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
::attributeWhitelistInternal( $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,
806 * either as a sequential array of valid attribute names or
807 * as an associative array where keys give valid attribute names
810 * @todo Check for legal values where the DTD limits things.
811 * @todo Check for unique id attribute :P
813 static function validateAttributes( $attribs, $whitelist ) {
814 if ( isset( $whitelist[0] ) ) {
815 // We would like to eventually deprecate calling this
816 // function with a sequential array, but for now just
818 $whitelist = array_flip( $whitelist );
820 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
823 foreach ( $attribs as $attribute => $value ) {
824 # Allow XML namespace declaration to allow RDFa
825 if ( preg_match( self
::XMLNS_ATTRIBUTE_PATTERN
, $attribute ) ) {
826 if ( !preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
827 $out[$attribute] = $value;
833 # Allow any attribute beginning with "data-"
835 # * Disallow data attributes used by MediaWiki code
836 # * Ensure that the attribute is not namespaced by banning
839 !preg_match( '/^data-[^:]*$/i', $attribute ) &&
840 !array_key_exists( $attribute, $whitelist )
841 ) || self
::isReservedDataAttribute( $attribute ) ) {
845 # Strip javascript "expression" from stylesheets.
846 # https://msdn.microsoft.com/en-us/library/ms537634.aspx
847 if ( $attribute == 'style' ) {
848 $value = self
::checkCss( $value );
851 # Escape HTML id attributes
852 if ( $attribute === 'id' ) {
853 $value = self
::escapeIdForAttribute( $value, self
::ID_PRIMARY
);
856 # Escape HTML id reference lists
857 if ( $attribute === 'aria-describedby'
858 ||
$attribute === 'aria-flowto'
859 ||
$attribute === 'aria-labelledby'
860 ||
$attribute === 'aria-owns'
862 $value = self
::escapeIdReferenceList( $value );
865 // RDFa and microdata properties allow URLs, URIs and/or CURIs.
866 // Check them for sanity.
867 if ( $attribute === 'rel' ||
$attribute === 'rev'
869 ||
$attribute === 'about' ||
$attribute === 'property'
870 ||
$attribute === 'resource' ||
$attribute === 'datatype'
871 ||
$attribute === 'typeof'
873 ||
$attribute === 'itemid' ||
$attribute === 'itemprop'
874 ||
$attribute === 'itemref' ||
$attribute === 'itemscope'
875 ||
$attribute === 'itemtype'
877 // Paranoia. Allow "simple" values but suppress javascript
878 if ( preg_match( self
::EVIL_URI_PATTERN
, $value ) ) {
883 # NOTE: even though elements using href/src are not allowed directly, supply
884 # validation code that can be used by tag hook handlers, etc
885 if ( $attribute === 'href' ||
$attribute === 'src' ||
$attribute === 'poster' ) {
886 if ( !preg_match( $hrefExp, $value ) ) {
887 continue; // drop any href or src attributes not using an allowed protocol.
888 // NOTE: this also drops all relative URLs
892 // If this attribute was previously set, override it.
893 // Output should only have one attribute of each name.
894 $out[$attribute] = $value;
897 # itemtype, itemid, itemref don't make sense without itemscope
898 if ( !array_key_exists( 'itemscope', $out ) ) {
899 unset( $out['itemtype'] );
900 unset( $out['itemid'] );
901 unset( $out['itemref'] );
903 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
909 * Given an attribute name, checks whether it is a reserved data attribute
910 * (such as data-mw-foo) which is unavailable to user-generated HTML so MediaWiki
911 * core and extension code can safely use it to communicate with frontend code.
912 * @param string $attr Attribute name.
915 public static function isReservedDataAttribute( $attr ) {
916 // data-ooui is reserved for ooui.
917 // data-mw and data-parsoid are reserved for parsoid.
918 // data-mw-<name here> is reserved for extensions (or core) if
919 // they need to communicate some data to the client and want to be
920 // sure that it isn't coming from an untrusted user.
921 // We ignore the possibility of namespaces since user-generated HTML
922 // can't use them anymore.
923 return (bool)preg_match( '/^data-(ooui|mw|parsoid)/i', $attr );
927 * Merge two sets of HTML attributes. Conflicting items in the second set
928 * will override those in the first, except for 'class' attributes which
929 * will be combined (if they're both strings).
931 * @todo implement merging for other attributes such as style
936 static function mergeAttributes( $a, $b ) {
937 $out = array_merge( $a, $b );
938 if ( isset( $a['class'] ) && isset( $b['class'] )
939 && is_string( $a['class'] ) && is_string( $b['class'] )
940 && $a['class'] !== $b['class']
942 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
943 -1, PREG_SPLIT_NO_EMPTY
);
944 $out['class'] = implode( ' ', array_unique( $classes ) );
950 * Normalize CSS into a format we can easily search for hostile input
951 * - decode character references
952 * - decode escape sequences
953 * - convert characters that IE6 interprets into ascii
954 * - remove comments, unless the entire value is one single comment
955 * @param string $value the css string
956 * @return string normalized css
958 public static function normalizeCss( $value ) {
959 // Decode character references like {
960 $value = self
::decodeCharReferences( $value );
962 // Decode escape sequences and line continuation
963 // See the grammar in the CSS 2 spec, appendix D.
964 // This has to be done AFTER decoding character references.
965 // This means it isn't possible for this function to return
966 // unsanitized escape sequences. It is possible to manufacture
967 // input that contains character references that decode to
968 // escape sequences that decode to character references, but
969 // it's OK for the return value to contain character references
970 // because the caller is supposed to escape those anyway.
972 if ( !$decodeRegex ) {
973 $space = '[\\x20\\t\\r\\n\\f]';
974 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
976 $decodeRegex = "/ $backslash
978 ($nl) | # 1. Line continuation
979 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
980 (.) | # 3. backslash cancelling special meaning
981 () | # 4. backslash at end of string
984 $value = preg_replace_callback( $decodeRegex,
985 [ __CLASS__
, 'cssDecodeCallback' ], $value );
987 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
988 $value = preg_replace_callback(
989 '/[!-[]-z]/u', // U+FF01 to U+FF5A, excluding U+FF3C (T60088)
990 function ( $matches ) {
991 $cp = UtfNormal\Utils
::utf8ToCodepoint( $matches[0] );
992 if ( $cp === false ) {
995 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
1000 // Convert more characters IE6 might treat as ascii
1001 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
1002 $value = str_replace(
1003 [ 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ],
1004 [ 'r', 'n', 'n', 'l', 'i', '(', '(' ],
1008 // Let the value through if it's nothing but a single comment, to
1009 // allow other functions which may reject it to pass some error
1011 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
1012 // Remove any comments; IE gets token splitting wrong
1013 // This must be done AFTER decoding character references and
1014 // escape sequences, because those steps can introduce comments
1015 // This step cannot introduce character references or escape
1016 // sequences, because it replaces comments with spaces rather
1017 // than removing them completely.
1018 $value = StringUtils
::delimiterReplace( '/*', '*/', ' ', $value );
1020 // Remove anything after a comment-start token, to guard against
1021 // incorrect client implementations.
1022 $commentPos = strpos( $value, '/*' );
1023 if ( $commentPos !== false ) {
1024 $value = substr( $value, 0, $commentPos );
1028 // S followed by repeat, iteration, or prolonged sound marks,
1029 // which IE will treat as "ss"
1030 $value = preg_replace(
1032 \xE3\x80\xB1 | # U+3031
1033 \xE3\x82\x9D | # U+309D
1034 \xE3\x83\xBC | # U+30FC
1035 \xE3\x83\xBD | # U+30FD
1036 \xEF\xB9\xBC | # U+FE7C
1037 \xEF\xB9\xBD | # U+FE7D
1038 \xEF\xBD\xB0 # U+FF70
1048 * Pick apart some CSS and check it for forbidden or unsafe structures.
1049 * Returns a sanitized string. This sanitized string will have
1050 * character references and escape sequences decoded and comments
1051 * stripped (unless it is itself one valid comment, in which case the value
1052 * will be passed through). If the input is just too evil, only a comment
1053 * complaining about evilness will be returned.
1055 * Currently URL references, 'expression', 'tps' are forbidden.
1057 * NOTE: Despite the fact that character references are decoded, the
1058 * returned string may contain character references given certain
1059 * clever input strings. These character references must
1060 * be escaped before the return value is embedded in HTML.
1062 * @param string $value
1065 static function checkCss( $value ) {
1066 $value = self
::normalizeCss( $value );
1068 // Reject problematic keywords and control characters
1069 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ||
1070 strpos( $value, UtfNormal\Constants
::UTF8_REPLACEMENT
) !== false ) {
1071 return '/* invalid control char */';
1072 } elseif ( preg_match(
1077 | -o-link-source\s*:
1082 | attr\s*\([^)]+[\s,]+url
1085 return '/* insecure input */';
1091 * @param array $matches
1094 static function cssDecodeCallback( $matches ) {
1095 if ( $matches[1] !== '' ) {
1096 // Line continuation
1098 } elseif ( $matches[2] !== '' ) {
1099 $char = UtfNormal\Utils
::codepointToUtf8( hexdec( $matches[2] ) );
1100 } elseif ( $matches[3] !== '' ) {
1101 $char = $matches[3];
1105 if ( $char == "\n" ||
$char == '"' ||
$char == "'" ||
$char == '\\' ) {
1106 // These characters need to be escaped in strings
1107 // Clean up the escape sequence to avoid parsing errors by clients
1108 return '\\' . dechex( ord( $char ) ) . ' ';
1110 // Decode unnecessary escape
1116 * Take a tag soup fragment listing an HTML element's attributes
1117 * and normalize it to well-formed XML, discarding unwanted attributes.
1118 * Output is safe for further wikitext processing, with escaping of
1119 * values that could trigger problems.
1121 * - Normalizes attribute names to lowercase
1122 * - Discards attributes not on a whitelist for the given element
1123 * - Turns broken or invalid entities into plaintext
1124 * - Double-quotes all attribute values
1125 * - Attributes without values are given the name as attribute
1126 * - Double attributes are discarded
1127 * - Unsafe style attributes are discarded
1128 * - Prepends space if there are attributes.
1129 * - (Optionally) Sorts attributes by name.
1131 * @param string $text
1132 * @param string $element
1133 * @param bool $sorted Whether to sort the attributes (default: false)
1136 static function fixTagAttributes( $text, $element, $sorted = false ) {
1137 if ( trim( $text ) == '' ) {
1141 $decoded = self
::decodeTagAttributes( $text );
1142 $stripped = self
::validateTagAttributes( $decoded, $element );
1148 return self
::safeEncodeTagAttributes( $stripped );
1152 * Encode an attribute value for HTML output.
1153 * @param string $text
1154 * @return string HTML-encoded text fragment
1156 static function encodeAttribute( $text ) {
1157 $encValue = htmlspecialchars( $text, ENT_QUOTES
);
1159 // Whitespace is normalized during attribute decoding,
1160 // so if we've been passed non-spaces we must encode them
1161 // ahead of time or they won't be preserved.
1162 $encValue = strtr( $encValue, [
1172 * Armor French spaces with a replacement character
1175 * @param string $text Text to armor
1176 * @param string $space Space character for the French spaces, defaults to ' '
1177 * @return string Armored text
1179 public static function armorFrenchSpaces( $text, $space = ' ' ) {
1180 // Replace $ with \$ and \ with \\
1181 $space = preg_replace( '#(?<!\\\\)(\\$|\\\\)#', '\\\\$1', $space );
1183 # French spaces, last one Guillemet-left
1184 # only if there is something before the space
1185 # and a non-word character after the punctuation.
1186 '/(\S) (?=[?:;!%»›](?!\w))/u' => "\\1$space",
1187 # French spaces, Guillemet-right
1188 '/([«‹]) /u' => "\\1$space",
1190 return preg_replace( array_keys( $fixtags ), array_values( $fixtags ), $text );
1194 * Encode an attribute value for HTML tags, with extra armoring
1195 * against further wiki processing.
1196 * @param string $text
1197 * @return string HTML-encoded text fragment
1199 static function safeEncodeAttribute( $text ) {
1200 $encValue = self
::encodeAttribute( $text );
1202 # Templates and links may be expanded in later parsing,
1203 # creating invalid or dangerous output. Suppress this.
1204 $encValue = strtr( $encValue, [
1205 '<' => '<', // This should never happen,
1206 '>' => '>', // we've received invalid input
1207 '"' => '"', // which should have been escaped.
1209 '}' => '}', // prevent unpaired language conversion syntax
1212 "''" => '''',
1213 'ISBN' => 'ISBN',
1215 'PMID' => 'PMID',
1220 # Armor against French spaces detection (T5158)
1221 $encValue = self
::armorFrenchSpaces( $encValue, ' ' );
1224 $encValue = preg_replace_callback(
1225 '/((?i)' . wfUrlProtocols() . ')/',
1226 function ( $matches ) {
1227 return str_replace( ':', ':', $matches[1] );
1234 * Given a value, escape it so that it can be used in an id attribute and
1235 * return it. This will use HTML5 validation, allowing anything but ASCII
1238 * To ensure we don't have to bother escaping anything, we also strip ', ".
1239 * TODO: Is this the best tactic?
1241 * We also strip # because it upsets IE, and % because it could be
1242 * ambiguous if it's part of something that looks like a percent escape
1243 * (which don't work reliably in fragments cross-browser).
1245 * @deprecated since 1.30, use one of this class' escapeIdFor*() functions
1247 * @see https://www.w3.org/TR/html401/types.html#type-name Valid characters
1248 * in the id and name attributes
1249 * @see https://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1251 * @see https://www.w3.org/TR/html5/dom.html#the-id-attribute
1252 * HTML5 definition of id attribute
1254 * @param string $id Id to escape
1255 * @param string|array $options String or array of strings (default is []):
1256 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1257 * so don't pay attention if the first character isn't valid at the
1258 * beginning of an id.
1261 static function escapeId( $id, $options = [] ) {
1262 $options = (array)$options;
1264 // HTML4-style escaping
1270 $id = urlencode( strtr( $id, ' ', '_' ) );
1271 $id = strtr( $id, $replace );
1273 if ( !preg_match( '/^[a-zA-Z]/', $id ) && !in_array( 'noninitial', $options ) ) {
1274 // Initial character must be a letter!
1281 * Given a section name or other user-generated or otherwise unsafe string, escapes it to be
1282 * a valid HTML id attribute.
1284 * WARNING: unlike escapeId(), the output of this function is not guaranteed to be HTML safe,
1285 * be sure to use proper escaping.
1287 * @param string $id String to escape
1288 * @param int $mode One of ID_* constants, specifying whether the primary or fallback encoding
1290 * @return string|bool Escaped ID or false if fallback encoding is requested but it's not
1295 public static function escapeIdForAttribute( $id, $mode = self
::ID_PRIMARY
) {
1296 global $wgFragmentMode;
1298 if ( !isset( $wgFragmentMode[$mode] ) ) {
1299 if ( $mode === self
::ID_PRIMARY
) {
1300 throw new UnexpectedValueException( '$wgFragmentMode is configured with no primary mode' );
1305 $internalMode = $wgFragmentMode[$mode];
1307 return self
::escapeIdInternal( $id, $internalMode );
1311 * Given a section name or other user-generated or otherwise unsafe string, escapes it to be
1312 * a valid URL fragment.
1314 * WARNING: unlike escapeId(), the output of this function is not guaranteed to be HTML safe,
1315 * be sure to use proper escaping.
1317 * @param string $id String to escape
1318 * @return string Escaped ID
1322 public static function escapeIdForLink( $id ) {
1323 global $wgFragmentMode;
1325 if ( !isset( $wgFragmentMode[self
::ID_PRIMARY
] ) ) {
1326 throw new UnexpectedValueException( '$wgFragmentMode is configured with no primary mode' );
1329 $mode = $wgFragmentMode[self
::ID_PRIMARY
];
1331 $id = self
::escapeIdInternal( $id, $mode );
1337 * Given a section name or other user-generated or otherwise unsafe string, escapes it to be
1338 * a valid URL fragment for external interwikis.
1340 * @param string $id String to escape
1341 * @return string Escaped ID
1345 public static function escapeIdForExternalInterwiki( $id ) {
1346 global $wgExternalInterwikiFragmentMode;
1348 $id = self
::escapeIdInternal( $id, $wgExternalInterwikiFragmentMode );
1354 * Helper for escapeIdFor*() functions. Performs most of the actual escaping.
1356 * @param string $id String to escape
1357 * @param string $mode One of modes from $wgFragmentMode
1360 private static function escapeIdInternal( $id, $mode ) {
1363 $id = str_replace( ' ', '_', $id );
1366 // This corresponds to 'noninitial' mode of the old escapeId()
1372 $id = urlencode( str_replace( ' ', '_', $id ) );
1373 $id = strtr( $id, $replace );
1376 throw new InvalidArgumentException( "Invalid mode '$mode' passed to '" . __METHOD__
);
1383 * Given a string containing a space delimited list of ids, escape each id
1384 * to match ids escaped by the escapeIdForAttribute() function.
1388 * @param string $referenceString Space delimited list of ids
1391 public static function escapeIdReferenceList( $referenceString ) {
1392 # Explode the space delimited list string into an array of tokens
1393 $references = preg_split( '/\s+/', "{$referenceString}", -1, PREG_SPLIT_NO_EMPTY
);
1395 # Escape each token as an id
1396 foreach ( $references as &$ref ) {
1397 $ref = self
::escapeIdForAttribute( $ref );
1400 # Merge the array back to a space delimited list string
1401 # If the array is empty, the result will be an empty string ('')
1402 $referenceString = implode( ' ', $references );
1404 return $referenceString;
1408 * Given a value, escape it so that it can be used as a CSS class and
1411 * @todo For extra validity, input should be validated UTF-8.
1413 * @see https://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1415 * @param string $class
1418 static function escapeClass( $class ) {
1419 // Convert ugly stuff to underscores and kill underscores in ugly places
1420 return rtrim( preg_replace(
1421 [ '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ],
1427 * Given HTML input, escape with htmlspecialchars but un-escape entities.
1428 * This allows (generally harmless) entities like   to survive.
1430 * @param string $html HTML to escape
1431 * @return string Escaped input
1433 static function escapeHtmlAllowEntities( $html ) {
1434 $html = self
::decodeCharReferences( $html );
1435 # It seems wise to escape ' as well as ", as a matter of course. Can't
1436 # hurt. Use ENT_SUBSTITUTE so that incorrectly truncated multibyte characters
1437 # don't cause the entire string to disappear.
1438 $html = htmlspecialchars( $html, ENT_QUOTES | ENT_SUBSTITUTE
);
1443 * Return an associative array of attribute names and values from
1444 * a partial tag string. Attribute names are forced to lowercase,
1445 * character references are decoded to UTF-8 text.
1447 * @param string $text
1450 public static function decodeTagAttributes( $text ) {
1451 if ( trim( $text ) == '' ) {
1456 if ( !preg_match_all(
1457 self
::getAttribsRegex(),
1460 PREG_SET_ORDER
) ) {
1465 foreach ( $pairs as $set ) {
1466 $attribute = strtolower( $set[1] );
1468 // Filter attribute names with unacceptable characters
1469 if ( !preg_match( self
::getAttribNameRegex(), $attribute ) ) {
1473 $value = self
::getTagAttributeCallback( $set );
1475 // Normalize whitespace
1476 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1477 $value = trim( $value );
1479 // Decode character references
1480 $attribs[$attribute] = self
::decodeCharReferences( $value );
1486 * Build a partial tag string from an associative array of attribute
1487 * names and values as returned by decodeTagAttributes.
1489 * @param array $assoc_array
1492 public static function safeEncodeTagAttributes( $assoc_array ) {
1494 foreach ( $assoc_array as $attribute => $value ) {
1495 $encAttribute = htmlspecialchars( $attribute );
1496 $encValue = self
::safeEncodeAttribute( $value );
1498 $attribs[] = "$encAttribute=\"$encValue\"";
1500 return count( $attribs ) ?
' ' . implode( ' ', $attribs ) : '';
1504 * Pick the appropriate attribute value from a match set from the
1505 * attribs regex matches.
1508 * @throws MWException When tag conditions are not met.
1511 private static function getTagAttributeCallback( $set ) {
1512 if ( isset( $set[5] ) ) {
1515 } elseif ( isset( $set[4] ) ) {
1518 } elseif ( isset( $set[3] ) ) {
1521 } elseif ( !isset( $set[2] ) ) {
1522 # In XHTML, attributes must have a value so return an empty string.
1523 # See "Empty attribute syntax",
1524 # https://www.w3.org/TR/html5/syntax.html#syntax-attribute-name
1527 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1532 * @param string $text
1535 private static function normalizeWhitespace( $text ) {
1536 return trim( preg_replace(
1537 '/(?:\r\n|[\x20\x0d\x0a\x09])+/',
1543 * Normalizes whitespace in a section name, such as might be returned
1544 * by Parser::stripSectionName(), for use in the id's that are used for
1547 * @param string $section
1550 static function normalizeSectionNameWhitespace( $section ) {
1551 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1555 * Ensure that any entities and character references are legal
1556 * for XML and XHTML specifically. Any stray bits will be
1557 * &-escaped to result in a valid text fragment.
1559 * a. named char refs can only be < > & ", others are
1560 * numericized (this way we're well-formed even without a DTD)
1561 * b. any numeric char refs must be legal chars, not invalid or forbidden
1562 * c. use lower cased "&#x", not "&#X"
1563 * d. fix or reject non-valid attributes
1565 * @param string $text
1569 static function normalizeCharReferences( $text ) {
1570 return preg_replace_callback(
1571 self
::CHAR_REFS_REGEX
,
1572 [ self
::class, 'normalizeCharReferencesCallback' ],
1577 * @param string $matches
1580 static function normalizeCharReferencesCallback( $matches ) {
1582 if ( $matches[1] != '' ) {
1583 $ret = self
::normalizeEntity( $matches[1] );
1584 } elseif ( $matches[2] != '' ) {
1585 $ret = self
::decCharReference( $matches[2] );
1586 } elseif ( $matches[3] != '' ) {
1587 $ret = self
::hexCharReference( $matches[3] );
1589 if ( is_null( $ret ) ) {
1590 return htmlspecialchars( $matches[0] );
1597 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1598 * return the equivalent numeric entity reference (except for the core <
1599 * > & "). If the entity is a MediaWiki-specific alias, returns
1600 * the HTML equivalent. Otherwise, returns HTML-escaped text of
1601 * pseudo-entity source (eg &foo;)
1603 * @param string $name
1606 static function normalizeEntity( $name ) {
1607 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1608 return '&' . self
::$htmlEntityAliases[$name] . ';';
1609 } elseif ( in_array( $name, [ 'lt', 'gt', 'amp', 'quot' ] ) ) {
1611 } elseif ( isset( self
::$htmlEntities[$name] ) ) {
1612 return '&#' . self
::$htmlEntities[$name] . ';';
1614 return "&$name;";
1619 * @param int $codepoint
1620 * @return null|string
1622 static function decCharReference( $codepoint ) {
1623 $point = intval( $codepoint );
1624 if ( self
::validateCodepoint( $point ) ) {
1625 return sprintf( '&#%d;', $point );
1632 * @param int $codepoint
1633 * @return null|string
1635 static function hexCharReference( $codepoint ) {
1636 $point = hexdec( $codepoint );
1637 if ( self
::validateCodepoint( $point ) ) {
1638 return sprintf( '&#x%x;', $point );
1645 * Returns true if a given Unicode codepoint is a valid character in
1646 * both HTML5 and XML.
1647 * @param int $codepoint
1650 private static function validateCodepoint( $codepoint ) {
1651 # U+000C is valid in HTML5 but not allowed in XML.
1652 # U+000D is valid in XML but not allowed in HTML5.
1653 # U+007F - U+009F are disallowed in HTML5 (control characters).
1654 return $codepoint == 0x09
1655 ||
$codepoint == 0x0a
1656 ||
( $codepoint >= 0x20 && $codepoint <= 0x7e )
1657 ||
( $codepoint >= 0xa0 && $codepoint <= 0xd7ff )
1658 ||
( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1659 ||
( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1663 * Decode any character references, numeric or named entities,
1664 * in the text and return a UTF-8 string.
1666 * @param string $text
1669 public static function decodeCharReferences( $text ) {
1670 return preg_replace_callback(
1671 self
::CHAR_REFS_REGEX
,
1672 [ self
::class, 'decodeCharReferencesCallback' ],
1677 * Decode any character references, numeric or named entities,
1678 * in the next and normalize the resulting string. (T16952)
1680 * This is useful for page titles, not for text to be displayed,
1681 * MediaWiki allows HTML entities to escape normalization as a feature.
1683 * @param string $text Already normalized, containing entities
1684 * @return string Still normalized, without entities
1686 public static function decodeCharReferencesAndNormalize( $text ) {
1687 $text = preg_replace_callback(
1688 self
::CHAR_REFS_REGEX
,
1689 [ self
::class, 'decodeCharReferencesCallback' ],
1696 return MediaWikiServices
::getInstance()->getContentLanguage()->normalize( $text );
1703 * @param string $matches
1706 static function decodeCharReferencesCallback( $matches ) {
1707 if ( $matches[1] != '' ) {
1708 return self
::decodeEntity( $matches[1] );
1709 } elseif ( $matches[2] != '' ) {
1710 return self
::decodeChar( intval( $matches[2] ) );
1711 } elseif ( $matches[3] != '' ) {
1712 return self
::decodeChar( hexdec( $matches[3] ) );
1714 # Last case should be an ampersand by itself
1719 * Return UTF-8 string for a codepoint if that is a valid
1720 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1721 * @param int $codepoint
1725 static function decodeChar( $codepoint ) {
1726 if ( self
::validateCodepoint( $codepoint ) ) {
1727 return UtfNormal\Utils
::codepointToUtf8( $codepoint );
1729 return UtfNormal\Constants
::UTF8_REPLACEMENT
;
1734 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1735 * return the UTF-8 encoding of that character. Otherwise, returns
1736 * pseudo-entity source (eg "&foo;")
1738 * @param string $name
1741 static function decodeEntity( $name ) {
1742 if ( isset( self
::$htmlEntityAliases[$name] ) ) {
1743 $name = self
::$htmlEntityAliases[$name];
1745 if ( isset( self
::$htmlEntities[$name] ) ) {
1746 return UtfNormal\Utils
::codepointToUtf8( self
::$htmlEntities[$name] );
1753 * Fetch the whitelist of acceptable attributes for a given element name.
1755 * @param string $element
1756 * @return array A sequential array of acceptable attribute names
1757 * @deprecated since 1.34; should be private
1759 static function attributeWhitelist( $element ) {
1760 wfDeprecated( __METHOD__
, '1.34' );
1761 $list = self
::setupAttributeWhitelist();
1762 return $list[$element] ??
[];
1766 * Fetch the whitelist of acceptable attributes for a given element name.
1768 * @param string $element
1769 * @return array An associative array where keys are acceptable attribute
1772 private static function attributeWhitelistInternal( $element ) {
1773 $list = self
::setupAttributeWhitelistInternal();
1774 return $list[$element] ??
[];
1778 * Foreach array key (an allowed HTML element), return an array
1779 * of allowed attributes
1781 * @deprecated since 1.34; should be private
1783 static function setupAttributeWhitelist() {
1784 wfDeprecated( __METHOD__
, '1.34' );
1785 $wlist = self
::setupAttributeWhitelistInternal();
1786 // This method is expected to return a sequential array as the
1787 // value for each HTML element key.
1788 return array_map( function ( $v ) {
1789 return array_keys( $v );
1794 * Foreach array key (an allowed HTML element), return an array
1795 * of allowed attributes
1796 * @return array An associative array: keys are HTML element names;
1797 * values are associative arrays where the keys are allowed attribute
1800 private static function setupAttributeWhitelistInternal() {
1803 if ( $whitelist !== null ) {
1807 // For lookup efficiency flip each attributes array so the keys are
1808 // the valid attributes.
1809 $merge = function ( $a, $b, $c = [] ) {
1810 return array_merge( $a, array_flip( $b ), array_flip( $c ) );
1812 $common = $merge( [], [
1830 # These attributes are specified in section 9 of
1831 # https://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1838 # Microdata. These are specified by
1839 # https://html.spec.whatwg.org/multipage/microdata.html#the-microdata-model
1847 $block = $merge( $common, [ 'align' ] );
1849 $tablealign = [ 'align', 'valign' ];
1857 'nowrap', # deprecated
1858 'width', # deprecated
1859 'height', # deprecated
1860 'bgcolor', # deprecated
1863 # Numbers refer to sections in HTML 4.01 standard describing the element.
1864 # See: https://www.w3.org/TR/html4/
1868 'center' => $common, # deprecated
1887 'strong' => $common,
1898 'blockquote' => $merge( $common, [ 'cite' ] ),
1899 'q' => $merge( $common, [ 'cite' ] ),
1909 'br' => $merge( $common, [ 'clear' ] ),
1911 # https://www.w3.org/TR/html5/text-level-semantics.html#the-wbr-element
1915 'pre' => $merge( $common, [ 'width' ] ),
1918 'ins' => $merge( $common, [ 'cite', 'datetime' ] ),
1919 'del' => $merge( $common, [ 'cite', 'datetime' ] ),
1922 'ul' => $merge( $common, [ 'type' ] ),
1923 'ol' => $merge( $common, [ 'type', 'start', 'reversed' ] ),
1924 'li' => $merge( $common, [ 'type', 'value' ] ),
1932 'table' => $merge( $common,
1933 [ 'summary', 'width', 'border', 'frame',
1934 'rules', 'cellspacing', 'cellpadding',
1939 'caption' => $block,
1947 'colgroup' => $merge( $common, [ 'span' ] ),
1948 'col' => $merge( $common, [ 'span' ] ),
1951 'tr' => $merge( $common, [ 'bgcolor' ], $tablealign ),
1954 'td' => $merge( $common, $tablecell, $tablealign ),
1955 'th' => $merge( $common, $tablecell, $tablealign ),
1958 # NOTE: <a> is not allowed directly, but the attrib
1959 # whitelist is used from the Parser object
1960 'a' => $merge( $common, [ 'href', 'rel', 'rev' ] ), # rel/rev esp. for RDFa
1963 # Not usually allowed, but may be used for extension-style hooks
1964 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1966 'img' => $merge( $common, [ 'alt', 'src', 'width', 'height', 'srcset' ] ),
1967 # Attributes for A/V tags added in T163583 / T133673
1968 'audio' => $merge( $common, [ 'controls', 'preload', 'width', 'height' ] ),
1969 'video' => $merge( $common, [ 'poster', 'controls', 'preload', 'width', 'height' ] ),
1970 'source' => $merge( $common, [ 'type', 'src' ] ),
1971 'track' => $merge( $common, [ 'type', 'src', 'srclang', 'kind', 'label' ] ),
1979 'strike' => $common,
1984 'font' => $merge( $common, [ 'size', 'color', 'face' ] ),
1988 'hr' => $merge( $common, [ 'width' ] ),
1990 # HTML Ruby annotation text module, simple ruby only.
1991 # https://www.w3.org/TR/html5/text-level-semantics.html#the-ruby-element
1996 'rt' => $common, # $merge( $common, [ 'rbspan' ] ),
1999 # MathML root element, where used for extensions
2000 # 'title' may not be 100% valid here; it's XHTML
2001 # https://www.w3.org/TR/REC-MathML/
2002 'math' => $merge( [], [ 'class', 'style', 'id', 'title' ] ),
2004 // HTML 5 section 4.5
2005 'figure' => $common,
2006 'figure-inline' => $common, # T118520
2007 'figcaption' => $common,
2009 # HTML 5 section 4.6
2012 # HTML5 elements, defined by:
2013 # https://html.spec.whatwg.org/multipage/semantics.html#the-data-element
2014 'data' => $merge( $common, [ 'value' ] ),
2015 'time' => $merge( $common, [ 'datetime' ] ),
2018 // meta and link are only permitted by removeHTMLtags when Microdata
2019 // is enabled so we don't bother adding a conditional to hide these
2020 // Also meta and link are only valid in WikiText as Microdata elements
2021 // (ie: validateTag rejects tags missing the attributes needed for Microdata)
2022 // So we don't bother including $common attributes that have no purpose.
2023 'meta' => $merge( [], [ 'itemprop', 'content' ] ),
2024 'link' => $merge( [], [ 'itemprop', 'href', 'title' ] ),
2031 * Take a fragment of (potentially invalid) HTML and return
2032 * a version with any tags removed, encoded as plain text.
2034 * Warning: this return value must be further escaped for literal
2035 * inclusion in HTML output as of 1.10!
2037 * @param string $html HTML fragment
2040 static function stripAllTags( $html ) {
2041 // Use RemexHtml to tokenize $html and extract the text
2042 $handler = new RemexStripTagHandler
;
2043 $tokenizer = new RemexHtml\Tokenizer\
Tokenizer( $handler, $html, [
2044 'ignoreErrors' => true,
2045 // don't ignore char refs, we want them to be decoded
2046 'ignoreNulls' => true,
2047 'skipPreprocess' => true,
2049 $tokenizer->execute();
2050 $text = $handler->getResult();
2052 $text = self
::normalizeWhitespace( $text );
2057 * Hack up a private DOCTYPE with HTML's standard entity declarations.
2058 * PHP 4 seemed to know these if you gave it an HTML doctype, but
2061 * Use for passing XHTML fragments to PHP's XML parsing functions
2065 static function hackDocType() {
2066 $out = "<!DOCTYPE html [\n";
2067 foreach ( self
::$htmlEntities as $entity => $codepoint ) {
2068 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
2075 * @param string $url
2076 * @return mixed|string
2078 static function cleanUrl( $url ) {
2079 # Normalize any HTML entities in input. They will be
2080 # re-escaped by makeExternalLink().
2081 $url = self
::decodeCharReferences( $url );
2083 # Escape any control characters introduced by the above step
2084 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
2085 [ __CLASS__
, 'cleanUrlCallback' ], $url );
2087 # Validate hostname portion
2089 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
2090 list( /* $whole */, $protocol, $host, $rest ) = $matches;
2092 // Characters that will be ignored in IDNs.
2093 // https://tools.ietf.org/html/rfc3454#section-3.1
2094 // Strip them before further processing so blacklists and such work.
2096 \\s| # general whitespace
2097 \xc2\xad| # 00ad SOFT HYPHEN
2098 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
2099 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
2100 \xe2\x81\xa0| # 2060 WORD JOINER
2101 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
2102 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
2103 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
2104 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
2105 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
2106 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
2107 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
2108 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
2111 $host = preg_replace( $strip, '', $host );
2113 // IPv6 host names are bracketed with []. Url-decode these.
2114 if ( substr_compare( "//%5B", $host, 0, 5 ) === 0 &&
2115 preg_match( '!^//%5B([0-9A-Fa-f:.]+)%5D((:\d+)?)$!', $host, $matches )
2117 $host = '//[' . $matches[1] . ']' . $matches[2];
2120 // @todo FIXME: Validate hostnames here
2122 return $protocol . $host . $rest;
2129 * @param array $matches
2132 static function cleanUrlCallback( $matches ) {
2133 return urlencode( $matches[0] );
2137 * Does a string look like an e-mail address?
2139 * This validates an email address using an HTML5 specification found at:
2140 * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
2141 * Which as of 2011-01-24 says:
2143 * A valid e-mail address is a string that matches the ABNF production
2144 * 1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
2145 * in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
2148 * This function is an implementation of the specification as requested in
2151 * Client-side forms will use the same standard validation rules via JS or
2152 * HTML 5 validation; additional restrictions can be enforced server-side
2153 * by extensions via the 'isValidEmailAddr' hook.
2155 * Note that this validation doesn't 100% match RFC 2822, but is believed
2156 * to be liberal enough for wide use. Some invalid addresses will still
2157 * pass validation here.
2161 * @param string $addr E-mail address
2164 public static function validateEmail( $addr ) {
2166 if ( !Hooks
::run( 'isValidEmailAddr', [ $addr, &$result ] ) ) {
2170 // Please note strings below are enclosed in brackets [], this make the
2171 // hyphen "-" a range indicator. Hence it is double backslashed below.
2173 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
2174 $rfc1034_ldh_str = "a-z0-9\\-";
2176 $html5_email_regexp = "/
2178 [$rfc5322_atext\\.]+ # user part which is liberal :p
2180 [$rfc1034_ldh_str]+ # First domain part
2181 (\\.[$rfc1034_ldh_str]+)* # Following part prefixed with a dot
2183 /ix"; // case Insensitive, eXtended
2185 return (bool)preg_match( $html5_email_regexp, $addr );