Merge "Make sure images that don't have an explicit alignment get aligned right"
[mediawiki.git] / includes / Sanitizer.php
blob1ea652320384c6aca74d26988e55e0e0d146e560
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 $attribsRegex;
328 * Regular expression to match HTML/XML attribute pairs within a tag.
329 * Allows some... latitude.
330 * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
332 static function getAttribsRegex() {
333 if ( self::$attribsRegex === null ) {
334 $attribFirst = '[:A-Z_a-z0-9]';
335 $attrib = '[:A-Z_a-z-.0-9]';
336 $space = '[\x09\x0a\x0d\x20]';
337 self::$attribsRegex =
338 "/(?:^|$space)({$attribFirst}{$attrib}*)
339 ($space*=$space*
341 # The attribute value: quoted or alone
342 \"([^<\"]*)\"
343 | '([^<']*)'
344 | ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
345 | (\#[0-9a-fA-F]+) # Technically wrong, but lots of
346 # colors are specified like this.
347 # We'll be normalizing it.
349 )?(?=$space|\$)/sx";
351 return self::$attribsRegex;
355 * Cleans up HTML, removes dangerous tags and attributes, and
356 * removes HTML comments
357 * @private
358 * @param string $text
359 * @param callable $processCallback Callback to do any variable or parameter
360 * replacements in HTML attribute values
361 * @param array|bool $args Arguments for the processing callback
362 * @param array $extratags For any extra tags to include
363 * @param array $removetags For any tags (default or extra) to exclude
364 * @return string
366 static function removeHTMLtags( $text, $processCallback = null,
367 $args = array(), $extratags = array(), $removetags = array()
369 global $wgUseTidy, $wgAllowMicrodataAttributes, $wgAllowImageTag;
371 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
372 $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
374 wfProfileIn( __METHOD__ );
376 // Base our staticInitialised variable off of the global config state so that if the globals
377 // are changed (like in the screwed up test system) we will re-initialise the settings.
378 $globalContext = implode( '-', compact( 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
379 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
381 $htmlpairsStatic = array( # Tags that must be closed
382 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
383 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
384 'strike', 'strong', 'tt', 'var', 'div', 'center',
385 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
386 'ruby', 'rt', 'rb', 'rp', 'p', 'span', 'abbr', 'dfn',
387 'kbd', 'samp', 'data', 'time', 'mark'
389 $htmlsingle = array(
390 'br', 'wbr', 'hr', 'li', 'dt', 'dd'
392 $htmlsingleonly = array( # Elements that cannot have close tags
393 'br', 'wbr', 'hr'
395 if ( $wgAllowMicrodataAttributes ) {
396 $htmlsingle[] = $htmlsingleonly[] = 'meta';
397 $htmlsingle[] = $htmlsingleonly[] = 'link';
399 $htmlnest = array( # Tags that can be nested--??
400 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
401 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
402 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
404 $tabletags = array( # Can only appear inside table, we will close them
405 'td', 'th', 'tr',
407 $htmllist = array( # Tags used by list
408 'ul', 'ol',
410 $listtags = array( # Tags that can appear in a list
411 'li',
414 if ( $wgAllowImageTag ) {
415 $htmlsingle[] = 'img';
416 $htmlsingleonly[] = 'img';
419 $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
420 $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
422 # Convert them all to hashtables for faster lookup
423 $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
424 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
425 foreach ( $vars as $var ) {
426 $$var = array_flip( $$var );
428 $staticInitialised = $globalContext;
430 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
431 $extratags = array_flip( $extratags );
432 $removetags = array_flip( $removetags );
433 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
434 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
436 # Remove HTML comments
437 $text = Sanitizer::removeHTMLcomments( $text );
438 $bits = explode( '<', $text );
439 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
440 if ( !$wgUseTidy ) {
441 $tagstack = $tablestack = array();
442 foreach ( $bits as $x ) {
443 $regs = array();
444 # $slash: Does the current element start with a '/'?
445 # $t: Current element name
446 # $params: String between element name and >
447 # $brace: Ending '>' or '/>'
448 # $rest: Everything until the next element of $bits
449 if ( preg_match( '!^(/?)([^\\s/>]+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
450 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
451 } else {
452 $slash = $t = $params = $brace = $rest = null;
455 $badtag = false;
456 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
457 # Check our stack
458 if ( $slash && isset( $htmlsingleonly[$t] ) ) {
459 $badtag = true;
460 } elseif ( $slash ) {
461 # Closing a tag... is it the one we just opened?
462 wfSuppressWarnings();
463 $ot = array_pop( $tagstack );
464 wfRestoreWarnings();
466 if ( $ot != $t ) {
467 if ( isset( $htmlsingleallowed[$ot] ) ) {
468 # Pop all elements with an optional close tag
469 # and see if we find a match below them
470 $optstack = array();
471 array_push( $optstack, $ot );
472 wfSuppressWarnings();
473 $ot = array_pop( $tagstack );
474 wfRestoreWarnings();
475 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
476 array_push( $optstack, $ot );
477 wfSuppressWarnings();
478 $ot = array_pop( $tagstack );
479 wfRestoreWarnings();
481 if ( $t != $ot ) {
482 # No match. Push the optional elements back again
483 $badtag = true;
484 wfSuppressWarnings();
485 $ot = array_pop( $optstack );
486 wfRestoreWarnings();
487 while ( $ot ) {
488 array_push( $tagstack, $ot );
489 wfSuppressWarnings();
490 $ot = array_pop( $optstack );
491 wfRestoreWarnings();
494 } else {
495 wfSuppressWarnings();
496 array_push( $tagstack, $ot );
497 wfRestoreWarnings();
499 # <li> can be nested in <ul> or <ol>, skip those cases:
500 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
501 $badtag = true;
504 } else {
505 if ( $t == 'table' ) {
506 $tagstack = array_pop( $tablestack );
509 $newparams = '';
510 } else {
511 # Keep track for later
512 if ( isset( $tabletags[$t] ) &&
513 !in_array( 'table', $tagstack ) ) {
514 $badtag = true;
515 } elseif ( in_array( $t, $tagstack ) &&
516 !isset( $htmlnest[$t] ) ) {
517 $badtag = true;
518 # Is it a self closed htmlpair ? (bug 5487)
519 } elseif ( $brace == '/>' &&
520 isset( $htmlpairs[$t] ) ) {
521 $badtag = true;
522 } elseif ( isset( $htmlsingleonly[$t] ) ) {
523 # Hack to force empty tag for unclosable elements
524 $brace = '/>';
525 } elseif ( isset( $htmlsingle[$t] ) ) {
526 # Hack to not close $htmlsingle tags
527 $brace = null;
528 # Still need to push this optionally-closed tag to
529 # the tag stack so that we can match end tags
530 # instead of marking them as bad.
531 array_push( $tagstack, $t );
532 } elseif ( isset( $tabletags[$t] )
533 && in_array( $t, $tagstack ) ) {
534 // New table tag but forgot to close the previous one
535 $text .= "</$t>";
536 } else {
537 if ( $t == 'table' ) {
538 array_push( $tablestack, $tagstack );
539 $tagstack = array();
541 array_push( $tagstack, $t );
544 # Replace any variables or template parameters with
545 # plaintext results.
546 if ( is_callable( $processCallback ) ) {
547 call_user_func_array( $processCallback, array( &$params, $args ) );
550 if ( !Sanitizer::validateTag( $params, $t ) ) {
551 $badtag = true;
554 # Strip non-approved attributes from the tag
555 $newparams = Sanitizer::fixTagAttributes( $params, $t );
557 if ( !$badtag ) {
558 $rest = str_replace( '>', '&gt;', $rest );
559 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
560 $text .= "<$slash$t$newparams$close>$rest";
561 continue;
564 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
566 # Close off any remaining tags
567 while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
568 $text .= "</$t>\n";
569 if ( $t == 'table' ) {
570 $tagstack = array_pop( $tablestack );
573 } else {
574 # this might be possible using tidy itself
575 foreach ( $bits as $x ) {
576 preg_match(
577 '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
579 $regs
582 wfSuppressWarnings();
583 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
584 wfRestoreWarnings();
586 $badtag = false;
587 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
588 if ( is_callable( $processCallback ) ) {
589 call_user_func_array( $processCallback, array( &$params, $args ) );
592 if ( !Sanitizer::validateTag( $params, $t ) ) {
593 $badtag = true;
596 $newparams = Sanitizer::fixTagAttributes( $params, $t );
597 if ( !$badtag ) {
598 $rest = str_replace( '>', '&gt;', $rest );
599 $text .= "<$slash$t$newparams$brace$rest";
600 continue;
603 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
606 wfProfileOut( __METHOD__ );
607 return $text;
611 * Remove '<!--', '-->', and everything between.
612 * To avoid leaving blank lines, when a comment is both preceded
613 * and followed by a newline (ignoring spaces), trim leading and
614 * trailing spaces and one of the newlines.
616 * @private
617 * @param string $text
618 * @return string
620 static function removeHTMLcomments( $text ) {
621 wfProfileIn( __METHOD__ );
622 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
623 $end = strpos( $text, '-->', $start + 4 );
624 if ( $end === false ) {
625 # Unterminated comment; bail out
626 break;
629 $end += 3;
631 # Trim space and newline if the comment is both
632 # preceded and followed by a newline
633 $spaceStart = max( $start - 1, 0 );
634 $spaceLen = $end - $spaceStart;
635 while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
636 $spaceStart--;
637 $spaceLen++;
639 while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
640 $spaceLen++;
642 if ( substr( $text, $spaceStart, 1 ) === "\n"
643 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
644 # Remove the comment, leading and trailing
645 # spaces, and leave only one newline.
646 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
647 } else {
648 # Remove just the comment.
649 $text = substr_replace( $text, '', $start, $end - $start );
652 wfProfileOut( __METHOD__ );
653 return $text;
657 * Takes attribute names and values for a tag and the tag name and
658 * validates that the tag is allowed to be present.
659 * This DOES NOT validate the attributes, nor does it validate the
660 * tags themselves. This method only handles the special circumstances
661 * where we may want to allow a tag within content but ONLY when it has
662 * specific attributes set.
664 * @param string $params
665 * @param string $element
666 * @return bool
668 static function validateTag( $params, $element ) {
669 $params = Sanitizer::decodeTagAttributes( $params );
671 if ( $element == 'meta' || $element == 'link' ) {
672 if ( !isset( $params['itemprop'] ) ) {
673 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
674 return false;
676 if ( $element == 'meta' && !isset( $params['content'] ) ) {
677 // <meta> must have a content="" for the itemprop
678 return false;
680 if ( $element == 'link' && !isset( $params['href'] ) ) {
681 // <link> must have an associated href=""
682 return false;
686 return true;
690 * Take an array of attribute names and values and normalize or discard
691 * illegal values for the given element type.
693 * - Discards attributes not on a whitelist for the given element
694 * - Unsafe style attributes are discarded
695 * - Invalid id attributes are re-encoded
697 * @param array $attribs
698 * @param string $element
699 * @return array
701 * @todo Check for legal values where the DTD limits things.
702 * @todo Check for unique id attribute :P
704 static function validateTagAttributes( $attribs, $element ) {
705 return Sanitizer::validateAttributes( $attribs,
706 Sanitizer::attributeWhitelist( $element ) );
710 * Take an array of attribute names and values and normalize or discard
711 * illegal values for the given whitelist.
713 * - Discards attributes not the given whitelist
714 * - Unsafe style attributes are discarded
715 * - Invalid id attributes are re-encoded
717 * @param array $attribs
718 * @param array $whitelist list of allowed attribute names
719 * @return array
721 * @todo Check for legal values where the DTD limits things.
722 * @todo Check for unique id attribute :P
724 static function validateAttributes( $attribs, $whitelist ) {
725 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
727 $whitelist = array_flip( $whitelist );
728 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
730 $out = array();
731 foreach ( $attribs as $attribute => $value ) {
732 #allow XML namespace declaration if RDFa is enabled
733 if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
734 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
735 $out[$attribute] = $value;
738 continue;
741 # Allow any attribute beginning with "data-"
742 if ( !preg_match( '/^data-/i', $attribute ) && !isset( $whitelist[$attribute] ) ) {
743 continue;
746 # Strip javascript "expression" from stylesheets.
747 # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
748 if ( $attribute == 'style' ) {
749 $value = Sanitizer::checkCss( $value );
752 if ( $attribute === 'id' ) {
753 $value = Sanitizer::escapeId( $value, 'noninitial' );
756 # WAI-ARIA
757 # http://www.w3.org/TR/wai-aria/
758 # http://www.whatwg.org/html/elements.html#wai-aria
759 # For now we only support role="presentation" until we work out what roles should be
760 # usable by content and we ensure that our code explicitly rejects patterns that
761 # violate HTML5's ARIA restrictions.
762 if ( $attribute === 'role' && $value !== 'presentation' ) {
763 continue;
766 // RDFa and microdata properties allow URLs, URIs and/or CURIs.
767 // Check them for sanity.
768 if ( $attribute === 'rel' || $attribute === 'rev'
769 # RDFa
770 || $attribute === 'about' || $attribute === 'property'
771 || $attribute === 'resource' || $attribute === 'datatype'
772 || $attribute === 'typeof'
773 # HTML5 microdata
774 || $attribute === 'itemid' || $attribute === 'itemprop'
775 || $attribute === 'itemref' || $attribute === 'itemscope'
776 || $attribute === 'itemtype'
778 //Paranoia. Allow "simple" values but suppress javascript
779 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
780 continue;
784 # NOTE: even though elements using href/src are not allowed directly, supply
785 # validation code that can be used by tag hook handlers, etc
786 if ( $attribute === 'href' || $attribute === 'src' ) {
787 if ( !preg_match( $hrefExp, $value ) ) {
788 continue; //drop any href or src attributes not using an allowed protocol.
789 // NOTE: this also drops all relative URLs
793 // If this attribute was previously set, override it.
794 // Output should only have one attribute of each name.
795 $out[$attribute] = $value;
798 if ( $wgAllowMicrodataAttributes ) {
799 # itemtype, itemid, itemref don't make sense without itemscope
800 if ( !array_key_exists( 'itemscope', $out ) ) {
801 unset( $out['itemtype'] );
802 unset( $out['itemid'] );
803 unset( $out['itemref'] );
805 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
807 return $out;
811 * Merge two sets of HTML attributes. Conflicting items in the second set
812 * will override those in the first, except for 'class' attributes which
813 * will be combined (if they're both strings).
815 * @todo implement merging for other attributes such as style
816 * @param array $a
817 * @param array $b
818 * @return array
820 static function mergeAttributes( $a, $b ) {
821 $out = array_merge( $a, $b );
822 if ( isset( $a['class'] ) && isset( $b['class'] )
823 && is_string( $a['class'] ) && is_string( $b['class'] )
824 && $a['class'] !== $b['class']
826 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
827 -1, PREG_SPLIT_NO_EMPTY );
828 $out['class'] = implode( ' ', array_unique( $classes ) );
830 return $out;
834 * Pick apart some CSS and check it for forbidden or unsafe structures.
835 * Returns a sanitized string. This sanitized string will have
836 * character references and escape sequences decoded and comments
837 * stripped (unless it is itself one valid comment, in which case the value
838 * will be passed through). If the input is just too evil, only a comment
839 * complaining about evilness will be returned.
841 * Currently URL references, 'expression', 'tps' are forbidden.
843 * NOTE: Despite the fact that character references are decoded, the
844 * returned string may contain character references given certain
845 * clever input strings. These character references must
846 * be escaped before the return value is embedded in HTML.
848 * @param string $value
849 * @return string
851 static function checkCss( $value ) {
852 // Decode character references like &#123;
853 $value = Sanitizer::decodeCharReferences( $value );
855 // Decode escape sequences and line continuation
856 // See the grammar in the CSS 2 spec, appendix D.
857 // This has to be done AFTER decoding character references.
858 // This means it isn't possible for this function to return
859 // unsanitized escape sequences. It is possible to manufacture
860 // input that contains character references that decode to
861 // escape sequences that decode to character references, but
862 // it's OK for the return value to contain character references
863 // because the caller is supposed to escape those anyway.
864 static $decodeRegex;
865 if ( !$decodeRegex ) {
866 $space = '[\\x20\\t\\r\\n\\f]';
867 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
868 $backslash = '\\\\';
869 $decodeRegex = "/ $backslash
871 ($nl) | # 1. Line continuation
872 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
873 (.) | # 3. backslash cancelling special meaning
874 () | # 4. backslash at end of string
875 )/xu";
877 $value = preg_replace_callback( $decodeRegex,
878 array( __CLASS__, 'cssDecodeCallback' ), $value );
880 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
881 $value = preg_replace_callback(
882 '/[!-[]-z]/u', // U+FF01 to U+FF5A, excluding U+FF3C (bug 58088)
883 function ( $matches ) {
884 $cp = utf8ToCodepoint( $matches[0] );
885 if ( $cp === false ) {
886 return '';
888 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
890 $value
893 // Convert more characters IE6 might treat as ascii
894 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
895 $value = str_replace(
896 array( 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ),
897 array( 'r', 'n', 'n', 'l', 'i', '(', '(' ),
898 $value
901 // Let the value through if it's nothing but a single comment, to
902 // allow other functions which may reject it to pass some error
903 // message through.
904 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
905 // Remove any comments; IE gets token splitting wrong
906 // This must be done AFTER decoding character references and
907 // escape sequences, because those steps can introduce comments
908 // This step cannot introduce character references or escape
909 // sequences, because it replaces comments with spaces rather
910 // than removing them completely.
911 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
913 // Remove anything after a comment-start token, to guard against
914 // incorrect client implementations.
915 $commentPos = strpos( $value, '/*' );
916 if ( $commentPos !== false ) {
917 $value = substr( $value, 0, $commentPos );
921 // S followed by repeat, iteration, or prolonged sound marks,
922 // which IE will treat as "ss"
923 $value = preg_replace(
924 '/s(?:
925 \xE3\x80\xB1 | # U+3031
926 \xE3\x82\x9D | # U+309D
927 \xE3\x83\xBC | # U+30FC
928 \xE3\x83\xBD | # U+30FD
929 \xEF\xB9\xBC | # U+FE7C
930 \xEF\xB9\xBD | # U+FE7D
931 \xEF\xBD\xB0 # U+FF70
932 )/ix',
933 'ss',
934 $value
937 // Reject problematic keywords and control characters
938 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ) {
939 return '/* invalid control char */';
940 } elseif ( preg_match(
941 '! expression
942 | filter\s*:
943 | accelerator\s*:
944 | -o-link\s*:
945 | -o-link-source\s*:
946 | -o-replace\s*:
947 | url\s*\(
948 | image\s*\(
949 | image-set\s*\(
950 !ix', $value ) ) {
951 return '/* insecure input */';
953 return $value;
957 * @param array $matches
958 * @return string
960 static function cssDecodeCallback( $matches ) {
961 if ( $matches[1] !== '' ) {
962 // Line continuation
963 return '';
964 } elseif ( $matches[2] !== '' ) {
965 $char = codepointToUtf8( hexdec( $matches[2] ) );
966 } elseif ( $matches[3] !== '' ) {
967 $char = $matches[3];
968 } else {
969 $char = '\\';
971 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
972 // These characters need to be escaped in strings
973 // Clean up the escape sequence to avoid parsing errors by clients
974 return '\\' . dechex( ord( $char ) ) . ' ';
975 } else {
976 // Decode unnecessary escape
977 return $char;
982 * Take a tag soup fragment listing an HTML element's attributes
983 * and normalize it to well-formed XML, discarding unwanted attributes.
984 * Output is safe for further wikitext processing, with escaping of
985 * values that could trigger problems.
987 * - Normalizes attribute names to lowercase
988 * - Discards attributes not on a whitelist for the given element
989 * - Turns broken or invalid entities into plaintext
990 * - Double-quotes all attribute values
991 * - Attributes without values are given the name as attribute
992 * - Double attributes are discarded
993 * - Unsafe style attributes are discarded
994 * - Prepends space if there are attributes.
996 * @param string $text
997 * @param string $element
998 * @return string
1000 static function fixTagAttributes( $text, $element ) {
1001 if ( trim( $text ) == '' ) {
1002 return '';
1005 $decoded = Sanitizer::decodeTagAttributes( $text );
1006 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1008 return Sanitizer::safeEncodeTagAttributes( $stripped );
1012 * Encode an attribute value for HTML output.
1013 * @param string $text
1014 * @return string HTML-encoded text fragment
1016 static function encodeAttribute( $text ) {
1017 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1019 // Whitespace is normalized during attribute decoding,
1020 // so if we've been passed non-spaces we must encode them
1021 // ahead of time or they won't be preserved.
1022 $encValue = strtr( $encValue, array(
1023 "\n" => '&#10;',
1024 "\r" => '&#13;',
1025 "\t" => '&#9;',
1026 ) );
1028 return $encValue;
1032 * Encode an attribute value for HTML tags, with extra armoring
1033 * against further wiki processing.
1034 * @param string $text
1035 * @return string HTML-encoded text fragment
1037 static function safeEncodeAttribute( $text ) {
1038 $encValue = Sanitizer::encodeAttribute( $text );
1040 # Templates and links may be expanded in later parsing,
1041 # creating invalid or dangerous output. Suppress this.
1042 $encValue = strtr( $encValue, array(
1043 '<' => '&lt;', // This should never happen,
1044 '>' => '&gt;', // we've received invalid input
1045 '"' => '&quot;', // which should have been escaped.
1046 '{' => '&#123;',
1047 '[' => '&#91;',
1048 "''" => '&#39;&#39;',
1049 'ISBN' => '&#73;SBN',
1050 'RFC' => '&#82;FC',
1051 'PMID' => '&#80;MID',
1052 '|' => '&#124;',
1053 '__' => '&#95;_',
1054 ) );
1056 # Stupid hack
1057 $encValue = preg_replace_callback(
1058 '/((?i)' . wfUrlProtocols() . ')/',
1059 array( 'Sanitizer', 'armorLinksCallback' ),
1060 $encValue );
1061 return $encValue;
1065 * Given a value, escape it so that it can be used in an id attribute and
1066 * return it. This will use HTML5 validation if $wgExperimentalHtmlIds is
1067 * true, allowing anything but ASCII whitespace. Otherwise it will use
1068 * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1069 * escaped with lots of dots.
1071 * To ensure we don't have to bother escaping anything, we also strip ', ",
1072 * & even if $wgExperimentalIds is true. TODO: Is this the best tactic?
1073 * We also strip # because it upsets IE, and % because it could be
1074 * ambiguous if it's part of something that looks like a percent escape
1075 * (which don't work reliably in fragments cross-browser).
1077 * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1078 * in the id and name attributes
1079 * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1080 * the id attribute
1081 * @see http://www.whatwg.org/html/elements.html#the-id-attribute
1082 * HTML5 definition of id attribute
1084 * @param string $id id to escape
1085 * @param string|array $options String or array of strings (default is array()):
1086 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1087 * so don't pay attention if the first character isn't valid at the
1088 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
1089 * false.
1090 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
1091 * if $wgExperimentalHtmlIds is used, so we can generate extra
1092 * anchors and links won't break.
1093 * @return string
1095 static function escapeId( $id, $options = array() ) {
1096 global $wgExperimentalHtmlIds;
1097 $options = (array)$options;
1099 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1100 $id = Sanitizer::decodeCharReferences( $id );
1101 $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1102 $id = trim( $id, '_' );
1103 if ( $id === '' ) {
1104 # Must have been all whitespace to start with.
1105 return '_';
1106 } else {
1107 return $id;
1111 # HTML4-style escaping
1112 static $replace = array(
1113 '%3A' => ':',
1114 '%' => '.'
1117 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1118 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1120 if ( !preg_match( '/^[a-zA-Z]/', $id )
1121 && !in_array( 'noninitial', $options ) ) {
1122 // Initial character must be a letter!
1123 $id = "x$id";
1125 return $id;
1129 * Given a value, escape it so that it can be used as a CSS class and
1130 * return it.
1132 * @todo For extra validity, input should be validated UTF-8.
1134 * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1136 * @param string $class
1137 * @return string
1139 static function escapeClass( $class ) {
1140 // Convert ugly stuff to underscores and kill underscores in ugly places
1141 return rtrim( preg_replace(
1142 array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1143 '_',
1144 $class ), '_' );
1148 * Given HTML input, escape with htmlspecialchars but un-escape entities.
1149 * This allows (generally harmless) entities like &#160; to survive.
1151 * @param string $html HTML to escape
1152 * @return string Escaped input
1154 static function escapeHtmlAllowEntities( $html ) {
1155 $html = Sanitizer::decodeCharReferences( $html );
1156 # It seems wise to escape ' as well as ", as a matter of course. Can't
1157 # hurt.
1158 $html = htmlspecialchars( $html, ENT_QUOTES );
1159 return $html;
1163 * Regex replace callback for armoring links against further processing.
1164 * @param array $matches
1165 * @return string
1167 private static function armorLinksCallback( $matches ) {
1168 return str_replace( ':', '&#58;', $matches[1] );
1172 * Return an associative array of attribute names and values from
1173 * a partial tag string. Attribute names are forces to lowercase,
1174 * character references are decoded to UTF-8 text.
1176 * @param string $text
1177 * @return array
1179 public static function decodeTagAttributes( $text ) {
1180 if ( trim( $text ) == '' ) {
1181 return array();
1184 $attribs = array();
1185 $pairs = array();
1186 if ( !preg_match_all(
1187 self::getAttribsRegex(),
1188 $text,
1189 $pairs,
1190 PREG_SET_ORDER ) ) {
1191 return $attribs;
1194 foreach ( $pairs as $set ) {
1195 $attribute = strtolower( $set[1] );
1196 $value = Sanitizer::getTagAttributeCallback( $set );
1198 // Normalize whitespace
1199 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1200 $value = trim( $value );
1202 // Decode character references
1203 $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1205 return $attribs;
1209 * Build a partial tag string from an associative array of attribute
1210 * names and values as returned by decodeTagAttributes.
1212 * @param array $assoc_array
1213 * @return string
1215 public static function safeEncodeTagAttributes( $assoc_array ) {
1216 $attribs = array();
1217 foreach ( $assoc_array as $attribute => $value ) {
1218 $encAttribute = htmlspecialchars( $attribute );
1219 $encValue = Sanitizer::safeEncodeAttribute( $value );
1221 $attribs[] = "$encAttribute=\"$encValue\"";
1223 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1227 * Pick the appropriate attribute value from a match set from the
1228 * attribs regex matches.
1230 * @param array $set
1231 * @throws MWException when tag conditions are not met.
1232 * @return string
1234 private static function getTagAttributeCallback( $set ) {
1235 if ( isset( $set[6] ) ) {
1236 # Illegal #XXXXXX color with no quotes.
1237 return $set[6];
1238 } elseif ( isset( $set[5] ) ) {
1239 # No quotes.
1240 return $set[5];
1241 } elseif ( isset( $set[4] ) ) {
1242 # Single-quoted
1243 return $set[4];
1244 } elseif ( isset( $set[3] ) ) {
1245 # Double-quoted
1246 return $set[3];
1247 } elseif ( !isset( $set[2] ) ) {
1248 # In XHTML, attributes must have a value.
1249 # For 'reduced' form, return explicitly the attribute name here.
1250 return $set[1];
1251 } else {
1252 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1257 * Normalize whitespace and character references in an XML source-
1258 * encoded text for an attribute value.
1260 * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1261 * but note that we're not returning the value, but are returning
1262 * XML source fragments that will be slapped into output.
1264 * @param string $text
1265 * @return string
1266 * @todo Remove, unused?
1268 private static function normalizeAttributeValue( $text ) {
1269 return str_replace( '"', '&quot;',
1270 self::normalizeWhitespace(
1271 Sanitizer::normalizeCharReferences( $text ) ) );
1275 * @param string $text
1276 * @return string
1278 private static function normalizeWhitespace( $text ) {
1279 return preg_replace(
1280 '/\r\n|[\x20\x0d\x0a\x09]/',
1281 ' ',
1282 $text );
1286 * Normalizes whitespace in a section name, such as might be returned
1287 * by Parser::stripSectionName(), for use in the id's that are used for
1288 * section links.
1290 * @param string $section
1291 * @return string
1293 static function normalizeSectionNameWhitespace( $section ) {
1294 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1298 * Ensure that any entities and character references are legal
1299 * for XML and XHTML specifically. Any stray bits will be
1300 * &amp;-escaped to result in a valid text fragment.
1302 * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1303 * numericized (this way we're well-formed even without a DTD)
1304 * b. any numeric char refs must be legal chars, not invalid or forbidden
1305 * c. use lower cased "&#x", not "&#X"
1306 * d. fix or reject non-valid attributes
1308 * @param string $text
1309 * @return string
1310 * @private
1312 static function normalizeCharReferences( $text ) {
1313 return preg_replace_callback(
1314 self::CHAR_REFS_REGEX,
1315 array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1316 $text );
1320 * @param string $matches
1321 * @return string
1323 static function normalizeCharReferencesCallback( $matches ) {
1324 $ret = null;
1325 if ( $matches[1] != '' ) {
1326 $ret = Sanitizer::normalizeEntity( $matches[1] );
1327 } elseif ( $matches[2] != '' ) {
1328 $ret = Sanitizer::decCharReference( $matches[2] );
1329 } elseif ( $matches[3] != '' ) {
1330 $ret = Sanitizer::hexCharReference( $matches[3] );
1332 if ( is_null( $ret ) ) {
1333 return htmlspecialchars( $matches[0] );
1334 } else {
1335 return $ret;
1340 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1341 * return the equivalent numeric entity reference (except for the core &lt;
1342 * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1343 * the HTML equivalent. Otherwise, returns HTML-escaped text of
1344 * pseudo-entity source (eg &amp;foo;)
1346 * @param string $name
1347 * @return string
1349 static function normalizeEntity( $name ) {
1350 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1351 return '&' . self::$htmlEntityAliases[$name] . ';';
1352 } elseif ( in_array( $name,
1353 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1354 return "&$name;";
1355 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1356 return '&#' . self::$htmlEntities[$name] . ';';
1357 } else {
1358 return "&amp;$name;";
1363 * @param int $codepoint
1364 * @return null|string
1366 static function decCharReference( $codepoint ) {
1367 $point = intval( $codepoint );
1368 if ( Sanitizer::validateCodepoint( $point ) ) {
1369 return sprintf( '&#%d;', $point );
1370 } else {
1371 return null;
1376 * @param int $codepoint
1377 * @return null|string
1379 static function hexCharReference( $codepoint ) {
1380 $point = hexdec( $codepoint );
1381 if ( Sanitizer::validateCodepoint( $point ) ) {
1382 return sprintf( '&#x%x;', $point );
1383 } else {
1384 return null;
1389 * Returns true if a given Unicode codepoint is a valid character in XML.
1390 * @param int $codepoint
1391 * @return bool
1393 private static function validateCodepoint( $codepoint ) {
1394 return $codepoint == 0x09
1395 || $codepoint == 0x0a
1396 || $codepoint == 0x0d
1397 || ( $codepoint >= 0x20 && $codepoint <= 0xd7ff )
1398 || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1399 || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1403 * Decode any character references, numeric or named entities,
1404 * in the text and return a UTF-8 string.
1406 * @param string $text
1407 * @return string
1409 public static function decodeCharReferences( $text ) {
1410 return preg_replace_callback(
1411 self::CHAR_REFS_REGEX,
1412 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1413 $text );
1417 * Decode any character references, numeric or named entities,
1418 * in the next and normalize the resulting string. (bug 14952)
1420 * This is useful for page titles, not for text to be displayed,
1421 * MediaWiki allows HTML entities to escape normalization as a feature.
1423 * @param string $text Already normalized, containing entities
1424 * @return string Still normalized, without entities
1426 public static function decodeCharReferencesAndNormalize( $text ) {
1427 global $wgContLang;
1428 $text = preg_replace_callback(
1429 self::CHAR_REFS_REGEX,
1430 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1431 $text, /* limit */ -1, $count );
1433 if ( $count ) {
1434 return $wgContLang->normalize( $text );
1435 } else {
1436 return $text;
1441 * @param string $matches
1442 * @return string
1444 static function decodeCharReferencesCallback( $matches ) {
1445 if ( $matches[1] != '' ) {
1446 return Sanitizer::decodeEntity( $matches[1] );
1447 } elseif ( $matches[2] != '' ) {
1448 return Sanitizer::decodeChar( intval( $matches[2] ) );
1449 } elseif ( $matches[3] != '' ) {
1450 return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1452 # Last case should be an ampersand by itself
1453 return $matches[0];
1457 * Return UTF-8 string for a codepoint if that is a valid
1458 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1459 * @param int $codepoint
1460 * @return string
1461 * @private
1463 static function decodeChar( $codepoint ) {
1464 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1465 return codepointToUtf8( $codepoint );
1466 } else {
1467 return UTF8_REPLACEMENT;
1472 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1473 * return the UTF-8 encoding of that character. Otherwise, returns
1474 * pseudo-entity source (eg "&foo;")
1476 * @param string $name
1477 * @return string
1479 static function decodeEntity( $name ) {
1480 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1481 $name = self::$htmlEntityAliases[$name];
1483 if ( isset( self::$htmlEntities[$name] ) ) {
1484 return codepointToUtf8( self::$htmlEntities[$name] );
1485 } else {
1486 return "&$name;";
1491 * Fetch the whitelist of acceptable attributes for a given element name.
1493 * @param string $element
1494 * @return array
1496 static function attributeWhitelist( $element ) {
1497 $list = Sanitizer::setupAttributeWhitelist();
1498 return isset( $list[$element] )
1499 ? $list[$element]
1500 : array();
1504 * Foreach array key (an allowed HTML element), return an array
1505 * of allowed attributes
1506 * @return array
1508 static function setupAttributeWhitelist() {
1509 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
1511 static $whitelist, $staticInitialised;
1512 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgAllowMicrodataAttributes' ) );
1514 if ( isset( $whitelist ) && $staticInitialised == $globalContext ) {
1515 return $whitelist;
1518 $common = array(
1519 # HTML
1520 'id',
1521 'class',
1522 'style',
1523 'lang',
1524 'dir',
1525 'title',
1527 # WAI-ARIA
1528 'role',
1531 if ( $wgAllowRdfaAttributes ) {
1532 # RDFa attributes as specified in section 9 of
1533 # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1534 $common = array_merge( $common, array(
1535 'about', 'property', 'resource', 'datatype', 'typeof',
1536 ) );
1539 if ( $wgAllowMicrodataAttributes ) {
1540 # add HTML5 microdata tags as specified by
1541 # http://www.whatwg.org/html/microdata.html#the-microdata-model
1542 $common = array_merge( $common, array(
1543 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1544 ) );
1547 $block = array_merge( $common, array( 'align' ) );
1548 $tablealign = array( 'align', 'valign' );
1549 $tablecell = array(
1550 'abbr',
1551 'axis',
1552 'headers',
1553 'scope',
1554 'rowspan',
1555 'colspan',
1556 'nowrap', # deprecated
1557 'width', # deprecated
1558 'height', # deprecated
1559 'bgcolor', # deprecated
1562 # Numbers refer to sections in HTML 4.01 standard describing the element.
1563 # See: http://www.w3.org/TR/html4/
1564 $whitelist = array(
1565 # 7.5.4
1566 'div' => $block,
1567 'center' => $common, # deprecated
1568 'span' => $common,
1570 # 7.5.5
1571 'h1' => $block,
1572 'h2' => $block,
1573 'h3' => $block,
1574 'h4' => $block,
1575 'h5' => $block,
1576 'h6' => $block,
1578 # 7.5.6
1579 # address
1581 # 8.2.4
1582 'bdo' => $common,
1584 # 9.2.1
1585 'em' => $common,
1586 'strong' => $common,
1587 'cite' => $common,
1588 'dfn' => $common,
1589 'code' => $common,
1590 'samp' => $common,
1591 'kbd' => $common,
1592 'var' => $common,
1593 'abbr' => $common,
1594 # acronym
1596 # 9.2.2
1597 'blockquote' => array_merge( $common, array( 'cite' ) ),
1598 'q' => array_merge( $common, array( 'cite' ) ),
1600 # 9.2.3
1601 'sub' => $common,
1602 'sup' => $common,
1604 # 9.3.1
1605 'p' => $block,
1607 # 9.3.2
1608 'br' => array_merge( $common, array( 'clear' ) ),
1610 # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1611 'wbr' => $common,
1613 # 9.3.4
1614 'pre' => array_merge( $common, array( 'width' ) ),
1616 # 9.4
1617 'ins' => array_merge( $common, array( 'cite', 'datetime' ) ),
1618 'del' => array_merge( $common, array( 'cite', 'datetime' ) ),
1620 # 10.2
1621 'ul' => array_merge( $common, array( 'type' ) ),
1622 'ol' => array_merge( $common, array( 'type', 'start' ) ),
1623 'li' => array_merge( $common, array( 'type', 'value' ) ),
1625 # 10.3
1626 'dl' => $common,
1627 'dd' => $common,
1628 'dt' => $common,
1630 # 11.2.1
1631 'table' => array_merge( $common,
1632 array( 'summary', 'width', 'border', 'frame',
1633 'rules', 'cellspacing', 'cellpadding',
1634 'align', 'bgcolor',
1635 ) ),
1637 # 11.2.2
1638 'caption' => $block,
1640 # 11.2.3
1641 'thead' => $common,
1642 'tfoot' => $common,
1643 'tbody' => $common,
1645 # 11.2.4
1646 'colgroup' => array_merge( $common, array( 'span' ) ),
1647 'col' => array_merge( $common, array( 'span' ) ),
1649 # 11.2.5
1650 'tr' => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1652 # 11.2.6
1653 'td' => array_merge( $common, $tablecell, $tablealign ),
1654 'th' => array_merge( $common, $tablecell, $tablealign ),
1656 # 12.2
1657 # NOTE: <a> is not allowed directly, but the attrib
1658 # whitelist is used from the Parser object
1659 'a' => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1661 # 13.2
1662 # Not usually allowed, but may be used for extension-style hooks
1663 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1664 # true
1665 'img' => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1667 # 15.2.1
1668 'tt' => $common,
1669 'b' => $common,
1670 'i' => $common,
1671 'big' => $common,
1672 'small' => $common,
1673 'strike' => $common,
1674 's' => $common,
1675 'u' => $common,
1677 # 15.2.2
1678 'font' => array_merge( $common, array( 'size', 'color', 'face' ) ),
1679 # basefont
1681 # 15.3
1682 'hr' => array_merge( $common, array( 'width' ) ),
1684 # HTML Ruby annotation text module, simple ruby only.
1685 # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
1686 'ruby' => $common,
1687 # rbc
1688 # rtc
1689 'rb' => $common,
1690 'rt' => $common, #array_merge( $common, array( 'rbspan' ) ),
1691 'rp' => $common,
1693 # MathML root element, where used for extensions
1694 # 'title' may not be 100% valid here; it's XHTML
1695 # http://www.w3.org/TR/REC-MathML/
1696 'math' => array( 'class', 'style', 'id', 'title' ),
1698 # HTML 5 section 4.6
1699 'bdi' => $common,
1701 # HTML5 elements, defined by:
1702 # http://www.whatwg.org/html/
1703 'data' => array_merge( $common, array( 'value' ) ),
1704 'time' => array_merge( $common, array( 'datetime' ) ),
1705 'mark' => $common,
1707 // meta and link are only permitted by removeHTMLtags when Microdata
1708 // is enabled so we don't bother adding a conditional to hide these
1709 // Also meta and link are only valid in WikiText as Microdata elements
1710 // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1711 // So we don't bother including $common attributes that have no purpose.
1712 'meta' => array( 'itemprop', 'content' ),
1713 'link' => array( 'itemprop', 'href' ),
1716 $staticInitialised = $globalContext;
1718 return $whitelist;
1722 * Take a fragment of (potentially invalid) HTML and return
1723 * a version with any tags removed, encoded as plain text.
1725 * Warning: this return value must be further escaped for literal
1726 * inclusion in HTML output as of 1.10!
1728 * @param string $text HTML fragment
1729 * @return string
1731 static function stripAllTags( $text ) {
1732 # Actual <tags>
1733 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1735 # Normalize &entities and whitespace
1736 $text = self::decodeCharReferences( $text );
1737 $text = self::normalizeWhitespace( $text );
1739 return $text;
1743 * Hack up a private DOCTYPE with HTML's standard entity declarations.
1744 * PHP 4 seemed to know these if you gave it an HTML doctype, but
1745 * PHP 5.1 doesn't.
1747 * Use for passing XHTML fragments to PHP's XML parsing functions
1749 * @return string
1751 static function hackDocType() {
1752 $out = "<!DOCTYPE html [\n";
1753 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1754 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1756 $out .= "]>\n";
1757 return $out;
1761 * @param string $url
1762 * @return mixed|string
1764 static function cleanUrl( $url ) {
1765 # Normalize any HTML entities in input. They will be
1766 # re-escaped by makeExternalLink().
1767 $url = Sanitizer::decodeCharReferences( $url );
1769 # Escape any control characters introduced by the above step
1770 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1771 array( __CLASS__, 'cleanUrlCallback' ), $url );
1773 # Validate hostname portion
1774 $matches = array();
1775 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1776 list( /* $whole */, $protocol, $host, $rest ) = $matches;
1778 // Characters that will be ignored in IDNs.
1779 // http://tools.ietf.org/html/3454#section-3.1
1780 // Strip them before further processing so blacklists and such work.
1781 $strip = "/
1782 \\s| # general whitespace
1783 \xc2\xad| # 00ad SOFT HYPHEN
1784 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1785 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1786 \xe2\x81\xa0| # 2060 WORD JOINER
1787 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1788 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
1789 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1790 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1791 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1792 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1793 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1794 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1795 /xuD";
1797 $host = preg_replace( $strip, '', $host );
1799 // @todo FIXME: Validate hostnames here
1801 return $protocol . $host . $rest;
1802 } else {
1803 return $url;
1808 * @param array $matches
1809 * @return string
1811 static function cleanUrlCallback( $matches ) {
1812 return urlencode( $matches[0] );
1816 * Does a string look like an e-mail address?
1818 * This validates an email address using an HTML5 specification found at:
1819 * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1820 * Which as of 2011-01-24 says:
1822 * A valid e-mail address is a string that matches the ABNF production
1823 * 1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1824 * in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1825 * 3.5.
1827 * This function is an implementation of the specification as requested in
1828 * bug 22449.
1830 * Client-side forms will use the same standard validation rules via JS or
1831 * HTML 5 validation; additional restrictions can be enforced server-side
1832 * by extensions via the 'isValidEmailAddr' hook.
1834 * Note that this validation doesn't 100% match RFC 2822, but is believed
1835 * to be liberal enough for wide use. Some invalid addresses will still
1836 * pass validation here.
1838 * @since 1.18
1840 * @param string $addr E-mail address
1841 * @return bool
1843 public static function validateEmail( $addr ) {
1844 $result = null;
1845 if ( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1846 return $result;
1849 // Please note strings below are enclosed in brackets [], this make the
1850 // hyphen "-" a range indicator. Hence it is double backslashed below.
1851 // See bug 26948
1852 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1853 $rfc1034_ldh_str = "a-z0-9\\-";
1855 $html5_email_regexp = "/
1856 ^ # start of string
1857 [$rfc5322_atext\\.]+ # user part which is liberal :p
1858 @ # 'apostrophe'
1859 [$rfc1034_ldh_str]+ # First domain part
1860 (\\.[$rfc1034_ldh_str]+)* # Following part prefixed with a dot
1861 $ # End of string
1862 /ix"; // case Insensitive, eXtended
1864 return (bool)preg_match( $html5_email_regexp, $addr );