rdbms: Avoid selectDB() call in LoadMonitor new connections
[mediawiki.git] / includes / Sanitizer.php
blobed09701d4bf714181d35663b34da2967f4e18755
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 * https://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 * Tells escapeUrlForHtml() to encode the ID using the wiki's primary encoding.
62 * @since 1.30
64 const ID_PRIMARY = 0;
66 /**
67 * Tells escapeUrlForHtml() to encode the ID using the fallback encoding, or return false
68 * if no fallback is configured.
70 * @since 1.30
72 const ID_FALLBACK = 1;
74 /**
75 * List of all named character entities defined in HTML 4.01
76 * https://www.w3.org/TR/html4/sgml/entities.html
77 * As well as &apos; which is only defined starting in XHTML1.
79 private static $htmlEntities = [
80 'Aacute' => 193,
81 'aacute' => 225,
82 'Acirc' => 194,
83 'acirc' => 226,
84 'acute' => 180,
85 'AElig' => 198,
86 'aelig' => 230,
87 'Agrave' => 192,
88 'agrave' => 224,
89 'alefsym' => 8501,
90 'Alpha' => 913,
91 'alpha' => 945,
92 'amp' => 38,
93 'and' => 8743,
94 'ang' => 8736,
95 'apos' => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
96 'Aring' => 197,
97 'aring' => 229,
98 'asymp' => 8776,
99 'Atilde' => 195,
100 'atilde' => 227,
101 'Auml' => 196,
102 'auml' => 228,
103 'bdquo' => 8222,
104 'Beta' => 914,
105 'beta' => 946,
106 'brvbar' => 166,
107 'bull' => 8226,
108 'cap' => 8745,
109 'Ccedil' => 199,
110 'ccedil' => 231,
111 'cedil' => 184,
112 'cent' => 162,
113 'Chi' => 935,
114 'chi' => 967,
115 'circ' => 710,
116 'clubs' => 9827,
117 'cong' => 8773,
118 'copy' => 169,
119 'crarr' => 8629,
120 'cup' => 8746,
121 'curren' => 164,
122 'dagger' => 8224,
123 'Dagger' => 8225,
124 'darr' => 8595,
125 'dArr' => 8659,
126 'deg' => 176,
127 'Delta' => 916,
128 'delta' => 948,
129 'diams' => 9830,
130 'divide' => 247,
131 'Eacute' => 201,
132 'eacute' => 233,
133 'Ecirc' => 202,
134 'ecirc' => 234,
135 'Egrave' => 200,
136 'egrave' => 232,
137 'empty' => 8709,
138 'emsp' => 8195,
139 'ensp' => 8194,
140 'Epsilon' => 917,
141 'epsilon' => 949,
142 'equiv' => 8801,
143 'Eta' => 919,
144 'eta' => 951,
145 'ETH' => 208,
146 'eth' => 240,
147 'Euml' => 203,
148 'euml' => 235,
149 'euro' => 8364,
150 'exist' => 8707,
151 'fnof' => 402,
152 'forall' => 8704,
153 'frac12' => 189,
154 'frac14' => 188,
155 'frac34' => 190,
156 'frasl' => 8260,
157 'Gamma' => 915,
158 'gamma' => 947,
159 'ge' => 8805,
160 'gt' => 62,
161 'harr' => 8596,
162 'hArr' => 8660,
163 'hearts' => 9829,
164 'hellip' => 8230,
165 'Iacute' => 205,
166 'iacute' => 237,
167 'Icirc' => 206,
168 'icirc' => 238,
169 'iexcl' => 161,
170 'Igrave' => 204,
171 'igrave' => 236,
172 'image' => 8465,
173 'infin' => 8734,
174 'int' => 8747,
175 'Iota' => 921,
176 'iota' => 953,
177 'iquest' => 191,
178 'isin' => 8712,
179 'Iuml' => 207,
180 'iuml' => 239,
181 'Kappa' => 922,
182 'kappa' => 954,
183 'Lambda' => 923,
184 'lambda' => 955,
185 'lang' => 9001,
186 'laquo' => 171,
187 'larr' => 8592,
188 'lArr' => 8656,
189 'lceil' => 8968,
190 'ldquo' => 8220,
191 'le' => 8804,
192 'lfloor' => 8970,
193 'lowast' => 8727,
194 'loz' => 9674,
195 'lrm' => 8206,
196 'lsaquo' => 8249,
197 'lsquo' => 8216,
198 'lt' => 60,
199 'macr' => 175,
200 'mdash' => 8212,
201 'micro' => 181,
202 'middot' => 183,
203 'minus' => 8722,
204 'Mu' => 924,
205 'mu' => 956,
206 'nabla' => 8711,
207 'nbsp' => 160,
208 'ndash' => 8211,
209 'ne' => 8800,
210 'ni' => 8715,
211 'not' => 172,
212 'notin' => 8713,
213 'nsub' => 8836,
214 'Ntilde' => 209,
215 'ntilde' => 241,
216 'Nu' => 925,
217 'nu' => 957,
218 'Oacute' => 211,
219 'oacute' => 243,
220 'Ocirc' => 212,
221 'ocirc' => 244,
222 'OElig' => 338,
223 'oelig' => 339,
224 'Ograve' => 210,
225 'ograve' => 242,
226 'oline' => 8254,
227 'Omega' => 937,
228 'omega' => 969,
229 'Omicron' => 927,
230 'omicron' => 959,
231 'oplus' => 8853,
232 'or' => 8744,
233 'ordf' => 170,
234 'ordm' => 186,
235 'Oslash' => 216,
236 'oslash' => 248,
237 'Otilde' => 213,
238 'otilde' => 245,
239 'otimes' => 8855,
240 'Ouml' => 214,
241 'ouml' => 246,
242 'para' => 182,
243 'part' => 8706,
244 'permil' => 8240,
245 'perp' => 8869,
246 'Phi' => 934,
247 'phi' => 966,
248 'Pi' => 928,
249 'pi' => 960,
250 'piv' => 982,
251 'plusmn' => 177,
252 'pound' => 163,
253 'prime' => 8242,
254 'Prime' => 8243,
255 'prod' => 8719,
256 'prop' => 8733,
257 'Psi' => 936,
258 'psi' => 968,
259 'quot' => 34,
260 'radic' => 8730,
261 'rang' => 9002,
262 'raquo' => 187,
263 'rarr' => 8594,
264 'rArr' => 8658,
265 'rceil' => 8969,
266 'rdquo' => 8221,
267 'real' => 8476,
268 'reg' => 174,
269 'rfloor' => 8971,
270 'Rho' => 929,
271 'rho' => 961,
272 'rlm' => 8207,
273 'rsaquo' => 8250,
274 'rsquo' => 8217,
275 'sbquo' => 8218,
276 'Scaron' => 352,
277 'scaron' => 353,
278 'sdot' => 8901,
279 'sect' => 167,
280 'shy' => 173,
281 'Sigma' => 931,
282 'sigma' => 963,
283 'sigmaf' => 962,
284 'sim' => 8764,
285 'spades' => 9824,
286 'sub' => 8834,
287 'sube' => 8838,
288 'sum' => 8721,
289 'sup' => 8835,
290 'sup1' => 185,
291 'sup2' => 178,
292 'sup3' => 179,
293 'supe' => 8839,
294 'szlig' => 223,
295 'Tau' => 932,
296 'tau' => 964,
297 'there4' => 8756,
298 'Theta' => 920,
299 'theta' => 952,
300 'thetasym' => 977,
301 'thinsp' => 8201,
302 'THORN' => 222,
303 'thorn' => 254,
304 'tilde' => 732,
305 'times' => 215,
306 'trade' => 8482,
307 'Uacute' => 218,
308 'uacute' => 250,
309 'uarr' => 8593,
310 'uArr' => 8657,
311 'Ucirc' => 219,
312 'ucirc' => 251,
313 'Ugrave' => 217,
314 'ugrave' => 249,
315 'uml' => 168,
316 'upsih' => 978,
317 'Upsilon' => 933,
318 'upsilon' => 965,
319 'Uuml' => 220,
320 'uuml' => 252,
321 'weierp' => 8472,
322 'Xi' => 926,
323 'xi' => 958,
324 'Yacute' => 221,
325 'yacute' => 253,
326 'yen' => 165,
327 'Yuml' => 376,
328 'yuml' => 255,
329 'Zeta' => 918,
330 'zeta' => 950,
331 'zwj' => 8205,
332 'zwnj' => 8204
336 * Character entity aliases accepted by MediaWiki
338 private static $htmlEntityAliases = [
339 'רלמ' => 'rlm',
340 'رلم' => 'rlm',
344 * Lazy-initialised attributes regex, see getAttribsRegex()
346 private static $attribsRegex;
349 * Regular expression to match HTML/XML attribute pairs within a tag.
350 * Allows some... latitude. Based on,
351 * https://www.w3.org/TR/html5/syntax.html#before-attribute-value-state
352 * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
353 * @return string
355 static function getAttribsRegex() {
356 if ( self::$attribsRegex === null ) {
357 $attribFirst = "[:_\p{L}\p{N}]";
358 $attrib = "[:_\.\-\p{L}\p{N}]";
359 $space = '[\x09\x0a\x0c\x0d\x20]';
360 self::$attribsRegex =
361 "/(?:^|$space)({$attribFirst}{$attrib}*)
362 ($space*=$space*
364 # The attribute value: quoted or alone
365 \"([^\"]*)(?:\"|\$)
366 | '([^']*)(?:'|\$)
367 | (((?!$space|>).)*)
369 )?(?=$space|\$)/sxu";
371 return self::$attribsRegex;
375 * Return the various lists of recognized tags
376 * @param array $extratags For any extra tags to include
377 * @param array $removetags For any tags (default or extra) to exclude
378 * @return array
380 public static function getRecognizedTagData( $extratags = [], $removetags = [] ) {
381 global $wgAllowImageTag;
383 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
384 $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
386 // Base our staticInitialised variable off of the global config state so that if the globals
387 // are changed (like in the screwed up test system) we will re-initialise the settings.
388 $globalContext = $wgAllowImageTag;
389 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
390 $htmlpairsStatic = [ # Tags that must be closed
391 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
392 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
393 'strike', 'strong', 'tt', 'var', 'div', 'center',
394 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
395 'ruby', 'rb', 'rp', 'rt', 'rtc', 'p', 'span', 'abbr', 'dfn',
396 'kbd', 'samp', 'data', 'time', 'mark'
398 $htmlsingle = [
399 'br', 'wbr', 'hr', 'li', 'dt', 'dd', 'meta', 'link'
402 # Elements that cannot have close tags. This is (not coincidentally)
403 # also the list of tags for which the HTML 5 parsing algorithm
404 # requires you to "acknowledge the token's self-closing flag", i.e.
405 # a self-closing tag like <br/> is not an HTML 5 parse error only
406 # for this list.
407 $htmlsingleonly = [
408 'br', 'wbr', 'hr', 'meta', 'link'
411 $htmlnest = [ # Tags that can be nested--??
412 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
413 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
414 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
416 $tabletags = [ # Can only appear inside table, we will close them
417 'td', 'th', 'tr',
419 $htmllist = [ # Tags used by list
420 'ul', 'ol',
422 $listtags = [ # Tags that can appear in a list
423 'li',
426 if ( $wgAllowImageTag ) {
427 $htmlsingle[] = 'img';
428 $htmlsingleonly[] = 'img';
431 $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
432 $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
434 # Convert them all to hashtables for faster lookup
435 $vars = [ 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
436 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' ];
437 foreach ( $vars as $var ) {
438 $$var = array_flip( $$var );
440 $staticInitialised = $globalContext;
443 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
444 $extratags = array_flip( $extratags );
445 $removetags = array_flip( $removetags );
446 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
447 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
449 return [
450 'htmlpairs' => $htmlpairs,
451 'htmlsingle' => $htmlsingle,
452 'htmlsingleonly' => $htmlsingleonly,
453 'htmlnest' => $htmlnest,
454 'tabletags' => $tabletags,
455 'htmllist' => $htmllist,
456 'listtags' => $listtags,
457 'htmlsingleallowed' => $htmlsingleallowed,
458 'htmlelements' => $htmlelements,
463 * Cleans up HTML, removes dangerous tags and attributes, and
464 * removes HTML comments
465 * @param string $text
466 * @param callable $processCallback Callback to do any variable or parameter
467 * replacements in HTML attribute values
468 * @param array|bool $args Arguments for the processing callback
469 * @param array $extratags For any extra tags to include
470 * @param array $removetags For any tags (default or extra) to exclude
471 * @param callable $warnCallback (Deprecated) Callback allowing the
472 * addition of a tracking category when bad input is encountered.
473 * DO NOT ADD NEW PARAMETERS AFTER $warnCallback, since it will be
474 * removed shortly.
475 * @return string
477 public static function removeHTMLtags( $text, $processCallback = null,
478 $args = [], $extratags = [], $removetags = [], $warnCallback = null
480 extract( self::getRecognizedTagData( $extratags, $removetags ) );
482 # Remove HTML comments
483 $text = self::removeHTMLcomments( $text );
484 $bits = explode( '<', $text );
485 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
486 if ( !MWTidy::isEnabled() ) {
487 $tagstack = $tablestack = [];
488 foreach ( $bits as $x ) {
489 $regs = [];
490 # $slash: Does the current element start with a '/'?
491 # $t: Current element name
492 # $params: String between element name and >
493 # $brace: Ending '>' or '/>'
494 # $rest: Everything until the next element of $bits
495 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
496 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
497 } else {
498 $slash = $t = $params = $brace = $rest = null;
501 $badtag = false;
502 $t = strtolower( $t );
503 if ( isset( $htmlelements[$t] ) ) {
504 # Check our stack
505 if ( $slash && isset( $htmlsingleonly[$t] ) ) {
506 $badtag = true;
507 } elseif ( $slash ) {
508 # Closing a tag... is it the one we just opened?
509 MediaWiki\suppressWarnings();
510 $ot = array_pop( $tagstack );
511 MediaWiki\restoreWarnings();
513 if ( $ot != $t ) {
514 if ( isset( $htmlsingleallowed[$ot] ) ) {
515 # Pop all elements with an optional close tag
516 # and see if we find a match below them
517 $optstack = [];
518 array_push( $optstack, $ot );
519 MediaWiki\suppressWarnings();
520 $ot = array_pop( $tagstack );
521 MediaWiki\restoreWarnings();
522 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
523 array_push( $optstack, $ot );
524 MediaWiki\suppressWarnings();
525 $ot = array_pop( $tagstack );
526 MediaWiki\restoreWarnings();
528 if ( $t != $ot ) {
529 # No match. Push the optional elements back again
530 $badtag = true;
531 MediaWiki\suppressWarnings();
532 $ot = array_pop( $optstack );
533 MediaWiki\restoreWarnings();
534 while ( $ot ) {
535 array_push( $tagstack, $ot );
536 MediaWiki\suppressWarnings();
537 $ot = array_pop( $optstack );
538 MediaWiki\restoreWarnings();
541 } else {
542 MediaWiki\suppressWarnings();
543 array_push( $tagstack, $ot );
544 MediaWiki\restoreWarnings();
546 # <li> can be nested in <ul> or <ol>, skip those cases:
547 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
548 $badtag = true;
551 } else {
552 if ( $t == 'table' ) {
553 $tagstack = array_pop( $tablestack );
556 $newparams = '';
557 } else {
558 # Keep track for later
559 if ( isset( $tabletags[$t] ) && !in_array( 'table', $tagstack ) ) {
560 $badtag = true;
561 } elseif ( in_array( $t, $tagstack ) && !isset( $htmlnest[$t] ) ) {
562 $badtag = true;
563 #  Is it a self closed htmlpair ? (T7487)
564 } elseif ( $brace == '/>' && isset( $htmlpairs[$t] ) ) {
565 // Eventually we'll just remove the self-closing
566 // slash, in order to be consistent with HTML5
567 // semantics.
568 // $brace = '>';
569 // For now, let's just warn authors to clean up.
570 if ( is_callable( $warnCallback ) ) {
571 call_user_func_array( $warnCallback, [ 'deprecated-self-close-category' ] );
573 $badtag = true;
574 } elseif ( isset( $htmlsingleonly[$t] ) ) {
575 # Hack to force empty tag for unclosable elements
576 $brace = '/>';
577 } elseif ( isset( $htmlsingle[$t] ) ) {
578 # Hack to not close $htmlsingle tags
579 $brace = null;
580 # Still need to push this optionally-closed tag to
581 # the tag stack so that we can match end tags
582 # instead of marking them as bad.
583 array_push( $tagstack, $t );
584 } elseif ( isset( $tabletags[$t] ) && in_array( $t, $tagstack ) ) {
585 // New table tag but forgot to close the previous one
586 $text .= "</$t>";
587 } else {
588 if ( $t == 'table' ) {
589 array_push( $tablestack, $tagstack );
590 $tagstack = [];
592 array_push( $tagstack, $t );
595 # Replace any variables or template parameters with
596 # plaintext results.
597 if ( is_callable( $processCallback ) ) {
598 call_user_func_array( $processCallback, [ &$params, $args ] );
601 if ( !self::validateTag( $params, $t ) ) {
602 $badtag = true;
605 # Strip non-approved attributes from the tag
606 $newparams = self::fixTagAttributes( $params, $t );
608 if ( !$badtag ) {
609 $rest = str_replace( '>', '&gt;', $rest );
610 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
611 $text .= "<$slash$t$newparams$close>$rest";
612 continue;
615 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
617 # Close off any remaining tags
618 while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
619 $text .= "</$t>\n";
620 if ( $t == 'table' ) {
621 $tagstack = array_pop( $tablestack );
624 } else {
625 # this might be possible using tidy itself
626 foreach ( $bits as $x ) {
627 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
628 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
630 $badtag = false;
631 $t = strtolower( $t );
632 if ( isset( $htmlelements[$t] ) ) {
633 if ( is_callable( $processCallback ) ) {
634 call_user_func_array( $processCallback, [ &$params, $args ] );
637 if ( $brace == '/>' && !( isset( $htmlsingle[$t] ) || isset( $htmlsingleonly[$t] ) ) ) {
638 // Eventually we'll just remove the self-closing
639 // slash, in order to be consistent with HTML5
640 // semantics.
641 // $brace = '>';
642 // For now, let's just warn authors to clean up.
643 if ( is_callable( $warnCallback ) ) {
644 call_user_func_array( $warnCallback, [ 'deprecated-self-close-category' ] );
647 if ( !self::validateTag( $params, $t ) ) {
648 $badtag = true;
651 $newparams = self::fixTagAttributes( $params, $t );
652 if ( !$badtag ) {
653 if ( $brace === '/>' && !isset( $htmlsingleonly[$t] ) ) {
654 # Interpret self-closing tags as empty tags even when
655 # HTML 5 would interpret them as start tags. Such input
656 # is commonly seen on Wikimedia wikis with this intention.
657 $brace = "></$t>";
660 $rest = str_replace( '>', '&gt;', $rest );
661 $text .= "<$slash$t$newparams$brace$rest";
662 continue;
666 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
669 return $text;
673 * Remove '<!--', '-->', and everything between.
674 * To avoid leaving blank lines, when a comment is both preceded
675 * and followed by a newline (ignoring spaces), trim leading and
676 * trailing spaces and one of the newlines.
678 * @param string $text
679 * @return string
681 public static function removeHTMLcomments( $text ) {
682 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
683 $end = strpos( $text, '-->', $start + 4 );
684 if ( $end === false ) {
685 # Unterminated comment; bail out
686 break;
689 $end += 3;
691 # Trim space and newline if the comment is both
692 # preceded and followed by a newline
693 $spaceStart = max( $start - 1, 0 );
694 $spaceLen = $end - $spaceStart;
695 while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
696 $spaceStart--;
697 $spaceLen++;
699 while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
700 $spaceLen++;
702 if ( substr( $text, $spaceStart, 1 ) === "\n"
703 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
704 # Remove the comment, leading and trailing
705 # spaces, and leave only one newline.
706 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
707 } else {
708 # Remove just the comment.
709 $text = substr_replace( $text, '', $start, $end - $start );
712 return $text;
716 * Takes attribute names and values for a tag and the tag name and
717 * validates that the tag is allowed to be present.
718 * This DOES NOT validate the attributes, nor does it validate the
719 * tags themselves. This method only handles the special circumstances
720 * where we may want to allow a tag within content but ONLY when it has
721 * specific attributes set.
723 * @param string $params
724 * @param string $element
725 * @return bool
727 static function validateTag( $params, $element ) {
728 $params = self::decodeTagAttributes( $params );
730 if ( $element == 'meta' || $element == 'link' ) {
731 if ( !isset( $params['itemprop'] ) ) {
732 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
733 return false;
735 if ( $element == 'meta' && !isset( $params['content'] ) ) {
736 // <meta> must have a content="" for the itemprop
737 return false;
739 if ( $element == 'link' && !isset( $params['href'] ) ) {
740 // <link> must have an associated href=""
741 return false;
745 return true;
749 * Take an array of attribute names and values and normalize or discard
750 * illegal values for the given element type.
752 * - Discards attributes not on a whitelist for the given element
753 * - Unsafe style attributes are discarded
754 * - Invalid id attributes are re-encoded
756 * @param array $attribs
757 * @param string $element
758 * @return array
760 * @todo Check for legal values where the DTD limits things.
761 * @todo Check for unique id attribute :P
763 static function validateTagAttributes( $attribs, $element ) {
764 return self::validateAttributes( $attribs,
765 self::attributeWhitelist( $element ) );
769 * Take an array of attribute names and values and normalize or discard
770 * illegal values for the given whitelist.
772 * - Discards attributes not on the given whitelist
773 * - Unsafe style attributes are discarded
774 * - Invalid id attributes are re-encoded
776 * @param array $attribs
777 * @param array $whitelist List of allowed attribute names
778 * @return array
780 * @todo Check for legal values where the DTD limits things.
781 * @todo Check for unique id attribute :P
783 static function validateAttributes( $attribs, $whitelist ) {
784 $whitelist = array_flip( $whitelist );
785 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
787 $out = [];
788 foreach ( $attribs as $attribute => $value ) {
789 # Allow XML namespace declaration to allow RDFa
790 if ( preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
791 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
792 $out[$attribute] = $value;
795 continue;
798 # Allow any attribute beginning with "data-"
799 # However:
800 # * Disallow data attributes used by MediaWiki code
801 # * Ensure that the attribute is not namespaced by banning
802 # colons.
803 if ( !preg_match( '/^data-[^:]*$/i', $attribute )
804 && !isset( $whitelist[$attribute] )
805 || self::isReservedDataAttribute( $attribute )
807 continue;
810 # Strip javascript "expression" from stylesheets.
811 # https://msdn.microsoft.com/en-us/library/ms537634.aspx
812 if ( $attribute == 'style' ) {
813 $value = self::checkCss( $value );
816 # Escape HTML id attributes
817 if ( $attribute === 'id' ) {
818 $value = self::escapeIdForAttribute( $value, self::ID_PRIMARY );
821 # Escape HTML id reference lists
822 if ( $attribute === 'aria-describedby'
823 || $attribute === 'aria-flowto'
824 || $attribute === 'aria-labelledby'
825 || $attribute === 'aria-owns'
827 $value = self::escapeIdReferenceList( $value, 'noninitial' );
830 // RDFa and microdata properties allow URLs, URIs and/or CURIs.
831 // Check them for sanity.
832 if ( $attribute === 'rel' || $attribute === 'rev'
833 # RDFa
834 || $attribute === 'about' || $attribute === 'property'
835 || $attribute === 'resource' || $attribute === 'datatype'
836 || $attribute === 'typeof'
837 # HTML5 microdata
838 || $attribute === 'itemid' || $attribute === 'itemprop'
839 || $attribute === 'itemref' || $attribute === 'itemscope'
840 || $attribute === 'itemtype'
842 // Paranoia. Allow "simple" values but suppress javascript
843 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
844 continue;
848 # NOTE: even though elements using href/src are not allowed directly, supply
849 # validation code that can be used by tag hook handlers, etc
850 if ( $attribute === 'href' || $attribute === 'src' || $attribute === 'poster' ) {
851 if ( !preg_match( $hrefExp, $value ) ) {
852 continue; // drop any href or src attributes not using an allowed protocol.
853 // NOTE: this also drops all relative URLs
857 // If this attribute was previously set, override it.
858 // Output should only have one attribute of each name.
859 $out[$attribute] = $value;
862 # itemtype, itemid, itemref don't make sense without itemscope
863 if ( !array_key_exists( 'itemscope', $out ) ) {
864 unset( $out['itemtype'] );
865 unset( $out['itemid'] );
866 unset( $out['itemref'] );
868 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
870 return $out;
874 * Given an attribute name, checks whether it is a reserved data attribute
875 * (such as data-mw-foo) which is unavailable to user-generated HTML so MediaWiki
876 * core and extension code can safely use it to communicate with frontend code.
877 * @param string $attr Attribute name.
878 * @return bool
880 public static function isReservedDataAttribute( $attr ) {
881 // data-ooui is reserved for ooui.
882 // data-mw and data-parsoid are reserved for parsoid.
883 // data-mw-<name here> is reserved for extensions (or core) if
884 // they need to communicate some data to the client and want to be
885 // sure that it isn't coming from an untrusted user.
886 // We ignore the possibility of namespaces since user-generated HTML
887 // can't use them anymore.
888 return (bool)preg_match( '/^data-(ooui|mw|parsoid)/i', $attr );
892 * Merge two sets of HTML attributes. Conflicting items in the second set
893 * will override those in the first, except for 'class' attributes which
894 * will be combined (if they're both strings).
896 * @todo implement merging for other attributes such as style
897 * @param array $a
898 * @param array $b
899 * @return array
901 static function mergeAttributes( $a, $b ) {
902 $out = array_merge( $a, $b );
903 if ( isset( $a['class'] ) && isset( $b['class'] )
904 && is_string( $a['class'] ) && is_string( $b['class'] )
905 && $a['class'] !== $b['class']
907 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
908 -1, PREG_SPLIT_NO_EMPTY );
909 $out['class'] = implode( ' ', array_unique( $classes ) );
911 return $out;
915 * Normalize CSS into a format we can easily search for hostile input
916 * - decode character references
917 * - decode escape sequences
918 * - convert characters that IE6 interprets into ascii
919 * - remove comments, unless the entire value is one single comment
920 * @param string $value the css string
921 * @return string normalized css
923 public static function normalizeCss( $value ) {
924 // Decode character references like &#123;
925 $value = self::decodeCharReferences( $value );
927 // Decode escape sequences and line continuation
928 // See the grammar in the CSS 2 spec, appendix D.
929 // This has to be done AFTER decoding character references.
930 // This means it isn't possible for this function to return
931 // unsanitized escape sequences. It is possible to manufacture
932 // input that contains character references that decode to
933 // escape sequences that decode to character references, but
934 // it's OK for the return value to contain character references
935 // because the caller is supposed to escape those anyway.
936 static $decodeRegex;
937 if ( !$decodeRegex ) {
938 $space = '[\\x20\\t\\r\\n\\f]';
939 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
940 $backslash = '\\\\';
941 $decodeRegex = "/ $backslash
943 ($nl) | # 1. Line continuation
944 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
945 (.) | # 3. backslash cancelling special meaning
946 () | # 4. backslash at end of string
947 )/xu";
949 $value = preg_replace_callback( $decodeRegex,
950 [ __CLASS__, 'cssDecodeCallback' ], $value );
952 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
953 $value = preg_replace_callback(
954 '/[!-[]-z]/u', // U+FF01 to U+FF5A, excluding U+FF3C (T60088)
955 function ( $matches ) {
956 $cp = UtfNormal\Utils::utf8ToCodepoint( $matches[0] );
957 if ( $cp === false ) {
958 return '';
960 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
962 $value
965 // Convert more characters IE6 might treat as ascii
966 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
967 $value = str_replace(
968 [ 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ],
969 [ 'r', 'n', 'n', 'l', 'i', '(', '(' ],
970 $value
973 // Let the value through if it's nothing but a single comment, to
974 // allow other functions which may reject it to pass some error
975 // message through.
976 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
977 // Remove any comments; IE gets token splitting wrong
978 // This must be done AFTER decoding character references and
979 // escape sequences, because those steps can introduce comments
980 // This step cannot introduce character references or escape
981 // sequences, because it replaces comments with spaces rather
982 // than removing them completely.
983 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
985 // Remove anything after a comment-start token, to guard against
986 // incorrect client implementations.
987 $commentPos = strpos( $value, '/*' );
988 if ( $commentPos !== false ) {
989 $value = substr( $value, 0, $commentPos );
993 // S followed by repeat, iteration, or prolonged sound marks,
994 // which IE will treat as "ss"
995 $value = preg_replace(
996 '/s(?:
997 \xE3\x80\xB1 | # U+3031
998 \xE3\x82\x9D | # U+309D
999 \xE3\x83\xBC | # U+30FC
1000 \xE3\x83\xBD | # U+30FD
1001 \xEF\xB9\xBC | # U+FE7C
1002 \xEF\xB9\xBD | # U+FE7D
1003 \xEF\xBD\xB0 # U+FF70
1004 )/ix',
1005 'ss',
1006 $value
1009 return $value;
1013 * Pick apart some CSS and check it for forbidden or unsafe structures.
1014 * Returns a sanitized string. This sanitized string will have
1015 * character references and escape sequences decoded and comments
1016 * stripped (unless it is itself one valid comment, in which case the value
1017 * will be passed through). If the input is just too evil, only a comment
1018 * complaining about evilness will be returned.
1020 * Currently URL references, 'expression', 'tps' are forbidden.
1022 * NOTE: Despite the fact that character references are decoded, the
1023 * returned string may contain character references given certain
1024 * clever input strings. These character references must
1025 * be escaped before the return value is embedded in HTML.
1027 * @param string $value
1028 * @return string
1030 static function checkCss( $value ) {
1031 $value = self::normalizeCss( $value );
1033 // Reject problematic keywords and control characters
1034 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ||
1035 strpos( $value, UtfNormal\Constants::UTF8_REPLACEMENT ) !== false ) {
1036 return '/* invalid control char */';
1037 } elseif ( preg_match(
1038 '! expression
1039 | filter\s*:
1040 | accelerator\s*:
1041 | -o-link\s*:
1042 | -o-link-source\s*:
1043 | -o-replace\s*:
1044 | url\s*\(
1045 | image\s*\(
1046 | image-set\s*\(
1047 | attr\s*\([^)]+[\s,]+url
1048 !ix', $value ) ) {
1049 return '/* insecure input */';
1051 return $value;
1055 * @param array $matches
1056 * @return string
1058 static function cssDecodeCallback( $matches ) {
1059 if ( $matches[1] !== '' ) {
1060 // Line continuation
1061 return '';
1062 } elseif ( $matches[2] !== '' ) {
1063 $char = UtfNormal\Utils::codepointToUtf8( hexdec( $matches[2] ) );
1064 } elseif ( $matches[3] !== '' ) {
1065 $char = $matches[3];
1066 } else {
1067 $char = '\\';
1069 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
1070 // These characters need to be escaped in strings
1071 // Clean up the escape sequence to avoid parsing errors by clients
1072 return '\\' . dechex( ord( $char ) ) . ' ';
1073 } else {
1074 // Decode unnecessary escape
1075 return $char;
1080 * Take a tag soup fragment listing an HTML element's attributes
1081 * and normalize it to well-formed XML, discarding unwanted attributes.
1082 * Output is safe for further wikitext processing, with escaping of
1083 * values that could trigger problems.
1085 * - Normalizes attribute names to lowercase
1086 * - Discards attributes not on a whitelist for the given element
1087 * - Turns broken or invalid entities into plaintext
1088 * - Double-quotes all attribute values
1089 * - Attributes without values are given the name as attribute
1090 * - Double attributes are discarded
1091 * - Unsafe style attributes are discarded
1092 * - Prepends space if there are attributes.
1093 * - (Optionally) Sorts attributes by name.
1095 * @param string $text
1096 * @param string $element
1097 * @param bool $sorted Whether to sort the attributes (default: false)
1098 * @return string
1100 static function fixTagAttributes( $text, $element, $sorted = false ) {
1101 if ( trim( $text ) == '' ) {
1102 return '';
1105 $decoded = self::decodeTagAttributes( $text );
1106 $stripped = self::validateTagAttributes( $decoded, $element );
1108 if ( $sorted ) {
1109 ksort( $stripped );
1112 return self::safeEncodeTagAttributes( $stripped );
1116 * Encode an attribute value for HTML output.
1117 * @param string $text
1118 * @return string HTML-encoded text fragment
1120 static function encodeAttribute( $text ) {
1121 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1123 // Whitespace is normalized during attribute decoding,
1124 // so if we've been passed non-spaces we must encode them
1125 // ahead of time or they won't be preserved.
1126 $encValue = strtr( $encValue, [
1127 "\n" => '&#10;',
1128 "\r" => '&#13;',
1129 "\t" => '&#9;',
1130 ] );
1132 return $encValue;
1136 * Encode an attribute value for HTML tags, with extra armoring
1137 * against further wiki processing.
1138 * @param string $text
1139 * @return string HTML-encoded text fragment
1141 static function safeEncodeAttribute( $text ) {
1142 $encValue = self::encodeAttribute( $text );
1144 # Templates and links may be expanded in later parsing,
1145 # creating invalid or dangerous output. Suppress this.
1146 $encValue = strtr( $encValue, [
1147 '<' => '&lt;', // This should never happen,
1148 '>' => '&gt;', // we've received invalid input
1149 '"' => '&quot;', // which should have been escaped.
1150 '{' => '&#123;',
1151 '}' => '&#125;', // prevent unpaired language conversion syntax
1152 '[' => '&#91;',
1153 "''" => '&#39;&#39;',
1154 'ISBN' => '&#73;SBN',
1155 'RFC' => '&#82;FC',
1156 'PMID' => '&#80;MID',
1157 '|' => '&#124;',
1158 '__' => '&#95;_',
1159 ] );
1161 # Stupid hack
1162 $encValue = preg_replace_callback(
1163 '/((?i)' . wfUrlProtocols() . ')/',
1164 [ 'Sanitizer', 'armorLinksCallback' ],
1165 $encValue );
1166 return $encValue;
1170 * Given a value, escape it so that it can be used in an id attribute and
1171 * return it. This will use HTML5 validation if $wgExperimentalHtmlIds is
1172 * true, allowing anything but ASCII whitespace. Otherwise it will use
1173 * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1174 * escaped with lots of dots.
1176 * To ensure we don't have to bother escaping anything, we also strip ', ",
1177 * & even if $wgExperimentalIds is true. TODO: Is this the best tactic?
1178 * We also strip # because it upsets IE, and % because it could be
1179 * ambiguous if it's part of something that looks like a percent escape
1180 * (which don't work reliably in fragments cross-browser).
1182 * @deprecated since 1.30, use one of this class' escapeIdFor*() functions
1184 * @see https://www.w3.org/TR/html401/types.html#type-name Valid characters
1185 * in the id and name attributes
1186 * @see https://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1187 * the id attribute
1188 * @see https://www.w3.org/TR/html5/dom.html#the-id-attribute
1189 * HTML5 definition of id attribute
1191 * @param string $id Id to escape
1192 * @param string|array $options String or array of strings (default is array()):
1193 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1194 * so don't pay attention if the first character isn't valid at the
1195 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
1196 * false.
1197 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
1198 * if $wgExperimentalHtmlIds is used, so we can generate extra
1199 * anchors and links won't break.
1200 * @return string
1202 static function escapeId( $id, $options = [] ) {
1203 global $wgExperimentalHtmlIds;
1204 $options = (array)$options;
1206 $id = self::decodeCharReferences( $id );
1208 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1209 $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1210 $id = trim( $id, '_' );
1211 if ( $id === '' ) {
1212 // Must have been all whitespace to start with.
1213 return '_';
1214 } else {
1215 return $id;
1219 // HTML4-style escaping
1220 static $replace = [
1221 '%3A' => ':',
1222 '%' => '.'
1225 $id = urlencode( strtr( $id, ' ', '_' ) );
1226 $id = strtr( $id, $replace );
1228 if ( !preg_match( '/^[a-zA-Z]/', $id ) && !in_array( 'noninitial', $options ) ) {
1229 // Initial character must be a letter!
1230 $id = "x$id";
1232 return $id;
1236 * Given a section name or other user-generated or otherwise unsafe string, escapes it to be
1237 * a valid HTML id attribute.
1239 * WARNING: unlike escapeId(), the output of this function is not guaranteed to be HTML safe,
1240 * be sure to use proper escaping.
1242 * @param string $id String to escape
1243 * @param int $mode One of ID_* constants, specifying whether the primary or fallback encoding
1244 * should be used.
1245 * @return string|bool Escaped ID or false if fallback encoding is requested but it's not
1246 * configured.
1248 * @since 1.30
1250 public static function escapeIdForAttribute( $id, $mode = self::ID_PRIMARY ) {
1251 global $wgFragmentMode;
1253 if ( !isset( $wgFragmentMode[$mode] ) ) {
1254 if ( $mode === self::ID_PRIMARY ) {
1255 throw new UnexpectedValueException( '$wgFragmentMode is configured with no primary mode' );
1257 return false;
1260 $internalMode = $wgFragmentMode[$mode];
1262 return self::escapeIdInternal( $id, $internalMode );
1266 * Given a section name or other user-generated or otherwise unsafe string, escapes it to be
1267 * a valid URL fragment.
1269 * WARNING: unlike escapeId(), the output of this function is not guaranteed to be HTML safe,
1270 * be sure to use proper escaping.
1272 * @param string $id String to escape
1273 * @return string Escaped ID
1275 * @since 1.30
1277 public static function escapeIdForLink( $id ) {
1278 global $wgFragmentMode;
1280 if ( !isset( $wgFragmentMode[self::ID_PRIMARY] ) ) {
1281 throw new UnexpectedValueException( '$wgFragmentMode is configured with no primary mode' );
1284 $mode = $wgFragmentMode[self::ID_PRIMARY];
1286 $id = self::escapeIdInternal( $id, $mode );
1287 $id = self::urlEscapeId( $id, $mode );
1289 return $id;
1293 * Given a section name or other user-generated or otherwise unsafe string, escapes it to be
1294 * a valid URL fragment for external interwikis.
1296 * @param string $id String to escape
1297 * @return string Escaped ID
1299 * @since 1.30
1301 public static function escapeIdForExternalInterwiki( $id ) {
1302 global $wgExternalInterwikiFragmentMode;
1304 $id = self::escapeIdInternal( $id, $wgExternalInterwikiFragmentMode );
1305 $id = self::urlEscapeId( $id, $wgExternalInterwikiFragmentMode );
1307 return $id;
1311 * Helper for escapeIdFor*() functions. URL-escapes the ID if needed.
1313 * @param string $id String to escape
1314 * @param string $mode One of modes from $wgFragmentMode
1315 * @return string
1317 private static function urlEscapeId( $id, $mode ) {
1318 if ( $mode === 'html5' ) {
1319 $id = urlencode( $id );
1320 $id = str_replace( '%3A', ':', $id );
1323 return $id;
1327 * Helper for escapeIdFor*() functions. Performs most of the actual escaping.
1329 * @param string $id String to escape
1330 * @param string $mode One of modes from $wgFragmentMode
1331 * @return string
1333 private static function escapeIdInternal( $id, $mode ) {
1334 $id = self::decodeCharReferences( $id );
1336 switch ( $mode ) {
1337 case 'html5':
1338 $id = str_replace( ' ', '_', $id );
1339 break;
1340 case 'legacy':
1341 // This corresponds to 'noninitial' mode of the old escapeId()
1342 static $replace = [
1343 '%3A' => ':',
1344 '%' => '.'
1347 $id = urlencode( str_replace( ' ', '_', $id ) );
1348 $id = strtr( $id, $replace );
1349 break;
1350 case 'html5-legacy':
1351 $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1352 $id = trim( $id, '_' );
1353 if ( $id === '' ) {
1354 // Must have been all whitespace to start with.
1355 $id = '_';
1357 break;
1358 default:
1359 throw new InvalidArgumentException( "Invalid mode '$mode' passed to '" . __METHOD__ );
1362 return $id;
1366 * Given a string containing a space delimited list of ids, escape each id
1367 * to match ids escaped by the escapeId() function.
1369 * @todo wfDeprecated() uses of $options in 1.31, remove completely in 1.32
1371 * @since 1.27
1373 * @param string $referenceString Space delimited list of ids
1374 * @param string|array $options Deprecated and does nothing.
1375 * @return string
1377 static function escapeIdReferenceList( $referenceString, $options = [] ) {
1378 # Explode the space delimited list string into an array of tokens
1379 $references = preg_split( '/\s+/', "{$referenceString}", -1, PREG_SPLIT_NO_EMPTY );
1381 # Escape each token as an id
1382 foreach ( $references as &$ref ) {
1383 $ref = self::escapeIdForAttribute( $ref );
1386 # Merge the array back to a space delimited list string
1387 # If the array is empty, the result will be an empty string ('')
1388 $referenceString = implode( ' ', $references );
1390 return $referenceString;
1394 * Given a value, escape it so that it can be used as a CSS class and
1395 * return it.
1397 * @todo For extra validity, input should be validated UTF-8.
1399 * @see https://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1401 * @param string $class
1402 * @return string
1404 static function escapeClass( $class ) {
1405 // Convert ugly stuff to underscores and kill underscores in ugly places
1406 return rtrim( preg_replace(
1407 [ '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ],
1408 '_',
1409 $class ), '_' );
1413 * Given HTML input, escape with htmlspecialchars but un-escape entities.
1414 * This allows (generally harmless) entities like &#160; to survive.
1416 * @param string $html HTML to escape
1417 * @return string Escaped input
1419 static function escapeHtmlAllowEntities( $html ) {
1420 $html = self::decodeCharReferences( $html );
1421 # It seems wise to escape ' as well as ", as a matter of course. Can't
1422 # hurt. Use ENT_SUBSTITUTE so that incorrectly truncated multibyte characters
1423 # don't cause the entire string to disappear.
1424 $html = htmlspecialchars( $html, ENT_QUOTES | ENT_SUBSTITUTE );
1425 return $html;
1429 * Regex replace callback for armoring links against further processing.
1430 * @param array $matches
1431 * @return string
1433 private static function armorLinksCallback( $matches ) {
1434 return str_replace( ':', '&#58;', $matches[1] );
1438 * Return an associative array of attribute names and values from
1439 * a partial tag string. Attribute names are forced to lowercase,
1440 * character references are decoded to UTF-8 text.
1442 * @param string $text
1443 * @return array
1445 public static function decodeTagAttributes( $text ) {
1446 if ( trim( $text ) == '' ) {
1447 return [];
1450 $attribs = [];
1451 $pairs = [];
1452 if ( !preg_match_all(
1453 self::getAttribsRegex(),
1454 $text,
1455 $pairs,
1456 PREG_SET_ORDER ) ) {
1457 return $attribs;
1460 foreach ( $pairs as $set ) {
1461 $attribute = strtolower( $set[1] );
1462 $value = self::getTagAttributeCallback( $set );
1464 // Normalize whitespace
1465 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1466 $value = trim( $value );
1468 // Decode character references
1469 $attribs[$attribute] = self::decodeCharReferences( $value );
1471 return $attribs;
1475 * Build a partial tag string from an associative array of attribute
1476 * names and values as returned by decodeTagAttributes.
1478 * @param array $assoc_array
1479 * @return string
1481 public static function safeEncodeTagAttributes( $assoc_array ) {
1482 $attribs = [];
1483 foreach ( $assoc_array as $attribute => $value ) {
1484 $encAttribute = htmlspecialchars( $attribute );
1485 $encValue = self::safeEncodeAttribute( $value );
1487 $attribs[] = "$encAttribute=\"$encValue\"";
1489 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1493 * Pick the appropriate attribute value from a match set from the
1494 * attribs regex matches.
1496 * @param array $set
1497 * @throws MWException When tag conditions are not met.
1498 * @return string
1500 private static function getTagAttributeCallback( $set ) {
1501 if ( isset( $set[5] ) ) {
1502 # No quotes.
1503 return $set[5];
1504 } elseif ( isset( $set[4] ) ) {
1505 # Single-quoted
1506 return $set[4];
1507 } elseif ( isset( $set[3] ) ) {
1508 # Double-quoted
1509 return $set[3];
1510 } elseif ( !isset( $set[2] ) ) {
1511 # In XHTML, attributes must have a value so return an empty string.
1512 # See "Empty attribute syntax",
1513 # https://www.w3.org/TR/html5/syntax.html#syntax-attribute-name
1514 return "";
1515 } else {
1516 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1521 * @param string $text
1522 * @return string
1524 private static function normalizeWhitespace( $text ) {
1525 return preg_replace(
1526 '/\r\n|[\x20\x0d\x0a\x09]/',
1527 ' ',
1528 $text );
1532 * Normalizes whitespace in a section name, such as might be returned
1533 * by Parser::stripSectionName(), for use in the id's that are used for
1534 * section links.
1536 * @param string $section
1537 * @return string
1539 static function normalizeSectionNameWhitespace( $section ) {
1540 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1544 * Ensure that any entities and character references are legal
1545 * for XML and XHTML specifically. Any stray bits will be
1546 * &amp;-escaped to result in a valid text fragment.
1548 * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1549 * numericized (this way we're well-formed even without a DTD)
1550 * b. any numeric char refs must be legal chars, not invalid or forbidden
1551 * c. use lower cased "&#x", not "&#X"
1552 * d. fix or reject non-valid attributes
1554 * @param string $text
1555 * @return string
1556 * @private
1558 static function normalizeCharReferences( $text ) {
1559 return preg_replace_callback(
1560 self::CHAR_REFS_REGEX,
1561 [ 'Sanitizer', 'normalizeCharReferencesCallback' ],
1562 $text );
1566 * @param string $matches
1567 * @return string
1569 static function normalizeCharReferencesCallback( $matches ) {
1570 $ret = null;
1571 if ( $matches[1] != '' ) {
1572 $ret = self::normalizeEntity( $matches[1] );
1573 } elseif ( $matches[2] != '' ) {
1574 $ret = self::decCharReference( $matches[2] );
1575 } elseif ( $matches[3] != '' ) {
1576 $ret = self::hexCharReference( $matches[3] );
1578 if ( is_null( $ret ) ) {
1579 return htmlspecialchars( $matches[0] );
1580 } else {
1581 return $ret;
1586 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1587 * return the equivalent numeric entity reference (except for the core &lt;
1588 * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1589 * the HTML equivalent. Otherwise, returns HTML-escaped text of
1590 * pseudo-entity source (eg &amp;foo;)
1592 * @param string $name
1593 * @return string
1595 static function normalizeEntity( $name ) {
1596 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1597 return '&' . self::$htmlEntityAliases[$name] . ';';
1598 } elseif ( in_array( $name, [ 'lt', 'gt', 'amp', 'quot' ] ) ) {
1599 return "&$name;";
1600 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1601 return '&#' . self::$htmlEntities[$name] . ';';
1602 } else {
1603 return "&amp;$name;";
1608 * @param int $codepoint
1609 * @return null|string
1611 static function decCharReference( $codepoint ) {
1612 $point = intval( $codepoint );
1613 if ( self::validateCodepoint( $point ) ) {
1614 return sprintf( '&#%d;', $point );
1615 } else {
1616 return null;
1621 * @param int $codepoint
1622 * @return null|string
1624 static function hexCharReference( $codepoint ) {
1625 $point = hexdec( $codepoint );
1626 if ( self::validateCodepoint( $point ) ) {
1627 return sprintf( '&#x%x;', $point );
1628 } else {
1629 return null;
1634 * Returns true if a given Unicode codepoint is a valid character in
1635 * both HTML5 and XML.
1636 * @param int $codepoint
1637 * @return bool
1639 private static function validateCodepoint( $codepoint ) {
1640 # U+000C is valid in HTML5 but not allowed in XML.
1641 # U+000D is valid in XML but not allowed in HTML5.
1642 # U+007F - U+009F are disallowed in HTML5 (control characters).
1643 return $codepoint == 0x09
1644 || $codepoint == 0x0a
1645 || ( $codepoint >= 0x20 && $codepoint <= 0x7e )
1646 || ( $codepoint >= 0xa0 && $codepoint <= 0xd7ff )
1647 || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1648 || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1652 * Decode any character references, numeric or named entities,
1653 * in the text and return a UTF-8 string.
1655 * @param string $text
1656 * @return string
1658 public static function decodeCharReferences( $text ) {
1659 return preg_replace_callback(
1660 self::CHAR_REFS_REGEX,
1661 [ 'Sanitizer', 'decodeCharReferencesCallback' ],
1662 $text );
1666 * Decode any character references, numeric or named entities,
1667 * in the next and normalize the resulting string. (T16952)
1669 * This is useful for page titles, not for text to be displayed,
1670 * MediaWiki allows HTML entities to escape normalization as a feature.
1672 * @param string $text Already normalized, containing entities
1673 * @return string Still normalized, without entities
1675 public static function decodeCharReferencesAndNormalize( $text ) {
1676 global $wgContLang;
1677 $text = preg_replace_callback(
1678 self::CHAR_REFS_REGEX,
1679 [ 'Sanitizer', 'decodeCharReferencesCallback' ],
1680 $text,
1681 -1, //limit
1682 $count
1685 if ( $count ) {
1686 return $wgContLang->normalize( $text );
1687 } else {
1688 return $text;
1693 * @param string $matches
1694 * @return string
1696 static function decodeCharReferencesCallback( $matches ) {
1697 if ( $matches[1] != '' ) {
1698 return self::decodeEntity( $matches[1] );
1699 } elseif ( $matches[2] != '' ) {
1700 return self::decodeChar( intval( $matches[2] ) );
1701 } elseif ( $matches[3] != '' ) {
1702 return self::decodeChar( hexdec( $matches[3] ) );
1704 # Last case should be an ampersand by itself
1705 return $matches[0];
1709 * Return UTF-8 string for a codepoint if that is a valid
1710 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1711 * @param int $codepoint
1712 * @return string
1713 * @private
1715 static function decodeChar( $codepoint ) {
1716 if ( self::validateCodepoint( $codepoint ) ) {
1717 return UtfNormal\Utils::codepointToUtf8( $codepoint );
1718 } else {
1719 return UtfNormal\Constants::UTF8_REPLACEMENT;
1724 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1725 * return the UTF-8 encoding of that character. Otherwise, returns
1726 * pseudo-entity source (eg "&foo;")
1728 * @param string $name
1729 * @return string
1731 static function decodeEntity( $name ) {
1732 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1733 $name = self::$htmlEntityAliases[$name];
1735 if ( isset( self::$htmlEntities[$name] ) ) {
1736 return UtfNormal\Utils::codepointToUtf8( self::$htmlEntities[$name] );
1737 } else {
1738 return "&$name;";
1743 * Fetch the whitelist of acceptable attributes for a given element name.
1745 * @param string $element
1746 * @return array
1748 static function attributeWhitelist( $element ) {
1749 $list = self::setupAttributeWhitelist();
1750 return isset( $list[$element] )
1751 ? $list[$element]
1752 : [];
1756 * Foreach array key (an allowed HTML element), return an array
1757 * of allowed attributes
1758 * @return array
1760 static function setupAttributeWhitelist() {
1761 static $whitelist;
1763 if ( $whitelist !== null ) {
1764 return $whitelist;
1767 $common = [
1768 # HTML
1769 'id',
1770 'class',
1771 'style',
1772 'lang',
1773 'dir',
1774 'title',
1776 # WAI-ARIA
1777 'aria-describedby',
1778 'aria-flowto',
1779 'aria-label',
1780 'aria-labelledby',
1781 'aria-owns',
1782 'role',
1784 # RDFa
1785 # These attributes are specified in section 9 of
1786 # https://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1787 'about',
1788 'property',
1789 'resource',
1790 'datatype',
1791 'typeof',
1793 # Microdata. These are specified by
1794 # https://html.spec.whatwg.org/multipage/microdata.html#the-microdata-model
1795 'itemid',
1796 'itemprop',
1797 'itemref',
1798 'itemscope',
1799 'itemtype',
1802 $block = array_merge( $common, [ 'align' ] );
1803 $tablealign = [ 'align', 'valign' ];
1804 $tablecell = [
1805 'abbr',
1806 'axis',
1807 'headers',
1808 'scope',
1809 'rowspan',
1810 'colspan',
1811 'nowrap', # deprecated
1812 'width', # deprecated
1813 'height', # deprecated
1814 'bgcolor', # deprecated
1817 # Numbers refer to sections in HTML 4.01 standard describing the element.
1818 # See: https://www.w3.org/TR/html4/
1819 $whitelist = [
1820 # 7.5.4
1821 'div' => $block,
1822 'center' => $common, # deprecated
1823 'span' => $common,
1825 # 7.5.5
1826 'h1' => $block,
1827 'h2' => $block,
1828 'h3' => $block,
1829 'h4' => $block,
1830 'h5' => $block,
1831 'h6' => $block,
1833 # 7.5.6
1834 # address
1836 # 8.2.4
1837 'bdo' => $common,
1839 # 9.2.1
1840 'em' => $common,
1841 'strong' => $common,
1842 'cite' => $common,
1843 'dfn' => $common,
1844 'code' => $common,
1845 'samp' => $common,
1846 'kbd' => $common,
1847 'var' => $common,
1848 'abbr' => $common,
1849 # acronym
1851 # 9.2.2
1852 'blockquote' => array_merge( $common, [ 'cite' ] ),
1853 'q' => array_merge( $common, [ 'cite' ] ),
1855 # 9.2.3
1856 'sub' => $common,
1857 'sup' => $common,
1859 # 9.3.1
1860 'p' => $block,
1862 # 9.3.2
1863 'br' => array_merge( $common, [ 'clear' ] ),
1865 # https://www.w3.org/TR/html5/text-level-semantics.html#the-wbr-element
1866 'wbr' => $common,
1868 # 9.3.4
1869 'pre' => array_merge( $common, [ 'width' ] ),
1871 # 9.4
1872 'ins' => array_merge( $common, [ 'cite', 'datetime' ] ),
1873 'del' => array_merge( $common, [ 'cite', 'datetime' ] ),
1875 # 10.2
1876 'ul' => array_merge( $common, [ 'type' ] ),
1877 'ol' => array_merge( $common, [ 'type', 'start', 'reversed' ] ),
1878 'li' => array_merge( $common, [ 'type', 'value' ] ),
1880 # 10.3
1881 'dl' => $common,
1882 'dd' => $common,
1883 'dt' => $common,
1885 # 11.2.1
1886 'table' => array_merge( $common,
1887 [ 'summary', 'width', 'border', 'frame',
1888 'rules', 'cellspacing', 'cellpadding',
1889 'align', 'bgcolor',
1890 ] ),
1892 # 11.2.2
1893 'caption' => $block,
1895 # 11.2.3
1896 'thead' => $common,
1897 'tfoot' => $common,
1898 'tbody' => $common,
1900 # 11.2.4
1901 'colgroup' => array_merge( $common, [ 'span' ] ),
1902 'col' => array_merge( $common, [ 'span' ] ),
1904 # 11.2.5
1905 'tr' => array_merge( $common, [ 'bgcolor' ], $tablealign ),
1907 # 11.2.6
1908 'td' => array_merge( $common, $tablecell, $tablealign ),
1909 'th' => array_merge( $common, $tablecell, $tablealign ),
1911 # 12.2
1912 # NOTE: <a> is not allowed directly, but the attrib
1913 # whitelist is used from the Parser object
1914 'a' => array_merge( $common, [ 'href', 'rel', 'rev' ] ), # rel/rev esp. for RDFa
1916 # 13.2
1917 # Not usually allowed, but may be used for extension-style hooks
1918 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1919 # true
1920 'img' => array_merge( $common, [ 'alt', 'src', 'width', 'height', 'srcset' ] ),
1922 'video' => array_merge( $common, [ 'poster', 'controls', 'preload', 'width', 'height' ] ),
1923 'source' => array_merge( $common, [ 'type', 'src' ] ),
1924 'track' => array_merge( $common, [ 'type', 'src', 'srclang', 'kind', 'label' ] ),
1926 # 15.2.1
1927 'tt' => $common,
1928 'b' => $common,
1929 'i' => $common,
1930 'big' => $common,
1931 'small' => $common,
1932 'strike' => $common,
1933 's' => $common,
1934 'u' => $common,
1936 # 15.2.2
1937 'font' => array_merge( $common, [ 'size', 'color', 'face' ] ),
1938 # basefont
1940 # 15.3
1941 'hr' => array_merge( $common, [ 'width' ] ),
1943 # HTML Ruby annotation text module, simple ruby only.
1944 # https://www.w3.org/TR/html5/text-level-semantics.html#the-ruby-element
1945 'ruby' => $common,
1946 # rbc
1947 'rb' => $common,
1948 'rp' => $common,
1949 'rt' => $common, # array_merge( $common, array( 'rbspan' ) ),
1950 'rtc' => $common,
1952 # MathML root element, where used for extensions
1953 # 'title' may not be 100% valid here; it's XHTML
1954 # https://www.w3.org/TR/REC-MathML/
1955 'math' => [ 'class', 'style', 'id', 'title' ],
1957 // HTML 5 section 4.5
1958 'figure' => $common,
1959 'figcaption' => $common,
1961 # HTML 5 section 4.6
1962 'bdi' => $common,
1964 # HTML5 elements, defined by:
1965 # https://html.spec.whatwg.org/multipage/semantics.html#the-data-element
1966 'data' => array_merge( $common, [ 'value' ] ),
1967 'time' => array_merge( $common, [ 'datetime' ] ),
1968 'mark' => $common,
1970 // meta and link are only permitted by removeHTMLtags when Microdata
1971 // is enabled so we don't bother adding a conditional to hide these
1972 // Also meta and link are only valid in WikiText as Microdata elements
1973 // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1974 // So we don't bother including $common attributes that have no purpose.
1975 'meta' => [ 'itemprop', 'content' ],
1976 'link' => [ 'itemprop', 'href', 'title' ],
1979 return $whitelist;
1983 * Take a fragment of (potentially invalid) HTML and return
1984 * a version with any tags removed, encoded as plain text.
1986 * Warning: this return value must be further escaped for literal
1987 * inclusion in HTML output as of 1.10!
1989 * @param string $text HTML fragment
1990 * @return string
1992 static function stripAllTags( $text ) {
1993 # Actual <tags>
1994 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1996 # Normalize &entities and whitespace
1997 $text = self::decodeCharReferences( $text );
1998 $text = self::normalizeWhitespace( $text );
2000 return $text;
2004 * Hack up a private DOCTYPE with HTML's standard entity declarations.
2005 * PHP 4 seemed to know these if you gave it an HTML doctype, but
2006 * PHP 5.1 doesn't.
2008 * Use for passing XHTML fragments to PHP's XML parsing functions
2010 * @return string
2012 static function hackDocType() {
2013 $out = "<!DOCTYPE html [\n";
2014 foreach ( self::$htmlEntities as $entity => $codepoint ) {
2015 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
2017 $out .= "]>\n";
2018 return $out;
2022 * @param string $url
2023 * @return mixed|string
2025 static function cleanUrl( $url ) {
2026 # Normalize any HTML entities in input. They will be
2027 # re-escaped by makeExternalLink().
2028 $url = self::decodeCharReferences( $url );
2030 # Escape any control characters introduced by the above step
2031 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
2032 [ __CLASS__, 'cleanUrlCallback' ], $url );
2034 # Validate hostname portion
2035 $matches = [];
2036 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
2037 list( /* $whole */, $protocol, $host, $rest ) = $matches;
2039 // Characters that will be ignored in IDNs.
2040 // https://tools.ietf.org/html/rfc3454#section-3.1
2041 // Strip them before further processing so blacklists and such work.
2042 $strip = "/
2043 \\s| # general whitespace
2044 \xc2\xad| # 00ad SOFT HYPHEN
2045 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
2046 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
2047 \xe2\x81\xa0| # 2060 WORD JOINER
2048 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
2049 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
2050 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
2051 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
2052 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
2053 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
2054 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
2055 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
2056 /xuD";
2058 $host = preg_replace( $strip, '', $host );
2060 // IPv6 host names are bracketed with []. Url-decode these.
2061 if ( substr_compare( "//%5B", $host, 0, 5 ) === 0 &&
2062 preg_match( '!^//%5B([0-9A-Fa-f:.]+)%5D((:\d+)?)$!', $host, $matches )
2064 $host = '//[' . $matches[1] . ']' . $matches[2];
2067 // @todo FIXME: Validate hostnames here
2069 return $protocol . $host . $rest;
2070 } else {
2071 return $url;
2076 * @param array $matches
2077 * @return string
2079 static function cleanUrlCallback( $matches ) {
2080 return urlencode( $matches[0] );
2084 * Does a string look like an e-mail address?
2086 * This validates an email address using an HTML5 specification found at:
2087 * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
2088 * Which as of 2011-01-24 says:
2090 * A valid e-mail address is a string that matches the ABNF production
2091 * 1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
2092 * in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
2093 * 3.5.
2095 * This function is an implementation of the specification as requested in
2096 * T24449.
2098 * Client-side forms will use the same standard validation rules via JS or
2099 * HTML 5 validation; additional restrictions can be enforced server-side
2100 * by extensions via the 'isValidEmailAddr' hook.
2102 * Note that this validation doesn't 100% match RFC 2822, but is believed
2103 * to be liberal enough for wide use. Some invalid addresses will still
2104 * pass validation here.
2106 * @since 1.18
2108 * @param string $addr E-mail address
2109 * @return bool
2111 public static function validateEmail( $addr ) {
2112 $result = null;
2113 if ( !Hooks::run( 'isValidEmailAddr', [ $addr, &$result ] ) ) {
2114 return $result;
2117 // Please note strings below are enclosed in brackets [], this make the
2118 // hyphen "-" a range indicator. Hence it is double backslashed below.
2119 // See T28948
2120 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
2121 $rfc1034_ldh_str = "a-z0-9\\-";
2123 $html5_email_regexp = "/
2124 ^ # start of string
2125 [$rfc5322_atext\\.]+ # user part which is liberal :p
2126 @ # 'apostrophe'
2127 [$rfc1034_ldh_str]+ # First domain part
2128 (\\.[$rfc1034_ldh_str]+)* # Following part prefixed with a dot
2129 $ # End of string
2130 /ix"; // case Insensitive, eXtended
2132 return (bool)preg_match( $html5_email_regexp, $addr );