includes/Sanitizer.php

   1 <?php
   2 /**
   3  * HTML sanitizer for %MediaWiki.
   4  *
   5  * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
   6  * https://www.mediawiki.org/
   7  *
   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.
  12  *
  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.
  17  *
  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
  22  *
  23  * @file
  24  * @ingroup Parser
  25  */
  26
  27 /**
  28  * HTML sanitizer for MediaWiki
  29  * @ingroup Parser
  30  */
  31 class Sanitizer {
  32         /**
  33          * Regular expression to match various types of character references in
  34          * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
  35          */
  36         const CHAR_REFS_REGEX =
  37                 '/&([A-Za-z0-9\x80-\xff]+);
  38                  |&\#([0-9]+);
  39                  |&\#[xX]([0-9A-Fa-f]+);
  40                  |(&)/x';
  41
  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
  49          */
  50         const EVIL_URI_PATTERN = '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
  51         const XMLNS_ATTRIBUTE_PATTERN = "/^xmlns:[:A-Z_a-z-.0-9]+$/";
  52
  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.
  57          */
  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
 312         );
 313
 314         /**
 315          * Character entity aliases accepted by MediaWiki
 316          */
 317         private static $htmlEntityAliases = array(
 318                 'רלמ' => 'rlm',
 319                 'رلم' => 'rlm',
 320         );
 321
 322         /**
 323          * Lazy-initialised attributes regex, see getAttribsRegex()
 324          */
 325         private static $attribsRegex;
 326
 327         /**
 328          * Regular expression to match HTML/XML attribute pairs within a tag.
 329          * Allows some... latitude.
 330          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 331          * @return string
 332          */
 333         static function getAttribsRegex() {
 334                 if ( self::$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*=$space*
 341                                         (?:
 342                                          # The attribute value: quoted or alone
 343                                           \"([^<\"]*)\"
 344                                          | '([^<']*)'
 345                                          |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
 346                                          |  (\#[0-9a-fA-F]+) # Technically wrong, but lots of
 347                                                                                  # colors are specified like this.
 348                                                                                  # We'll be normalizing it.
 349                                         )
 350                                 )?(?=$space|\$)/sx";
 351                 }
 352                 return self::$attribsRegex;
 353         }
 354
 355         /**
 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
 366          */
 367         static function removeHTMLtags( $text, $processCallback = null,
 368                 $args = array(), $extratags = array(), $removetags = array()
 369         ) {
 370                 global $wgUseTidy, $wgAllowMicrodataAttributes, $wgAllowImageTag;
 371
 372                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 373                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 374
 375                 wfProfileIn( __METHOD__ );
 376
 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 ) {
 381
 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'
 389                         );
 390                         $htmlsingle = array(
 391                                 'br', 'wbr', 'hr', 'li', 'dt', 'dd'
 392                         );
 393                         $htmlsingleonly = array( # Elements that cannot have close tags
 394                                 'br', 'wbr', 'hr'
 395                         );
 396                         if ( $wgAllowMicrodataAttributes ) {
 397                                 $htmlsingle[] = $htmlsingleonly[] = 'meta';
 398                                 $htmlsingle[] = $htmlsingleonly[] = 'link';
 399                         }
 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'
 404                         );
 405                         $tabletags = array( # Can only appear inside table, we will close them
 406                                 'td', 'th', 'tr',
 407                         );
 408                         $htmllist = array( # Tags used by list
 409                                 'ul', 'ol',
 410                         );
 411                         $listtags = array( # Tags that can appear in a list
 412                                 'li',
 413                         );
 414
 415                         if ( $wgAllowImageTag ) {
 416                                 $htmlsingle[] = 'img';
 417                                 $htmlsingleonly[] = 'img';
 418                         }
 419
 420                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 421                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 422
 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 );
 428                         }
 429                         $staticInitialised = $globalContext;
 430                 }
 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 );
 436
 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;
 454                                 }
 455
 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();
 466
 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();
 481                                                                 }
 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();
 493                                                                         }
 494                                                                 }
 495                                                         } else {
 496                                                                 wfSuppressWarnings();
 497                                                                 array_push( $tagstack, $ot );
 498                                                                 wfRestoreWarnings();
 499
 500                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 501                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 502                                                                         $badtag = true;
 503                                                                 }
 504                                                         }
 505                                                 } else {
 506                                                         if ( $t == 'table' ) {
 507                                                                 $tagstack = array_pop( $tablestack );
 508                                                         }
 509                                                 }
 510                                                 $newparams = '';
 511                                         } else {
 512                                                 # Keep track for later
 513                                                 if ( isset( $tabletags[$t] ) &&
 514                                                 !in_array( 'table', $tagstack ) ) {
 515                                                         $badtag = true;
 516                                                 } elseif ( in_array( $t, $tagstack ) &&
 517                                                 !isset( $htmlnest[$t] ) ) {
 518                                                         $badtag = true;
 519                                                 # Is it a self closed htmlpair ? (bug 5487)
 520                                                 } elseif ( $brace == '/>' &&
 521                                                 isset( $htmlpairs[$t] ) ) {
 522                                                         $badtag = true;
 523                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 524                                                         # Hack to force empty tag for unclosable elements
 525                                                         $brace = '/>';
 526                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 527                                                         # Hack to not close $htmlsingle tags
 528                                                         $brace = null;
 529                                                         # Still need to push this optionally-closed tag to
 530                                                         # the tag stack so that we can match end tags
 531                                                         # instead of marking them as bad.
 532                                                         array_push( $tagstack, $t );
 533                                                 } elseif ( isset( $tabletags[$t] )
 534                                                 && in_array( $t, $tagstack ) ) {
 535                                                         // New table tag but forgot to close the previous one
 536                                                         $text .= "</$t>";
 537                                                 } else {
 538                                                         if ( $t == 'table' ) {
 539                                                                 array_push( $tablestack, $tagstack );
 540                                                                 $tagstack = array();
 541                                                         }
 542                                                         array_push( $tagstack, $t );
 543                                                 }
 544
 545                                                 # Replace any variables or template parameters with
 546                                                 # plaintext results.
 547                                                 if ( is_callable( $processCallback ) ) {
 548                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 549                                                 }
 550
 551                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 552                                                         $badtag = true;
 553                                                 }
 554
 555                                                 # Strip non-approved attributes from the tag
 556                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 557                                         }
 558                                         if ( !$badtag ) {
 559                                                 $rest = str_replace( '>', '&gt;', $rest );
 560                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 561                                                 $text .= "<$slash$t$newparams$close>$rest";
 562                                                 continue;
 563                                         }
 564                                 }
 565                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 566                         }
 567                         # Close off any remaining tags
 568                         while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
 569                                 $text .= "</$t>\n";
 570                                 if ( $t == 'table' ) {
 571                                         $tagstack = array_pop( $tablestack );
 572                                 }
 573                         }
 574                 } else {
 575                         # this might be possible using tidy itself
 576                         foreach ( $bits as $x ) {
 577                                 preg_match(
 578                                         '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 579                                         $x,
 580                                         $regs
 581                                 );
 582
 583                                 wfSuppressWarnings();
 584                                 list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 585                                 wfRestoreWarnings();
 586
 587                                 $badtag = false;
 588                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 589                                         if ( is_callable( $processCallback ) ) {
 590                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 591                                         }
 592
 593                                         if ( !Sanitizer::validateTag( $params, $t ) ) {
 594                                                 $badtag = true;
 595                                         }
 596
 597                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 598                                         if ( !$badtag ) {
 599                                                 $rest = str_replace( '>', '&gt;', $rest );
 600                                                 $text .= "<$slash$t$newparams$brace$rest";
 601                                                 continue;
 602                                         }
 603                                 }
 604                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 605                         }
 606                 }
 607                 wfProfileOut( __METHOD__ );
 608                 return $text;
 609         }
 610
 611         /**
 612          * Remove '<!--', '-->', and everything between.
 613          * To avoid leaving blank lines, when a comment is both preceded
 614          * and followed by a newline (ignoring spaces), trim leading and
 615          * trailing spaces and one of the newlines.
 616          *
 617          * @private
 618          * @param string $text
 619          * @return string
 620          */
 621         static function removeHTMLcomments( $text ) {
 622                 wfProfileIn( __METHOD__ );
 623                 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
 624                         $end = strpos( $text, '-->', $start + 4 );
 625                         if ( $end === false ) {
 626                                 # Unterminated comment; bail out
 627                                 break;
 628                         }
 629
 630                         $end += 3;
 631
 632                         # Trim space and newline if the comment is both
 633                         # preceded and followed by a newline
 634                         $spaceStart = max( $start - 1, 0 );
 635                         $spaceLen = $end - $spaceStart;
 636                         while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
 637                                 $spaceStart--;
 638                                 $spaceLen++;
 639                         }
 640                         while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
 641                                 $spaceLen++;
 642                         }
 643                         if ( substr( $text, $spaceStart, 1 ) === "\n"
 644                                 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
 645                                 # Remove the comment, leading and trailing
 646                                 # spaces, and leave only one newline.
 647                                 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
 648                         } else {
 649                                 # Remove just the comment.
 650                                 $text = substr_replace( $text, '', $start, $end - $start );
 651                         }
 652                 }
 653                 wfProfileOut( __METHOD__ );
 654                 return $text;
 655         }
 656
 657         /**
 658          * Takes attribute names and values for a tag and the tag name and
 659          * validates that the tag is allowed to be present.
 660          * This DOES NOT validate the attributes, nor does it validate the
 661          * tags themselves. This method only handles the special circumstances
 662          * where we may want to allow a tag within content but ONLY when it has
 663          * specific attributes set.
 664          *
 665          * @param string $params
 666          * @param string $element
 667          * @return bool
 668          */
 669         static function validateTag( $params, $element ) {
 670                 $params = Sanitizer::decodeTagAttributes( $params );
 671
 672                 if ( $element == 'meta' || $element == 'link' ) {
 673                         if ( !isset( $params['itemprop'] ) ) {
 674                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 675                                 return false;
 676                         }
 677                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 678                                 // <meta> must have a content="" for the itemprop
 679                                 return false;
 680                         }
 681                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 682                                 // <link> must have an associated href=""
 683                                 return false;
 684                         }
 685                 }
 686
 687                 return true;
 688         }
 689
 690         /**
 691          * Take an array of attribute names and values and normalize or discard
 692          * illegal values for the given element type.
 693          *
 694          * - Discards attributes not on a whitelist for the given element
 695          * - Unsafe style attributes are discarded
 696          * - Invalid id attributes are re-encoded
 697          *
 698          * @param array $attribs
 699          * @param string $element
 700          * @return array
 701          *
 702          * @todo Check for legal values where the DTD limits things.
 703          * @todo Check for unique id attribute :P
 704          */
 705         static function validateTagAttributes( $attribs, $element ) {
 706                 return Sanitizer::validateAttributes( $attribs,
 707                         Sanitizer::attributeWhitelist( $element ) );
 708         }
 709
 710         /**
 711          * Take an array of attribute names and values and normalize or discard
 712          * illegal values for the given whitelist.
 713          *
 714          * - Discards attributes not the given whitelist
 715          * - Unsafe style attributes are discarded
 716          * - Invalid id attributes are re-encoded
 717          *
 718          * @param array $attribs
 719          * @param array $whitelist List of allowed attribute names
 720          * @return array
 721          *
 722          * @todo Check for legal values where the DTD limits things.
 723          * @todo Check for unique id attribute :P
 724          */
 725         static function validateAttributes( $attribs, $whitelist ) {
 726                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
 727
 728                 $whitelist = array_flip( $whitelist );
 729                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 730
 731                 $out = array();
 732                 foreach ( $attribs as $attribute => $value ) {
 733                         #allow XML namespace declaration if RDFa is enabled
 734                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 735                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 736                                         $out[$attribute] = $value;
 737                                 }
 738
 739                                 continue;
 740                         }
 741
 742                         # Allow any attribute beginning with "data-"
 743                         if ( !preg_match( '/^data-/i', $attribute ) && !isset( $whitelist[$attribute] ) ) {
 744                                 continue;
 745                         }
 746
 747                         # Strip javascript "expression" from stylesheets.
 748                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 749                         if ( $attribute == 'style' ) {
 750                                 $value = Sanitizer::checkCss( $value );
 751                         }
 752
 753                         if ( $attribute === 'id' ) {
 754                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 755                         }
 756
 757                         # WAI-ARIA
 758                         # http://www.w3.org/TR/wai-aria/
 759                         # http://www.whatwg.org/html/elements.html#wai-aria
 760                         # For now we only support role="presentation" until we work out what roles should be
 761                         # usable by content and we ensure that our code explicitly rejects patterns that
 762                         # violate HTML5's ARIA restrictions.
 763                         if ( $attribute === 'role' && $value !== 'presentation' ) {
 764                                 continue;
 765                         }
 766
 767                         // RDFa and microdata properties allow URLs, URIs and/or CURIs.
 768                         // Check them for sanity.
 769                         if ( $attribute === 'rel' || $attribute === 'rev'
 770                                 # RDFa
 771                                 || $attribute === 'about' || $attribute === 'property'
 772                                 || $attribute === 'resource' || $attribute === 'datatype'
 773                                 || $attribute === 'typeof'
 774                                 # HTML5 microdata
 775                                 || $attribute === 'itemid' || $attribute === 'itemprop'
 776                                 || $attribute === 'itemref' || $attribute === 'itemscope'
 777                                 || $attribute === 'itemtype'
 778                         ) {
 779                                 //Paranoia. Allow "simple" values but suppress javascript
 780                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 781                                         continue;
 782                                 }
 783                         }
 784
 785                         # NOTE: even though elements using href/src are not allowed directly, supply
 786                         #       validation code that can be used by tag hook handlers, etc
 787                         if ( $attribute === 'href' || $attribute === 'src' ) {
 788                                 if ( !preg_match( $hrefExp, $value ) ) {
 789                                         continue; //drop any href or src attributes not using an allowed protocol.
 790                                         // NOTE: this also drops all relative URLs
 791                                 }
 792                         }
 793
 794                         // If this attribute was previously set, override it.
 795                         // Output should only have one attribute of each name.
 796                         $out[$attribute] = $value;
 797                 }
 798
 799                 if ( $wgAllowMicrodataAttributes ) {
 800                         # itemtype, itemid, itemref don't make sense without itemscope
 801                         if ( !array_key_exists( 'itemscope', $out ) ) {
 802                                 unset( $out['itemtype'] );
 803                                 unset( $out['itemid'] );
 804                                 unset( $out['itemref'] );
 805                         }
 806                         # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 807                 }
 808                 return $out;
 809         }
 810
 811         /**
 812          * Merge two sets of HTML attributes.  Conflicting items in the second set
 813          * will override those in the first, except for 'class' attributes which
 814          * will be combined (if they're both strings).
 815          *
 816          * @todo implement merging for other attributes such as style
 817          * @param array $a
 818          * @param array $b
 819          * @return array
 820          */
 821         static function mergeAttributes( $a, $b ) {
 822                 $out = array_merge( $a, $b );
 823                 if ( isset( $a['class'] ) && isset( $b['class'] )
 824                         && is_string( $a['class'] ) && is_string( $b['class'] )
 825                         && $a['class'] !== $b['class']
 826                 ) {
 827                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 828                                 -1, PREG_SPLIT_NO_EMPTY );
 829                         $out['class'] = implode( ' ', array_unique( $classes ) );
 830                 }
 831                 return $out;
 832         }
 833
 834         /**
 835          * Pick apart some CSS and check it for forbidden or unsafe structures.
 836          * Returns a sanitized string. This sanitized string will have
 837          * character references and escape sequences decoded and comments
 838          * stripped (unless it is itself one valid comment, in which case the value
 839          * will be passed through). If the input is just too evil, only a comment
 840          * complaining about evilness will be returned.
 841          *
 842          * Currently URL references, 'expression', 'tps' are forbidden.
 843          *
 844          * NOTE: Despite the fact that character references are decoded, the
 845          * returned string may contain character references given certain
 846          * clever input strings. These character references must
 847          * be escaped before the return value is embedded in HTML.
 848          *
 849          * @param string $value
 850          * @return string
 851          */
 852         static function checkCss( $value ) {
 853                 // Decode character references like &#123;
 854                 $value = Sanitizer::decodeCharReferences( $value );
 855
 856                 // Decode escape sequences and line continuation
 857                 // See the grammar in the CSS 2 spec, appendix D.
 858                 // This has to be done AFTER decoding character references.
 859                 // This means it isn't possible for this function to return
 860                 // unsanitized escape sequences. It is possible to manufacture
 861                 // input that contains character references that decode to
 862                 // escape sequences that decode to character references, but
 863                 // it's OK for the return value to contain character references
 864                 // because the caller is supposed to escape those anyway.
 865                 static $decodeRegex;
 866                 if ( !$decodeRegex ) {
 867                         $space = '[\\x20\\t\\r\\n\\f]';
 868                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 869                         $backslash = '\\\\';
 870                         $decodeRegex = "/ $backslash
 871                                 (?:
 872                                         ($nl) |  # 1. Line continuation
 873                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 874                                         (.) | # 3. backslash cancelling special meaning
 875                                         () | # 4. backslash at end of string
 876                                 )/xu";
 877                 }
 878                 $value = preg_replace_callback( $decodeRegex,
 879                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 880
 881                 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
 882                 $value = preg_replace_callback(
 883                         '/[！-［］-ｚ]/u', // U+FF01 to U+FF5A, excluding U+FF3C (bug 58088)
 884                         function ( $matches ) {
 885                                 $cp = utf8ToCodepoint( $matches[0] );
 886                                 if ( $cp === false ) {
 887                                         return '';
 888                                 }
 889                                 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
 890                         },
 891                         $value
 892                 );
 893
 894                 // Convert more characters IE6 might treat as ascii
 895                 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
 896                 $value = str_replace(
 897                         array( 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ),
 898                         array( 'r', 'n', 'n', 'l', 'i', '(', '(' ),
 899                         $value
 900                 );
 901
 902                 // Let the value through if it's nothing but a single comment, to
 903                 // allow other functions which may reject it to pass some error
 904                 // message through.
 905                 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
 906                         // Remove any comments; IE gets token splitting wrong
 907                         // This must be done AFTER decoding character references and
 908                         // escape sequences, because those steps can introduce comments
 909                         // This step cannot introduce character references or escape
 910                         // sequences, because it replaces comments with spaces rather
 911                         // than removing them completely.
 912                         $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 913
 914                         // Remove anything after a comment-start token, to guard against
 915                         // incorrect client implementations.
 916                         $commentPos = strpos( $value, '/*' );
 917                         if ( $commentPos !== false ) {
 918                                 $value = substr( $value, 0, $commentPos );
 919                         }
 920                 }
 921
 922                 // S followed by repeat, iteration, or prolonged sound marks,
 923                 // which IE will treat as "ss"
 924                 $value = preg_replace(
 925                         '/s(?:
 926                                 \xE3\x80\xB1 | # U+3031
 927                                 \xE3\x82\x9D | # U+309D
 928                                 \xE3\x83\xBC | # U+30FC
 929                                 \xE3\x83\xBD | # U+30FD
 930                                 \xEF\xB9\xBC | # U+FE7C
 931                                 \xEF\xB9\xBD | # U+FE7D
 932                                 \xEF\xBD\xB0   # U+FF70
 933                         )/ix',
 934                         'ss',
 935                         $value
 936                 );
 937
 938                 // Reject problematic keywords and control characters
 939                 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ) {
 940                         return '/* invalid control char */';
 941                 } elseif ( preg_match(
 942                         '! expression
 943                                 | filter\s*:
 944                                 | accelerator\s*:
 945                                 | -o-link\s*:
 946                                 | -o-link-source\s*:
 947                                 | -o-replace\s*:
 948                                 | url\s*\(
 949                                 | image\s*\(
 950                                 | image-set\s*\(
 951                         !ix', $value ) ) {
 952                         return '/* insecure input */';
 953                 }
 954                 return $value;
 955         }
 956
 957         /**
 958          * @param array $matches
 959          * @return string
 960          */
 961         static function cssDecodeCallback( $matches ) {
 962                 if ( $matches[1] !== '' ) {
 963                         // Line continuation
 964                         return '';
 965                 } elseif ( $matches[2] !== '' ) {
 966                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 967                 } elseif ( $matches[3] !== '' ) {
 968                         $char = $matches[3];
 969                 } else {
 970                         $char = '\\';
 971                 }
 972                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 973                         // These characters need to be escaped in strings
 974                         // Clean up the escape sequence to avoid parsing errors by clients
 975                         return '\\' . dechex( ord( $char ) ) . ' ';
 976                 } else {
 977                         // Decode unnecessary escape
 978                         return $char;
 979                 }
 980         }
 981
 982         /**
 983          * Take a tag soup fragment listing an HTML element's attributes
 984          * and normalize it to well-formed XML, discarding unwanted attributes.
 985          * Output is safe for further wikitext processing, with escaping of
 986          * values that could trigger problems.
 987          *
 988          * - Normalizes attribute names to lowercase
 989          * - Discards attributes not on a whitelist for the given element
 990          * - Turns broken or invalid entities into plaintext
 991          * - Double-quotes all attribute values
 992          * - Attributes without values are given the name as attribute
 993          * - Double attributes are discarded
 994          * - Unsafe style attributes are discarded
 995          * - Prepends space if there are attributes.
 996          *
 997          * @param string $text
 998          * @param string $element
 999          * @return string
1000          */
1001         static function fixTagAttributes( $text, $element ) {
1002                 if ( trim( $text ) == '' ) {
1003                         return '';
1004                 }
1005
1006                 $decoded = Sanitizer::decodeTagAttributes( $text );
1007                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1008
1009                 return Sanitizer::safeEncodeTagAttributes( $stripped );
1010         }
1011
1012         /**
1013          * Encode an attribute value for HTML output.
1014          * @param string $text
1015          * @return string HTML-encoded text fragment
1016          */
1017         static function encodeAttribute( $text ) {
1018                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1019
1020                 // Whitespace is normalized during attribute decoding,
1021                 // so if we've been passed non-spaces we must encode them
1022                 // ahead of time or they won't be preserved.
1023                 $encValue = strtr( $encValue, array(
1024                         "\n" => '&#10;',
1025                         "\r" => '&#13;',
1026                         "\t" => '&#9;',
1027                 ) );
1028
1029                 return $encValue;
1030         }
1031
1032         /**
1033          * Encode an attribute value for HTML tags, with extra armoring
1034          * against further wiki processing.
1035          * @param string $text
1036          * @return string HTML-encoded text fragment
1037          */
1038         static function safeEncodeAttribute( $text ) {
1039                 $encValue = Sanitizer::encodeAttribute( $text );
1040
1041                 # Templates and links may be expanded in later parsing,
1042                 # creating invalid or dangerous output. Suppress this.
1043                 $encValue = strtr( $encValue, array(
1044                         '<'    => '&lt;',   // This should never happen,
1045                         '>'    => '&gt;',   // we've received invalid input
1046                         '"'    => '&quot;', // which should have been escaped.
1047                         '{'    => '&#123;',
1048                         '['    => '&#91;',
1049                         "''"   => '&#39;&#39;',
1050                         'ISBN' => '&#73;SBN',
1051                         'RFC'  => '&#82;FC',
1052                         'PMID' => '&#80;MID',
1053                         '|'    => '&#124;',
1054                         '__'   => '&#95;_',
1055                 ) );
1056
1057                 # Stupid hack
1058                 $encValue = preg_replace_callback(
1059                         '/((?i)' . wfUrlProtocols() . ')/',
1060                         array( 'Sanitizer', 'armorLinksCallback' ),
1061                         $encValue );
1062                 return $encValue;
1063         }
1064
1065         /**
1066          * Given a value, escape it so that it can be used in an id attribute and
1067          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1068          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1069          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1070          * escaped with lots of dots.
1071          *
1072          * To ensure we don't have to bother escaping anything, we also strip ', ",
1073          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1074          * We also strip # because it upsets IE, and % because it could be
1075          * ambiguous if it's part of something that looks like a percent escape
1076          * (which don't work reliably in fragments cross-browser).
1077          *
1078          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1079          *   in the id and name attributes
1080          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1081          *   the id attribute
1082          * @see http://www.whatwg.org/html/elements.html#the-id-attribute
1083          *   HTML5 definition of id attribute
1084          *
1085          * @param string $id Id to escape
1086          * @param string|array $options String or array of strings (default is array()):
1087          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1088          *       so don't pay attention if the first character isn't valid at the
1089          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1090          *       false.
1091          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1092          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1093          *       anchors and links won't break.
1094          * @return string
1095          */
1096         static function escapeId( $id, $options = array() ) {
1097                 global $wgExperimentalHtmlIds;
1098                 $options = (array)$options;
1099
1100                 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1101                         $id = Sanitizer::decodeCharReferences( $id );
1102                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1103                         $id = trim( $id, '_' );
1104                         if ( $id === '' ) {
1105                                 # Must have been all whitespace to start with.
1106                                 return '_';
1107                         } else {
1108                                 return $id;
1109                         }
1110                 }
1111
1112                 # HTML4-style escaping
1113                 static $replace = array(
1114                         '%3A' => ':',
1115                         '%' => '.'
1116                 );
1117
1118                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1119                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1120
1121                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1122                 && !in_array( 'noninitial', $options ) ) {
1123                         // Initial character must be a letter!
1124                         $id = "x$id";
1125                 }
1126                 return $id;
1127         }
1128
1129         /**
1130          * Given a value, escape it so that it can be used as a CSS class and
1131          * return it.
1132          *
1133          * @todo For extra validity, input should be validated UTF-8.
1134          *
1135          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1136          *
1137          * @param string $class
1138          * @return string
1139          */
1140         static function escapeClass( $class ) {
1141                 // Convert ugly stuff to underscores and kill underscores in ugly places
1142                 return rtrim( preg_replace(
1143                         array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1144                         '_',
1145                         $class ), '_' );
1146         }
1147
1148         /**
1149          * Given HTML input, escape with htmlspecialchars but un-escape entities.
1150          * This allows (generally harmless) entities like &#160; to survive.
1151          *
1152          * @param string $html HTML to escape
1153          * @return string Escaped input
1154          */
1155         static function escapeHtmlAllowEntities( $html ) {
1156                 $html = Sanitizer::decodeCharReferences( $html );
1157                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1158                 # hurt.
1159                 $html = htmlspecialchars( $html, ENT_QUOTES );
1160                 return $html;
1161         }
1162
1163         /**
1164          * Regex replace callback for armoring links against further processing.
1165          * @param array $matches
1166          * @return string
1167          */
1168         private static function armorLinksCallback( $matches ) {
1169                 return str_replace( ':', '&#58;', $matches[1] );
1170         }
1171
1172         /**
1173          * Return an associative array of attribute names and values from
1174          * a partial tag string. Attribute names are forces to lowercase,
1175          * character references are decoded to UTF-8 text.
1176          *
1177          * @param string $text
1178          * @return array
1179          */
1180         public static function decodeTagAttributes( $text ) {
1181                 if ( trim( $text ) == '' ) {
1182                         return array();
1183                 }
1184
1185                 $attribs = array();
1186                 $pairs = array();
1187                 if ( !preg_match_all(
1188                         self::getAttribsRegex(),
1189                         $text,
1190                         $pairs,
1191                         PREG_SET_ORDER ) ) {
1192                         return $attribs;
1193                 }
1194
1195                 foreach ( $pairs as $set ) {
1196                         $attribute = strtolower( $set[1] );
1197                         $value = Sanitizer::getTagAttributeCallback( $set );
1198
1199                         // Normalize whitespace
1200                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1201                         $value = trim( $value );
1202
1203                         // Decode character references
1204                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1205                 }
1206                 return $attribs;
1207         }
1208
1209         /**
1210          * Build a partial tag string from an associative array of attribute
1211          * names and values as returned by decodeTagAttributes.
1212          *
1213          * @param array $assoc_array
1214          * @return string
1215          */
1216         public static function safeEncodeTagAttributes( $assoc_array ) {
1217                 $attribs = array();
1218                 foreach ( $assoc_array as $attribute => $value ) {
1219                         $encAttribute = htmlspecialchars( $attribute );
1220                         $encValue = Sanitizer::safeEncodeAttribute( $value );
1221
1222                         $attribs[] = "$encAttribute=\"$encValue\"";
1223                 }
1224                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1225         }
1226
1227         /**
1228          * Pick the appropriate attribute value from a match set from the
1229          * attribs regex matches.
1230          *
1231          * @param array $set
1232          * @throws MWException When tag conditions are not met.
1233          * @return string
1234          */
1235         private static function getTagAttributeCallback( $set ) {
1236                 if ( isset( $set[6] ) ) {
1237                         # Illegal #XXXXXX color with no quotes.
1238                         return $set[6];
1239                 } elseif ( isset( $set[5] ) ) {
1240                         # No quotes.
1241                         return $set[5];
1242                 } elseif ( isset( $set[4] ) ) {
1243                         # Single-quoted
1244                         return $set[4];
1245                 } elseif ( isset( $set[3] ) ) {
1246                         # Double-quoted
1247                         return $set[3];
1248                 } elseif ( !isset( $set[2] ) ) {
1249                         # In XHTML, attributes must have a value.
1250                         # For 'reduced' form, return explicitly the attribute name here.
1251                         return $set[1];
1252                 } else {
1253                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1254                 }
1255         }
1256
1257         /**
1258          * Normalize whitespace and character references in an XML source-
1259          * encoded text for an attribute value.
1260          *
1261          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1262          * but note that we're not returning the value, but are returning
1263          * XML source fragments that will be slapped into output.
1264          *
1265          * @param string $text
1266          * @return string
1267          * @todo Remove, unused?
1268          */
1269         private static function normalizeAttributeValue( $text ) {
1270                 return str_replace( '"', '&quot;',
1271                         self::normalizeWhitespace(
1272                                 Sanitizer::normalizeCharReferences( $text ) ) );
1273         }
1274
1275         /**
1276          * @param string $text
1277          * @return string
1278          */
1279         private static function normalizeWhitespace( $text ) {
1280                 return preg_replace(
1281                         '/\r\n|[\x20\x0d\x0a\x09]/',
1282                         ' ',
1283                         $text );
1284         }
1285
1286         /**
1287          * Normalizes whitespace in a section name, such as might be returned
1288          * by Parser::stripSectionName(), for use in the id's that are used for
1289          * section links.
1290          *
1291          * @param string $section
1292          * @return string
1293          */
1294         static function normalizeSectionNameWhitespace( $section ) {
1295                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1296         }
1297
1298         /**
1299          * Ensure that any entities and character references are legal
1300          * for XML and XHTML specifically. Any stray bits will be
1301          * &amp;-escaped to result in a valid text fragment.
1302          *
1303          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1304          *   numericized (this way we're well-formed even without a DTD)
1305          * b. any numeric char refs must be legal chars, not invalid or forbidden
1306          * c. use lower cased "&#x", not "&#X"
1307          * d. fix or reject non-valid attributes
1308          *
1309          * @param string $text
1310          * @return string
1311          * @private
1312          */
1313         static function normalizeCharReferences( $text ) {
1314                 return preg_replace_callback(
1315                         self::CHAR_REFS_REGEX,
1316                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1317                         $text );
1318         }
1319
1320         /**
1321          * @param string $matches
1322          * @return string
1323          */
1324         static function normalizeCharReferencesCallback( $matches ) {
1325                 $ret = null;
1326                 if ( $matches[1] != '' ) {
1327                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1328                 } elseif ( $matches[2] != '' ) {
1329                         $ret = Sanitizer::decCharReference( $matches[2] );
1330                 } elseif ( $matches[3] != '' ) {
1331                         $ret = Sanitizer::hexCharReference( $matches[3] );
1332                 }
1333                 if ( is_null( $ret ) ) {
1334                         return htmlspecialchars( $matches[0] );
1335                 } else {
1336                         return $ret;
1337                 }
1338         }
1339
1340         /**
1341          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1342          * return the equivalent numeric entity reference (except for the core &lt;
1343          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1344          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1345          * pseudo-entity source (eg &amp;foo;)
1346          *
1347          * @param string $name
1348          * @return string
1349          */
1350         static function normalizeEntity( $name ) {
1351                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1352                         return '&' . self::$htmlEntityAliases[$name] . ';';
1353                 } elseif ( in_array( $name,
1354                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1355                         return "&$name;";
1356                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1357                         return '&#' . self::$htmlEntities[$name] . ';';
1358                 } else {
1359                         return "&amp;$name;";
1360                 }
1361         }
1362
1363         /**
1364          * @param int $codepoint
1365          * @return null|string
1366          */
1367         static function decCharReference( $codepoint ) {
1368                 $point = intval( $codepoint );
1369                 if ( Sanitizer::validateCodepoint( $point ) ) {
1370                         return sprintf( '&#%d;', $point );
1371                 } else {
1372                         return null;
1373                 }
1374         }
1375
1376         /**
1377          * @param int $codepoint
1378          * @return null|string
1379          */
1380         static function hexCharReference( $codepoint ) {
1381                 $point = hexdec( $codepoint );
1382                 if ( Sanitizer::validateCodepoint( $point ) ) {
1383                         return sprintf( '&#x%x;', $point );
1384                 } else {
1385                         return null;
1386                 }
1387         }
1388
1389         /**
1390          * Returns true if a given Unicode codepoint is a valid character in XML.
1391          * @param int $codepoint
1392          * @return bool
1393          */
1394         private static function validateCodepoint( $codepoint ) {
1395                 return $codepoint == 0x09
1396                         || $codepoint == 0x0a
1397                         || $codepoint == 0x0d
1398                         || ( $codepoint >= 0x20 && $codepoint <= 0xd7ff )
1399                         || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1400                         || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1401         }
1402
1403         /**
1404          * Decode any character references, numeric or named entities,
1405          * in the text and return a UTF-8 string.
1406          *
1407          * @param string $text
1408          * @return string
1409          */
1410         public static function decodeCharReferences( $text ) {
1411                 return preg_replace_callback(
1412                         self::CHAR_REFS_REGEX,
1413                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1414                         $text );
1415         }
1416
1417         /**
1418          * Decode any character references, numeric or named entities,
1419          * in the next and normalize the resulting string. (bug 14952)
1420          *
1421          * This is useful for page titles, not for text to be displayed,
1422          * MediaWiki allows HTML entities to escape normalization as a feature.
1423          *
1424          * @param string $text Already normalized, containing entities
1425          * @return string Still normalized, without entities
1426          */
1427         public static function decodeCharReferencesAndNormalize( $text ) {
1428                 global $wgContLang;
1429                 $text = preg_replace_callback(
1430                         self::CHAR_REFS_REGEX,
1431                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1432                         $text, /* limit */ -1, $count );
1433
1434                 if ( $count ) {
1435                         return $wgContLang->normalize( $text );
1436                 } else {
1437                         return $text;
1438                 }
1439         }
1440
1441         /**
1442          * @param string $matches
1443          * @return string
1444          */
1445         static function decodeCharReferencesCallback( $matches ) {
1446                 if ( $matches[1] != '' ) {
1447                         return Sanitizer::decodeEntity( $matches[1] );
1448                 } elseif ( $matches[2] != '' ) {
1449                         return Sanitizer::decodeChar( intval( $matches[2] ) );
1450                 } elseif ( $matches[3] != '' ) {
1451                         return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1452                 }
1453                 # Last case should be an ampersand by itself
1454                 return $matches[0];
1455         }
1456
1457         /**
1458          * Return UTF-8 string for a codepoint if that is a valid
1459          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1460          * @param int $codepoint
1461          * @return string
1462          * @private
1463          */
1464         static function decodeChar( $codepoint ) {
1465                 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1466                         return codepointToUtf8( $codepoint );
1467                 } else {
1468                         return UTF8_REPLACEMENT;
1469                 }
1470         }
1471
1472         /**
1473          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1474          * return the UTF-8 encoding of that character. Otherwise, returns
1475          * pseudo-entity source (eg "&foo;")
1476          *
1477          * @param string $name
1478          * @return string
1479          */
1480         static function decodeEntity( $name ) {
1481                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1482                         $name = self::$htmlEntityAliases[$name];
1483                 }
1484                 if ( isset( self::$htmlEntities[$name] ) ) {
1485                         return codepointToUtf8( self::$htmlEntities[$name] );
1486                 } else {
1487                         return "&$name;";
1488                 }
1489         }
1490
1491         /**
1492          * Fetch the whitelist of acceptable attributes for a given element name.
1493          *
1494          * @param string $element
1495          * @return array
1496          */
1497         static function attributeWhitelist( $element ) {
1498                 $list = Sanitizer::setupAttributeWhitelist();
1499                 return isset( $list[$element] )
1500                         ? $list[$element]
1501                         : array();
1502         }
1503
1504         /**
1505          * Foreach array key (an allowed HTML element), return an array
1506          * of allowed attributes
1507          * @return array
1508          */
1509         static function setupAttributeWhitelist() {
1510                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
1511                 static $whitelist, $staticInitialised;
1512
1513                 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgAllowMicrodataAttributes' ) );
1514
1515                 if ( $whitelist !== null && $staticInitialised == $globalContext ) {
1516                         return $whitelist;
1517                 }
1518
1519                 $common = array(
1520                         # HTML
1521                         'id',
1522                         'class',
1523                         'style',
1524                         'lang',
1525                         'dir',
1526                         'title',
1527
1528                         # WAI-ARIA
1529                         'role',
1530                 );
1531
1532                 if ( $wgAllowRdfaAttributes ) {
1533                         # RDFa attributes as specified in section 9 of
1534                         # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1535                         $common = array_merge( $common, array(
1536                                 'about', 'property', 'resource', 'datatype', 'typeof',
1537                         ) );
1538                 }
1539
1540                 if ( $wgAllowMicrodataAttributes ) {
1541                         # add HTML5 microdata tags as specified by
1542                         # http://www.whatwg.org/html/microdata.html#the-microdata-model
1543                         $common = array_merge( $common, array(
1544                                 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1545                         ) );
1546                 }
1547
1548                 $block = array_merge( $common, array( 'align' ) );
1549                 $tablealign = array( 'align', 'valign' );
1550                 $tablecell = array(
1551                         'abbr',
1552                         'axis',
1553                         'headers',
1554                         'scope',
1555                         'rowspan',
1556                         'colspan',
1557                         'nowrap', # deprecated
1558                         'width', # deprecated
1559                         'height', # deprecated
1560                         'bgcolor', # deprecated
1561                 );
1562
1563                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1564                 # See: http://www.w3.org/TR/html4/
1565                 $whitelist = array(
1566                         # 7.5.4
1567                         'div'        => $block,
1568                         'center'     => $common, # deprecated
1569                         'span'       => $common,
1570
1571                         # 7.5.5
1572                         'h1'         => $block,
1573                         'h2'         => $block,
1574                         'h3'         => $block,
1575                         'h4'         => $block,
1576                         'h5'         => $block,
1577                         'h6'         => $block,
1578
1579                         # 7.5.6
1580                         # address
1581
1582                         # 8.2.4
1583                         'bdo'        => $common,
1584
1585                         # 9.2.1
1586                         'em'         => $common,
1587                         'strong'     => $common,
1588                         'cite'       => $common,
1589                         'dfn'        => $common,
1590                         'code'       => $common,
1591                         'samp'       => $common,
1592                         'kbd'        => $common,
1593                         'var'        => $common,
1594                         'abbr'       => $common,
1595                         # acronym
1596
1597                         # 9.2.2
1598                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1599                         'q'          => array_merge( $common, array( 'cite' ) ),
1600
1601                         # 9.2.3
1602                         'sub'        => $common,
1603                         'sup'        => $common,
1604
1605                         # 9.3.1
1606                         'p'          => $block,
1607
1608                         # 9.3.2
1609                         'br'         => array_merge( $common, array( 'clear' ) ),
1610
1611                         # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1612                         'wbr'        => $common,
1613
1614                         # 9.3.4
1615                         'pre'        => array_merge( $common, array( 'width' ) ),
1616
1617                         # 9.4
1618                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1619                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1620
1621                         # 10.2
1622                         'ul'         => array_merge( $common, array( 'type' ) ),
1623                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1624                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1625
1626                         # 10.3
1627                         'dl'         => $common,
1628                         'dd'         => $common,
1629                         'dt'         => $common,
1630
1631                         # 11.2.1
1632                         'table'      => array_merge( $common,
1633                                                                 array( 'summary', 'width', 'border', 'frame',
1634                                                                                 'rules', 'cellspacing', 'cellpadding',
1635                                                                                 'align', 'bgcolor',
1636                                                                 ) ),
1637
1638                         # 11.2.2
1639                         'caption'    => $block,
1640
1641                         # 11.2.3
1642                         'thead'      => $common,
1643                         'tfoot'      => $common,
1644                         'tbody'      => $common,
1645
1646                         # 11.2.4
1647                         'colgroup'   => array_merge( $common, array( 'span' ) ),
1648                         'col'        => array_merge( $common, array( 'span' ) ),
1649
1650                         # 11.2.5
1651                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1652
1653                         # 11.2.6
1654                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1655                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1656
1657                         # 12.2
1658                         # NOTE: <a> is not allowed directly, but the attrib
1659                         # whitelist is used from the Parser object
1660                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1661
1662                         # 13.2
1663                         # Not usually allowed, but may be used for extension-style hooks
1664                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1665                         # true
1666                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1667
1668                         # 15.2.1
1669                         'tt'         => $common,
1670                         'b'          => $common,
1671                         'i'          => $common,
1672                         'big'        => $common,
1673                         'small'      => $common,
1674                         'strike'     => $common,
1675                         's'          => $common,
1676                         'u'          => $common,
1677
1678                         # 15.2.2
1679                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1680                         # basefont
1681
1682                         # 15.3
1683                         'hr'         => array_merge( $common, array( 'width' ) ),
1684
1685                         # HTML Ruby annotation text module, simple ruby only.
1686                         # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
1687                         'ruby'       => $common,
1688                         # rbc
1689                         'rb'         => $common,
1690                         'rp'         => $common,
1691                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1692                         'rtc'         => $common,
1693
1694                         # MathML root element, where used for extensions
1695                         # 'title' may not be 100% valid here; it's XHTML
1696                         # http://www.w3.org/TR/REC-MathML/
1697                         'math'       => array( 'class', 'style', 'id', 'title' ),
1698
1699                         # HTML 5 section 4.6
1700                         'bdi' => $common,
1701
1702                         # HTML5 elements, defined by:
1703                         # http://www.whatwg.org/html/
1704                         'data' => array_merge( $common, array( 'value' ) ),
1705                         'time' => array_merge( $common, array( 'datetime' ) ),
1706                         'mark' => $common,
1707
1708                         // meta and link are only permitted by removeHTMLtags when Microdata
1709                         // is enabled so we don't bother adding a conditional to hide these
1710                         // Also meta and link are only valid in WikiText as Microdata elements
1711                         // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1712                         // So we don't bother including $common attributes that have no purpose.
1713                         'meta' => array( 'itemprop', 'content' ),
1714                         'link' => array( 'itemprop', 'href' ),
1715                 );
1716
1717                 $staticInitialised = $globalContext;
1718
1719                 return $whitelist;
1720         }
1721
1722         /**
1723          * Take a fragment of (potentially invalid) HTML and return
1724          * a version with any tags removed, encoded as plain text.
1725          *
1726          * Warning: this return value must be further escaped for literal
1727          * inclusion in HTML output as of 1.10!
1728          *
1729          * @param string $text HTML fragment
1730          * @return string
1731          */
1732         static function stripAllTags( $text ) {
1733                 # Actual <tags>
1734                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1735
1736                 # Normalize &entities and whitespace
1737                 $text = self::decodeCharReferences( $text );
1738                 $text = self::normalizeWhitespace( $text );
1739
1740                 return $text;
1741         }
1742
1743         /**
1744          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1745          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1746          * PHP 5.1 doesn't.
1747          *
1748          * Use for passing XHTML fragments to PHP's XML parsing functions
1749          *
1750          * @return string
1751          */
1752         static function hackDocType() {
1753                 $out = "<!DOCTYPE html [\n";
1754                 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1755                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1756                 }
1757                 $out .= "]>\n";
1758                 return $out;
1759         }
1760
1761         /**
1762          * @param string $url
1763          * @return mixed|string
1764          */
1765         static function cleanUrl( $url ) {
1766                 # Normalize any HTML entities in input. They will be
1767                 # re-escaped by makeExternalLink().
1768                 $url = Sanitizer::decodeCharReferences( $url );
1769
1770                 # Escape any control characters introduced by the above step
1771                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1772                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1773
1774                 # Validate hostname portion
1775                 $matches = array();
1776                 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1777                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1778
1779                         // Characters that will be ignored in IDNs.
1780                         // http://tools.ietf.org/html/3454#section-3.1
1781                         // Strip them before further processing so blacklists and such work.
1782                         $strip = "/
1783                                 \\s|          # general whitespace
1784                                 \xc2\xad|     # 00ad SOFT HYPHEN
1785                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1786                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1787                                 \xe2\x81\xa0| # 2060 WORD JOINER
1788                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1789                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1790                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1791                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1792                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1793                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1794                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1795                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1796                                 /xuD";
1797
1798                         $host = preg_replace( $strip, '', $host );
1799
1800                         // @todo FIXME: Validate hostnames here
1801
1802                         return $protocol . $host . $rest;
1803                 } else {
1804                         return $url;
1805                 }
1806         }
1807
1808         /**
1809          * @param array $matches
1810          * @return string
1811          */
1812         static function cleanUrlCallback( $matches ) {
1813                 return urlencode( $matches[0] );
1814         }
1815
1816         /**
1817          * Does a string look like an e-mail address?
1818          *
1819          * This validates an email address using an HTML5 specification found at:
1820          * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1821          * Which as of 2011-01-24 says:
1822          *
1823          *   A valid e-mail address is a string that matches the ABNF production
1824          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1825          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1826          *   3.5.
1827          *
1828          * This function is an implementation of the specification as requested in
1829          * bug 22449.
1830          *
1831          * Client-side forms will use the same standard validation rules via JS or
1832          * HTML 5 validation; additional restrictions can be enforced server-side
1833          * by extensions via the 'isValidEmailAddr' hook.
1834          *
1835          * Note that this validation doesn't 100% match RFC 2822, but is believed
1836          * to be liberal enough for wide use. Some invalid addresses will still
1837          * pass validation here.
1838          *
1839          * @since 1.18
1840          *
1841          * @param string $addr E-mail address
1842          * @return bool
1843          */
1844         public static function validateEmail( $addr ) {
1845                 $result = null;
1846                 if ( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1847                         return $result;
1848                 }
1849
1850                 // Please note strings below are enclosed in brackets [], this make the
1851                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1852                 // See bug 26948
1853                 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1854                 $rfc1034_ldh_str = "a-z0-9\\-";
1855
1856                 $html5_email_regexp = "/
1857                 ^                      # start of string
1858                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1859                 @                      # 'apostrophe'
1860                 [$rfc1034_ldh_str]+       # First domain part
1861                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1862                 $                      # End of string
1863                 /ix"; // case Insensitive, eXtended
1864
1865                 return (bool)preg_match( $html5_email_regexp, $addr );
1866         }
1867 }