Fix missing commit() flag in postgres savepoint class
[mediawiki.git] / includes / Sanitizer.php
blob40696583760d9b3b8ce518da1011365e89a3d4e9
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 $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 = $wgAllowImageTag;
374 if ( !$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', 'meta', 'link'
387 # Elements that cannot have close tags. This is (not coincidentally)
388 # also the list of tags for which the HTML 5 parsing algorithm
389 # requires you to "acknowledge the token's self-closing flag", i.e.
390 # a self-closing tag like <br/> is not an HTML 5 parse error only
391 # for this list.
392 $htmlsingleonly = [
393 'br', 'wbr', 'hr', 'meta', 'link'
396 $htmlnest = [ # Tags that can be nested--??
397 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
398 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
399 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
401 $tabletags = [ # Can only appear inside table, we will close them
402 'td', 'th', 'tr',
404 $htmllist = [ # Tags used by list
405 'ul', 'ol',
407 $listtags = [ # Tags that can appear in a list
408 'li',
411 if ( $wgAllowImageTag ) {
412 $htmlsingle[] = 'img';
413 $htmlsingleonly[] = 'img';
416 $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
417 $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
419 # Convert them all to hashtables for faster lookup
420 $vars = [ 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
421 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' ];
422 foreach ( $vars as $var ) {
423 $$var = array_flip( $$var );
425 $staticInitialised = $globalContext;
428 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
429 $extratags = array_flip( $extratags );
430 $removetags = array_flip( $removetags );
431 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
432 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
434 return [
435 'htmlpairs' => $htmlpairs,
436 'htmlsingle' => $htmlsingle,
437 'htmlsingleonly' => $htmlsingleonly,
438 'htmlnest' => $htmlnest,
439 'tabletags' => $tabletags,
440 'htmllist' => $htmllist,
441 'listtags' => $listtags,
442 'htmlsingleallowed' => $htmlsingleallowed,
443 'htmlelements' => $htmlelements,
448 * Cleans up HTML, removes dangerous tags and attributes, and
449 * removes HTML comments
450 * @param string $text
451 * @param callable $processCallback Callback to do any variable or parameter
452 * replacements in HTML attribute values
453 * @param array|bool $args Arguments for the processing callback
454 * @param array $extratags For any extra tags to include
455 * @param array $removetags For any tags (default or extra) to exclude
456 * @param callable $warnCallback (Deprecated) Callback allowing the
457 * addition of a tracking category when bad input is encountered.
458 * DO NOT ADD NEW PARAMETERS AFTER $warnCallback, since it will be
459 * removed shortly.
460 * @return string
462 public static function removeHTMLtags( $text, $processCallback = null,
463 $args = [], $extratags = [], $removetags = [], $warnCallback = null
465 extract( self::getRecognizedTagData( $extratags, $removetags ) );
467 # Remove HTML comments
468 $text = Sanitizer::removeHTMLcomments( $text );
469 $bits = explode( '<', $text );
470 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
471 if ( !MWTidy::isEnabled() ) {
472 $tagstack = $tablestack = [];
473 foreach ( $bits as $x ) {
474 $regs = [];
475 # $slash: Does the current element start with a '/'?
476 # $t: Current element name
477 # $params: String between element name and >
478 # $brace: Ending '>' or '/>'
479 # $rest: Everything until the next element of $bits
480 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
481 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
482 } else {
483 $slash = $t = $params = $brace = $rest = null;
486 $badtag = false;
487 $t = strtolower( $t );
488 if ( isset( $htmlelements[$t] ) ) {
489 # Check our stack
490 if ( $slash && isset( $htmlsingleonly[$t] ) ) {
491 $badtag = true;
492 } elseif ( $slash ) {
493 # Closing a tag... is it the one we just opened?
494 MediaWiki\suppressWarnings();
495 $ot = array_pop( $tagstack );
496 MediaWiki\restoreWarnings();
498 if ( $ot != $t ) {
499 if ( isset( $htmlsingleallowed[$ot] ) ) {
500 # Pop all elements with an optional close tag
501 # and see if we find a match below them
502 $optstack = [];
503 array_push( $optstack, $ot );
504 MediaWiki\suppressWarnings();
505 $ot = array_pop( $tagstack );
506 MediaWiki\restoreWarnings();
507 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
508 array_push( $optstack, $ot );
509 MediaWiki\suppressWarnings();
510 $ot = array_pop( $tagstack );
511 MediaWiki\restoreWarnings();
513 if ( $t != $ot ) {
514 # No match. Push the optional elements back again
515 $badtag = true;
516 MediaWiki\suppressWarnings();
517 $ot = array_pop( $optstack );
518 MediaWiki\restoreWarnings();
519 while ( $ot ) {
520 array_push( $tagstack, $ot );
521 MediaWiki\suppressWarnings();
522 $ot = array_pop( $optstack );
523 MediaWiki\restoreWarnings();
526 } else {
527 MediaWiki\suppressWarnings();
528 array_push( $tagstack, $ot );
529 MediaWiki\restoreWarnings();
531 # <li> can be nested in <ul> or <ol>, skip those cases:
532 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
533 $badtag = true;
536 } else {
537 if ( $t == 'table' ) {
538 $tagstack = array_pop( $tablestack );
541 $newparams = '';
542 } else {
543 # Keep track for later
544 if ( isset( $tabletags[$t] ) && !in_array( 'table', $tagstack ) ) {
545 $badtag = true;
546 } elseif ( in_array( $t, $tagstack ) && !isset( $htmlnest[$t] ) ) {
547 $badtag = true;
548 #  Is it a self closed htmlpair ? (bug 5487)
549 } elseif ( $brace == '/>' && isset( $htmlpairs[$t] ) ) {
550 // Eventually we'll just remove the self-closing
551 // slash, in order to be consistent with HTML5
552 // semantics.
553 // $brace = '>';
554 // For now, let's just warn authors to clean up.
555 if ( is_callable( $warnCallback ) ) {
556 call_user_func_array( $warnCallback, [ 'deprecated-self-close-category' ] );
558 $badtag = true;
559 } elseif ( isset( $htmlsingleonly[$t] ) ) {
560 # Hack to force empty tag for unclosable elements
561 $brace = '/>';
562 } elseif ( isset( $htmlsingle[$t] ) ) {
563 # Hack to not close $htmlsingle tags
564 $brace = null;
565 # Still need to push this optionally-closed tag to
566 # the tag stack so that we can match end tags
567 # instead of marking them as bad.
568 array_push( $tagstack, $t );
569 } elseif ( isset( $tabletags[$t] ) && in_array( $t, $tagstack ) ) {
570 // New table tag but forgot to close the previous one
571 $text .= "</$t>";
572 } else {
573 if ( $t == 'table' ) {
574 array_push( $tablestack, $tagstack );
575 $tagstack = [];
577 array_push( $tagstack, $t );
580 # Replace any variables or template parameters with
581 # plaintext results.
582 if ( is_callable( $processCallback ) ) {
583 call_user_func_array( $processCallback, [ &$params, $args ] );
586 if ( !Sanitizer::validateTag( $params, $t ) ) {
587 $badtag = true;
590 # Strip non-approved attributes from the tag
591 $newparams = Sanitizer::fixTagAttributes( $params, $t );
593 if ( !$badtag ) {
594 $rest = str_replace( '>', '&gt;', $rest );
595 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
596 $text .= "<$slash$t$newparams$close>$rest";
597 continue;
600 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
602 # Close off any remaining tags
603 while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
604 $text .= "</$t>\n";
605 if ( $t == 'table' ) {
606 $tagstack = array_pop( $tablestack );
609 } else {
610 # this might be possible using tidy itself
611 foreach ( $bits as $x ) {
612 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
613 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
615 $badtag = false;
616 $t = strtolower( $t );
617 if ( isset( $htmlelements[$t] ) ) {
618 if ( is_callable( $processCallback ) ) {
619 call_user_func_array( $processCallback, [ &$params, $args ] );
622 if ( $brace == '/>' && !( isset( $htmlsingle[$t] ) || isset( $htmlsingleonly[$t] ) ) ) {
623 // Eventually we'll just remove the self-closing
624 // slash, in order to be consistent with HTML5
625 // semantics.
626 // $brace = '>';
627 // For now, let's just warn authors to clean up.
628 if ( is_callable( $warnCallback ) ) {
629 call_user_func_array( $warnCallback, [ 'deprecated-self-close-category' ] );
632 if ( !Sanitizer::validateTag( $params, $t ) ) {
633 $badtag = true;
636 $newparams = Sanitizer::fixTagAttributes( $params, $t );
637 if ( !$badtag ) {
638 if ( $brace === '/>' && !isset( $htmlsingleonly[$t] ) ) {
639 # Interpret self-closing tags as empty tags even when
640 # HTML 5 would interpret them as start tags. Such input
641 # is commonly seen on Wikimedia wikis with this intention.
642 $brace = "></$t>";
645 $rest = str_replace( '>', '&gt;', $rest );
646 $text .= "<$slash$t$newparams$brace$rest";
647 continue;
651 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
654 return $text;
658 * Remove '<!--', '-->', and everything between.
659 * To avoid leaving blank lines, when a comment is both preceded
660 * and followed by a newline (ignoring spaces), trim leading and
661 * trailing spaces and one of the newlines.
663 * @param string $text
664 * @return string
666 public static function removeHTMLcomments( $text ) {
667 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
668 $end = strpos( $text, '-->', $start + 4 );
669 if ( $end === false ) {
670 # Unterminated comment; bail out
671 break;
674 $end += 3;
676 # Trim space and newline if the comment is both
677 # preceded and followed by a newline
678 $spaceStart = max( $start - 1, 0 );
679 $spaceLen = $end - $spaceStart;
680 while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
681 $spaceStart--;
682 $spaceLen++;
684 while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
685 $spaceLen++;
687 if ( substr( $text, $spaceStart, 1 ) === "\n"
688 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
689 # Remove the comment, leading and trailing
690 # spaces, and leave only one newline.
691 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
692 } else {
693 # Remove just the comment.
694 $text = substr_replace( $text, '', $start, $end - $start );
697 return $text;
701 * Takes attribute names and values for a tag and the tag name and
702 * validates that the tag is allowed to be present.
703 * This DOES NOT validate the attributes, nor does it validate the
704 * tags themselves. This method only handles the special circumstances
705 * where we may want to allow a tag within content but ONLY when it has
706 * specific attributes set.
708 * @param string $params
709 * @param string $element
710 * @return bool
712 static function validateTag( $params, $element ) {
713 $params = Sanitizer::decodeTagAttributes( $params );
715 if ( $element == 'meta' || $element == 'link' ) {
716 if ( !isset( $params['itemprop'] ) ) {
717 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
718 return false;
720 if ( $element == 'meta' && !isset( $params['content'] ) ) {
721 // <meta> must have a content="" for the itemprop
722 return false;
724 if ( $element == 'link' && !isset( $params['href'] ) ) {
725 // <link> must have an associated href=""
726 return false;
730 return true;
734 * Take an array of attribute names and values and normalize or discard
735 * illegal values for the given element type.
737 * - Discards attributes not on a whitelist for the given element
738 * - Unsafe style attributes are discarded
739 * - Invalid id attributes are re-encoded
741 * @param array $attribs
742 * @param string $element
743 * @return array
745 * @todo Check for legal values where the DTD limits things.
746 * @todo Check for unique id attribute :P
748 static function validateTagAttributes( $attribs, $element ) {
749 return Sanitizer::validateAttributes( $attribs,
750 Sanitizer::attributeWhitelist( $element ) );
754 * Take an array of attribute names and values and normalize or discard
755 * illegal values for the given whitelist.
757 * - Discards attributes not on the given whitelist
758 * - Unsafe style attributes are discarded
759 * - Invalid id attributes are re-encoded
761 * @param array $attribs
762 * @param array $whitelist List of allowed attribute names
763 * @return array
765 * @todo Check for legal values where the DTD limits things.
766 * @todo Check for unique id attribute :P
768 static function validateAttributes( $attribs, $whitelist ) {
769 $whitelist = array_flip( $whitelist );
770 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
772 $out = [];
773 foreach ( $attribs as $attribute => $value ) {
774 # Allow XML namespace declaration to allow RDFa
775 if ( preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
776 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
777 $out[$attribute] = $value;
780 continue;
783 # Allow any attribute beginning with "data-"
784 # However:
785 # * data-ooui is reserved for ooui
786 # * data-mw and data-parsoid are reserved for parsoid
787 # * data-mw-<name here> is reserved for extensions (or core) if
788 # they need to communicate some data to the client and want to be
789 # sure that it isn't coming from an untrusted user.
790 # * Ensure that the attribute is not namespaced by banning
791 # colons.
792 if ( !preg_match( '/^data-(?!ooui|mw|parsoid)[^:]*$/i', $attribute )
793 && !isset( $whitelist[$attribute] )
795 continue;
798 # Strip javascript "expression" from stylesheets.
799 # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
800 if ( $attribute == 'style' ) {
801 $value = Sanitizer::checkCss( $value );
804 # Escape HTML id attributes
805 if ( $attribute === 'id' ) {
806 $value = Sanitizer::escapeId( $value, 'noninitial' );
809 # Escape HTML id reference lists
810 if ( $attribute === 'aria-describedby'
811 || $attribute === 'aria-flowto'
812 || $attribute === 'aria-labelledby'
813 || $attribute === 'aria-owns'
815 $value = Sanitizer::escapeIdReferenceList( $value, 'noninitial' );
818 // RDFa and microdata properties allow URLs, URIs and/or CURIs.
819 // Check them for sanity.
820 if ( $attribute === 'rel' || $attribute === 'rev'
821 # RDFa
822 || $attribute === 'about' || $attribute === 'property'
823 || $attribute === 'resource' || $attribute === 'datatype'
824 || $attribute === 'typeof'
825 # HTML5 microdata
826 || $attribute === 'itemid' || $attribute === 'itemprop'
827 || $attribute === 'itemref' || $attribute === 'itemscope'
828 || $attribute === 'itemtype'
830 // Paranoia. Allow "simple" values but suppress javascript
831 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
832 continue;
836 # NOTE: even though elements using href/src are not allowed directly, supply
837 # validation code that can be used by tag hook handlers, etc
838 if ( $attribute === 'href' || $attribute === 'src' ) {
839 if ( !preg_match( $hrefExp, $value ) ) {
840 continue; // drop any href or src attributes not using an allowed protocol.
841 // NOTE: this also drops all relative URLs
845 // If this attribute was previously set, override it.
846 // Output should only have one attribute of each name.
847 $out[$attribute] = $value;
850 # itemtype, itemid, itemref don't make sense without itemscope
851 if ( !array_key_exists( 'itemscope', $out ) ) {
852 unset( $out['itemtype'] );
853 unset( $out['itemid'] );
854 unset( $out['itemref'] );
856 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
858 return $out;
862 * Merge two sets of HTML attributes. Conflicting items in the second set
863 * will override those in the first, except for 'class' attributes which
864 * will be combined (if they're both strings).
866 * @todo implement merging for other attributes such as style
867 * @param array $a
868 * @param array $b
869 * @return array
871 static function mergeAttributes( $a, $b ) {
872 $out = array_merge( $a, $b );
873 if ( isset( $a['class'] ) && isset( $b['class'] )
874 && is_string( $a['class'] ) && is_string( $b['class'] )
875 && $a['class'] !== $b['class']
877 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
878 -1, PREG_SPLIT_NO_EMPTY );
879 $out['class'] = implode( ' ', array_unique( $classes ) );
881 return $out;
885 * Normalize CSS into a format we can easily search for hostile input
886 * - decode character references
887 * - decode escape sequences
888 * - convert characters that IE6 interprets into ascii
889 * - remove comments, unless the entire value is one single comment
890 * @param string $value the css string
891 * @return string normalized css
893 public static function normalizeCss( $value ) {
895 // Decode character references like &#123;
896 $value = Sanitizer::decodeCharReferences( $value );
898 // Decode escape sequences and line continuation
899 // See the grammar in the CSS 2 spec, appendix D.
900 // This has to be done AFTER decoding character references.
901 // This means it isn't possible for this function to return
902 // unsanitized escape sequences. It is possible to manufacture
903 // input that contains character references that decode to
904 // escape sequences that decode to character references, but
905 // it's OK for the return value to contain character references
906 // because the caller is supposed to escape those anyway.
907 static $decodeRegex;
908 if ( !$decodeRegex ) {
909 $space = '[\\x20\\t\\r\\n\\f]';
910 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
911 $backslash = '\\\\';
912 $decodeRegex = "/ $backslash
914 ($nl) | # 1. Line continuation
915 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
916 (.) | # 3. backslash cancelling special meaning
917 () | # 4. backslash at end of string
918 )/xu";
920 $value = preg_replace_callback( $decodeRegex,
921 [ __CLASS__, 'cssDecodeCallback' ], $value );
923 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
924 $value = preg_replace_callback(
925 '/[!-[]-z]/u', // U+FF01 to U+FF5A, excluding U+FF3C (bug 58088)
926 function ( $matches ) {
927 $cp = UtfNormal\Utils::utf8ToCodepoint( $matches[0] );
928 if ( $cp === false ) {
929 return '';
931 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
933 $value
936 // Convert more characters IE6 might treat as ascii
937 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
938 $value = str_replace(
939 [ 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ],
940 [ 'r', 'n', 'n', 'l', 'i', '(', '(' ],
941 $value
944 // Let the value through if it's nothing but a single comment, to
945 // allow other functions which may reject it to pass some error
946 // message through.
947 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
948 // Remove any comments; IE gets token splitting wrong
949 // This must be done AFTER decoding character references and
950 // escape sequences, because those steps can introduce comments
951 // This step cannot introduce character references or escape
952 // sequences, because it replaces comments with spaces rather
953 // than removing them completely.
954 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
956 // Remove anything after a comment-start token, to guard against
957 // incorrect client implementations.
958 $commentPos = strpos( $value, '/*' );
959 if ( $commentPos !== false ) {
960 $value = substr( $value, 0, $commentPos );
964 // S followed by repeat, iteration, or prolonged sound marks,
965 // which IE will treat as "ss"
966 $value = preg_replace(
967 '/s(?:
968 \xE3\x80\xB1 | # U+3031
969 \xE3\x82\x9D | # U+309D
970 \xE3\x83\xBC | # U+30FC
971 \xE3\x83\xBD | # U+30FD
972 \xEF\xB9\xBC | # U+FE7C
973 \xEF\xB9\xBD | # U+FE7D
974 \xEF\xBD\xB0 # U+FF70
975 )/ix',
976 'ss',
977 $value
980 return $value;
984 * Pick apart some CSS and check it for forbidden or unsafe structures.
985 * Returns a sanitized string. This sanitized string will have
986 * character references and escape sequences decoded and comments
987 * stripped (unless it is itself one valid comment, in which case the value
988 * will be passed through). If the input is just too evil, only a comment
989 * complaining about evilness will be returned.
991 * Currently URL references, 'expression', 'tps' are forbidden.
993 * NOTE: Despite the fact that character references are decoded, the
994 * returned string may contain character references given certain
995 * clever input strings. These character references must
996 * be escaped before the return value is embedded in HTML.
998 * @param string $value
999 * @return string
1001 static function checkCss( $value ) {
1002 $value = self::normalizeCss( $value );
1004 // Reject problematic keywords and control characters
1005 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ||
1006 strpos( $value, UtfNormal\Constants::UTF8_REPLACEMENT ) !== false ) {
1007 return '/* invalid control char */';
1008 } elseif ( preg_match(
1009 '! expression
1010 | filter\s*:
1011 | accelerator\s*:
1012 | -o-link\s*:
1013 | -o-link-source\s*:
1014 | -o-replace\s*:
1015 | url\s*\(
1016 | image\s*\(
1017 | image-set\s*\(
1018 | attr\s*\([^)]+[\s,]+url
1019 !ix', $value ) ) {
1020 return '/* insecure input */';
1022 return $value;
1026 * @param array $matches
1027 * @return string
1029 static function cssDecodeCallback( $matches ) {
1030 if ( $matches[1] !== '' ) {
1031 // Line continuation
1032 return '';
1033 } elseif ( $matches[2] !== '' ) {
1034 $char = UtfNormal\Utils::codepointToUtf8( hexdec( $matches[2] ) );
1035 } elseif ( $matches[3] !== '' ) {
1036 $char = $matches[3];
1037 } else {
1038 $char = '\\';
1040 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
1041 // These characters need to be escaped in strings
1042 // Clean up the escape sequence to avoid parsing errors by clients
1043 return '\\' . dechex( ord( $char ) ) . ' ';
1044 } else {
1045 // Decode unnecessary escape
1046 return $char;
1051 * Take a tag soup fragment listing an HTML element's attributes
1052 * and normalize it to well-formed XML, discarding unwanted attributes.
1053 * Output is safe for further wikitext processing, with escaping of
1054 * values that could trigger problems.
1056 * - Normalizes attribute names to lowercase
1057 * - Discards attributes not on a whitelist for the given element
1058 * - Turns broken or invalid entities into plaintext
1059 * - Double-quotes all attribute values
1060 * - Attributes without values are given the name as attribute
1061 * - Double attributes are discarded
1062 * - Unsafe style attributes are discarded
1063 * - Prepends space if there are attributes.
1064 * - (Optionally) Sorts attributes by name.
1066 * @param string $text
1067 * @param string $element
1068 * @param bool $sorted Whether to sort the attributes (default: false)
1069 * @return string
1071 static function fixTagAttributes( $text, $element, $sorted = false ) {
1072 if ( trim( $text ) == '' ) {
1073 return '';
1076 $decoded = Sanitizer::decodeTagAttributes( $text );
1077 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1079 if ( $sorted ) {
1080 ksort( $stripped );
1083 return Sanitizer::safeEncodeTagAttributes( $stripped );
1087 * Encode an attribute value for HTML output.
1088 * @param string $text
1089 * @return string HTML-encoded text fragment
1091 static function encodeAttribute( $text ) {
1092 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1094 // Whitespace is normalized during attribute decoding,
1095 // so if we've been passed non-spaces we must encode them
1096 // ahead of time or they won't be preserved.
1097 $encValue = strtr( $encValue, [
1098 "\n" => '&#10;',
1099 "\r" => '&#13;',
1100 "\t" => '&#9;',
1101 ] );
1103 return $encValue;
1107 * Encode an attribute value for HTML tags, with extra armoring
1108 * against further wiki processing.
1109 * @param string $text
1110 * @return string HTML-encoded text fragment
1112 static function safeEncodeAttribute( $text ) {
1113 $encValue = Sanitizer::encodeAttribute( $text );
1115 # Templates and links may be expanded in later parsing,
1116 # creating invalid or dangerous output. Suppress this.
1117 $encValue = strtr( $encValue, [
1118 '<' => '&lt;', // This should never happen,
1119 '>' => '&gt;', // we've received invalid input
1120 '"' => '&quot;', // which should have been escaped.
1121 '{' => '&#123;',
1122 '[' => '&#91;',
1123 "''" => '&#39;&#39;',
1124 'ISBN' => '&#73;SBN',
1125 'RFC' => '&#82;FC',
1126 'PMID' => '&#80;MID',
1127 '|' => '&#124;',
1128 '__' => '&#95;_',
1129 ] );
1131 # Stupid hack
1132 $encValue = preg_replace_callback(
1133 '/((?i)' . wfUrlProtocols() . ')/',
1134 [ 'Sanitizer', 'armorLinksCallback' ],
1135 $encValue );
1136 return $encValue;
1140 * Given a value, escape it so that it can be used in an id attribute and
1141 * return it. This will use HTML5 validation if $wgExperimentalHtmlIds is
1142 * true, allowing anything but ASCII whitespace. Otherwise it will use
1143 * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1144 * escaped with lots of dots.
1146 * To ensure we don't have to bother escaping anything, we also strip ', ",
1147 * & even if $wgExperimentalIds is true. TODO: Is this the best tactic?
1148 * We also strip # because it upsets IE, and % because it could be
1149 * ambiguous if it's part of something that looks like a percent escape
1150 * (which don't work reliably in fragments cross-browser).
1152 * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1153 * in the id and name attributes
1154 * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1155 * the id attribute
1156 * @see http://www.whatwg.org/html/elements.html#the-id-attribute
1157 * HTML5 definition of id attribute
1159 * @param string $id Id to escape
1160 * @param string|array $options String or array of strings (default is array()):
1161 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1162 * so don't pay attention if the first character isn't valid at the
1163 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
1164 * false.
1165 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
1166 * if $wgExperimentalHtmlIds is used, so we can generate extra
1167 * anchors and links won't break.
1168 * @return string
1170 static function escapeId( $id, $options = [] ) {
1171 global $wgExperimentalHtmlIds;
1172 $options = (array)$options;
1174 $id = Sanitizer::decodeCharReferences( $id );
1176 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1177 $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1178 $id = trim( $id, '_' );
1179 if ( $id === '' ) {
1180 // Must have been all whitespace to start with.
1181 return '_';
1182 } else {
1183 return $id;
1187 // HTML4-style escaping
1188 static $replace = [
1189 '%3A' => ':',
1190 '%' => '.'
1193 $id = urlencode( strtr( $id, ' ', '_' ) );
1194 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1196 if ( !preg_match( '/^[a-zA-Z]/', $id ) && !in_array( 'noninitial', $options ) ) {
1197 // Initial character must be a letter!
1198 $id = "x$id";
1200 return $id;
1204 * Given a string containing a space delimited list of ids, escape each id
1205 * to match ids escaped by the escapeId() function.
1207 * @since 1.27
1209 * @param string $referenceString Space delimited list of ids
1210 * @param string|array $options String or array of strings (default is array()):
1211 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1212 * so don't pay attention if the first character isn't valid at the
1213 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
1214 * false.
1215 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
1216 * if $wgExperimentalHtmlIds is used, so we can generate extra
1217 * anchors and links won't break.
1218 * @return string
1220 static function escapeIdReferenceList( $referenceString, $options = [] ) {
1221 # Explode the space delimited list string into an array of tokens
1222 $references = preg_split( '/\s+/', "{$referenceString}", -1, PREG_SPLIT_NO_EMPTY );
1224 # Escape each token as an id
1225 foreach ( $references as &$ref ) {
1226 $ref = Sanitizer::escapeId( $ref, $options );
1229 # Merge the array back to a space delimited list string
1230 # If the array is empty, the result will be an empty string ('')
1231 $referenceString = implode( ' ', $references );
1233 return $referenceString;
1237 * Given a value, escape it so that it can be used as a CSS class and
1238 * return it.
1240 * @todo For extra validity, input should be validated UTF-8.
1242 * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1244 * @param string $class
1245 * @return string
1247 static function escapeClass( $class ) {
1248 // Convert ugly stuff to underscores and kill underscores in ugly places
1249 return rtrim( preg_replace(
1250 [ '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ],
1251 '_',
1252 $class ), '_' );
1256 * Given HTML input, escape with htmlspecialchars but un-escape entities.
1257 * This allows (generally harmless) entities like &#160; to survive.
1259 * @param string $html HTML to escape
1260 * @return string Escaped input
1262 static function escapeHtmlAllowEntities( $html ) {
1263 $html = Sanitizer::decodeCharReferences( $html );
1264 # It seems wise to escape ' as well as ", as a matter of course. Can't
1265 # hurt.
1266 $html = htmlspecialchars( $html, ENT_QUOTES );
1267 return $html;
1271 * Regex replace callback for armoring links against further processing.
1272 * @param array $matches
1273 * @return string
1275 private static function armorLinksCallback( $matches ) {
1276 return str_replace( ':', '&#58;', $matches[1] );
1280 * Return an associative array of attribute names and values from
1281 * a partial tag string. Attribute names are forced to lowercase,
1282 * character references are decoded to UTF-8 text.
1284 * @param string $text
1285 * @return array
1287 public static function decodeTagAttributes( $text ) {
1288 if ( trim( $text ) == '' ) {
1289 return [];
1292 $attribs = [];
1293 $pairs = [];
1294 if ( !preg_match_all(
1295 self::getAttribsRegex(),
1296 $text,
1297 $pairs,
1298 PREG_SET_ORDER ) ) {
1299 return $attribs;
1302 foreach ( $pairs as $set ) {
1303 $attribute = strtolower( $set[1] );
1304 $value = Sanitizer::getTagAttributeCallback( $set );
1306 // Normalize whitespace
1307 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1308 $value = trim( $value );
1310 // Decode character references
1311 $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1313 return $attribs;
1317 * Build a partial tag string from an associative array of attribute
1318 * names and values as returned by decodeTagAttributes.
1320 * @param array $assoc_array
1321 * @return string
1323 public static function safeEncodeTagAttributes( $assoc_array ) {
1324 $attribs = [];
1325 foreach ( $assoc_array as $attribute => $value ) {
1326 $encAttribute = htmlspecialchars( $attribute );
1327 $encValue = Sanitizer::safeEncodeAttribute( $value );
1329 $attribs[] = "$encAttribute=\"$encValue\"";
1331 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1335 * Pick the appropriate attribute value from a match set from the
1336 * attribs regex matches.
1338 * @param array $set
1339 * @throws MWException When tag conditions are not met.
1340 * @return string
1342 private static function getTagAttributeCallback( $set ) {
1343 if ( isset( $set[5] ) ) {
1344 # No quotes.
1345 return $set[5];
1346 } elseif ( isset( $set[4] ) ) {
1347 # Single-quoted
1348 return $set[4];
1349 } elseif ( isset( $set[3] ) ) {
1350 # Double-quoted
1351 return $set[3];
1352 } elseif ( !isset( $set[2] ) ) {
1353 # In XHTML, attributes must have a value so return an empty string.
1354 # See "Empty attribute syntax",
1355 # http://www.w3.org/TR/html5/syntax.html#syntax-attribute-name
1356 return "";
1357 } else {
1358 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1363 * @param string $text
1364 * @return string
1366 private static function normalizeWhitespace( $text ) {
1367 return preg_replace(
1368 '/\r\n|[\x20\x0d\x0a\x09]/',
1369 ' ',
1370 $text );
1374 * Normalizes whitespace in a section name, such as might be returned
1375 * by Parser::stripSectionName(), for use in the id's that are used for
1376 * section links.
1378 * @param string $section
1379 * @return string
1381 static function normalizeSectionNameWhitespace( $section ) {
1382 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1386 * Ensure that any entities and character references are legal
1387 * for XML and XHTML specifically. Any stray bits will be
1388 * &amp;-escaped to result in a valid text fragment.
1390 * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1391 * numericized (this way we're well-formed even without a DTD)
1392 * b. any numeric char refs must be legal chars, not invalid or forbidden
1393 * c. use lower cased "&#x", not "&#X"
1394 * d. fix or reject non-valid attributes
1396 * @param string $text
1397 * @return string
1398 * @private
1400 static function normalizeCharReferences( $text ) {
1401 return preg_replace_callback(
1402 self::CHAR_REFS_REGEX,
1403 [ 'Sanitizer', 'normalizeCharReferencesCallback' ],
1404 $text );
1408 * @param string $matches
1409 * @return string
1411 static function normalizeCharReferencesCallback( $matches ) {
1412 $ret = null;
1413 if ( $matches[1] != '' ) {
1414 $ret = Sanitizer::normalizeEntity( $matches[1] );
1415 } elseif ( $matches[2] != '' ) {
1416 $ret = Sanitizer::decCharReference( $matches[2] );
1417 } elseif ( $matches[3] != '' ) {
1418 $ret = Sanitizer::hexCharReference( $matches[3] );
1420 if ( is_null( $ret ) ) {
1421 return htmlspecialchars( $matches[0] );
1422 } else {
1423 return $ret;
1428 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1429 * return the equivalent numeric entity reference (except for the core &lt;
1430 * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1431 * the HTML equivalent. Otherwise, returns HTML-escaped text of
1432 * pseudo-entity source (eg &amp;foo;)
1434 * @param string $name
1435 * @return string
1437 static function normalizeEntity( $name ) {
1438 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1439 return '&' . self::$htmlEntityAliases[$name] . ';';
1440 } elseif ( in_array( $name, [ 'lt', 'gt', 'amp', 'quot' ] ) ) {
1441 return "&$name;";
1442 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1443 return '&#' . self::$htmlEntities[$name] . ';';
1444 } else {
1445 return "&amp;$name;";
1450 * @param int $codepoint
1451 * @return null|string
1453 static function decCharReference( $codepoint ) {
1454 $point = intval( $codepoint );
1455 if ( Sanitizer::validateCodepoint( $point ) ) {
1456 return sprintf( '&#%d;', $point );
1457 } else {
1458 return null;
1463 * @param int $codepoint
1464 * @return null|string
1466 static function hexCharReference( $codepoint ) {
1467 $point = hexdec( $codepoint );
1468 if ( Sanitizer::validateCodepoint( $point ) ) {
1469 return sprintf( '&#x%x;', $point );
1470 } else {
1471 return null;
1476 * Returns true if a given Unicode codepoint is a valid character in
1477 * both HTML5 and XML.
1478 * @param int $codepoint
1479 * @return bool
1481 private static function validateCodepoint( $codepoint ) {
1482 # U+000C is valid in HTML5 but not allowed in XML.
1483 # U+000D is valid in XML but not allowed in HTML5.
1484 # U+007F - U+009F are disallowed in HTML5 (control characters).
1485 return $codepoint == 0x09
1486 || $codepoint == 0x0a
1487 || ( $codepoint >= 0x20 && $codepoint <= 0x7e )
1488 || ( $codepoint >= 0xa0 && $codepoint <= 0xd7ff )
1489 || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1490 || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1494 * Decode any character references, numeric or named entities,
1495 * in the text and return a UTF-8 string.
1497 * @param string $text
1498 * @return string
1500 public static function decodeCharReferences( $text ) {
1501 return preg_replace_callback(
1502 self::CHAR_REFS_REGEX,
1503 [ 'Sanitizer', 'decodeCharReferencesCallback' ],
1504 $text );
1508 * Decode any character references, numeric or named entities,
1509 * in the next and normalize the resulting string. (bug 14952)
1511 * This is useful for page titles, not for text to be displayed,
1512 * MediaWiki allows HTML entities to escape normalization as a feature.
1514 * @param string $text Already normalized, containing entities
1515 * @return string Still normalized, without entities
1517 public static function decodeCharReferencesAndNormalize( $text ) {
1518 global $wgContLang;
1519 $text = preg_replace_callback(
1520 self::CHAR_REFS_REGEX,
1521 [ 'Sanitizer', 'decodeCharReferencesCallback' ],
1522 $text, /* limit */ -1, $count );
1524 if ( $count ) {
1525 return $wgContLang->normalize( $text );
1526 } else {
1527 return $text;
1532 * @param string $matches
1533 * @return string
1535 static function decodeCharReferencesCallback( $matches ) {
1536 if ( $matches[1] != '' ) {
1537 return Sanitizer::decodeEntity( $matches[1] );
1538 } elseif ( $matches[2] != '' ) {
1539 return Sanitizer::decodeChar( intval( $matches[2] ) );
1540 } elseif ( $matches[3] != '' ) {
1541 return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1543 # Last case should be an ampersand by itself
1544 return $matches[0];
1548 * Return UTF-8 string for a codepoint if that is a valid
1549 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1550 * @param int $codepoint
1551 * @return string
1552 * @private
1554 static function decodeChar( $codepoint ) {
1555 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1556 return UtfNormal\Utils::codepointToUtf8( $codepoint );
1557 } else {
1558 return UtfNormal\Constants::UTF8_REPLACEMENT;
1563 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1564 * return the UTF-8 encoding of that character. Otherwise, returns
1565 * pseudo-entity source (eg "&foo;")
1567 * @param string $name
1568 * @return string
1570 static function decodeEntity( $name ) {
1571 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1572 $name = self::$htmlEntityAliases[$name];
1574 if ( isset( self::$htmlEntities[$name] ) ) {
1575 return UtfNormal\Utils::codepointToUtf8( self::$htmlEntities[$name] );
1576 } else {
1577 return "&$name;";
1582 * Fetch the whitelist of acceptable attributes for a given element name.
1584 * @param string $element
1585 * @return array
1587 static function attributeWhitelist( $element ) {
1588 $list = Sanitizer::setupAttributeWhitelist();
1589 return isset( $list[$element] )
1590 ? $list[$element]
1591 : [];
1595 * Foreach array key (an allowed HTML element), return an array
1596 * of allowed attributes
1597 * @return array
1599 static function setupAttributeWhitelist() {
1600 static $whitelist;
1602 if ( $whitelist !== null ) {
1603 return $whitelist;
1606 $common = [
1607 # HTML
1608 'id',
1609 'class',
1610 'style',
1611 'lang',
1612 'dir',
1613 'title',
1615 # WAI-ARIA
1616 'aria-describedby',
1617 'aria-flowto',
1618 'aria-label',
1619 'aria-labelledby',
1620 'aria-owns',
1621 'role',
1623 # RDFa
1624 # These attributes are specified in section 9 of
1625 # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1626 'about',
1627 'property',
1628 'resource',
1629 'datatype',
1630 'typeof',
1632 # Microdata. These are specified by
1633 # http://www.whatwg.org/html/microdata.html#the-microdata-model
1634 'itemid',
1635 'itemprop',
1636 'itemref',
1637 'itemscope',
1638 'itemtype',
1641 $block = array_merge( $common, [ 'align' ] );
1642 $tablealign = [ 'align', 'valign' ];
1643 $tablecell = [
1644 'abbr',
1645 'axis',
1646 'headers',
1647 'scope',
1648 'rowspan',
1649 'colspan',
1650 'nowrap', # deprecated
1651 'width', # deprecated
1652 'height', # deprecated
1653 'bgcolor', # deprecated
1656 # Numbers refer to sections in HTML 4.01 standard describing the element.
1657 # See: http://www.w3.org/TR/html4/
1658 $whitelist = [
1659 # 7.5.4
1660 'div' => $block,
1661 'center' => $common, # deprecated
1662 'span' => $common,
1664 # 7.5.5
1665 'h1' => $block,
1666 'h2' => $block,
1667 'h3' => $block,
1668 'h4' => $block,
1669 'h5' => $block,
1670 'h6' => $block,
1672 # 7.5.6
1673 # address
1675 # 8.2.4
1676 'bdo' => $common,
1678 # 9.2.1
1679 'em' => $common,
1680 'strong' => $common,
1681 'cite' => $common,
1682 'dfn' => $common,
1683 'code' => $common,
1684 'samp' => $common,
1685 'kbd' => $common,
1686 'var' => $common,
1687 'abbr' => $common,
1688 # acronym
1690 # 9.2.2
1691 'blockquote' => array_merge( $common, [ 'cite' ] ),
1692 'q' => array_merge( $common, [ 'cite' ] ),
1694 # 9.2.3
1695 'sub' => $common,
1696 'sup' => $common,
1698 # 9.3.1
1699 'p' => $block,
1701 # 9.3.2
1702 'br' => array_merge( $common, [ 'clear' ] ),
1704 # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1705 'wbr' => $common,
1707 # 9.3.4
1708 'pre' => array_merge( $common, [ 'width' ] ),
1710 # 9.4
1711 'ins' => array_merge( $common, [ 'cite', 'datetime' ] ),
1712 'del' => array_merge( $common, [ 'cite', 'datetime' ] ),
1714 # 10.2
1715 'ul' => array_merge( $common, [ 'type' ] ),
1716 'ol' => array_merge( $common, [ 'type', 'start', 'reversed' ] ),
1717 'li' => array_merge( $common, [ 'type', 'value' ] ),
1719 # 10.3
1720 'dl' => $common,
1721 'dd' => $common,
1722 'dt' => $common,
1724 # 11.2.1
1725 'table' => array_merge( $common,
1726 [ 'summary', 'width', 'border', 'frame',
1727 'rules', 'cellspacing', 'cellpadding',
1728 'align', 'bgcolor',
1729 ] ),
1731 # 11.2.2
1732 'caption' => $block,
1734 # 11.2.3
1735 'thead' => $common,
1736 'tfoot' => $common,
1737 'tbody' => $common,
1739 # 11.2.4
1740 'colgroup' => array_merge( $common, [ 'span' ] ),
1741 'col' => array_merge( $common, [ 'span' ] ),
1743 # 11.2.5
1744 'tr' => array_merge( $common, [ 'bgcolor' ], $tablealign ),
1746 # 11.2.6
1747 'td' => array_merge( $common, $tablecell, $tablealign ),
1748 'th' => array_merge( $common, $tablecell, $tablealign ),
1750 # 12.2
1751 # NOTE: <a> is not allowed directly, but the attrib
1752 # whitelist is used from the Parser object
1753 'a' => array_merge( $common, [ 'href', 'rel', 'rev' ] ), # rel/rev esp. for RDFa
1755 # 13.2
1756 # Not usually allowed, but may be used for extension-style hooks
1757 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1758 # true
1759 'img' => array_merge( $common, [ 'alt', 'src', 'width', 'height' ] ),
1761 # 15.2.1
1762 'tt' => $common,
1763 'b' => $common,
1764 'i' => $common,
1765 'big' => $common,
1766 'small' => $common,
1767 'strike' => $common,
1768 's' => $common,
1769 'u' => $common,
1771 # 15.2.2
1772 'font' => array_merge( $common, [ 'size', 'color', 'face' ] ),
1773 # basefont
1775 # 15.3
1776 'hr' => array_merge( $common, [ 'width' ] ),
1778 # HTML Ruby annotation text module, simple ruby only.
1779 # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
1780 'ruby' => $common,
1781 # rbc
1782 'rb' => $common,
1783 'rp' => $common,
1784 'rt' => $common, # array_merge( $common, array( 'rbspan' ) ),
1785 'rtc' => $common,
1787 # MathML root element, where used for extensions
1788 # 'title' may not be 100% valid here; it's XHTML
1789 # http://www.w3.org/TR/REC-MathML/
1790 'math' => [ 'class', 'style', 'id', 'title' ],
1792 # HTML 5 section 4.6
1793 'bdi' => $common,
1795 # HTML5 elements, defined by:
1796 # http://www.whatwg.org/html/
1797 'data' => array_merge( $common, [ 'value' ] ),
1798 'time' => array_merge( $common, [ 'datetime' ] ),
1799 'mark' => $common,
1801 // meta and link are only permitted by removeHTMLtags when Microdata
1802 // is enabled so we don't bother adding a conditional to hide these
1803 // Also meta and link are only valid in WikiText as Microdata elements
1804 // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1805 // So we don't bother including $common attributes that have no purpose.
1806 'meta' => [ 'itemprop', 'content' ],
1807 'link' => [ 'itemprop', 'href' ],
1810 return $whitelist;
1814 * Take a fragment of (potentially invalid) HTML and return
1815 * a version with any tags removed, encoded as plain text.
1817 * Warning: this return value must be further escaped for literal
1818 * inclusion in HTML output as of 1.10!
1820 * @param string $text HTML fragment
1821 * @return string
1823 static function stripAllTags( $text ) {
1824 # Actual <tags>
1825 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1827 # Normalize &entities and whitespace
1828 $text = self::decodeCharReferences( $text );
1829 $text = self::normalizeWhitespace( $text );
1831 return $text;
1835 * Hack up a private DOCTYPE with HTML's standard entity declarations.
1836 * PHP 4 seemed to know these if you gave it an HTML doctype, but
1837 * PHP 5.1 doesn't.
1839 * Use for passing XHTML fragments to PHP's XML parsing functions
1841 * @return string
1843 static function hackDocType() {
1844 $out = "<!DOCTYPE html [\n";
1845 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1846 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1848 $out .= "]>\n";
1849 return $out;
1853 * @param string $url
1854 * @return mixed|string
1856 static function cleanUrl( $url ) {
1857 # Normalize any HTML entities in input. They will be
1858 # re-escaped by makeExternalLink().
1859 $url = Sanitizer::decodeCharReferences( $url );
1861 # Escape any control characters introduced by the above step
1862 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1863 [ __CLASS__, 'cleanUrlCallback' ], $url );
1865 # Validate hostname portion
1866 $matches = [];
1867 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1868 list( /* $whole */, $protocol, $host, $rest ) = $matches;
1870 // Characters that will be ignored in IDNs.
1871 // https://tools.ietf.org/html/rfc3454#section-3.1
1872 // Strip them before further processing so blacklists and such work.
1873 $strip = "/
1874 \\s| # general whitespace
1875 \xc2\xad| # 00ad SOFT HYPHEN
1876 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1877 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1878 \xe2\x81\xa0| # 2060 WORD JOINER
1879 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1880 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
1881 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1882 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1883 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1884 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1885 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1886 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1887 /xuD";
1889 $host = preg_replace( $strip, '', $host );
1891 // IPv6 host names are bracketed with []. Url-decode these.
1892 if ( substr_compare( "//%5B", $host, 0, 5 ) === 0 &&
1893 preg_match( '!^//%5B([0-9A-Fa-f:.]+)%5D((:\d+)?)$!', $host, $matches )
1895 $host = '//[' . $matches[1] . ']' . $matches[2];
1898 // @todo FIXME: Validate hostnames here
1900 return $protocol . $host . $rest;
1901 } else {
1902 return $url;
1907 * @param array $matches
1908 * @return string
1910 static function cleanUrlCallback( $matches ) {
1911 return urlencode( $matches[0] );
1915 * Does a string look like an e-mail address?
1917 * This validates an email address using an HTML5 specification found at:
1918 * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1919 * Which as of 2011-01-24 says:
1921 * A valid e-mail address is a string that matches the ABNF production
1922 * 1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1923 * in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1924 * 3.5.
1926 * This function is an implementation of the specification as requested in
1927 * bug 22449.
1929 * Client-side forms will use the same standard validation rules via JS or
1930 * HTML 5 validation; additional restrictions can be enforced server-side
1931 * by extensions via the 'isValidEmailAddr' hook.
1933 * Note that this validation doesn't 100% match RFC 2822, but is believed
1934 * to be liberal enough for wide use. Some invalid addresses will still
1935 * pass validation here.
1937 * @since 1.18
1939 * @param string $addr E-mail address
1940 * @return bool
1942 public static function validateEmail( $addr ) {
1943 $result = null;
1944 if ( !Hooks::run( 'isValidEmailAddr', [ $addr, &$result ] ) ) {
1945 return $result;
1948 // Please note strings below are enclosed in brackets [], this make the
1949 // hyphen "-" a range indicator. Hence it is double backslashed below.
1950 // See bug 26948
1951 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1952 $rfc1034_ldh_str = "a-z0-9\\-";
1954 $html5_email_regexp = "/
1955 ^ # start of string
1956 [$rfc5322_atext\\.]+ # user part which is liberal :p
1957 @ # 'apostrophe'
1958 [$rfc1034_ldh_str]+ # First domain part
1959 (\\.[$rfc1034_ldh_str]+)* # Following part prefixed with a dot
1960 $ # End of string
1961 /ix"; // case Insensitive, eXtended
1963 return (bool)preg_match( $html5_email_regexp, $addr );