User::isSafeToLoad() should return false if MW_NO_SESSION
[mediawiki.git] / includes / Sanitizer.php
blobd52bc07324a5075a46f38fff1b144fef6c285a3a
1 <?php
2 /**
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
23 * @file
24 * @ingroup Parser
27 /**
28 * HTML sanitizer for MediaWiki
29 * @ingroup Parser
31 class Sanitizer {
32 /**
33 * Regular expression to match various types of character references in
34 * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
36 const CHAR_REFS_REGEX =
37 '/&([A-Za-z0-9\x80-\xff]+);
38 |&\#([0-9]+);
39 |&\#[xX]([0-9A-Fa-f]+);
40 |(&)/x';
42 /**
43 * Acceptable tag name charset from HTML5 parsing spec
44 * http://www.w3.org/TR/html5/syntax.html#tag-open-state
46 const ELEMENT_BITS_REGEX = '!^(/?)([A-Za-z][^\t\n\v />\0]*+)([^>]*?)(/?>)([^<]*)$!';
48 /**
49 * Blacklist for evil uris like javascript:
50 * WARNING: DO NOT use this in any place that actually requires blacklisting
51 * for security reasons. There are NUMEROUS[1] ways to bypass blacklisting, the
52 * only way to be secure from javascript: uri based xss vectors is to whitelist
53 * things that you know are safe and deny everything else.
54 * [1]: http://ha.ckers.org/xss.html
56 const EVIL_URI_PATTERN = '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
57 const XMLNS_ATTRIBUTE_PATTERN = "/^xmlns:[:A-Z_a-z-.0-9]+$/";
59 /**
60 * List of all named character entities defined in HTML 4.01
61 * http://www.w3.org/TR/html4/sgml/entities.html
62 * As well as &apos; which is only defined starting in XHTML1.
64 private static $htmlEntities = [
65 'Aacute' => 193,
66 'aacute' => 225,
67 'Acirc' => 194,
68 'acirc' => 226,
69 'acute' => 180,
70 'AElig' => 198,
71 'aelig' => 230,
72 'Agrave' => 192,
73 'agrave' => 224,
74 'alefsym' => 8501,
75 'Alpha' => 913,
76 'alpha' => 945,
77 'amp' => 38,
78 'and' => 8743,
79 'ang' => 8736,
80 'apos' => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
81 'Aring' => 197,
82 'aring' => 229,
83 'asymp' => 8776,
84 'Atilde' => 195,
85 'atilde' => 227,
86 'Auml' => 196,
87 'auml' => 228,
88 'bdquo' => 8222,
89 'Beta' => 914,
90 'beta' => 946,
91 'brvbar' => 166,
92 'bull' => 8226,
93 'cap' => 8745,
94 'Ccedil' => 199,
95 'ccedil' => 231,
96 'cedil' => 184,
97 'cent' => 162,
98 'Chi' => 935,
99 'chi' => 967,
100 'circ' => 710,
101 'clubs' => 9827,
102 'cong' => 8773,
103 'copy' => 169,
104 'crarr' => 8629,
105 'cup' => 8746,
106 'curren' => 164,
107 'dagger' => 8224,
108 'Dagger' => 8225,
109 'darr' => 8595,
110 'dArr' => 8659,
111 'deg' => 176,
112 'Delta' => 916,
113 'delta' => 948,
114 'diams' => 9830,
115 'divide' => 247,
116 'Eacute' => 201,
117 'eacute' => 233,
118 'Ecirc' => 202,
119 'ecirc' => 234,
120 'Egrave' => 200,
121 'egrave' => 232,
122 'empty' => 8709,
123 'emsp' => 8195,
124 'ensp' => 8194,
125 'Epsilon' => 917,
126 'epsilon' => 949,
127 'equiv' => 8801,
128 'Eta' => 919,
129 'eta' => 951,
130 'ETH' => 208,
131 'eth' => 240,
132 'Euml' => 203,
133 'euml' => 235,
134 'euro' => 8364,
135 'exist' => 8707,
136 'fnof' => 402,
137 'forall' => 8704,
138 'frac12' => 189,
139 'frac14' => 188,
140 'frac34' => 190,
141 'frasl' => 8260,
142 'Gamma' => 915,
143 'gamma' => 947,
144 'ge' => 8805,
145 'gt' => 62,
146 'harr' => 8596,
147 'hArr' => 8660,
148 'hearts' => 9829,
149 'hellip' => 8230,
150 'Iacute' => 205,
151 'iacute' => 237,
152 'Icirc' => 206,
153 'icirc' => 238,
154 'iexcl' => 161,
155 'Igrave' => 204,
156 'igrave' => 236,
157 'image' => 8465,
158 'infin' => 8734,
159 'int' => 8747,
160 'Iota' => 921,
161 'iota' => 953,
162 'iquest' => 191,
163 'isin' => 8712,
164 'Iuml' => 207,
165 'iuml' => 239,
166 'Kappa' => 922,
167 'kappa' => 954,
168 'Lambda' => 923,
169 'lambda' => 955,
170 'lang' => 9001,
171 'laquo' => 171,
172 'larr' => 8592,
173 'lArr' => 8656,
174 'lceil' => 8968,
175 'ldquo' => 8220,
176 'le' => 8804,
177 'lfloor' => 8970,
178 'lowast' => 8727,
179 'loz' => 9674,
180 'lrm' => 8206,
181 'lsaquo' => 8249,
182 'lsquo' => 8216,
183 'lt' => 60,
184 'macr' => 175,
185 'mdash' => 8212,
186 'micro' => 181,
187 'middot' => 183,
188 'minus' => 8722,
189 'Mu' => 924,
190 'mu' => 956,
191 'nabla' => 8711,
192 'nbsp' => 160,
193 'ndash' => 8211,
194 'ne' => 8800,
195 'ni' => 8715,
196 'not' => 172,
197 'notin' => 8713,
198 'nsub' => 8836,
199 'Ntilde' => 209,
200 'ntilde' => 241,
201 'Nu' => 925,
202 'nu' => 957,
203 'Oacute' => 211,
204 'oacute' => 243,
205 'Ocirc' => 212,
206 'ocirc' => 244,
207 'OElig' => 338,
208 'oelig' => 339,
209 'Ograve' => 210,
210 'ograve' => 242,
211 'oline' => 8254,
212 'Omega' => 937,
213 'omega' => 969,
214 'Omicron' => 927,
215 'omicron' => 959,
216 'oplus' => 8853,
217 'or' => 8744,
218 'ordf' => 170,
219 'ordm' => 186,
220 'Oslash' => 216,
221 'oslash' => 248,
222 'Otilde' => 213,
223 'otilde' => 245,
224 'otimes' => 8855,
225 'Ouml' => 214,
226 'ouml' => 246,
227 'para' => 182,
228 'part' => 8706,
229 'permil' => 8240,
230 'perp' => 8869,
231 'Phi' => 934,
232 'phi' => 966,
233 'Pi' => 928,
234 'pi' => 960,
235 'piv' => 982,
236 'plusmn' => 177,
237 'pound' => 163,
238 'prime' => 8242,
239 'Prime' => 8243,
240 'prod' => 8719,
241 'prop' => 8733,
242 'Psi' => 936,
243 'psi' => 968,
244 'quot' => 34,
245 'radic' => 8730,
246 'rang' => 9002,
247 'raquo' => 187,
248 'rarr' => 8594,
249 'rArr' => 8658,
250 'rceil' => 8969,
251 'rdquo' => 8221,
252 'real' => 8476,
253 'reg' => 174,
254 'rfloor' => 8971,
255 'Rho' => 929,
256 'rho' => 961,
257 'rlm' => 8207,
258 'rsaquo' => 8250,
259 'rsquo' => 8217,
260 'sbquo' => 8218,
261 'Scaron' => 352,
262 'scaron' => 353,
263 'sdot' => 8901,
264 'sect' => 167,
265 'shy' => 173,
266 'Sigma' => 931,
267 'sigma' => 963,
268 'sigmaf' => 962,
269 'sim' => 8764,
270 'spades' => 9824,
271 'sub' => 8834,
272 'sube' => 8838,
273 'sum' => 8721,
274 'sup' => 8835,
275 'sup1' => 185,
276 'sup2' => 178,
277 'sup3' => 179,
278 'supe' => 8839,
279 'szlig' => 223,
280 'Tau' => 932,
281 'tau' => 964,
282 'there4' => 8756,
283 'Theta' => 920,
284 'theta' => 952,
285 'thetasym' => 977,
286 'thinsp' => 8201,
287 'THORN' => 222,
288 'thorn' => 254,
289 'tilde' => 732,
290 'times' => 215,
291 'trade' => 8482,
292 'Uacute' => 218,
293 'uacute' => 250,
294 'uarr' => 8593,
295 'uArr' => 8657,
296 'Ucirc' => 219,
297 'ucirc' => 251,
298 'Ugrave' => 217,
299 'ugrave' => 249,
300 'uml' => 168,
301 'upsih' => 978,
302 'Upsilon' => 933,
303 'upsilon' => 965,
304 'Uuml' => 220,
305 'uuml' => 252,
306 'weierp' => 8472,
307 'Xi' => 926,
308 'xi' => 958,
309 'Yacute' => 221,
310 'yacute' => 253,
311 'yen' => 165,
312 'Yuml' => 376,
313 'yuml' => 255,
314 'Zeta' => 918,
315 'zeta' => 950,
316 'zwj' => 8205,
317 'zwnj' => 8204
321 * Character entity aliases accepted by MediaWiki
323 private static $htmlEntityAliases = [
324 'רלמ' => 'rlm',
325 'رلم' => 'rlm',
329 * Lazy-initialised attributes regex, see getAttribsRegex()
331 private static $attribsRegex;
334 * Regular expression to match HTML/XML attribute pairs within a tag.
335 * Allows some... latitude. Based on,
336 * http://www.w3.org/TR/html5/syntax.html#before-attribute-value-state
337 * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
338 * @return string
340 static function getAttribsRegex() {
341 if ( self::$attribsRegex === null ) {
342 $attribFirst = '[:A-Z_a-z0-9]';
343 $attrib = '[:A-Z_a-z-.0-9]';
344 $space = '[\x09\x0a\x0c\x0d\x20]';
345 self::$attribsRegex =
346 "/(?:^|$space)({$attribFirst}{$attrib}*)
347 ($space*=$space*
349 # The attribute value: quoted or alone
350 \"([^\"]*)(?:\"|\$)
351 | '([^']*)(?:'|\$)
352 | (((?!$space|>).)*)
354 )?(?=$space|\$)/sx";
356 return self::$attribsRegex;
360 * Return the various lists of recognized tags
361 * @param array $extratags For any extra tags to include
362 * @param array $removetags For any tags (default or extra) to exclude
363 * @return array
365 public static function getRecognizedTagData( $extratags = [], $removetags = [] ) {
366 global $wgAllowMicrodataAttributes, $wgAllowImageTag;
368 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
369 $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
371 // Base our staticInitialised variable off of the global config state so that if the globals
372 // are changed (like in the screwed up test system) we will re-initialise the settings.
373 $globalContext = implode( '-', compact( 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
374 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
375 $htmlpairsStatic = [ # Tags that must be closed
376 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
377 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
378 'strike', 'strong', 'tt', 'var', 'div', 'center',
379 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
380 'ruby', 'rb', 'rp', 'rt', 'rtc', 'p', 'span', 'abbr', 'dfn',
381 'kbd', 'samp', 'data', 'time', 'mark'
383 $htmlsingle = [
384 'br', 'wbr', 'hr', 'li', 'dt', 'dd'
386 $htmlsingleonly = [ # Elements that cannot have close tags
387 'br', 'wbr', 'hr'
389 if ( $wgAllowMicrodataAttributes ) {
390 $htmlsingle[] = $htmlsingleonly[] = 'meta';
391 $htmlsingle[] = $htmlsingleonly[] = 'link';
393 $htmlnest = [ # Tags that can be nested--??
394 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
395 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
396 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
398 $tabletags = [ # Can only appear inside table, we will close them
399 'td', 'th', 'tr',
401 $htmllist = [ # Tags used by list
402 'ul', 'ol',
404 $listtags = [ # Tags that can appear in a list
405 'li',
408 if ( $wgAllowImageTag ) {
409 $htmlsingle[] = 'img';
410 $htmlsingleonly[] = 'img';
413 $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
414 $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
416 # Convert them all to hashtables for faster lookup
417 $vars = [ 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
418 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' ];
419 foreach ( $vars as $var ) {
420 $$var = array_flip( $$var );
422 $staticInitialised = $globalContext;
425 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
426 $extratags = array_flip( $extratags );
427 $removetags = array_flip( $removetags );
428 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
429 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
431 return [
432 'htmlpairs' => $htmlpairs,
433 'htmlsingle' => $htmlsingle,
434 'htmlsingleonly' => $htmlsingleonly,
435 'htmlnest' => $htmlnest,
436 'tabletags' => $tabletags,
437 'htmllist' => $htmllist,
438 'listtags' => $listtags,
439 'htmlsingleallowed' => $htmlsingleallowed,
440 'htmlelements' => $htmlelements,
445 * Cleans up HTML, removes dangerous tags and attributes, and
446 * removes HTML comments
447 * @param string $text
448 * @param callable $processCallback Callback to do any variable or parameter
449 * replacements in HTML attribute values
450 * @param array|bool $args Arguments for the processing callback
451 * @param array $extratags For any extra tags to include
452 * @param array $removetags For any tags (default or extra) to exclude
453 * @return string
455 public static function removeHTMLtags( $text, $processCallback = null,
456 $args = [], $extratags = [], $removetags = []
458 extract( self::getRecognizedTagData( $extratags, $removetags ) );
460 # Remove HTML comments
461 $text = Sanitizer::removeHTMLcomments( $text );
462 $bits = explode( '<', $text );
463 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
464 if ( !MWTidy::isEnabled() ) {
465 $tagstack = $tablestack = [];
466 foreach ( $bits as $x ) {
467 $regs = [];
468 # $slash: Does the current element start with a '/'?
469 # $t: Current element name
470 # $params: String between element name and >
471 # $brace: Ending '>' or '/>'
472 # $rest: Everything until the next element of $bits
473 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
474 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
475 } else {
476 $slash = $t = $params = $brace = $rest = null;
479 $badtag = false;
480 $t = strtolower( $t );
481 if ( isset( $htmlelements[$t] ) ) {
482 # Check our stack
483 if ( $slash && isset( $htmlsingleonly[$t] ) ) {
484 $badtag = true;
485 } elseif ( $slash ) {
486 # Closing a tag... is it the one we just opened?
487 MediaWiki\suppressWarnings();
488 $ot = array_pop( $tagstack );
489 MediaWiki\restoreWarnings();
491 if ( $ot != $t ) {
492 if ( isset( $htmlsingleallowed[$ot] ) ) {
493 # Pop all elements with an optional close tag
494 # and see if we find a match below them
495 $optstack = [];
496 array_push( $optstack, $ot );
497 MediaWiki\suppressWarnings();
498 $ot = array_pop( $tagstack );
499 MediaWiki\restoreWarnings();
500 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
501 array_push( $optstack, $ot );
502 MediaWiki\suppressWarnings();
503 $ot = array_pop( $tagstack );
504 MediaWiki\restoreWarnings();
506 if ( $t != $ot ) {
507 # No match. Push the optional elements back again
508 $badtag = true;
509 MediaWiki\suppressWarnings();
510 $ot = array_pop( $optstack );
511 MediaWiki\restoreWarnings();
512 while ( $ot ) {
513 array_push( $tagstack, $ot );
514 MediaWiki\suppressWarnings();
515 $ot = array_pop( $optstack );
516 MediaWiki\restoreWarnings();
519 } else {
520 MediaWiki\suppressWarnings();
521 array_push( $tagstack, $ot );
522 MediaWiki\restoreWarnings();
524 # <li> can be nested in <ul> or <ol>, skip those cases:
525 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
526 $badtag = true;
529 } else {
530 if ( $t == 'table' ) {
531 $tagstack = array_pop( $tablestack );
534 $newparams = '';
535 } else {
536 # Keep track for later
537 if ( isset( $tabletags[$t] ) && !in_array( 'table', $tagstack ) ) {
538 $badtag = true;
539 } elseif ( in_array( $t, $tagstack ) && !isset( $htmlnest[$t] ) ) {
540 $badtag = true;
541 #  Is it a self closed htmlpair ? (bug 5487)
542 } elseif ( $brace == '/>' && isset( $htmlpairs[$t] ) ) {
543 $badtag = true;
544 } elseif ( isset( $htmlsingleonly[$t] ) ) {
545 # Hack to force empty tag for unclosable elements
546 $brace = '/>';
547 } elseif ( isset( $htmlsingle[$t] ) ) {
548 # Hack to not close $htmlsingle tags
549 $brace = null;
550 # Still need to push this optionally-closed tag to
551 # the tag stack so that we can match end tags
552 # instead of marking them as bad.
553 array_push( $tagstack, $t );
554 } elseif ( isset( $tabletags[$t] ) && in_array( $t, $tagstack ) ) {
555 // New table tag but forgot to close the previous one
556 $text .= "</$t>";
557 } else {
558 if ( $t == 'table' ) {
559 array_push( $tablestack, $tagstack );
560 $tagstack = [];
562 array_push( $tagstack, $t );
565 # Replace any variables or template parameters with
566 # plaintext results.
567 if ( is_callable( $processCallback ) ) {
568 call_user_func_array( $processCallback, [ &$params, $args ] );
571 if ( !Sanitizer::validateTag( $params, $t ) ) {
572 $badtag = true;
575 # Strip non-approved attributes from the tag
576 $newparams = Sanitizer::fixTagAttributes( $params, $t );
578 if ( !$badtag ) {
579 $rest = str_replace( '>', '&gt;', $rest );
580 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
581 $text .= "<$slash$t$newparams$close>$rest";
582 continue;
585 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
587 # Close off any remaining tags
588 while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
589 $text .= "</$t>\n";
590 if ( $t == 'table' ) {
591 $tagstack = array_pop( $tablestack );
594 } else {
595 # this might be possible using tidy itself
596 foreach ( $bits as $x ) {
597 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
598 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
600 $badtag = false;
601 $t = strtolower( $t );
602 if ( isset( $htmlelements[$t] ) ) {
603 if ( is_callable( $processCallback ) ) {
604 call_user_func_array( $processCallback, [ &$params, $args ] );
607 if ( !Sanitizer::validateTag( $params, $t ) ) {
608 $badtag = true;
611 $newparams = Sanitizer::fixTagAttributes( $params, $t );
612 if ( !$badtag ) {
613 $rest = str_replace( '>', '&gt;', $rest );
614 $text .= "<$slash$t$newparams$brace$rest";
615 continue;
619 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
622 return $text;
626 * Remove '<!--', '-->', and everything between.
627 * To avoid leaving blank lines, when a comment is both preceded
628 * and followed by a newline (ignoring spaces), trim leading and
629 * trailing spaces and one of the newlines.
631 * @param string $text
632 * @return string
634 public static function removeHTMLcomments( $text ) {
635 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
636 $end = strpos( $text, '-->', $start + 4 );
637 if ( $end === false ) {
638 # Unterminated comment; bail out
639 break;
642 $end += 3;
644 # Trim space and newline if the comment is both
645 # preceded and followed by a newline
646 $spaceStart = max( $start - 1, 0 );
647 $spaceLen = $end - $spaceStart;
648 while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
649 $spaceStart--;
650 $spaceLen++;
652 while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
653 $spaceLen++;
655 if ( substr( $text, $spaceStart, 1 ) === "\n"
656 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
657 # Remove the comment, leading and trailing
658 # spaces, and leave only one newline.
659 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
660 } else {
661 # Remove just the comment.
662 $text = substr_replace( $text, '', $start, $end - $start );
665 return $text;
669 * Takes attribute names and values for a tag and the tag name and
670 * validates that the tag is allowed to be present.
671 * This DOES NOT validate the attributes, nor does it validate the
672 * tags themselves. This method only handles the special circumstances
673 * where we may want to allow a tag within content but ONLY when it has
674 * specific attributes set.
676 * @param string $params
677 * @param string $element
678 * @return bool
680 static function validateTag( $params, $element ) {
681 $params = Sanitizer::decodeTagAttributes( $params );
683 if ( $element == 'meta' || $element == 'link' ) {
684 if ( !isset( $params['itemprop'] ) ) {
685 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
686 return false;
688 if ( $element == 'meta' && !isset( $params['content'] ) ) {
689 // <meta> must have a content="" for the itemprop
690 return false;
692 if ( $element == 'link' && !isset( $params['href'] ) ) {
693 // <link> must have an associated href=""
694 return false;
698 return true;
702 * Take an array of attribute names and values and normalize or discard
703 * illegal values for the given element type.
705 * - Discards attributes not on a whitelist for the given element
706 * - Unsafe style attributes are discarded
707 * - Invalid id attributes are re-encoded
709 * @param array $attribs
710 * @param string $element
711 * @return array
713 * @todo Check for legal values where the DTD limits things.
714 * @todo Check for unique id attribute :P
716 static function validateTagAttributes( $attribs, $element ) {
717 return Sanitizer::validateAttributes( $attribs,
718 Sanitizer::attributeWhitelist( $element ) );
722 * Take an array of attribute names and values and normalize or discard
723 * illegal values for the given whitelist.
725 * - Discards attributes not on the given whitelist
726 * - Unsafe style attributes are discarded
727 * - Invalid id attributes are re-encoded
729 * @param array $attribs
730 * @param array $whitelist List of allowed attribute names
731 * @return array
733 * @todo Check for legal values where the DTD limits things.
734 * @todo Check for unique id attribute :P
736 static function validateAttributes( $attribs, $whitelist ) {
737 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
739 $whitelist = array_flip( $whitelist );
740 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
742 $out = [];
743 foreach ( $attribs as $attribute => $value ) {
744 # allow XML namespace declaration if RDFa is enabled
745 if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
746 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
747 $out[$attribute] = $value;
750 continue;
753 # Allow any attribute beginning with "data-"
754 # However:
755 # * data-ooui is reserved for ooui
756 # * data-mw and data-parsoid are reserved for parsoid
757 # * data-mw-<name here> is reserved for extensions (or core) if
758 # they need to communicate some data to the client and want to be
759 # sure that it isn't coming from an untrusted user.
760 # * Ensure that the attribute is not namespaced by banning
761 # colons.
762 if ( !preg_match( '/^data-(?!ooui|mw|parsoid)[^:]*$/i', $attribute )
763 && !isset( $whitelist[$attribute] )
765 continue;
768 # Strip javascript "expression" from stylesheets.
769 # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
770 if ( $attribute == 'style' ) {
771 $value = Sanitizer::checkCss( $value );
774 # Escape HTML id attributes
775 if ( $attribute === 'id' ) {
776 $value = Sanitizer::escapeId( $value, 'noninitial' );
779 # Escape HTML id reference lists
780 if ( $attribute === 'aria-describedby'
781 || $attribute === 'aria-flowto'
782 || $attribute === 'aria-labelledby'
783 || $attribute === 'aria-owns'
785 $value = Sanitizer::escapeIdReferenceList( $value, 'noninitial' );
788 // RDFa and microdata properties allow URLs, URIs and/or CURIs.
789 // Check them for sanity.
790 if ( $attribute === 'rel' || $attribute === 'rev'
791 # RDFa
792 || $attribute === 'about' || $attribute === 'property'
793 || $attribute === 'resource' || $attribute === 'datatype'
794 || $attribute === 'typeof'
795 # HTML5 microdata
796 || $attribute === 'itemid' || $attribute === 'itemprop'
797 || $attribute === 'itemref' || $attribute === 'itemscope'
798 || $attribute === 'itemtype'
800 // Paranoia. Allow "simple" values but suppress javascript
801 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
802 continue;
806 # NOTE: even though elements using href/src are not allowed directly, supply
807 # validation code that can be used by tag hook handlers, etc
808 if ( $attribute === 'href' || $attribute === 'src' ) {
809 if ( !preg_match( $hrefExp, $value ) ) {
810 continue; // drop any href or src attributes not using an allowed protocol.
811 // NOTE: this also drops all relative URLs
815 // If this attribute was previously set, override it.
816 // Output should only have one attribute of each name.
817 $out[$attribute] = $value;
820 if ( $wgAllowMicrodataAttributes ) {
821 # itemtype, itemid, itemref don't make sense without itemscope
822 if ( !array_key_exists( 'itemscope', $out ) ) {
823 unset( $out['itemtype'] );
824 unset( $out['itemid'] );
825 unset( $out['itemref'] );
827 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
829 return $out;
833 * Merge two sets of HTML attributes. Conflicting items in the second set
834 * will override those in the first, except for 'class' attributes which
835 * will be combined (if they're both strings).
837 * @todo implement merging for other attributes such as style
838 * @param array $a
839 * @param array $b
840 * @return array
842 static function mergeAttributes( $a, $b ) {
843 $out = array_merge( $a, $b );
844 if ( isset( $a['class'] ) && isset( $b['class'] )
845 && is_string( $a['class'] ) && is_string( $b['class'] )
846 && $a['class'] !== $b['class']
848 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
849 -1, PREG_SPLIT_NO_EMPTY );
850 $out['class'] = implode( ' ', array_unique( $classes ) );
852 return $out;
856 * Normalize CSS into a format we can easily search for hostile input
857 * - decode character references
858 * - decode escape sequences
859 * - convert characters that IE6 interprets into ascii
860 * - remove comments, unless the entire value is one single comment
861 * @param string $value the css string
862 * @return string normalized css
864 public static function normalizeCss( $value ) {
866 // Decode character references like &#123;
867 $value = Sanitizer::decodeCharReferences( $value );
869 // Decode escape sequences and line continuation
870 // See the grammar in the CSS 2 spec, appendix D.
871 // This has to be done AFTER decoding character references.
872 // This means it isn't possible for this function to return
873 // unsanitized escape sequences. It is possible to manufacture
874 // input that contains character references that decode to
875 // escape sequences that decode to character references, but
876 // it's OK for the return value to contain character references
877 // because the caller is supposed to escape those anyway.
878 static $decodeRegex;
879 if ( !$decodeRegex ) {
880 $space = '[\\x20\\t\\r\\n\\f]';
881 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
882 $backslash = '\\\\';
883 $decodeRegex = "/ $backslash
885 ($nl) | # 1. Line continuation
886 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
887 (.) | # 3. backslash cancelling special meaning
888 () | # 4. backslash at end of string
889 )/xu";
891 $value = preg_replace_callback( $decodeRegex,
892 [ __CLASS__, 'cssDecodeCallback' ], $value );
894 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
895 $value = preg_replace_callback(
896 '/[!-[]-z]/u', // U+FF01 to U+FF5A, excluding U+FF3C (bug 58088)
897 function ( $matches ) {
898 $cp = UtfNormal\Utils::utf8ToCodepoint( $matches[0] );
899 if ( $cp === false ) {
900 return '';
902 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
904 $value
907 // Convert more characters IE6 might treat as ascii
908 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
909 $value = str_replace(
910 [ 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ],
911 [ 'r', 'n', 'n', 'l', 'i', '(', '(' ],
912 $value
915 // Let the value through if it's nothing but a single comment, to
916 // allow other functions which may reject it to pass some error
917 // message through.
918 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
919 // Remove any comments; IE gets token splitting wrong
920 // This must be done AFTER decoding character references and
921 // escape sequences, because those steps can introduce comments
922 // This step cannot introduce character references or escape
923 // sequences, because it replaces comments with spaces rather
924 // than removing them completely.
925 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
927 // Remove anything after a comment-start token, to guard against
928 // incorrect client implementations.
929 $commentPos = strpos( $value, '/*' );
930 if ( $commentPos !== false ) {
931 $value = substr( $value, 0, $commentPos );
935 // S followed by repeat, iteration, or prolonged sound marks,
936 // which IE will treat as "ss"
937 $value = preg_replace(
938 '/s(?:
939 \xE3\x80\xB1 | # U+3031
940 \xE3\x82\x9D | # U+309D
941 \xE3\x83\xBC | # U+30FC
942 \xE3\x83\xBD | # U+30FD
943 \xEF\xB9\xBC | # U+FE7C
944 \xEF\xB9\xBD | # U+FE7D
945 \xEF\xBD\xB0 # U+FF70
946 )/ix',
947 'ss',
948 $value
951 return $value;
955 * Pick apart some CSS and check it for forbidden or unsafe structures.
956 * Returns a sanitized string. This sanitized string will have
957 * character references and escape sequences decoded and comments
958 * stripped (unless it is itself one valid comment, in which case the value
959 * will be passed through). If the input is just too evil, only a comment
960 * complaining about evilness will be returned.
962 * Currently URL references, 'expression', 'tps' are forbidden.
964 * NOTE: Despite the fact that character references are decoded, the
965 * returned string may contain character references given certain
966 * clever input strings. These character references must
967 * be escaped before the return value is embedded in HTML.
969 * @param string $value
970 * @return string
972 static function checkCss( $value ) {
973 $value = self::normalizeCss( $value );
975 // Reject problematic keywords and control characters
976 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ||
977 strpos( $value, UtfNormal\Constants::UTF8_REPLACEMENT ) !== false ) {
978 return '/* invalid control char */';
979 } elseif ( preg_match(
980 '! expression
981 | filter\s*:
982 | accelerator\s*:
983 | -o-link\s*:
984 | -o-link-source\s*:
985 | -o-replace\s*:
986 | url\s*\(
987 | image\s*\(
988 | image-set\s*\(
989 !ix', $value ) ) {
990 return '/* insecure input */';
992 return $value;
996 * @param array $matches
997 * @return string
999 static function cssDecodeCallback( $matches ) {
1000 if ( $matches[1] !== '' ) {
1001 // Line continuation
1002 return '';
1003 } elseif ( $matches[2] !== '' ) {
1004 $char = UtfNormal\Utils::codepointToUtf8( hexdec( $matches[2] ) );
1005 } elseif ( $matches[3] !== '' ) {
1006 $char = $matches[3];
1007 } else {
1008 $char = '\\';
1010 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
1011 // These characters need to be escaped in strings
1012 // Clean up the escape sequence to avoid parsing errors by clients
1013 return '\\' . dechex( ord( $char ) ) . ' ';
1014 } else {
1015 // Decode unnecessary escape
1016 return $char;
1021 * Take a tag soup fragment listing an HTML element's attributes
1022 * and normalize it to well-formed XML, discarding unwanted attributes.
1023 * Output is safe for further wikitext processing, with escaping of
1024 * values that could trigger problems.
1026 * - Normalizes attribute names to lowercase
1027 * - Discards attributes not on a whitelist for the given element
1028 * - Turns broken or invalid entities into plaintext
1029 * - Double-quotes all attribute values
1030 * - Attributes without values are given the name as attribute
1031 * - Double attributes are discarded
1032 * - Unsafe style attributes are discarded
1033 * - Prepends space if there are attributes.
1035 * @param string $text
1036 * @param string $element
1037 * @return string
1039 static function fixTagAttributes( $text, $element ) {
1040 if ( trim( $text ) == '' ) {
1041 return '';
1044 $decoded = Sanitizer::decodeTagAttributes( $text );
1045 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1047 return Sanitizer::safeEncodeTagAttributes( $stripped );
1051 * Encode an attribute value for HTML output.
1052 * @param string $text
1053 * @return string HTML-encoded text fragment
1055 static function encodeAttribute( $text ) {
1056 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1058 // Whitespace is normalized during attribute decoding,
1059 // so if we've been passed non-spaces we must encode them
1060 // ahead of time or they won't be preserved.
1061 $encValue = strtr( $encValue, [
1062 "\n" => '&#10;',
1063 "\r" => '&#13;',
1064 "\t" => '&#9;',
1065 ] );
1067 return $encValue;
1071 * Encode an attribute value for HTML tags, with extra armoring
1072 * against further wiki processing.
1073 * @param string $text
1074 * @return string HTML-encoded text fragment
1076 static function safeEncodeAttribute( $text ) {
1077 $encValue = Sanitizer::encodeAttribute( $text );
1079 # Templates and links may be expanded in later parsing,
1080 # creating invalid or dangerous output. Suppress this.
1081 $encValue = strtr( $encValue, [
1082 '<' => '&lt;', // This should never happen,
1083 '>' => '&gt;', // we've received invalid input
1084 '"' => '&quot;', // which should have been escaped.
1085 '{' => '&#123;',
1086 '[' => '&#91;',
1087 "''" => '&#39;&#39;',
1088 'ISBN' => '&#73;SBN',
1089 'RFC' => '&#82;FC',
1090 'PMID' => '&#80;MID',
1091 '|' => '&#124;',
1092 '__' => '&#95;_',
1093 ] );
1095 # Stupid hack
1096 $encValue = preg_replace_callback(
1097 '/((?i)' . wfUrlProtocols() . ')/',
1098 [ 'Sanitizer', 'armorLinksCallback' ],
1099 $encValue );
1100 return $encValue;
1104 * Given a value, escape it so that it can be used in an id attribute and
1105 * return it. This will use HTML5 validation if $wgExperimentalHtmlIds is
1106 * true, allowing anything but ASCII whitespace. Otherwise it will use
1107 * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1108 * escaped with lots of dots.
1110 * To ensure we don't have to bother escaping anything, we also strip ', ",
1111 * & even if $wgExperimentalIds is true. TODO: Is this the best tactic?
1112 * We also strip # because it upsets IE, and % because it could be
1113 * ambiguous if it's part of something that looks like a percent escape
1114 * (which don't work reliably in fragments cross-browser).
1116 * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1117 * in the id and name attributes
1118 * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1119 * the id attribute
1120 * @see http://www.whatwg.org/html/elements.html#the-id-attribute
1121 * HTML5 definition of id attribute
1123 * @param string $id Id to escape
1124 * @param string|array $options String or array of strings (default is array()):
1125 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1126 * so don't pay attention if the first character isn't valid at the
1127 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
1128 * false.
1129 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
1130 * if $wgExperimentalHtmlIds is used, so we can generate extra
1131 * anchors and links won't break.
1132 * @return string
1134 static function escapeId( $id, $options = [] ) {
1135 global $wgExperimentalHtmlIds;
1136 $options = (array)$options;
1138 $id = Sanitizer::decodeCharReferences( $id );
1140 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1141 $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1142 $id = trim( $id, '_' );
1143 if ( $id === '' ) {
1144 // Must have been all whitespace to start with.
1145 return '_';
1146 } else {
1147 return $id;
1151 // HTML4-style escaping
1152 static $replace = [
1153 '%3A' => ':',
1154 '%' => '.'
1157 $id = urlencode( strtr( $id, ' ', '_' ) );
1158 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1160 if ( !preg_match( '/^[a-zA-Z]/', $id ) && !in_array( 'noninitial', $options ) ) {
1161 // Initial character must be a letter!
1162 $id = "x$id";
1164 return $id;
1168 * Given a string containing a space delimited list of ids, escape each id
1169 * to match ids escaped by the escapeId() function.
1171 * @since 1.27
1173 * @param string $referenceString Space delimited list of ids
1174 * @param string|array $options String or array of strings (default is array()):
1175 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1176 * so don't pay attention if the first character isn't valid at the
1177 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
1178 * false.
1179 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
1180 * if $wgExperimentalHtmlIds is used, so we can generate extra
1181 * anchors and links won't break.
1182 * @return string
1184 static function escapeIdReferenceList( $referenceString, $options = [] ) {
1185 # Explode the space delimited list string into an array of tokens
1186 $references = preg_split( '/\s+/', "{$referenceString}", -1, PREG_SPLIT_NO_EMPTY );
1188 # Escape each token as an id
1189 foreach ( $references as &$ref ) {
1190 $ref = Sanitizer::escapeId( $ref, $options );
1193 # Merge the array back to a space delimited list string
1194 # If the array is empty, the result will be an empty string ('')
1195 $referenceString = implode( ' ', $references );
1197 return $referenceString;
1201 * Given a value, escape it so that it can be used as a CSS class and
1202 * return it.
1204 * @todo For extra validity, input should be validated UTF-8.
1206 * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1208 * @param string $class
1209 * @return string
1211 static function escapeClass( $class ) {
1212 // Convert ugly stuff to underscores and kill underscores in ugly places
1213 return rtrim( preg_replace(
1214 [ '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ],
1215 '_',
1216 $class ), '_' );
1220 * Given HTML input, escape with htmlspecialchars but un-escape entities.
1221 * This allows (generally harmless) entities like &#160; to survive.
1223 * @param string $html HTML to escape
1224 * @return string Escaped input
1226 static function escapeHtmlAllowEntities( $html ) {
1227 $html = Sanitizer::decodeCharReferences( $html );
1228 # It seems wise to escape ' as well as ", as a matter of course. Can't
1229 # hurt.
1230 $html = htmlspecialchars( $html, ENT_QUOTES );
1231 return $html;
1235 * Regex replace callback for armoring links against further processing.
1236 * @param array $matches
1237 * @return string
1239 private static function armorLinksCallback( $matches ) {
1240 return str_replace( ':', '&#58;', $matches[1] );
1244 * Return an associative array of attribute names and values from
1245 * a partial tag string. Attribute names are forced to lowercase,
1246 * character references are decoded to UTF-8 text.
1248 * @param string $text
1249 * @return array
1251 public static function decodeTagAttributes( $text ) {
1252 if ( trim( $text ) == '' ) {
1253 return [];
1256 $attribs = [];
1257 $pairs = [];
1258 if ( !preg_match_all(
1259 self::getAttribsRegex(),
1260 $text,
1261 $pairs,
1262 PREG_SET_ORDER ) ) {
1263 return $attribs;
1266 foreach ( $pairs as $set ) {
1267 $attribute = strtolower( $set[1] );
1268 $value = Sanitizer::getTagAttributeCallback( $set );
1270 // Normalize whitespace
1271 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1272 $value = trim( $value );
1274 // Decode character references
1275 $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1277 return $attribs;
1281 * Build a partial tag string from an associative array of attribute
1282 * names and values as returned by decodeTagAttributes.
1284 * @param array $assoc_array
1285 * @return string
1287 public static function safeEncodeTagAttributes( $assoc_array ) {
1288 $attribs = [];
1289 foreach ( $assoc_array as $attribute => $value ) {
1290 $encAttribute = htmlspecialchars( $attribute );
1291 $encValue = Sanitizer::safeEncodeAttribute( $value );
1293 $attribs[] = "$encAttribute=\"$encValue\"";
1295 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1299 * Pick the appropriate attribute value from a match set from the
1300 * attribs regex matches.
1302 * @param array $set
1303 * @throws MWException When tag conditions are not met.
1304 * @return string
1306 private static function getTagAttributeCallback( $set ) {
1307 if ( isset( $set[5] ) ) {
1308 # No quotes.
1309 return $set[5];
1310 } elseif ( isset( $set[4] ) ) {
1311 # Single-quoted
1312 return $set[4];
1313 } elseif ( isset( $set[3] ) ) {
1314 # Double-quoted
1315 return $set[3];
1316 } elseif ( !isset( $set[2] ) ) {
1317 # In XHTML, attributes must have a value so return an empty string.
1318 # See "Empty attribute syntax",
1319 # http://www.w3.org/TR/html5/syntax.html#syntax-attribute-name
1320 return "";
1321 } else {
1322 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1327 * @param string $text
1328 * @return string
1330 private static function normalizeWhitespace( $text ) {
1331 return preg_replace(
1332 '/\r\n|[\x20\x0d\x0a\x09]/',
1333 ' ',
1334 $text );
1338 * Normalizes whitespace in a section name, such as might be returned
1339 * by Parser::stripSectionName(), for use in the id's that are used for
1340 * section links.
1342 * @param string $section
1343 * @return string
1345 static function normalizeSectionNameWhitespace( $section ) {
1346 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1350 * Ensure that any entities and character references are legal
1351 * for XML and XHTML specifically. Any stray bits will be
1352 * &amp;-escaped to result in a valid text fragment.
1354 * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1355 * numericized (this way we're well-formed even without a DTD)
1356 * b. any numeric char refs must be legal chars, not invalid or forbidden
1357 * c. use lower cased "&#x", not "&#X"
1358 * d. fix or reject non-valid attributes
1360 * @param string $text
1361 * @return string
1362 * @private
1364 static function normalizeCharReferences( $text ) {
1365 return preg_replace_callback(
1366 self::CHAR_REFS_REGEX,
1367 [ 'Sanitizer', 'normalizeCharReferencesCallback' ],
1368 $text );
1372 * @param string $matches
1373 * @return string
1375 static function normalizeCharReferencesCallback( $matches ) {
1376 $ret = null;
1377 if ( $matches[1] != '' ) {
1378 $ret = Sanitizer::normalizeEntity( $matches[1] );
1379 } elseif ( $matches[2] != '' ) {
1380 $ret = Sanitizer::decCharReference( $matches[2] );
1381 } elseif ( $matches[3] != '' ) {
1382 $ret = Sanitizer::hexCharReference( $matches[3] );
1384 if ( is_null( $ret ) ) {
1385 return htmlspecialchars( $matches[0] );
1386 } else {
1387 return $ret;
1392 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1393 * return the equivalent numeric entity reference (except for the core &lt;
1394 * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1395 * the HTML equivalent. Otherwise, returns HTML-escaped text of
1396 * pseudo-entity source (eg &amp;foo;)
1398 * @param string $name
1399 * @return string
1401 static function normalizeEntity( $name ) {
1402 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1403 return '&' . self::$htmlEntityAliases[$name] . ';';
1404 } elseif ( in_array( $name, [ 'lt', 'gt', 'amp', 'quot' ] ) ) {
1405 return "&$name;";
1406 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1407 return '&#' . self::$htmlEntities[$name] . ';';
1408 } else {
1409 return "&amp;$name;";
1414 * @param int $codepoint
1415 * @return null|string
1417 static function decCharReference( $codepoint ) {
1418 $point = intval( $codepoint );
1419 if ( Sanitizer::validateCodepoint( $point ) ) {
1420 return sprintf( '&#%d;', $point );
1421 } else {
1422 return null;
1427 * @param int $codepoint
1428 * @return null|string
1430 static function hexCharReference( $codepoint ) {
1431 $point = hexdec( $codepoint );
1432 if ( Sanitizer::validateCodepoint( $point ) ) {
1433 return sprintf( '&#x%x;', $point );
1434 } else {
1435 return null;
1440 * Returns true if a given Unicode codepoint is a valid character in
1441 * both HTML5 and XML.
1442 * @param int $codepoint
1443 * @return bool
1445 private static function validateCodepoint( $codepoint ) {
1446 # U+000C is valid in HTML5 but not allowed in XML.
1447 # U+000D is valid in XML but not allowed in HTML5.
1448 # U+007F - U+009F are disallowed in HTML5 (control characters).
1449 return $codepoint == 0x09
1450 || $codepoint == 0x0a
1451 || ( $codepoint >= 0x20 && $codepoint <= 0x7e )
1452 || ( $codepoint >= 0xa0 && $codepoint <= 0xd7ff )
1453 || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1454 || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1458 * Decode any character references, numeric or named entities,
1459 * in the text and return a UTF-8 string.
1461 * @param string $text
1462 * @return string
1464 public static function decodeCharReferences( $text ) {
1465 return preg_replace_callback(
1466 self::CHAR_REFS_REGEX,
1467 [ 'Sanitizer', 'decodeCharReferencesCallback' ],
1468 $text );
1472 * Decode any character references, numeric or named entities,
1473 * in the next and normalize the resulting string. (bug 14952)
1475 * This is useful for page titles, not for text to be displayed,
1476 * MediaWiki allows HTML entities to escape normalization as a feature.
1478 * @param string $text Already normalized, containing entities
1479 * @return string Still normalized, without entities
1481 public static function decodeCharReferencesAndNormalize( $text ) {
1482 global $wgContLang;
1483 $text = preg_replace_callback(
1484 self::CHAR_REFS_REGEX,
1485 [ 'Sanitizer', 'decodeCharReferencesCallback' ],
1486 $text, /* limit */ -1, $count );
1488 if ( $count ) {
1489 return $wgContLang->normalize( $text );
1490 } else {
1491 return $text;
1496 * @param string $matches
1497 * @return string
1499 static function decodeCharReferencesCallback( $matches ) {
1500 if ( $matches[1] != '' ) {
1501 return Sanitizer::decodeEntity( $matches[1] );
1502 } elseif ( $matches[2] != '' ) {
1503 return Sanitizer::decodeChar( intval( $matches[2] ) );
1504 } elseif ( $matches[3] != '' ) {
1505 return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1507 # Last case should be an ampersand by itself
1508 return $matches[0];
1512 * Return UTF-8 string for a codepoint if that is a valid
1513 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1514 * @param int $codepoint
1515 * @return string
1516 * @private
1518 static function decodeChar( $codepoint ) {
1519 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1520 return UtfNormal\Utils::codepointToUtf8( $codepoint );
1521 } else {
1522 return UtfNormal\Constants::UTF8_REPLACEMENT;
1527 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1528 * return the UTF-8 encoding of that character. Otherwise, returns
1529 * pseudo-entity source (eg "&foo;")
1531 * @param string $name
1532 * @return string
1534 static function decodeEntity( $name ) {
1535 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1536 $name = self::$htmlEntityAliases[$name];
1538 if ( isset( self::$htmlEntities[$name] ) ) {
1539 return UtfNormal\Utils::codepointToUtf8( self::$htmlEntities[$name] );
1540 } else {
1541 return "&$name;";
1546 * Fetch the whitelist of acceptable attributes for a given element name.
1548 * @param string $element
1549 * @return array
1551 static function attributeWhitelist( $element ) {
1552 $list = Sanitizer::setupAttributeWhitelist();
1553 return isset( $list[$element] )
1554 ? $list[$element]
1555 : [];
1559 * Foreach array key (an allowed HTML element), return an array
1560 * of allowed attributes
1561 * @return array
1563 static function setupAttributeWhitelist() {
1564 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
1565 static $whitelist, $staticInitialised;
1567 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgAllowMicrodataAttributes' ) );
1569 if ( $whitelist !== null && $staticInitialised == $globalContext ) {
1570 return $whitelist;
1573 $common = [
1574 # HTML
1575 'id',
1576 'class',
1577 'style',
1578 'lang',
1579 'dir',
1580 'title',
1582 # WAI-ARIA
1583 'aria-describedby',
1584 'aria-flowto',
1585 'aria-label',
1586 'aria-labelledby',
1587 'aria-owns',
1588 'role',
1591 if ( $wgAllowRdfaAttributes ) {
1592 # RDFa attributes as specified in section 9 of
1593 # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1594 $common = array_merge( $common, [
1595 'about', 'property', 'resource', 'datatype', 'typeof',
1596 ] );
1599 if ( $wgAllowMicrodataAttributes ) {
1600 # add HTML5 microdata tags as specified by
1601 # http://www.whatwg.org/html/microdata.html#the-microdata-model
1602 $common = array_merge( $common, [
1603 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1604 ] );
1607 $block = array_merge( $common, [ 'align' ] );
1608 $tablealign = [ 'align', 'valign' ];
1609 $tablecell = [
1610 'abbr',
1611 'axis',
1612 'headers',
1613 'scope',
1614 'rowspan',
1615 'colspan',
1616 'nowrap', # deprecated
1617 'width', # deprecated
1618 'height', # deprecated
1619 'bgcolor', # deprecated
1622 # Numbers refer to sections in HTML 4.01 standard describing the element.
1623 # See: http://www.w3.org/TR/html4/
1624 $whitelist = [
1625 # 7.5.4
1626 'div' => $block,
1627 'center' => $common, # deprecated
1628 'span' => $common,
1630 # 7.5.5
1631 'h1' => $block,
1632 'h2' => $block,
1633 'h3' => $block,
1634 'h4' => $block,
1635 'h5' => $block,
1636 'h6' => $block,
1638 # 7.5.6
1639 # address
1641 # 8.2.4
1642 'bdo' => $common,
1644 # 9.2.1
1645 'em' => $common,
1646 'strong' => $common,
1647 'cite' => $common,
1648 'dfn' => $common,
1649 'code' => $common,
1650 'samp' => $common,
1651 'kbd' => $common,
1652 'var' => $common,
1653 'abbr' => $common,
1654 # acronym
1656 # 9.2.2
1657 'blockquote' => array_merge( $common, [ 'cite' ] ),
1658 'q' => array_merge( $common, [ 'cite' ] ),
1660 # 9.2.3
1661 'sub' => $common,
1662 'sup' => $common,
1664 # 9.3.1
1665 'p' => $block,
1667 # 9.3.2
1668 'br' => array_merge( $common, [ 'clear' ] ),
1670 # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1671 'wbr' => $common,
1673 # 9.3.4
1674 'pre' => array_merge( $common, [ 'width' ] ),
1676 # 9.4
1677 'ins' => array_merge( $common, [ 'cite', 'datetime' ] ),
1678 'del' => array_merge( $common, [ 'cite', 'datetime' ] ),
1680 # 10.2
1681 'ul' => array_merge( $common, [ 'type' ] ),
1682 'ol' => array_merge( $common, [ 'type', 'start', 'reversed' ] ),
1683 'li' => array_merge( $common, [ 'type', 'value' ] ),
1685 # 10.3
1686 'dl' => $common,
1687 'dd' => $common,
1688 'dt' => $common,
1690 # 11.2.1
1691 'table' => array_merge( $common,
1692 [ 'summary', 'width', 'border', 'frame',
1693 'rules', 'cellspacing', 'cellpadding',
1694 'align', 'bgcolor',
1695 ] ),
1697 # 11.2.2
1698 'caption' => $block,
1700 # 11.2.3
1701 'thead' => $common,
1702 'tfoot' => $common,
1703 'tbody' => $common,
1705 # 11.2.4
1706 'colgroup' => array_merge( $common, [ 'span' ] ),
1707 'col' => array_merge( $common, [ 'span' ] ),
1709 # 11.2.5
1710 'tr' => array_merge( $common, [ 'bgcolor' ], $tablealign ),
1712 # 11.2.6
1713 'td' => array_merge( $common, $tablecell, $tablealign ),
1714 'th' => array_merge( $common, $tablecell, $tablealign ),
1716 # 12.2
1717 # NOTE: <a> is not allowed directly, but the attrib
1718 # whitelist is used from the Parser object
1719 'a' => array_merge( $common, [ 'href', 'rel', 'rev' ] ), # rel/rev esp. for RDFa
1721 # 13.2
1722 # Not usually allowed, but may be used for extension-style hooks
1723 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1724 # true
1725 'img' => array_merge( $common, [ 'alt', 'src', 'width', 'height' ] ),
1727 # 15.2.1
1728 'tt' => $common,
1729 'b' => $common,
1730 'i' => $common,
1731 'big' => $common,
1732 'small' => $common,
1733 'strike' => $common,
1734 's' => $common,
1735 'u' => $common,
1737 # 15.2.2
1738 'font' => array_merge( $common, [ 'size', 'color', 'face' ] ),
1739 # basefont
1741 # 15.3
1742 'hr' => array_merge( $common, [ 'width' ] ),
1744 # HTML Ruby annotation text module, simple ruby only.
1745 # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
1746 'ruby' => $common,
1747 # rbc
1748 'rb' => $common,
1749 'rp' => $common,
1750 'rt' => $common, # array_merge( $common, array( 'rbspan' ) ),
1751 'rtc' => $common,
1753 # MathML root element, where used for extensions
1754 # 'title' may not be 100% valid here; it's XHTML
1755 # http://www.w3.org/TR/REC-MathML/
1756 'math' => [ 'class', 'style', 'id', 'title' ],
1758 # HTML 5 section 4.6
1759 'bdi' => $common,
1761 # HTML5 elements, defined by:
1762 # http://www.whatwg.org/html/
1763 'data' => array_merge( $common, [ 'value' ] ),
1764 'time' => array_merge( $common, [ 'datetime' ] ),
1765 'mark' => $common,
1767 // meta and link are only permitted by removeHTMLtags when Microdata
1768 // is enabled so we don't bother adding a conditional to hide these
1769 // Also meta and link are only valid in WikiText as Microdata elements
1770 // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1771 // So we don't bother including $common attributes that have no purpose.
1772 'meta' => [ 'itemprop', 'content' ],
1773 'link' => [ 'itemprop', 'href' ],
1776 $staticInitialised = $globalContext;
1778 return $whitelist;
1782 * Take a fragment of (potentially invalid) HTML and return
1783 * a version with any tags removed, encoded as plain text.
1785 * Warning: this return value must be further escaped for literal
1786 * inclusion in HTML output as of 1.10!
1788 * @param string $text HTML fragment
1789 * @return string
1791 static function stripAllTags( $text ) {
1792 # Actual <tags>
1793 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1795 # Normalize &entities and whitespace
1796 $text = self::decodeCharReferences( $text );
1797 $text = self::normalizeWhitespace( $text );
1799 return $text;
1803 * Hack up a private DOCTYPE with HTML's standard entity declarations.
1804 * PHP 4 seemed to know these if you gave it an HTML doctype, but
1805 * PHP 5.1 doesn't.
1807 * Use for passing XHTML fragments to PHP's XML parsing functions
1809 * @return string
1811 static function hackDocType() {
1812 $out = "<!DOCTYPE html [\n";
1813 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1814 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1816 $out .= "]>\n";
1817 return $out;
1821 * @param string $url
1822 * @return mixed|string
1824 static function cleanUrl( $url ) {
1825 # Normalize any HTML entities in input. They will be
1826 # re-escaped by makeExternalLink().
1827 $url = Sanitizer::decodeCharReferences( $url );
1829 # Escape any control characters introduced by the above step
1830 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1831 [ __CLASS__, 'cleanUrlCallback' ], $url );
1833 # Validate hostname portion
1834 $matches = [];
1835 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1836 list( /* $whole */, $protocol, $host, $rest ) = $matches;
1838 // Characters that will be ignored in IDNs.
1839 // http://tools.ietf.org/html/3454#section-3.1
1840 // Strip them before further processing so blacklists and such work.
1841 $strip = "/
1842 \\s| # general whitespace
1843 \xc2\xad| # 00ad SOFT HYPHEN
1844 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1845 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1846 \xe2\x81\xa0| # 2060 WORD JOINER
1847 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1848 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
1849 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1850 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1851 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1852 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1853 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1854 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1855 /xuD";
1857 $host = preg_replace( $strip, '', $host );
1859 // IPv6 host names are bracketed with []. Url-decode these.
1860 if ( substr_compare( "//%5B", $host, 0, 5 ) === 0 &&
1861 preg_match( '!^//%5B([0-9A-Fa-f:.]+)%5D((:\d+)?)$!', $host, $matches )
1863 $host = '//[' . $matches[1] . ']' . $matches[2];
1866 // @todo FIXME: Validate hostnames here
1868 return $protocol . $host . $rest;
1869 } else {
1870 return $url;
1875 * @param array $matches
1876 * @return string
1878 static function cleanUrlCallback( $matches ) {
1879 return urlencode( $matches[0] );
1883 * Does a string look like an e-mail address?
1885 * This validates an email address using an HTML5 specification found at:
1886 * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1887 * Which as of 2011-01-24 says:
1889 * A valid e-mail address is a string that matches the ABNF production
1890 * 1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1891 * in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1892 * 3.5.
1894 * This function is an implementation of the specification as requested in
1895 * bug 22449.
1897 * Client-side forms will use the same standard validation rules via JS or
1898 * HTML 5 validation; additional restrictions can be enforced server-side
1899 * by extensions via the 'isValidEmailAddr' hook.
1901 * Note that this validation doesn't 100% match RFC 2822, but is believed
1902 * to be liberal enough for wide use. Some invalid addresses will still
1903 * pass validation here.
1905 * @since 1.18
1907 * @param string $addr E-mail address
1908 * @return bool
1910 public static function validateEmail( $addr ) {
1911 $result = null;
1912 if ( !Hooks::run( 'isValidEmailAddr', [ $addr, &$result ] ) ) {
1913 return $result;
1916 // Please note strings below are enclosed in brackets [], this make the
1917 // hyphen "-" a range indicator. Hence it is double backslashed below.
1918 // See bug 26948
1919 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1920 $rfc1034_ldh_str = "a-z0-9\\-";
1922 $html5_email_regexp = "/
1923 ^ # start of string
1924 [$rfc5322_atext\\.]+ # user part which is liberal :p
1925 @ # 'apostrophe'
1926 [$rfc1034_ldh_str]+ # First domain part
1927 (\\.[$rfc1034_ldh_str]+)* # Following part prefixed with a dot
1928 $ # End of string
1929 /ix"; // case Insensitive, eXtended
1931 return (bool)preg_match( $html5_email_regexp, $addr );