Don't rely on $wgTitle in WebRequest
[mediawiki.git] / includes / Sanitizer.php
blobbca2f67e91386be4a76d2d06f7a0a17e0fb883fc
1 <?php
2 /**
3 * HTML sanitizer for %MediaWiki.
5 * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
6 * https://www.mediawiki.org/
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 2 of the License, or
11 * (at your option) any later version.
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License along
19 * with this program; if not, write to the Free Software Foundation, Inc.,
20 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
21 * http://www.gnu.org/copyleft/gpl.html
23 * @file
24 * @ingroup Parser
27 /**
28 * HTML sanitizer for MediaWiki
29 * @ingroup Parser
31 class Sanitizer {
32 /**
33 * Regular expression to match various types of character references in
34 * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
36 const CHAR_REFS_REGEX =
37 '/&([A-Za-z0-9\x80-\xff]+);
38 |&\#([0-9]+);
39 |&\#[xX]([0-9A-Fa-f]+);
40 |(&)/x';
42 /**
43 * Blacklist for evil uris like javascript:
44 * WARNING: DO NOT use this in any place that actually requires blacklisting
45 * for security reasons. There are NUMEROUS[1] ways to bypass blacklisting, the
46 * only way to be secure from javascript: uri based xss vectors is to whitelist
47 * things that you know are safe and deny everything else.
48 * [1]: http://ha.ckers.org/xss.html
50 const EVIL_URI_PATTERN = '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
51 const XMLNS_ATTRIBUTE_PATTERN = "/^xmlns:[:A-Z_a-z-.0-9]+$/";
53 /**
54 * List of all named character entities defined in HTML 4.01
55 * http://www.w3.org/TR/html4/sgml/entities.html
56 * As well as &apos; which is only defined starting in XHTML1.
58 private static $htmlEntities = array(
59 'Aacute' => 193,
60 'aacute' => 225,
61 'Acirc' => 194,
62 'acirc' => 226,
63 'acute' => 180,
64 'AElig' => 198,
65 'aelig' => 230,
66 'Agrave' => 192,
67 'agrave' => 224,
68 'alefsym' => 8501,
69 'Alpha' => 913,
70 'alpha' => 945,
71 'amp' => 38,
72 'and' => 8743,
73 'ang' => 8736,
74 'apos' => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
75 'Aring' => 197,
76 'aring' => 229,
77 'asymp' => 8776,
78 'Atilde' => 195,
79 'atilde' => 227,
80 'Auml' => 196,
81 'auml' => 228,
82 'bdquo' => 8222,
83 'Beta' => 914,
84 'beta' => 946,
85 'brvbar' => 166,
86 'bull' => 8226,
87 'cap' => 8745,
88 'Ccedil' => 199,
89 'ccedil' => 231,
90 'cedil' => 184,
91 'cent' => 162,
92 'Chi' => 935,
93 'chi' => 967,
94 'circ' => 710,
95 'clubs' => 9827,
96 'cong' => 8773,
97 'copy' => 169,
98 'crarr' => 8629,
99 'cup' => 8746,
100 'curren' => 164,
101 'dagger' => 8224,
102 'Dagger' => 8225,
103 'darr' => 8595,
104 'dArr' => 8659,
105 'deg' => 176,
106 'Delta' => 916,
107 'delta' => 948,
108 'diams' => 9830,
109 'divide' => 247,
110 'Eacute' => 201,
111 'eacute' => 233,
112 'Ecirc' => 202,
113 'ecirc' => 234,
114 'Egrave' => 200,
115 'egrave' => 232,
116 'empty' => 8709,
117 'emsp' => 8195,
118 'ensp' => 8194,
119 'Epsilon' => 917,
120 'epsilon' => 949,
121 'equiv' => 8801,
122 'Eta' => 919,
123 'eta' => 951,
124 'ETH' => 208,
125 'eth' => 240,
126 'Euml' => 203,
127 'euml' => 235,
128 'euro' => 8364,
129 'exist' => 8707,
130 'fnof' => 402,
131 'forall' => 8704,
132 'frac12' => 189,
133 'frac14' => 188,
134 'frac34' => 190,
135 'frasl' => 8260,
136 'Gamma' => 915,
137 'gamma' => 947,
138 'ge' => 8805,
139 'gt' => 62,
140 'harr' => 8596,
141 'hArr' => 8660,
142 'hearts' => 9829,
143 'hellip' => 8230,
144 'Iacute' => 205,
145 'iacute' => 237,
146 'Icirc' => 206,
147 'icirc' => 238,
148 'iexcl' => 161,
149 'Igrave' => 204,
150 'igrave' => 236,
151 'image' => 8465,
152 'infin' => 8734,
153 'int' => 8747,
154 'Iota' => 921,
155 'iota' => 953,
156 'iquest' => 191,
157 'isin' => 8712,
158 'Iuml' => 207,
159 'iuml' => 239,
160 'Kappa' => 922,
161 'kappa' => 954,
162 'Lambda' => 923,
163 'lambda' => 955,
164 'lang' => 9001,
165 'laquo' => 171,
166 'larr' => 8592,
167 'lArr' => 8656,
168 'lceil' => 8968,
169 'ldquo' => 8220,
170 'le' => 8804,
171 'lfloor' => 8970,
172 'lowast' => 8727,
173 'loz' => 9674,
174 'lrm' => 8206,
175 'lsaquo' => 8249,
176 'lsquo' => 8216,
177 'lt' => 60,
178 'macr' => 175,
179 'mdash' => 8212,
180 'micro' => 181,
181 'middot' => 183,
182 'minus' => 8722,
183 'Mu' => 924,
184 'mu' => 956,
185 'nabla' => 8711,
186 'nbsp' => 160,
187 'ndash' => 8211,
188 'ne' => 8800,
189 'ni' => 8715,
190 'not' => 172,
191 'notin' => 8713,
192 'nsub' => 8836,
193 'Ntilde' => 209,
194 'ntilde' => 241,
195 'Nu' => 925,
196 'nu' => 957,
197 'Oacute' => 211,
198 'oacute' => 243,
199 'Ocirc' => 212,
200 'ocirc' => 244,
201 'OElig' => 338,
202 'oelig' => 339,
203 'Ograve' => 210,
204 'ograve' => 242,
205 'oline' => 8254,
206 'Omega' => 937,
207 'omega' => 969,
208 'Omicron' => 927,
209 'omicron' => 959,
210 'oplus' => 8853,
211 'or' => 8744,
212 'ordf' => 170,
213 'ordm' => 186,
214 'Oslash' => 216,
215 'oslash' => 248,
216 'Otilde' => 213,
217 'otilde' => 245,
218 'otimes' => 8855,
219 'Ouml' => 214,
220 'ouml' => 246,
221 'para' => 182,
222 'part' => 8706,
223 'permil' => 8240,
224 'perp' => 8869,
225 'Phi' => 934,
226 'phi' => 966,
227 'Pi' => 928,
228 'pi' => 960,
229 'piv' => 982,
230 'plusmn' => 177,
231 'pound' => 163,
232 'prime' => 8242,
233 'Prime' => 8243,
234 'prod' => 8719,
235 'prop' => 8733,
236 'Psi' => 936,
237 'psi' => 968,
238 'quot' => 34,
239 'radic' => 8730,
240 'rang' => 9002,
241 'raquo' => 187,
242 'rarr' => 8594,
243 'rArr' => 8658,
244 'rceil' => 8969,
245 'rdquo' => 8221,
246 'real' => 8476,
247 'reg' => 174,
248 'rfloor' => 8971,
249 'Rho' => 929,
250 'rho' => 961,
251 'rlm' => 8207,
252 'rsaquo' => 8250,
253 'rsquo' => 8217,
254 'sbquo' => 8218,
255 'Scaron' => 352,
256 'scaron' => 353,
257 'sdot' => 8901,
258 'sect' => 167,
259 'shy' => 173,
260 'Sigma' => 931,
261 'sigma' => 963,
262 'sigmaf' => 962,
263 'sim' => 8764,
264 'spades' => 9824,
265 'sub' => 8834,
266 'sube' => 8838,
267 'sum' => 8721,
268 'sup' => 8835,
269 'sup1' => 185,
270 'sup2' => 178,
271 'sup3' => 179,
272 'supe' => 8839,
273 'szlig' => 223,
274 'Tau' => 932,
275 'tau' => 964,
276 'there4' => 8756,
277 'Theta' => 920,
278 'theta' => 952,
279 'thetasym' => 977,
280 'thinsp' => 8201,
281 'THORN' => 222,
282 'thorn' => 254,
283 'tilde' => 732,
284 'times' => 215,
285 'trade' => 8482,
286 'Uacute' => 218,
287 'uacute' => 250,
288 'uarr' => 8593,
289 'uArr' => 8657,
290 'Ucirc' => 219,
291 'ucirc' => 251,
292 'Ugrave' => 217,
293 'ugrave' => 249,
294 'uml' => 168,
295 'upsih' => 978,
296 'Upsilon' => 933,
297 'upsilon' => 965,
298 'Uuml' => 220,
299 'uuml' => 252,
300 'weierp' => 8472,
301 'Xi' => 926,
302 'xi' => 958,
303 'Yacute' => 221,
304 'yacute' => 253,
305 'yen' => 165,
306 'Yuml' => 376,
307 'yuml' => 255,
308 'Zeta' => 918,
309 'zeta' => 950,
310 'zwj' => 8205,
311 'zwnj' => 8204
315 * Character entity aliases accepted by MediaWiki
317 private static $htmlEntityAliases = array(
318 'רלמ' => 'rlm',
319 'رلم' => 'rlm',
323 * Lazy-initialised attributes regex, see getAttribsRegex()
325 private static $attribsRegex;
328 * Regular expression to match HTML/XML attribute pairs within a tag.
329 * Allows some... latitude.
330 * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
331 * @return string
333 static function getAttribsRegex() {
334 if ( self::$attribsRegex === null ) {
335 $attribFirst = '[:A-Z_a-z0-9]';
336 $attrib = '[:A-Z_a-z-.0-9]';
337 $space = '[\x09\x0a\x0d\x20]';
338 self::$attribsRegex =
339 "/(?:^|$space)({$attribFirst}{$attrib}*)
340 ($space*=$space*
342 # The attribute value: quoted or alone
343 \"([^<\"]*)\"
344 | '([^<']*)'
345 | ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
346 | (\#[0-9a-fA-F]+) # Technically wrong, but lots of
347 # colors are specified like this.
348 # We'll be normalizing it.
350 )?(?=$space|\$)/sx";
352 return self::$attribsRegex;
356 * Cleans up HTML, removes dangerous tags and attributes, and
357 * removes HTML comments
358 * @private
359 * @param string $text
360 * @param callable $processCallback Callback to do any variable or parameter
361 * replacements in HTML attribute values
362 * @param array|bool $args Arguments for the processing callback
363 * @param array $extratags For any extra tags to include
364 * @param array $removetags For any tags (default or extra) to exclude
365 * @return string
367 static function removeHTMLtags( $text, $processCallback = null,
368 $args = array(), $extratags = array(), $removetags = array()
370 global $wgUseTidy, $wgAllowMicrodataAttributes, $wgAllowImageTag;
372 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
373 $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
375 wfProfileIn( __METHOD__ );
377 // Base our staticInitialised variable off of the global config state so that if the globals
378 // are changed (like in the screwed up test system) we will re-initialise the settings.
379 $globalContext = implode( '-', compact( 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
380 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
382 $htmlpairsStatic = array( # Tags that must be closed
383 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
384 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
385 'strike', 'strong', 'tt', 'var', 'div', 'center',
386 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
387 'ruby', 'rb', 'rp', 'rt', 'rtc', 'p', 'span', 'abbr', 'dfn',
388 'kbd', 'samp', 'data', 'time', 'mark'
390 $htmlsingle = array(
391 'br', 'wbr', 'hr', 'li', 'dt', 'dd'
393 $htmlsingleonly = array( # Elements that cannot have close tags
394 'br', 'wbr', 'hr'
396 if ( $wgAllowMicrodataAttributes ) {
397 $htmlsingle[] = $htmlsingleonly[] = 'meta';
398 $htmlsingle[] = $htmlsingleonly[] = 'link';
400 $htmlnest = array( # Tags that can be nested--??
401 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
402 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
403 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
405 $tabletags = array( # Can only appear inside table, we will close them
406 'td', 'th', 'tr',
408 $htmllist = array( # Tags used by list
409 'ul', 'ol',
411 $listtags = array( # Tags that can appear in a list
412 'li',
415 if ( $wgAllowImageTag ) {
416 $htmlsingle[] = 'img';
417 $htmlsingleonly[] = 'img';
420 $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
421 $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
423 # Convert them all to hashtables for faster lookup
424 $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
425 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
426 foreach ( $vars as $var ) {
427 $$var = array_flip( $$var );
429 $staticInitialised = $globalContext;
431 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
432 $extratags = array_flip( $extratags );
433 $removetags = array_flip( $removetags );
434 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
435 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
437 # Remove HTML comments
438 $text = Sanitizer::removeHTMLcomments( $text );
439 $bits = explode( '<', $text );
440 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
441 if ( !$wgUseTidy ) {
442 $tagstack = $tablestack = array();
443 foreach ( $bits as $x ) {
444 $regs = array();
445 # $slash: Does the current element start with a '/'?
446 # $t: Current element name
447 # $params: String between element name and >
448 # $brace: Ending '>' or '/>'
449 # $rest: Everything until the next element of $bits
450 if ( preg_match( '!^(/?)([^\\s/>]+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
451 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
452 } else {
453 $slash = $t = $params = $brace = $rest = null;
456 $badtag = false;
457 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
458 # Check our stack
459 if ( $slash && isset( $htmlsingleonly[$t] ) ) {
460 $badtag = true;
461 } elseif ( $slash ) {
462 # Closing a tag... is it the one we just opened?
463 wfSuppressWarnings();
464 $ot = array_pop( $tagstack );
465 wfRestoreWarnings();
467 if ( $ot != $t ) {
468 if ( isset( $htmlsingleallowed[$ot] ) ) {
469 # Pop all elements with an optional close tag
470 # and see if we find a match below them
471 $optstack = array();
472 array_push( $optstack, $ot );
473 wfSuppressWarnings();
474 $ot = array_pop( $tagstack );
475 wfRestoreWarnings();
476 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
477 array_push( $optstack, $ot );
478 wfSuppressWarnings();
479 $ot = array_pop( $tagstack );
480 wfRestoreWarnings();
482 if ( $t != $ot ) {
483 # No match. Push the optional elements back again
484 $badtag = true;
485 wfSuppressWarnings();
486 $ot = array_pop( $optstack );
487 wfRestoreWarnings();
488 while ( $ot ) {
489 array_push( $tagstack, $ot );
490 wfSuppressWarnings();
491 $ot = array_pop( $optstack );
492 wfRestoreWarnings();
495 } else {
496 wfSuppressWarnings();
497 array_push( $tagstack, $ot );
498 wfRestoreWarnings();
500 # <li> can be nested in <ul> or <ol>, skip those cases:
501 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
502 $badtag = true;
505 } else {
506 if ( $t == 'table' ) {
507 $tagstack = array_pop( $tablestack );
510 $newparams = '';
511 } else {
512 # Keep track for later
513 if ( isset( $tabletags[$t] ) &&
514 !in_array( 'table', $tagstack ) ) {
515 $badtag = true;
516 } elseif ( in_array( $t, $tagstack ) &&
517 !isset( $htmlnest[$t] ) ) {
518 $badtag = true;
519 # Is it a self closed htmlpair ? (bug 5487)
520 } elseif ( $brace == '/>' &&
521 isset( $htmlpairs[$t] ) ) {
522 $badtag = true;
523 } elseif ( isset( $htmlsingleonly[$t] ) ) {
524 # Hack to force empty tag for unclosable elements
525 $brace = '/>';
526 } elseif ( isset( $htmlsingle[$t] ) ) {
527 # Hack to not close $htmlsingle tags
528 $brace = null;
529 # Still need to push this optionally-closed tag to
530 # the tag stack so that we can match end tags
531 # instead of marking them as bad.
532 array_push( $tagstack, $t );
533 } elseif ( isset( $tabletags[$t] )
534 && in_array( $t, $tagstack ) ) {
535 // New table tag but forgot to close the previous one
536 $text .= "</$t>";
537 } else {
538 if ( $t == 'table' ) {
539 array_push( $tablestack, $tagstack );
540 $tagstack = array();
542 array_push( $tagstack, $t );
545 # Replace any variables or template parameters with
546 # plaintext results.
547 if ( is_callable( $processCallback ) ) {
548 call_user_func_array( $processCallback, array( &$params, $args ) );
551 if ( !Sanitizer::validateTag( $params, $t ) ) {
552 $badtag = true;
555 # Strip non-approved attributes from the tag
556 $newparams = Sanitizer::fixTagAttributes( $params, $t );
558 if ( !$badtag ) {
559 $rest = str_replace( '>', '&gt;', $rest );
560 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
561 $text .= "<$slash$t$newparams$close>$rest";
562 continue;
565 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
567 # Close off any remaining tags
568 while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
569 $text .= "</$t>\n";
570 if ( $t == 'table' ) {
571 $tagstack = array_pop( $tablestack );
574 } else {
575 # this might be possible using tidy itself
576 foreach ( $bits as $x ) {
577 preg_match(
578 '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
580 $regs
583 wfSuppressWarnings();
584 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
585 wfRestoreWarnings();
587 $badtag = false;
588 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
589 if ( is_callable( $processCallback ) ) {
590 call_user_func_array( $processCallback, array( &$params, $args ) );
593 if ( !Sanitizer::validateTag( $params, $t ) ) {
594 $badtag = true;
597 $newparams = Sanitizer::fixTagAttributes( $params, $t );
598 if ( !$badtag ) {
599 $rest = str_replace( '>', '&gt;', $rest );
600 $text .= "<$slash$t$newparams$brace$rest";
601 continue;
604 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
607 wfProfileOut( __METHOD__ );
608 return $text;
612 * Remove '<!--', '-->', and everything between.
613 * To avoid leaving blank lines, when a comment is both preceded
614 * and followed by a newline (ignoring spaces), trim leading and
615 * trailing spaces and one of the newlines.
617 * @private
618 * @param string $text
619 * @return string
621 static function removeHTMLcomments( $text ) {
622 wfProfileIn( __METHOD__ );
623 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
624 $end = strpos( $text, '-->', $start + 4 );
625 if ( $end === false ) {
626 # Unterminated comment; bail out
627 break;
630 $end += 3;
632 # Trim space and newline if the comment is both
633 # preceded and followed by a newline
634 $spaceStart = max( $start - 1, 0 );
635 $spaceLen = $end - $spaceStart;
636 while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
637 $spaceStart--;
638 $spaceLen++;
640 while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
641 $spaceLen++;
643 if ( substr( $text, $spaceStart, 1 ) === "\n"
644 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
645 # Remove the comment, leading and trailing
646 # spaces, and leave only one newline.
647 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
648 } else {
649 # Remove just the comment.
650 $text = substr_replace( $text, '', $start, $end - $start );
653 wfProfileOut( __METHOD__ );
654 return $text;
658 * Takes attribute names and values for a tag and the tag name and
659 * validates that the tag is allowed to be present.
660 * This DOES NOT validate the attributes, nor does it validate the
661 * tags themselves. This method only handles the special circumstances
662 * where we may want to allow a tag within content but ONLY when it has
663 * specific attributes set.
665 * @param string $params
666 * @param string $element
667 * @return bool
669 static function validateTag( $params, $element ) {
670 $params = Sanitizer::decodeTagAttributes( $params );
672 if ( $element == 'meta' || $element == 'link' ) {
673 if ( !isset( $params['itemprop'] ) ) {
674 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
675 return false;
677 if ( $element == 'meta' && !isset( $params['content'] ) ) {
678 // <meta> must have a content="" for the itemprop
679 return false;
681 if ( $element == 'link' && !isset( $params['href'] ) ) {
682 // <link> must have an associated href=""
683 return false;
687 return true;
691 * Take an array of attribute names and values and normalize or discard
692 * illegal values for the given element type.
694 * - Discards attributes not on a whitelist for the given element
695 * - Unsafe style attributes are discarded
696 * - Invalid id attributes are re-encoded
698 * @param array $attribs
699 * @param string $element
700 * @return array
702 * @todo Check for legal values where the DTD limits things.
703 * @todo Check for unique id attribute :P
705 static function validateTagAttributes( $attribs, $element ) {
706 return Sanitizer::validateAttributes( $attribs,
707 Sanitizer::attributeWhitelist( $element ) );
711 * Take an array of attribute names and values and normalize or discard
712 * illegal values for the given whitelist.
714 * - Discards attributes not the given whitelist
715 * - Unsafe style attributes are discarded
716 * - Invalid id attributes are re-encoded
718 * @param array $attribs
719 * @param array $whitelist List of allowed attribute names
720 * @return array
722 * @todo Check for legal values where the DTD limits things.
723 * @todo Check for unique id attribute :P
725 static function validateAttributes( $attribs, $whitelist ) {
726 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
728 $whitelist = array_flip( $whitelist );
729 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
731 $out = array();
732 foreach ( $attribs as $attribute => $value ) {
733 #allow XML namespace declaration if RDFa is enabled
734 if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
735 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
736 $out[$attribute] = $value;
739 continue;
742 # Allow any attribute beginning with "data-"
743 if ( !preg_match( '/^data-/i', $attribute ) && !isset( $whitelist[$attribute] ) ) {
744 continue;
747 # Strip javascript "expression" from stylesheets.
748 # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
749 if ( $attribute == 'style' ) {
750 $value = Sanitizer::checkCss( $value );
753 if ( $attribute === 'id' ) {
754 $value = Sanitizer::escapeId( $value, 'noninitial' );
757 # WAI-ARIA
758 # http://www.w3.org/TR/wai-aria/
759 # http://www.whatwg.org/html/elements.html#wai-aria
760 # For now we only support role="presentation" until we work out what roles should be
761 # usable by content and we ensure that our code explicitly rejects patterns that
762 # violate HTML5's ARIA restrictions.
763 if ( $attribute === 'role' && $value !== 'presentation' ) {
764 continue;
767 // RDFa and microdata properties allow URLs, URIs and/or CURIs.
768 // Check them for sanity.
769 if ( $attribute === 'rel' || $attribute === 'rev'
770 # RDFa
771 || $attribute === 'about' || $attribute === 'property'
772 || $attribute === 'resource' || $attribute === 'datatype'
773 || $attribute === 'typeof'
774 # HTML5 microdata
775 || $attribute === 'itemid' || $attribute === 'itemprop'
776 || $attribute === 'itemref' || $attribute === 'itemscope'
777 || $attribute === 'itemtype'
779 //Paranoia. Allow "simple" values but suppress javascript
780 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
781 continue;
785 # NOTE: even though elements using href/src are not allowed directly, supply
786 # validation code that can be used by tag hook handlers, etc
787 if ( $attribute === 'href' || $attribute === 'src' ) {
788 if ( !preg_match( $hrefExp, $value ) ) {
789 continue; //drop any href or src attributes not using an allowed protocol.
790 // NOTE: this also drops all relative URLs
794 // If this attribute was previously set, override it.
795 // Output should only have one attribute of each name.
796 $out[$attribute] = $value;
799 if ( $wgAllowMicrodataAttributes ) {
800 # itemtype, itemid, itemref don't make sense without itemscope
801 if ( !array_key_exists( 'itemscope', $out ) ) {
802 unset( $out['itemtype'] );
803 unset( $out['itemid'] );
804 unset( $out['itemref'] );
806 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
808 return $out;
812 * Merge two sets of HTML attributes. Conflicting items in the second set
813 * will override those in the first, except for 'class' attributes which
814 * will be combined (if they're both strings).
816 * @todo implement merging for other attributes such as style
817 * @param array $a
818 * @param array $b
819 * @return array
821 static function mergeAttributes( $a, $b ) {
822 $out = array_merge( $a, $b );
823 if ( isset( $a['class'] ) && isset( $b['class'] )
824 && is_string( $a['class'] ) && is_string( $b['class'] )
825 && $a['class'] !== $b['class']
827 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
828 -1, PREG_SPLIT_NO_EMPTY );
829 $out['class'] = implode( ' ', array_unique( $classes ) );
831 return $out;
835 * Normalize CSS into a format we can easily search for hostile input
836 * - decode character references
837 * - decode escape sequences
838 * - convert characters that IE6 interprets into ascii
839 * - remove comments, unless the entire value is one single comment
840 * @param string $value the css string
841 * @return string normalized css
843 public static function normalizeCss( $value ) {
845 // Decode character references like &#123;
846 $value = Sanitizer::decodeCharReferences( $value );
848 // Decode escape sequences and line continuation
849 // See the grammar in the CSS 2 spec, appendix D.
850 // This has to be done AFTER decoding character references.
851 // This means it isn't possible for this function to return
852 // unsanitized escape sequences. It is possible to manufacture
853 // input that contains character references that decode to
854 // escape sequences that decode to character references, but
855 // it's OK for the return value to contain character references
856 // because the caller is supposed to escape those anyway.
857 static $decodeRegex;
858 if ( !$decodeRegex ) {
859 $space = '[\\x20\\t\\r\\n\\f]';
860 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
861 $backslash = '\\\\';
862 $decodeRegex = "/ $backslash
864 ($nl) | # 1. Line continuation
865 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
866 (.) | # 3. backslash cancelling special meaning
867 () | # 4. backslash at end of string
868 )/xu";
870 $value = preg_replace_callback( $decodeRegex,
871 array( __CLASS__, 'cssDecodeCallback' ), $value );
873 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
874 $value = preg_replace_callback(
875 '/[!-[]-z]/u', // U+FF01 to U+FF5A, excluding U+FF3C (bug 58088)
876 function ( $matches ) {
877 $cp = utf8ToCodepoint( $matches[0] );
878 if ( $cp === false ) {
879 return '';
881 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
883 $value
886 // Convert more characters IE6 might treat as ascii
887 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
888 $value = str_replace(
889 array( 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ),
890 array( 'r', 'n', 'n', 'l', 'i', '(', '(' ),
891 $value
894 // Let the value through if it's nothing but a single comment, to
895 // allow other functions which may reject it to pass some error
896 // message through.
897 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
898 // Remove any comments; IE gets token splitting wrong
899 // This must be done AFTER decoding character references and
900 // escape sequences, because those steps can introduce comments
901 // This step cannot introduce character references or escape
902 // sequences, because it replaces comments with spaces rather
903 // than removing them completely.
904 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
906 // Remove anything after a comment-start token, to guard against
907 // incorrect client implementations.
908 $commentPos = strpos( $value, '/*' );
909 if ( $commentPos !== false ) {
910 $value = substr( $value, 0, $commentPos );
914 // S followed by repeat, iteration, or prolonged sound marks,
915 // which IE will treat as "ss"
916 $value = preg_replace(
917 '/s(?:
918 \xE3\x80\xB1 | # U+3031
919 \xE3\x82\x9D | # U+309D
920 \xE3\x83\xBC | # U+30FC
921 \xE3\x83\xBD | # U+30FD
922 \xEF\xB9\xBC | # U+FE7C
923 \xEF\xB9\xBD | # U+FE7D
924 \xEF\xBD\xB0 # U+FF70
925 )/ix',
926 'ss',
927 $value
930 return $value;
935 * Pick apart some CSS and check it for forbidden or unsafe structures.
936 * Returns a sanitized string. This sanitized string will have
937 * character references and escape sequences decoded and comments
938 * stripped (unless it is itself one valid comment, in which case the value
939 * will be passed through). If the input is just too evil, only a comment
940 * complaining about evilness will be returned.
942 * Currently URL references, 'expression', 'tps' are forbidden.
944 * NOTE: Despite the fact that character references are decoded, the
945 * returned string may contain character references given certain
946 * clever input strings. These character references must
947 * be escaped before the return value is embedded in HTML.
949 * @param string $value
950 * @return string
952 static function checkCss( $value ) {
953 $value = self::normalizeCss( $value );
955 // Reject problematic keywords and control characters
956 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ) {
957 return '/* invalid control char */';
958 } elseif ( preg_match(
959 '! expression
960 | filter\s*:
961 | accelerator\s*:
962 | -o-link\s*:
963 | -o-link-source\s*:
964 | -o-replace\s*:
965 | url\s*\(
966 | image\s*\(
967 | image-set\s*\(
968 !ix', $value ) ) {
969 return '/* insecure input */';
971 return $value;
975 * @param array $matches
976 * @return string
978 static function cssDecodeCallback( $matches ) {
979 if ( $matches[1] !== '' ) {
980 // Line continuation
981 return '';
982 } elseif ( $matches[2] !== '' ) {
983 $char = codepointToUtf8( hexdec( $matches[2] ) );
984 } elseif ( $matches[3] !== '' ) {
985 $char = $matches[3];
986 } else {
987 $char = '\\';
989 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
990 // These characters need to be escaped in strings
991 // Clean up the escape sequence to avoid parsing errors by clients
992 return '\\' . dechex( ord( $char ) ) . ' ';
993 } else {
994 // Decode unnecessary escape
995 return $char;
1000 * Take a tag soup fragment listing an HTML element's attributes
1001 * and normalize it to well-formed XML, discarding unwanted attributes.
1002 * Output is safe for further wikitext processing, with escaping of
1003 * values that could trigger problems.
1005 * - Normalizes attribute names to lowercase
1006 * - Discards attributes not on a whitelist for the given element
1007 * - Turns broken or invalid entities into plaintext
1008 * - Double-quotes all attribute values
1009 * - Attributes without values are given the name as attribute
1010 * - Double attributes are discarded
1011 * - Unsafe style attributes are discarded
1012 * - Prepends space if there are attributes.
1014 * @param string $text
1015 * @param string $element
1016 * @return string
1018 static function fixTagAttributes( $text, $element ) {
1019 if ( trim( $text ) == '' ) {
1020 return '';
1023 $decoded = Sanitizer::decodeTagAttributes( $text );
1024 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1026 return Sanitizer::safeEncodeTagAttributes( $stripped );
1030 * Encode an attribute value for HTML output.
1031 * @param string $text
1032 * @return string HTML-encoded text fragment
1034 static function encodeAttribute( $text ) {
1035 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1037 // Whitespace is normalized during attribute decoding,
1038 // so if we've been passed non-spaces we must encode them
1039 // ahead of time or they won't be preserved.
1040 $encValue = strtr( $encValue, array(
1041 "\n" => '&#10;',
1042 "\r" => '&#13;',
1043 "\t" => '&#9;',
1044 ) );
1046 return $encValue;
1050 * Encode an attribute value for HTML tags, with extra armoring
1051 * against further wiki processing.
1052 * @param string $text
1053 * @return string HTML-encoded text fragment
1055 static function safeEncodeAttribute( $text ) {
1056 $encValue = Sanitizer::encodeAttribute( $text );
1058 # Templates and links may be expanded in later parsing,
1059 # creating invalid or dangerous output. Suppress this.
1060 $encValue = strtr( $encValue, array(
1061 '<' => '&lt;', // This should never happen,
1062 '>' => '&gt;', // we've received invalid input
1063 '"' => '&quot;', // which should have been escaped.
1064 '{' => '&#123;',
1065 '[' => '&#91;',
1066 "''" => '&#39;&#39;',
1067 'ISBN' => '&#73;SBN',
1068 'RFC' => '&#82;FC',
1069 'PMID' => '&#80;MID',
1070 '|' => '&#124;',
1071 '__' => '&#95;_',
1072 ) );
1074 # Stupid hack
1075 $encValue = preg_replace_callback(
1076 '/((?i)' . wfUrlProtocols() . ')/',
1077 array( 'Sanitizer', 'armorLinksCallback' ),
1078 $encValue );
1079 return $encValue;
1083 * Given a value, escape it so that it can be used in an id attribute and
1084 * return it. This will use HTML5 validation if $wgExperimentalHtmlIds is
1085 * true, allowing anything but ASCII whitespace. Otherwise it will use
1086 * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1087 * escaped with lots of dots.
1089 * To ensure we don't have to bother escaping anything, we also strip ', ",
1090 * & even if $wgExperimentalIds is true. TODO: Is this the best tactic?
1091 * We also strip # because it upsets IE, and % because it could be
1092 * ambiguous if it's part of something that looks like a percent escape
1093 * (which don't work reliably in fragments cross-browser).
1095 * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1096 * in the id and name attributes
1097 * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1098 * the id attribute
1099 * @see http://www.whatwg.org/html/elements.html#the-id-attribute
1100 * HTML5 definition of id attribute
1102 * @param string $id Id to escape
1103 * @param string|array $options String or array of strings (default is array()):
1104 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1105 * so don't pay attention if the first character isn't valid at the
1106 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
1107 * false.
1108 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
1109 * if $wgExperimentalHtmlIds is used, so we can generate extra
1110 * anchors and links won't break.
1111 * @return string
1113 static function escapeId( $id, $options = array() ) {
1114 global $wgExperimentalHtmlIds;
1115 $options = (array)$options;
1117 $id = Sanitizer::decodeCharReferences( $id );
1119 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1120 $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1121 $id = trim( $id, '_' );
1122 if ( $id === '' ) {
1123 # Must have been all whitespace to start with.
1124 return '_';
1125 } else {
1126 return $id;
1130 # HTML4-style escaping
1131 static $replace = array(
1132 '%3A' => ':',
1133 '%' => '.'
1136 $id = urlencode( strtr( $id, ' ', '_' ) );
1137 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1139 if ( !preg_match( '/^[a-zA-Z]/', $id )
1140 && !in_array( 'noninitial', $options ) ) {
1141 // Initial character must be a letter!
1142 $id = "x$id";
1144 return $id;
1148 * Given a value, escape it so that it can be used as a CSS class and
1149 * return it.
1151 * @todo For extra validity, input should be validated UTF-8.
1153 * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1155 * @param string $class
1156 * @return string
1158 static function escapeClass( $class ) {
1159 // Convert ugly stuff to underscores and kill underscores in ugly places
1160 return rtrim( preg_replace(
1161 array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1162 '_',
1163 $class ), '_' );
1167 * Given HTML input, escape with htmlspecialchars but un-escape entities.
1168 * This allows (generally harmless) entities like &#160; to survive.
1170 * @param string $html HTML to escape
1171 * @return string Escaped input
1173 static function escapeHtmlAllowEntities( $html ) {
1174 $html = Sanitizer::decodeCharReferences( $html );
1175 # It seems wise to escape ' as well as ", as a matter of course. Can't
1176 # hurt.
1177 $html = htmlspecialchars( $html, ENT_QUOTES );
1178 return $html;
1182 * Regex replace callback for armoring links against further processing.
1183 * @param array $matches
1184 * @return string
1186 private static function armorLinksCallback( $matches ) {
1187 return str_replace( ':', '&#58;', $matches[1] );
1191 * Return an associative array of attribute names and values from
1192 * a partial tag string. Attribute names are forces to lowercase,
1193 * character references are decoded to UTF-8 text.
1195 * @param string $text
1196 * @return array
1198 public static function decodeTagAttributes( $text ) {
1199 if ( trim( $text ) == '' ) {
1200 return array();
1203 $attribs = array();
1204 $pairs = array();
1205 if ( !preg_match_all(
1206 self::getAttribsRegex(),
1207 $text,
1208 $pairs,
1209 PREG_SET_ORDER ) ) {
1210 return $attribs;
1213 foreach ( $pairs as $set ) {
1214 $attribute = strtolower( $set[1] );
1215 $value = Sanitizer::getTagAttributeCallback( $set );
1217 // Normalize whitespace
1218 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1219 $value = trim( $value );
1221 // Decode character references
1222 $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1224 return $attribs;
1228 * Build a partial tag string from an associative array of attribute
1229 * names and values as returned by decodeTagAttributes.
1231 * @param array $assoc_array
1232 * @return string
1234 public static function safeEncodeTagAttributes( $assoc_array ) {
1235 $attribs = array();
1236 foreach ( $assoc_array as $attribute => $value ) {
1237 $encAttribute = htmlspecialchars( $attribute );
1238 $encValue = Sanitizer::safeEncodeAttribute( $value );
1240 $attribs[] = "$encAttribute=\"$encValue\"";
1242 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1246 * Pick the appropriate attribute value from a match set from the
1247 * attribs regex matches.
1249 * @param array $set
1250 * @throws MWException When tag conditions are not met.
1251 * @return string
1253 private static function getTagAttributeCallback( $set ) {
1254 if ( isset( $set[6] ) ) {
1255 # Illegal #XXXXXX color with no quotes.
1256 return $set[6];
1257 } elseif ( isset( $set[5] ) ) {
1258 # No quotes.
1259 return $set[5];
1260 } elseif ( isset( $set[4] ) ) {
1261 # Single-quoted
1262 return $set[4];
1263 } elseif ( isset( $set[3] ) ) {
1264 # Double-quoted
1265 return $set[3];
1266 } elseif ( !isset( $set[2] ) ) {
1267 # In XHTML, attributes must have a value.
1268 # For 'reduced' form, return explicitly the attribute name here.
1269 return $set[1];
1270 } else {
1271 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1276 * Normalize whitespace and character references in an XML source-
1277 * encoded text for an attribute value.
1279 * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1280 * but note that we're not returning the value, but are returning
1281 * XML source fragments that will be slapped into output.
1283 * @param string $text
1284 * @return string
1285 * @todo Remove, unused?
1287 private static function normalizeAttributeValue( $text ) {
1288 return str_replace( '"', '&quot;',
1289 self::normalizeWhitespace(
1290 Sanitizer::normalizeCharReferences( $text ) ) );
1294 * @param string $text
1295 * @return string
1297 private static function normalizeWhitespace( $text ) {
1298 return preg_replace(
1299 '/\r\n|[\x20\x0d\x0a\x09]/',
1300 ' ',
1301 $text );
1305 * Normalizes whitespace in a section name, such as might be returned
1306 * by Parser::stripSectionName(), for use in the id's that are used for
1307 * section links.
1309 * @param string $section
1310 * @return string
1312 static function normalizeSectionNameWhitespace( $section ) {
1313 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1317 * Ensure that any entities and character references are legal
1318 * for XML and XHTML specifically. Any stray bits will be
1319 * &amp;-escaped to result in a valid text fragment.
1321 * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1322 * numericized (this way we're well-formed even without a DTD)
1323 * b. any numeric char refs must be legal chars, not invalid or forbidden
1324 * c. use lower cased "&#x", not "&#X"
1325 * d. fix or reject non-valid attributes
1327 * @param string $text
1328 * @return string
1329 * @private
1331 static function normalizeCharReferences( $text ) {
1332 return preg_replace_callback(
1333 self::CHAR_REFS_REGEX,
1334 array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1335 $text );
1339 * @param string $matches
1340 * @return string
1342 static function normalizeCharReferencesCallback( $matches ) {
1343 $ret = null;
1344 if ( $matches[1] != '' ) {
1345 $ret = Sanitizer::normalizeEntity( $matches[1] );
1346 } elseif ( $matches[2] != '' ) {
1347 $ret = Sanitizer::decCharReference( $matches[2] );
1348 } elseif ( $matches[3] != '' ) {
1349 $ret = Sanitizer::hexCharReference( $matches[3] );
1351 if ( is_null( $ret ) ) {
1352 return htmlspecialchars( $matches[0] );
1353 } else {
1354 return $ret;
1359 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1360 * return the equivalent numeric entity reference (except for the core &lt;
1361 * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1362 * the HTML equivalent. Otherwise, returns HTML-escaped text of
1363 * pseudo-entity source (eg &amp;foo;)
1365 * @param string $name
1366 * @return string
1368 static function normalizeEntity( $name ) {
1369 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1370 return '&' . self::$htmlEntityAliases[$name] . ';';
1371 } elseif ( in_array( $name,
1372 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1373 return "&$name;";
1374 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1375 return '&#' . self::$htmlEntities[$name] . ';';
1376 } else {
1377 return "&amp;$name;";
1382 * @param int $codepoint
1383 * @return null|string
1385 static function decCharReference( $codepoint ) {
1386 $point = intval( $codepoint );
1387 if ( Sanitizer::validateCodepoint( $point ) ) {
1388 return sprintf( '&#%d;', $point );
1389 } else {
1390 return null;
1395 * @param int $codepoint
1396 * @return null|string
1398 static function hexCharReference( $codepoint ) {
1399 $point = hexdec( $codepoint );
1400 if ( Sanitizer::validateCodepoint( $point ) ) {
1401 return sprintf( '&#x%x;', $point );
1402 } else {
1403 return null;
1408 * Returns true if a given Unicode codepoint is a valid character in XML.
1409 * @param int $codepoint
1410 * @return bool
1412 private static function validateCodepoint( $codepoint ) {
1413 return $codepoint == 0x09
1414 || $codepoint == 0x0a
1415 || $codepoint == 0x0d
1416 || ( $codepoint >= 0x20 && $codepoint <= 0xd7ff )
1417 || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1418 || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1422 * Decode any character references, numeric or named entities,
1423 * in the text and return a UTF-8 string.
1425 * @param string $text
1426 * @return string
1428 public static function decodeCharReferences( $text ) {
1429 return preg_replace_callback(
1430 self::CHAR_REFS_REGEX,
1431 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1432 $text );
1436 * Decode any character references, numeric or named entities,
1437 * in the next and normalize the resulting string. (bug 14952)
1439 * This is useful for page titles, not for text to be displayed,
1440 * MediaWiki allows HTML entities to escape normalization as a feature.
1442 * @param string $text Already normalized, containing entities
1443 * @return string Still normalized, without entities
1445 public static function decodeCharReferencesAndNormalize( $text ) {
1446 global $wgContLang;
1447 $text = preg_replace_callback(
1448 self::CHAR_REFS_REGEX,
1449 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1450 $text, /* limit */ -1, $count );
1452 if ( $count ) {
1453 return $wgContLang->normalize( $text );
1454 } else {
1455 return $text;
1460 * @param string $matches
1461 * @return string
1463 static function decodeCharReferencesCallback( $matches ) {
1464 if ( $matches[1] != '' ) {
1465 return Sanitizer::decodeEntity( $matches[1] );
1466 } elseif ( $matches[2] != '' ) {
1467 return Sanitizer::decodeChar( intval( $matches[2] ) );
1468 } elseif ( $matches[3] != '' ) {
1469 return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1471 # Last case should be an ampersand by itself
1472 return $matches[0];
1476 * Return UTF-8 string for a codepoint if that is a valid
1477 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1478 * @param int $codepoint
1479 * @return string
1480 * @private
1482 static function decodeChar( $codepoint ) {
1483 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1484 return codepointToUtf8( $codepoint );
1485 } else {
1486 return UTF8_REPLACEMENT;
1491 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1492 * return the UTF-8 encoding of that character. Otherwise, returns
1493 * pseudo-entity source (eg "&foo;")
1495 * @param string $name
1496 * @return string
1498 static function decodeEntity( $name ) {
1499 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1500 $name = self::$htmlEntityAliases[$name];
1502 if ( isset( self::$htmlEntities[$name] ) ) {
1503 return codepointToUtf8( self::$htmlEntities[$name] );
1504 } else {
1505 return "&$name;";
1510 * Fetch the whitelist of acceptable attributes for a given element name.
1512 * @param string $element
1513 * @return array
1515 static function attributeWhitelist( $element ) {
1516 $list = Sanitizer::setupAttributeWhitelist();
1517 return isset( $list[$element] )
1518 ? $list[$element]
1519 : array();
1523 * Foreach array key (an allowed HTML element), return an array
1524 * of allowed attributes
1525 * @return array
1527 static function setupAttributeWhitelist() {
1528 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
1529 static $whitelist, $staticInitialised;
1531 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgAllowMicrodataAttributes' ) );
1533 if ( $whitelist !== null && $staticInitialised == $globalContext ) {
1534 return $whitelist;
1537 $common = array(
1538 # HTML
1539 'id',
1540 'class',
1541 'style',
1542 'lang',
1543 'dir',
1544 'title',
1546 # WAI-ARIA
1547 'role',
1550 if ( $wgAllowRdfaAttributes ) {
1551 # RDFa attributes as specified in section 9 of
1552 # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1553 $common = array_merge( $common, array(
1554 'about', 'property', 'resource', 'datatype', 'typeof',
1555 ) );
1558 if ( $wgAllowMicrodataAttributes ) {
1559 # add HTML5 microdata tags as specified by
1560 # http://www.whatwg.org/html/microdata.html#the-microdata-model
1561 $common = array_merge( $common, array(
1562 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1563 ) );
1566 $block = array_merge( $common, array( 'align' ) );
1567 $tablealign = array( 'align', 'valign' );
1568 $tablecell = array(
1569 'abbr',
1570 'axis',
1571 'headers',
1572 'scope',
1573 'rowspan',
1574 'colspan',
1575 'nowrap', # deprecated
1576 'width', # deprecated
1577 'height', # deprecated
1578 'bgcolor', # deprecated
1581 # Numbers refer to sections in HTML 4.01 standard describing the element.
1582 # See: http://www.w3.org/TR/html4/
1583 $whitelist = array(
1584 # 7.5.4
1585 'div' => $block,
1586 'center' => $common, # deprecated
1587 'span' => $common,
1589 # 7.5.5
1590 'h1' => $block,
1591 'h2' => $block,
1592 'h3' => $block,
1593 'h4' => $block,
1594 'h5' => $block,
1595 'h6' => $block,
1597 # 7.5.6
1598 # address
1600 # 8.2.4
1601 'bdo' => $common,
1603 # 9.2.1
1604 'em' => $common,
1605 'strong' => $common,
1606 'cite' => $common,
1607 'dfn' => $common,
1608 'code' => $common,
1609 'samp' => $common,
1610 'kbd' => $common,
1611 'var' => $common,
1612 'abbr' => $common,
1613 # acronym
1615 # 9.2.2
1616 'blockquote' => array_merge( $common, array( 'cite' ) ),
1617 'q' => array_merge( $common, array( 'cite' ) ),
1619 # 9.2.3
1620 'sub' => $common,
1621 'sup' => $common,
1623 # 9.3.1
1624 'p' => $block,
1626 # 9.3.2
1627 'br' => array_merge( $common, array( 'clear' ) ),
1629 # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1630 'wbr' => $common,
1632 # 9.3.4
1633 'pre' => array_merge( $common, array( 'width' ) ),
1635 # 9.4
1636 'ins' => array_merge( $common, array( 'cite', 'datetime' ) ),
1637 'del' => array_merge( $common, array( 'cite', 'datetime' ) ),
1639 # 10.2
1640 'ul' => array_merge( $common, array( 'type' ) ),
1641 'ol' => array_merge( $common, array( 'type', 'start' ) ),
1642 'li' => array_merge( $common, array( 'type', 'value' ) ),
1644 # 10.3
1645 'dl' => $common,
1646 'dd' => $common,
1647 'dt' => $common,
1649 # 11.2.1
1650 'table' => array_merge( $common,
1651 array( 'summary', 'width', 'border', 'frame',
1652 'rules', 'cellspacing', 'cellpadding',
1653 'align', 'bgcolor',
1654 ) ),
1656 # 11.2.2
1657 'caption' => $block,
1659 # 11.2.3
1660 'thead' => $common,
1661 'tfoot' => $common,
1662 'tbody' => $common,
1664 # 11.2.4
1665 'colgroup' => array_merge( $common, array( 'span' ) ),
1666 'col' => array_merge( $common, array( 'span' ) ),
1668 # 11.2.5
1669 'tr' => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1671 # 11.2.6
1672 'td' => array_merge( $common, $tablecell, $tablealign ),
1673 'th' => array_merge( $common, $tablecell, $tablealign ),
1675 # 12.2
1676 # NOTE: <a> is not allowed directly, but the attrib
1677 # whitelist is used from the Parser object
1678 'a' => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1680 # 13.2
1681 # Not usually allowed, but may be used for extension-style hooks
1682 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1683 # true
1684 'img' => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1686 # 15.2.1
1687 'tt' => $common,
1688 'b' => $common,
1689 'i' => $common,
1690 'big' => $common,
1691 'small' => $common,
1692 'strike' => $common,
1693 's' => $common,
1694 'u' => $common,
1696 # 15.2.2
1697 'font' => array_merge( $common, array( 'size', 'color', 'face' ) ),
1698 # basefont
1700 # 15.3
1701 'hr' => array_merge( $common, array( 'width' ) ),
1703 # HTML Ruby annotation text module, simple ruby only.
1704 # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
1705 'ruby' => $common,
1706 # rbc
1707 'rb' => $common,
1708 'rp' => $common,
1709 'rt' => $common, #array_merge( $common, array( 'rbspan' ) ),
1710 'rtc' => $common,
1712 # MathML root element, where used for extensions
1713 # 'title' may not be 100% valid here; it's XHTML
1714 # http://www.w3.org/TR/REC-MathML/
1715 'math' => array( 'class', 'style', 'id', 'title' ),
1717 # HTML 5 section 4.6
1718 'bdi' => $common,
1720 # HTML5 elements, defined by:
1721 # http://www.whatwg.org/html/
1722 'data' => array_merge( $common, array( 'value' ) ),
1723 'time' => array_merge( $common, array( 'datetime' ) ),
1724 'mark' => $common,
1726 // meta and link are only permitted by removeHTMLtags when Microdata
1727 // is enabled so we don't bother adding a conditional to hide these
1728 // Also meta and link are only valid in WikiText as Microdata elements
1729 // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1730 // So we don't bother including $common attributes that have no purpose.
1731 'meta' => array( 'itemprop', 'content' ),
1732 'link' => array( 'itemprop', 'href' ),
1735 $staticInitialised = $globalContext;
1737 return $whitelist;
1741 * Take a fragment of (potentially invalid) HTML and return
1742 * a version with any tags removed, encoded as plain text.
1744 * Warning: this return value must be further escaped for literal
1745 * inclusion in HTML output as of 1.10!
1747 * @param string $text HTML fragment
1748 * @return string
1750 static function stripAllTags( $text ) {
1751 # Actual <tags>
1752 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1754 # Normalize &entities and whitespace
1755 $text = self::decodeCharReferences( $text );
1756 $text = self::normalizeWhitespace( $text );
1758 return $text;
1762 * Hack up a private DOCTYPE with HTML's standard entity declarations.
1763 * PHP 4 seemed to know these if you gave it an HTML doctype, but
1764 * PHP 5.1 doesn't.
1766 * Use for passing XHTML fragments to PHP's XML parsing functions
1768 * @return string
1770 static function hackDocType() {
1771 $out = "<!DOCTYPE html [\n";
1772 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1773 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1775 $out .= "]>\n";
1776 return $out;
1780 * @param string $url
1781 * @return mixed|string
1783 static function cleanUrl( $url ) {
1784 # Normalize any HTML entities in input. They will be
1785 # re-escaped by makeExternalLink().
1786 $url = Sanitizer::decodeCharReferences( $url );
1788 # Escape any control characters introduced by the above step
1789 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1790 array( __CLASS__, 'cleanUrlCallback' ), $url );
1792 # Validate hostname portion
1793 $matches = array();
1794 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1795 list( /* $whole */, $protocol, $host, $rest ) = $matches;
1797 // Characters that will be ignored in IDNs.
1798 // http://tools.ietf.org/html/3454#section-3.1
1799 // Strip them before further processing so blacklists and such work.
1800 $strip = "/
1801 \\s| # general whitespace
1802 \xc2\xad| # 00ad SOFT HYPHEN
1803 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1804 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1805 \xe2\x81\xa0| # 2060 WORD JOINER
1806 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1807 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
1808 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1809 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1810 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1811 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1812 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1813 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1814 /xuD";
1816 $host = preg_replace( $strip, '', $host );
1818 // @todo FIXME: Validate hostnames here
1820 return $protocol . $host . $rest;
1821 } else {
1822 return $url;
1827 * @param array $matches
1828 * @return string
1830 static function cleanUrlCallback( $matches ) {
1831 return urlencode( $matches[0] );
1835 * Does a string look like an e-mail address?
1837 * This validates an email address using an HTML5 specification found at:
1838 * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1839 * Which as of 2011-01-24 says:
1841 * A valid e-mail address is a string that matches the ABNF production
1842 * 1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1843 * in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1844 * 3.5.
1846 * This function is an implementation of the specification as requested in
1847 * bug 22449.
1849 * Client-side forms will use the same standard validation rules via JS or
1850 * HTML 5 validation; additional restrictions can be enforced server-side
1851 * by extensions via the 'isValidEmailAddr' hook.
1853 * Note that this validation doesn't 100% match RFC 2822, but is believed
1854 * to be liberal enough for wide use. Some invalid addresses will still
1855 * pass validation here.
1857 * @since 1.18
1859 * @param string $addr E-mail address
1860 * @return bool
1862 public static function validateEmail( $addr ) {
1863 $result = null;
1864 if ( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1865 return $result;
1868 // Please note strings below are enclosed in brackets [], this make the
1869 // hyphen "-" a range indicator. Hence it is double backslashed below.
1870 // See bug 26948
1871 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1872 $rfc1034_ldh_str = "a-z0-9\\-";
1874 $html5_email_regexp = "/
1875 ^ # start of string
1876 [$rfc5322_atext\\.]+ # user part which is liberal :p
1877 @ # 'apostrophe'
1878 [$rfc1034_ldh_str]+ # First domain part
1879 (\\.[$rfc1034_ldh_str]+)* # Following part prefixed with a dot
1880 $ # End of string
1881 /ix"; // case Insensitive, eXtended
1883 return (bool)preg_match( $html5_email_regexp, $addr );