string to disappear.
1431 $html = htmlspecialchars( $html, ENT_QUOTES | ENT_SUBSTITUTE
);
1436 * Return an associative array of attribute names and values from
1437 * a partial tag string. Attribute names are forced to lowercase,
1438 * character references are decoded to UTF-8 text.
1440 * @param string $text
1443 public static function decodeTagAttributes( $text ) {
1444 if ( trim( $text ) == '' ) {
1449 if ( !preg_match_all(
1450 self
::getAttribsRegex(),
1453 PREG_SET_ORDER
) ) {
1458 foreach ( $pairs as $set ) {
1459 $attribute = strtolower( $set[1] );
1461 // Filter attribute names with unacceptable characters
1462 if ( !preg_match( self
::getAttribNameRegex(), $attribute ) ) {
1466 $value = self
::getTagAttributeCallback( $set );
1468 // Normalize whitespace
1469 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1470 $value = trim( $value );
1472 // Decode character references
1473 $attribs[$attribute] = self
::decodeCharReferences( $value );
1479 * Build a partial tag string from an associative array of attribute
1480 * names and values as returned by decodeTagAttributes.
1482 * @param array $assoc_array
1485 public static function safeEncodeTagAttributes( $assoc_array ) {
1487 foreach ( $assoc_array as $attribute => $value ) {
1488 $encAttribute = htmlspecialchars( $attribute );
1489 $encValue = self
::safeEncodeAttribute( $value );
1491 $attribs[] = "$encAttribute=\"$encValue\"";
1493 return count( $attribs ) ?
' ' . implode( ' ', $attribs ) : '';
1497 * Pick the appropriate attribute value from a match set from the
1498 * attribs regex matches.
1501 * @throws MWException When tag conditions are not met.
1504 private static function getTagAttributeCallback( $set ) {
1505 if ( isset( $set[5] ) ) {
1508 } elseif ( isset( $set[4] ) ) {
1511 } elseif ( isset( $set[3] ) ) {
1514 } elseif ( !isset( $set[2] ) ) {
1515 # In XHTML, attributes must have a value so return an empty string.
1516 # See "Empty attribute syntax",
1517 # https://www.w3.org/TR/html5/syntax.html#syntax-attribute-name
1520 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1525 * @param string $text
1528 private static function normalizeWhitespace( $text ) {
1529 return trim( preg_replace(
1530 '/(?:\r\n|[\x20\x0d\x0a\x09])+/',
1536 * Normalizes whitespace in a section name, such as might be returned
1537 * by Parser::stripSectionName(), for use in the id's that are used for
1540 * @param string $section
1543 static function normalizeSectionNameWhitespace( $section ) {
1544 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1548 * Ensure that any entities and character references are legal
1549 * for XML and XHTML specifically. Any stray bits will be
1550 * &-escaped to result in a valid text fragment.
1552 * a. named char refs can only be < > & ", others are
1553 * numericized (this way we're well-formed even without a DTD)
1554 * b. any numeric char refs must be legal chars, not invalid or forbidden
1555 * c. use lower cased "&#x", not "&#X"
1556 * d. fix or reject non-valid attributes
1558 * @param string $text
1562 static function normalizeCharReferences( $text ) {
1563 return preg_replace_callback(
1564 self
::CHAR_REFS_REGEX
,
1565 [ self
::class, 'normalizeCharReferencesCallback' ],
1570 * @param string $matches
1573 static function normalizeCharReferencesCallback( $matches ) {
1575 if ( $matches[1] != '' ) {
1576 $ret = self
::normalizeEntity( $matches[1] );
1577 } elseif ( $matches[2] != '' ) {
1578 $ret = self
::decCharReference( $matches[2] );
1579 } elseif ( $matches[3] != '' ) {
1580 $ret = self
::hexCharReference( $matches[3] );
1582 if ( is_null( $ret ) ) {
1583 return htmlspecialchars( $matches[0] );
1590 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1591 * return the equivalent numeric entity reference (except for the core <
1592 * > & "). If the entity is a MediaWiki-specific alias, returns
1593 * the HTML equivalent. Otherwise, returns HTML-escaped text of
1594 * pseudo-entity source (eg &foo;)
1596 * @param string $name
1599 static function normalizeEntity( $name ) {
1600 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1601 return '&' . self
::$htmlEntityAliases[$name] . ';';
1602 } elseif ( in_array( $name, [ 'lt', 'gt', 'amp', 'quot' ] ) ) {
1604 } elseif ( isset( self
::$htmlEntities[$name] ) ) {
1605 return '&#' . self
::$htmlEntities[$name] . ';';
1607 return "&$name;";
1612 * @param int $codepoint
1613 * @return null|string
1615 static function decCharReference( $codepoint ) {
1616 $point = intval( $codepoint );
1617 if ( self
::validateCodepoint( $point ) ) {
1618 return sprintf( '&#%d;', $point );
1625 * @param int $codepoint
1626 * @return null|string
1628 static function hexCharReference( $codepoint ) {
1629 $point = hexdec( $codepoint );
1630 if ( self
::validateCodepoint( $point ) ) {
1631 return sprintf( '&#x%x;', $point );
1638 * Returns true if a given Unicode codepoint is a valid character in
1639 * both HTML5 and XML.
1640 * @param int $codepoint
1643 private static function validateCodepoint( $codepoint ) {
1644 # U+000C is valid in HTML5 but not allowed in XML.
1645 # U+000D is valid in XML but not allowed in HTML5.
1646 # U+007F - U+009F are disallowed in HTML5 (control characters).
1647 return $codepoint == 0x09
1648 ||
$codepoint == 0x0a
1649 ||
( $codepoint >= 0x20 && $codepoint <= 0x7e )
1650 ||
( $codepoint >= 0xa0 && $codepoint <= 0xd7ff )
1651 ||
( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1652 ||
( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1656 * Decode any character references, numeric or named entities,
1657 * in the text and return a UTF-8 string.
1659 * @param string $text
1662 public static function decodeCharReferences( $text ) {
1663 return preg_replace_callback(
1664 self
::CHAR_REFS_REGEX
,
1665 [ self
::class, 'decodeCharReferencesCallback' ],
1670 * Decode any character references, numeric or named entities,
1671 * in the next and normalize the resulting string. (T16952)
1673 * This is useful for page titles, not for text to be displayed,
1674 * MediaWiki allows HTML entities to escape normalization as a feature.
1676 * @param string $text Already normalized, containing entities
1677 * @return string Still normalized, without entities
1679 public static function decodeCharReferencesAndNormalize( $text ) {
1680 $text = preg_replace_callback(
1681 self
::CHAR_REFS_REGEX
,
1682 [ self
::class, 'decodeCharReferencesCallback' ],
1689 return MediaWikiServices
::getInstance()->getContentLanguage()->normalize( $text );
1696 * @param string $matches
1699 static function decodeCharReferencesCallback( $matches ) {
1700 if ( $matches[1] != '' ) {
1701 return self
::decodeEntity( $matches[1] );
1702 } elseif ( $matches[2] != '' ) {
1703 return self
::decodeChar( intval( $matches[2] ) );
1704 } elseif ( $matches[3] != '' ) {
1705 return self
::decodeChar( hexdec( $matches[3] ) );
1707 # Last case should be an ampersand by itself
1712 * Return UTF-8 string for a codepoint if that is a valid
1713 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1714 * @param int $codepoint
1718 static function decodeChar( $codepoint ) {
1719 if ( self
::validateCodepoint( $codepoint ) ) {
1720 return UtfNormal\Utils
::codepointToUtf8( $codepoint );
1722 return UtfNormal\Constants
::UTF8_REPLACEMENT
;
1727 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1728 * return the UTF-8 encoding of that character. Otherwise, returns
1729 * pseudo-entity source (eg "&foo;")
1731 * @param string $name
1734 static function decodeEntity( $name ) {
1735 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1736 $name = self
::$htmlEntityAliases[$name];
1738 if ( isset( self
::$htmlEntities[$name] ) ) {
1739 return UtfNormal\Utils
::codepointToUtf8( self
::$htmlEntities[$name] );
1746 * Fetch the whitelist of acceptable attributes for a given element name.
1748 * @param string $element
1751 static function attributeWhitelist( $element ) {
1752 $list = self
::setupAttributeWhitelist();
1753 return $list[$element] ??
[];
1757 * Foreach array key (an allowed HTML element), return an array
1758 * of allowed attributes
1761 static function setupAttributeWhitelist() {
1764 if ( $whitelist !== null ) {
1786 # These attributes are specified in section 9 of
1787 # https://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1794 # Microdata. These are specified by
1795 # https://html.spec.whatwg.org/multipage/microdata.html#the-microdata-model
1803 $block = array_merge( $common, [ 'align' ] );
1804 $tablealign = [ 'align', 'valign' ];
1812 'nowrap', # deprecated
1813 'width', # deprecated
1814 'height', # deprecated
1815 'bgcolor', # deprecated
1818 # Numbers refer to sections in HTML 4.01 standard describing the element.
1819 # See: https://www.w3.org/TR/html4/
1823 'center' => $common, # deprecated
1842 'strong' => $common,
1853 'blockquote' => array_merge( $common, [ 'cite' ] ),
1854 'q' => array_merge( $common, [ 'cite' ] ),
1864 'br' => array_merge( $common, [ 'clear' ] ),
1866 # https://www.w3.org/TR/html5/text-level-semantics.html#the-wbr-element
1870 'pre' => array_merge( $common, [ 'width' ] ),
1873 'ins' => array_merge( $common, [ 'cite', 'datetime' ] ),
1874 'del' => array_merge( $common, [ 'cite', 'datetime' ] ),
1877 'ul' => array_merge( $common, [ 'type' ] ),
1878 'ol' => array_merge( $common, [ 'type', 'start', 'reversed' ] ),
1879 'li' => array_merge( $common, [ 'type', 'value' ] ),
1887 'table' => array_merge( $common,
1888 [ 'summary', 'width', 'border', 'frame',
1889 'rules', 'cellspacing', 'cellpadding',
1894 'caption' => $block,
1902 'colgroup' => array_merge( $common, [ 'span' ] ),
1903 'col' => array_merge( $common, [ 'span' ] ),
1906 'tr' => array_merge( $common, [ 'bgcolor' ], $tablealign ),
1909 'td' => array_merge( $common, $tablecell, $tablealign ),
1910 'th' => array_merge( $common, $tablecell, $tablealign ),
1913 # NOTE: <a> is not allowed directly, but the attrib
1914 # whitelist is used from the Parser object
1915 'a' => array_merge( $common, [ 'href', 'rel', 'rev' ] ), # rel/rev esp. for RDFa
1918 # Not usually allowed, but may be used for extension-style hooks
1919 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1921 'img' => array_merge( $common, [ 'alt', 'src', 'width', 'height', 'srcset' ] ),
1922 # Attributes for A/V tags added in T163583 / T133673
1923 'audio' => array_merge( $common, [ 'controls', 'preload', 'width', 'height' ] ),
1924 'video' => array_merge( $common, [ 'poster', 'controls', 'preload', 'width', 'height' ] ),
1925 'source' => array_merge( $common, [ 'type', 'src' ] ),
1926 'track' => array_merge( $common, [ 'type', 'src', 'srclang', 'kind', 'label' ] ),
1934 'strike' => $common,
1939 'font' => array_merge( $common, [ 'size', 'color', 'face' ] ),
1943 'hr' => array_merge( $common, [ 'width' ] ),
1945 # HTML Ruby annotation text module, simple ruby only.
1946 # https://www.w3.org/TR/html5/text-level-semantics.html#the-ruby-element
1951 'rt' => $common, # array_merge( $common, array( 'rbspan' ) ),
1954 # MathML root element, where used for extensions
1955 # 'title' may not be 100% valid here; it's XHTML
1956 # https://www.w3.org/TR/REC-MathML/
1957 'math' => [ 'class', 'style', 'id', 'title' ],
1959 // HTML 5 section 4.5
1960 'figure' => $common,
1961 'figure-inline' => $common, # T118520
1962 'figcaption' => $common,
1964 # HTML 5 section 4.6
1967 # HTML5 elements, defined by:
1968 # https://html.spec.whatwg.org/multipage/semantics.html#the-data-element
1969 'data' => array_merge( $common, [ 'value' ] ),
1970 'time' => array_merge( $common, [ 'datetime' ] ),
1973 // meta and link are only permitted by removeHTMLtags when Microdata
1974 // is enabled so we don't bother adding a conditional to hide these
1975 // Also meta and link are only valid in WikiText as Microdata elements
1976 // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1977 // So we don't bother including $common attributes that have no purpose.
1978 'meta' => [ 'itemprop', 'content' ],
1979 'link' => [ 'itemprop', 'href', 'title' ],
1986 * Take a fragment of (potentially invalid) HTML and return
1987 * a version with any tags removed, encoded as plain text.
1989 * Warning: this return value must be further escaped for literal
1990 * inclusion in HTML output as of 1.10!
1992 * @param string $html HTML fragment
1995 static function stripAllTags( $html ) {
1996 // Use RemexHtml to tokenize $html and extract the text
1997 $handler = new RemexStripTagHandler
;
1998 $tokenizer = new RemexHtml\Tokenizer\
Tokenizer( $handler, $html, [
1999 'ignoreErrors' => true,
2000 // don't ignore char refs, we want them to be decoded
2001 'ignoreNulls' => true,
2002 'skipPreprocess' => true,
2004 $tokenizer->execute();
2005 $text = $handler->getResult();
2007 $text = self
::normalizeWhitespace( $text );
2012 * Hack up a private DOCTYPE with HTML's standard entity declarations.
2013 * PHP 4 seemed to know these if you gave it an HTML doctype, but
2016 * Use for passing XHTML fragments to PHP's XML parsing functions
2020 static function hackDocType() {
2021 $out = "<!DOCTYPE html [\n";
2022 foreach ( self
::$htmlEntities as $entity => $codepoint ) {
2023 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
2030 * @param string $url
2031 * @return mixed|string
2033 static function cleanUrl( $url ) {
2034 # Normalize any HTML entities in input. They will be
2035 # re-escaped by makeExternalLink().
2036 $url = self
::decodeCharReferences( $url );
2038 # Escape any control characters introduced by the above step
2039 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
2040 [ __CLASS__
, 'cleanUrlCallback' ], $url );
2042 # Validate hostname portion
2044 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
2045 list( /* $whole */, $protocol, $host, $rest ) = $matches;
2047 // Characters that will be ignored in IDNs.
2048 // https://tools.ietf.org/html/rfc3454#section-3.1
2049 // Strip them before further processing so blacklists and such work.
2051 \\s| # general whitespace
2052 \xc2\xad| # 00ad SOFT HYPHEN
2053 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
2054 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
2055 \xe2\x81\xa0| # 2060 WORD JOINER
2056 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
2057 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
2058 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
2059 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
2060 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
2061 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
2062 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
2063 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
2066 $host = preg_replace( $strip, '', $host );
2068 // IPv6 host names are bracketed with []. Url-decode these.
2069 if ( substr_compare( "//%5B", $host, 0, 5 ) === 0 &&
2070 preg_match( '!^//%5B([0-9A-Fa-f:.]+)%5D((:\d+)?)$!', $host, $matches )
2072 $host = '//[' . $matches[1] . ']' . $matches[2];
2075 // @todo FIXME: Validate hostnames here
2077 return $protocol . $host . $rest;
2084 * @param array $matches
2087 static function cleanUrlCallback( $matches ) {
2088 return urlencode( $matches[0] );
2092 * Does a string look like an e-mail address?
2094 * This validates an email address using an HTML5 specification found at:
2095 * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
2096 * Which as of 2011-01-24 says:
2098 * A valid e-mail address is a string that matches the ABNF production
2099 * 1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
2100 * in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
2103 * This function is an implementation of the specification as requested in
2106 * Client-side forms will use the same standard validation rules via JS or
2107 * HTML 5 validation; additional restrictions can be enforced server-side
2108 * by extensions via the 'isValidEmailAddr' hook.
2110 * Note that this validation doesn't 100% match RFC 2822, but is believed
2111 * to be liberal enough for wide use. Some invalid addresses will still
2112 * pass validation here.
2116 * @param string $addr E-mail address
2119 public static function validateEmail( $addr ) {
2121 if ( !Hooks
::run( 'isValidEmailAddr', [ $addr, &$result ] ) ) {
2125 // Please note strings below are enclosed in brackets [], this make the
2126 // hyphen "-" a range indicator. Hence it is double backslashed below.
2128 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
2129 $rfc1034_ldh_str = "a-z0-9\\-";
2131 $html5_email_regexp = "/
2133 [$rfc5322_atext\\.]+ # user part which is liberal :p
2135 [$rfc1034_ldh_str]+ # First domain part
2136 (\\.[$rfc1034_ldh_str]+)* # Following part prefixed with a dot
2138 /ix"; // case Insensitive, eXtended
2140 return (bool)preg_match( $html5_email_regexp, $addr );