Use PHPUnit_Framework_TestCase in a few utils/ tests
[mediawiki.git] / includes / Sanitizer.php
blob11c30a25912e6b49f839780bc659890ff015efdc
1 <?php
2 /**
3 * HTML sanitizer for %MediaWiki.
5 * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
6 * https://www.mediawiki.org/
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 2 of the License, or
11 * (at your option) any later version.
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License along
19 * with this program; if not, write to the Free Software Foundation, Inc.,
20 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
21 * http://www.gnu.org/copyleft/gpl.html
23 * @file
24 * @ingroup Parser
27 /**
28 * HTML sanitizer for MediaWiki
29 * @ingroup Parser
31 class Sanitizer {
32 /**
33 * Regular expression to match various types of character references in
34 * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
36 const CHAR_REFS_REGEX =
37 '/&([A-Za-z0-9\x80-\xff]+);
38 |&\#([0-9]+);
39 |&\#[xX]([0-9A-Fa-f]+);
40 |(&)/x';
42 /**
43 * Blacklist for evil uris like javascript:
44 * WARNING: DO NOT use this in any place that actually requires blacklisting
45 * for security reasons. There are NUMEROUS[1] ways to bypass blacklisting, the
46 * only way to be secure from javascript: uri based xss vectors is to whitelist
47 * things that you know are safe and deny everything else.
48 * [1]: http://ha.ckers.org/xss.html
50 const EVIL_URI_PATTERN = '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
51 const XMLNS_ATTRIBUTE_PATTERN = "/^xmlns:[:A-Z_a-z-.0-9]+$/";
53 /**
54 * List of all named character entities defined in HTML 4.01
55 * http://www.w3.org/TR/html4/sgml/entities.html
56 * As well as &apos; which is only defined starting in XHTML1.
58 private static $htmlEntities = array(
59 'Aacute' => 193,
60 'aacute' => 225,
61 'Acirc' => 194,
62 'acirc' => 226,
63 'acute' => 180,
64 'AElig' => 198,
65 'aelig' => 230,
66 'Agrave' => 192,
67 'agrave' => 224,
68 'alefsym' => 8501,
69 'Alpha' => 913,
70 'alpha' => 945,
71 'amp' => 38,
72 'and' => 8743,
73 'ang' => 8736,
74 'apos' => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
75 'Aring' => 197,
76 'aring' => 229,
77 'asymp' => 8776,
78 'Atilde' => 195,
79 'atilde' => 227,
80 'Auml' => 196,
81 'auml' => 228,
82 'bdquo' => 8222,
83 'Beta' => 914,
84 'beta' => 946,
85 'brvbar' => 166,
86 'bull' => 8226,
87 'cap' => 8745,
88 'Ccedil' => 199,
89 'ccedil' => 231,
90 'cedil' => 184,
91 'cent' => 162,
92 'Chi' => 935,
93 'chi' => 967,
94 'circ' => 710,
95 'clubs' => 9827,
96 'cong' => 8773,
97 'copy' => 169,
98 'crarr' => 8629,
99 'cup' => 8746,
100 'curren' => 164,
101 'dagger' => 8224,
102 'Dagger' => 8225,
103 'darr' => 8595,
104 'dArr' => 8659,
105 'deg' => 176,
106 'Delta' => 916,
107 'delta' => 948,
108 'diams' => 9830,
109 'divide' => 247,
110 'Eacute' => 201,
111 'eacute' => 233,
112 'Ecirc' => 202,
113 'ecirc' => 234,
114 'Egrave' => 200,
115 'egrave' => 232,
116 'empty' => 8709,
117 'emsp' => 8195,
118 'ensp' => 8194,
119 'Epsilon' => 917,
120 'epsilon' => 949,
121 'equiv' => 8801,
122 'Eta' => 919,
123 'eta' => 951,
124 'ETH' => 208,
125 'eth' => 240,
126 'Euml' => 203,
127 'euml' => 235,
128 'euro' => 8364,
129 'exist' => 8707,
130 'fnof' => 402,
131 'forall' => 8704,
132 'frac12' => 189,
133 'frac14' => 188,
134 'frac34' => 190,
135 'frasl' => 8260,
136 'Gamma' => 915,
137 'gamma' => 947,
138 'ge' => 8805,
139 'gt' => 62,
140 'harr' => 8596,
141 'hArr' => 8660,
142 'hearts' => 9829,
143 'hellip' => 8230,
144 'Iacute' => 205,
145 'iacute' => 237,
146 'Icirc' => 206,
147 'icirc' => 238,
148 'iexcl' => 161,
149 'Igrave' => 204,
150 'igrave' => 236,
151 'image' => 8465,
152 'infin' => 8734,
153 'int' => 8747,
154 'Iota' => 921,
155 'iota' => 953,
156 'iquest' => 191,
157 'isin' => 8712,
158 'Iuml' => 207,
159 'iuml' => 239,
160 'Kappa' => 922,
161 'kappa' => 954,
162 'Lambda' => 923,
163 'lambda' => 955,
164 'lang' => 9001,
165 'laquo' => 171,
166 'larr' => 8592,
167 'lArr' => 8656,
168 'lceil' => 8968,
169 'ldquo' => 8220,
170 'le' => 8804,
171 'lfloor' => 8970,
172 'lowast' => 8727,
173 'loz' => 9674,
174 'lrm' => 8206,
175 'lsaquo' => 8249,
176 'lsquo' => 8216,
177 'lt' => 60,
178 'macr' => 175,
179 'mdash' => 8212,
180 'micro' => 181,
181 'middot' => 183,
182 'minus' => 8722,
183 'Mu' => 924,
184 'mu' => 956,
185 'nabla' => 8711,
186 'nbsp' => 160,
187 'ndash' => 8211,
188 'ne' => 8800,
189 'ni' => 8715,
190 'not' => 172,
191 'notin' => 8713,
192 'nsub' => 8836,
193 'Ntilde' => 209,
194 'ntilde' => 241,
195 'Nu' => 925,
196 'nu' => 957,
197 'Oacute' => 211,
198 'oacute' => 243,
199 'Ocirc' => 212,
200 'ocirc' => 244,
201 'OElig' => 338,
202 'oelig' => 339,
203 'Ograve' => 210,
204 'ograve' => 242,
205 'oline' => 8254,
206 'Omega' => 937,
207 'omega' => 969,
208 'Omicron' => 927,
209 'omicron' => 959,
210 'oplus' => 8853,
211 'or' => 8744,
212 'ordf' => 170,
213 'ordm' => 186,
214 'Oslash' => 216,
215 'oslash' => 248,
216 'Otilde' => 213,
217 'otilde' => 245,
218 'otimes' => 8855,
219 'Ouml' => 214,
220 'ouml' => 246,
221 'para' => 182,
222 'part' => 8706,
223 'permil' => 8240,
224 'perp' => 8869,
225 'Phi' => 934,
226 'phi' => 966,
227 'Pi' => 928,
228 'pi' => 960,
229 'piv' => 982,
230 'plusmn' => 177,
231 'pound' => 163,
232 'prime' => 8242,
233 'Prime' => 8243,
234 'prod' => 8719,
235 'prop' => 8733,
236 'Psi' => 936,
237 'psi' => 968,
238 'quot' => 34,
239 'radic' => 8730,
240 'rang' => 9002,
241 'raquo' => 187,
242 'rarr' => 8594,
243 'rArr' => 8658,
244 'rceil' => 8969,
245 'rdquo' => 8221,
246 'real' => 8476,
247 'reg' => 174,
248 'rfloor' => 8971,
249 'Rho' => 929,
250 'rho' => 961,
251 'rlm' => 8207,
252 'rsaquo' => 8250,
253 'rsquo' => 8217,
254 'sbquo' => 8218,
255 'Scaron' => 352,
256 'scaron' => 353,
257 'sdot' => 8901,
258 'sect' => 167,
259 'shy' => 173,
260 'Sigma' => 931,
261 'sigma' => 963,
262 'sigmaf' => 962,
263 'sim' => 8764,
264 'spades' => 9824,
265 'sub' => 8834,
266 'sube' => 8838,
267 'sum' => 8721,
268 'sup' => 8835,
269 'sup1' => 185,
270 'sup2' => 178,
271 'sup3' => 179,
272 'supe' => 8839,
273 'szlig' => 223,
274 'Tau' => 932,
275 'tau' => 964,
276 'there4' => 8756,
277 'Theta' => 920,
278 'theta' => 952,
279 'thetasym' => 977,
280 'thinsp' => 8201,
281 'THORN' => 222,
282 'thorn' => 254,
283 'tilde' => 732,
284 'times' => 215,
285 'trade' => 8482,
286 'Uacute' => 218,
287 'uacute' => 250,
288 'uarr' => 8593,
289 'uArr' => 8657,
290 'Ucirc' => 219,
291 'ucirc' => 251,
292 'Ugrave' => 217,
293 'ugrave' => 249,
294 'uml' => 168,
295 'upsih' => 978,
296 'Upsilon' => 933,
297 'upsilon' => 965,
298 'Uuml' => 220,
299 'uuml' => 252,
300 'weierp' => 8472,
301 'Xi' => 926,
302 'xi' => 958,
303 'Yacute' => 221,
304 'yacute' => 253,
305 'yen' => 165,
306 'Yuml' => 376,
307 'yuml' => 255,
308 'Zeta' => 918,
309 'zeta' => 950,
310 'zwj' => 8205,
311 'zwnj' => 8204
315 * Character entity aliases accepted by MediaWiki
317 private static $htmlEntityAliases = array(
318 'רלמ' => 'rlm',
319 'رلم' => 'rlm',
323 * Lazy-initialised attributes regex, see getAttribsRegex()
325 private static $attribsRegex;
328 * Regular expression to match HTML/XML attribute pairs within a tag.
329 * Allows some... latitude.
330 * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
331 * @return string
333 static function getAttribsRegex() {
334 if ( self::$attribsRegex === null ) {
335 $attribFirst = '[:A-Z_a-z0-9]';
336 $attrib = '[:A-Z_a-z-.0-9]';
337 $space = '[\x09\x0a\x0d\x20]';
338 self::$attribsRegex =
339 "/(?:^|$space)({$attribFirst}{$attrib}*)
340 ($space*=$space*
342 # The attribute value: quoted or alone
343 \"([^<\"]*)\"
344 | '([^<']*)'
345 | ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
346 | (\#[0-9a-fA-F]+) # Technically wrong, but lots of
347 # colors are specified like this.
348 # We'll be normalizing it.
350 )?(?=$space|\$)/sx";
352 return self::$attribsRegex;
356 * Cleans up HTML, removes dangerous tags and attributes, and
357 * removes HTML comments
358 * @private
359 * @param string $text
360 * @param callable $processCallback Callback to do any variable or parameter
361 * replacements in HTML attribute values
362 * @param array|bool $args Arguments for the processing callback
363 * @param array $extratags For any extra tags to include
364 * @param array $removetags For any tags (default or extra) to exclude
365 * @return string
367 static function removeHTMLtags( $text, $processCallback = null,
368 $args = array(), $extratags = array(), $removetags = array()
370 global $wgUseTidy, $wgAllowMicrodataAttributes, $wgAllowImageTag;
372 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
373 $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
375 wfProfileIn( __METHOD__ );
377 // Base our staticInitialised variable off of the global config state so that if the globals
378 // are changed (like in the screwed up test system) we will re-initialise the settings.
379 $globalContext = implode( '-', compact( 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
380 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
382 $htmlpairsStatic = array( # Tags that must be closed
383 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
384 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
385 'strike', 'strong', 'tt', 'var', 'div', 'center',
386 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
387 'ruby', 'rb', 'rp', 'rt', 'rtc', 'p', 'span', 'abbr', 'dfn',
388 'kbd', 'samp', 'data', 'time', 'mark'
390 $htmlsingle = array(
391 'br', 'wbr', 'hr', 'li', 'dt', 'dd'
393 $htmlsingleonly = array( # Elements that cannot have close tags
394 'br', 'wbr', 'hr'
396 if ( $wgAllowMicrodataAttributes ) {
397 $htmlsingle[] = $htmlsingleonly[] = 'meta';
398 $htmlsingle[] = $htmlsingleonly[] = 'link';
400 $htmlnest = array( # Tags that can be nested--??
401 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
402 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
403 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
405 $tabletags = array( # Can only appear inside table, we will close them
406 'td', 'th', 'tr',
408 $htmllist = array( # Tags used by list
409 'ul', 'ol',
411 $listtags = array( # Tags that can appear in a list
412 'li',
415 if ( $wgAllowImageTag ) {
416 $htmlsingle[] = 'img';
417 $htmlsingleonly[] = 'img';
420 $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
421 $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
423 # Convert them all to hashtables for faster lookup
424 $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
425 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
426 foreach ( $vars as $var ) {
427 $$var = array_flip( $$var );
429 $staticInitialised = $globalContext;
431 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
432 $extratags = array_flip( $extratags );
433 $removetags = array_flip( $removetags );
434 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
435 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
437 # Remove HTML comments
438 $text = Sanitizer::removeHTMLcomments( $text );
439 $bits = explode( '<', $text );
440 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
441 if ( !$wgUseTidy ) {
442 $tagstack = $tablestack = array();
443 foreach ( $bits as $x ) {
444 $regs = array();
445 # $slash: Does the current element start with a '/'?
446 # $t: Current element name
447 # $params: String between element name and >
448 # $brace: Ending '>' or '/>'
449 # $rest: Everything until the next element of $bits
450 if ( preg_match( '!^(/?)([^\\s/>]+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
451 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
452 } else {
453 $slash = $t = $params = $brace = $rest = null;
456 $badtag = false;
457 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
458 # Check our stack
459 if ( $slash && isset( $htmlsingleonly[$t] ) ) {
460 $badtag = true;
461 } elseif ( $slash ) {
462 # Closing a tag... is it the one we just opened?
463 wfSuppressWarnings();
464 $ot = array_pop( $tagstack );
465 wfRestoreWarnings();
467 if ( $ot != $t ) {
468 if ( isset( $htmlsingleallowed[$ot] ) ) {
469 # Pop all elements with an optional close tag
470 # and see if we find a match below them
471 $optstack = array();
472 array_push( $optstack, $ot );
473 wfSuppressWarnings();
474 $ot = array_pop( $tagstack );
475 wfRestoreWarnings();
476 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
477 array_push( $optstack, $ot );
478 wfSuppressWarnings();
479 $ot = array_pop( $tagstack );
480 wfRestoreWarnings();
482 if ( $t != $ot ) {
483 # No match. Push the optional elements back again
484 $badtag = true;
485 wfSuppressWarnings();
486 $ot = array_pop( $optstack );
487 wfRestoreWarnings();
488 while ( $ot ) {
489 array_push( $tagstack, $ot );
490 wfSuppressWarnings();
491 $ot = array_pop( $optstack );
492 wfRestoreWarnings();
495 } else {
496 wfSuppressWarnings();
497 array_push( $tagstack, $ot );
498 wfRestoreWarnings();
500 # <li> can be nested in <ul> or <ol>, skip those cases:
501 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
502 $badtag = true;
505 } else {
506 if ( $t == 'table' ) {
507 $tagstack = array_pop( $tablestack );
510 $newparams = '';
511 } else {
512 # Keep track for later
513 if ( isset( $tabletags[$t] ) && !in_array( 'table', $tagstack ) ) {
514 $badtag = true;
515 } elseif ( in_array( $t, $tagstack ) && !isset( $htmlnest[$t] ) ) {
516 $badtag = true;
517 # Is it a self closed htmlpair ? (bug 5487)
518 } elseif ( $brace == '/>' && isset( $htmlpairs[$t] ) ) {
519 $badtag = true;
520 } elseif ( isset( $htmlsingleonly[$t] ) ) {
521 # Hack to force empty tag for unclosable elements
522 $brace = '/>';
523 } elseif ( isset( $htmlsingle[$t] ) ) {
524 # Hack to not close $htmlsingle tags
525 $brace = null;
526 # Still need to push this optionally-closed tag to
527 # the tag stack so that we can match end tags
528 # instead of marking them as bad.
529 array_push( $tagstack, $t );
530 } elseif ( isset( $tabletags[$t] ) && in_array( $t, $tagstack ) ) {
531 // New table tag but forgot to close the previous one
532 $text .= "</$t>";
533 } else {
534 if ( $t == 'table' ) {
535 array_push( $tablestack, $tagstack );
536 $tagstack = array();
538 array_push( $tagstack, $t );
541 # Replace any variables or template parameters with
542 # plaintext results.
543 if ( is_callable( $processCallback ) ) {
544 call_user_func_array( $processCallback, array( &$params, $args ) );
547 if ( !Sanitizer::validateTag( $params, $t ) ) {
548 $badtag = true;
551 # Strip non-approved attributes from the tag
552 $newparams = Sanitizer::fixTagAttributes( $params, $t );
554 if ( !$badtag ) {
555 $rest = str_replace( '>', '&gt;', $rest );
556 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
557 $text .= "<$slash$t$newparams$close>$rest";
558 continue;
561 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
563 # Close off any remaining tags
564 while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
565 $text .= "</$t>\n";
566 if ( $t == 'table' ) {
567 $tagstack = array_pop( $tablestack );
570 } else {
571 # this might be possible using tidy itself
572 foreach ( $bits as $x ) {
573 preg_match(
574 '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
576 $regs
579 wfSuppressWarnings();
580 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
581 wfRestoreWarnings();
583 $badtag = false;
584 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
585 if ( is_callable( $processCallback ) ) {
586 call_user_func_array( $processCallback, array( &$params, $args ) );
589 if ( !Sanitizer::validateTag( $params, $t ) ) {
590 $badtag = true;
593 $newparams = Sanitizer::fixTagAttributes( $params, $t );
594 if ( !$badtag ) {
595 $rest = str_replace( '>', '&gt;', $rest );
596 $text .= "<$slash$t$newparams$brace$rest";
597 continue;
600 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
603 wfProfileOut( __METHOD__ );
604 return $text;
608 * Remove '<!--', '-->', and everything between.
609 * To avoid leaving blank lines, when a comment is both preceded
610 * and followed by a newline (ignoring spaces), trim leading and
611 * trailing spaces and one of the newlines.
613 * @private
614 * @param string $text
615 * @return string
617 static function removeHTMLcomments( $text ) {
618 wfProfileIn( __METHOD__ );
619 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
620 $end = strpos( $text, '-->', $start + 4 );
621 if ( $end === false ) {
622 # Unterminated comment; bail out
623 break;
626 $end += 3;
628 # Trim space and newline if the comment is both
629 # preceded and followed by a newline
630 $spaceStart = max( $start - 1, 0 );
631 $spaceLen = $end - $spaceStart;
632 while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
633 $spaceStart--;
634 $spaceLen++;
636 while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
637 $spaceLen++;
639 if ( substr( $text, $spaceStart, 1 ) === "\n"
640 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
641 # Remove the comment, leading and trailing
642 # spaces, and leave only one newline.
643 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
644 } else {
645 # Remove just the comment.
646 $text = substr_replace( $text, '', $start, $end - $start );
649 wfProfileOut( __METHOD__ );
650 return $text;
654 * Takes attribute names and values for a tag and the tag name and
655 * validates that the tag is allowed to be present.
656 * This DOES NOT validate the attributes, nor does it validate the
657 * tags themselves. This method only handles the special circumstances
658 * where we may want to allow a tag within content but ONLY when it has
659 * specific attributes set.
661 * @param string $params
662 * @param string $element
663 * @return bool
665 static function validateTag( $params, $element ) {
666 $params = Sanitizer::decodeTagAttributes( $params );
668 if ( $element == 'meta' || $element == 'link' ) {
669 if ( !isset( $params['itemprop'] ) ) {
670 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
671 return false;
673 if ( $element == 'meta' && !isset( $params['content'] ) ) {
674 // <meta> must have a content="" for the itemprop
675 return false;
677 if ( $element == 'link' && !isset( $params['href'] ) ) {
678 // <link> must have an associated href=""
679 return false;
683 return true;
687 * Take an array of attribute names and values and normalize or discard
688 * illegal values for the given element type.
690 * - Discards attributes not on a whitelist for the given element
691 * - Unsafe style attributes are discarded
692 * - Invalid id attributes are re-encoded
694 * @param array $attribs
695 * @param string $element
696 * @return array
698 * @todo Check for legal values where the DTD limits things.
699 * @todo Check for unique id attribute :P
701 static function validateTagAttributes( $attribs, $element ) {
702 return Sanitizer::validateAttributes( $attribs,
703 Sanitizer::attributeWhitelist( $element ) );
707 * Take an array of attribute names and values and normalize or discard
708 * illegal values for the given whitelist.
710 * - Discards attributes not the given whitelist
711 * - Unsafe style attributes are discarded
712 * - Invalid id attributes are re-encoded
714 * @param array $attribs
715 * @param array $whitelist List of allowed attribute names
716 * @return array
718 * @todo Check for legal values where the DTD limits things.
719 * @todo Check for unique id attribute :P
721 static function validateAttributes( $attribs, $whitelist ) {
722 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
724 $whitelist = array_flip( $whitelist );
725 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
727 $out = array();
728 foreach ( $attribs as $attribute => $value ) {
729 #allow XML namespace declaration if RDFa is enabled
730 if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
731 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
732 $out[$attribute] = $value;
735 continue;
738 # Allow any attribute beginning with "data-"
739 if ( !preg_match( '/^data-/i', $attribute ) && !isset( $whitelist[$attribute] ) ) {
740 continue;
743 # Strip javascript "expression" from stylesheets.
744 # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
745 if ( $attribute == 'style' ) {
746 $value = Sanitizer::checkCss( $value );
749 if ( $attribute === 'id' ) {
750 $value = Sanitizer::escapeId( $value, 'noninitial' );
753 # WAI-ARIA
754 # http://www.w3.org/TR/wai-aria/
755 # http://www.whatwg.org/html/elements.html#wai-aria
756 # For now we only support role="presentation" until we work out what roles should be
757 # usable by content and we ensure that our code explicitly rejects patterns that
758 # violate HTML5's ARIA restrictions.
759 if ( $attribute === 'role' && $value !== 'presentation' ) {
760 continue;
763 // RDFa and microdata properties allow URLs, URIs and/or CURIs.
764 // Check them for sanity.
765 if ( $attribute === 'rel' || $attribute === 'rev'
766 # RDFa
767 || $attribute === 'about' || $attribute === 'property'
768 || $attribute === 'resource' || $attribute === 'datatype'
769 || $attribute === 'typeof'
770 # HTML5 microdata
771 || $attribute === 'itemid' || $attribute === 'itemprop'
772 || $attribute === 'itemref' || $attribute === 'itemscope'
773 || $attribute === 'itemtype'
775 //Paranoia. Allow "simple" values but suppress javascript
776 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
777 continue;
781 # NOTE: even though elements using href/src are not allowed directly, supply
782 # validation code that can be used by tag hook handlers, etc
783 if ( $attribute === 'href' || $attribute === 'src' ) {
784 if ( !preg_match( $hrefExp, $value ) ) {
785 continue; //drop any href or src attributes not using an allowed protocol.
786 // NOTE: this also drops all relative URLs
790 // If this attribute was previously set, override it.
791 // Output should only have one attribute of each name.
792 $out[$attribute] = $value;
795 if ( $wgAllowMicrodataAttributes ) {
796 # itemtype, itemid, itemref don't make sense without itemscope
797 if ( !array_key_exists( 'itemscope', $out ) ) {
798 unset( $out['itemtype'] );
799 unset( $out['itemid'] );
800 unset( $out['itemref'] );
802 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
804 return $out;
808 * Merge two sets of HTML attributes. Conflicting items in the second set
809 * will override those in the first, except for 'class' attributes which
810 * will be combined (if they're both strings).
812 * @todo implement merging for other attributes such as style
813 * @param array $a
814 * @param array $b
815 * @return array
817 static function mergeAttributes( $a, $b ) {
818 $out = array_merge( $a, $b );
819 if ( isset( $a['class'] ) && isset( $b['class'] )
820 && is_string( $a['class'] ) && is_string( $b['class'] )
821 && $a['class'] !== $b['class']
823 $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
824 -1, PREG_SPLIT_NO_EMPTY );
825 $out['class'] = implode( ' ', array_unique( $classes ) );
827 return $out;
831 * Normalize CSS into a format we can easily search for hostile input
832 * - decode character references
833 * - decode escape sequences
834 * - convert characters that IE6 interprets into ascii
835 * - remove comments, unless the entire value is one single comment
836 * @param string $value the css string
837 * @return string normalized css
839 public static function normalizeCss( $value ) {
841 // Decode character references like &#123;
842 $value = Sanitizer::decodeCharReferences( $value );
844 // Decode escape sequences and line continuation
845 // See the grammar in the CSS 2 spec, appendix D.
846 // This has to be done AFTER decoding character references.
847 // This means it isn't possible for this function to return
848 // unsanitized escape sequences. It is possible to manufacture
849 // input that contains character references that decode to
850 // escape sequences that decode to character references, but
851 // it's OK for the return value to contain character references
852 // because the caller is supposed to escape those anyway.
853 static $decodeRegex;
854 if ( !$decodeRegex ) {
855 $space = '[\\x20\\t\\r\\n\\f]';
856 $nl = '(?:\\n|\\r\\n|\\r|\\f)';
857 $backslash = '\\\\';
858 $decodeRegex = "/ $backslash
860 ($nl) | # 1. Line continuation
861 ([0-9A-Fa-f]{1,6})$space? | # 2. character number
862 (.) | # 3. backslash cancelling special meaning
863 () | # 4. backslash at end of string
864 )/xu";
866 $value = preg_replace_callback( $decodeRegex,
867 array( __CLASS__, 'cssDecodeCallback' ), $value );
869 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
870 $value = preg_replace_callback(
871 '/[!-[]-z]/u', // U+FF01 to U+FF5A, excluding U+FF3C (bug 58088)
872 function ( $matches ) {
873 $cp = utf8ToCodepoint( $matches[0] );
874 if ( $cp === false ) {
875 return '';
877 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
879 $value
882 // Convert more characters IE6 might treat as ascii
883 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
884 $value = str_replace(
885 array( 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ),
886 array( 'r', 'n', 'n', 'l', 'i', '(', '(' ),
887 $value
890 // Let the value through if it's nothing but a single comment, to
891 // allow other functions which may reject it to pass some error
892 // message through.
893 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
894 // Remove any comments; IE gets token splitting wrong
895 // This must be done AFTER decoding character references and
896 // escape sequences, because those steps can introduce comments
897 // This step cannot introduce character references or escape
898 // sequences, because it replaces comments with spaces rather
899 // than removing them completely.
900 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
902 // Remove anything after a comment-start token, to guard against
903 // incorrect client implementations.
904 $commentPos = strpos( $value, '/*' );
905 if ( $commentPos !== false ) {
906 $value = substr( $value, 0, $commentPos );
910 // S followed by repeat, iteration, or prolonged sound marks,
911 // which IE will treat as "ss"
912 $value = preg_replace(
913 '/s(?:
914 \xE3\x80\xB1 | # U+3031
915 \xE3\x82\x9D | # U+309D
916 \xE3\x83\xBC | # U+30FC
917 \xE3\x83\xBD | # U+30FD
918 \xEF\xB9\xBC | # U+FE7C
919 \xEF\xB9\xBD | # U+FE7D
920 \xEF\xBD\xB0 # U+FF70
921 )/ix',
922 'ss',
923 $value
926 return $value;
931 * Pick apart some CSS and check it for forbidden or unsafe structures.
932 * Returns a sanitized string. This sanitized string will have
933 * character references and escape sequences decoded and comments
934 * stripped (unless it is itself one valid comment, in which case the value
935 * will be passed through). If the input is just too evil, only a comment
936 * complaining about evilness will be returned.
938 * Currently URL references, 'expression', 'tps' are forbidden.
940 * NOTE: Despite the fact that character references are decoded, the
941 * returned string may contain character references given certain
942 * clever input strings. These character references must
943 * be escaped before the return value is embedded in HTML.
945 * @param string $value
946 * @return string
948 static function checkCss( $value ) {
949 $value = self::normalizeCss( $value );
951 // Reject problematic keywords and control characters
952 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ) {
953 return '/* invalid control char */';
954 } elseif ( preg_match(
955 '! expression
956 | filter\s*:
957 | accelerator\s*:
958 | -o-link\s*:
959 | -o-link-source\s*:
960 | -o-replace\s*:
961 | url\s*\(
962 | image\s*\(
963 | image-set\s*\(
964 !ix', $value ) ) {
965 return '/* insecure input */';
967 return $value;
971 * @param array $matches
972 * @return string
974 static function cssDecodeCallback( $matches ) {
975 if ( $matches[1] !== '' ) {
976 // Line continuation
977 return '';
978 } elseif ( $matches[2] !== '' ) {
979 $char = codepointToUtf8( hexdec( $matches[2] ) );
980 } elseif ( $matches[3] !== '' ) {
981 $char = $matches[3];
982 } else {
983 $char = '\\';
985 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
986 // These characters need to be escaped in strings
987 // Clean up the escape sequence to avoid parsing errors by clients
988 return '\\' . dechex( ord( $char ) ) . ' ';
989 } else {
990 // Decode unnecessary escape
991 return $char;
996 * Take a tag soup fragment listing an HTML element's attributes
997 * and normalize it to well-formed XML, discarding unwanted attributes.
998 * Output is safe for further wikitext processing, with escaping of
999 * values that could trigger problems.
1001 * - Normalizes attribute names to lowercase
1002 * - Discards attributes not on a whitelist for the given element
1003 * - Turns broken or invalid entities into plaintext
1004 * - Double-quotes all attribute values
1005 * - Attributes without values are given the name as attribute
1006 * - Double attributes are discarded
1007 * - Unsafe style attributes are discarded
1008 * - Prepends space if there are attributes.
1010 * @param string $text
1011 * @param string $element
1012 * @return string
1014 static function fixTagAttributes( $text, $element ) {
1015 if ( trim( $text ) == '' ) {
1016 return '';
1019 $decoded = Sanitizer::decodeTagAttributes( $text );
1020 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1022 return Sanitizer::safeEncodeTagAttributes( $stripped );
1026 * Encode an attribute value for HTML output.
1027 * @param string $text
1028 * @return string HTML-encoded text fragment
1030 static function encodeAttribute( $text ) {
1031 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1033 // Whitespace is normalized during attribute decoding,
1034 // so if we've been passed non-spaces we must encode them
1035 // ahead of time or they won't be preserved.
1036 $encValue = strtr( $encValue, array(
1037 "\n" => '&#10;',
1038 "\r" => '&#13;',
1039 "\t" => '&#9;',
1040 ) );
1042 return $encValue;
1046 * Encode an attribute value for HTML tags, with extra armoring
1047 * against further wiki processing.
1048 * @param string $text
1049 * @return string HTML-encoded text fragment
1051 static function safeEncodeAttribute( $text ) {
1052 $encValue = Sanitizer::encodeAttribute( $text );
1054 # Templates and links may be expanded in later parsing,
1055 # creating invalid or dangerous output. Suppress this.
1056 $encValue = strtr( $encValue, array(
1057 '<' => '&lt;', // This should never happen,
1058 '>' => '&gt;', // we've received invalid input
1059 '"' => '&quot;', // which should have been escaped.
1060 '{' => '&#123;',
1061 '[' => '&#91;',
1062 "''" => '&#39;&#39;',
1063 'ISBN' => '&#73;SBN',
1064 'RFC' => '&#82;FC',
1065 'PMID' => '&#80;MID',
1066 '|' => '&#124;',
1067 '__' => '&#95;_',
1068 ) );
1070 # Stupid hack
1071 $encValue = preg_replace_callback(
1072 '/((?i)' . wfUrlProtocols() . ')/',
1073 array( 'Sanitizer', 'armorLinksCallback' ),
1074 $encValue );
1075 return $encValue;
1079 * Given a value, escape it so that it can be used in an id attribute and
1080 * return it. This will use HTML5 validation if $wgExperimentalHtmlIds is
1081 * true, allowing anything but ASCII whitespace. Otherwise it will use
1082 * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1083 * escaped with lots of dots.
1085 * To ensure we don't have to bother escaping anything, we also strip ', ",
1086 * & even if $wgExperimentalIds is true. TODO: Is this the best tactic?
1087 * We also strip # because it upsets IE, and % because it could be
1088 * ambiguous if it's part of something that looks like a percent escape
1089 * (which don't work reliably in fragments cross-browser).
1091 * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1092 * in the id and name attributes
1093 * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1094 * the id attribute
1095 * @see http://www.whatwg.org/html/elements.html#the-id-attribute
1096 * HTML5 definition of id attribute
1098 * @param string $id Id to escape
1099 * @param string|array $options String or array of strings (default is array()):
1100 * 'noninitial': This is a non-initial fragment of an id, not a full id,
1101 * so don't pay attention if the first character isn't valid at the
1102 * beginning of an id. Only matters if $wgExperimentalHtmlIds is
1103 * false.
1104 * 'legacy': Behave the way the old HTML 4-based ID escaping worked even
1105 * if $wgExperimentalHtmlIds is used, so we can generate extra
1106 * anchors and links won't break.
1107 * @return string
1109 static function escapeId( $id, $options = array() ) {
1110 global $wgExperimentalHtmlIds;
1111 $options = (array)$options;
1113 $id = Sanitizer::decodeCharReferences( $id );
1115 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1116 $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1117 $id = trim( $id, '_' );
1118 if ( $id === '' ) {
1119 // Must have been all whitespace to start with.
1120 return '_';
1121 } else {
1122 return $id;
1126 // HTML4-style escaping
1127 static $replace = array(
1128 '%3A' => ':',
1129 '%' => '.'
1132 $id = urlencode( strtr( $id, ' ', '_' ) );
1133 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1135 if ( !preg_match( '/^[a-zA-Z]/', $id ) && !in_array( 'noninitial', $options ) ) {
1136 // Initial character must be a letter!
1137 $id = "x$id";
1139 return $id;
1143 * Given a value, escape it so that it can be used as a CSS class and
1144 * return it.
1146 * @todo For extra validity, input should be validated UTF-8.
1148 * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1150 * @param string $class
1151 * @return string
1153 static function escapeClass( $class ) {
1154 // Convert ugly stuff to underscores and kill underscores in ugly places
1155 return rtrim( preg_replace(
1156 array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1157 '_',
1158 $class ), '_' );
1162 * Given HTML input, escape with htmlspecialchars but un-escape entities.
1163 * This allows (generally harmless) entities like &#160; to survive.
1165 * @param string $html HTML to escape
1166 * @return string Escaped input
1168 static function escapeHtmlAllowEntities( $html ) {
1169 $html = Sanitizer::decodeCharReferences( $html );
1170 # It seems wise to escape ' as well as ", as a matter of course. Can't
1171 # hurt.
1172 $html = htmlspecialchars( $html, ENT_QUOTES );
1173 return $html;
1177 * Regex replace callback for armoring links against further processing.
1178 * @param array $matches
1179 * @return string
1181 private static function armorLinksCallback( $matches ) {
1182 return str_replace( ':', '&#58;', $matches[1] );
1186 * Return an associative array of attribute names and values from
1187 * a partial tag string. Attribute names are forces to lowercase,
1188 * character references are decoded to UTF-8 text.
1190 * @param string $text
1191 * @return array
1193 public static function decodeTagAttributes( $text ) {
1194 if ( trim( $text ) == '' ) {
1195 return array();
1198 $attribs = array();
1199 $pairs = array();
1200 if ( !preg_match_all(
1201 self::getAttribsRegex(),
1202 $text,
1203 $pairs,
1204 PREG_SET_ORDER ) ) {
1205 return $attribs;
1208 foreach ( $pairs as $set ) {
1209 $attribute = strtolower( $set[1] );
1210 $value = Sanitizer::getTagAttributeCallback( $set );
1212 // Normalize whitespace
1213 $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1214 $value = trim( $value );
1216 // Decode character references
1217 $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1219 return $attribs;
1223 * Build a partial tag string from an associative array of attribute
1224 * names and values as returned by decodeTagAttributes.
1226 * @param array $assoc_array
1227 * @return string
1229 public static function safeEncodeTagAttributes( $assoc_array ) {
1230 $attribs = array();
1231 foreach ( $assoc_array as $attribute => $value ) {
1232 $encAttribute = htmlspecialchars( $attribute );
1233 $encValue = Sanitizer::safeEncodeAttribute( $value );
1235 $attribs[] = "$encAttribute=\"$encValue\"";
1237 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1241 * Pick the appropriate attribute value from a match set from the
1242 * attribs regex matches.
1244 * @param array $set
1245 * @throws MWException When tag conditions are not met.
1246 * @return string
1248 private static function getTagAttributeCallback( $set ) {
1249 if ( isset( $set[6] ) ) {
1250 # Illegal #XXXXXX color with no quotes.
1251 return $set[6];
1252 } elseif ( isset( $set[5] ) ) {
1253 # No quotes.
1254 return $set[5];
1255 } elseif ( isset( $set[4] ) ) {
1256 # Single-quoted
1257 return $set[4];
1258 } elseif ( isset( $set[3] ) ) {
1259 # Double-quoted
1260 return $set[3];
1261 } elseif ( !isset( $set[2] ) ) {
1262 # In XHTML, attributes must have a value.
1263 # For 'reduced' form, return explicitly the attribute name here.
1264 return $set[1];
1265 } else {
1266 throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1271 * Normalize whitespace and character references in an XML source-
1272 * encoded text for an attribute value.
1274 * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1275 * but note that we're not returning the value, but are returning
1276 * XML source fragments that will be slapped into output.
1278 * @param string $text
1279 * @return string
1280 * @todo Remove, unused?
1282 private static function normalizeAttributeValue( $text ) {
1283 return str_replace( '"', '&quot;',
1284 self::normalizeWhitespace(
1285 Sanitizer::normalizeCharReferences( $text ) ) );
1289 * @param string $text
1290 * @return string
1292 private static function normalizeWhitespace( $text ) {
1293 return preg_replace(
1294 '/\r\n|[\x20\x0d\x0a\x09]/',
1295 ' ',
1296 $text );
1300 * Normalizes whitespace in a section name, such as might be returned
1301 * by Parser::stripSectionName(), for use in the id's that are used for
1302 * section links.
1304 * @param string $section
1305 * @return string
1307 static function normalizeSectionNameWhitespace( $section ) {
1308 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1312 * Ensure that any entities and character references are legal
1313 * for XML and XHTML specifically. Any stray bits will be
1314 * &amp;-escaped to result in a valid text fragment.
1316 * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1317 * numericized (this way we're well-formed even without a DTD)
1318 * b. any numeric char refs must be legal chars, not invalid or forbidden
1319 * c. use lower cased "&#x", not "&#X"
1320 * d. fix or reject non-valid attributes
1322 * @param string $text
1323 * @return string
1324 * @private
1326 static function normalizeCharReferences( $text ) {
1327 return preg_replace_callback(
1328 self::CHAR_REFS_REGEX,
1329 array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1330 $text );
1334 * @param string $matches
1335 * @return string
1337 static function normalizeCharReferencesCallback( $matches ) {
1338 $ret = null;
1339 if ( $matches[1] != '' ) {
1340 $ret = Sanitizer::normalizeEntity( $matches[1] );
1341 } elseif ( $matches[2] != '' ) {
1342 $ret = Sanitizer::decCharReference( $matches[2] );
1343 } elseif ( $matches[3] != '' ) {
1344 $ret = Sanitizer::hexCharReference( $matches[3] );
1346 if ( is_null( $ret ) ) {
1347 return htmlspecialchars( $matches[0] );
1348 } else {
1349 return $ret;
1354 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1355 * return the equivalent numeric entity reference (except for the core &lt;
1356 * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1357 * the HTML equivalent. Otherwise, returns HTML-escaped text of
1358 * pseudo-entity source (eg &amp;foo;)
1360 * @param string $name
1361 * @return string
1363 static function normalizeEntity( $name ) {
1364 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1365 return '&' . self::$htmlEntityAliases[$name] . ';';
1366 } elseif ( in_array( $name, array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1367 return "&$name;";
1368 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1369 return '&#' . self::$htmlEntities[$name] . ';';
1370 } else {
1371 return "&amp;$name;";
1376 * @param int $codepoint
1377 * @return null|string
1379 static function decCharReference( $codepoint ) {
1380 $point = intval( $codepoint );
1381 if ( Sanitizer::validateCodepoint( $point ) ) {
1382 return sprintf( '&#%d;', $point );
1383 } else {
1384 return null;
1389 * @param int $codepoint
1390 * @return null|string
1392 static function hexCharReference( $codepoint ) {
1393 $point = hexdec( $codepoint );
1394 if ( Sanitizer::validateCodepoint( $point ) ) {
1395 return sprintf( '&#x%x;', $point );
1396 } else {
1397 return null;
1402 * Returns true if a given Unicode codepoint is a valid character in XML.
1403 * @param int $codepoint
1404 * @return bool
1406 private static function validateCodepoint( $codepoint ) {
1407 return $codepoint == 0x09
1408 || $codepoint == 0x0a
1409 || $codepoint == 0x0d
1410 || ( $codepoint >= 0x20 && $codepoint <= 0xd7ff )
1411 || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1412 || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1416 * Decode any character references, numeric or named entities,
1417 * in the text and return a UTF-8 string.
1419 * @param string $text
1420 * @return string
1422 public static function decodeCharReferences( $text ) {
1423 return preg_replace_callback(
1424 self::CHAR_REFS_REGEX,
1425 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1426 $text );
1430 * Decode any character references, numeric or named entities,
1431 * in the next and normalize the resulting string. (bug 14952)
1433 * This is useful for page titles, not for text to be displayed,
1434 * MediaWiki allows HTML entities to escape normalization as a feature.
1436 * @param string $text Already normalized, containing entities
1437 * @return string Still normalized, without entities
1439 public static function decodeCharReferencesAndNormalize( $text ) {
1440 global $wgContLang;
1441 $text = preg_replace_callback(
1442 self::CHAR_REFS_REGEX,
1443 array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1444 $text, /* limit */ -1, $count );
1446 if ( $count ) {
1447 return $wgContLang->normalize( $text );
1448 } else {
1449 return $text;
1454 * @param string $matches
1455 * @return string
1457 static function decodeCharReferencesCallback( $matches ) {
1458 if ( $matches[1] != '' ) {
1459 return Sanitizer::decodeEntity( $matches[1] );
1460 } elseif ( $matches[2] != '' ) {
1461 return Sanitizer::decodeChar( intval( $matches[2] ) );
1462 } elseif ( $matches[3] != '' ) {
1463 return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1465 # Last case should be an ampersand by itself
1466 return $matches[0];
1470 * Return UTF-8 string for a codepoint if that is a valid
1471 * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1472 * @param int $codepoint
1473 * @return string
1474 * @private
1476 static function decodeChar( $codepoint ) {
1477 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1478 return codepointToUtf8( $codepoint );
1479 } else {
1480 return UTF8_REPLACEMENT;
1485 * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1486 * return the UTF-8 encoding of that character. Otherwise, returns
1487 * pseudo-entity source (eg "&foo;")
1489 * @param string $name
1490 * @return string
1492 static function decodeEntity( $name ) {
1493 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1494 $name = self::$htmlEntityAliases[$name];
1496 if ( isset( self::$htmlEntities[$name] ) ) {
1497 return codepointToUtf8( self::$htmlEntities[$name] );
1498 } else {
1499 return "&$name;";
1504 * Fetch the whitelist of acceptable attributes for a given element name.
1506 * @param string $element
1507 * @return array
1509 static function attributeWhitelist( $element ) {
1510 $list = Sanitizer::setupAttributeWhitelist();
1511 return isset( $list[$element] )
1512 ? $list[$element]
1513 : array();
1517 * Foreach array key (an allowed HTML element), return an array
1518 * of allowed attributes
1519 * @return array
1521 static function setupAttributeWhitelist() {
1522 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
1523 static $whitelist, $staticInitialised;
1525 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgAllowMicrodataAttributes' ) );
1527 if ( $whitelist !== null && $staticInitialised == $globalContext ) {
1528 return $whitelist;
1531 $common = array(
1532 # HTML
1533 'id',
1534 'class',
1535 'style',
1536 'lang',
1537 'dir',
1538 'title',
1540 # WAI-ARIA
1541 'role',
1544 if ( $wgAllowRdfaAttributes ) {
1545 # RDFa attributes as specified in section 9 of
1546 # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1547 $common = array_merge( $common, array(
1548 'about', 'property', 'resource', 'datatype', 'typeof',
1549 ) );
1552 if ( $wgAllowMicrodataAttributes ) {
1553 # add HTML5 microdata tags as specified by
1554 # http://www.whatwg.org/html/microdata.html#the-microdata-model
1555 $common = array_merge( $common, array(
1556 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1557 ) );
1560 $block = array_merge( $common, array( 'align' ) );
1561 $tablealign = array( 'align', 'valign' );
1562 $tablecell = array(
1563 'abbr',
1564 'axis',
1565 'headers',
1566 'scope',
1567 'rowspan',
1568 'colspan',
1569 'nowrap', # deprecated
1570 'width', # deprecated
1571 'height', # deprecated
1572 'bgcolor', # deprecated
1575 # Numbers refer to sections in HTML 4.01 standard describing the element.
1576 # See: http://www.w3.org/TR/html4/
1577 $whitelist = array(
1578 # 7.5.4
1579 'div' => $block,
1580 'center' => $common, # deprecated
1581 'span' => $common,
1583 # 7.5.5
1584 'h1' => $block,
1585 'h2' => $block,
1586 'h3' => $block,
1587 'h4' => $block,
1588 'h5' => $block,
1589 'h6' => $block,
1591 # 7.5.6
1592 # address
1594 # 8.2.4
1595 'bdo' => $common,
1597 # 9.2.1
1598 'em' => $common,
1599 'strong' => $common,
1600 'cite' => $common,
1601 'dfn' => $common,
1602 'code' => $common,
1603 'samp' => $common,
1604 'kbd' => $common,
1605 'var' => $common,
1606 'abbr' => $common,
1607 # acronym
1609 # 9.2.2
1610 'blockquote' => array_merge( $common, array( 'cite' ) ),
1611 'q' => array_merge( $common, array( 'cite' ) ),
1613 # 9.2.3
1614 'sub' => $common,
1615 'sup' => $common,
1617 # 9.3.1
1618 'p' => $block,
1620 # 9.3.2
1621 'br' => array_merge( $common, array( 'clear' ) ),
1623 # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1624 'wbr' => $common,
1626 # 9.3.4
1627 'pre' => array_merge( $common, array( 'width' ) ),
1629 # 9.4
1630 'ins' => array_merge( $common, array( 'cite', 'datetime' ) ),
1631 'del' => array_merge( $common, array( 'cite', 'datetime' ) ),
1633 # 10.2
1634 'ul' => array_merge( $common, array( 'type' ) ),
1635 'ol' => array_merge( $common, array( 'type', 'start' ) ),
1636 'li' => array_merge( $common, array( 'type', 'value' ) ),
1638 # 10.3
1639 'dl' => $common,
1640 'dd' => $common,
1641 'dt' => $common,
1643 # 11.2.1
1644 'table' => array_merge( $common,
1645 array( 'summary', 'width', 'border', 'frame',
1646 'rules', 'cellspacing', 'cellpadding',
1647 'align', 'bgcolor',
1648 ) ),
1650 # 11.2.2
1651 'caption' => $block,
1653 # 11.2.3
1654 'thead' => $common,
1655 'tfoot' => $common,
1656 'tbody' => $common,
1658 # 11.2.4
1659 'colgroup' => array_merge( $common, array( 'span' ) ),
1660 'col' => array_merge( $common, array( 'span' ) ),
1662 # 11.2.5
1663 'tr' => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1665 # 11.2.6
1666 'td' => array_merge( $common, $tablecell, $tablealign ),
1667 'th' => array_merge( $common, $tablecell, $tablealign ),
1669 # 12.2
1670 # NOTE: <a> is not allowed directly, but the attrib
1671 # whitelist is used from the Parser object
1672 'a' => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1674 # 13.2
1675 # Not usually allowed, but may be used for extension-style hooks
1676 # such as <math> when it is rasterized, or if $wgAllowImageTag is
1677 # true
1678 'img' => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1680 # 15.2.1
1681 'tt' => $common,
1682 'b' => $common,
1683 'i' => $common,
1684 'big' => $common,
1685 'small' => $common,
1686 'strike' => $common,
1687 's' => $common,
1688 'u' => $common,
1690 # 15.2.2
1691 'font' => array_merge( $common, array( 'size', 'color', 'face' ) ),
1692 # basefont
1694 # 15.3
1695 'hr' => array_merge( $common, array( 'width' ) ),
1697 # HTML Ruby annotation text module, simple ruby only.
1698 # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
1699 'ruby' => $common,
1700 # rbc
1701 'rb' => $common,
1702 'rp' => $common,
1703 'rt' => $common, #array_merge( $common, array( 'rbspan' ) ),
1704 'rtc' => $common,
1706 # MathML root element, where used for extensions
1707 # 'title' may not be 100% valid here; it's XHTML
1708 # http://www.w3.org/TR/REC-MathML/
1709 'math' => array( 'class', 'style', 'id', 'title' ),
1711 # HTML 5 section 4.6
1712 'bdi' => $common,
1714 # HTML5 elements, defined by:
1715 # http://www.whatwg.org/html/
1716 'data' => array_merge( $common, array( 'value' ) ),
1717 'time' => array_merge( $common, array( 'datetime' ) ),
1718 'mark' => $common,
1720 // meta and link are only permitted by removeHTMLtags when Microdata
1721 // is enabled so we don't bother adding a conditional to hide these
1722 // Also meta and link are only valid in WikiText as Microdata elements
1723 // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1724 // So we don't bother including $common attributes that have no purpose.
1725 'meta' => array( 'itemprop', 'content' ),
1726 'link' => array( 'itemprop', 'href' ),
1729 $staticInitialised = $globalContext;
1731 return $whitelist;
1735 * Take a fragment of (potentially invalid) HTML and return
1736 * a version with any tags removed, encoded as plain text.
1738 * Warning: this return value must be further escaped for literal
1739 * inclusion in HTML output as of 1.10!
1741 * @param string $text HTML fragment
1742 * @return string
1744 static function stripAllTags( $text ) {
1745 # Actual <tags>
1746 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1748 # Normalize &entities and whitespace
1749 $text = self::decodeCharReferences( $text );
1750 $text = self::normalizeWhitespace( $text );
1752 return $text;
1756 * Hack up a private DOCTYPE with HTML's standard entity declarations.
1757 * PHP 4 seemed to know these if you gave it an HTML doctype, but
1758 * PHP 5.1 doesn't.
1760 * Use for passing XHTML fragments to PHP's XML parsing functions
1762 * @return string
1764 static function hackDocType() {
1765 $out = "<!DOCTYPE html [\n";
1766 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1767 $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1769 $out .= "]>\n";
1770 return $out;
1774 * @param string $url
1775 * @return mixed|string
1777 static function cleanUrl( $url ) {
1778 # Normalize any HTML entities in input. They will be
1779 # re-escaped by makeExternalLink().
1780 $url = Sanitizer::decodeCharReferences( $url );
1782 # Escape any control characters introduced by the above step
1783 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1784 array( __CLASS__, 'cleanUrlCallback' ), $url );
1786 # Validate hostname portion
1787 $matches = array();
1788 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1789 list( /* $whole */, $protocol, $host, $rest ) = $matches;
1791 // Characters that will be ignored in IDNs.
1792 // http://tools.ietf.org/html/3454#section-3.1
1793 // Strip them before further processing so blacklists and such work.
1794 $strip = "/
1795 \\s| # general whitespace
1796 \xc2\xad| # 00ad SOFT HYPHEN
1797 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1798 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1799 \xe2\x81\xa0| # 2060 WORD JOINER
1800 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1801 \xcd\x8f| # 034f COMBINING GRAPHEME JOINER
1802 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1803 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1804 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1805 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1806 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1807 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1808 /xuD";
1810 $host = preg_replace( $strip, '', $host );
1812 // @todo FIXME: Validate hostnames here
1814 return $protocol . $host . $rest;
1815 } else {
1816 return $url;
1821 * @param array $matches
1822 * @return string
1824 static function cleanUrlCallback( $matches ) {
1825 return urlencode( $matches[0] );
1829 * Does a string look like an e-mail address?
1831 * This validates an email address using an HTML5 specification found at:
1832 * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1833 * Which as of 2011-01-24 says:
1835 * A valid e-mail address is a string that matches the ABNF production
1836 * 1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1837 * in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1838 * 3.5.
1840 * This function is an implementation of the specification as requested in
1841 * bug 22449.
1843 * Client-side forms will use the same standard validation rules via JS or
1844 * HTML 5 validation; additional restrictions can be enforced server-side
1845 * by extensions via the 'isValidEmailAddr' hook.
1847 * Note that this validation doesn't 100% match RFC 2822, but is believed
1848 * to be liberal enough for wide use. Some invalid addresses will still
1849 * pass validation here.
1851 * @since 1.18
1853 * @param string $addr E-mail address
1854 * @return bool
1856 public static function validateEmail( $addr ) {
1857 $result = null;
1858 if ( !Hooks::run( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1859 return $result;
1862 // Please note strings below are enclosed in brackets [], this make the
1863 // hyphen "-" a range indicator. Hence it is double backslashed below.
1864 // See bug 26948
1865 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1866 $rfc1034_ldh_str = "a-z0-9\\-";
1868 $html5_email_regexp = "/
1869 ^ # start of string
1870 [$rfc5322_atext\\.]+ # user part which is liberal :p
1871 @ # 'apostrophe'
1872 [$rfc1034_ldh_str]+ # First domain part
1873 (\\.[$rfc1034_ldh_str]+)* # Following part prefixed with a dot
1874 $ # End of string
1875 /ix"; // case Insensitive, eXtended
1877 return (bool)preg_match( $html5_email_regexp, $addr );