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                 $id = Sanitizer::decodeCharReferences( $id );
1101
1102                 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1103                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1104                         $id = trim( $id, '_' );
1105                         if ( $id === '' ) {
1106                                 # Must have been all whitespace to start with.
1107                                 return '_';
1108                         } else {
1109                                 return $id;
1110                         }
1111                 }
1112
1113                 # HTML4-style escaping
1114                 static $replace = array(
1115                         '%3A' => ':',
1116                         '%' => '.'
1117                 );
1118
1119                 $id = urlencode( strtr( $id, ' ', '_' ) );
1120                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1121
1122                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1123                 && !in_array( 'noninitial', $options ) ) {
1124                         // Initial character must be a letter!
1125                         $id = "x$id";
1126                 }
1127                 return $id;
1128         }
1129
1130         /**
1131          * Given a value, escape it so that it can be used as a CSS class and
1132          * return it.
1133          *
1134          * @todo For extra validity, input should be validated UTF-8.
1135          *
1136          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1137          *
1138          * @param string $class
1139          * @return string
1140          */
1141         static function escapeClass( $class ) {
1142                 // Convert ugly stuff to underscores and kill underscores in ugly places
1143                 return rtrim( preg_replace(
1144                         array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1145                         '_',
1146                         $class ), '_' );
1147         }
1148
1149         /**
1150          * Given HTML input, escape with htmlspecialchars but un-escape entities.
1151          * This allows (generally harmless) entities like &#160; to survive.
1152          *
1153          * @param string $html HTML to escape
1154          * @return string Escaped input
1155          */
1156         static function escapeHtmlAllowEntities( $html ) {
1157                 $html = Sanitizer::decodeCharReferences( $html );
1158                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1159                 # hurt.
1160                 $html = htmlspecialchars( $html, ENT_QUOTES );
1161                 return $html;
1162         }
1163
1164         /**
1165          * Regex replace callback for armoring links against further processing.
1166          * @param array $matches
1167          * @return string
1168          */
1169         private static function armorLinksCallback( $matches ) {
1170                 return str_replace( ':', '&#58;', $matches[1] );
1171         }
1172
1173         /**
1174          * Return an associative array of attribute names and values from
1175          * a partial tag string. Attribute names are forces to lowercase,
1176          * character references are decoded to UTF-8 text.
1177          *
1178          * @param string $text
1179          * @return array
1180          */
1181         public static function decodeTagAttributes( $text ) {
1182                 if ( trim( $text ) == '' ) {
1183                         return array();
1184                 }
1185
1186                 $attribs = array();
1187                 $pairs = array();
1188                 if ( !preg_match_all(
1189                         self::getAttribsRegex(),
1190                         $text,
1191                         $pairs,
1192                         PREG_SET_ORDER ) ) {
1193                         return $attribs;
1194                 }
1195
1196                 foreach ( $pairs as $set ) {
1197                         $attribute = strtolower( $set[1] );
1198                         $value = Sanitizer::getTagAttributeCallback( $set );
1199
1200                         // Normalize whitespace
1201                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1202                         $value = trim( $value );
1203
1204                         // Decode character references
1205                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1206                 }
1207                 return $attribs;
1208         }
1209
1210         /**
1211          * Build a partial tag string from an associative array of attribute
1212          * names and values as returned by decodeTagAttributes.
1213          *
1214          * @param array $assoc_array
1215          * @return string
1216          */
1217         public static function safeEncodeTagAttributes( $assoc_array ) {
1218                 $attribs = array();
1219                 foreach ( $assoc_array as $attribute => $value ) {
1220                         $encAttribute = htmlspecialchars( $attribute );
1221                         $encValue = Sanitizer::safeEncodeAttribute( $value );
1222
1223                         $attribs[] = "$encAttribute=\"$encValue\"";
1224                 }
1225                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1226         }
1227
1228         /**
1229          * Pick the appropriate attribute value from a match set from the
1230          * attribs regex matches.
1231          *
1232          * @param array $set
1233          * @throws MWException When tag conditions are not met.
1234          * @return string
1235          */
1236         private static function getTagAttributeCallback( $set ) {
1237                 if ( isset( $set[6] ) ) {
1238                         # Illegal #XXXXXX color with no quotes.
1239                         return $set[6];
1240                 } elseif ( isset( $set[5] ) ) {
1241                         # No quotes.
1242                         return $set[5];
1243                 } elseif ( isset( $set[4] ) ) {
1244                         # Single-quoted
1245                         return $set[4];
1246                 } elseif ( isset( $set[3] ) ) {
1247                         # Double-quoted
1248                         return $set[3];
1249                 } elseif ( !isset( $set[2] ) ) {
1250                         # In XHTML, attributes must have a value.
1251                         # For 'reduced' form, return explicitly the attribute name here.
1252                         return $set[1];
1253                 } else {
1254                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1255                 }
1256         }
1257
1258         /**
1259          * Normalize whitespace and character references in an XML source-
1260          * encoded text for an attribute value.
1261          *
1262          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1263          * but note that we're not returning the value, but are returning
1264          * XML source fragments that will be slapped into output.
1265          *
1266          * @param string $text
1267          * @return string
1268          * @todo Remove, unused?
1269          */
1270         private static function normalizeAttributeValue( $text ) {
1271                 return str_replace( '"', '&quot;',
1272                         self::normalizeWhitespace(
1273                                 Sanitizer::normalizeCharReferences( $text ) ) );
1274         }
1275
1276         /**
1277          * @param string $text
1278          * @return string
1279          */
1280         private static function normalizeWhitespace( $text ) {
1281                 return preg_replace(
1282                         '/\r\n|[\x20\x0d\x0a\x09]/',
1283                         ' ',
1284                         $text );
1285         }
1286
1287         /**
1288          * Normalizes whitespace in a section name, such as might be returned
1289          * by Parser::stripSectionName(), for use in the id's that are used for
1290          * section links.
1291          *
1292          * @param string $section
1293          * @return string
1294          */
1295         static function normalizeSectionNameWhitespace( $section ) {
1296                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1297         }
1298
1299         /**
1300          * Ensure that any entities and character references are legal
1301          * for XML and XHTML specifically. Any stray bits will be
1302          * &amp;-escaped to result in a valid text fragment.
1303          *
1304          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1305          *   numericized (this way we're well-formed even without a DTD)
1306          * b. any numeric char refs must be legal chars, not invalid or forbidden
1307          * c. use lower cased "&#x", not "&#X"
1308          * d. fix or reject non-valid attributes
1309          *
1310          * @param string $text
1311          * @return string
1312          * @private
1313          */
1314         static function normalizeCharReferences( $text ) {
1315                 return preg_replace_callback(
1316                         self::CHAR_REFS_REGEX,
1317                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1318                         $text );
1319         }
1320
1321         /**
1322          * @param string $matches
1323          * @return string
1324          */
1325         static function normalizeCharReferencesCallback( $matches ) {
1326                 $ret = null;
1327                 if ( $matches[1] != '' ) {
1328                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1329                 } elseif ( $matches[2] != '' ) {
1330                         $ret = Sanitizer::decCharReference( $matches[2] );
1331                 } elseif ( $matches[3] != '' ) {
1332                         $ret = Sanitizer::hexCharReference( $matches[3] );
1333                 }
1334                 if ( is_null( $ret ) ) {
1335                         return htmlspecialchars( $matches[0] );
1336                 } else {
1337                         return $ret;
1338                 }
1339         }
1340
1341         /**
1342          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1343          * return the equivalent numeric entity reference (except for the core &lt;
1344          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1345          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1346          * pseudo-entity source (eg &amp;foo;)
1347          *
1348          * @param string $name
1349          * @return string
1350          */
1351         static function normalizeEntity( $name ) {
1352                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1353                         return '&' . self::$htmlEntityAliases[$name] . ';';
1354                 } elseif ( in_array( $name,
1355                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1356                         return "&$name;";
1357                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1358                         return '&#' . self::$htmlEntities[$name] . ';';
1359                 } else {
1360                         return "&amp;$name;";
1361                 }
1362         }
1363
1364         /**
1365          * @param int $codepoint
1366          * @return null|string
1367          */
1368         static function decCharReference( $codepoint ) {
1369                 $point = intval( $codepoint );
1370                 if ( Sanitizer::validateCodepoint( $point ) ) {
1371                         return sprintf( '&#%d;', $point );
1372                 } else {
1373                         return null;
1374                 }
1375         }
1376
1377         /**
1378          * @param int $codepoint
1379          * @return null|string
1380          */
1381         static function hexCharReference( $codepoint ) {
1382                 $point = hexdec( $codepoint );
1383                 if ( Sanitizer::validateCodepoint( $point ) ) {
1384                         return sprintf( '&#x%x;', $point );
1385                 } else {
1386                         return null;
1387                 }
1388         }
1389
1390         /**
1391          * Returns true if a given Unicode codepoint is a valid character in XML.
1392          * @param int $codepoint
1393          * @return bool
1394          */
1395         private static function validateCodepoint( $codepoint ) {
1396                 return $codepoint == 0x09
1397                         || $codepoint == 0x0a
1398                         || $codepoint == 0x0d
1399                         || ( $codepoint >= 0x20 && $codepoint <= 0xd7ff )
1400                         || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1401                         || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1402         }
1403
1404         /**
1405          * Decode any character references, numeric or named entities,
1406          * in the text and return a UTF-8 string.
1407          *
1408          * @param string $text
1409          * @return string
1410          */
1411         public static function decodeCharReferences( $text ) {
1412                 return preg_replace_callback(
1413                         self::CHAR_REFS_REGEX,
1414                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1415                         $text );
1416         }
1417
1418         /**
1419          * Decode any character references, numeric or named entities,
1420          * in the next and normalize the resulting string. (bug 14952)
1421          *
1422          * This is useful for page titles, not for text to be displayed,
1423          * MediaWiki allows HTML entities to escape normalization as a feature.
1424          *
1425          * @param string $text Already normalized, containing entities
1426          * @return string Still normalized, without entities
1427          */
1428         public static function decodeCharReferencesAndNormalize( $text ) {
1429                 global $wgContLang;
1430                 $text = preg_replace_callback(
1431                         self::CHAR_REFS_REGEX,
1432                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1433                         $text, /* limit */ -1, $count );
1434
1435                 if ( $count ) {
1436                         return $wgContLang->normalize( $text );
1437                 } else {
1438                         return $text;
1439                 }
1440         }
1441
1442         /**
1443          * @param string $matches
1444          * @return string
1445          */
1446         static function decodeCharReferencesCallback( $matches ) {
1447                 if ( $matches[1] != '' ) {
1448                         return Sanitizer::decodeEntity( $matches[1] );
1449                 } elseif ( $matches[2] != '' ) {
1450                         return Sanitizer::decodeChar( intval( $matches[2] ) );
1451                 } elseif ( $matches[3] != '' ) {
1452                         return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1453                 }
1454                 # Last case should be an ampersand by itself
1455                 return $matches[0];
1456         }
1457
1458         /**
1459          * Return UTF-8 string for a codepoint if that is a valid
1460          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1461          * @param int $codepoint
1462          * @return string
1463          * @private
1464          */
1465         static function decodeChar( $codepoint ) {
1466                 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1467                         return codepointToUtf8( $codepoint );
1468                 } else {
1469                         return UTF8_REPLACEMENT;
1470                 }
1471         }
1472
1473         /**
1474          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1475          * return the UTF-8 encoding of that character. Otherwise, returns
1476          * pseudo-entity source (eg "&foo;")
1477          *
1478          * @param string $name
1479          * @return string
1480          */
1481         static function decodeEntity( $name ) {
1482                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1483                         $name = self::$htmlEntityAliases[$name];
1484                 }
1485                 if ( isset( self::$htmlEntities[$name] ) ) {
1486                         return codepointToUtf8( self::$htmlEntities[$name] );
1487                 } else {
1488                         return "&$name;";
1489                 }
1490         }
1491
1492         /**
1493          * Fetch the whitelist of acceptable attributes for a given element name.
1494          *
1495          * @param string $element
1496          * @return array
1497          */
1498         static function attributeWhitelist( $element ) {
1499                 $list = Sanitizer::setupAttributeWhitelist();
1500                 return isset( $list[$element] )
1501                         ? $list[$element]
1502                         : array();
1503         }
1504
1505         /**
1506          * Foreach array key (an allowed HTML element), return an array
1507          * of allowed attributes
1508          * @return array
1509          */
1510         static function setupAttributeWhitelist() {
1511                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
1512                 static $whitelist, $staticInitialised;
1513
1514                 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgAllowMicrodataAttributes' ) );
1515
1516                 if ( $whitelist !== null && $staticInitialised == $globalContext ) {
1517                         return $whitelist;
1518                 }
1519
1520                 $common = array(
1521                         # HTML
1522                         'id',
1523                         'class',
1524                         'style',
1525                         'lang',
1526                         'dir',
1527                         'title',
1528
1529                         # WAI-ARIA
1530                         'role',
1531                 );
1532
1533                 if ( $wgAllowRdfaAttributes ) {
1534                         # RDFa attributes as specified in section 9 of
1535                         # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1536                         $common = array_merge( $common, array(
1537                                 'about', 'property', 'resource', 'datatype', 'typeof',
1538                         ) );
1539                 }
1540
1541                 if ( $wgAllowMicrodataAttributes ) {
1542                         # add HTML5 microdata tags as specified by
1543                         # http://www.whatwg.org/html/microdata.html#the-microdata-model
1544                         $common = array_merge( $common, array(
1545                                 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1546                         ) );
1547                 }
1548
1549                 $block = array_merge( $common, array( 'align' ) );
1550                 $tablealign = array( 'align', 'valign' );
1551                 $tablecell = array(
1552                         'abbr',
1553                         'axis',
1554                         'headers',
1555                         'scope',
1556                         'rowspan',
1557                         'colspan',
1558                         'nowrap', # deprecated
1559                         'width', # deprecated
1560                         'height', # deprecated
1561                         'bgcolor', # deprecated
1562                 );
1563
1564                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1565                 # See: http://www.w3.org/TR/html4/
1566                 $whitelist = array(
1567                         # 7.5.4
1568                         'div'        => $block,
1569                         'center'     => $common, # deprecated
1570                         'span'       => $common,
1571
1572                         # 7.5.5
1573                         'h1'         => $block,
1574                         'h2'         => $block,
1575                         'h3'         => $block,
1576                         'h4'         => $block,
1577                         'h5'         => $block,
1578                         'h6'         => $block,
1579
1580                         # 7.5.6
1581                         # address
1582
1583                         # 8.2.4
1584                         'bdo'        => $common,
1585
1586                         # 9.2.1
1587                         'em'         => $common,
1588                         'strong'     => $common,
1589                         'cite'       => $common,
1590                         'dfn'        => $common,
1591                         'code'       => $common,
1592                         'samp'       => $common,
1593                         'kbd'        => $common,
1594                         'var'        => $common,
1595                         'abbr'       => $common,
1596                         # acronym
1597
1598                         # 9.2.2
1599                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1600                         'q'          => array_merge( $common, array( 'cite' ) ),
1601
1602                         # 9.2.3
1603                         'sub'        => $common,
1604                         'sup'        => $common,
1605
1606                         # 9.3.1
1607                         'p'          => $block,
1608
1609                         # 9.3.2
1610                         'br'         => array_merge( $common, array( 'clear' ) ),
1611
1612                         # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1613                         'wbr'        => $common,
1614
1615                         # 9.3.4
1616                         'pre'        => array_merge( $common, array( 'width' ) ),
1617
1618                         # 9.4
1619                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1620                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1621
1622                         # 10.2
1623                         'ul'         => array_merge( $common, array( 'type' ) ),
1624                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1625                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1626
1627                         # 10.3
1628                         'dl'         => $common,
1629                         'dd'         => $common,
1630                         'dt'         => $common,
1631
1632                         # 11.2.1
1633                         'table'      => array_merge( $common,
1634                                                                 array( 'summary', 'width', 'border', 'frame',
1635                                                                                 'rules', 'cellspacing', 'cellpadding',
1636                                                                                 'align', 'bgcolor',
1637                                                                 ) ),
1638
1639                         # 11.2.2
1640                         'caption'    => $block,
1641
1642                         # 11.2.3
1643                         'thead'      => $common,
1644                         'tfoot'      => $common,
1645                         'tbody'      => $common,
1646
1647                         # 11.2.4
1648                         'colgroup'   => array_merge( $common, array( 'span' ) ),
1649                         'col'        => array_merge( $common, array( 'span' ) ),
1650
1651                         # 11.2.5
1652                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1653
1654                         # 11.2.6
1655                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1656                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1657
1658                         # 12.2
1659                         # NOTE: <a> is not allowed directly, but the attrib
1660                         # whitelist is used from the Parser object
1661                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1662
1663                         # 13.2
1664                         # Not usually allowed, but may be used for extension-style hooks
1665                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1666                         # true
1667                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1668
1669                         # 15.2.1
1670                         'tt'         => $common,
1671                         'b'          => $common,
1672                         'i'          => $common,
1673                         'big'        => $common,
1674                         'small'      => $common,
1675                         'strike'     => $common,
1676                         's'          => $common,
1677                         'u'          => $common,
1678
1679                         # 15.2.2
1680                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1681                         # basefont
1682
1683                         # 15.3
1684                         'hr'         => array_merge( $common, array( 'width' ) ),
1685
1686                         # HTML Ruby annotation text module, simple ruby only.
1687                         # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
1688                         'ruby'       => $common,
1689                         # rbc
1690                         'rb'         => $common,
1691                         'rp'         => $common,
1692                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1693                         'rtc'         => $common,
1694
1695                         # MathML root element, where used for extensions
1696                         # 'title' may not be 100% valid here; it's XHTML
1697                         # http://www.w3.org/TR/REC-MathML/
1698                         'math'       => array( 'class', 'style', 'id', 'title' ),
1699
1700                         # HTML 5 section 4.6
1701                         'bdi' => $common,
1702
1703                         # HTML5 elements, defined by:
1704                         # http://www.whatwg.org/html/
1705                         'data' => array_merge( $common, array( 'value' ) ),
1706                         'time' => array_merge( $common, array( 'datetime' ) ),
1707                         'mark' => $common,
1708
1709                         // meta and link are only permitted by removeHTMLtags when Microdata
1710                         // is enabled so we don't bother adding a conditional to hide these
1711                         // Also meta and link are only valid in WikiText as Microdata elements
1712                         // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1713                         // So we don't bother including $common attributes that have no purpose.
1714                         'meta' => array( 'itemprop', 'content' ),
1715                         'link' => array( 'itemprop', 'href' ),
1716                 );
1717
1718                 $staticInitialised = $globalContext;
1719
1720                 return $whitelist;
1721         }
1722
1723         /**
1724          * Take a fragment of (potentially invalid) HTML and return
1725          * a version with any tags removed, encoded as plain text.
1726          *
1727          * Warning: this return value must be further escaped for literal
1728          * inclusion in HTML output as of 1.10!
1729          *
1730          * @param string $text HTML fragment
1731          * @return string
1732          */
1733         static function stripAllTags( $text ) {
1734                 # Actual <tags>
1735                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1736
1737                 # Normalize &entities and whitespace
1738                 $text = self::decodeCharReferences( $text );
1739                 $text = self::normalizeWhitespace( $text );
1740
1741                 return $text;
1742         }
1743
1744         /**
1745          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1746          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1747          * PHP 5.1 doesn't.
1748          *
1749          * Use for passing XHTML fragments to PHP's XML parsing functions
1750          *
1751          * @return string
1752          */
1753         static function hackDocType() {
1754                 $out = "<!DOCTYPE html [\n";
1755                 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1756                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1757                 }
1758                 $out .= "]>\n";
1759                 return $out;
1760         }
1761
1762         /**
1763          * @param string $url
1764          * @return mixed|string
1765          */
1766         static function cleanUrl( $url ) {
1767                 # Normalize any HTML entities in input. They will be
1768                 # re-escaped by makeExternalLink().
1769                 $url = Sanitizer::decodeCharReferences( $url );
1770
1771                 # Escape any control characters introduced by the above step
1772                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1773                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1774
1775                 # Validate hostname portion
1776                 $matches = array();
1777                 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1778                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1779
1780                         // Characters that will be ignored in IDNs.
1781                         // http://tools.ietf.org/html/3454#section-3.1
1782                         // Strip them before further processing so blacklists and such work.
1783                         $strip = "/
1784                                 \\s|          # general whitespace
1785                                 \xc2\xad|     # 00ad SOFT HYPHEN
1786                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1787                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1788                                 \xe2\x81\xa0| # 2060 WORD JOINER
1789                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1790                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1791                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1792                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1793                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1794                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1795                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1796                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1797                                 /xuD";
1798
1799                         $host = preg_replace( $strip, '', $host );
1800
1801                         // @todo FIXME: Validate hostnames here
1802
1803                         return $protocol . $host . $rest;
1804                 } else {
1805                         return $url;
1806                 }
1807         }
1808
1809         /**
1810          * @param array $matches
1811          * @return string
1812          */
1813         static function cleanUrlCallback( $matches ) {
1814                 return urlencode( $matches[0] );
1815         }
1816
1817         /**
1818          * Does a string look like an e-mail address?
1819          *
1820          * This validates an email address using an HTML5 specification found at:
1821          * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1822          * Which as of 2011-01-24 says:
1823          *
1824          *   A valid e-mail address is a string that matches the ABNF production
1825          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1826          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1827          *   3.5.
1828          *
1829          * This function is an implementation of the specification as requested in
1830          * bug 22449.
1831          *
1832          * Client-side forms will use the same standard validation rules via JS or
1833          * HTML 5 validation; additional restrictions can be enforced server-side
1834          * by extensions via the 'isValidEmailAddr' hook.
1835          *
1836          * Note that this validation doesn't 100% match RFC 2822, but is believed
1837          * to be liberal enough for wide use. Some invalid addresses will still
1838          * pass validation here.
1839          *
1840          * @since 1.18
1841          *
1842          * @param string $addr E-mail address
1843          * @return bool
1844          */
1845         public static function validateEmail( $addr ) {
1846                 $result = null;
1847                 if ( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1848                         return $result;
1849                 }
1850
1851                 // Please note strings below are enclosed in brackets [], this make the
1852                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1853                 // See bug 26948
1854                 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1855                 $rfc1034_ldh_str = "a-z0-9\\-";
1856
1857                 $html5_email_regexp = "/
1858                 ^                      # start of string
1859                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1860                 @                      # 'apostrophe'
1861                 [$rfc1034_ldh_str]+       # First domain part
1862                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1863                 $                      # End of string
1864                 /ix"; // case Insensitive, eXtended
1865
1866                 return (bool)preg_match( $html5_email_regexp, $addr );
1867         }
1868 }