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;
 368
 369                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 370                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 371
 372                 wfProfileIn( __METHOD__ );
 373
 374                 if ( !$staticInitialised ) {
 375
 376                         $htmlpairsStatic = array( # Tags that must be closed
 377                                 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 378                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 379                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 380                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 381                                 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'abbr', 'dfn',
 382                                 'kbd', 'samp'
 383                         );
 384                         $htmlsingle = array(
 385                                 'br', 'hr', 'li', 'dt', 'dd'
 386                         );
 387                         $htmlsingleonly = array( # Elements that cannot have close tags
 388                                 'br', 'hr'
 389                         );
 390                         $htmlnest = array( # Tags that can be nested--??
 391                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 392                                 'dl', 'font', 'big', 'small', 'sub', 'sup', 'span'
 393                         );
 394                         $tabletags = array( # Can only appear inside table, we will close them
 395                                 'td', 'th', 'tr',
 396                         );
 397                         $htmllist = array( # Tags used by list
 398                                 'ul','ol',
 399                         );
 400                         $listtags = array( # Tags that can appear in a list
 401                                 'li',
 402                         );
 403
 404                         global $wgAllowImageTag;
 405                         if ( $wgAllowImageTag ) {
 406                                 $htmlsingle[] = 'img';
 407                                 $htmlsingleonly[] = 'img';
 408                         }
 409
 410                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 411                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 412
 413                         # Convert them all to hashtables for faster lookup
 414                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 415                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 416                         foreach ( $vars as $var ) {
 417                                 $$var = array_flip( $$var );
 418                         }
 419                         $staticInitialised = true;
 420                 }
 421                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 422                 $extratags = array_flip( $extratags );
 423                 $removetags = array_flip( $removetags );
 424                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 425                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ) , $removetags );
 426
 427                 # Remove HTML comments
 428                 $text = Sanitizer::removeHTMLcomments( $text );
 429                 $bits = explode( '<', $text );
 430                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 431                 if ( !$wgUseTidy ) {
 432                         $tagstack = $tablestack = array();
 433                         foreach ( $bits as $x ) {
 434                                 $regs = array();
 435                                 # $slash: Does the current element start with a '/'?
 436                                 # $t: Current element name
 437                                 # $params: String between element name and >
 438                                 # $brace: Ending '>' or '/>'
 439                                 # $rest: Everything until the next element of $bits
 440                                 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 441                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 442                                 } else {
 443                                         $slash = $t = $params = $brace = $rest = null;
 444                                 }
 445
 446                                 $badtag = false;
 447                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 448                                         # Check our stack
 449                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 450                                                 $badtag = true;
 451                                         } elseif ( $slash ) {
 452                                                 # Closing a tag... is it the one we just opened?
 453                                                 $ot = @array_pop( $tagstack );
 454                                                 if ( $ot != $t ) {
 455                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 456                                                                 # Pop all elements with an optional close tag
 457                                                                 # and see if we find a match below them
 458                                                                 $optstack = array();
 459                                                                 array_push( $optstack, $ot );
 460                                                                 wfSuppressWarnings();
 461                                                                 $ot = array_pop( $tagstack );
 462                                                                 wfRestoreWarnings();
 463                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 464                                                                         array_push( $optstack, $ot );
 465                                                                         wfSuppressWarnings();
 466                                                                         $ot = array_pop( $tagstack );
 467                                                                         wfRestoreWarnings();
 468                                                                 }
 469                                                                 if ( $t != $ot ) {
 470                                                                         # No match. Push the optional elements back again
 471                                                                         $badtag = true;
 472                                                                         wfSuppressWarnings();
 473                                                                         $ot = array_pop( $optstack );
 474                                                                         wfRestoreWarnings();
 475                                                                         while ( $ot ) {
 476                                                                                 array_push( $tagstack, $ot );
 477                                                                                 wfSuppressWarnings();
 478                                                                                 $ot = array_pop( $optstack );
 479                                                                                 wfRestoreWarnings();
 480                                                                         }
 481                                                                 }
 482                                                         } else {
 483                                                                 @array_push( $tagstack, $ot );
 484                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 485                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 486                                                                         $badtag = true;
 487                                                                 }
 488                                                         }
 489                                                 } else {
 490                                                         if ( $t == 'table' ) {
 491                                                                 $tagstack = array_pop( $tablestack );
 492                                                         }
 493                                                 }
 494                                                 $newparams = '';
 495                                         } else {
 496                                                 # Keep track for later
 497                                                 if ( isset( $tabletags[$t] ) &&
 498                                                 !in_array( 'table', $tagstack ) ) {
 499                                                         $badtag = true;
 500                                                 } elseif ( in_array( $t, $tagstack ) &&
 501                                                 !isset( $htmlnest [$t ] ) ) {
 502                                                         $badtag = true;
 503                                                 # Is it a self closed htmlpair ? (bug 5487)
 504                                                 } elseif ( $brace == '/>' &&
 505                                                 isset( $htmlpairs[$t] ) ) {
 506                                                         $badtag = true;
 507                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 508                                                         # Hack to force empty tag for uncloseable elements
 509                                                         $brace = '/>';
 510                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 511                                                         # Hack to not close $htmlsingle tags
 512                                                         $brace = null;
 513                                                 } elseif ( isset( $tabletags[$t] )
 514                                                 && in_array( $t, $tagstack ) ) {
 515                                                         // New table tag but forgot to close the previous one
 516                                                         $text .= "</$t>";
 517                                                 } else {
 518                                                         if ( $t == 'table' ) {
 519                                                                 array_push( $tablestack, $tagstack );
 520                                                                 $tagstack = array();
 521                                                         }
 522                                                         array_push( $tagstack, $t );
 523                                                 }
 524
 525                                                 # Replace any variables or template parameters with
 526                                                 # plaintext results.
 527                                                 if( is_callable( $processCallback ) ) {
 528                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 529                                                 }
 530
 531                                                 # Strip non-approved attributes from the tag
 532                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 533                                         }
 534                                         if ( !$badtag ) {
 535                                                 $rest = str_replace( '>', '&gt;', $rest );
 536                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 537                                                 $text .= "<$slash$t$newparams$close>$rest";
 538                                                 continue;
 539                                         }
 540                                 }
 541                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 542                         }
 543                         # Close off any remaining tags
 544                         while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
 545                                 $text .= "</$t>\n";
 546                                 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
 547                         }
 548                 } else {
 549                         # this might be possible using tidy itself
 550                         foreach ( $bits as $x ) {
 551                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 552                                 $x, $regs );
 553                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 554                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 555                                         if( is_callable( $processCallback ) ) {
 556                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 557                                         }
 558                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 559                                         $rest = str_replace( '>', '&gt;', $rest );
 560                                         $text .= "<$slash$t$newparams$brace$rest";
 561                                 } else {
 562                                         $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 563                                 }
 564                         }
 565                 }
 566                 wfProfileOut( __METHOD__ );
 567                 return $text;
 568         }
 569
 570         /**
 571          * Remove '<!--', '-->', and everything between.
 572          * To avoid leaving blank lines, when a comment is both preceded
 573          * and followed by a newline (ignoring spaces), trim leading and
 574          * trailing spaces and one of the newlines.
 575          *
 576          * @private
 577          * @param $text String
 578          * @return string
 579          */
 580         static function removeHTMLcomments( $text ) {
 581                 wfProfileIn( __METHOD__ );
 582                 while (($start = strpos($text, '<!--')) !== false) {
 583                         $end = strpos($text, '-->', $start + 4);
 584                         if ($end === false) {
 585                                 # Unterminated comment; bail out
 586                                 break;
 587                         }
 588
 589                         $end += 3;
 590
 591                         # Trim space and newline if the comment is both
 592                         # preceded and followed by a newline
 593                         $spaceStart = max($start - 1, 0);
 594                         $spaceLen = $end - $spaceStart;
 595                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 596                                 $spaceStart--;
 597                                 $spaceLen++;
 598                         }
 599                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 600                                 $spaceLen++;
 601                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 602                                 # Remove the comment, leading and trailing
 603                                 # spaces, and leave only one newline.
 604                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 605                         }
 606                         else {
 607                                 # Remove just the comment.
 608                                 $text = substr_replace($text, '', $start, $end - $start);
 609                         }
 610                 }
 611                 wfProfileOut( __METHOD__ );
 612                 return $text;
 613         }
 614
 615         /**
 616          * Take an array of attribute names and values and fix some deprecated values
 617          * for the given element type.
 618          * This does not validate properties, so you should ensure that you call
 619          * validateTagAttributes AFTER this to ensure that the resulting style rule
 620          * this may add is safe.
 621          *
 622          * - Converts most presentational attributes like align into inline css
 623          *
 624          * @param $attribs Array
 625          * @param $element String
 626          * @return Array
 627          */
 628         static function fixDeprecatedAttributes( $attribs, $element ) {
 629                 global $wgHtml5, $wgCleanupPresentationalAttributes;
 630
 631                 // presentational attributes were removed from html5, we can leave them
 632                 // in when html5 is turned off
 633                 if ( !$wgHtml5 || !$wgCleanupPresentationalAttributes ) {
 634                         return $attribs;
 635                 }
 636
 637                 $table = array( 'table' );
 638                 $cells = array( 'td', 'th' );
 639                 $colls = array( 'col', 'colgroup' );
 640                 $tblocks = array( 'tbody', 'tfoot', 'thead' );
 641                 $h = array( 'h1', 'h2', 'h3', 'h4', 'h5', 'h6' );
 642
 643                 $presentationalAttribs = array(
 644                         'align' => array( 'text-align', array_merge( array( 'caption', 'hr', 'div', 'p', 'tr' ), $table, $cells, $colls, $tblocks, $h ) ),
 645                         'clear' => array( 'clear', array( 'br' ) ),
 646                         'height' => array( 'height', $cells ),
 647                         'nowrap' => array( 'white-space', $cells ),
 648                         'size' => array( 'height', array( 'hr' ) ),
 649                         'type' => array( 'list-style-type', array( 'li', 'ol', 'ul' ) ),
 650                         'valign' => array( 'vertical-align', array_merge( $cells, $colls, $tblocks ) ),
 651                         'width' => array( 'width', array_merge( array( 'hr', 'pre' ), $table, $cells, $colls ) ),
 652                 );
 653
 654                 // Ensure that any upper case or mixed case attributes are converted to lowercase
 655                 foreach ( $attribs as $attribute => $value ) {
 656                         if ( $attribute !== strtolower( $attribute ) && array_key_exists( strtolower( $attribute ), $presentationalAttribs ) ) {
 657                                 $attribs[strtolower( $attribute )] = $value;
 658                                 unset( $attribs[$attribute] );
 659                         }
 660                 }
 661
 662                 $style = "";
 663                 foreach ( $presentationalAttribs as $attribute => $info ) {
 664                         list( $property, $elements ) = $info;
 665
 666                         // Skip if this attribute is not relevant to this element
 667                         if ( !in_array( $element, $elements ) ) {
 668                                 continue;
 669                         }
 670
 671                         // Skip if the attribute is not used
 672                         if ( !array_key_exists( $attribute, $attribs ) ) {
 673                                 continue;
 674                         }
 675
 676                         $value = $attribs[$attribute];
 677
 678                         // For nowrap the value should be nowrap instead of whatever text is in the value
 679                         if ( $attribute === 'nowrap' ) {
 680                                 $value = 'nowrap';
 681                         }
 682
 683                         // clear="all" is clear: both; in css
 684                         if ( $attribute === 'clear' && strtolower( $value ) === 'all' ) {
 685                                 $value = 'both';
 686                         }
 687
 688                         // Size based properties should have px applied to them if they have no unit
 689                         if ( in_array( $attribute, array( 'height', 'width', 'size' ) ) ) {
 690                                 if ( preg_match( '/^[\d.]+$/', $value ) ) {
 691                                         $value = "{$value}px";
 692                                 }
 693                         }
 694
 695                         if ( $attribute === 'align' && !in_array( $element, $cells ) ) {
 696                                 if ( $value === 'center' ) {
 697                                         $style .= ' margin-left: auto;';
 698                                         $property = 'margin-right';
 699                                         $value = 'auto';
 700                                 } else {
 701                                         $property = 'float';
 702                                 }
 703                         }
 704
 705                         $style .= " $property: $value;";
 706
 707                         unset( $attribs[$attribute] );
 708                 }
 709
 710                 if ( $style ) {
 711                         // Prepend our style rules so that they can be overridden by user css
 712                         if ( isset($attribs['style']) ) {
 713                                 $style .= " " . $attribs['style'];
 714                         }
 715                         $attribs['style'] = trim($style);
 716                 }
 717
 718                 return $attribs;
 719         }
 720
 721         /**
 722          * Take an array of attribute names and values and normalize or discard
 723          * illegal values for the given element type.
 724          *
 725          * - Discards attributes not on a whitelist for the given element
 726          * - Unsafe style attributes are discarded
 727          * - Invalid id attributes are reencoded
 728          *
 729          * @param $attribs Array
 730          * @param $element String
 731          * @return Array
 732          *
 733          * @todo Check for legal values where the DTD limits things.
 734          * @todo Check for unique id attribute :P
 735          */
 736         static function validateTagAttributes( $attribs, $element ) {
 737                 return Sanitizer::validateAttributes( $attribs,
 738                         Sanitizer::attributeWhitelist( $element ) );
 739         }
 740
 741         /**
 742          * Take an array of attribute names and values and normalize or discard
 743          * illegal values for the given whitelist.
 744          *
 745          * - Discards attributes not the given whitelist
 746          * - Unsafe style attributes are discarded
 747          * - Invalid id attributes are reencoded
 748          *
 749          * @param $attribs Array
 750          * @param $whitelist Array: list of allowed attribute names
 751          * @return Array
 752          *
 753          * @todo Check for legal values where the DTD limits things.
 754          * @todo Check for unique id attribute :P
 755          */
 756         static function validateAttributes( $attribs, $whitelist ) {
 757                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
 758
 759                 $whitelist = array_flip( $whitelist );
 760                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 761
 762                 $out = array();
 763                 foreach( $attribs as $attribute => $value ) {
 764                         #allow XML namespace declaration if RDFa is enabled
 765                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 766                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 767                                         $out[$attribute] = $value;
 768                                 }
 769
 770                                 continue;
 771                         }
 772
 773                         # Allow any attribute beginning with "data-", if in HTML5 mode
 774                         if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
 775                                 continue;
 776                         }
 777
 778                         # Strip javascript "expression" from stylesheets.
 779                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 780                         if( $attribute == 'style' ) {
 781                                 $value = Sanitizer::checkCss( $value );
 782                         }
 783
 784                         if ( $attribute === 'id' ) {
 785                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 786                         }
 787
 788                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 789                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 790                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 791                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 792                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 793                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 794
 795                                 //Paranoia. Allow "simple" values but suppress javascript
 796                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 797                                         continue;
 798                                 }
 799                         }
 800
 801                         # NOTE: even though elements using href/src are not allowed directly, supply
 802                         #       validation code that can be used by tag hook handlers, etc
 803                         if ( $attribute === 'href' || $attribute === 'src' ) {
 804                                 if ( !preg_match( $hrefExp, $value ) ) {
 805                                         continue; //drop any href or src attributes not using an allowed protocol.
 806                                                   //NOTE: this also drops all relative URLs
 807                                 }
 808                         }
 809
 810                         // If this attribute was previously set, override it.
 811                         // Output should only have one attribute of each name.
 812                         $out[$attribute] = $value;
 813                 }
 814
 815                 if ( $wgAllowMicrodataAttributes ) {
 816                         # itemtype, itemid, itemref don't make sense without itemscope
 817                         if ( !array_key_exists( 'itemscope', $out ) ) {
 818                                 unset( $out['itemtype'] );
 819                                 unset( $out['itemid'] );
 820                                 unset( $out['itemref'] );
 821                         }
 822                         # TODO: Strip itemprop if we aren't descendants of an itemscope.
 823                 }
 824                 return $out;
 825         }
 826
 827         /**
 828          * Merge two sets of HTML attributes.  Conflicting items in the second set
 829          * will override those in the first, except for 'class' attributes which
 830          * will be combined (if they're both strings).
 831          *
 832          * @todo implement merging for other attributes such as style
 833          * @param $a Array
 834          * @param $b Array
 835          * @return array
 836          */
 837         static function mergeAttributes( $a, $b ) {
 838                 $out = array_merge( $a, $b );
 839                 if( isset( $a['class'] ) && isset( $b['class'] )
 840                 && is_string( $a['class'] ) && is_string( $b['class'] )
 841                 && $a['class'] !== $b['class'] ) {
 842                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 843                                 -1, PREG_SPLIT_NO_EMPTY );
 844                         $out['class'] = implode( ' ', array_unique( $classes ) );
 845                 }
 846                 return $out;
 847         }
 848
 849         /**
 850          * Pick apart some CSS and check it for forbidden or unsafe structures.
 851          * Returns a sanitized string. This sanitized string will have
 852          * character references and escape sequences decoded, and comments
 853          * stripped. If the input is just too evil, only a comment complaining
 854          * about evilness will be returned.
 855          *
 856          * Currently URL references, 'expression', 'tps' are forbidden.
 857          *
 858          * NOTE: Despite the fact that character references are decoded, the
 859          * returned string may contain character references given certain
 860          * clever input strings. These character references must
 861          * be escaped before the return value is embedded in HTML.
 862          *
 863          * @param $value String
 864          * @return String
 865          */
 866         static function checkCss( $value ) {
 867                 // Decode character references like &#123;
 868                 $value = Sanitizer::decodeCharReferences( $value );
 869
 870                 // Decode escape sequences and line continuation
 871                 // See the grammar in the CSS 2 spec, appendix D.
 872                 // This has to be done AFTER decoding character references.
 873                 // This means it isn't possible for this function to return
 874                 // unsanitized escape sequences. It is possible to manufacture
 875                 // input that contains character references that decode to
 876                 // escape sequences that decode to character references, but
 877                 // it's OK for the return value to contain character references
 878                 // because the caller is supposed to escape those anyway.
 879                 static $decodeRegex;
 880                 if ( !$decodeRegex ) {
 881                         $space = '[\\x20\\t\\r\\n\\f]';
 882                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 883                         $backslash = '\\\\';
 884                         $decodeRegex = "/ $backslash
 885                                 (?:
 886                                         ($nl) |  # 1. Line continuation
 887                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 888                                         (.) | # 3. backslash cancelling special meaning
 889                                         () | # 4. backslash at end of string
 890                                 )/xu";
 891                 }
 892                 $value = preg_replace_callback( $decodeRegex,
 893                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 894
 895                 // Remove any comments; IE gets token splitting wrong
 896                 // This must be done AFTER decoding character references and
 897                 // escape sequences, because those steps can introduce comments
 898                 // This step cannot introduce character references or escape
 899                 // sequences, because it replaces comments with spaces rather
 900                 // than removing them completely.
 901                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 902
 903                 // Remove anything after a comment-start token, to guard against
 904                 // incorrect client implementations.
 905                 $commentPos = strpos( $value, '/*' );
 906                 if ( $commentPos !== false ) {
 907                         $value = substr( $value, 0, $commentPos );
 908                 }
 909
 910                 // Reject problematic keywords and control characters
 911                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 912                         return '/* invalid control char */';
 913                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
 914                         return '/* insecure input */';
 915                 }
 916                 return $value;
 917         }
 918
 919         /**
 920          * @param $matches array
 921          * @return String
 922          */
 923         static function cssDecodeCallback( $matches ) {
 924                 if ( $matches[1] !== '' ) {
 925                         // Line continuation
 926                         return '';
 927                 } elseif ( $matches[2] !== '' ) {
 928                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 929                 } elseif ( $matches[3] !== '' ) {
 930                         $char = $matches[3];
 931                 } else {
 932                         $char = '\\';
 933                 }
 934                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 935                         // These characters need to be escaped in strings
 936                         // Clean up the escape sequence to avoid parsing errors by clients
 937                         return '\\' . dechex( ord( $char ) ) . ' ';
 938                 } else {
 939                         // Decode unnecessary escape
 940                         return $char;
 941                 }
 942         }
 943
 944         /**
 945          * Take a tag soup fragment listing an HTML element's attributes
 946          * and normalize it to well-formed XML, discarding unwanted attributes.
 947          * Output is safe for further wikitext processing, with escaping of
 948          * values that could trigger problems.
 949          *
 950          * - Normalizes attribute names to lowercase
 951          * - Discards attributes not on a whitelist for the given element
 952          * - Turns broken or invalid entities into plaintext
 953          * - Double-quotes all attribute values
 954          * - Attributes without values are given the name as attribute
 955          * - Double attributes are discarded
 956          * - Unsafe style attributes are discarded
 957          * - Prepends space if there are attributes.
 958          *
 959          * @param $text String
 960          * @param $element String
 961          * @return String
 962          */
 963         static function fixTagAttributes( $text, $element ) {
 964                 if( trim( $text ) == '' ) {
 965                         return '';
 966                 }
 967
 968                 $decoded = Sanitizer::decodeTagAttributes( $text );
 969                 $decoded = Sanitizer::fixDeprecatedAttributes( $decoded, $element );
 970                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
 971
 972                 $attribs = array();
 973                 foreach( $stripped as $attribute => $value ) {
 974                         $encAttribute = htmlspecialchars( $attribute );
 975                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 976
 977                         $attribs[] = "$encAttribute=\"$encValue\"";
 978                 }
 979                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 980         }
 981
 982         /**
 983          * Encode an attribute value for HTML output.
 984          * @param $text String
 985          * @return HTML-encoded text fragment
 986          */
 987         static function encodeAttribute( $text ) {
 988                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 989
 990                 // Whitespace is normalized during attribute decoding,
 991                 // so if we've been passed non-spaces we must encode them
 992                 // ahead of time or they won't be preserved.
 993                 $encValue = strtr( $encValue, array(
 994                         "\n" => '&#10;',
 995                         "\r" => '&#13;',
 996                         "\t" => '&#9;',
 997                 ) );
 998
 999                 return $encValue;
1000         }
1001
1002         /**
1003          * Encode an attribute value for HTML tags, with extra armoring
1004          * against further wiki processing.
1005          * @param $text String
1006          * @return HTML-encoded text fragment
1007          */
1008         static function safeEncodeAttribute( $text ) {
1009                 $encValue = Sanitizer::encodeAttribute( $text );
1010
1011                 # Templates and links may be expanded in later parsing,
1012                 # creating invalid or dangerous output. Suppress this.
1013                 $encValue = strtr( $encValue, array(
1014                         '<'    => '&lt;',   // This should never happen,
1015                         '>'    => '&gt;',   // we've received invalid input
1016                         '"'    => '&quot;', // which should have been escaped.
1017                         '{'    => '&#123;',
1018                         '['    => '&#91;',
1019                         "''"   => '&#39;&#39;',
1020                         'ISBN' => '&#73;SBN',
1021                         'RFC'  => '&#82;FC',
1022                         'PMID' => '&#80;MID',
1023                         '|'    => '&#124;',
1024                         '__'   => '&#95;_',
1025                 ) );
1026
1027                 # Stupid hack
1028                 $encValue = preg_replace_callback(
1029                         '/(' . wfUrlProtocols() . ')/',
1030                         array( 'Sanitizer', 'armorLinksCallback' ),
1031                         $encValue );
1032                 return $encValue;
1033         }
1034
1035         /**
1036          * Given a value, escape it so that it can be used in an id attribute and
1037          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1038          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1039          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1040          * escaped with lots of dots.
1041          *
1042          * To ensure we don't have to bother escaping anything, we also strip ', ",
1043          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1044          * We also strip # because it upsets IE, and % because it could be
1045          * ambiguous if it's part of something that looks like a percent escape
1046          * (which don't work reliably in fragments cross-browser).
1047          *
1048          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1049          *                                                          in the id and
1050          *                                                          name attributes
1051          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1052          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1053          *   HTML5 definition of id attribute
1054          *
1055          * @param $id String: id to escape
1056          * @param $options Mixed: string or array of strings (default is array()):
1057          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1058          *       so don't pay attention if the first character isn't valid at the
1059          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1060          *       false.
1061          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1062          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1063          *       anchors and links won't break.
1064          * @return String
1065          */
1066         static function escapeId( $id, $options = array() ) {
1067                 global $wgHtml5, $wgExperimentalHtmlIds;
1068                 $options = (array)$options;
1069
1070                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1071                         $id = Sanitizer::decodeCharReferences( $id );
1072                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1073                         $id = trim( $id, '_' );
1074                         if ( $id === '' ) {
1075                                 # Must have been all whitespace to start with.
1076                                 return '_';
1077                         } else {
1078                                 return $id;
1079                         }
1080                 }
1081
1082                 # HTML4-style escaping
1083                 static $replace = array(
1084                         '%3A' => ':',
1085                         '%' => '.'
1086                 );
1087
1088                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1089                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1090
1091                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1092                 && !in_array( 'noninitial', $options ) )  {
1093                         // Initial character must be a letter!
1094                         $id = "x$id";
1095                 }
1096                 return $id;
1097         }
1098
1099         /**
1100          * Given a value, escape it so that it can be used as a CSS class and
1101          * return it.
1102          *
1103          * @todo For extra validity, input should be validated UTF-8.
1104          *
1105          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1106          *
1107          * @param $class String
1108          * @return String
1109          */
1110         static function escapeClass( $class ) {
1111                 // Convert ugly stuff to underscores and kill underscores in ugly places
1112                 return rtrim(preg_replace(
1113                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
1114                         '_',
1115                         $class ), '_');
1116         }
1117
1118         /**
1119          * Given HTML input, escape with htmlspecialchars but un-escape entites.
1120          * This allows (generally harmless) entities like &#160; to survive.
1121          *
1122          * @param $html String to escape
1123          * @return String: escaped input
1124          */
1125         static function escapeHtmlAllowEntities( $html ) {
1126                 $html = Sanitizer::decodeCharReferences( $html );
1127                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1128                 # hurt.
1129                 $html = htmlspecialchars( $html, ENT_QUOTES );
1130                 return $html;
1131         }
1132
1133         /**
1134          * Regex replace callback for armoring links against further processing.
1135          * @param $matches Array
1136          * @return string
1137          */
1138         private static function armorLinksCallback( $matches ) {
1139                 return str_replace( ':', '&#58;', $matches[1] );
1140         }
1141
1142         /**
1143          * Return an associative array of attribute names and values from
1144          * a partial tag string. Attribute names are forces to lowercase,
1145          * character references are decoded to UTF-8 text.
1146          *
1147          * @param $text String
1148          * @return Array
1149          */
1150         public static function decodeTagAttributes( $text ) {
1151                 if( trim( $text ) == '' ) {
1152                         return array();
1153                 }
1154
1155                 $attribs = array();
1156                 $pairs = array();
1157                 if( !preg_match_all(
1158                         self::getAttribsRegex(),
1159                         $text,
1160                         $pairs,
1161                         PREG_SET_ORDER ) ) {
1162                         return $attribs;
1163                 }
1164
1165                 foreach( $pairs as $set ) {
1166                         $attribute = strtolower( $set[1] );
1167                         $value = Sanitizer::getTagAttributeCallback( $set );
1168
1169                         // Normalize whitespace
1170                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1171                         $value = trim( $value );
1172
1173                         // Decode character references
1174                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1175                 }
1176                 return $attribs;
1177         }
1178
1179         /**
1180          * Pick the appropriate attribute value from a match set from the
1181          * attribs regex matches.
1182          *
1183          * @param $set Array
1184          * @return String
1185          */
1186         private static function getTagAttributeCallback( $set ) {
1187                 if( isset( $set[6] ) ) {
1188                         # Illegal #XXXXXX color with no quotes.
1189                         return $set[6];
1190                 } elseif( isset( $set[5] ) ) {
1191                         # No quotes.
1192                         return $set[5];
1193                 } elseif( isset( $set[4] ) ) {
1194                         # Single-quoted
1195                         return $set[4];
1196                 } elseif( isset( $set[3] ) ) {
1197                         # Double-quoted
1198                         return $set[3];
1199                 } elseif( !isset( $set[2] ) ) {
1200                         # In XHTML, attributes must have a value.
1201                         # For 'reduced' form, return explicitly the attribute name here.
1202                         return $set[1];
1203                 } else {
1204                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1205                 }
1206         }
1207
1208         /**
1209          * Normalize whitespace and character references in an XML source-
1210          * encoded text for an attribute value.
1211          *
1212          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1213          * but note that we're not returning the value, but are returning
1214          * XML source fragments that will be slapped into output.
1215          *
1216          * @param $text String
1217          * @return String
1218          */
1219         private static function normalizeAttributeValue( $text ) {
1220                 return str_replace( '"', '&quot;',
1221                         self::normalizeWhitespace(
1222                                 Sanitizer::normalizeCharReferences( $text ) ) );
1223         }
1224
1225         /**
1226          * @param $text string
1227          * @return mixed
1228          */
1229         private static function normalizeWhitespace( $text ) {
1230                 return preg_replace(
1231                         '/\r\n|[\x20\x0d\x0a\x09]/',
1232                         ' ',
1233                         $text );
1234         }
1235
1236         /**
1237          * Normalizes whitespace in a section name, such as might be returned
1238          * by Parser::stripSectionName(), for use in the id's that are used for
1239          * section links.
1240          *
1241          * @param $section String
1242          * @return String
1243          */
1244         static function normalizeSectionNameWhitespace( $section ) {
1245                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1246         }
1247
1248         /**
1249          * Ensure that any entities and character references are legal
1250          * for XML and XHTML specifically. Any stray bits will be
1251          * &amp;-escaped to result in a valid text fragment.
1252          *
1253          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1254          *   numericized (this way we're well-formed even without a DTD)
1255          * b. any numeric char refs must be legal chars, not invalid or forbidden
1256          * c. use lower cased "&#x", not "&#X"
1257          * d. fix or reject non-valid attributes
1258          *
1259          * @param $text String
1260          * @return String
1261          * @private
1262          */
1263         static function normalizeCharReferences( $text ) {
1264                 return preg_replace_callback(
1265                         self::CHAR_REFS_REGEX,
1266                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1267                         $text );
1268         }
1269         /**
1270          * @param $matches String
1271          * @return String
1272          */
1273         static function normalizeCharReferencesCallback( $matches ) {
1274                 $ret = null;
1275                 if( $matches[1] != '' ) {
1276                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1277                 } elseif( $matches[2] != '' ) {
1278                         $ret = Sanitizer::decCharReference( $matches[2] );
1279                 } elseif( $matches[3] != ''  ) {
1280                         $ret = Sanitizer::hexCharReference( $matches[3] );
1281                 }
1282                 if( is_null( $ret ) ) {
1283                         return htmlspecialchars( $matches[0] );
1284                 } else {
1285                         return $ret;
1286                 }
1287         }
1288
1289         /**
1290          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1291          * return the equivalent numeric entity reference (except for the core &lt;
1292          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1293          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1294          * pseudo-entity source (eg &amp;foo;)
1295          *
1296          * @param $name String
1297          * @return String
1298          */
1299         static function normalizeEntity( $name ) {
1300                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1301                         return '&' . self::$htmlEntityAliases[$name] . ';';
1302                 } elseif ( in_array( $name,
1303                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1304                         return "&$name;";
1305                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1306                         return '&#' . self::$htmlEntities[$name] . ';';
1307                 } else {
1308                         return "&amp;$name;";
1309                 }
1310         }
1311
1312         /**
1313          * @param $codepoint
1314          * @return null|string
1315          */
1316         static function decCharReference( $codepoint ) {
1317                 $point = intval( $codepoint );
1318                 if( Sanitizer::validateCodepoint( $point ) ) {
1319                         return sprintf( '&#%d;', $point );
1320                 } else {
1321                         return null;
1322                 }
1323         }
1324
1325         /**
1326          * @param $codepoint
1327          * @return null|string
1328          */
1329         static function hexCharReference( $codepoint ) {
1330                 $point = hexdec( $codepoint );
1331                 if( Sanitizer::validateCodepoint( $point ) ) {
1332                         return sprintf( '&#x%x;', $point );
1333                 } else {
1334                         return null;
1335                 }
1336         }
1337
1338         /**
1339          * Returns true if a given Unicode codepoint is a valid character in XML.
1340          * @param $codepoint Integer
1341          * @return Boolean
1342          */
1343         private static function validateCodepoint( $codepoint ) {
1344                 return ($codepoint ==    0x09)
1345                         || ($codepoint ==    0x0a)
1346                         || ($codepoint ==    0x0d)
1347                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1348                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1349                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1350         }
1351
1352         /**
1353          * Decode any character references, numeric or named entities,
1354          * in the text and return a UTF-8 string.
1355          *
1356          * @param $text String
1357          * @return String
1358          */
1359         public static function decodeCharReferences( $text ) {
1360                 return preg_replace_callback(
1361                         self::CHAR_REFS_REGEX,
1362                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1363                         $text );
1364         }
1365
1366         /**
1367          * Decode any character references, numeric or named entities,
1368          * in the next and normalize the resulting string. (bug 14952)
1369          *
1370          * This is useful for page titles, not for text to be displayed,
1371          * MediaWiki allows HTML entities to escape normalization as a feature.
1372          *
1373          * @param $text String (already normalized, containing entities)
1374          * @return String (still normalized, without entities)
1375          */
1376         public static function decodeCharReferencesAndNormalize( $text ) {
1377                 global $wgContLang;
1378                 $text = preg_replace_callback(
1379                         self::CHAR_REFS_REGEX,
1380                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1381                         $text, /* limit */ -1, $count );
1382
1383                 if ( $count ) {
1384                         return $wgContLang->normalize( $text );
1385                 } else {
1386                         return $text;
1387                 }
1388         }
1389
1390         /**
1391          * @param $matches String
1392          * @return String
1393          */
1394         static function decodeCharReferencesCallback( $matches ) {
1395                 if( $matches[1] != '' ) {
1396                         return Sanitizer::decodeEntity( $matches[1] );
1397                 } elseif( $matches[2] != '' ) {
1398                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1399                 } elseif( $matches[3] != ''  ) {
1400                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1401                 }
1402                 # Last case should be an ampersand by itself
1403                 return $matches[0];
1404         }
1405
1406         /**
1407          * Return UTF-8 string for a codepoint if that is a valid
1408          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1409          * @param $codepoint Integer
1410          * @return String
1411          * @private
1412          */
1413         static function decodeChar( $codepoint ) {
1414                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1415                         return codepointToUtf8( $codepoint );
1416                 } else {
1417                         return UTF8_REPLACEMENT;
1418                 }
1419         }
1420
1421         /**
1422          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1423          * return the UTF-8 encoding of that character. Otherwise, returns
1424          * pseudo-entity source (eg "&foo;")
1425          *
1426          * @param $name String
1427          * @return String
1428          */
1429         static function decodeEntity( $name ) {
1430                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1431                         $name = self::$htmlEntityAliases[$name];
1432                 }
1433                 if( isset( self::$htmlEntities[$name] ) ) {
1434                         return codepointToUtf8( self::$htmlEntities[$name] );
1435                 } else {
1436                         return "&$name;";
1437                 }
1438         }
1439
1440         /**
1441          * Fetch the whitelist of acceptable attributes for a given element name.
1442          *
1443          * @param $element String
1444          * @return Array
1445          */
1446         static function attributeWhitelist( $element ) {
1447                 static $list;
1448                 if( !isset( $list ) ) {
1449                         $list = Sanitizer::setupAttributeWhitelist();
1450                 }
1451                 return isset( $list[$element] )
1452                         ? $list[$element]
1453                         : array();
1454         }
1455
1456         /**
1457          * Foreach array key (an allowed HTML element), return an array
1458          * of allowed attributes
1459          * @return Array
1460          */
1461         static function setupAttributeWhitelist() {
1462                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1463
1464                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1465
1466                 if ( $wgAllowRdfaAttributes ) {
1467                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1468                         $common = array_merge( $common, array(
1469                             'about', 'property', 'resource', 'datatype', 'typeof',
1470                         ) );
1471                 }
1472
1473                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1474                         # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1475                         $common = array_merge( $common, array(
1476                             'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1477                         ) );
1478                 }
1479
1480                 $block = array_merge( $common, array( 'align' ) );
1481                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1482                 $tablecell = array( 'abbr',
1483                                     'axis',
1484                                     'headers',
1485                                     'scope',
1486                                     'rowspan',
1487                                     'colspan',
1488                                     'nowrap', # deprecated
1489                                     'width',  # deprecated
1490                                     'height', # deprecated
1491                                     'bgcolor' # deprecated
1492                                     );
1493
1494                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1495                 # See: http://www.w3.org/TR/html4/
1496                 $whitelist = array (
1497                         # 7.5.4
1498                         'div'        => $block,
1499                         'center'     => $common, # deprecated
1500                         'span'       => $block, # ??
1501
1502                         # 7.5.5
1503                         'h1'         => $block,
1504                         'h2'         => $block,
1505                         'h3'         => $block,
1506                         'h4'         => $block,
1507                         'h5'         => $block,
1508                         'h6'         => $block,
1509
1510                         # 7.5.6
1511                         # address
1512
1513                         # 8.2.4
1514                         # bdo
1515
1516                         # 9.2.1
1517                         'em'         => $common,
1518                         'strong'     => $common,
1519                         'cite'       => $common,
1520                         'dfn'        => $common,
1521                         'code'       => $common,
1522                         'samp'       => $common,
1523                         'kbd'        => $common,
1524                         'var'        => $common,
1525                         'abbr'       => $common,
1526                         # acronym
1527
1528                         # 9.2.2
1529                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1530                         # q
1531
1532                         # 9.2.3
1533                         'sub'        => $common,
1534                         'sup'        => $common,
1535
1536                         # 9.3.1
1537                         'p'          => $block,
1538
1539                         # 9.3.2
1540                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1541
1542                         # 9.3.4
1543                         'pre'        => array_merge( $common, array( 'width' ) ),
1544
1545                         # 9.4
1546                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1547                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1548
1549                         # 10.2
1550                         'ul'         => array_merge( $common, array( 'type' ) ),
1551                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1552                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1553
1554                         # 10.3
1555                         'dl'         => $common,
1556                         'dd'         => $common,
1557                         'dt'         => $common,
1558
1559                         # 11.2.1
1560                         'table'      => array_merge( $common,
1561                                                                 array( 'summary', 'width', 'border', 'frame',
1562                                                                                 'rules', 'cellspacing', 'cellpadding',
1563                                                                                 'align', 'bgcolor',
1564                                                                 ) ),
1565
1566                         # 11.2.2
1567                         'caption'    => array_merge( $common, array( 'align' ) ),
1568
1569                         # 11.2.3
1570                         'thead'      => array_merge( $common, $tablealign ),
1571                         'tfoot'      => array_merge( $common, $tablealign ),
1572                         'tbody'      => array_merge( $common, $tablealign ),
1573
1574                         # 11.2.4
1575                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1576                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1577
1578                         # 11.2.5
1579                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1580
1581                         # 11.2.6
1582                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1583                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1584
1585                         # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1586                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1587
1588                         # 13.2
1589                         # Not usually allowed, but may be used for extension-style hooks
1590                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1591                         # true
1592                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1593
1594                         # 15.2.1
1595                         'tt'         => $common,
1596                         'b'          => $common,
1597                         'i'          => $common,
1598                         'big'        => $common,
1599                         'small'      => $common,
1600                         'strike'     => $common,
1601                         's'          => $common,
1602                         'u'          => $common,
1603
1604                         # 15.2.2
1605                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1606                         # basefont
1607
1608                         # 15.3
1609                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1610
1611                         # XHTML Ruby annotation text module, simple ruby only.
1612                         # http://www.w3c.org/TR/ruby/
1613                         'ruby'       => $common,
1614                         # rbc
1615                         # rtc
1616                         'rb'         => $common,
1617                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1618                         'rp'         => $common,
1619
1620                         # MathML root element, where used for extensions
1621                         # 'title' may not be 100% valid here; it's XHTML
1622                         # http://www.w3.org/TR/REC-MathML/
1623                         'math'       => array( 'class', 'style', 'id', 'title' ),
1624
1625                         # HTML 5 section 4.6
1626                         'bdi' => $common,
1627
1628                         );
1629                 return $whitelist;
1630         }
1631
1632         /**
1633          * Take a fragment of (potentially invalid) HTML and return
1634          * a version with any tags removed, encoded as plain text.
1635          *
1636          * Warning: this return value must be further escaped for literal
1637          * inclusion in HTML output as of 1.10!
1638          *
1639          * @param $text String: HTML fragment
1640          * @return String
1641          */
1642         static function stripAllTags( $text ) {
1643                 # Actual <tags>
1644                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1645
1646                 # Normalize &entities and whitespace
1647                 $text = self::decodeCharReferences( $text );
1648                 $text = self::normalizeWhitespace( $text );
1649
1650                 return $text;
1651         }
1652
1653         /**
1654          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1655          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1656          * PHP 5.1 doesn't.
1657          *
1658          * Use for passing XHTML fragments to PHP's XML parsing functions
1659          *
1660          * @return String
1661          */
1662         static function hackDocType() {
1663                 $out = "<!DOCTYPE html [\n";
1664                 foreach( self::$htmlEntities as $entity => $codepoint ) {
1665                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1666                 }
1667                 $out .= "]>\n";
1668                 return $out;
1669         }
1670
1671         /**
1672          * @param $url string
1673          * @return mixed|string
1674          */
1675         static function cleanUrl( $url ) {
1676                 # Normalize any HTML entities in input. They will be
1677                 # re-escaped by makeExternalLink().
1678                 $url = Sanitizer::decodeCharReferences( $url );
1679
1680                 # Escape any control characters introduced by the above step
1681                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1682                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1683
1684                 # Validate hostname portion
1685                 $matches = array();
1686                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1687                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1688
1689                         // Characters that will be ignored in IDNs.
1690                         // http://tools.ietf.org/html/3454#section-3.1
1691                         // Strip them before further processing so blacklists and such work.
1692                         $strip = "/
1693                                 \\s|          # general whitespace
1694                                 \xc2\xad|     # 00ad SOFT HYPHEN
1695                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1696                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1697                                 \xe2\x81\xa0| # 2060 WORD JOINER
1698                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1699                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1700                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1701                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1702                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1703                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1704                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1705                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1706                                 /xuD";
1707
1708                         $host = preg_replace( $strip, '', $host );
1709
1710                         // @todo FIXME: Validate hostnames here
1711
1712                         return $protocol . $host . $rest;
1713                 } else {
1714                         return $url;
1715                 }
1716         }
1717
1718         /**
1719          * @param $matches array
1720          * @return string
1721          */
1722         static function cleanUrlCallback( $matches ) {
1723                 return urlencode( $matches[0] );
1724         }
1725
1726         /**
1727          * Does a string look like an e-mail address?
1728          *
1729          * This validates an email address using an HTML5 specification found at:
1730          * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1731          * Which as of 2011-01-24 says:
1732          *
1733          *   A valid e-mail address is a string that matches the ABNF production
1734          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1735          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1736          *   3.5.
1737          *
1738          * This function is an implementation of the specification as requested in
1739          * bug 22449.
1740          *
1741          * Client-side forms will use the same standard validation rules via JS or
1742          * HTML 5 validation; additional restrictions can be enforced server-side
1743          * by extensions via the 'isValidEmailAddr' hook.
1744          *
1745          * Note that this validation doesn't 100% match RFC 2822, but is believed
1746          * to be liberal enough for wide use. Some invalid addresses will still
1747          * pass validation here.
1748          *
1749          * @since 1.18
1750          *
1751          * @param $addr String E-mail address
1752          * @return Bool
1753          */
1754         public static function validateEmail( $addr ) {
1755                 $result = null;
1756                 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1757                         return $result;
1758                 }
1759
1760                 // Please note strings below are enclosed in brackets [], this make the
1761                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1762                 // See bug 26948
1763                 $rfc5322_atext   = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1764                 $rfc1034_ldh_str = "a-z0-9\\-" ;
1765
1766                 $HTML5_email_regexp = "/
1767                 ^                      # start of string
1768                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1769                 @                      # 'apostrophe'
1770                 [$rfc1034_ldh_str]+       # First domain part
1771                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1772                 $                      # End of string
1773                 /ix" ; // case Insensitive, eXtended
1774
1775                 return (bool) preg_match( $HTML5_email_regexp, $addr );
1776         }
1777 }