includes/Sanitizer.php

   1 <?php
   2 /**
   3  * XHTML sanitizer for %MediaWiki.
   4  *
   5  * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
   6  * http://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  * XHTML 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          * @private
  58          */
  59         static $htmlEntities = array(
  60                 'Aacute'   => 193,
  61                 'aacute'   => 225,
  62                 'Acirc'    => 194,
  63                 'acirc'    => 226,
  64                 'acute'    => 180,
  65                 'AElig'    => 198,
  66                 'aelig'    => 230,
  67                 'Agrave'   => 192,
  68                 'agrave'   => 224,
  69                 'alefsym'  => 8501,
  70                 'Alpha'    => 913,
  71                 'alpha'    => 945,
  72                 'amp'      => 38,
  73                 'and'      => 8743,
  74                 'ang'      => 8736,
  75                 'apos'     => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
  76                 'Aring'    => 197,
  77                 'aring'    => 229,
  78                 'asymp'    => 8776,
  79                 'Atilde'   => 195,
  80                 'atilde'   => 227,
  81                 'Auml'     => 196,
  82                 'auml'     => 228,
  83                 'bdquo'    => 8222,
  84                 'Beta'     => 914,
  85                 'beta'     => 946,
  86                 'brvbar'   => 166,
  87                 'bull'     => 8226,
  88                 'cap'      => 8745,
  89                 'Ccedil'   => 199,
  90                 'ccedil'   => 231,
  91                 'cedil'    => 184,
  92                 'cent'     => 162,
  93                 'Chi'      => 935,
  94                 'chi'      => 967,
  95                 'circ'     => 710,
  96                 'clubs'    => 9827,
  97                 'cong'     => 8773,
  98                 'copy'     => 169,
  99                 'crarr'    => 8629,
 100                 'cup'      => 8746,
 101                 'curren'   => 164,
 102                 'dagger'   => 8224,
 103                 'Dagger'   => 8225,
 104                 'darr'     => 8595,
 105                 'dArr'     => 8659,
 106                 'deg'      => 176,
 107                 'Delta'    => 916,
 108                 'delta'    => 948,
 109                 'diams'    => 9830,
 110                 'divide'   => 247,
 111                 'Eacute'   => 201,
 112                 'eacute'   => 233,
 113                 'Ecirc'    => 202,
 114                 'ecirc'    => 234,
 115                 'Egrave'   => 200,
 116                 'egrave'   => 232,
 117                 'empty'    => 8709,
 118                 'emsp'     => 8195,
 119                 'ensp'     => 8194,
 120                 'Epsilon'  => 917,
 121                 'epsilon'  => 949,
 122                 'equiv'    => 8801,
 123                 'Eta'      => 919,
 124                 'eta'      => 951,
 125                 'ETH'      => 208,
 126                 'eth'      => 240,
 127                 'Euml'     => 203,
 128                 'euml'     => 235,
 129                 'euro'     => 8364,
 130                 'exist'    => 8707,
 131                 'fnof'     => 402,
 132                 'forall'   => 8704,
 133                 'frac12'   => 189,
 134                 'frac14'   => 188,
 135                 'frac34'   => 190,
 136                 'frasl'    => 8260,
 137                 'Gamma'    => 915,
 138                 'gamma'    => 947,
 139                 'ge'       => 8805,
 140                 'gt'       => 62,
 141                 'harr'     => 8596,
 142                 'hArr'     => 8660,
 143                 'hearts'   => 9829,
 144                 'hellip'   => 8230,
 145                 'Iacute'   => 205,
 146                 'iacute'   => 237,
 147                 'Icirc'    => 206,
 148                 'icirc'    => 238,
 149                 'iexcl'    => 161,
 150                 'Igrave'   => 204,
 151                 'igrave'   => 236,
 152                 'image'    => 8465,
 153                 'infin'    => 8734,
 154                 'int'      => 8747,
 155                 'Iota'     => 921,
 156                 'iota'     => 953,
 157                 'iquest'   => 191,
 158                 'isin'     => 8712,
 159                 'Iuml'     => 207,
 160                 'iuml'     => 239,
 161                 'Kappa'    => 922,
 162                 'kappa'    => 954,
 163                 'Lambda'   => 923,
 164                 'lambda'   => 955,
 165                 'lang'     => 9001,
 166                 'laquo'    => 171,
 167                 'larr'     => 8592,
 168                 'lArr'     => 8656,
 169                 'lceil'    => 8968,
 170                 'ldquo'    => 8220,
 171                 'le'       => 8804,
 172                 'lfloor'   => 8970,
 173                 'lowast'   => 8727,
 174                 'loz'      => 9674,
 175                 'lrm'      => 8206,
 176                 'lsaquo'   => 8249,
 177                 'lsquo'    => 8216,
 178                 'lt'       => 60,
 179                 'macr'     => 175,
 180                 'mdash'    => 8212,
 181                 'micro'    => 181,
 182                 'middot'   => 183,
 183                 'minus'    => 8722,
 184                 'Mu'       => 924,
 185                 'mu'       => 956,
 186                 'nabla'    => 8711,
 187                 'nbsp'     => 160,
 188                 'ndash'    => 8211,
 189                 'ne'       => 8800,
 190                 'ni'       => 8715,
 191                 'not'      => 172,
 192                 'notin'    => 8713,
 193                 'nsub'     => 8836,
 194                 'Ntilde'   => 209,
 195                 'ntilde'   => 241,
 196                 'Nu'       => 925,
 197                 'nu'       => 957,
 198                 'Oacute'   => 211,
 199                 'oacute'   => 243,
 200                 'Ocirc'    => 212,
 201                 'ocirc'    => 244,
 202                 'OElig'    => 338,
 203                 'oelig'    => 339,
 204                 'Ograve'   => 210,
 205                 'ograve'   => 242,
 206                 'oline'    => 8254,
 207                 'Omega'    => 937,
 208                 'omega'    => 969,
 209                 'Omicron'  => 927,
 210                 'omicron'  => 959,
 211                 'oplus'    => 8853,
 212                 'or'       => 8744,
 213                 'ordf'     => 170,
 214                 'ordm'     => 186,
 215                 'Oslash'   => 216,
 216                 'oslash'   => 248,
 217                 'Otilde'   => 213,
 218                 'otilde'   => 245,
 219                 'otimes'   => 8855,
 220                 'Ouml'     => 214,
 221                 'ouml'     => 246,
 222                 'para'     => 182,
 223                 'part'     => 8706,
 224                 'permil'   => 8240,
 225                 'perp'     => 8869,
 226                 'Phi'      => 934,
 227                 'phi'      => 966,
 228                 'Pi'       => 928,
 229                 'pi'       => 960,
 230                 'piv'      => 982,
 231                 'plusmn'   => 177,
 232                 'pound'    => 163,
 233                 'prime'    => 8242,
 234                 'Prime'    => 8243,
 235                 'prod'     => 8719,
 236                 'prop'     => 8733,
 237                 'Psi'      => 936,
 238                 'psi'      => 968,
 239                 'quot'     => 34,
 240                 'radic'    => 8730,
 241                 'rang'     => 9002,
 242                 'raquo'    => 187,
 243                 'rarr'     => 8594,
 244                 'rArr'     => 8658,
 245                 'rceil'    => 8969,
 246                 'rdquo'    => 8221,
 247                 'real'     => 8476,
 248                 'reg'      => 174,
 249                 'rfloor'   => 8971,
 250                 'Rho'      => 929,
 251                 'rho'      => 961,
 252                 'rlm'      => 8207,
 253                 'rsaquo'   => 8250,
 254                 'rsquo'    => 8217,
 255                 'sbquo'    => 8218,
 256                 'Scaron'   => 352,
 257                 'scaron'   => 353,
 258                 'sdot'     => 8901,
 259                 'sect'     => 167,
 260                 'shy'      => 173,
 261                 'Sigma'    => 931,
 262                 'sigma'    => 963,
 263                 'sigmaf'   => 962,
 264                 'sim'      => 8764,
 265                 'spades'   => 9824,
 266                 'sub'      => 8834,
 267                 'sube'     => 8838,
 268                 'sum'      => 8721,
 269                 'sup'      => 8835,
 270                 'sup1'     => 185,
 271                 'sup2'     => 178,
 272                 'sup3'     => 179,
 273                 'supe'     => 8839,
 274                 'szlig'    => 223,
 275                 'Tau'      => 932,
 276                 'tau'      => 964,
 277                 'there4'   => 8756,
 278                 'Theta'    => 920,
 279                 'theta'    => 952,
 280                 'thetasym' => 977,
 281                 'thinsp'   => 8201,
 282                 'THORN'    => 222,
 283                 'thorn'    => 254,
 284                 'tilde'    => 732,
 285                 'times'    => 215,
 286                 'trade'    => 8482,
 287                 'Uacute'   => 218,
 288                 'uacute'   => 250,
 289                 'uarr'     => 8593,
 290                 'uArr'     => 8657,
 291                 'Ucirc'    => 219,
 292                 'ucirc'    => 251,
 293                 'Ugrave'   => 217,
 294                 'ugrave'   => 249,
 295                 'uml'      => 168,
 296                 'upsih'    => 978,
 297                 'Upsilon'  => 933,
 298                 'upsilon'  => 965,
 299                 'Uuml'     => 220,
 300                 'uuml'     => 252,
 301                 'weierp'   => 8472,
 302                 'Xi'       => 926,
 303                 'xi'       => 958,
 304                 'Yacute'   => 221,
 305                 'yacute'   => 253,
 306                 'yen'      => 165,
 307                 'Yuml'     => 376,
 308                 'yuml'     => 255,
 309                 'Zeta'     => 918,
 310                 'zeta'     => 950,
 311                 'zwj'      => 8205,
 312                 'zwnj'     => 8204
 313         );
 314
 315         /**
 316          * Character entity aliases accepted by MediaWiki
 317          */
 318         static $htmlEntityAliases = array(
 319                 'רלמ' => 'rlm',
 320                 'رلم' => 'rlm',
 321         );
 322
 323         /**
 324          * Lazy-initialised attributes regex, see getAttribsRegex()
 325          */
 326         static $attribsRegex;
 327
 328         /**
 329          * Regular expression to match HTML/XML attribute pairs within a tag.
 330          * Allows some... latitude.
 331          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 332          */
 333         static function getAttribsRegex() {
 334                 if ( self::$attribsRegex === null ) {
 335                         $attribFirst = '[:A-Z_a-z0-9]';
 336                         $attrib = '[:A-Z_a-z-.0-9]';
 337                         $space = '[\x09\x0a\x0d\x20]';
 338                         self::$attribsRegex =
 339                                 "/(?:^|$space)({$attribFirst}{$attrib}*)
 340                                   ($space*=$space*
 341                                         (?:
 342                                          # The attribute value: quoted or alone
 343                                           \"([^<\"]*)\"
 344                                          | '([^<']*)'
 345                                          |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
 346                                          |  (\#[0-9a-fA-F]+) # Technically wrong, but lots of
 347                                                                                  # colors are specified like this.
 348                                                                                  # We'll be normalizing it.
 349                                         )
 350                                 )?(?=$space|\$)/sx";
 351                 }
 352                 return self::$attribsRegex;
 353         }
 354
 355         /**
 356          * Cleans up HTML, removes dangerous tags and attributes, and
 357          * removes HTML comments
 358          * @private
 359          * @param $text String
 360          * @param $processCallback Callback to do any variable or parameter replacements in HTML attribute values
 361          * @param $args Array for the processing callback
 362          * @param $extratags Array for any extra tags to include
 363          * @param $removetags Array for any tags (default or extra) to exclude
 364          * @return string
 365          */
 366         static function removeHTMLtags( $text, $processCallback = null, $args = array(), $extratags = array(), $removetags = array() ) {
 367                 global $wgUseTidy, $wgHtml5, $wgAllowMicrodataAttributes, $wgAllowImageTag;
 368
 369                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 370                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 371
 372                 wfProfileIn( __METHOD__ );
 373
 374                 // Base our staticInitialised variable off of the global config state so that if the globals
 375                 // are changed (like in the secrewed up test system) we will re-initialise the settings.
 376                 $globalContext = implode( '-', compact( 'wgHtml5', 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
 377                 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
 378
 379                         $htmlpairsStatic = array( # Tags that must be closed
 380                                 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 381                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 382                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 383                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 384                                 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'abbr', 'dfn',
 385                                 'kbd', 'samp'
 386                         );
 387                         if ( $wgHtml5 ) {
 388                                 $htmlpairsStatic = array_merge( $htmlpairsStatic, array( 'data', 'time' ) );
 389                         }
 390                         $htmlsingle = array(
 391                                 'br', 'hr', 'li', 'dt', 'dd'
 392                         );
 393                         $htmlsingleonly = array( # Elements that cannot have close tags
 394                                 'br', 'hr'
 395                         );
 396                         if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
 397                                 $htmlsingle[] = $htmlsingleonly[] = 'meta';
 398                                 $htmlsingle[] = $htmlsingleonly[] = 'link';
 399                         }
 400                         $htmlnest = array( # Tags that can be nested--??
 401                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 402                                 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span'
 403                         );
 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 $bits
 449                                 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $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 uncloseable 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' ) { $tagstack = array_pop( $tablestack ); }
 564                         }
 565                 } else {
 566                         # this might be possible using tidy itself
 567                         foreach ( $bits as $x ) {
 568                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 569                                 $x, $regs );
 570                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 571                                 $badtag = false;
 572                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 573                                         if( is_callable( $processCallback ) ) {
 574                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 575                                         }
 576
 577                                         if ( !Sanitizer::validateTag( $params, $t ) ) {
 578                                                 $badtag = true;
 579                                         }
 580
 581                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 582                                         if ( !$badtag ) {
 583                                                 $rest = str_replace( '>', '&gt;', $rest );
 584                                                 $text .= "<$slash$t$newparams$brace$rest";
 585                                                 continue;
 586                                         }
 587                                 }
 588                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 589                         }
 590                 }
 591                 wfProfileOut( __METHOD__ );
 592                 return $text;
 593         }
 594
 595         /**
 596          * Remove '<!--', '-->', and everything between.
 597          * To avoid leaving blank lines, when a comment is both preceded
 598          * and followed by a newline (ignoring spaces), trim leading and
 599          * trailing spaces and one of the newlines.
 600          *
 601          * @private
 602          * @param $text String
 603          * @return string
 604          */
 605         static function removeHTMLcomments( $text ) {
 606                 wfProfileIn( __METHOD__ );
 607                 while (($start = strpos($text, '<!--')) !== false) {
 608                         $end = strpos($text, '-->', $start + 4);
 609                         if ($end === false) {
 610                                 # Unterminated comment; bail out
 611                                 break;
 612                         }
 613
 614                         $end += 3;
 615
 616                         # Trim space and newline if the comment is both
 617                         # preceded and followed by a newline
 618                         $spaceStart = max($start - 1, 0);
 619                         $spaceLen = $end - $spaceStart;
 620                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 621                                 $spaceStart--;
 622                                 $spaceLen++;
 623                         }
 624                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 625                                 $spaceLen++;
 626                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 627                                 # Remove the comment, leading and trailing
 628                                 # spaces, and leave only one newline.
 629                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 630                         }
 631                         else {
 632                                 # Remove just the comment.
 633                                 $text = substr_replace($text, '', $start, $end - $start);
 634                         }
 635                 }
 636                 wfProfileOut( __METHOD__ );
 637                 return $text;
 638         }
 639
 640         /**
 641          * Take an array of attribute names and values and fix some deprecated values
 642          * for the given element type.
 643          * This does not validate properties, so you should ensure that you call
 644          * validateTagAttributes AFTER this to ensure that the resulting style rule
 645          * this may add is safe.
 646          *
 647          * - Converts most presentational attributes like align into inline css
 648          *
 649          * @param $attribs Array
 650          * @param $element String
 651          * @return Array
 652          */
 653         static function fixDeprecatedAttributes( $attribs, $element ) {
 654                 global $wgHtml5, $wgCleanupPresentationalAttributes;
 655
 656                 // presentational attributes were removed from html5, we can leave them
 657                 // in when html5 is turned off
 658                 if ( !$wgHtml5 || !$wgCleanupPresentationalAttributes ) {
 659                         return $attribs;
 660                 }
 661
 662                 $table = array( 'table' );
 663                 $cells = array( 'td', 'th' );
 664                 $colls = array( 'col', 'colgroup' );
 665                 $tblocks = array( 'tbody', 'tfoot', 'thead' );
 666                 $h = array( 'h1', 'h2', 'h3', 'h4', 'h5', 'h6' );
 667
 668                 $presentationalAttribs = array(
 669                         'align' => array( 'text-align', array_merge( array( 'caption', 'hr', 'div', 'p', 'tr' ), $table, $cells, $colls, $tblocks, $h ) ),
 670                         'clear' => array( 'clear', array( 'br' ) ),
 671                         'height' => array( 'height', $cells ),
 672                         'nowrap' => array( 'white-space', $cells ),
 673                         'size' => array( 'height', array( 'hr' ) ),
 674                         'type' => array( 'list-style-type', array( 'li', 'ol', 'ul' ) ),
 675                         'valign' => array( 'vertical-align', array_merge( $cells, $colls, $tblocks ) ),
 676                         'width' => array( 'width', array_merge( array( 'hr', 'pre' ), $table, $cells, $colls ) ),
 677                 );
 678
 679                 // Ensure that any upper case or mixed case attributes are converted to lowercase
 680                 foreach ( $attribs as $attribute => $value ) {
 681                         if ( $attribute !== strtolower( $attribute ) && array_key_exists( strtolower( $attribute ), $presentationalAttribs ) ) {
 682                                 $attribs[strtolower( $attribute )] = $value;
 683                                 unset( $attribs[$attribute] );
 684                         }
 685                 }
 686
 687                 $style = "";
 688                 foreach ( $presentationalAttribs as $attribute => $info ) {
 689                         list( $property, $elements ) = $info;
 690
 691                         // Skip if this attribute is not relevant to this element
 692                         if ( !in_array( $element, $elements ) ) {
 693                                 continue;
 694                         }
 695
 696                         // Skip if the attribute is not used
 697                         if ( !array_key_exists( $attribute, $attribs ) ) {
 698                                 continue;
 699                         }
 700
 701                         $value = $attribs[$attribute];
 702
 703                         // For nowrap the value should be nowrap instead of whatever text is in the value
 704                         if ( $attribute === 'nowrap' ) {
 705                                 $value = 'nowrap';
 706                         }
 707
 708                         // clear="all" is clear: both; in css
 709                         if ( $attribute === 'clear' && strtolower( $value ) === 'all' ) {
 710                                 $value = 'both';
 711                         }
 712
 713                         // Size based properties should have px applied to them if they have no unit
 714                         if ( in_array( $attribute, array( 'height', 'width', 'size' ) ) ) {
 715                                 if ( preg_match( '/^[\d.]+$/', $value ) ) {
 716                                         $value = "{$value}px";
 717                                 }
 718                         }
 719
 720                         // Table align is special, it's about block alignment instead of
 721                         // content align (see also bug 40306)
 722                         if ( $attribute === 'align' && in_array( $element, $table ) ) {
 723                                 if ( $value === 'center' ) {
 724                                         $style .= ' margin-left: auto;';
 725                                         $property = 'margin-right';
 726                                         $value = 'auto';
 727                                 } else {
 728                                         $property = 'float';
 729                                 }
 730                         }
 731
 732                         $style .= " $property: $value;";
 733
 734                         unset( $attribs[$attribute] );
 735                 }
 736
 737                 if ( $style ) {
 738                         // Prepend our style rules so that they can be overridden by user css
 739                         if ( isset($attribs['style']) ) {
 740                                 $style .= " " . $attribs['style'];
 741                         }
 742                         $attribs['style'] = trim($style);
 743                 }
 744
 745                 return $attribs;
 746         }
 747
 748         /**
 749          * Takes attribute names and values for a tag and the tah name and
 750          * validates that the tag is allowed to be present.
 751          * This DOES NOT validate the attributes, nor does it validate the
 752          * tags themselves. This method only handles the special circumstances
 753          * where we may want to allow a tag within content but ONLY when it has
 754          * specific attributes set.
 755          *
 756          * @param $
 757          */
 758         static function validateTag( $params, $element ) {
 759                 $params = Sanitizer::decodeTagAttributes( $params );
 760
 761                 if ( $element == 'meta' || $element == 'link' ) {
 762                         if ( !isset( $params['itemprop'] ) ) {
 763                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 764                                 return false;
 765                         }
 766                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 767                                 // <meta> must have a content="" for the itemprop
 768                                 return false;
 769                         }
 770                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 771                                 // <link> must have an associated href=""
 772                                 return false;
 773                         }
 774                 }
 775
 776                 return true;
 777         }
 778
 779         /**
 780          * Take an array of attribute names and values and normalize or discard
 781          * illegal values for the given element type.
 782          *
 783          * - Discards attributes not on a whitelist for the given element
 784          * - Unsafe style attributes are discarded
 785          * - Invalid id attributes are reencoded
 786          *
 787          * @param $attribs Array
 788          * @param $element String
 789          * @return Array
 790          *
 791          * @todo Check for legal values where the DTD limits things.
 792          * @todo Check for unique id attribute :P
 793          */
 794         static function validateTagAttributes( $attribs, $element ) {
 795                 return Sanitizer::validateAttributes( $attribs,
 796                         Sanitizer::attributeWhitelist( $element ) );
 797         }
 798
 799         /**
 800          * Take an array of attribute names and values and normalize or discard
 801          * illegal values for the given whitelist.
 802          *
 803          * - Discards attributes not the given whitelist
 804          * - Unsafe style attributes are discarded
 805          * - Invalid id attributes are reencoded
 806          *
 807          * @param $attribs Array
 808          * @param $whitelist Array: list of allowed attribute names
 809          * @return Array
 810          *
 811          * @todo Check for legal values where the DTD limits things.
 812          * @todo Check for unique id attribute :P
 813          */
 814         static function validateAttributes( $attribs, $whitelist ) {
 815                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
 816
 817                 $whitelist = array_flip( $whitelist );
 818                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 819
 820                 $out = array();
 821                 foreach( $attribs as $attribute => $value ) {
 822                         #allow XML namespace declaration if RDFa is enabled
 823                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 824                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 825                                         $out[$attribute] = $value;
 826                                 }
 827
 828                                 continue;
 829                         }
 830
 831                         # Allow any attribute beginning with "data-", if in HTML5 mode
 832                         if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
 833                                 continue;
 834                         }
 835
 836                         # Strip javascript "expression" from stylesheets.
 837                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 838                         if( $attribute == 'style' ) {
 839                                 $value = Sanitizer::checkCss( $value );
 840                         }
 841
 842                         if ( $attribute === 'id' ) {
 843                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 844                         }
 845
 846                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 847                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 848                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 849                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 850                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 851                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 852
 853                                 //Paranoia. Allow "simple" values but suppress javascript
 854                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 855                                         continue;
 856                                 }
 857                         }
 858
 859                         # NOTE: even though elements using href/src are not allowed directly, supply
 860                         #       validation code that can be used by tag hook handlers, etc
 861                         if ( $attribute === 'href' || $attribute === 'src' ) {
 862                                 if ( !preg_match( $hrefExp, $value ) ) {
 863                                         continue; //drop any href or src attributes not using an allowed protocol.
 864                                                   //NOTE: this also drops all relative URLs
 865                                 }
 866                         }
 867
 868                         // If this attribute was previously set, override it.
 869                         // Output should only have one attribute of each name.
 870                         $out[$attribute] = $value;
 871                 }
 872
 873                 if ( $wgAllowMicrodataAttributes ) {
 874                         # itemtype, itemid, itemref don't make sense without itemscope
 875                         if ( !array_key_exists( 'itemscope', $out ) ) {
 876                                 unset( $out['itemtype'] );
 877                                 unset( $out['itemid'] );
 878                                 unset( $out['itemref'] );
 879                         }
 880                         # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 881                 }
 882                 return $out;
 883         }
 884
 885         /**
 886          * Merge two sets of HTML attributes.  Conflicting items in the second set
 887          * will override those in the first, except for 'class' attributes which
 888          * will be combined (if they're both strings).
 889          *
 890          * @todo implement merging for other attributes such as style
 891          * @param $a Array
 892          * @param $b Array
 893          * @return array
 894          */
 895         static function mergeAttributes( $a, $b ) {
 896                 $out = array_merge( $a, $b );
 897                 if( isset( $a['class'] ) && isset( $b['class'] )
 898                 && is_string( $a['class'] ) && is_string( $b['class'] )
 899                 && $a['class'] !== $b['class'] ) {
 900                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 901                                 -1, PREG_SPLIT_NO_EMPTY );
 902                         $out['class'] = implode( ' ', array_unique( $classes ) );
 903                 }
 904                 return $out;
 905         }
 906
 907         /**
 908          * Pick apart some CSS and check it for forbidden or unsafe structures.
 909          * Returns a sanitized string. This sanitized string will have
 910          * character references and escape sequences decoded, and comments
 911          * stripped. If the input is just too evil, only a comment complaining
 912          * about evilness will be returned.
 913          *
 914          * Currently URL references, 'expression', 'tps' are forbidden.
 915          *
 916          * NOTE: Despite the fact that character references are decoded, the
 917          * returned string may contain character references given certain
 918          * clever input strings. These character references must
 919          * be escaped before the return value is embedded in HTML.
 920          *
 921          * @param $value String
 922          * @return String
 923          */
 924         static function checkCss( $value ) {
 925                 // Decode character references like &#123;
 926                 $value = Sanitizer::decodeCharReferences( $value );
 927
 928                 // Decode escape sequences and line continuation
 929                 // See the grammar in the CSS 2 spec, appendix D.
 930                 // This has to be done AFTER decoding character references.
 931                 // This means it isn't possible for this function to return
 932                 // unsanitized escape sequences. It is possible to manufacture
 933                 // input that contains character references that decode to
 934                 // escape sequences that decode to character references, but
 935                 // it's OK for the return value to contain character references
 936                 // because the caller is supposed to escape those anyway.
 937                 static $decodeRegex;
 938                 if ( !$decodeRegex ) {
 939                         $space = '[\\x20\\t\\r\\n\\f]';
 940                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 941                         $backslash = '\\\\';
 942                         $decodeRegex = "/ $backslash
 943                                 (?:
 944                                         ($nl) |  # 1. Line continuation
 945                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 946                                         (.) | # 3. backslash cancelling special meaning
 947                                         () | # 4. backslash at end of string
 948                                 )/xu";
 949                 }
 950                 $value = preg_replace_callback( $decodeRegex,
 951                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 952
 953                 // Remove any comments; IE gets token splitting wrong
 954                 // This must be done AFTER decoding character references and
 955                 // escape sequences, because those steps can introduce comments
 956                 // This step cannot introduce character references or escape
 957                 // sequences, because it replaces comments with spaces rather
 958                 // than removing them completely.
 959                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 960
 961                 // Remove anything after a comment-start token, to guard against
 962                 // incorrect client implementations.
 963                 $commentPos = strpos( $value, '/*' );
 964                 if ( $commentPos !== false ) {
 965                         $value = substr( $value, 0, $commentPos );
 966                 }
 967
 968                 // Reject problematic keywords and control characters
 969                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 970                         return '/* invalid control char */';
 971                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( | image\s*\( | image-set\s*\( !ix', $value ) ) {
 972                         return '/* insecure input */';
 973                 }
 974                 return $value;
 975         }
 976
 977         /**
 978          * @param $matches array
 979          * @return String
 980          */
 981         static function cssDecodeCallback( $matches ) {
 982                 if ( $matches[1] !== '' ) {
 983                         // Line continuation
 984                         return '';
 985                 } elseif ( $matches[2] !== '' ) {
 986                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 987                 } elseif ( $matches[3] !== '' ) {
 988                         $char = $matches[3];
 989                 } else {
 990                         $char = '\\';
 991                 }
 992                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 993                         // These characters need to be escaped in strings
 994                         // Clean up the escape sequence to avoid parsing errors by clients
 995                         return '\\' . dechex( ord( $char ) ) . ' ';
 996                 } else {
 997                         // Decode unnecessary escape
 998                         return $char;
 999                 }
1000         }
1001
1002         /**
1003          * Take a tag soup fragment listing an HTML element's attributes
1004          * and normalize it to well-formed XML, discarding unwanted attributes.
1005          * Output is safe for further wikitext processing, with escaping of
1006          * values that could trigger problems.
1007          *
1008          * - Normalizes attribute names to lowercase
1009          * - Discards attributes not on a whitelist for the given element
1010          * - Turns broken or invalid entities into plaintext
1011          * - Double-quotes all attribute values
1012          * - Attributes without values are given the name as attribute
1013          * - Double attributes are discarded
1014          * - Unsafe style attributes are discarded
1015          * - Prepends space if there are attributes.
1016          *
1017          * @param $text String
1018          * @param $element String
1019          * @return String
1020          */
1021         static function fixTagAttributes( $text, $element ) {
1022                 if( trim( $text ) == '' ) {
1023                         return '';
1024                 }
1025
1026                 $decoded = Sanitizer::decodeTagAttributes( $text );
1027                 $decoded = Sanitizer::fixDeprecatedAttributes( $decoded, $element );
1028                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1029
1030                 $attribs = array();
1031                 foreach( $stripped as $attribute => $value ) {
1032                         $encAttribute = htmlspecialchars( $attribute );
1033                         $encValue = Sanitizer::safeEncodeAttribute( $value );
1034
1035                         $attribs[] = "$encAttribute=\"$encValue\"";
1036                 }
1037                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1038         }
1039
1040         /**
1041          * Encode an attribute value for HTML output.
1042          * @param $text String
1043          * @return HTML-encoded text fragment
1044          */
1045         static function encodeAttribute( $text ) {
1046                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1047
1048                 // Whitespace is normalized during attribute decoding,
1049                 // so if we've been passed non-spaces we must encode them
1050                 // ahead of time or they won't be preserved.
1051                 $encValue = strtr( $encValue, array(
1052                         "\n" => '&#10;',
1053                         "\r" => '&#13;',
1054                         "\t" => '&#9;',
1055                 ) );
1056
1057                 return $encValue;
1058         }
1059
1060         /**
1061          * Encode an attribute value for HTML tags, with extra armoring
1062          * against further wiki processing.
1063          * @param $text String
1064          * @return HTML-encoded text fragment
1065          */
1066         static function safeEncodeAttribute( $text ) {
1067                 $encValue = Sanitizer::encodeAttribute( $text );
1068
1069                 # Templates and links may be expanded in later parsing,
1070                 # creating invalid or dangerous output. Suppress this.
1071                 $encValue = strtr( $encValue, array(
1072                         '<'    => '&lt;',   // This should never happen,
1073                         '>'    => '&gt;',   // we've received invalid input
1074                         '"'    => '&quot;', // which should have been escaped.
1075                         '{'    => '&#123;',
1076                         '['    => '&#91;',
1077                         "''"   => '&#39;&#39;',
1078                         'ISBN' => '&#73;SBN',
1079                         'RFC'  => '&#82;FC',
1080                         'PMID' => '&#80;MID',
1081                         '|'    => '&#124;',
1082                         '__'   => '&#95;_',
1083                 ) );
1084
1085                 # Stupid hack
1086                 $encValue = preg_replace_callback(
1087                         '/((?i)' . wfUrlProtocols() . ')/',
1088                         array( 'Sanitizer', 'armorLinksCallback' ),
1089                         $encValue );
1090                 return $encValue;
1091         }
1092
1093         /**
1094          * Given a value, escape it so that it can be used in an id attribute and
1095          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1096          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1097          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1098          * escaped with lots of dots.
1099          *
1100          * To ensure we don't have to bother escaping anything, we also strip ', ",
1101          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1102          * We also strip # because it upsets IE, and % because it could be
1103          * ambiguous if it's part of something that looks like a percent escape
1104          * (which don't work reliably in fragments cross-browser).
1105          *
1106          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1107          *                                                          in the id and
1108          *                                                          name attributes
1109          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1110          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1111          *   HTML5 definition of id attribute
1112          *
1113          * @param $id String: id to escape
1114          * @param $options Mixed: string or array of strings (default is array()):
1115          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1116          *       so don't pay attention if the first character isn't valid at the
1117          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1118          *       false.
1119          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1120          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1121          *       anchors and links won't break.
1122          * @return String
1123          */
1124         static function escapeId( $id, $options = array() ) {
1125                 global $wgHtml5, $wgExperimentalHtmlIds;
1126                 $options = (array)$options;
1127
1128                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1129                         $id = Sanitizer::decodeCharReferences( $id );
1130                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1131                         $id = trim( $id, '_' );
1132                         if ( $id === '' ) {
1133                                 # Must have been all whitespace to start with.
1134                                 return '_';
1135                         } else {
1136                                 return $id;
1137                         }
1138                 }
1139
1140                 # HTML4-style escaping
1141                 static $replace = array(
1142                         '%3A' => ':',
1143                         '%' => '.'
1144                 );
1145
1146                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1147                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1148
1149                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1150                 && !in_array( 'noninitial', $options ) )  {
1151                         // Initial character must be a letter!
1152                         $id = "x$id";
1153                 }
1154                 return $id;
1155         }
1156
1157         /**
1158          * Given a value, escape it so that it can be used as a CSS class and
1159          * return it.
1160          *
1161          * @todo For extra validity, input should be validated UTF-8.
1162          *
1163          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1164          *
1165          * @param $class String
1166          * @return String
1167          */
1168         static function escapeClass( $class ) {
1169                 // Convert ugly stuff to underscores and kill underscores in ugly places
1170                 return rtrim(preg_replace(
1171                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
1172                         '_',
1173                         $class ), '_');
1174         }
1175
1176         /**
1177          * Given HTML input, escape with htmlspecialchars but un-escape entites.
1178          * This allows (generally harmless) entities like &#160; to survive.
1179          *
1180          * @param $html String to escape
1181          * @return String: escaped input
1182          */
1183         static function escapeHtmlAllowEntities( $html ) {
1184                 $html = Sanitizer::decodeCharReferences( $html );
1185                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1186                 # hurt.
1187                 $html = htmlspecialchars( $html, ENT_QUOTES );
1188                 return $html;
1189         }
1190
1191         /**
1192          * Regex replace callback for armoring links against further processing.
1193          * @param $matches Array
1194          * @return string
1195          */
1196         private static function armorLinksCallback( $matches ) {
1197                 return str_replace( ':', '&#58;', $matches[1] );
1198         }
1199
1200         /**
1201          * Return an associative array of attribute names and values from
1202          * a partial tag string. Attribute names are forces to lowercase,
1203          * character references are decoded to UTF-8 text.
1204          *
1205          * @param $text String
1206          * @return Array
1207          */
1208         public static function decodeTagAttributes( $text ) {
1209                 if( trim( $text ) == '' ) {
1210                         return array();
1211                 }
1212
1213                 $attribs = array();
1214                 $pairs = array();
1215                 if( !preg_match_all(
1216                         self::getAttribsRegex(),
1217                         $text,
1218                         $pairs,
1219                         PREG_SET_ORDER ) ) {
1220                         return $attribs;
1221                 }
1222
1223                 foreach( $pairs as $set ) {
1224                         $attribute = strtolower( $set[1] );
1225                         $value = Sanitizer::getTagAttributeCallback( $set );
1226
1227                         // Normalize whitespace
1228                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1229                         $value = trim( $value );
1230
1231                         // Decode character references
1232                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1233                 }
1234                 return $attribs;
1235         }
1236
1237         /**
1238          * Pick the appropriate attribute value from a match set from the
1239          * attribs regex matches.
1240          *
1241          * @param $set Array
1242          * @throws MWException
1243          * @return String
1244          */
1245         private static function getTagAttributeCallback( $set ) {
1246                 if( isset( $set[6] ) ) {
1247                         # Illegal #XXXXXX color with no quotes.
1248                         return $set[6];
1249                 } elseif( isset( $set[5] ) ) {
1250                         # No quotes.
1251                         return $set[5];
1252                 } elseif( isset( $set[4] ) ) {
1253                         # Single-quoted
1254                         return $set[4];
1255                 } elseif( isset( $set[3] ) ) {
1256                         # Double-quoted
1257                         return $set[3];
1258                 } elseif( !isset( $set[2] ) ) {
1259                         # In XHTML, attributes must have a value.
1260                         # For 'reduced' form, return explicitly the attribute name here.
1261                         return $set[1];
1262                 } else {
1263                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1264                 }
1265         }
1266
1267         /**
1268          * Normalize whitespace and character references in an XML source-
1269          * encoded text for an attribute value.
1270          *
1271          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1272          * but note that we're not returning the value, but are returning
1273          * XML source fragments that will be slapped into output.
1274          *
1275          * @param $text String
1276          * @return String
1277          */
1278         private static function normalizeAttributeValue( $text ) {
1279                 return str_replace( '"', '&quot;',
1280                         self::normalizeWhitespace(
1281                                 Sanitizer::normalizeCharReferences( $text ) ) );
1282         }
1283
1284         /**
1285          * @param $text string
1286          * @return mixed
1287          */
1288         private static function normalizeWhitespace( $text ) {
1289                 return preg_replace(
1290                         '/\r\n|[\x20\x0d\x0a\x09]/',
1291                         ' ',
1292                         $text );
1293         }
1294
1295         /**
1296          * Normalizes whitespace in a section name, such as might be returned
1297          * by Parser::stripSectionName(), for use in the id's that are used for
1298          * section links.
1299          *
1300          * @param $section String
1301          * @return String
1302          */
1303         static function normalizeSectionNameWhitespace( $section ) {
1304                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1305         }
1306
1307         /**
1308          * Ensure that any entities and character references are legal
1309          * for XML and XHTML specifically. Any stray bits will be
1310          * &amp;-escaped to result in a valid text fragment.
1311          *
1312          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1313          *   numericized (this way we're well-formed even without a DTD)
1314          * b. any numeric char refs must be legal chars, not invalid or forbidden
1315          * c. use lower cased "&#x", not "&#X"
1316          * d. fix or reject non-valid attributes
1317          *
1318          * @param $text String
1319          * @return String
1320          * @private
1321          */
1322         static function normalizeCharReferences( $text ) {
1323                 return preg_replace_callback(
1324                         self::CHAR_REFS_REGEX,
1325                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1326                         $text );
1327         }
1328         /**
1329          * @param $matches String
1330          * @return String
1331          */
1332         static function normalizeCharReferencesCallback( $matches ) {
1333                 $ret = null;
1334                 if( $matches[1] != '' ) {
1335                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1336                 } elseif( $matches[2] != '' ) {
1337                         $ret = Sanitizer::decCharReference( $matches[2] );
1338                 } elseif( $matches[3] != ''  ) {
1339                         $ret = Sanitizer::hexCharReference( $matches[3] );
1340                 }
1341                 if( is_null( $ret ) ) {
1342                         return htmlspecialchars( $matches[0] );
1343                 } else {
1344                         return $ret;
1345                 }
1346         }
1347
1348         /**
1349          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1350          * return the equivalent numeric entity reference (except for the core &lt;
1351          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1352          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1353          * pseudo-entity source (eg &amp;foo;)
1354          *
1355          * @param $name String
1356          * @return String
1357          */
1358         static function normalizeEntity( $name ) {
1359                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1360                         return '&' . self::$htmlEntityAliases[$name] . ';';
1361                 } elseif ( in_array( $name,
1362                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1363                         return "&$name;";
1364                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1365                         return '&#' . self::$htmlEntities[$name] . ';';
1366                 } else {
1367                         return "&amp;$name;";
1368                 }
1369         }
1370
1371         /**
1372          * @param $codepoint
1373          * @return null|string
1374          */
1375         static function decCharReference( $codepoint ) {
1376                 $point = intval( $codepoint );
1377                 if( Sanitizer::validateCodepoint( $point ) ) {
1378                         return sprintf( '&#%d;', $point );
1379                 } else {
1380                         return null;
1381                 }
1382         }
1383
1384         /**
1385          * @param $codepoint
1386          * @return null|string
1387          */
1388         static function hexCharReference( $codepoint ) {
1389                 $point = hexdec( $codepoint );
1390                 if( Sanitizer::validateCodepoint( $point ) ) {
1391                         return sprintf( '&#x%x;', $point );
1392                 } else {
1393                         return null;
1394                 }
1395         }
1396
1397         /**
1398          * Returns true if a given Unicode codepoint is a valid character in XML.
1399          * @param $codepoint Integer
1400          * @return Boolean
1401          */
1402         private static function validateCodepoint( $codepoint ) {
1403                 return ($codepoint ==    0x09)
1404                         || ($codepoint ==    0x0a)
1405                         || ($codepoint ==    0x0d)
1406                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1407                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1408                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1409         }
1410
1411         /**
1412          * Decode any character references, numeric or named entities,
1413          * in the text and return a UTF-8 string.
1414          *
1415          * @param $text String
1416          * @return String
1417          */
1418         public static function decodeCharReferences( $text ) {
1419                 return preg_replace_callback(
1420                         self::CHAR_REFS_REGEX,
1421                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1422                         $text );
1423         }
1424
1425         /**
1426          * Decode any character references, numeric or named entities,
1427          * in the next and normalize the resulting string. (bug 14952)
1428          *
1429          * This is useful for page titles, not for text to be displayed,
1430          * MediaWiki allows HTML entities to escape normalization as a feature.
1431          *
1432          * @param $text String (already normalized, containing entities)
1433          * @return String (still normalized, without entities)
1434          */
1435         public static function decodeCharReferencesAndNormalize( $text ) {
1436                 global $wgContLang;
1437                 $text = preg_replace_callback(
1438                         self::CHAR_REFS_REGEX,
1439                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1440                         $text, /* limit */ -1, $count );
1441
1442                 if ( $count ) {
1443                         return $wgContLang->normalize( $text );
1444                 } else {
1445                         return $text;
1446                 }
1447         }
1448
1449         /**
1450          * @param $matches String
1451          * @return String
1452          */
1453         static function decodeCharReferencesCallback( $matches ) {
1454                 if( $matches[1] != '' ) {
1455                         return Sanitizer::decodeEntity( $matches[1] );
1456                 } elseif( $matches[2] != '' ) {
1457                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1458                 } elseif( $matches[3] != ''  ) {
1459                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1460                 }
1461                 # Last case should be an ampersand by itself
1462                 return $matches[0];
1463         }
1464
1465         /**
1466          * Return UTF-8 string for a codepoint if that is a valid
1467          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1468          * @param $codepoint Integer
1469          * @return String
1470          * @private
1471          */
1472         static function decodeChar( $codepoint ) {
1473                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1474                         return codepointToUtf8( $codepoint );
1475                 } else {
1476                         return UTF8_REPLACEMENT;
1477                 }
1478         }
1479
1480         /**
1481          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1482          * return the UTF-8 encoding of that character. Otherwise, returns
1483          * pseudo-entity source (eg "&foo;")
1484          *
1485          * @param $name String
1486          * @return String
1487          */
1488         static function decodeEntity( $name ) {
1489                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1490                         $name = self::$htmlEntityAliases[$name];
1491                 }
1492                 if( isset( self::$htmlEntities[$name] ) ) {
1493                         return codepointToUtf8( self::$htmlEntities[$name] );
1494                 } else {
1495                         return "&$name;";
1496                 }
1497         }
1498
1499         /**
1500          * Fetch the whitelist of acceptable attributes for a given element name.
1501          *
1502          * @param $element String
1503          * @return Array
1504          */
1505         static function attributeWhitelist( $element ) {
1506                 $list = Sanitizer::setupAttributeWhitelist();
1507                 return isset( $list[$element] )
1508                         ? $list[$element]
1509                         : array();
1510         }
1511
1512         /**
1513          * Foreach array key (an allowed HTML element), return an array
1514          * of allowed attributes
1515          * @return Array
1516          */
1517         static function setupAttributeWhitelist() {
1518                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1519
1520                 static $whitelist, $staticInitialised;
1521                 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgHtml5', 'wgAllowMicrodataAttributes' ) );
1522
1523                 if ( isset( $whitelist ) && $staticInitialised == $globalContext ) {
1524                         return $whitelist;
1525                 }
1526
1527                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1528
1529                 if ( $wgAllowRdfaAttributes ) {
1530                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1531                         $common = array_merge( $common, array(
1532                             'about', 'property', 'resource', 'datatype', 'typeof',
1533                         ) );
1534                 }
1535
1536                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1537                         # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1538                         $common = array_merge( $common, array(
1539                             'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1540                         ) );
1541                 }
1542
1543                 $block = array_merge( $common, array( 'align' ) );
1544                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1545                 $tablecell = array( 'abbr',
1546                                     'axis',
1547                                     'headers',
1548                                     'scope',
1549                                     'rowspan',
1550                                     'colspan',
1551                                     'nowrap', # deprecated
1552                                     'width',  # deprecated
1553                                     'height', # deprecated
1554                                     'bgcolor' # deprecated
1555                                     );
1556
1557                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1558                 # See: http://www.w3.org/TR/html4/
1559                 $whitelist = array(
1560                         # 7.5.4
1561                         'div'        => $block,
1562                         'center'     => $common, # deprecated
1563                         'span'       => $block, # ??
1564
1565                         # 7.5.5
1566                         'h1'         => $block,
1567                         'h2'         => $block,
1568                         'h3'         => $block,
1569                         'h4'         => $block,
1570                         'h5'         => $block,
1571                         'h6'         => $block,
1572
1573                         # 7.5.6
1574                         # address
1575
1576                         # 8.2.4
1577                         # bdo
1578
1579                         # 9.2.1
1580                         'em'         => $common,
1581                         'strong'     => $common,
1582                         'cite'       => $common,
1583                         'dfn'        => $common,
1584                         'code'       => $common,
1585                         'samp'       => $common,
1586                         'kbd'        => $common,
1587                         'var'        => $common,
1588                         'abbr'       => $common,
1589                         # acronym
1590
1591                         # 9.2.2
1592                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1593                         # q
1594
1595                         # 9.2.3
1596                         'sub'        => $common,
1597                         'sup'        => $common,
1598
1599                         # 9.3.1
1600                         'p'          => $block,
1601
1602                         # 9.3.2
1603                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1604
1605                         # 9.3.4
1606                         'pre'        => array_merge( $common, array( 'width' ) ),
1607
1608                         # 9.4
1609                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1610                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1611
1612                         # 10.2
1613                         'ul'         => array_merge( $common, array( 'type' ) ),
1614                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1615                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1616
1617                         # 10.3
1618                         'dl'         => $common,
1619                         'dd'         => $common,
1620                         'dt'         => $common,
1621
1622                         # 11.2.1
1623                         'table'      => array_merge( $common,
1624                                                                 array( 'summary', 'width', 'border', 'frame',
1625                                                                                 'rules', 'cellspacing', 'cellpadding',
1626                                                                                 'align', 'bgcolor',
1627                                                                 ) ),
1628
1629                         # 11.2.2
1630                         'caption'    => array_merge( $common, array( 'align' ) ),
1631
1632                         # 11.2.3
1633                         'thead'      => array_merge( $common, $tablealign ),
1634                         'tfoot'      => array_merge( $common, $tablealign ),
1635                         'tbody'      => array_merge( $common, $tablealign ),
1636
1637                         # 11.2.4
1638                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1639                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1640
1641                         # 11.2.5
1642                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1643
1644                         # 11.2.6
1645                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1646                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1647
1648                         # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1649                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1650
1651                         # 13.2
1652                         # Not usually allowed, but may be used for extension-style hooks
1653                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1654                         # true
1655                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1656
1657                         # 15.2.1
1658                         'tt'         => $common,
1659                         'b'          => $common,
1660                         'i'          => $common,
1661                         'big'        => $common,
1662                         'small'      => $common,
1663                         'strike'     => $common,
1664                         's'          => $common,
1665                         'u'          => $common,
1666
1667                         # 15.2.2
1668                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1669                         # basefont
1670
1671                         # 15.3
1672                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1673
1674                         # XHTML Ruby annotation text module, simple ruby only.
1675                         # http://www.w3c.org/TR/ruby/
1676                         'ruby'       => $common,
1677                         # rbc
1678                         # rtc
1679                         'rb'         => $common,
1680                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1681                         'rp'         => $common,
1682
1683                         # MathML root element, where used for extensions
1684                         # 'title' may not be 100% valid here; it's XHTML
1685                         # http://www.w3.org/TR/REC-MathML/
1686                         'math'       => array( 'class', 'style', 'id', 'title' ),
1687
1688                         # HTML 5 section 4.6
1689                         'bdi' => $common,
1690
1691                 );
1692
1693                 if ( $wgHtml5 ) {
1694                         # HTML5 elements, defined by:
1695                         # http://www.whatwg.org/specs/web-apps/current-work/multipage/
1696                         $whitelist += array(
1697                                 'data' => array_merge( $common, array( 'value' ) ),
1698                                 'time' => array_merge( $common, array( 'datetime' ) ),
1699
1700                                 // meta and link are only permitted by removeHTMLtags when Microdata
1701                                 // is enabled so we don't bother adding a conditional to hide these
1702                                 // Also meta and link are only valid in WikiText as Microdata elements
1703                                 // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1704                                 // So we don't bother including $common attributes that have no purpose.
1705                                 'meta' => array( 'itemprop', 'content' ),
1706                                 'link' => array( 'itemprop', 'href' ),
1707                         );
1708                 }
1709
1710                 $staticInitialised = $globalContext;
1711
1712                 return $whitelist;
1713         }
1714
1715         /**
1716          * Take a fragment of (potentially invalid) HTML and return
1717          * a version with any tags removed, encoded as plain text.
1718          *
1719          * Warning: this return value must be further escaped for literal
1720          * inclusion in HTML output as of 1.10!
1721          *
1722          * @param $text String: HTML fragment
1723          * @return String
1724          */
1725         static function stripAllTags( $text ) {
1726                 # Actual <tags>
1727                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1728
1729                 # Normalize &entities and whitespace
1730                 $text = self::decodeCharReferences( $text );
1731                 $text = self::normalizeWhitespace( $text );
1732
1733                 return $text;
1734         }
1735
1736         /**
1737          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1738          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1739          * PHP 5.1 doesn't.
1740          *
1741          * Use for passing XHTML fragments to PHP's XML parsing functions
1742          *
1743          * @return String
1744          */
1745         static function hackDocType() {
1746                 $out = "<!DOCTYPE html [\n";
1747                 foreach( self::$htmlEntities as $entity => $codepoint ) {
1748                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1749                 }
1750                 $out .= "]>\n";
1751                 return $out;
1752         }
1753
1754         /**
1755          * @param $url string
1756          * @return mixed|string
1757          */
1758         static function cleanUrl( $url ) {
1759                 # Normalize any HTML entities in input. They will be
1760                 # re-escaped by makeExternalLink().
1761                 $url = Sanitizer::decodeCharReferences( $url );
1762
1763                 # Escape any control characters introduced by the above step
1764                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1765                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1766
1767                 # Validate hostname portion
1768                 $matches = array();
1769                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1770                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1771
1772                         // Characters that will be ignored in IDNs.
1773                         // http://tools.ietf.org/html/3454#section-3.1
1774                         // Strip them before further processing so blacklists and such work.
1775                         $strip = "/
1776                                 \\s|          # general whitespace
1777                                 \xc2\xad|     # 00ad SOFT HYPHEN
1778                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1779                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1780                                 \xe2\x81\xa0| # 2060 WORD JOINER
1781                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1782                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1783                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1784                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1785                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1786                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1787                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1788                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1789                                 /xuD";
1790
1791                         $host = preg_replace( $strip, '', $host );
1792
1793                         // @todo FIXME: Validate hostnames here
1794
1795                         return $protocol . $host . $rest;
1796                 } else {
1797                         return $url;
1798                 }
1799         }
1800
1801         /**
1802          * @param $matches array
1803          * @return string
1804          */
1805         static function cleanUrlCallback( $matches ) {
1806                 return urlencode( $matches[0] );
1807         }
1808
1809         /**
1810          * Does a string look like an e-mail address?
1811          *
1812          * This validates an email address using an HTML5 specification found at:
1813          * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1814          * Which as of 2011-01-24 says:
1815          *
1816          *   A valid e-mail address is a string that matches the ABNF production
1817          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1818          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1819          *   3.5.
1820          *
1821          * This function is an implementation of the specification as requested in
1822          * bug 22449.
1823          *
1824          * Client-side forms will use the same standard validation rules via JS or
1825          * HTML 5 validation; additional restrictions can be enforced server-side
1826          * by extensions via the 'isValidEmailAddr' hook.
1827          *
1828          * Note that this validation doesn't 100% match RFC 2822, but is believed
1829          * to be liberal enough for wide use. Some invalid addresses will still
1830          * pass validation here.
1831          *
1832          * @since 1.18
1833          *
1834          * @param $addr String E-mail address
1835          * @return Bool
1836          */
1837         public static function validateEmail( $addr ) {
1838                 $result = null;
1839                 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1840                         return $result;
1841                 }
1842
1843                 // Please note strings below are enclosed in brackets [], this make the
1844                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1845                 // See bug 26948
1846                 $rfc5322_atext   = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1847                 $rfc1034_ldh_str = "a-z0-9\\-" ;
1848
1849                 $HTML5_email_regexp = "/
1850                 ^                      # start of string
1851                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1852                 @                      # 'apostrophe'
1853                 [$rfc1034_ldh_str]+       # First domain part
1854                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1855                 $                      # End of string
1856                 /ix" ; // case Insensitive, eXtended
1857
1858                 return (bool) preg_match( $HTML5_email_regexp, $addr );
1859         }
1860 }