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                         // Table align is special, it's about block alignment instead of
 696                         // content align (see also bug 40306)
 697                         if ( $attribute === 'align' && in_array( $element, $table ) ) {
 698                                 if ( $value === 'center' ) {
 699                                         $style .= ' margin-left: auto;';
 700                                         $property = 'margin-right';
 701                                         $value = 'auto';
 702                                 } else {
 703                                         $property = 'float';
 704                                 }
 705                         }
 706
 707                         $style .= " $property: $value;";
 708
 709                         unset( $attribs[$attribute] );
 710                 }
 711
 712                 if ( $style ) {
 713                         // Prepend our style rules so that they can be overridden by user css
 714                         if ( isset($attribs['style']) ) {
 715                                 $style .= " " . $attribs['style'];
 716                         }
 717                         $attribs['style'] = trim($style);
 718                 }
 719
 720                 return $attribs;
 721         }
 722
 723         /**
 724          * Take an array of attribute names and values and normalize or discard
 725          * illegal values for the given element type.
 726          *
 727          * - Discards attributes not on a whitelist for the given element
 728          * - Unsafe style attributes are discarded
 729          * - Invalid id attributes are reencoded
 730          *
 731          * @param $attribs Array
 732          * @param $element String
 733          * @return Array
 734          *
 735          * @todo Check for legal values where the DTD limits things.
 736          * @todo Check for unique id attribute :P
 737          */
 738         static function validateTagAttributes( $attribs, $element ) {
 739                 return Sanitizer::validateAttributes( $attribs,
 740                         Sanitizer::attributeWhitelist( $element ) );
 741         }
 742
 743         /**
 744          * Take an array of attribute names and values and normalize or discard
 745          * illegal values for the given whitelist.
 746          *
 747          * - Discards attributes not the given whitelist
 748          * - Unsafe style attributes are discarded
 749          * - Invalid id attributes are reencoded
 750          *
 751          * @param $attribs Array
 752          * @param $whitelist Array: list of allowed attribute names
 753          * @return Array
 754          *
 755          * @todo Check for legal values where the DTD limits things.
 756          * @todo Check for unique id attribute :P
 757          */
 758         static function validateAttributes( $attribs, $whitelist ) {
 759                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
 760
 761                 $whitelist = array_flip( $whitelist );
 762                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 763
 764                 $out = array();
 765                 foreach( $attribs as $attribute => $value ) {
 766                         #allow XML namespace declaration if RDFa is enabled
 767                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 768                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 769                                         $out[$attribute] = $value;
 770                                 }
 771
 772                                 continue;
 773                         }
 774
 775                         # Allow any attribute beginning with "data-", if in HTML5 mode
 776                         if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
 777                                 continue;
 778                         }
 779
 780                         # Strip javascript "expression" from stylesheets.
 781                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 782                         if( $attribute == 'style' ) {
 783                                 $value = Sanitizer::checkCss( $value );
 784                         }
 785
 786                         if ( $attribute === 'id' ) {
 787                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 788                         }
 789
 790                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 791                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 792                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 793                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 794                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 795                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 796
 797                                 //Paranoia. Allow "simple" values but suppress javascript
 798                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 799                                         continue;
 800                                 }
 801                         }
 802
 803                         # NOTE: even though elements using href/src are not allowed directly, supply
 804                         #       validation code that can be used by tag hook handlers, etc
 805                         if ( $attribute === 'href' || $attribute === 'src' ) {
 806                                 if ( !preg_match( $hrefExp, $value ) ) {
 807                                         continue; //drop any href or src attributes not using an allowed protocol.
 808                                                   //NOTE: this also drops all relative URLs
 809                                 }
 810                         }
 811
 812                         // If this attribute was previously set, override it.
 813                         // Output should only have one attribute of each name.
 814                         $out[$attribute] = $value;
 815                 }
 816
 817                 if ( $wgAllowMicrodataAttributes ) {
 818                         # itemtype, itemid, itemref don't make sense without itemscope
 819                         if ( !array_key_exists( 'itemscope', $out ) ) {
 820                                 unset( $out['itemtype'] );
 821                                 unset( $out['itemid'] );
 822                                 unset( $out['itemref'] );
 823                         }
 824                         # TODO: Strip itemprop if we aren't descendants of an itemscope.
 825                 }
 826                 return $out;
 827         }
 828
 829         /**
 830          * Merge two sets of HTML attributes.  Conflicting items in the second set
 831          * will override those in the first, except for 'class' attributes which
 832          * will be combined (if they're both strings).
 833          *
 834          * @todo implement merging for other attributes such as style
 835          * @param $a Array
 836          * @param $b Array
 837          * @return array
 838          */
 839         static function mergeAttributes( $a, $b ) {
 840                 $out = array_merge( $a, $b );
 841                 if( isset( $a['class'] ) && isset( $b['class'] )
 842                 && is_string( $a['class'] ) && is_string( $b['class'] )
 843                 && $a['class'] !== $b['class'] ) {
 844                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 845                                 -1, PREG_SPLIT_NO_EMPTY );
 846                         $out['class'] = implode( ' ', array_unique( $classes ) );
 847                 }
 848                 return $out;
 849         }
 850
 851         /**
 852          * Pick apart some CSS and check it for forbidden or unsafe structures.
 853          * Returns a sanitized string. This sanitized string will have
 854          * character references and escape sequences decoded, and comments
 855          * stripped. If the input is just too evil, only a comment complaining
 856          * about evilness will be returned.
 857          *
 858          * Currently URL references, 'expression', 'tps' are forbidden.
 859          *
 860          * NOTE: Despite the fact that character references are decoded, the
 861          * returned string may contain character references given certain
 862          * clever input strings. These character references must
 863          * be escaped before the return value is embedded in HTML.
 864          *
 865          * @param $value String
 866          * @return String
 867          */
 868         static function checkCss( $value ) {
 869                 // Decode character references like &#123;
 870                 $value = Sanitizer::decodeCharReferences( $value );
 871
 872                 // Decode escape sequences and line continuation
 873                 // See the grammar in the CSS 2 spec, appendix D.
 874                 // This has to be done AFTER decoding character references.
 875                 // This means it isn't possible for this function to return
 876                 // unsanitized escape sequences. It is possible to manufacture
 877                 // input that contains character references that decode to
 878                 // escape sequences that decode to character references, but
 879                 // it's OK for the return value to contain character references
 880                 // because the caller is supposed to escape those anyway.
 881                 static $decodeRegex;
 882                 if ( !$decodeRegex ) {
 883                         $space = '[\\x20\\t\\r\\n\\f]';
 884                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 885                         $backslash = '\\\\';
 886                         $decodeRegex = "/ $backslash
 887                                 (?:
 888                                         ($nl) |  # 1. Line continuation
 889                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 890                                         (.) | # 3. backslash cancelling special meaning
 891                                         () | # 4. backslash at end of string
 892                                 )/xu";
 893                 }
 894                 $value = preg_replace_callback( $decodeRegex,
 895                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 896
 897                 // Remove any comments; IE gets token splitting wrong
 898                 // This must be done AFTER decoding character references and
 899                 // escape sequences, because those steps can introduce comments
 900                 // This step cannot introduce character references or escape
 901                 // sequences, because it replaces comments with spaces rather
 902                 // than removing them completely.
 903                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 904
 905                 // Remove anything after a comment-start token, to guard against
 906                 // incorrect client implementations.
 907                 $commentPos = strpos( $value, '/*' );
 908                 if ( $commentPos !== false ) {
 909                         $value = substr( $value, 0, $commentPos );
 910                 }
 911
 912                 // Reject problematic keywords and control characters
 913                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 914                         return '/* invalid control char */';
 915                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
 916                         return '/* insecure input */';
 917                 }
 918                 return $value;
 919         }
 920
 921         /**
 922          * @param $matches array
 923          * @return String
 924          */
 925         static function cssDecodeCallback( $matches ) {
 926                 if ( $matches[1] !== '' ) {
 927                         // Line continuation
 928                         return '';
 929                 } elseif ( $matches[2] !== '' ) {
 930                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 931                 } elseif ( $matches[3] !== '' ) {
 932                         $char = $matches[3];
 933                 } else {
 934                         $char = '\\';
 935                 }
 936                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 937                         // These characters need to be escaped in strings
 938                         // Clean up the escape sequence to avoid parsing errors by clients
 939                         return '\\' . dechex( ord( $char ) ) . ' ';
 940                 } else {
 941                         // Decode unnecessary escape
 942                         return $char;
 943                 }
 944         }
 945
 946         /**
 947          * Take a tag soup fragment listing an HTML element's attributes
 948          * and normalize it to well-formed XML, discarding unwanted attributes.
 949          * Output is safe for further wikitext processing, with escaping of
 950          * values that could trigger problems.
 951          *
 952          * - Normalizes attribute names to lowercase
 953          * - Discards attributes not on a whitelist for the given element
 954          * - Turns broken or invalid entities into plaintext
 955          * - Double-quotes all attribute values
 956          * - Attributes without values are given the name as attribute
 957          * - Double attributes are discarded
 958          * - Unsafe style attributes are discarded
 959          * - Prepends space if there are attributes.
 960          *
 961          * @param $text String
 962          * @param $element String
 963          * @return String
 964          */
 965         static function fixTagAttributes( $text, $element ) {
 966                 if( trim( $text ) == '' ) {
 967                         return '';
 968                 }
 969
 970                 $decoded = Sanitizer::decodeTagAttributes( $text );
 971                 $decoded = Sanitizer::fixDeprecatedAttributes( $decoded, $element );
 972                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
 973
 974                 $attribs = array();
 975                 foreach( $stripped as $attribute => $value ) {
 976                         $encAttribute = htmlspecialchars( $attribute );
 977                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 978
 979                         $attribs[] = "$encAttribute=\"$encValue\"";
 980                 }
 981                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 982         }
 983
 984         /**
 985          * Encode an attribute value for HTML output.
 986          * @param $text String
 987          * @return HTML-encoded text fragment
 988          */
 989         static function encodeAttribute( $text ) {
 990                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 991
 992                 // Whitespace is normalized during attribute decoding,
 993                 // so if we've been passed non-spaces we must encode them
 994                 // ahead of time or they won't be preserved.
 995                 $encValue = strtr( $encValue, array(
 996                         "\n" => '&#10;',
 997                         "\r" => '&#13;',
 998                         "\t" => '&#9;',
 999                 ) );
1000
1001                 return $encValue;
1002         }
1003
1004         /**
1005          * Encode an attribute value for HTML tags, with extra armoring
1006          * against further wiki processing.
1007          * @param $text String
1008          * @return HTML-encoded text fragment
1009          */
1010         static function safeEncodeAttribute( $text ) {
1011                 $encValue = Sanitizer::encodeAttribute( $text );
1012
1013                 # Templates and links may be expanded in later parsing,
1014                 # creating invalid or dangerous output. Suppress this.
1015                 $encValue = strtr( $encValue, array(
1016                         '<'    => '&lt;',   // This should never happen,
1017                         '>'    => '&gt;',   // we've received invalid input
1018                         '"'    => '&quot;', // which should have been escaped.
1019                         '{'    => '&#123;',
1020                         '['    => '&#91;',
1021                         "''"   => '&#39;&#39;',
1022                         'ISBN' => '&#73;SBN',
1023                         'RFC'  => '&#82;FC',
1024                         'PMID' => '&#80;MID',
1025                         '|'    => '&#124;',
1026                         '__'   => '&#95;_',
1027                 ) );
1028
1029                 # Stupid hack
1030                 $encValue = preg_replace_callback(
1031                         '/((?i)' . wfUrlProtocols() . ')/',
1032                         array( 'Sanitizer', 'armorLinksCallback' ),
1033                         $encValue );
1034                 return $encValue;
1035         }
1036
1037         /**
1038          * Given a value, escape it so that it can be used in an id attribute and
1039          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1040          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1041          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1042          * escaped with lots of dots.
1043          *
1044          * To ensure we don't have to bother escaping anything, we also strip ', ",
1045          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1046          * We also strip # because it upsets IE, and % because it could be
1047          * ambiguous if it's part of something that looks like a percent escape
1048          * (which don't work reliably in fragments cross-browser).
1049          *
1050          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1051          *                                                          in the id and
1052          *                                                          name attributes
1053          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1054          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1055          *   HTML5 definition of id attribute
1056          *
1057          * @param $id String: id to escape
1058          * @param $options Mixed: string or array of strings (default is array()):
1059          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1060          *       so don't pay attention if the first character isn't valid at the
1061          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1062          *       false.
1063          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1064          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1065          *       anchors and links won't break.
1066          * @return String
1067          */
1068         static function escapeId( $id, $options = array() ) {
1069                 global $wgHtml5, $wgExperimentalHtmlIds;
1070                 $options = (array)$options;
1071
1072                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1073                         $id = Sanitizer::decodeCharReferences( $id );
1074                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1075                         $id = trim( $id, '_' );
1076                         if ( $id === '' ) {
1077                                 # Must have been all whitespace to start with.
1078                                 return '_';
1079                         } else {
1080                                 return $id;
1081                         }
1082                 }
1083
1084                 # HTML4-style escaping
1085                 static $replace = array(
1086                         '%3A' => ':',
1087                         '%' => '.'
1088                 );
1089
1090                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1091                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1092
1093                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1094                 && !in_array( 'noninitial', $options ) )  {
1095                         // Initial character must be a letter!
1096                         $id = "x$id";
1097                 }
1098                 return $id;
1099         }
1100
1101         /**
1102          * Given a value, escape it so that it can be used as a CSS class and
1103          * return it.
1104          *
1105          * @todo For extra validity, input should be validated UTF-8.
1106          *
1107          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1108          *
1109          * @param $class String
1110          * @return String
1111          */
1112         static function escapeClass( $class ) {
1113                 // Convert ugly stuff to underscores and kill underscores in ugly places
1114                 return rtrim(preg_replace(
1115                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
1116                         '_',
1117                         $class ), '_');
1118         }
1119
1120         /**
1121          * Given HTML input, escape with htmlspecialchars but un-escape entites.
1122          * This allows (generally harmless) entities like &#160; to survive.
1123          *
1124          * @param $html String to escape
1125          * @return String: escaped input
1126          */
1127         static function escapeHtmlAllowEntities( $html ) {
1128                 $html = Sanitizer::decodeCharReferences( $html );
1129                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1130                 # hurt.
1131                 $html = htmlspecialchars( $html, ENT_QUOTES );
1132                 return $html;
1133         }
1134
1135         /**
1136          * Regex replace callback for armoring links against further processing.
1137          * @param $matches Array
1138          * @return string
1139          */
1140         private static function armorLinksCallback( $matches ) {
1141                 return str_replace( ':', '&#58;', $matches[1] );
1142         }
1143
1144         /**
1145          * Return an associative array of attribute names and values from
1146          * a partial tag string. Attribute names are forces to lowercase,
1147          * character references are decoded to UTF-8 text.
1148          *
1149          * @param $text String
1150          * @return Array
1151          */
1152         public static function decodeTagAttributes( $text ) {
1153                 if( trim( $text ) == '' ) {
1154                         return array();
1155                 }
1156
1157                 $attribs = array();
1158                 $pairs = array();
1159                 if( !preg_match_all(
1160                         self::getAttribsRegex(),
1161                         $text,
1162                         $pairs,
1163                         PREG_SET_ORDER ) ) {
1164                         return $attribs;
1165                 }
1166
1167                 foreach( $pairs as $set ) {
1168                         $attribute = strtolower( $set[1] );
1169                         $value = Sanitizer::getTagAttributeCallback( $set );
1170
1171                         // Normalize whitespace
1172                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1173                         $value = trim( $value );
1174
1175                         // Decode character references
1176                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1177                 }
1178                 return $attribs;
1179         }
1180
1181         /**
1182          * Pick the appropriate attribute value from a match set from the
1183          * attribs regex matches.
1184          *
1185          * @param $set Array
1186          * @return String
1187          */
1188         private static function getTagAttributeCallback( $set ) {
1189                 if( isset( $set[6] ) ) {
1190                         # Illegal #XXXXXX color with no quotes.
1191                         return $set[6];
1192                 } elseif( isset( $set[5] ) ) {
1193                         # No quotes.
1194                         return $set[5];
1195                 } elseif( isset( $set[4] ) ) {
1196                         # Single-quoted
1197                         return $set[4];
1198                 } elseif( isset( $set[3] ) ) {
1199                         # Double-quoted
1200                         return $set[3];
1201                 } elseif( !isset( $set[2] ) ) {
1202                         # In XHTML, attributes must have a value.
1203                         # For 'reduced' form, return explicitly the attribute name here.
1204                         return $set[1];
1205                 } else {
1206                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1207                 }
1208         }
1209
1210         /**
1211          * Normalize whitespace and character references in an XML source-
1212          * encoded text for an attribute value.
1213          *
1214          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1215          * but note that we're not returning the value, but are returning
1216          * XML source fragments that will be slapped into output.
1217          *
1218          * @param $text String
1219          * @return String
1220          */
1221         private static function normalizeAttributeValue( $text ) {
1222                 return str_replace( '"', '&quot;',
1223                         self::normalizeWhitespace(
1224                                 Sanitizer::normalizeCharReferences( $text ) ) );
1225         }
1226
1227         /**
1228          * @param $text string
1229          * @return mixed
1230          */
1231         private static function normalizeWhitespace( $text ) {
1232                 return preg_replace(
1233                         '/\r\n|[\x20\x0d\x0a\x09]/',
1234                         ' ',
1235                         $text );
1236         }
1237
1238         /**
1239          * Normalizes whitespace in a section name, such as might be returned
1240          * by Parser::stripSectionName(), for use in the id's that are used for
1241          * section links.
1242          *
1243          * @param $section String
1244          * @return String
1245          */
1246         static function normalizeSectionNameWhitespace( $section ) {
1247                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1248         }
1249
1250         /**
1251          * Ensure that any entities and character references are legal
1252          * for XML and XHTML specifically. Any stray bits will be
1253          * &amp;-escaped to result in a valid text fragment.
1254          *
1255          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1256          *   numericized (this way we're well-formed even without a DTD)
1257          * b. any numeric char refs must be legal chars, not invalid or forbidden
1258          * c. use lower cased "&#x", not "&#X"
1259          * d. fix or reject non-valid attributes
1260          *
1261          * @param $text String
1262          * @return String
1263          * @private
1264          */
1265         static function normalizeCharReferences( $text ) {
1266                 return preg_replace_callback(
1267                         self::CHAR_REFS_REGEX,
1268                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1269                         $text );
1270         }
1271         /**
1272          * @param $matches String
1273          * @return String
1274          */
1275         static function normalizeCharReferencesCallback( $matches ) {
1276                 $ret = null;
1277                 if( $matches[1] != '' ) {
1278                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1279                 } elseif( $matches[2] != '' ) {
1280                         $ret = Sanitizer::decCharReference( $matches[2] );
1281                 } elseif( $matches[3] != ''  ) {
1282                         $ret = Sanitizer::hexCharReference( $matches[3] );
1283                 }
1284                 if( is_null( $ret ) ) {
1285                         return htmlspecialchars( $matches[0] );
1286                 } else {
1287                         return $ret;
1288                 }
1289         }
1290
1291         /**
1292          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1293          * return the equivalent numeric entity reference (except for the core &lt;
1294          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1295          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1296          * pseudo-entity source (eg &amp;foo;)
1297          *
1298          * @param $name String
1299          * @return String
1300          */
1301         static function normalizeEntity( $name ) {
1302                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1303                         return '&' . self::$htmlEntityAliases[$name] . ';';
1304                 } elseif ( in_array( $name,
1305                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1306                         return "&$name;";
1307                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1308                         return '&#' . self::$htmlEntities[$name] . ';';
1309                 } else {
1310                         return "&amp;$name;";
1311                 }
1312         }
1313
1314         /**
1315          * @param $codepoint
1316          * @return null|string
1317          */
1318         static function decCharReference( $codepoint ) {
1319                 $point = intval( $codepoint );
1320                 if( Sanitizer::validateCodepoint( $point ) ) {
1321                         return sprintf( '&#%d;', $point );
1322                 } else {
1323                         return null;
1324                 }
1325         }
1326
1327         /**
1328          * @param $codepoint
1329          * @return null|string
1330          */
1331         static function hexCharReference( $codepoint ) {
1332                 $point = hexdec( $codepoint );
1333                 if( Sanitizer::validateCodepoint( $point ) ) {
1334                         return sprintf( '&#x%x;', $point );
1335                 } else {
1336                         return null;
1337                 }
1338         }
1339
1340         /**
1341          * Returns true if a given Unicode codepoint is a valid character in XML.
1342          * @param $codepoint Integer
1343          * @return Boolean
1344          */
1345         private static function validateCodepoint( $codepoint ) {
1346                 return ($codepoint ==    0x09)
1347                         || ($codepoint ==    0x0a)
1348                         || ($codepoint ==    0x0d)
1349                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1350                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1351                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1352         }
1353
1354         /**
1355          * Decode any character references, numeric or named entities,
1356          * in the text and return a UTF-8 string.
1357          *
1358          * @param $text String
1359          * @return String
1360          */
1361         public static function decodeCharReferences( $text ) {
1362                 return preg_replace_callback(
1363                         self::CHAR_REFS_REGEX,
1364                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1365                         $text );
1366         }
1367
1368         /**
1369          * Decode any character references, numeric or named entities,
1370          * in the next and normalize the resulting string. (bug 14952)
1371          *
1372          * This is useful for page titles, not for text to be displayed,
1373          * MediaWiki allows HTML entities to escape normalization as a feature.
1374          *
1375          * @param $text String (already normalized, containing entities)
1376          * @return String (still normalized, without entities)
1377          */
1378         public static function decodeCharReferencesAndNormalize( $text ) {
1379                 global $wgContLang;
1380                 $text = preg_replace_callback(
1381                         self::CHAR_REFS_REGEX,
1382                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1383                         $text, /* limit */ -1, $count );
1384
1385                 if ( $count ) {
1386                         return $wgContLang->normalize( $text );
1387                 } else {
1388                         return $text;
1389                 }
1390         }
1391
1392         /**
1393          * @param $matches String
1394          * @return String
1395          */
1396         static function decodeCharReferencesCallback( $matches ) {
1397                 if( $matches[1] != '' ) {
1398                         return Sanitizer::decodeEntity( $matches[1] );
1399                 } elseif( $matches[2] != '' ) {
1400                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1401                 } elseif( $matches[3] != ''  ) {
1402                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1403                 }
1404                 # Last case should be an ampersand by itself
1405                 return $matches[0];
1406         }
1407
1408         /**
1409          * Return UTF-8 string for a codepoint if that is a valid
1410          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1411          * @param $codepoint Integer
1412          * @return String
1413          * @private
1414          */
1415         static function decodeChar( $codepoint ) {
1416                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1417                         return codepointToUtf8( $codepoint );
1418                 } else {
1419                         return UTF8_REPLACEMENT;
1420                 }
1421         }
1422
1423         /**
1424          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1425          * return the UTF-8 encoding of that character. Otherwise, returns
1426          * pseudo-entity source (eg "&foo;")
1427          *
1428          * @param $name String
1429          * @return String
1430          */
1431         static function decodeEntity( $name ) {
1432                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1433                         $name = self::$htmlEntityAliases[$name];
1434                 }
1435                 if( isset( self::$htmlEntities[$name] ) ) {
1436                         return codepointToUtf8( self::$htmlEntities[$name] );
1437                 } else {
1438                         return "&$name;";
1439                 }
1440         }
1441
1442         /**
1443          * Fetch the whitelist of acceptable attributes for a given element name.
1444          *
1445          * @param $element String
1446          * @return Array
1447          */
1448         static function attributeWhitelist( $element ) {
1449                 static $list;
1450                 if( !isset( $list ) ) {
1451                         $list = Sanitizer::setupAttributeWhitelist();
1452                 }
1453                 return isset( $list[$element] )
1454                         ? $list[$element]
1455                         : array();
1456         }
1457
1458         /**
1459          * Foreach array key (an allowed HTML element), return an array
1460          * of allowed attributes
1461          * @return Array
1462          */
1463         static function setupAttributeWhitelist() {
1464                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1465
1466                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1467
1468                 if ( $wgAllowRdfaAttributes ) {
1469                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1470                         $common = array_merge( $common, array(
1471                             'about', 'property', 'resource', 'datatype', 'typeof',
1472                         ) );
1473                 }
1474
1475                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1476                         # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1477                         $common = array_merge( $common, array(
1478                             'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1479                         ) );
1480                 }
1481
1482                 $block = array_merge( $common, array( 'align' ) );
1483                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1484                 $tablecell = array( 'abbr',
1485                                     'axis',
1486                                     'headers',
1487                                     'scope',
1488                                     'rowspan',
1489                                     'colspan',
1490                                     'nowrap', # deprecated
1491                                     'width',  # deprecated
1492                                     'height', # deprecated
1493                                     'bgcolor' # deprecated
1494                                     );
1495
1496                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1497                 # See: http://www.w3.org/TR/html4/
1498                 $whitelist = array (
1499                         # 7.5.4
1500                         'div'        => $block,
1501                         'center'     => $common, # deprecated
1502                         'span'       => $block, # ??
1503
1504                         # 7.5.5
1505                         'h1'         => $block,
1506                         'h2'         => $block,
1507                         'h3'         => $block,
1508                         'h4'         => $block,
1509                         'h5'         => $block,
1510                         'h6'         => $block,
1511
1512                         # 7.5.6
1513                         # address
1514
1515                         # 8.2.4
1516                         # bdo
1517
1518                         # 9.2.1
1519                         'em'         => $common,
1520                         'strong'     => $common,
1521                         'cite'       => $common,
1522                         'dfn'        => $common,
1523                         'code'       => $common,
1524                         'samp'       => $common,
1525                         'kbd'        => $common,
1526                         'var'        => $common,
1527                         'abbr'       => $common,
1528                         # acronym
1529
1530                         # 9.2.2
1531                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1532                         # q
1533
1534                         # 9.2.3
1535                         'sub'        => $common,
1536                         'sup'        => $common,
1537
1538                         # 9.3.1
1539                         'p'          => $block,
1540
1541                         # 9.3.2
1542                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1543
1544                         # 9.3.4
1545                         'pre'        => array_merge( $common, array( 'width' ) ),
1546
1547                         # 9.4
1548                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1549                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1550
1551                         # 10.2
1552                         'ul'         => array_merge( $common, array( 'type' ) ),
1553                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1554                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1555
1556                         # 10.3
1557                         'dl'         => $common,
1558                         'dd'         => $common,
1559                         'dt'         => $common,
1560
1561                         # 11.2.1
1562                         'table'      => array_merge( $common,
1563                                                                 array( 'summary', 'width', 'border', 'frame',
1564                                                                                 'rules', 'cellspacing', 'cellpadding',
1565                                                                                 'align', 'bgcolor',
1566                                                                 ) ),
1567
1568                         # 11.2.2
1569                         'caption'    => array_merge( $common, array( 'align' ) ),
1570
1571                         # 11.2.3
1572                         'thead'      => array_merge( $common, $tablealign ),
1573                         'tfoot'      => array_merge( $common, $tablealign ),
1574                         'tbody'      => array_merge( $common, $tablealign ),
1575
1576                         # 11.2.4
1577                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1578                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1579
1580                         # 11.2.5
1581                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1582
1583                         # 11.2.6
1584                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1585                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1586
1587                         # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1588                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1589
1590                         # 13.2
1591                         # Not usually allowed, but may be used for extension-style hooks
1592                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1593                         # true
1594                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1595
1596                         # 15.2.1
1597                         'tt'         => $common,
1598                         'b'          => $common,
1599                         'i'          => $common,
1600                         'big'        => $common,
1601                         'small'      => $common,
1602                         'strike'     => $common,
1603                         's'          => $common,
1604                         'u'          => $common,
1605
1606                         # 15.2.2
1607                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1608                         # basefont
1609
1610                         # 15.3
1611                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1612
1613                         # XHTML Ruby annotation text module, simple ruby only.
1614                         # http://www.w3c.org/TR/ruby/
1615                         'ruby'       => $common,
1616                         # rbc
1617                         # rtc
1618                         'rb'         => $common,
1619                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1620                         'rp'         => $common,
1621
1622                         # MathML root element, where used for extensions
1623                         # 'title' may not be 100% valid here; it's XHTML
1624                         # http://www.w3.org/TR/REC-MathML/
1625                         'math'       => array( 'class', 'style', 'id', 'title' ),
1626
1627                         # HTML 5 section 4.6
1628                         'bdi' => $common,
1629
1630                         );
1631                 return $whitelist;
1632         }
1633
1634         /**
1635          * Take a fragment of (potentially invalid) HTML and return
1636          * a version with any tags removed, encoded as plain text.
1637          *
1638          * Warning: this return value must be further escaped for literal
1639          * inclusion in HTML output as of 1.10!
1640          *
1641          * @param $text String: HTML fragment
1642          * @return String
1643          */
1644         static function stripAllTags( $text ) {
1645                 # Actual <tags>
1646                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1647
1648                 # Normalize &entities and whitespace
1649                 $text = self::decodeCharReferences( $text );
1650                 $text = self::normalizeWhitespace( $text );
1651
1652                 return $text;
1653         }
1654
1655         /**
1656          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1657          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1658          * PHP 5.1 doesn't.
1659          *
1660          * Use for passing XHTML fragments to PHP's XML parsing functions
1661          *
1662          * @return String
1663          */
1664         static function hackDocType() {
1665                 $out = "<!DOCTYPE html [\n";
1666                 foreach( self::$htmlEntities as $entity => $codepoint ) {
1667                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1668                 }
1669                 $out .= "]>\n";
1670                 return $out;
1671         }
1672
1673         /**
1674          * @param $url string
1675          * @return mixed|string
1676          */
1677         static function cleanUrl( $url ) {
1678                 # Normalize any HTML entities in input. They will be
1679                 # re-escaped by makeExternalLink().
1680                 $url = Sanitizer::decodeCharReferences( $url );
1681
1682                 # Escape any control characters introduced by the above step
1683                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1684                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1685
1686                 # Validate hostname portion
1687                 $matches = array();
1688                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1689                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1690
1691                         // Characters that will be ignored in IDNs.
1692                         // http://tools.ietf.org/html/3454#section-3.1
1693                         // Strip them before further processing so blacklists and such work.
1694                         $strip = "/
1695                                 \\s|          # general whitespace
1696                                 \xc2\xad|     # 00ad SOFT HYPHEN
1697                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1698                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1699                                 \xe2\x81\xa0| # 2060 WORD JOINER
1700                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1701                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1702                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1703                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1704                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1705                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1706                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1707                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1708                                 /xuD";
1709
1710                         $host = preg_replace( $strip, '', $host );
1711
1712                         // @todo FIXME: Validate hostnames here
1713
1714                         return $protocol . $host . $rest;
1715                 } else {
1716                         return $url;
1717                 }
1718         }
1719
1720         /**
1721          * @param $matches array
1722          * @return string
1723          */
1724         static function cleanUrlCallback( $matches ) {
1725                 return urlencode( $matches[0] );
1726         }
1727
1728         /**
1729          * Does a string look like an e-mail address?
1730          *
1731          * This validates an email address using an HTML5 specification found at:
1732          * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1733          * Which as of 2011-01-24 says:
1734          *
1735          *   A valid e-mail address is a string that matches the ABNF production
1736          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1737          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1738          *   3.5.
1739          *
1740          * This function is an implementation of the specification as requested in
1741          * bug 22449.
1742          *
1743          * Client-side forms will use the same standard validation rules via JS or
1744          * HTML 5 validation; additional restrictions can be enforced server-side
1745          * by extensions via the 'isValidEmailAddr' hook.
1746          *
1747          * Note that this validation doesn't 100% match RFC 2822, but is believed
1748          * to be liberal enough for wide use. Some invalid addresses will still
1749          * pass validation here.
1750          *
1751          * @since 1.18
1752          *
1753          * @param $addr String E-mail address
1754          * @return Bool
1755          */
1756         public static function validateEmail( $addr ) {
1757                 $result = null;
1758                 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1759                         return $result;
1760                 }
1761
1762                 // Please note strings below are enclosed in brackets [], this make the
1763                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1764                 // See bug 26948
1765                 $rfc5322_atext   = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1766                 $rfc1034_ldh_str = "a-z0-9\\-" ;
1767
1768                 $HTML5_email_regexp = "/
1769                 ^                      # start of string
1770                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1771                 @                      # 'apostrophe'
1772                 [$rfc1034_ldh_str]+       # First domain part
1773                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1774                 $                      # End of string
1775                 /ix" ; // case Insensitive, eXtended
1776
1777                 return (bool) preg_match( $HTML5_email_regexp, $addr );
1778         }
1779 }