PHPSessionHandler: Implement SessionHandlerInterface
[mediawiki.git] / includes / Sanitizer.php
blob60c949893823b5f7c8d97140b770e86e442cc5c3
1 <?php
2 /**
3 * HTML sanitizer for %MediaWiki.
5 * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
6 * https://www.mediawiki.org/
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 2 of the License, or
11 * (at your option) any later version.
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License along
19 * with this program; if not, write to the Free Software Foundation, Inc.,
20 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
21 * http://www.gnu.org/copyleft/gpl.html
23 * @file
24 * @ingroup Parser
27 /**
28 * HTML sanitizer for MediaWiki
29 * @ingroup Parser
31 class Sanitizer {
32 /**
33 * Regular expression to match various types of character references in
34 * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
36 const CHAR_REFS_REGEX =
37 '/&([A-Za-z0-9\x80-\xff]+);
38 |&\#([0-9]+);
39 |&\#[xX]([0-9A-Fa-f]+);
40 |(&)/x';
42 /**
43 * Acceptable tag name charset from HTML5 parsing spec
44 * http://www.w3.org/TR/html5/syntax.html#tag-open-state
46 const ELEMENT_BITS_REGEX = '!^(/?)([A-Za-z][^\t\n\v />\0]*+)([^>]*?)(/?>)([^<]*)$!';
48 /**
49 * Blacklist for evil uris like javascript:
50 * WARNING: DO NOT use this in any place that actually requires blacklisting
51 * for security reasons. There are NUMEROUS[1] ways to bypass blacklisting, the
52 * only way to be secure from javascript: uri based xss vectors is to whitelist
53 * things that you know are safe and deny everything else.
54 * [1]: http://ha.ckers.org/xss.html
56 const EVIL_URI_PATTERN = '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
57 const XMLNS_ATTRIBUTE_PATTERN = "/^xmlns:[:A-Z_a-z-.0-9]+$/";
59 /**
60 * List of all named character entities defined in HTML 4.01
61 * http://www.w3.org/TR/html4/sgml/entities.html
62 * As well as &apos; which is only defined starting in XHTML1.
64 private static $htmlEntities = array(
65 'Aacute' => 193,
66 'aacute' => 225,
67 'Acirc' => 194,
68 'acirc' => 226,
69 'acute' => 180,
70 'AElig' => 198,
71 'aelig' => 230,
72 'Agrave' => 192,
73 'agrave' => 224,
74 'alefsym' => 8501,
75 'Alpha' => 913,
76 'alpha' => 945,
77 'amp' => 38,
78 'and' => 8743,
79 'ang' => 8736,
80 'apos' => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
81 'Aring' => 197,
82 'aring' => 229,
83 'asymp' => 8776,
84 'Atilde' => 195,
85 'atilde' => 227,
86 'Auml' => 196,
87 'auml' => 228,
88 'bdquo' => 8222,
89 'Beta' => 914,
90 'beta' => 946,
91 'brvbar' => 166,
92 'bull' => 8226,
93 'cap' => 8745,
94 'Ccedil' => 199,
95 'ccedil' => 231,
96 'cedil' => 184,
97 'cent' => 162,
98 'Chi' => 935,
99 'chi' => 967,
100 'circ' => 710,
101 'clubs' => 9827,
102 'cong' => 8773,
103 'copy' => 169,
104 'crarr' => 8629,
105 'cup' => 8746,
106 'curren' => 164,
107 'dagger' => 8224,
108 'Dagger' => 8225,
109 'darr' => 8595,
110 'dArr' => 8659,
111 'deg' => 176,
112 'Delta' => 916,
113 'delta' => 948,
114 'diams' => 9830,
115 'divide' => 247,
116 'Eacute' => 201,
117 'eacute' => 233,
118 'Ecirc' => 202,
119 'ecirc' => 234,
120 'Egrave' => 200,
121 'egrave' => 232,
122 'empty' => 8709,
123 'emsp' => 8195,
124 'ensp' => 8194,
125 'Epsilon' => 917,
126 'epsilon' => 949,
127 'equiv' => 8801,
128 'Eta' => 919,
129 'eta' => 951,
130 'ETH' => 208,
131 'eth' => 240,
132 'Euml' => 203,
133 'euml' => 235,
134 'euro' => 8364,
135 'exist' => 8707,
136 'fnof' => 402,
137 'forall' => 8704,
138 'frac12' => 189,
139 'frac14' => 188,
140 'frac34' => 190,
141 'frasl' => 8260,
142 'Gamma' => 915,
143 'gamma' => 947,
144 'ge' => 8805,
145 'gt' => 62,
146 'harr' => 8596,
147 'hArr' => 8660,
148 'hearts' => 9829,
149 'hellip' => 8230,
150 'Iacute' => 205,
151 'iacute' => 237,
152 'Icirc' => 206,
153 'icirc' => 238,
154 'iexcl' => 161,
155 'Igrave' => 204,
156 'igrave' => 236,
157 'image' => 8465,
158 'infin' => 8734,
159 'int' => 8747,
160 'Iota' => 921,
161 'iota' => 953,
162 'iquest' => 191,
163 'isin' => 8712,
164 'Iuml' => 207,
165 'iuml' => 239,
166 'Kappa' => 922,
167 'kappa' => 954,
168 'Lambda' => 923,
169 'lambda' => 955,
170 'lang' => 9001,
171 'laquo' => 171,
172 'larr' => 8592,
173 'lArr' => 8656,
174 'lceil' => 8968,
175 'ldquo' => 8220,
176 'le' => 8804,
177 'lfloor' => 8970,
178 'lowast' => 8727,
179 'loz' => 9674,
180 'lrm' => 8206,
181 'lsaquo' => 8249,
182 'lsquo' => 8216,
183 'lt' => 60,
184 'macr' => 175,
185 'mdash' => 8212,
186 'micro' => 181,
187 'middot' => 183,
188 'minus' => 8722,
189 'Mu' => 924,
190 'mu' => 956,
191 'nabla' => 8711,
192 'nbsp' => 160,
193 'ndash' => 8211,
194 'ne' => 8800,
195 'ni' => 8715,
196 'not' => 172,
197 'notin' => 8713,
198 'nsub' => 8836,
199 'Ntilde' => 209,
200 'ntilde' => 241,
201 'Nu' => 925,
202 'nu' => 957,
203 'Oacute' => 211,
204 'oacute' => 243,
205 'Ocirc' => 212,
206 'ocirc' => 244,
207 'OElig' => 338,
208 'oelig' => 339,
209 'Ograve' => 210,
210 'ograve' => 242,
211 'oline' => 8254,
212 'Omega' => 937,
213 'omega' => 969,
214 'Omicron' => 927,
215 'omicron' => 959,
216 'oplus' => 8853,
217 'or' => 8744,
218 'ordf' => 170,
219 'ordm' => 186,
220 'Oslash' => 216,
221 'oslash' => 248,
222 'Otilde' => 213,
223 'otilde' => 245,
224 'otimes' => 8855,
225 'Ouml' => 214,
226 'ouml' => 246,
227 'para' => 182,
228 'part' => 8706,
229 'permil' => 8240,
230 'perp' => 8869,
231 'Phi' => 934,
232 'phi' => 966,
233 'Pi' => 928,
234 'pi' => 960,
235 'piv' => 982,
236 'plusmn' => 177,
237 'pound' => 163,
238 'prime' => 8242,
239 'Prime' => 8243,
240 'prod' => 8719,
241 'prop' => 8733,
242 'Psi' => 936,
243 'psi' => 968,
244 'quot' => 34,
245 'radic' => 8730,
246 'rang' => 9002,
247 'raquo' => 187,
248 'rarr' => 8594,
249 'rArr' => 8658,
250 'rceil' => 8969,
251 'rdquo' => 8221,
252 'real' => 8476,
253 'reg' => 174,
254 'rfloor' => 8971,
255 'Rho' => 929,
256 'rho' => 961,
257 'rlm' => 8207,
258 'rsaquo' => 8250,
259 'rsquo' => 8217,
260 'sbquo' => 8218,
261 'Scaron' => 352,
262 'scaron' => 353,
263 'sdot' => 8901,
264 'sect' => 167,
265 'shy' => 173,
266 'Sigma' => 931,
267 'sigma' => 963,
268 'sigmaf' => 962,
269 'sim' => 8764,
270 'spades' => 9824,
271 'sub' => 8834,
272 'sube' => 8838,
273 'sum' => 8721,
274 'sup' => 8835,
275 'sup1' => 185,
276 'sup2' => 178,
277 'sup3' => 179,
278 'supe' => 8839,
279 'szlig' => 223,
280 'Tau' => 932,
281 'tau' => 964,
282 'there4' => 8756,
283 'Theta' => 920,
284 'theta' => 952,
285 'thetasym' => 977,
286 'thinsp' => 8201,
287 'THORN' => 222,
288 'thorn' => 254,
289 'tilde' => 732,
290 'times' => 215,
291 'trade' => 8482,
292 'Uacute' => 218,
293 'uacute' => 250,
294 'uarr' => 8593,
295 'uArr' => 8657,
296 'Ucirc' => 219,
297 'ucirc' => 251,
298 'Ugrave' => 217,
299 'ugrave' => 249,
300 'uml' => 168,
301 'upsih' => 978,
302 'Upsilon' => 933,
303 'upsilon' => 965,
304 'Uuml' => 220,
305 'uuml' => 252,
306 'weierp' => 8472,
307 'Xi' => 926,
308 'xi' => 958,
309 'Yacute' => 221,
310 'yacute' => 253,
311 'yen' => 165,
312 'Yuml' => 376,
313 'yuml' => 255,
314 'Zeta' => 918,
315 'zeta' => 950,
316 'zwj' => 8205,
317 'zwnj' => 8204
321 * Character entity aliases accepted by MediaWiki
323 private static $htmlEntityAliases = array(
324 'רלמ' => 'rlm',
325 'رلم' => 'rlm',
329 * Lazy-initialised attributes regex, see getAttribsRegex()
331 private static $attribsRegex;
334 * Regular expression to match HTML/XML attribute pairs within a tag.
335 * Allows some... latitude.
336 * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
337 * @return string
339 static function getAttribsRegex() {
340 if ( self::$attribsRegex === null ) {
341 $attribFirst = '[:A-Z_a-z0-9]';
342 $attrib = '[:A-Z_a-z-.0-9]';
343 $space = '[\x09\x0a\x0d\x20]';
344 self::$attribsRegex =
345 "/(?:^|$space)({$attribFirst}{$attrib}*)
346 ($space*=$space*
348 # The attribute value: quoted or alone
349 \"([^<\"]*)(?:\"|\$)
350 | '([^<']*)(?:'|\$)
351 | ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
353 )?(?=$space|\$)/sx";
355 return self::$attribsRegex;
359 * Return the various lists of recognized tags
360 * @param array $extratags For any extra tags to include
361 * @param array $removetags For any tags (default or extra) to exclude
362 * @return array
364 public static function getRecognizedTagData( $extratags = array(), $removetags = array() ) {
365 global $wgAllowMicrodataAttributes, $wgAllowImageTag;
367 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
368 $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
370 // Base our staticInitialised variable off of the global config state so that if the globals
371 // are changed (like in the screwed up test system) we will re-initialise the settings.
372 $globalContext = implode( '-', compact( 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
373 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
374 $htmlpairsStatic = array( # Tags that must be closed
375 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
376 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
377 'strike', 'strong', 'tt', 'var', 'div', 'center',
378 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
379 'ruby', 'rb', 'rp', 'rt', 'rtc', 'p', 'span', 'abbr', 'dfn',
380 'kbd', 'samp', 'data', 'time', 'mark'
382 $htmlsingle = array(
383 'br', 'wbr', 'hr', 'li', 'dt', 'dd'
385 $htmlsingleonly = array( # Elements that cannot have close tags
386 'br', 'wbr', 'hr'
388 if ( $wgAllowMicrodataAttributes ) {
389 $htmlsingle[] = $htmlsingleonly[] = 'meta';
390 $htmlsingle[] = $htmlsingleonly[] = 'link';
392 $htmlnest = array( # Tags that can be nested--??
393 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
394 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
395 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
397 $tabletags = array( # Can only appear inside table, we will close them
398 'td', 'th', 'tr',
400 $htmllist = array( # Tags used by list
401 'ul', 'ol',
403 $listtags = array( # Tags that can appear in a list
404 'li',
407 if ( $wgAllowImageTag ) {
408 $htmlsingle[] = 'img';
409 $htmlsingleonly[] = 'img';
412 $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
413 $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
415 # Convert them all to hashtables for faster lookup
416 $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
417 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
418 foreach ( $vars as $var ) {
419 $$var = array_flip( $$var );
421 $staticInitialised = $globalContext;
424 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
425 $extratags = array_flip( $extratags );
426 $removetags = array_flip( $removetags );
427 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
428 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
430 return array(
431 'htmlpairs' => $htmlpairs,
432 'htmlsingle' => $htmlsingle,
433 'htmlsingleonly' => $htmlsingleonly,
434 'htmlnest' => $htmlnest,
435 'tabletags' => $tabletags,
436 'htmllist' => $htmllist,
437 'listtags' => $listtags,
438 'htmlsingleallowed' => $htmlsingleallowed,
439 'htmlelements' => $htmlelements,
444 * Cleans up HTML, removes dangerous tags and attributes, and
445 * removes HTML comments
446 * @param string $text
447 * @param callable $processCallback Callback to do any variable or parameter
448 * replacements in HTML attribute values
449 * @param array|bool $args Arguments for the processing callback
450 * @param array $extratags For any extra tags to include
451 * @param array $removetags For any tags (default or extra) to exclude
452 * @return string
454 public static function removeHTMLtags( $text, $processCallback = null,
455 $args = array(), $extratags = array(), $removetags = array()
457 extract( self::getRecognizedTagData( $extratags, $removetags ) );
459 # Remove HTML comments
460 $text = Sanitizer::removeHTMLcomments( $text );
461 $bits = explode( '<', $text );
462 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
463 if ( !MWTidy::isEnabled() ) {
464 $tagstack = $tablestack = array();
465 foreach ( $bits as $x ) {
466 $regs = array();
467 # $slash: Does the current element start with a '/'?
468 # $t: Current element name
469 # $params: String between element name and >
470 # $brace: Ending '>' or '/>'
471 # $rest: Everything until the next element of $bits
472 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
473 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
474 } else {
475 $slash = $t = $params = $brace = $rest = null;
478 $badtag = false;
479 $t = strtolower( $t );
480 if ( isset( $htmlelements[$t] ) ) {
481 # Check our stack
482 if ( $slash && isset( $htmlsingleonly[$t] ) ) {
483 $badtag = true;
484 } elseif ( $slash ) {
485 # Closing a tag... is it the one we just opened?
486 MediaWiki\suppressWarnings();
487 $ot = array_pop( $tagstack );
488 MediaWiki\restoreWarnings();
490 if ( $ot != $t ) {
491 if ( isset( $htmlsingleallowed[$ot] ) ) {
492 # Pop all elements with an optional close tag
493 # and see if we find a match below them
494 $optstack = array();
495 array_push( $optstack, $ot );
496 MediaWiki\suppressWarnings();
497 $ot = array_pop( $tagstack );
498 MediaWiki\restoreWarnings();
499 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
500 array_push( $optstack, $ot );
501 MediaWiki\suppressWarnings();
502 $ot = array_pop( $tagstack );
503 MediaWiki\restoreWarnings();
505 if ( $t != $ot ) {
506 # No match. Push the optional elements back again
507 $badtag = true;
508 MediaWiki\suppressWarnings();
509 $ot = array_pop( $optstack );
510 MediaWiki\restoreWarnings();
511 while ( $ot ) {
512 array_push( $tagstack, $ot );
513 MediaWiki\suppressWarnings();
514 $ot = array_pop( $optstack );
515 MediaWiki\restoreWarnings();
518 } else {
519 MediaWiki\suppressWarnings();
520 array_push( $tagstack, $ot );
521 MediaWiki\restoreWarnings();
523 # <li> can be nested in <ul> or <ol>, skip those cases:
524 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
525 $badtag = true;
528 } else {
529 if ( $t == 'table' ) {
530 $tagstack = array_pop( $tablestack );
533 $newparams = '';
534 } else {
535 # Keep track for later
536 if ( isset( $tabletags[$t] ) && !in_array( 'table', $tagstack ) ) {
537 $badtag = true;
538 } elseif ( in_array( $t, $tagstack ) && !isset( $htmlnest[$t] ) ) {
539 $badtag = true;
540 #  Is it a self closed htmlpair ? (bug 5487)
541 } elseif ( $brace == '/>' && isset( $htmlpairs[$t] ) ) {
542 $badtag = true;
543 } elseif ( isset( $htmlsingleonly[$t] ) ) {
544 # Hack to force empty tag for unclosable elements
545 $brace = '/>';
546 } elseif ( isset( $htmlsingle[$t] ) ) {
547 # Hack to not close $htmlsingle tags
548 $brace = null;
549 # Still need to push this optionally-closed tag to
550 # the tag stack so that we can match end tags
551 # instead of marking them as bad.
552 array_push( $tagstack, $t );
553 } elseif ( isset( $tabletags[$t] ) && in_array( $t, $tagstack ) ) {
554 // New table tag but forgot to close the previous one
555 $text .= "</$t>";
556 } else {
557 if ( $t == 'table' ) {
558 array_push( $tablestack, $tagstack );
559 $tagstack = array();
561 array_push( $tagstack, $t );
564 # Replace any variables or template parameters with
565 # plaintext results.
566 if ( is_callable( $processCallback ) ) {
567 call_user_func_array( $processCallback, array( &$params, $args ) );
570 if ( !Sanitizer::validateTag( $params, $t ) ) {
571 $badtag = true;
574 # Strip non-approved attributes from the tag
575 $newparams = Sanitizer::fixTagAttributes( $params, $t );
577 if ( !$badtag ) {
578 $rest = str_replace( '>', '&gt;', $rest );
579 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
580 $text .= "<$slash$t$newparams$close>$rest";
581 continue;
584 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
586 # Close off any remaining tags
587 while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
588 $text .= "</$t>\n";
589 if ( $t == 'table' ) {
590 $tagstack = array_pop( $tablestack );
593 } else {
594 # this might be possible using tidy itself
595 foreach ( $bits as $x ) {
596 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
597 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
599 $badtag = false;
600 $t = strtolower( $t );
601 if ( isset( $htmlelements[$t] ) ) {
602 if ( is_callable( $processCallback ) ) {
603 call_user_func_array( $processCallback, array( &$params, $args ) );
606 if ( !Sanitizer::validateTag( $params, $t ) ) {
607 $badtag = true;
610 $newparams = Sanitizer::fixTagAttributes( $params, $t );
611 if ( !$badtag ) {
612 $rest = str_replace( '>', '&gt;', $rest );
613 $text .= "<$slash$t$newparams$brace$rest";
614 continue;
618 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
621 return $text;
625 * Remove '<!--', '-->', and everything between.
626 * To avoid leaving blank lines, when a comment is both preceded
627 * and followed by a newline (ignoring spaces), trim leading and
628 * trailing spaces and one of the newlines.
630 * @param string $text
631 * @return string
633 public static function removeHTMLcomments( $text ) {
634 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
635 $end = strpos( $text, '-->', $start + 4 );
636 if ( $end === false ) {
637 # Unterminated comment; bail out
638 break;
641 $end += 3;
643 # Trim space and newline if the comment is both
644 # preceded and followed by a newline
645 $spaceStart = max( $start - 1, 0 );
646 $spaceLen = $end - $spaceStart;
647 while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
648 $spaceStart--;
649 $spaceLen++;
651 while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
652 $spaceLen++;
654 if ( substr( $text, $spaceStart, 1 ) === "\n"
655 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
656 # Remove the comment, leading and trailing
657 # spaces, and leave only one newline.
658 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
659 } else {
660 # Remove just the comment.
661 $text = substr_replace( $text, '', $start, $end - $start );
664 return $text;
668 * Takes attribute names and values for a tag and the tag name and
669 * validates that the tag is allowed to be present.
670 * This DOES NOT validate the attributes, nor does it validate the
671 * tags themselves. This method only handles the special circumstances
672 * where we may want to allow a tag within content but ONLY when it has
673 * specific attributes set.
675 * @param string $params
676 * @param string $element
677 * @return bool
679 static function validateTag( $params, $element ) {
680 $params = Sanitizer::decodeTagAttributes( $params );
682 if ( $element == 'meta' || $element == 'link' ) {
683 if ( !isset( $params['itemprop'] ) ) {
684 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
685 return false;
687 if ( $element == 'meta' && !isset( $params['content'] ) ) {
688 // <meta> must have a content="" for the itemprop
689 return false;
691 if ( $element == 'link' && !isset( $params['href'] ) ) {
692 // <link> must have an associated href=""
693 return false;
697 return true;
701 * Take an array of attribute names and values and normalize or discard
702 * illegal values for the given element type.
704 * - Discards attributes not on a whitelist for the given element
705 * - Unsafe style attributes are discarded
706 * - Invalid id attributes are re-encoded
708 * @param array $attribs
709 * @param string $element
710 * @return array
712 * @todo Check for legal values where the DTD limits things.
713 * @todo Check for unique id attribute :P
715 static function validateTagAttributes( $attribs, $element ) {
716 return Sanitizer::validateAttributes( $attribs,
717 Sanitizer::attributeWhitelist( $element ) );
721 * Take an array of attribute names and values and normalize or discard
722 * illegal values for the given whitelist.
724 * - Discards attributes not on the given whitelist
725 * - Unsafe style attributes are discarded
726 * - Invalid id attributes are re-encoded
728 * @param array $attribs
729 * @param array $whitelist List of allowed attribute names
730 * @return array
732 * @todo Check for legal values where the DTD limits things.
733 * @todo Check for unique id attribute :P
735 static function validateAttributes( $attribs, $whitelist ) {
736 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
738 $whitelist = array_flip( $whitelist );
739 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
741 $out = array();
742 foreach ( $attribs as $attribute => $value ) {
743 # allow XML namespace declaration if RDFa is enabled
744 if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
745 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
746 $out[$attribute] = $value;
749 continue;
752 # Allow any attribute beginning with "data-"
753 # However:
754 # * data-ooui is reserved for ooui
755 # * data-mw and data-parsoid are reserved for parsoid
756 # * data-mw-<name here> is reserved for extensions (or core) if
757 # they need to communicate some data to the client and want to be
758 # sure that it isn't coming from an untrusted user.
759 # * Ensure that the attribute is not namespaced by banning
760 # colons.
761 if ( !preg_match( '/^data-(?!ooui|mw|parsoid)[^:]*$/i', $attribute )
762 && !isset( $whitelist[$attribute] )
764 continue;
767 # Strip javascript "expression" from stylesheets.
768 # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
769 if ( $attribute == 'style' ) {
770 $value = Sanitizer::checkCss( $value );
773 # Escape HTML id attributes
774 if ( $attribute === 'id' ) {
775 $value = Sanitizer::escapeId( $value, 'noninitial' );
778 # Escape HTML id reference lists
779 if ( $attribute === 'aria-describedby'
780 || $attribute === 'aria-flowto'
781 || $attribute === 'aria-labelledby'
782 || $attribute === 'aria-owns'
784 $value = Sanitizer::escapeIdReferenceList( $value, 'noninitial' );
787 // RDFa and microdata properties allow URLs, URIs and/or CURIs.
788 // Check them for sanity.
789 if ( $attribute === 'rel' || $attribute === 'rev'
790 # RDFa
791 || $attribute === 'about' || $attribute === 'property'
792 || $attribute === 'resource' || $attribute === 'datatype'
793 || $attribute === 'typeof'
794 # HTML5 microdata
795 || $attribute === 'itemid' || $attribute === 'itemprop'
796 || $attribute === 'itemref' || $attribute === 'itemscope'
797 || $attribute === 'itemtype'
799 // Paranoia. Allow "simple" values but suppress javascript
800 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
801 continue;
805 # NOTE: even though elements using href/src are not allowed directly, supply
806 # validation code that can be used by tag hook handlers, etc
807 if ( $attribute === 'href' || $attribute === 'src' ) {
808 if ( !preg_match( $hrefExp, $value ) ) {
809 continue; // drop any href or src attributes not using an allowed protocol.
810 // NOTE: this also drops all relative URLs
814 // If this attribute was previously set, override it.
815 // Output should only have one attribute of each name.
816 $out[$attribute] = $value;
819 if ( $wgAllowMicrodataAttributes ) {
820 # itemtype, itemid, itemref don't make sense without itemscope
821 if ( !array_key_exists( 'itemscope', $out ) ) {
822 unset( $out['itemtype'] );
823 unset( $out['itemid'] );
824 unset( $out['itemref'] );
826 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
828 return $out;
832 * Merge two sets of HTML attributes. Conflicting items in the second set
833 * will override those in the first, except for 'class' attributes which
834 * will be combined (if they're both strings).
836 * @todo implement merging for other attributes such as style
837 * @param array $a
838 * @param array $b
839 * @return array
841 static function mergeAttributes( $a, $b ) {
842 $out = array_merge( $a, $b );
843 if ( isset( $a['class'] ) && isset( $b['class'] )
844 && is_string( $a['class'] ) && is_string( $b['class'] )
845 && $a['class'] !== $b['class']
847 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
848 -1, PREG_SPLIT_NO_EMPTY );
849 $out['class'] = implode( ' ', array_unique( $classes ) );
851 return $out;
855 * Normalize CSS into a format we can easily search for hostile input
856 * - decode character references
857 * - decode escape sequences
858 * - convert characters that IE6 interprets into ascii
859 * - remove comments, unless the entire value is one single comment
860 * @param string $value the css string
861 * @return string normalized css
863 public static function normalizeCss( $value ) {
865 // Decode character references like &#123;
866 $value = Sanitizer::decodeCharReferences( $value );
868 // Decode escape sequences and line continuation
869 // See the grammar in the CSS 2 spec, appendix D.
870 // This has to be done AFTER decoding character references.
871 // This means it isn't possible for this function to return
872 // unsanitized escape sequences. It is possible to manufacture
873 // input that contains character references that decode to
874 // escape sequences that decode to character references, but
875 // it's OK for the return value to contain character references
876 // because the caller is supposed to escape those anyway.
877 static $decodeRegex;
878 if ( !$decodeRegex ) {
879 $space = '[\\x20\\t\\r\\n\\f]';
880 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
881 $backslash = '\\\\';
882 $decodeRegex = "/ $backslash
884 ($nl) | # 1. Line continuation
885 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
886 (.) | # 3. backslash cancelling special meaning
887 () | # 4. backslash at end of string
888 )/xu";
890 $value = preg_replace_callback( $decodeRegex,
891 array( __CLASS__, 'cssDecodeCallback' ), $value );
893 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
894 $value = preg_replace_callback(
895 '/[!-[]-z]/u', // U+FF01 to U+FF5A, excluding U+FF3C (bug 58088)
896 function ( $matches ) {
897 $cp = UtfNormal\Utils::utf8ToCodepoint( $matches[0] );
898 if ( $cp === false ) {
899 return '';
901 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
903 $value
906 // Convert more characters IE6 might treat as ascii
907 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
908 $value = str_replace(
909 array( 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ),
910 array( 'r', 'n', 'n', 'l', 'i', '(', '(' ),
911 $value
914 // Let the value through if it's nothing but a single comment, to
915 // allow other functions which may reject it to pass some error
916 // message through.
917 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
918 // Remove any comments; IE gets token splitting wrong
919 // This must be done AFTER decoding character references and
920 // escape sequences, because those steps can introduce comments
921 // This step cannot introduce character references or escape
922 // sequences, because it replaces comments with spaces rather
923 // than removing them completely.
924 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
926 // Remove anything after a comment-start token, to guard against
927 // incorrect client implementations.
928 $commentPos = strpos( $value, '/*' );
929 if ( $commentPos !== false ) {
930 $value = substr( $value, 0, $commentPos );
934 // S followed by repeat, iteration, or prolonged sound marks,
935 // which IE will treat as "ss"
936 $value = preg_replace(
937 '/s(?:
938 \xE3\x80\xB1 | # U+3031
939 \xE3\x82\x9D | # U+309D
940 \xE3\x83\xBC | # U+30FC
941 \xE3\x83\xBD | # U+30FD
942 \xEF\xB9\xBC | # U+FE7C
943 \xEF\xB9\xBD | # U+FE7D
944 \xEF\xBD\xB0 # U+FF70
945 )/ix',
946 'ss',
947 $value
950 return $value;
954 * Pick apart some CSS and check it for forbidden or unsafe structures.
955 * Returns a sanitized string. This sanitized string will have
956 * character references and escape sequences decoded and comments
957 * stripped (unless it is itself one valid comment, in which case the value
958 * will be passed through). If the input is just too evil, only a comment
959 * complaining about evilness will be returned.
961 * Currently URL references, 'expression', 'tps' are forbidden.
963 * NOTE: Despite the fact that character references are decoded, the
964 * returned string may contain character references given certain
965 * clever input strings. These character references must
966 * be escaped before the return value is embedded in HTML.
968 * @param string $value
969 * @return string
971 static function checkCss( $value ) {
972 $value = self::normalizeCss( $value );
974 // Reject problematic keywords and control characters
975 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ||
976 strpos( $value, UtfNormal\Constants::UTF8_REPLACEMENT ) !== false ) {
977 return '/* invalid control char */';
978 } elseif ( preg_match(
979 '! expression
980 | filter\s*:
981 | accelerator\s*:
982 | -o-link\s*:
983 | -o-link-source\s*:
984 | -o-replace\s*:
985 | url\s*\(
986 | image\s*\(
987 | image-set\s*\(
988 !ix', $value ) ) {
989 return '/* insecure input */';
991 return $value;
995 * @param array $matches
996 * @return string
998 static function cssDecodeCallback( $matches ) {
999 if ( $matches[1] !== '' ) {
1000 // Line continuation
1001 return '';
1002 } elseif ( $matches[2] !== '' ) {
1003 $char = UtfNormal\Utils::codepointToUtf8( hexdec( $matches[2] ) );
1004 } elseif ( $matches[3] !== '' ) {
1005 $char = $matches[3];
1006 } else {
1007 $char = '\\';
1009 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
1010 // These characters need to be escaped in strings
1011 // Clean up the escape sequence to avoid parsing errors by clients
1012 return '\\' . dechex( ord( $char ) ) . ' ';
1013 } else {
1014 // Decode unnecessary escape
1015 return $char;
1020 * Take a tag soup fragment listing an HTML element's attributes
1021 * and normalize it to well-formed XML, discarding unwanted attributes.
1022 * Output is safe for further wikitext processing, with escaping of
1023 * values that could trigger problems.
1025 * - Normalizes attribute names to lowercase
1026 * - Discards attributes not on a whitelist for the given element
1027 * - Turns broken or invalid entities into plaintext
1028 * - Double-quotes all attribute values
1029 * - Attributes without values are given the name as attribute
1030 * - Double attributes are discarded
1031 * - Unsafe style attributes are discarded
1032 * - Prepends space if there are attributes.
1034 * @param string $text
1035 * @param string $element
1036 * @return string
1038 static function fixTagAttributes( $text, $element ) {
1039 if ( trim( $text ) == '' ) {
1040 return '';
1043 $decoded = Sanitizer::decodeTagAttributes( $text );
1044 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1046 return Sanitizer::safeEncodeTagAttributes( $stripped );
1050 * Encode an attribute value for HTML output.
1051 * @param string $text
1052 * @return string HTML-encoded text fragment
1054 static function encodeAttribute( $text ) {
1055 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1057 // Whitespace is normalized during attribute decoding,
1058 // so if we've been passed non-spaces we must encode them
1059 // ahead of time or they won't be preserved.
1060 $encValue = strtr( $encValue, array(
1061 "\n" => '&#10;',
1062 "\r" => '&#13;',
1063 "\t" => '&#9;',
1064 ) );
1066 return $encValue;
1070 * Encode an attribute value for HTML tags, with extra armoring
1071 * against further wiki processing.
1072 * @param string $text
1073 * @return string HTML-encoded text fragment
1075 static function safeEncodeAttribute( $text ) {
1076 $encValue = Sanitizer::encodeAttribute( $text );
1078 # Templates and links may be expanded in later parsing,
1079 # creating invalid or dangerous output. Suppress this.
1080 $encValue = strtr( $encValue, array(
1081 '<' => '&lt;', // This should never happen,
1082 '>' => '&gt;', // we've received invalid input
1083 '"' => '&quot;', // which should have been escaped.
1084 '{' => '&#123;',
1085 '[' => '&#91;',
1086 "''" => '&#39;&#39;',
1087 'ISBN' => '&#73;SBN',
1088 'RFC' => '&#82;FC',
1089 'PMID' => '&#80;MID',
1090 '|' => '&#124;',
1091 '__' => '&#95;_',
1092 ) );
1094 # Stupid hack
1095 $encValue = preg_replace_callback(
1096 '/((?i)' . wfUrlProtocols() . ')/',
1097 array( 'Sanitizer', 'armorLinksCallback' ),
1098 $encValue );
1099 return $encValue;
1103 * Given a value, escape it so that it can be used in an id attribute and
1104 * return it. This will use HTML5 validation if $wgExperimentalHtmlIds is
1105 * true, allowing anything but ASCII whitespace. Otherwise it will use
1106 * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1107 * escaped with lots of dots.
1109 * To ensure we don't have to bother escaping anything, we also strip ', ",
1110 * & even if $wgExperimentalIds is true. TODO: Is this the best tactic?
1111 * We also strip # because it upsets IE, and % because it could be
1112 * ambiguous if it's part of something that looks like a percent escape
1113 * (which don't work reliably in fragments cross-browser).
1115 * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1116 * in the id and name attributes
1117 * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1118 * the id attribute
1119 * @see http://www.whatwg.org/html/elements.html#the-id-attribute
1120 * HTML5 definition of id attribute
1122 * @param string $id Id to escape
1123 * @param string|array $options String or array of strings (default is array()):
1124 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1125 * so don't pay attention if the first character isn't valid at the
1126 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
1127 * false.
1128 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
1129 * if $wgExperimentalHtmlIds is used, so we can generate extra
1130 * anchors and links won't break.
1131 * @return string
1133 static function escapeId( $id, $options = array() ) {
1134 global $wgExperimentalHtmlIds;
1135 $options = (array)$options;
1137 $id = Sanitizer::decodeCharReferences( $id );
1139 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1140 $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1141 $id = trim( $id, '_' );
1142 if ( $id === '' ) {
1143 // Must have been all whitespace to start with.
1144 return '_';
1145 } else {
1146 return $id;
1150 // HTML4-style escaping
1151 static $replace = array(
1152 '%3A' => ':',
1153 '%' => '.'
1156 $id = urlencode( strtr( $id, ' ', '_' ) );
1157 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1159 if ( !preg_match( '/^[a-zA-Z]/', $id ) && !in_array( 'noninitial', $options ) ) {
1160 // Initial character must be a letter!
1161 $id = "x$id";
1163 return $id;
1167 * Given a string containing a space delimited list of ids, escape each id
1168 * to match ids escaped by the escapeId() function.
1170 * @since 1.27
1172 * @param string $referenceString Space delimited list of ids
1173 * @param string|array $options String or array of strings (default is array()):
1174 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1175 * so don't pay attention if the first character isn't valid at the
1176 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
1177 * false.
1178 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
1179 * if $wgExperimentalHtmlIds is used, so we can generate extra
1180 * anchors and links won't break.
1181 * @return string
1183 static function escapeIdReferenceList( $referenceString, $options = array() ) {
1184 # Explode the space delimited list string into an array of tokens
1185 $references = preg_split( '/\s+/', "{$referenceString}", -1, PREG_SPLIT_NO_EMPTY );
1187 # Escape each token as an id
1188 foreach ( $references as &$ref ) {
1189 $ref = Sanitizer::escapeId( $ref, $options );
1192 # Merge the array back to a space delimited list string
1193 # If the array is empty, the result will be an empty string ('')
1194 $referenceString = implode( ' ', $references );
1196 return $referenceString;
1200 * Given a value, escape it so that it can be used as a CSS class and
1201 * return it.
1203 * @todo For extra validity, input should be validated UTF-8.
1205 * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1207 * @param string $class
1208 * @return string
1210 static function escapeClass( $class ) {
1211 // Convert ugly stuff to underscores and kill underscores in ugly places
1212 return rtrim( preg_replace(
1213 array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1214 '_',
1215 $class ), '_' );
1219 * Given HTML input, escape with htmlspecialchars but un-escape entities.
1220 * This allows (generally harmless) entities like &#160; to survive.
1222 * @param string $html HTML to escape
1223 * @return string Escaped input
1225 static function escapeHtmlAllowEntities( $html ) {
1226 $html = Sanitizer::decodeCharReferences( $html );
1227 # It seems wise to escape ' as well as ", as a matter of course. Can't
1228 # hurt.
1229 $html = htmlspecialchars( $html, ENT_QUOTES );
1230 return $html;
1234 * Regex replace callback for armoring links against further processing.
1235 * @param array $matches
1236 * @return string
1238 private static function armorLinksCallback( $matches ) {
1239 return str_replace( ':', '&#58;', $matches[1] );
1243 * Return an associative array of attribute names and values from
1244 * a partial tag string. Attribute names are forced to lowercase,
1245 * character references are decoded to UTF-8 text.
1247 * @param string $text
1248 * @return array
1250 public static function decodeTagAttributes( $text ) {
1251 if ( trim( $text ) == '' ) {
1252 return array();
1255 $attribs = array();
1256 $pairs = array();
1257 if ( !preg_match_all(
1258 self::getAttribsRegex(),
1259 $text,
1260 $pairs,
1261 PREG_SET_ORDER ) ) {
1262 return $attribs;
1265 foreach ( $pairs as $set ) {
1266 $attribute = strtolower( $set[1] );
1267 $value = Sanitizer::getTagAttributeCallback( $set );
1269 // Normalize whitespace
1270 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1271 $value = trim( $value );
1273 // Decode character references
1274 $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1276 return $attribs;
1280 * Build a partial tag string from an associative array of attribute
1281 * names and values as returned by decodeTagAttributes.
1283 * @param array $assoc_array
1284 * @return string
1286 public static function safeEncodeTagAttributes( $assoc_array ) {
1287 $attribs = array();
1288 foreach ( $assoc_array as $attribute => $value ) {
1289 $encAttribute = htmlspecialchars( $attribute );
1290 $encValue = Sanitizer::safeEncodeAttribute( $value );
1292 $attribs[] = "$encAttribute=\"$encValue\"";
1294 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1298 * Pick the appropriate attribute value from a match set from the
1299 * attribs regex matches.
1301 * @param array $set
1302 * @throws MWException When tag conditions are not met.
1303 * @return string
1305 private static function getTagAttributeCallback( $set ) {
1306 if ( isset( $set[5] ) ) {
1307 # No quotes.
1308 return $set[5];
1309 } elseif ( isset( $set[4] ) ) {
1310 # Single-quoted
1311 return $set[4];
1312 } elseif ( isset( $set[3] ) ) {
1313 # Double-quoted
1314 return $set[3];
1315 } elseif ( !isset( $set[2] ) ) {
1316 # In XHTML, attributes must have a value so return an empty string.
1317 # See "Empty attribute syntax",
1318 # http://www.w3.org/TR/html5/syntax.html#syntax-attribute-name
1319 return "";
1320 } else {
1321 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1326 * @param string $text
1327 * @return string
1329 private static function normalizeWhitespace( $text ) {
1330 return preg_replace(
1331 '/\r\n|[\x20\x0d\x0a\x09]/',
1332 ' ',
1333 $text );
1337 * Normalizes whitespace in a section name, such as might be returned
1338 * by Parser::stripSectionName(), for use in the id's that are used for
1339 * section links.
1341 * @param string $section
1342 * @return string
1344 static function normalizeSectionNameWhitespace( $section ) {
1345 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1349 * Ensure that any entities and character references are legal
1350 * for XML and XHTML specifically. Any stray bits will be
1351 * &amp;-escaped to result in a valid text fragment.
1353 * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1354 * numericized (this way we're well-formed even without a DTD)
1355 * b. any numeric char refs must be legal chars, not invalid or forbidden
1356 * c. use lower cased "&#x", not "&#X"
1357 * d. fix or reject non-valid attributes
1359 * @param string $text
1360 * @return string
1361 * @private
1363 static function normalizeCharReferences( $text ) {
1364 return preg_replace_callback(
1365 self::CHAR_REFS_REGEX,
1366 array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1367 $text );
1371 * @param string $matches
1372 * @return string
1374 static function normalizeCharReferencesCallback( $matches ) {
1375 $ret = null;
1376 if ( $matches[1] != '' ) {
1377 $ret = Sanitizer::normalizeEntity( $matches[1] );
1378 } elseif ( $matches[2] != '' ) {
1379 $ret = Sanitizer::decCharReference( $matches[2] );
1380 } elseif ( $matches[3] != '' ) {
1381 $ret = Sanitizer::hexCharReference( $matches[3] );
1383 if ( is_null( $ret ) ) {
1384 return htmlspecialchars( $matches[0] );
1385 } else {
1386 return $ret;
1391 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1392 * return the equivalent numeric entity reference (except for the core &lt;
1393 * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1394 * the HTML equivalent. Otherwise, returns HTML-escaped text of
1395 * pseudo-entity source (eg &amp;foo;)
1397 * @param string $name
1398 * @return string
1400 static function normalizeEntity( $name ) {
1401 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1402 return '&' . self::$htmlEntityAliases[$name] . ';';
1403 } elseif ( in_array( $name, array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1404 return "&$name;";
1405 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1406 return '&#' . self::$htmlEntities[$name] . ';';
1407 } else {
1408 return "&amp;$name;";
1413 * @param int $codepoint
1414 * @return null|string
1416 static function decCharReference( $codepoint ) {
1417 $point = intval( $codepoint );
1418 if ( Sanitizer::validateCodepoint( $point ) ) {
1419 return sprintf( '&#%d;', $point );
1420 } else {
1421 return null;
1426 * @param int $codepoint
1427 * @return null|string
1429 static function hexCharReference( $codepoint ) {
1430 $point = hexdec( $codepoint );
1431 if ( Sanitizer::validateCodepoint( $point ) ) {
1432 return sprintf( '&#x%x;', $point );
1433 } else {
1434 return null;
1439 * Returns true if a given Unicode codepoint is a valid character in
1440 * both HTML5 and XML.
1441 * @param int $codepoint
1442 * @return bool
1444 private static function validateCodepoint( $codepoint ) {
1445 # U+000C is valid in HTML5 but not allowed in XML.
1446 # U+000D is valid in XML but not allowed in HTML5.
1447 # U+007F - U+009F are disallowed in HTML5 (control characters).
1448 return $codepoint == 0x09
1449 || $codepoint == 0x0a
1450 || ( $codepoint >= 0x20 && $codepoint <= 0x7e )
1451 || ( $codepoint >= 0xa0 && $codepoint <= 0xd7ff )
1452 || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1453 || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1457 * Decode any character references, numeric or named entities,
1458 * in the text and return a UTF-8 string.
1460 * @param string $text
1461 * @return string
1463 public static function decodeCharReferences( $text ) {
1464 return preg_replace_callback(
1465 self::CHAR_REFS_REGEX,
1466 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1467 $text );
1471 * Decode any character references, numeric or named entities,
1472 * in the next and normalize the resulting string. (bug 14952)
1474 * This is useful for page titles, not for text to be displayed,
1475 * MediaWiki allows HTML entities to escape normalization as a feature.
1477 * @param string $text Already normalized, containing entities
1478 * @return string Still normalized, without entities
1480 public static function decodeCharReferencesAndNormalize( $text ) {
1481 global $wgContLang;
1482 $text = preg_replace_callback(
1483 self::CHAR_REFS_REGEX,
1484 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1485 $text, /* limit */ -1, $count );
1487 if ( $count ) {
1488 return $wgContLang->normalize( $text );
1489 } else {
1490 return $text;
1495 * @param string $matches
1496 * @return string
1498 static function decodeCharReferencesCallback( $matches ) {
1499 if ( $matches[1] != '' ) {
1500 return Sanitizer::decodeEntity( $matches[1] );
1501 } elseif ( $matches[2] != '' ) {
1502 return Sanitizer::decodeChar( intval( $matches[2] ) );
1503 } elseif ( $matches[3] != '' ) {
1504 return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1506 # Last case should be an ampersand by itself
1507 return $matches[0];
1511 * Return UTF-8 string for a codepoint if that is a valid
1512 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1513 * @param int $codepoint
1514 * @return string
1515 * @private
1517 static function decodeChar( $codepoint ) {
1518 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1519 return UtfNormal\Utils::codepointToUtf8( $codepoint );
1520 } else {
1521 return UtfNormal\Constants::UTF8_REPLACEMENT;
1526 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1527 * return the UTF-8 encoding of that character. Otherwise, returns
1528 * pseudo-entity source (eg "&foo;")
1530 * @param string $name
1531 * @return string
1533 static function decodeEntity( $name ) {
1534 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1535 $name = self::$htmlEntityAliases[$name];
1537 if ( isset( self::$htmlEntities[$name] ) ) {
1538 return UtfNormal\Utils::codepointToUtf8( self::$htmlEntities[$name] );
1539 } else {
1540 return "&$name;";
1545 * Fetch the whitelist of acceptable attributes for a given element name.
1547 * @param string $element
1548 * @return array
1550 static function attributeWhitelist( $element ) {
1551 $list = Sanitizer::setupAttributeWhitelist();
1552 return isset( $list[$element] )
1553 ? $list[$element]
1554 : array();
1558 * Foreach array key (an allowed HTML element), return an array
1559 * of allowed attributes
1560 * @return array
1562 static function setupAttributeWhitelist() {
1563 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
1564 static $whitelist, $staticInitialised;
1566 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgAllowMicrodataAttributes' ) );
1568 if ( $whitelist !== null && $staticInitialised == $globalContext ) {
1569 return $whitelist;
1572 $common = array(
1573 # HTML
1574 'id',
1575 'class',
1576 'style',
1577 'lang',
1578 'dir',
1579 'title',
1581 # WAI-ARIA
1582 'aria-describedby',
1583 'aria-flowto',
1584 'aria-label',
1585 'aria-labelledby',
1586 'aria-owns',
1587 'role',
1590 if ( $wgAllowRdfaAttributes ) {
1591 # RDFa attributes as specified in section 9 of
1592 # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1593 $common = array_merge( $common, array(
1594 'about', 'property', 'resource', 'datatype', 'typeof',
1595 ) );
1598 if ( $wgAllowMicrodataAttributes ) {
1599 # add HTML5 microdata tags as specified by
1600 # http://www.whatwg.org/html/microdata.html#the-microdata-model
1601 $common = array_merge( $common, array(
1602 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1603 ) );
1606 $block = array_merge( $common, array( 'align' ) );
1607 $tablealign = array( 'align', 'valign' );
1608 $tablecell = array(
1609 'abbr',
1610 'axis',
1611 'headers',
1612 'scope',
1613 'rowspan',
1614 'colspan',
1615 'nowrap', # deprecated
1616 'width', # deprecated
1617 'height', # deprecated
1618 'bgcolor', # deprecated
1621 # Numbers refer to sections in HTML 4.01 standard describing the element.
1622 # See: http://www.w3.org/TR/html4/
1623 $whitelist = array(
1624 # 7.5.4
1625 'div' => $block,
1626 'center' => $common, # deprecated
1627 'span' => $common,
1629 # 7.5.5
1630 'h1' => $block,
1631 'h2' => $block,
1632 'h3' => $block,
1633 'h4' => $block,
1634 'h5' => $block,
1635 'h6' => $block,
1637 # 7.5.6
1638 # address
1640 # 8.2.4
1641 'bdo' => $common,
1643 # 9.2.1
1644 'em' => $common,
1645 'strong' => $common,
1646 'cite' => $common,
1647 'dfn' => $common,
1648 'code' => $common,
1649 'samp' => $common,
1650 'kbd' => $common,
1651 'var' => $common,
1652 'abbr' => $common,
1653 # acronym
1655 # 9.2.2
1656 'blockquote' => array_merge( $common, array( 'cite' ) ),
1657 'q' => array_merge( $common, array( 'cite' ) ),
1659 # 9.2.3
1660 'sub' => $common,
1661 'sup' => $common,
1663 # 9.3.1
1664 'p' => $block,
1666 # 9.3.2
1667 'br' => array_merge( $common, array( 'clear' ) ),
1669 # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1670 'wbr' => $common,
1672 # 9.3.4
1673 'pre' => array_merge( $common, array( 'width' ) ),
1675 # 9.4
1676 'ins' => array_merge( $common, array( 'cite', 'datetime' ) ),
1677 'del' => array_merge( $common, array( 'cite', 'datetime' ) ),
1679 # 10.2
1680 'ul' => array_merge( $common, array( 'type' ) ),
1681 'ol' => array_merge( $common, array( 'type', 'start', 'reversed' ) ),
1682 'li' => array_merge( $common, array( 'type', 'value' ) ),
1684 # 10.3
1685 'dl' => $common,
1686 'dd' => $common,
1687 'dt' => $common,
1689 # 11.2.1
1690 'table' => array_merge( $common,
1691 array( 'summary', 'width', 'border', 'frame',
1692 'rules', 'cellspacing', 'cellpadding',
1693 'align', 'bgcolor',
1694 ) ),
1696 # 11.2.2
1697 'caption' => $block,
1699 # 11.2.3
1700 'thead' => $common,
1701 'tfoot' => $common,
1702 'tbody' => $common,
1704 # 11.2.4
1705 'colgroup' => array_merge( $common, array( 'span' ) ),
1706 'col' => array_merge( $common, array( 'span' ) ),
1708 # 11.2.5
1709 'tr' => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1711 # 11.2.6
1712 'td' => array_merge( $common, $tablecell, $tablealign ),
1713 'th' => array_merge( $common, $tablecell, $tablealign ),
1715 # 12.2
1716 # NOTE: <a> is not allowed directly, but the attrib
1717 # whitelist is used from the Parser object
1718 'a' => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1720 # 13.2
1721 # Not usually allowed, but may be used for extension-style hooks
1722 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1723 # true
1724 'img' => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1726 # 15.2.1
1727 'tt' => $common,
1728 'b' => $common,
1729 'i' => $common,
1730 'big' => $common,
1731 'small' => $common,
1732 'strike' => $common,
1733 's' => $common,
1734 'u' => $common,
1736 # 15.2.2
1737 'font' => array_merge( $common, array( 'size', 'color', 'face' ) ),
1738 # basefont
1740 # 15.3
1741 'hr' => array_merge( $common, array( 'width' ) ),
1743 # HTML Ruby annotation text module, simple ruby only.
1744 # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
1745 'ruby' => $common,
1746 # rbc
1747 'rb' => $common,
1748 'rp' => $common,
1749 'rt' => $common, # array_merge( $common, array( 'rbspan' ) ),
1750 'rtc' => $common,
1752 # MathML root element, where used for extensions
1753 # 'title' may not be 100% valid here; it's XHTML
1754 # http://www.w3.org/TR/REC-MathML/
1755 'math' => array( 'class', 'style', 'id', 'title' ),
1757 # HTML 5 section 4.6
1758 'bdi' => $common,
1760 # HTML5 elements, defined by:
1761 # http://www.whatwg.org/html/
1762 'data' => array_merge( $common, array( 'value' ) ),
1763 'time' => array_merge( $common, array( 'datetime' ) ),
1764 'mark' => $common,
1766 // meta and link are only permitted by removeHTMLtags when Microdata
1767 // is enabled so we don't bother adding a conditional to hide these
1768 // Also meta and link are only valid in WikiText as Microdata elements
1769 // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1770 // So we don't bother including $common attributes that have no purpose.
1771 'meta' => array( 'itemprop', 'content' ),
1772 'link' => array( 'itemprop', 'href' ),
1775 $staticInitialised = $globalContext;
1777 return $whitelist;
1781 * Take a fragment of (potentially invalid) HTML and return
1782 * a version with any tags removed, encoded as plain text.
1784 * Warning: this return value must be further escaped for literal
1785 * inclusion in HTML output as of 1.10!
1787 * @param string $text HTML fragment
1788 * @return string
1790 static function stripAllTags( $text ) {
1791 # Actual <tags>
1792 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1794 # Normalize &entities and whitespace
1795 $text = self::decodeCharReferences( $text );
1796 $text = self::normalizeWhitespace( $text );
1798 return $text;
1802 * Hack up a private DOCTYPE with HTML's standard entity declarations.
1803 * PHP 4 seemed to know these if you gave it an HTML doctype, but
1804 * PHP 5.1 doesn't.
1806 * Use for passing XHTML fragments to PHP's XML parsing functions
1808 * @return string
1810 static function hackDocType() {
1811 $out = "<!DOCTYPE html [\n";
1812 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1813 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1815 $out .= "]>\n";
1816 return $out;
1820 * @param string $url
1821 * @return mixed|string
1823 static function cleanUrl( $url ) {
1824 # Normalize any HTML entities in input. They will be
1825 # re-escaped by makeExternalLink().
1826 $url = Sanitizer::decodeCharReferences( $url );
1828 # Escape any control characters introduced by the above step
1829 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1830 array( __CLASS__, 'cleanUrlCallback' ), $url );
1832 # Validate hostname portion
1833 $matches = array();
1834 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1835 list( /* $whole */, $protocol, $host, $rest ) = $matches;
1837 // Characters that will be ignored in IDNs.
1838 // http://tools.ietf.org/html/3454#section-3.1
1839 // Strip them before further processing so blacklists and such work.
1840 $strip = "/
1841 \\s| # general whitespace
1842 \xc2\xad| # 00ad SOFT HYPHEN
1843 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1844 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1845 \xe2\x81\xa0| # 2060 WORD JOINER
1846 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1847 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
1848 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1849 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1850 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1851 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1852 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1853 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1854 /xuD";
1856 $host = preg_replace( $strip, '', $host );
1858 // IPv6 host names are bracketed with []. Url-decode these.
1859 if ( substr_compare( "//%5B", $host, 0, 5 ) === 0 &&
1860 preg_match( '!^//%5B([0-9A-Fa-f:.]+)%5D((:\d+)?)$!', $host, $matches )
1862 $host = '//[' . $matches[1] . ']' . $matches[2];
1865 // @todo FIXME: Validate hostnames here
1867 return $protocol . $host . $rest;
1868 } else {
1869 return $url;
1874 * @param array $matches
1875 * @return string
1877 static function cleanUrlCallback( $matches ) {
1878 return urlencode( $matches[0] );
1882 * Does a string look like an e-mail address?
1884 * This validates an email address using an HTML5 specification found at:
1885 * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1886 * Which as of 2011-01-24 says:
1888 * A valid e-mail address is a string that matches the ABNF production
1889 * 1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1890 * in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1891 * 3.5.
1893 * This function is an implementation of the specification as requested in
1894 * bug 22449.
1896 * Client-side forms will use the same standard validation rules via JS or
1897 * HTML 5 validation; additional restrictions can be enforced server-side
1898 * by extensions via the 'isValidEmailAddr' hook.
1900 * Note that this validation doesn't 100% match RFC 2822, but is believed
1901 * to be liberal enough for wide use. Some invalid addresses will still
1902 * pass validation here.
1904 * @since 1.18
1906 * @param string $addr E-mail address
1907 * @return bool
1909 public static function validateEmail( $addr ) {
1910 $result = null;
1911 if ( !Hooks::run( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1912 return $result;
1915 // Please note strings below are enclosed in brackets [], this make the
1916 // hyphen "-" a range indicator. Hence it is double backslashed below.
1917 // See bug 26948
1918 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1919 $rfc1034_ldh_str = "a-z0-9\\-";
1921 $html5_email_regexp = "/
1922 ^ # start of string
1923 [$rfc5322_atext\\.]+ # user part which is liberal :p
1924 @ # 'apostrophe'
1925 [$rfc1034_ldh_str]+ # First domain part
1926 (\\.[$rfc1034_ldh_str]+)* # Following part prefixed with a dot
1927 $ # End of string
1928 /ix"; // case Insensitive, eXtended
1930 return (bool)preg_match( $html5_email_regexp, $addr );