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 $attribsRegex;
 326
 327         /**
 328          * Regular expression to match HTML/XML attribute pairs within a tag.
 329          * Allows some... latitude.
 330          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 331          */
 332         static function getAttribsRegex() {
 333                 if ( self::$attribsRegex === null ) {
 334                         $attribFirst = '[:A-Z_a-z0-9]';
 335                         $attrib = '[:A-Z_a-z-.0-9]';
 336                         $space = '[\x09\x0a\x0d\x20]';
 337                         self::$attribsRegex =
 338                                 "/(?:^|$space)({$attribFirst}{$attrib}*)
 339                                   ($space*=$space*
 340                                         (?:
 341                                          # The attribute value: quoted or alone
 342                                           \"([^<\"]*)\"
 343                                          | '([^<']*)'
 344                                          |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
 345                                          |  (\#[0-9a-fA-F]+) # Technically wrong, but lots of
 346                                                                                  # colors are specified like this.
 347                                                                                  # We'll be normalizing it.
 348                                         )
 349                                 )?(?=$space|\$)/sx";
 350                 }
 351                 return self::$attribsRegex;
 352         }
 353
 354         /**
 355          * Cleans up HTML, removes dangerous tags and attributes, and
 356          * removes HTML comments
 357          * @private
 358          * @param string $text
 359          * @param callable $processCallback Callback to do any variable or parameter
 360          *   replacements in HTML attribute values
 361          * @param array $args Arguments for the processing callback
 362          * @param array $extratags For any extra tags to include
 363          * @param array $removetags For any tags (default or extra) to exclude
 364          * @return string
 365          */
 366         static function removeHTMLtags( $text, $processCallback = null,
 367                 $args = array(), $extratags = array(), $removetags = array()
 368         ) {
 369                 global $wgUseTidy, $wgAllowMicrodataAttributes, $wgAllowImageTag;
 370
 371                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 372                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 373
 374                 wfProfileIn( __METHOD__ );
 375
 376                 // Base our staticInitialised variable off of the global config state so that if the globals
 377                 // are changed (like in the screwed up test system) we will re-initialise the settings.
 378                 $globalContext = implode( '-', compact( 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
 379                 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
 380
 381                         $htmlpairsStatic = array( # Tags that must be closed
 382                                 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 383                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 384                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 385                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 386                                 'ruby', 'rt', 'rb', 'rp', 'p', 'span', 'abbr', 'dfn',
 387                                 'kbd', 'samp', 'data', 'time', 'mark'
 388                         );
 389                         $htmlsingle = array(
 390                                 'br', 'wbr', 'hr', 'li', 'dt', 'dd'
 391                         );
 392                         $htmlsingleonly = array( # Elements that cannot have close tags
 393                                 'br', 'wbr', 'hr'
 394                         );
 395                         if ( $wgAllowMicrodataAttributes ) {
 396                                 $htmlsingle[] = $htmlsingleonly[] = 'meta';
 397                                 $htmlsingle[] = $htmlsingleonly[] = 'link';
 398                         }
 399                         $htmlnest = array( # Tags that can be nested--??
 400                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 401                                 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
 402                                 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
 403                         );
 404                         $tabletags = array( # Can only appear inside table, we will close them
 405                                 'td', 'th', 'tr',
 406                         );
 407                         $htmllist = array( # Tags used by list
 408                                 'ul', 'ol',
 409                         );
 410                         $listtags = array( # Tags that can appear in a list
 411                                 'li',
 412                         );
 413
 414                         if ( $wgAllowImageTag ) {
 415                                 $htmlsingle[] = 'img';
 416                                 $htmlsingleonly[] = 'img';
 417                         }
 418
 419                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 420                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 421
 422                         # Convert them all to hashtables for faster lookup
 423                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 424                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 425                         foreach ( $vars as $var ) {
 426                                 $$var = array_flip( $$var );
 427                         }
 428                         $staticInitialised = $globalContext;
 429                 }
 430                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 431                 $extratags = array_flip( $extratags );
 432                 $removetags = array_flip( $removetags );
 433                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 434                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
 435
 436                 # Remove HTML comments
 437                 $text = Sanitizer::removeHTMLcomments( $text );
 438                 $bits = explode( '<', $text );
 439                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 440                 if ( !$wgUseTidy ) {
 441                         $tagstack = $tablestack = array();
 442                         foreach ( $bits as $x ) {
 443                                 $regs = array();
 444                                 # $slash: Does the current element start with a '/'?
 445                                 # $t: Current element name
 446                                 # $params: String between element name and >
 447                                 # $brace: Ending '>' or '/>'
 448                                 # $rest: Everything until the next element of $bits
 449                                 if ( preg_match( '!^(/?)([^\\s/>]+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 450                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 451                                 } else {
 452                                         $slash = $t = $params = $brace = $rest = null;
 453                                 }
 454
 455                                 $badtag = false;
 456                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 457                                         # Check our stack
 458                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 459                                                 $badtag = true;
 460                                         } elseif ( $slash ) {
 461                                                 # Closing a tag... is it the one we just opened?
 462                                                 $ot = @array_pop( $tagstack );
 463                                                 if ( $ot != $t ) {
 464                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 465                                                                 # Pop all elements with an optional close tag
 466                                                                 # and see if we find a match below them
 467                                                                 $optstack = array();
 468                                                                 array_push( $optstack, $ot );
 469                                                                 wfSuppressWarnings();
 470                                                                 $ot = array_pop( $tagstack );
 471                                                                 wfRestoreWarnings();
 472                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 473                                                                         array_push( $optstack, $ot );
 474                                                                         wfSuppressWarnings();
 475                                                                         $ot = array_pop( $tagstack );
 476                                                                         wfRestoreWarnings();
 477                                                                 }
 478                                                                 if ( $t != $ot ) {
 479                                                                         # No match. Push the optional elements back again
 480                                                                         $badtag = true;
 481                                                                         wfSuppressWarnings();
 482                                                                         $ot = array_pop( $optstack );
 483                                                                         wfRestoreWarnings();
 484                                                                         while ( $ot ) {
 485                                                                                 array_push( $tagstack, $ot );
 486                                                                                 wfSuppressWarnings();
 487                                                                                 $ot = array_pop( $optstack );
 488                                                                                 wfRestoreWarnings();
 489                                                                         }
 490                                                                 }
 491                                                         } else {
 492                                                                 @array_push( $tagstack, $ot );
 493                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 494                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 495                                                                         $badtag = true;
 496                                                                 }
 497                                                         }
 498                                                 } else {
 499                                                         if ( $t == 'table' ) {
 500                                                                 $tagstack = array_pop( $tablestack );
 501                                                         }
 502                                                 }
 503                                                 $newparams = '';
 504                                         } else {
 505                                                 # Keep track for later
 506                                                 if ( isset( $tabletags[$t] ) &&
 507                                                 !in_array( 'table', $tagstack ) ) {
 508                                                         $badtag = true;
 509                                                 } elseif ( in_array( $t, $tagstack ) &&
 510                                                 !isset( $htmlnest[$t] ) ) {
 511                                                         $badtag = true;
 512                                                 # Is it a self closed htmlpair ? (bug 5487)
 513                                                 } elseif ( $brace == '/>' &&
 514                                                 isset( $htmlpairs[$t] ) ) {
 515                                                         $badtag = true;
 516                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 517                                                         # Hack to force empty tag for unclosable elements
 518                                                         $brace = '/>';
 519                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 520                                                         # Hack to not close $htmlsingle tags
 521                                                         $brace = null;
 522                                                         # Still need to push this optionally-closed tag to
 523                                                         # the tag stack so that we can match end tags
 524                                                         # instead of marking them as bad.
 525                                                         array_push( $tagstack, $t );
 526                                                 } elseif ( isset( $tabletags[$t] )
 527                                                 && in_array( $t, $tagstack ) ) {
 528                                                         // New table tag but forgot to close the previous one
 529                                                         $text .= "</$t>";
 530                                                 } else {
 531                                                         if ( $t == 'table' ) {
 532                                                                 array_push( $tablestack, $tagstack );
 533                                                                 $tagstack = array();
 534                                                         }
 535                                                         array_push( $tagstack, $t );
 536                                                 }
 537
 538                                                 # Replace any variables or template parameters with
 539                                                 # plaintext results.
 540                                                 if ( is_callable( $processCallback ) ) {
 541                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 542                                                 }
 543
 544                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 545                                                         $badtag = true;
 546                                                 }
 547
 548                                                 # Strip non-approved attributes from the tag
 549                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 550                                         }
 551                                         if ( !$badtag ) {
 552                                                 $rest = str_replace( '>', '&gt;', $rest );
 553                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 554                                                 $text .= "<$slash$t$newparams$close>$rest";
 555                                                 continue;
 556                                         }
 557                                 }
 558                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 559                         }
 560                         # Close off any remaining tags
 561                         while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
 562                                 $text .= "</$t>\n";
 563                                 if ( $t == 'table' ) {
 564                                         $tagstack = array_pop( $tablestack );
 565                                 }
 566                         }
 567                 } else {
 568                         # this might be possible using tidy itself
 569                         foreach ( $bits as $x ) {
 570                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 571                                 $x, $regs );
 572                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 573                                 $badtag = false;
 574                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 575                                         if ( is_callable( $processCallback ) ) {
 576                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 577                                         }
 578
 579                                         if ( !Sanitizer::validateTag( $params, $t ) ) {
 580                                                 $badtag = true;
 581                                         }
 582
 583                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 584                                         if ( !$badtag ) {
 585                                                 $rest = str_replace( '>', '&gt;', $rest );
 586                                                 $text .= "<$slash$t$newparams$brace$rest";
 587                                                 continue;
 588                                         }
 589                                 }
 590                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 591                         }
 592                 }
 593                 wfProfileOut( __METHOD__ );
 594                 return $text;
 595         }
 596
 597         /**
 598          * Remove '<!--', '-->', and everything between.
 599          * To avoid leaving blank lines, when a comment is both preceded
 600          * and followed by a newline (ignoring spaces), trim leading and
 601          * trailing spaces and one of the newlines.
 602          *
 603          * @private
 604          * @param string $text
 605          * @return string
 606          */
 607         static function removeHTMLcomments( $text ) {
 608                 wfProfileIn( __METHOD__ );
 609                 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
 610                         $end = strpos( $text, '-->', $start + 4 );
 611                         if ( $end === false ) {
 612                                 # Unterminated comment; bail out
 613                                 break;
 614                         }
 615
 616                         $end += 3;
 617
 618                         # Trim space and newline if the comment is both
 619                         # preceded and followed by a newline
 620                         $spaceStart = max( $start - 1, 0 );
 621                         $spaceLen = $end - $spaceStart;
 622                         while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
 623                                 $spaceStart--;
 624                                 $spaceLen++;
 625                         }
 626                         while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
 627                                 $spaceLen++;
 628                         }
 629                         if ( substr( $text, $spaceStart, 1 ) === "\n"
 630                                 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
 631                                 # Remove the comment, leading and trailing
 632                                 # spaces, and leave only one newline.
 633                                 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
 634                         } else {
 635                                 # Remove just the comment.
 636                                 $text = substr_replace( $text, '', $start, $end - $start );
 637                         }
 638                 }
 639                 wfProfileOut( __METHOD__ );
 640                 return $text;
 641         }
 642
 643         /**
 644          * Takes attribute names and values for a tag and the tag name and
 645          * validates that the tag is allowed to be present.
 646          * This DOES NOT validate the attributes, nor does it validate the
 647          * tags themselves. This method only handles the special circumstances
 648          * where we may want to allow a tag within content but ONLY when it has
 649          * specific attributes set.
 650          *
 651          * @param string $params
 652          * @param string $element
 653          * @return bool
 654          */
 655         static function validateTag( $params, $element ) {
 656                 $params = Sanitizer::decodeTagAttributes( $params );
 657
 658                 if ( $element == 'meta' || $element == 'link' ) {
 659                         if ( !isset( $params['itemprop'] ) ) {
 660                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 661                                 return false;
 662                         }
 663                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 664                                 // <meta> must have a content="" for the itemprop
 665                                 return false;
 666                         }
 667                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 668                                 // <link> must have an associated href=""
 669                                 return false;
 670                         }
 671                 }
 672
 673                 return true;
 674         }
 675
 676         /**
 677          * Take an array of attribute names and values and normalize or discard
 678          * illegal values for the given element type.
 679          *
 680          * - Discards attributes not on a whitelist for the given element
 681          * - Unsafe style attributes are discarded
 682          * - Invalid id attributes are re-encoded
 683          *
 684          * @param array $attribs
 685          * @param string $element
 686          * @return array
 687          *
 688          * @todo Check for legal values where the DTD limits things.
 689          * @todo Check for unique id attribute :P
 690          */
 691         static function validateTagAttributes( $attribs, $element ) {
 692                 return Sanitizer::validateAttributes( $attribs,
 693                         Sanitizer::attributeWhitelist( $element ) );
 694         }
 695
 696         /**
 697          * Take an array of attribute names and values and normalize or discard
 698          * illegal values for the given whitelist.
 699          *
 700          * - Discards attributes not the given whitelist
 701          * - Unsafe style attributes are discarded
 702          * - Invalid id attributes are re-encoded
 703          *
 704          * @param array $attribs
 705          * @param array $whitelist list of allowed attribute names
 706          * @return array
 707          *
 708          * @todo Check for legal values where the DTD limits things.
 709          * @todo Check for unique id attribute :P
 710          */
 711         static function validateAttributes( $attribs, $whitelist ) {
 712                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
 713
 714                 $whitelist = array_flip( $whitelist );
 715                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 716
 717                 $out = array();
 718                 foreach ( $attribs as $attribute => $value ) {
 719                         #allow XML namespace declaration if RDFa is enabled
 720                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 721                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 722                                         $out[$attribute] = $value;
 723                                 }
 724
 725                                 continue;
 726                         }
 727
 728                         # Allow any attribute beginning with "data-"
 729                         if ( !preg_match( '/^data-/i', $attribute ) && !isset( $whitelist[$attribute] ) ) {
 730                                 continue;
 731                         }
 732
 733                         # Strip javascript "expression" from stylesheets.
 734                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 735                         if ( $attribute == 'style' ) {
 736                                 $value = Sanitizer::checkCss( $value );
 737                         }
 738
 739                         if ( $attribute === 'id' ) {
 740                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 741                         }
 742
 743                         # WAI-ARIA
 744                         # http://www.w3.org/TR/wai-aria/
 745                         # http://www.whatwg.org/html/elements.html#wai-aria
 746                         # For now we only support role="presentation" until we work out what roles should be
 747                         # usable by content and we ensure that our code explicitly rejects patterns that
 748                         # violate HTML5's ARIA restrictions.
 749                         if ( $attribute === 'role' && $value !== 'presentation' ) {
 750                                 continue;
 751                         }
 752
 753                         // RDFa and microdata properties allow URLs, URIs and/or CURIs.
 754                         // Check them for sanity.
 755                         if ( $attribute === 'rel' || $attribute === 'rev'
 756                                 # RDFa
 757                                 || $attribute === 'about' || $attribute === 'property'
 758                                 || $attribute === 'resource' || $attribute === 'datatype'
 759                                 || $attribute === 'typeof'
 760                                 # HTML5 microdata
 761                                 || $attribute === 'itemid' || $attribute === 'itemprop'
 762                                 || $attribute === 'itemref' || $attribute === 'itemscope'
 763                                 || $attribute === 'itemtype'
 764                         ) {
 765                                 //Paranoia. Allow "simple" values but suppress javascript
 766                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 767                                         continue;
 768                                 }
 769                         }
 770
 771                         # NOTE: even though elements using href/src are not allowed directly, supply
 772                         #       validation code that can be used by tag hook handlers, etc
 773                         if ( $attribute === 'href' || $attribute === 'src' ) {
 774                                 if ( !preg_match( $hrefExp, $value ) ) {
 775                                         continue; //drop any href or src attributes not using an allowed protocol.
 776                                         // NOTE: this also drops all relative URLs
 777                                 }
 778                         }
 779
 780                         // If this attribute was previously set, override it.
 781                         // Output should only have one attribute of each name.
 782                         $out[$attribute] = $value;
 783                 }
 784
 785                 if ( $wgAllowMicrodataAttributes ) {
 786                         # itemtype, itemid, itemref don't make sense without itemscope
 787                         if ( !array_key_exists( 'itemscope', $out ) ) {
 788                                 unset( $out['itemtype'] );
 789                                 unset( $out['itemid'] );
 790                                 unset( $out['itemref'] );
 791                         }
 792                         # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 793                 }
 794                 return $out;
 795         }
 796
 797         /**
 798          * Merge two sets of HTML attributes.  Conflicting items in the second set
 799          * will override those in the first, except for 'class' attributes which
 800          * will be combined (if they're both strings).
 801          *
 802          * @todo implement merging for other attributes such as style
 803          * @param array $a
 804          * @param array $b
 805          * @return array
 806          */
 807         static function mergeAttributes( $a, $b ) {
 808                 $out = array_merge( $a, $b );
 809                 if ( isset( $a['class'] ) && isset( $b['class'] )
 810                         && is_string( $a['class'] ) && is_string( $b['class'] )
 811                         && $a['class'] !== $b['class']
 812                 ) {
 813                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 814                                 -1, PREG_SPLIT_NO_EMPTY );
 815                         $out['class'] = implode( ' ', array_unique( $classes ) );
 816                 }
 817                 return $out;
 818         }
 819
 820         /**
 821          * Pick apart some CSS and check it for forbidden or unsafe structures.
 822          * Returns a sanitized string. This sanitized string will have
 823          * character references and escape sequences decoded and comments
 824          * stripped (unless it is itself one valid comment, in which case the value
 825          * will be passed through). If the input is just too evil, only a comment
 826          * complaining about evilness will be returned.
 827          *
 828          * Currently URL references, 'expression', 'tps' are forbidden.
 829          *
 830          * NOTE: Despite the fact that character references are decoded, the
 831          * returned string may contain character references given certain
 832          * clever input strings. These character references must
 833          * be escaped before the return value is embedded in HTML.
 834          *
 835          * @param string $value
 836          * @return string
 837          */
 838         static function checkCss( $value ) {
 839                 // Decode character references like &#123;
 840                 $value = Sanitizer::decodeCharReferences( $value );
 841
 842                 // Decode escape sequences and line continuation
 843                 // See the grammar in the CSS 2 spec, appendix D.
 844                 // This has to be done AFTER decoding character references.
 845                 // This means it isn't possible for this function to return
 846                 // unsanitized escape sequences. It is possible to manufacture
 847                 // input that contains character references that decode to
 848                 // escape sequences that decode to character references, but
 849                 // it's OK for the return value to contain character references
 850                 // because the caller is supposed to escape those anyway.
 851                 static $decodeRegex;
 852                 if ( !$decodeRegex ) {
 853                         $space = '[\\x20\\t\\r\\n\\f]';
 854                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 855                         $backslash = '\\\\';
 856                         $decodeRegex = "/ $backslash
 857                                 (?:
 858                                         ($nl) |  # 1. Line continuation
 859                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 860                                         (.) | # 3. backslash cancelling special meaning
 861                                         () | # 4. backslash at end of string
 862                                 )/xu";
 863                 }
 864                 $value = preg_replace_callback( $decodeRegex,
 865                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 866
 867                 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
 868                 $value = preg_replace_callback(
 869                         '/[！-［］-ｚ]/u', // U+FF01 to U+FF5A, excluding U+FF3C (bug 58088)
 870                         function ( $matches ) {
 871                                 $cp = utf8ToCodepoint( $matches[0] );
 872                                 if ( $cp === false ) {
 873                                         return '';
 874                                 }
 875                                 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
 876                         },
 877                         $value
 878                 );
 879
 880                 // Convert more characters IE6 might treat as ascii
 881                 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
 882                 $value = str_replace(
 883                         array( 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ),
 884                         array( 'r', 'n', 'n', 'l', 'i', '(', '(' ),
 885                         $value
 886                 );
 887
 888                 // Let the value through if it's nothing but a single comment, to
 889                 // allow other functions which may reject it to pass some error
 890                 // message through.
 891                 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
 892                         // Remove any comments; IE gets token splitting wrong
 893                         // This must be done AFTER decoding character references and
 894                         // escape sequences, because those steps can introduce comments
 895                         // This step cannot introduce character references or escape
 896                         // sequences, because it replaces comments with spaces rather
 897                         // than removing them completely.
 898                         $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 899
 900                         // Remove anything after a comment-start token, to guard against
 901                         // incorrect client implementations.
 902                         $commentPos = strpos( $value, '/*' );
 903                         if ( $commentPos !== false ) {
 904                                 $value = substr( $value, 0, $commentPos );
 905                         }
 906                 }
 907
 908                 // S followed by repeat, iteration, or prolonged sound marks,
 909                 // which IE will treat as "ss"
 910                 $value = preg_replace(
 911                         '/s(?:
 912                                 \xE3\x80\xB1 | # U+3031
 913                                 \xE3\x82\x9D | # U+309D
 914                                 \xE3\x83\xBC | # U+30FC
 915                                 \xE3\x83\xBD | # U+30FD
 916                                 \xEF\xB9\xBC | # U+FE7C
 917                                 \xEF\xB9\xBD | # U+FE7D
 918                                 \xEF\xBD\xB0   # U+FF70
 919                         )/ix',
 920                         'ss',
 921                         $value
 922                 );
 923
 924                 // Reject problematic keywords and control characters
 925                 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ) {
 926                         return '/* invalid control char */';
 927                 } elseif ( preg_match(
 928                         '! expression
 929                                 | filter\s*:
 930                                 | accelerator\s*:
 931                                 | -o-link\s*:
 932                                 | -o-link-source\s*:
 933                                 | -o-replace\s*:
 934                                 | url\s*\(
 935                                 | image\s*\(
 936                                 | image-set\s*\(
 937                         !ix', $value ) ) {
 938                         return '/* insecure input */';
 939                 }
 940                 return $value;
 941         }
 942
 943         /**
 944          * @param array $matches
 945          * @return string
 946          */
 947         static function cssDecodeCallback( $matches ) {
 948                 if ( $matches[1] !== '' ) {
 949                         // Line continuation
 950                         return '';
 951                 } elseif ( $matches[2] !== '' ) {
 952                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 953                 } elseif ( $matches[3] !== '' ) {
 954                         $char = $matches[3];
 955                 } else {
 956                         $char = '\\';
 957                 }
 958                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 959                         // These characters need to be escaped in strings
 960                         // Clean up the escape sequence to avoid parsing errors by clients
 961                         return '\\' . dechex( ord( $char ) ) . ' ';
 962                 } else {
 963                         // Decode unnecessary escape
 964                         return $char;
 965                 }
 966         }
 967
 968         /**
 969          * Take a tag soup fragment listing an HTML element's attributes
 970          * and normalize it to well-formed XML, discarding unwanted attributes.
 971          * Output is safe for further wikitext processing, with escaping of
 972          * values that could trigger problems.
 973          *
 974          * - Normalizes attribute names to lowercase
 975          * - Discards attributes not on a whitelist for the given element
 976          * - Turns broken or invalid entities into plaintext
 977          * - Double-quotes all attribute values
 978          * - Attributes without values are given the name as attribute
 979          * - Double attributes are discarded
 980          * - Unsafe style attributes are discarded
 981          * - Prepends space if there are attributes.
 982          *
 983          * @param string $text
 984          * @param string $element
 985          * @return string
 986          */
 987         static function fixTagAttributes( $text, $element ) {
 988                 if ( trim( $text ) == '' ) {
 989                         return '';
 990                 }
 991
 992                 $decoded = Sanitizer::decodeTagAttributes( $text );
 993                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
 994
 995                 return Sanitizer::safeEncodeTagAttributes( $stripped );
 996         }
 997
 998         /**
 999          * Encode an attribute value for HTML output.
1000          * @param string $text
1001          * @return string HTML-encoded text fragment
1002          */
1003         static function encodeAttribute( $text ) {
1004                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1005
1006                 // Whitespace is normalized during attribute decoding,
1007                 // so if we've been passed non-spaces we must encode them
1008                 // ahead of time or they won't be preserved.
1009                 $encValue = strtr( $encValue, array(
1010                         "\n" => '&#10;',
1011                         "\r" => '&#13;',
1012                         "\t" => '&#9;',
1013                 ) );
1014
1015                 return $encValue;
1016         }
1017
1018         /**
1019          * Encode an attribute value for HTML tags, with extra armoring
1020          * against further wiki processing.
1021          * @param string $text
1022          * @return string HTML-encoded text fragment
1023          */
1024         static function safeEncodeAttribute( $text ) {
1025                 $encValue = Sanitizer::encodeAttribute( $text );
1026
1027                 # Templates and links may be expanded in later parsing,
1028                 # creating invalid or dangerous output. Suppress this.
1029                 $encValue = strtr( $encValue, array(
1030                         '<'    => '&lt;',   // This should never happen,
1031                         '>'    => '&gt;',   // we've received invalid input
1032                         '"'    => '&quot;', // which should have been escaped.
1033                         '{'    => '&#123;',
1034                         '['    => '&#91;',
1035                         "''"   => '&#39;&#39;',
1036                         'ISBN' => '&#73;SBN',
1037                         'RFC'  => '&#82;FC',
1038                         'PMID' => '&#80;MID',
1039                         '|'    => '&#124;',
1040                         '__'   => '&#95;_',
1041                 ) );
1042
1043                 # Stupid hack
1044                 $encValue = preg_replace_callback(
1045                         '/((?i)' . wfUrlProtocols() . ')/',
1046                         array( 'Sanitizer', 'armorLinksCallback' ),
1047                         $encValue );
1048                 return $encValue;
1049         }
1050
1051         /**
1052          * Given a value, escape it so that it can be used in an id attribute and
1053          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1054          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1055          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1056          * escaped with lots of dots.
1057          *
1058          * To ensure we don't have to bother escaping anything, we also strip ', ",
1059          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1060          * We also strip # because it upsets IE, and % because it could be
1061          * ambiguous if it's part of something that looks like a percent escape
1062          * (which don't work reliably in fragments cross-browser).
1063          *
1064          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1065          *   in the id and name attributes
1066          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1067          *   the id attribute
1068          * @see http://www.whatwg.org/html/elements.html#the-id-attribute
1069          *   HTML5 definition of id attribute
1070          *
1071          * @param string $id id to escape
1072          * @param string|array $options String or array of strings (default is array()):
1073          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1074          *       so don't pay attention if the first character isn't valid at the
1075          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1076          *       false.
1077          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1078          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1079          *       anchors and links won't break.
1080          * @return string
1081          */
1082         static function escapeId( $id, $options = array() ) {
1083                 global $wgExperimentalHtmlIds;
1084                 $options = (array)$options;
1085
1086                 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1087                         $id = Sanitizer::decodeCharReferences( $id );
1088                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1089                         $id = trim( $id, '_' );
1090                         if ( $id === '' ) {
1091                                 # Must have been all whitespace to start with.
1092                                 return '_';
1093                         } else {
1094                                 return $id;
1095                         }
1096                 }
1097
1098                 # HTML4-style escaping
1099                 static $replace = array(
1100                         '%3A' => ':',
1101                         '%' => '.'
1102                 );
1103
1104                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1105                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1106
1107                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1108                 && !in_array( 'noninitial', $options ) ) {
1109                         // Initial character must be a letter!
1110                         $id = "x$id";
1111                 }
1112                 return $id;
1113         }
1114
1115         /**
1116          * Given a value, escape it so that it can be used as a CSS class and
1117          * return it.
1118          *
1119          * @todo For extra validity, input should be validated UTF-8.
1120          *
1121          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1122          *
1123          * @param string $class
1124          * @return string
1125          */
1126         static function escapeClass( $class ) {
1127                 // Convert ugly stuff to underscores and kill underscores in ugly places
1128                 return rtrim( preg_replace(
1129                         array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1130                         '_',
1131                         $class ), '_' );
1132         }
1133
1134         /**
1135          * Given HTML input, escape with htmlspecialchars but un-escape entities.
1136          * This allows (generally harmless) entities like &#160; to survive.
1137          *
1138          * @param string $html HTML to escape
1139          * @return string Escaped input
1140          */
1141         static function escapeHtmlAllowEntities( $html ) {
1142                 $html = Sanitizer::decodeCharReferences( $html );
1143                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1144                 # hurt.
1145                 $html = htmlspecialchars( $html, ENT_QUOTES );
1146                 return $html;
1147         }
1148
1149         /**
1150          * Regex replace callback for armoring links against further processing.
1151          * @param array $matches
1152          * @return string
1153          */
1154         private static function armorLinksCallback( $matches ) {
1155                 return str_replace( ':', '&#58;', $matches[1] );
1156         }
1157
1158         /**
1159          * Return an associative array of attribute names and values from
1160          * a partial tag string. Attribute names are forces to lowercase,
1161          * character references are decoded to UTF-8 text.
1162          *
1163          * @param string $text
1164          * @return array
1165          */
1166         public static function decodeTagAttributes( $text ) {
1167                 if ( trim( $text ) == '' ) {
1168                         return array();
1169                 }
1170
1171                 $attribs = array();
1172                 $pairs = array();
1173                 if ( !preg_match_all(
1174                         self::getAttribsRegex(),
1175                         $text,
1176                         $pairs,
1177                         PREG_SET_ORDER ) ) {
1178                         return $attribs;
1179                 }
1180
1181                 foreach ( $pairs as $set ) {
1182                         $attribute = strtolower( $set[1] );
1183                         $value = Sanitizer::getTagAttributeCallback( $set );
1184
1185                         // Normalize whitespace
1186                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1187                         $value = trim( $value );
1188
1189                         // Decode character references
1190                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1191                 }
1192                 return $attribs;
1193         }
1194
1195         /**
1196          * Build a partial tag string from an associative array of attribute
1197          * names and values as returned by decodeTagAttributes.
1198          *
1199          * @param array $assoc_array
1200          * @return string
1201          */
1202         public static function safeEncodeTagAttributes( $assoc_array ) {
1203                 $attribs = array();
1204                 foreach ( $assoc_array as $attribute => $value ) {
1205                         $encAttribute = htmlspecialchars( $attribute );
1206                         $encValue = Sanitizer::safeEncodeAttribute( $value );
1207
1208                         $attribs[] = "$encAttribute=\"$encValue\"";
1209                 }
1210                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1211         }
1212
1213         /**
1214          * Pick the appropriate attribute value from a match set from the
1215          * attribs regex matches.
1216          *
1217          * @param array $set
1218          * @throws MWException when tag conditions are not met.
1219          * @return string
1220          */
1221         private static function getTagAttributeCallback( $set ) {
1222                 if ( isset( $set[6] ) ) {
1223                         # Illegal #XXXXXX color with no quotes.
1224                         return $set[6];
1225                 } elseif ( isset( $set[5] ) ) {
1226                         # No quotes.
1227                         return $set[5];
1228                 } elseif ( isset( $set[4] ) ) {
1229                         # Single-quoted
1230                         return $set[4];
1231                 } elseif ( isset( $set[3] ) ) {
1232                         # Double-quoted
1233                         return $set[3];
1234                 } elseif ( !isset( $set[2] ) ) {
1235                         # In XHTML, attributes must have a value.
1236                         # For 'reduced' form, return explicitly the attribute name here.
1237                         return $set[1];
1238                 } else {
1239                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1240                 }
1241         }
1242
1243         /**
1244          * Normalize whitespace and character references in an XML source-
1245          * encoded text for an attribute value.
1246          *
1247          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1248          * but note that we're not returning the value, but are returning
1249          * XML source fragments that will be slapped into output.
1250          *
1251          * @param string $text
1252          * @return string
1253          * @todo Remove, unused?
1254          */
1255         private static function normalizeAttributeValue( $text ) {
1256                 return str_replace( '"', '&quot;',
1257                         self::normalizeWhitespace(
1258                                 Sanitizer::normalizeCharReferences( $text ) ) );
1259         }
1260
1261         /**
1262          * @param string $text
1263          * @return string
1264          */
1265         private static function normalizeWhitespace( $text ) {
1266                 return preg_replace(
1267                         '/\r\n|[\x20\x0d\x0a\x09]/',
1268                         ' ',
1269                         $text );
1270         }
1271
1272         /**
1273          * Normalizes whitespace in a section name, such as might be returned
1274          * by Parser::stripSectionName(), for use in the id's that are used for
1275          * section links.
1276          *
1277          * @param string $section
1278          * @return string
1279          */
1280         static function normalizeSectionNameWhitespace( $section ) {
1281                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1282         }
1283
1284         /**
1285          * Ensure that any entities and character references are legal
1286          * for XML and XHTML specifically. Any stray bits will be
1287          * &amp;-escaped to result in a valid text fragment.
1288          *
1289          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1290          *   numericized (this way we're well-formed even without a DTD)
1291          * b. any numeric char refs must be legal chars, not invalid or forbidden
1292          * c. use lower cased "&#x", not "&#X"
1293          * d. fix or reject non-valid attributes
1294          *
1295          * @param string $text
1296          * @return string
1297          * @private
1298          */
1299         static function normalizeCharReferences( $text ) {
1300                 return preg_replace_callback(
1301                         self::CHAR_REFS_REGEX,
1302                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1303                         $text );
1304         }
1305
1306         /**
1307          * @param string $matches
1308          * @return string
1309          */
1310         static function normalizeCharReferencesCallback( $matches ) {
1311                 $ret = null;
1312                 if ( $matches[1] != '' ) {
1313                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1314                 } elseif ( $matches[2] != '' ) {
1315                         $ret = Sanitizer::decCharReference( $matches[2] );
1316                 } elseif ( $matches[3] != '' ) {
1317                         $ret = Sanitizer::hexCharReference( $matches[3] );
1318                 }
1319                 if ( is_null( $ret ) ) {
1320                         return htmlspecialchars( $matches[0] );
1321                 } else {
1322                         return $ret;
1323                 }
1324         }
1325
1326         /**
1327          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1328          * return the equivalent numeric entity reference (except for the core &lt;
1329          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1330          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1331          * pseudo-entity source (eg &amp;foo;)
1332          *
1333          * @param string $name
1334          * @return string
1335          */
1336         static function normalizeEntity( $name ) {
1337                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1338                         return '&' . self::$htmlEntityAliases[$name] . ';';
1339                 } elseif ( in_array( $name,
1340                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1341                         return "&$name;";
1342                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1343                         return '&#' . self::$htmlEntities[$name] . ';';
1344                 } else {
1345                         return "&amp;$name;";
1346                 }
1347         }
1348
1349         /**
1350          * @param int $codepoint
1351          * @return null|string
1352          */
1353         static function decCharReference( $codepoint ) {
1354                 $point = intval( $codepoint );
1355                 if ( Sanitizer::validateCodepoint( $point ) ) {
1356                         return sprintf( '&#%d;', $point );
1357                 } else {
1358                         return null;
1359                 }
1360         }
1361
1362         /**
1363          * @param int $codepoint
1364          * @return null|string
1365          */
1366         static function hexCharReference( $codepoint ) {
1367                 $point = hexdec( $codepoint );
1368                 if ( Sanitizer::validateCodepoint( $point ) ) {
1369                         return sprintf( '&#x%x;', $point );
1370                 } else {
1371                         return null;
1372                 }
1373         }
1374
1375         /**
1376          * Returns true if a given Unicode codepoint is a valid character in XML.
1377          * @param int $codepoint
1378          * @return bool
1379          */
1380         private static function validateCodepoint( $codepoint ) {
1381                 return $codepoint == 0x09
1382                         || $codepoint == 0x0a
1383                         || $codepoint == 0x0d
1384                         || ( $codepoint >= 0x20 && $codepoint <= 0xd7ff )
1385                         || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1386                         || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1387         }
1388
1389         /**
1390          * Decode any character references, numeric or named entities,
1391          * in the text and return a UTF-8 string.
1392          *
1393          * @param string $text
1394          * @return string
1395          */
1396         public static function decodeCharReferences( $text ) {
1397                 return preg_replace_callback(
1398                         self::CHAR_REFS_REGEX,
1399                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1400                         $text );
1401         }
1402
1403         /**
1404          * Decode any character references, numeric or named entities,
1405          * in the next and normalize the resulting string. (bug 14952)
1406          *
1407          * This is useful for page titles, not for text to be displayed,
1408          * MediaWiki allows HTML entities to escape normalization as a feature.
1409          *
1410          * @param string $text Already normalized, containing entities
1411          * @return string Still normalized, without entities
1412          */
1413         public static function decodeCharReferencesAndNormalize( $text ) {
1414                 global $wgContLang;
1415                 $text = preg_replace_callback(
1416                         self::CHAR_REFS_REGEX,
1417                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1418                         $text, /* limit */ -1, $count );
1419
1420                 if ( $count ) {
1421                         return $wgContLang->normalize( $text );
1422                 } else {
1423                         return $text;
1424                 }
1425         }
1426
1427         /**
1428          * @param string $matches
1429          * @return string
1430          */
1431         static function decodeCharReferencesCallback( $matches ) {
1432                 if ( $matches[1] != '' ) {
1433                         return Sanitizer::decodeEntity( $matches[1] );
1434                 } elseif ( $matches[2] != '' ) {
1435                         return Sanitizer::decodeChar( intval( $matches[2] ) );
1436                 } elseif ( $matches[3] != '' ) {
1437                         return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1438                 }
1439                 # Last case should be an ampersand by itself
1440                 return $matches[0];
1441         }
1442
1443         /**
1444          * Return UTF-8 string for a codepoint if that is a valid
1445          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1446          * @param int $codepoint
1447          * @return string
1448          * @private
1449          */
1450         static function decodeChar( $codepoint ) {
1451                 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1452                         return codepointToUtf8( $codepoint );
1453                 } else {
1454                         return UTF8_REPLACEMENT;
1455                 }
1456         }
1457
1458         /**
1459          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1460          * return the UTF-8 encoding of that character. Otherwise, returns
1461          * pseudo-entity source (eg "&foo;")
1462          *
1463          * @param string $name
1464          * @return string
1465          */
1466         static function decodeEntity( $name ) {
1467                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1468                         $name = self::$htmlEntityAliases[$name];
1469                 }
1470                 if ( isset( self::$htmlEntities[$name] ) ) {
1471                         return codepointToUtf8( self::$htmlEntities[$name] );
1472                 } else {
1473                         return "&$name;";
1474                 }
1475         }
1476
1477         /**
1478          * Fetch the whitelist of acceptable attributes for a given element name.
1479          *
1480          * @param string $element
1481          * @return array
1482          */
1483         static function attributeWhitelist( $element ) {
1484                 $list = Sanitizer::setupAttributeWhitelist();
1485                 return isset( $list[$element] )
1486                         ? $list[$element]
1487                         : array();
1488         }
1489
1490         /**
1491          * Foreach array key (an allowed HTML element), return an array
1492          * of allowed attributes
1493          * @return array
1494          */
1495         static function setupAttributeWhitelist() {
1496                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
1497
1498                 static $whitelist, $staticInitialised;
1499                 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgAllowMicrodataAttributes' ) );
1500
1501                 if ( isset( $whitelist ) && $staticInitialised == $globalContext ) {
1502                         return $whitelist;
1503                 }
1504
1505                 $common = array(
1506                         # HTML
1507                         'id',
1508                         'class',
1509                         'style',
1510                         'lang',
1511                         'dir',
1512                         'title',
1513
1514                         # WAI-ARIA
1515                         'role',
1516                 );
1517
1518                 if ( $wgAllowRdfaAttributes ) {
1519                         # RDFa attributes as specified in section 9 of
1520                         # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1521                         $common = array_merge( $common, array(
1522                                 'about', 'property', 'resource', 'datatype', 'typeof',
1523                         ) );
1524                 }
1525
1526                 if ( $wgAllowMicrodataAttributes ) {
1527                         # add HTML5 microdata tags as specified by
1528                         # http://www.whatwg.org/html/microdata.html#the-microdata-model
1529                         $common = array_merge( $common, array(
1530                                 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1531                         ) );
1532                 }
1533
1534                 $block = array_merge( $common, array( 'align' ) );
1535                 $tablealign = array( 'align', 'valign' );
1536                 $tablecell = array(
1537                         'abbr',
1538                         'axis',
1539                         'headers',
1540                         'scope',
1541                         'rowspan',
1542                         'colspan',
1543                         'nowrap', # deprecated
1544                         'width', # deprecated
1545                         'height', # deprecated
1546                         'bgcolor', # deprecated
1547                 );
1548
1549                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1550                 # See: http://www.w3.org/TR/html4/
1551                 $whitelist = array(
1552                         # 7.5.4
1553                         'div'        => $block,
1554                         'center'     => $common, # deprecated
1555                         'span'       => $common,
1556
1557                         # 7.5.5
1558                         'h1'         => $block,
1559                         'h2'         => $block,
1560                         'h3'         => $block,
1561                         'h4'         => $block,
1562                         'h5'         => $block,
1563                         'h6'         => $block,
1564
1565                         # 7.5.6
1566                         # address
1567
1568                         # 8.2.4
1569                         'bdo'        => $common,
1570
1571                         # 9.2.1
1572                         'em'         => $common,
1573                         'strong'     => $common,
1574                         'cite'       => $common,
1575                         'dfn'        => $common,
1576                         'code'       => $common,
1577                         'samp'       => $common,
1578                         'kbd'        => $common,
1579                         'var'        => $common,
1580                         'abbr'       => $common,
1581                         # acronym
1582
1583                         # 9.2.2
1584                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1585                         'q'          => array_merge( $common, array( 'cite' ) ),
1586
1587                         # 9.2.3
1588                         'sub'        => $common,
1589                         'sup'        => $common,
1590
1591                         # 9.3.1
1592                         'p'          => $block,
1593
1594                         # 9.3.2
1595                         'br'         => array_merge( $common, array( 'clear' ) ),
1596
1597                         # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1598                         'wbr'        => $common,
1599
1600                         # 9.3.4
1601                         'pre'        => array_merge( $common, array( 'width' ) ),
1602
1603                         # 9.4
1604                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1605                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1606
1607                         # 10.2
1608                         'ul'         => array_merge( $common, array( 'type' ) ),
1609                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1610                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1611
1612                         # 10.3
1613                         'dl'         => $common,
1614                         'dd'         => $common,
1615                         'dt'         => $common,
1616
1617                         # 11.2.1
1618                         'table'      => array_merge( $common,
1619                                                                 array( 'summary', 'width', 'border', 'frame',
1620                                                                                 'rules', 'cellspacing', 'cellpadding',
1621                                                                                 'align', 'bgcolor',
1622                                                                 ) ),
1623
1624                         # 11.2.2
1625                         'caption'    => $block,
1626
1627                         # 11.2.3
1628                         'thead'      => $common,
1629                         'tfoot'      => $common,
1630                         'tbody'      => $common,
1631
1632                         # 11.2.4
1633                         'colgroup'   => array_merge( $common, array( 'span' ) ),
1634                         'col'        => array_merge( $common, array( 'span' ) ),
1635
1636                         # 11.2.5
1637                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1638
1639                         # 11.2.6
1640                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1641                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1642
1643                         # 12.2
1644                         # NOTE: <a> is not allowed directly, but the attrib
1645                         # whitelist is used from the Parser object
1646                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1647
1648                         # 13.2
1649                         # Not usually allowed, but may be used for extension-style hooks
1650                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1651                         # true
1652                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1653
1654                         # 15.2.1
1655                         'tt'         => $common,
1656                         'b'          => $common,
1657                         'i'          => $common,
1658                         'big'        => $common,
1659                         'small'      => $common,
1660                         'strike'     => $common,
1661                         's'          => $common,
1662                         'u'          => $common,
1663
1664                         # 15.2.2
1665                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1666                         # basefont
1667
1668                         # 15.3
1669                         'hr'         => array_merge( $common, array( 'width' ) ),
1670
1671                         # HTML Ruby annotation text module, simple ruby only.
1672                         # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
1673                         'ruby'       => $common,
1674                         # rbc
1675                         # rtc
1676                         'rb'         => $common,
1677                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1678                         'rp'         => $common,
1679
1680                         # MathML root element, where used for extensions
1681                         # 'title' may not be 100% valid here; it's XHTML
1682                         # http://www.w3.org/TR/REC-MathML/
1683                         'math'       => array( 'class', 'style', 'id', 'title' ),
1684
1685                         # HTML 5 section 4.6
1686                         'bdi' => $common,
1687
1688                         # HTML5 elements, defined by:
1689                         # http://www.whatwg.org/html/
1690                         'data' => array_merge( $common, array( 'value' ) ),
1691                         'time' => array_merge( $common, array( 'datetime' ) ),
1692                         'mark' => $common,
1693
1694                         // meta and link are only permitted by removeHTMLtags when Microdata
1695                         // is enabled so we don't bother adding a conditional to hide these
1696                         // Also meta and link are only valid in WikiText as Microdata elements
1697                         // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1698                         // So we don't bother including $common attributes that have no purpose.
1699                         'meta' => array( 'itemprop', 'content' ),
1700                         'link' => array( 'itemprop', 'href' ),
1701                 );
1702
1703                 $staticInitialised = $globalContext;
1704
1705                 return $whitelist;
1706         }
1707
1708         /**
1709          * Take a fragment of (potentially invalid) HTML and return
1710          * a version with any tags removed, encoded as plain text.
1711          *
1712          * Warning: this return value must be further escaped for literal
1713          * inclusion in HTML output as of 1.10!
1714          *
1715          * @param string $text HTML fragment
1716          * @return string
1717          */
1718         static function stripAllTags( $text ) {
1719                 # Actual <tags>
1720                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1721
1722                 # Normalize &entities and whitespace
1723                 $text = self::decodeCharReferences( $text );
1724                 $text = self::normalizeWhitespace( $text );
1725
1726                 return $text;
1727         }
1728
1729         /**
1730          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1731          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1732          * PHP 5.1 doesn't.
1733          *
1734          * Use for passing XHTML fragments to PHP's XML parsing functions
1735          *
1736          * @return string
1737          */
1738         static function hackDocType() {
1739                 $out = "<!DOCTYPE html [\n";
1740                 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1741                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1742                 }
1743                 $out .= "]>\n";
1744                 return $out;
1745         }
1746
1747         /**
1748          * @param string $url
1749          * @return mixed|string
1750          */
1751         static function cleanUrl( $url ) {
1752                 # Normalize any HTML entities in input. They will be
1753                 # re-escaped by makeExternalLink().
1754                 $url = Sanitizer::decodeCharReferences( $url );
1755
1756                 # Escape any control characters introduced by the above step
1757                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1758                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1759
1760                 # Validate hostname portion
1761                 $matches = array();
1762                 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1763                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1764
1765                         // Characters that will be ignored in IDNs.
1766                         // http://tools.ietf.org/html/3454#section-3.1
1767                         // Strip them before further processing so blacklists and such work.
1768                         $strip = "/
1769                                 \\s|          # general whitespace
1770                                 \xc2\xad|     # 00ad SOFT HYPHEN
1771                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1772                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1773                                 \xe2\x81\xa0| # 2060 WORD JOINER
1774                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1775                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1776                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1777                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1778                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1779                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1780                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1781                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1782                                 /xuD";
1783
1784                         $host = preg_replace( $strip, '', $host );
1785
1786                         // @todo FIXME: Validate hostnames here
1787
1788                         return $protocol . $host . $rest;
1789                 } else {
1790                         return $url;
1791                 }
1792         }
1793
1794         /**
1795          * @param array $matches
1796          * @return string
1797          */
1798         static function cleanUrlCallback( $matches ) {
1799                 return urlencode( $matches[0] );
1800         }
1801
1802         /**
1803          * Does a string look like an e-mail address?
1804          *
1805          * This validates an email address using an HTML5 specification found at:
1806          * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1807          * Which as of 2011-01-24 says:
1808          *
1809          *   A valid e-mail address is a string that matches the ABNF production
1810          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1811          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1812          *   3.5.
1813          *
1814          * This function is an implementation of the specification as requested in
1815          * bug 22449.
1816          *
1817          * Client-side forms will use the same standard validation rules via JS or
1818          * HTML 5 validation; additional restrictions can be enforced server-side
1819          * by extensions via the 'isValidEmailAddr' hook.
1820          *
1821          * Note that this validation doesn't 100% match RFC 2822, but is believed
1822          * to be liberal enough for wide use. Some invalid addresses will still
1823          * pass validation here.
1824          *
1825          * @since 1.18
1826          *
1827          * @param string $addr E-mail address
1828          * @return bool
1829          */
1830         public static function validateEmail( $addr ) {
1831                 $result = null;
1832                 if ( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1833                         return $result;
1834                 }
1835
1836                 // Please note strings below are enclosed in brackets [], this make the
1837                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1838                 // See bug 26948
1839                 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1840                 $rfc1034_ldh_str = "a-z0-9\\-";
1841
1842                 $html5_email_regexp = "/
1843                 ^                      # start of string
1844                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1845                 @                      # 'apostrophe'
1846                 [$rfc1034_ldh_str]+       # First domain part
1847                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1848                 $                      # End of string
1849                 /ix"; // case Insensitive, eXtended
1850
1851                 return (bool)preg_match( $html5_email_regexp, $addr );
1852         }
1853 }