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 array $args for the processing callback
 362          * @param array $extratags for any extra tags to include
 363          * @param array $removetags for any tags (default or extra) to exclude
 364          * @return string
 365          */
 366         static function removeHTMLtags( $text, $processCallback = null, $args = array(), $extratags = array(), $removetags = array() ) {
 367                 global $wgUseTidy, $wgHtml5, $wgAllowMicrodataAttributes, $wgAllowImageTag;
 368
 369                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 370                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 371
 372                 wfProfileIn( __METHOD__ );
 373
 374                 // Base our staticInitialised variable off of the global config state so that if the globals
 375                 // are changed (like in the screwed up test system) we will re-initialise the settings.
 376                 $globalContext = implode( '-', compact( 'wgHtml5', 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
 377                 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
 378
 379                         $htmlpairsStatic = array( # Tags that must be closed
 380                                 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 381                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 382                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 383                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 384                                 'ruby', 'rt', 'rb', 'rp', 'p', 'span', 'abbr', 'dfn',
 385                                 'kbd', 'samp'
 386                         );
 387                         if ( $wgHtml5 ) {
 388                                 $htmlpairsStatic = array_merge( $htmlpairsStatic, array( 'data', 'time', 'mark' ) );
 389                         }
 390                         $htmlsingle = array(
 391                                 'br', 'hr', 'li', 'dt', 'dd'
 392                         );
 393                         $htmlsingleonly = array( # Elements that cannot have close tags
 394                                 'br', 'hr'
 395                         );
 396                         if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
 397                                 $htmlsingle[] = $htmlsingleonly[] = 'meta';
 398                                 $htmlsingle[] = $htmlsingleonly[] = 'link';
 399                         }
 400                         $htmlnest = array( # Tags that can be nested--??
 401                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 402                                 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span'
 403                         );
 404                         $tabletags = array( # Can only appear inside table, we will close them
 405                                 'td', 'th', 'tr',
 406                         );
 407                         $htmllist = array( # Tags used by list
 408                                 'ul', 'ol',
 409                         );
 410                         $listtags = array( # Tags that can appear in a list
 411                                 'li',
 412                         );
 413
 414                         if ( $wgAllowImageTag ) {
 415                                 $htmlsingle[] = 'img';
 416                                 $htmlsingleonly[] = 'img';
 417                         }
 418
 419                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 420                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 421
 422                         # Convert them all to hashtables for faster lookup
 423                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 424                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 425                         foreach ( $vars as $var ) {
 426                                 $$var = array_flip( $$var );
 427                         }
 428                         $staticInitialised = $globalContext;
 429                 }
 430                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 431                 $extratags = array_flip( $extratags );
 432                 $removetags = array_flip( $removetags );
 433                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 434                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
 435
 436                 # Remove HTML comments
 437                 $text = Sanitizer::removeHTMLcomments( $text );
 438                 $bits = explode( '<', $text );
 439                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 440                 if ( !$wgUseTidy ) {
 441                         $tagstack = $tablestack = array();
 442                         foreach ( $bits as $x ) {
 443                                 $regs = array();
 444                                 # $slash: Does the current element start with a '/'?
 445                                 # $t: Current element name
 446                                 # $params: String between element name and >
 447                                 # $brace: Ending '>' or '/>'
 448                                 # $rest: Everything until the next element of $bits
 449                                 if ( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 450                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 451                                 } else {
 452                                         $slash = $t = $params = $brace = $rest = null;
 453                                 }
 454
 455                                 $badtag = false;
 456                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 457                                         # Check our stack
 458                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 459                                                 $badtag = true;
 460                                         } elseif ( $slash ) {
 461                                                 # Closing a tag... is it the one we just opened?
 462                                                 $ot = @array_pop( $tagstack );
 463                                                 if ( $ot != $t ) {
 464                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 465                                                                 # Pop all elements with an optional close tag
 466                                                                 # and see if we find a match below them
 467                                                                 $optstack = array();
 468                                                                 array_push( $optstack, $ot );
 469                                                                 wfSuppressWarnings();
 470                                                                 $ot = array_pop( $tagstack );
 471                                                                 wfRestoreWarnings();
 472                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 473                                                                         array_push( $optstack, $ot );
 474                                                                         wfSuppressWarnings();
 475                                                                         $ot = array_pop( $tagstack );
 476                                                                         wfRestoreWarnings();
 477                                                                 }
 478                                                                 if ( $t != $ot ) {
 479                                                                         # No match. Push the optional elements back again
 480                                                                         $badtag = true;
 481                                                                         wfSuppressWarnings();
 482                                                                         $ot = array_pop( $optstack );
 483                                                                         wfRestoreWarnings();
 484                                                                         while ( $ot ) {
 485                                                                                 array_push( $tagstack, $ot );
 486                                                                                 wfSuppressWarnings();
 487                                                                                 $ot = array_pop( $optstack );
 488                                                                                 wfRestoreWarnings();
 489                                                                         }
 490                                                                 }
 491                                                         } else {
 492                                                                 @array_push( $tagstack, $ot );
 493                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 494                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 495                                                                         $badtag = true;
 496                                                                 }
 497                                                         }
 498                                                 } else {
 499                                                         if ( $t == 'table' ) {
 500                                                                 $tagstack = array_pop( $tablestack );
 501                                                         }
 502                                                 }
 503                                                 $newparams = '';
 504                                         } else {
 505                                                 # Keep track for later
 506                                                 if ( isset( $tabletags[$t] ) &&
 507                                                 !in_array( 'table', $tagstack ) ) {
 508                                                         $badtag = true;
 509                                                 } elseif ( in_array( $t, $tagstack ) &&
 510                                                 !isset( $htmlnest[$t] ) ) {
 511                                                         $badtag = true;
 512                                                 # Is it a self closed htmlpair ? (bug 5487)
 513                                                 } elseif ( $brace == '/>' &&
 514                                                 isset( $htmlpairs[$t] ) ) {
 515                                                         $badtag = true;
 516                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 517                                                         # Hack to force empty tag for unclosable elements
 518                                                         $brace = '/>';
 519                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 520                                                         # Hack to not close $htmlsingle tags
 521                                                         $brace = null;
 522                                                         # Still need to push this optionally-closed tag to
 523                                                         # the tag stack so that we can match end tags
 524                                                         # instead of marking them as bad.
 525                                                         array_push( $tagstack, $t );
 526                                                 } elseif ( isset( $tabletags[$t] )
 527                                                 && in_array( $t, $tagstack ) ) {
 528                                                         // New table tag but forgot to close the previous one
 529                                                         $text .= "</$t>";
 530                                                 } else {
 531                                                         if ( $t == 'table' ) {
 532                                                                 array_push( $tablestack, $tagstack );
 533                                                                 $tagstack = array();
 534                                                         }
 535                                                         array_push( $tagstack, $t );
 536                                                 }
 537
 538                                                 # Replace any variables or template parameters with
 539                                                 # plaintext results.
 540                                                 if ( is_callable( $processCallback ) ) {
 541                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 542                                                 }
 543
 544                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 545                                                         $badtag = true;
 546                                                 }
 547
 548                                                 # Strip non-approved attributes from the tag
 549                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 550                                         }
 551                                         if ( !$badtag ) {
 552                                                 $rest = str_replace( '>', '&gt;', $rest );
 553                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 554                                                 $text .= "<$slash$t$newparams$close>$rest";
 555                                                 continue;
 556                                         }
 557                                 }
 558                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 559                         }
 560                         # Close off any remaining tags
 561                         while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
 562                                 $text .= "</$t>\n";
 563                                 if ( $t == 'table' ) {
 564                                         $tagstack = array_pop( $tablestack );
 565                                 }
 566                         }
 567                 } else {
 568                         # this might be possible using tidy itself
 569                         foreach ( $bits as $x ) {
 570                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 571                                 $x, $regs );
 572                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 573                                 $badtag = false;
 574                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 575                                         if ( is_callable( $processCallback ) ) {
 576                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 577                                         }
 578
 579                                         if ( !Sanitizer::validateTag( $params, $t ) ) {
 580                                                 $badtag = true;
 581                                         }
 582
 583                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 584                                         if ( !$badtag ) {
 585                                                 $rest = str_replace( '>', '&gt;', $rest );
 586                                                 $text .= "<$slash$t$newparams$brace$rest";
 587                                                 continue;
 588                                         }
 589                                 }
 590                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 591                         }
 592                 }
 593                 wfProfileOut( __METHOD__ );
 594                 return $text;
 595         }
 596
 597         /**
 598          * Remove '<!--', '-->', and everything between.
 599          * To avoid leaving blank lines, when a comment is both preceded
 600          * and followed by a newline (ignoring spaces), trim leading and
 601          * trailing spaces and one of the newlines.
 602          *
 603          * @private
 604          * @param $text String
 605          * @return string
 606          */
 607         static function removeHTMLcomments( $text ) {
 608                 wfProfileIn( __METHOD__ );
 609                 while ( ($start = strpos( $text, '<!--' ) ) !== false ) {
 610                         $end = strpos( $text, '-->', $start + 4 );
 611                         if ( $end === false ) {
 612                                 # Unterminated comment; bail out
 613                                 break;
 614                         }
 615
 616                         $end += 3;
 617
 618                         # Trim space and newline if the comment is both
 619                         # preceded and followed by a newline
 620                         $spaceStart = max( $start - 1, 0 );
 621                         $spaceLen = $end - $spaceStart;
 622                         while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
 623                                 $spaceStart--;
 624                                 $spaceLen++;
 625                         }
 626                         while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
 627                                 $spaceLen++;
 628                         }
 629                         if ( substr( $text, $spaceStart, 1 ) === "\n" and substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
 630                                 # Remove the comment, leading and trailing
 631                                 # spaces, and leave only one newline.
 632                                 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
 633                         }
 634                         else {
 635                                 # Remove just the comment.
 636                                 $text = substr_replace( $text, '', $start, $end - $start );
 637                         }
 638                 }
 639                 wfProfileOut( __METHOD__ );
 640                 return $text;
 641         }
 642
 643         /**
 644          * Takes attribute names and values for a tag and the tag name and
 645          * validates that the tag is allowed to be present.
 646          * This DOES NOT validate the attributes, nor does it validate the
 647          * tags themselves. This method only handles the special circumstances
 648          * where we may want to allow a tag within content but ONLY when it has
 649          * specific attributes set.
 650          *
 651          * @param $params
 652          * @param $element
 653          * @return bool
 654          */
 655         static function validateTag( $params, $element ) {
 656                 $params = Sanitizer::decodeTagAttributes( $params );
 657
 658                 if ( $element == 'meta' || $element == 'link' ) {
 659                         if ( !isset( $params['itemprop'] ) ) {
 660                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 661                                 return false;
 662                         }
 663                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 664                                 // <meta> must have a content="" for the itemprop
 665                                 return false;
 666                         }
 667                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 668                                 // <link> must have an associated href=""
 669                                 return false;
 670                         }
 671                 }
 672
 673                 return true;
 674         }
 675
 676         /**
 677          * Take an array of attribute names and values and normalize or discard
 678          * illegal values for the given element type.
 679          *
 680          * - Discards attributes not on a whitelist for the given element
 681          * - Unsafe style attributes are discarded
 682          * - Invalid id attributes are re-encoded
 683          *
 684          * @param $attribs Array
 685          * @param $element String
 686          * @return Array
 687          *
 688          * @todo Check for legal values where the DTD limits things.
 689          * @todo Check for unique id attribute :P
 690          */
 691         static function validateTagAttributes( $attribs, $element ) {
 692                 return Sanitizer::validateAttributes( $attribs,
 693                         Sanitizer::attributeWhitelist( $element ) );
 694         }
 695
 696         /**
 697          * Take an array of attribute names and values and normalize or discard
 698          * illegal values for the given whitelist.
 699          *
 700          * - Discards attributes not the given whitelist
 701          * - Unsafe style attributes are discarded
 702          * - Invalid id attributes are re-encoded
 703          *
 704          * @param $attribs Array
 705          * @param array $whitelist list of allowed attribute names
 706          * @return Array
 707          *
 708          * @todo Check for legal values where the DTD limits things.
 709          * @todo Check for unique id attribute :P
 710          */
 711         static function validateAttributes( $attribs, $whitelist ) {
 712                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
 713
 714                 $whitelist = array_flip( $whitelist );
 715                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 716
 717                 $out = array();
 718                 foreach ( $attribs as $attribute => $value ) {
 719                         #allow XML namespace declaration if RDFa is enabled
 720                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 721                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 722                                         $out[$attribute] = $value;
 723                                 }
 724
 725                                 continue;
 726                         }
 727
 728                         # Allow any attribute beginning with "data-", if in HTML5 mode
 729                         if ( !( $wgHtml5 && preg_match( '/^data-/i', $attribute ) ) && !isset( $whitelist[$attribute] ) ) {
 730                                 continue;
 731                         }
 732
 733                         # Strip javascript "expression" from stylesheets.
 734                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 735                         if ( $attribute == 'style' ) {
 736                                 $value = Sanitizer::checkCss( $value );
 737                         }
 738
 739                         if ( $attribute === 'id' ) {
 740                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 741                         }
 742
 743                         # WAI-ARIA
 744                         # http://www.w3.org/TR/wai-aria/
 745                         # http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#wai-aria
 746                         # For now we only support role="presentation" until we work out what roles should be
 747                         # usable by content and we ensure that our code explicitly rejects patterns that
 748                         # violate HTML5's ARIA restrictions.
 749                         if ( $attribute === 'role' && $value !== 'presentation' ) {
 750                                 continue;
 751                         }
 752
 753                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 754                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 755                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 756                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 757                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 758                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 759
 760                                 //Paranoia. Allow "simple" values but suppress javascript
 761                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 762                                         continue;
 763                                 }
 764                         }
 765
 766                         # NOTE: even though elements using href/src are not allowed directly, supply
 767                         #       validation code that can be used by tag hook handlers, etc
 768                         if ( $attribute === 'href' || $attribute === 'src' ) {
 769                                 if ( !preg_match( $hrefExp, $value ) ) {
 770                                         continue; //drop any href or src attributes not using an allowed protocol.
 771                                                   //NOTE: this also drops all relative URLs
 772                                 }
 773                         }
 774
 775                         // If this attribute was previously set, override it.
 776                         // Output should only have one attribute of each name.
 777                         $out[$attribute] = $value;
 778                 }
 779
 780                 if ( $wgAllowMicrodataAttributes ) {
 781                         # itemtype, itemid, itemref don't make sense without itemscope
 782                         if ( !array_key_exists( 'itemscope', $out ) ) {
 783                                 unset( $out['itemtype'] );
 784                                 unset( $out['itemid'] );
 785                                 unset( $out['itemref'] );
 786                         }
 787                         # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 788                 }
 789                 return $out;
 790         }
 791
 792         /**
 793          * Merge two sets of HTML attributes.  Conflicting items in the second set
 794          * will override those in the first, except for 'class' attributes which
 795          * will be combined (if they're both strings).
 796          *
 797          * @todo implement merging for other attributes such as style
 798          * @param $a Array
 799          * @param $b Array
 800          * @return array
 801          */
 802         static function mergeAttributes( $a, $b ) {
 803                 $out = array_merge( $a, $b );
 804                 if ( isset( $a['class'] ) && isset( $b['class'] )
 805                         && is_string( $a['class'] ) && is_string( $b['class'] )
 806                         && $a['class'] !== $b['class']
 807                 ) {
 808                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 809                                 -1, PREG_SPLIT_NO_EMPTY );
 810                         $out['class'] = implode( ' ', array_unique( $classes ) );
 811                 }
 812                 return $out;
 813         }
 814
 815         /**
 816          * Pick apart some CSS and check it for forbidden or unsafe structures.
 817          * Returns a sanitized string. This sanitized string will have
 818          * character references and escape sequences decoded, and comments
 819          * stripped. If the input is just too evil, only a comment complaining
 820          * about evilness will be returned.
 821          *
 822          * Currently URL references, 'expression', 'tps' are forbidden.
 823          *
 824          * NOTE: Despite the fact that character references are decoded, the
 825          * returned string may contain character references given certain
 826          * clever input strings. These character references must
 827          * be escaped before the return value is embedded in HTML.
 828          *
 829          * @param $value String
 830          * @return String
 831          */
 832         static function checkCss( $value ) {
 833                 // Decode character references like &#123;
 834                 $value = Sanitizer::decodeCharReferences( $value );
 835
 836                 // Decode escape sequences and line continuation
 837                 // See the grammar in the CSS 2 spec, appendix D.
 838                 // This has to be done AFTER decoding character references.
 839                 // This means it isn't possible for this function to return
 840                 // unsanitized escape sequences. It is possible to manufacture
 841                 // input that contains character references that decode to
 842                 // escape sequences that decode to character references, but
 843                 // it's OK for the return value to contain character references
 844                 // because the caller is supposed to escape those anyway.
 845                 static $decodeRegex;
 846                 if ( !$decodeRegex ) {
 847                         $space = '[\\x20\\t\\r\\n\\f]';
 848                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 849                         $backslash = '\\\\';
 850                         $decodeRegex = "/ $backslash
 851                                 (?:
 852                                         ($nl) |  # 1. Line continuation
 853                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 854                                         (.) | # 3. backslash cancelling special meaning
 855                                         () | # 4. backslash at end of string
 856                                 )/xu";
 857                 }
 858                 $value = preg_replace_callback( $decodeRegex,
 859                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 860
 861                 // Remove any comments; IE gets token splitting wrong
 862                 // This must be done AFTER decoding character references and
 863                 // escape sequences, because those steps can introduce comments
 864                 // This step cannot introduce character references or escape
 865                 // sequences, because it replaces comments with spaces rather
 866                 // than removing them completely.
 867                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 868
 869                 // Remove anything after a comment-start token, to guard against
 870                 // incorrect client implementations.
 871                 $commentPos = strpos( $value, '/*' );
 872                 if ( $commentPos !== false ) {
 873                         $value = substr( $value, 0, $commentPos );
 874                 }
 875
 876                 // Reject problematic keywords and control characters
 877                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 878                         return '/* invalid control char */';
 879                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( | image\s*\( | image-set\s*\( !ix', $value ) ) {
 880                         return '/* insecure input */';
 881                 }
 882                 return $value;
 883         }
 884
 885         /**
 886          * @param $matches array
 887          * @return String
 888          */
 889         static function cssDecodeCallback( $matches ) {
 890                 if ( $matches[1] !== '' ) {
 891                         // Line continuation
 892                         return '';
 893                 } elseif ( $matches[2] !== '' ) {
 894                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 895                 } elseif ( $matches[3] !== '' ) {
 896                         $char = $matches[3];
 897                 } else {
 898                         $char = '\\';
 899                 }
 900                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 901                         // These characters need to be escaped in strings
 902                         // Clean up the escape sequence to avoid parsing errors by clients
 903                         return '\\' . dechex( ord( $char ) ) . ' ';
 904                 } else {
 905                         // Decode unnecessary escape
 906                         return $char;
 907                 }
 908         }
 909
 910         /**
 911          * Take a tag soup fragment listing an HTML element's attributes
 912          * and normalize it to well-formed XML, discarding unwanted attributes.
 913          * Output is safe for further wikitext processing, with escaping of
 914          * values that could trigger problems.
 915          *
 916          * - Normalizes attribute names to lowercase
 917          * - Discards attributes not on a whitelist for the given element
 918          * - Turns broken or invalid entities into plaintext
 919          * - Double-quotes all attribute values
 920          * - Attributes without values are given the name as attribute
 921          * - Double attributes are discarded
 922          * - Unsafe style attributes are discarded
 923          * - Prepends space if there are attributes.
 924          *
 925          * @param $text String
 926          * @param $element String
 927          * @return String
 928          */
 929         static function fixTagAttributes( $text, $element ) {
 930                 if ( trim( $text ) == '' ) {
 931                         return '';
 932                 }
 933
 934                 $decoded = Sanitizer::decodeTagAttributes( $text );
 935                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
 936
 937                 $attribs = array();
 938                 foreach ( $stripped as $attribute => $value ) {
 939                         $encAttribute = htmlspecialchars( $attribute );
 940                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 941
 942                         $attribs[] = "$encAttribute=\"$encValue\"";
 943                 }
 944                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 945         }
 946
 947         /**
 948          * Encode an attribute value for HTML output.
 949          * @param $text String
 950          * @return HTML-encoded text fragment
 951          */
 952         static function encodeAttribute( $text ) {
 953                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 954
 955                 // Whitespace is normalized during attribute decoding,
 956                 // so if we've been passed non-spaces we must encode them
 957                 // ahead of time or they won't be preserved.
 958                 $encValue = strtr( $encValue, array(
 959                         "\n" => '&#10;',
 960                         "\r" => '&#13;',
 961                         "\t" => '&#9;',
 962                 ) );
 963
 964                 return $encValue;
 965         }
 966
 967         /**
 968          * Encode an attribute value for HTML tags, with extra armoring
 969          * against further wiki processing.
 970          * @param $text String
 971          * @return HTML-encoded text fragment
 972          */
 973         static function safeEncodeAttribute( $text ) {
 974                 $encValue = Sanitizer::encodeAttribute( $text );
 975
 976                 # Templates and links may be expanded in later parsing,
 977                 # creating invalid or dangerous output. Suppress this.
 978                 $encValue = strtr( $encValue, array(
 979                         '<'    => '&lt;',   // This should never happen,
 980                         '>'    => '&gt;',   // we've received invalid input
 981                         '"'    => '&quot;', // which should have been escaped.
 982                         '{'    => '&#123;',
 983                         '['    => '&#91;',
 984                         "''"   => '&#39;&#39;',
 985                         'ISBN' => '&#73;SBN',
 986                         'RFC'  => '&#82;FC',
 987                         'PMID' => '&#80;MID',
 988                         '|'    => '&#124;',
 989                         '__'   => '&#95;_',
 990                 ) );
 991
 992                 # Stupid hack
 993                 $encValue = preg_replace_callback(
 994                         '/((?i)' . wfUrlProtocols() . ')/',
 995                         array( 'Sanitizer', 'armorLinksCallback' ),
 996                         $encValue );
 997                 return $encValue;
 998         }
 999
1000         /**
1001          * Given a value, escape it so that it can be used in an id attribute and
1002          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1003          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1004          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1005          * escaped with lots of dots.
1006          *
1007          * To ensure we don't have to bother escaping anything, we also strip ', ",
1008          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1009          * We also strip # because it upsets IE, and % because it could be
1010          * ambiguous if it's part of something that looks like a percent escape
1011          * (which don't work reliably in fragments cross-browser).
1012          *
1013          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1014          *                                                          in the id and
1015          *                                                          name attributes
1016          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1017          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1018          *   HTML5 definition of id attribute
1019          *
1020          * @param string $id id to escape
1021          * @param $options Mixed: string or array of strings (default is array()):
1022          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1023          *       so don't pay attention if the first character isn't valid at the
1024          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1025          *       false.
1026          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1027          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1028          *       anchors and links won't break.
1029          * @return String
1030          */
1031         static function escapeId( $id, $options = array() ) {
1032                 global $wgHtml5, $wgExperimentalHtmlIds;
1033                 $options = (array)$options;
1034
1035                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1036                         $id = Sanitizer::decodeCharReferences( $id );
1037                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1038                         $id = trim( $id, '_' );
1039                         if ( $id === '' ) {
1040                                 # Must have been all whitespace to start with.
1041                                 return '_';
1042                         } else {
1043                                 return $id;
1044                         }
1045                 }
1046
1047                 # HTML4-style escaping
1048                 static $replace = array(
1049                         '%3A' => ':',
1050                         '%' => '.'
1051                 );
1052
1053                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1054                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1055
1056                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1057                 && !in_array( 'noninitial', $options ) ) {
1058                         // Initial character must be a letter!
1059                         $id = "x$id";
1060                 }
1061                 return $id;
1062         }
1063
1064         /**
1065          * Given a value, escape it so that it can be used as a CSS class and
1066          * return it.
1067          *
1068          * @todo For extra validity, input should be validated UTF-8.
1069          *
1070          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1071          *
1072          * @param $class String
1073          * @return String
1074          */
1075         static function escapeClass( $class ) {
1076                 // Convert ugly stuff to underscores and kill underscores in ugly places
1077                 return rtrim( preg_replace(
1078                         array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1079                         '_',
1080                         $class ), '_' );
1081         }
1082
1083         /**
1084          * Given HTML input, escape with htmlspecialchars but un-escape entities.
1085          * This allows (generally harmless) entities like &#160; to survive.
1086          *
1087          * @param string $html to escape
1088          * @return String: escaped input
1089          */
1090         static function escapeHtmlAllowEntities( $html ) {
1091                 $html = Sanitizer::decodeCharReferences( $html );
1092                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1093                 # hurt.
1094                 $html = htmlspecialchars( $html, ENT_QUOTES );
1095                 return $html;
1096         }
1097
1098         /**
1099          * Regex replace callback for armoring links against further processing.
1100          * @param $matches Array
1101          * @return string
1102          */
1103         private static function armorLinksCallback( $matches ) {
1104                 return str_replace( ':', '&#58;', $matches[1] );
1105         }
1106
1107         /**
1108          * Return an associative array of attribute names and values from
1109          * a partial tag string. Attribute names are forces to lowercase,
1110          * character references are decoded to UTF-8 text.
1111          *
1112          * @param $text String
1113          * @return Array
1114          */
1115         public static function decodeTagAttributes( $text ) {
1116                 if ( trim( $text ) == '' ) {
1117                         return array();
1118                 }
1119
1120                 $attribs = array();
1121                 $pairs = array();
1122                 if ( !preg_match_all(
1123                         self::getAttribsRegex(),
1124                         $text,
1125                         $pairs,
1126                         PREG_SET_ORDER ) ) {
1127                         return $attribs;
1128                 }
1129
1130                 foreach ( $pairs as $set ) {
1131                         $attribute = strtolower( $set[1] );
1132                         $value = Sanitizer::getTagAttributeCallback( $set );
1133
1134                         // Normalize whitespace
1135                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1136                         $value = trim( $value );
1137
1138                         // Decode character references
1139                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1140                 }
1141                 return $attribs;
1142         }
1143
1144         /**
1145          * Pick the appropriate attribute value from a match set from the
1146          * attribs regex matches.
1147          *
1148          * @param $set Array
1149          * @throws MWException
1150          * @return String
1151          */
1152         private static function getTagAttributeCallback( $set ) {
1153                 if ( isset( $set[6] ) ) {
1154                         # Illegal #XXXXXX color with no quotes.
1155                         return $set[6];
1156                 } elseif ( isset( $set[5] ) ) {
1157                         # No quotes.
1158                         return $set[5];
1159                 } elseif ( isset( $set[4] ) ) {
1160                         # Single-quoted
1161                         return $set[4];
1162                 } elseif ( isset( $set[3] ) ) {
1163                         # Double-quoted
1164                         return $set[3];
1165                 } elseif ( !isset( $set[2] ) ) {
1166                         # In XHTML, attributes must have a value.
1167                         # For 'reduced' form, return explicitly the attribute name here.
1168                         return $set[1];
1169                 } else {
1170                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1171                 }
1172         }
1173
1174         /**
1175          * Normalize whitespace and character references in an XML source-
1176          * encoded text for an attribute value.
1177          *
1178          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1179          * but note that we're not returning the value, but are returning
1180          * XML source fragments that will be slapped into output.
1181          *
1182          * @param $text String
1183          * @return String
1184          */
1185         private static function normalizeAttributeValue( $text ) {
1186                 return str_replace( '"', '&quot;',
1187                         self::normalizeWhitespace(
1188                                 Sanitizer::normalizeCharReferences( $text ) ) );
1189         }
1190
1191         /**
1192          * @param $text string
1193          * @return mixed
1194          */
1195         private static function normalizeWhitespace( $text ) {
1196                 return preg_replace(
1197                         '/\r\n|[\x20\x0d\x0a\x09]/',
1198                         ' ',
1199                         $text );
1200         }
1201
1202         /**
1203          * Normalizes whitespace in a section name, such as might be returned
1204          * by Parser::stripSectionName(), for use in the id's that are used for
1205          * section links.
1206          *
1207          * @param $section String
1208          * @return String
1209          */
1210         static function normalizeSectionNameWhitespace( $section ) {
1211                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1212         }
1213
1214         /**
1215          * Ensure that any entities and character references are legal
1216          * for XML and XHTML specifically. Any stray bits will be
1217          * &amp;-escaped to result in a valid text fragment.
1218          *
1219          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1220          *   numericized (this way we're well-formed even without a DTD)
1221          * b. any numeric char refs must be legal chars, not invalid or forbidden
1222          * c. use lower cased "&#x", not "&#X"
1223          * d. fix or reject non-valid attributes
1224          *
1225          * @param $text String
1226          * @return String
1227          * @private
1228          */
1229         static function normalizeCharReferences( $text ) {
1230                 return preg_replace_callback(
1231                         self::CHAR_REFS_REGEX,
1232                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1233                         $text );
1234         }
1235         /**
1236          * @param $matches String
1237          * @return String
1238          */
1239         static function normalizeCharReferencesCallback( $matches ) {
1240                 $ret = null;
1241                 if ( $matches[1] != '' ) {
1242                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1243                 } elseif ( $matches[2] != '' ) {
1244                         $ret = Sanitizer::decCharReference( $matches[2] );
1245                 } elseif ( $matches[3] != '' ) {
1246                         $ret = Sanitizer::hexCharReference( $matches[3] );
1247                 }
1248                 if ( is_null( $ret ) ) {
1249                         return htmlspecialchars( $matches[0] );
1250                 } else {
1251                         return $ret;
1252                 }
1253         }
1254
1255         /**
1256          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1257          * return the equivalent numeric entity reference (except for the core &lt;
1258          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1259          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1260          * pseudo-entity source (eg &amp;foo;)
1261          *
1262          * @param $name String
1263          * @return String
1264          */
1265         static function normalizeEntity( $name ) {
1266                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1267                         return '&' . self::$htmlEntityAliases[$name] . ';';
1268                 } elseif ( in_array( $name,
1269                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1270                         return "&$name;";
1271                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1272                         return '&#' . self::$htmlEntities[$name] . ';';
1273                 } else {
1274                         return "&amp;$name;";
1275                 }
1276         }
1277
1278         /**
1279          * @param $codepoint
1280          * @return null|string
1281          */
1282         static function decCharReference( $codepoint ) {
1283                 $point = intval( $codepoint );
1284                 if ( Sanitizer::validateCodepoint( $point ) ) {
1285                         return sprintf( '&#%d;', $point );
1286                 } else {
1287                         return null;
1288                 }
1289         }
1290
1291         /**
1292          * @param $codepoint
1293          * @return null|string
1294          */
1295         static function hexCharReference( $codepoint ) {
1296                 $point = hexdec( $codepoint );
1297                 if ( Sanitizer::validateCodepoint( $point ) ) {
1298                         return sprintf( '&#x%x;', $point );
1299                 } else {
1300                         return null;
1301                 }
1302         }
1303
1304         /**
1305          * Returns true if a given Unicode codepoint is a valid character in XML.
1306          * @param $codepoint Integer
1307          * @return Boolean
1308          */
1309         private static function validateCodepoint( $codepoint ) {
1310                 return ($codepoint ==    0x09)
1311                         || ($codepoint ==    0x0a)
1312                         || ($codepoint ==    0x0d)
1313                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1314                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1315                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1316         }
1317
1318         /**
1319          * Decode any character references, numeric or named entities,
1320          * in the text and return a UTF-8 string.
1321          *
1322          * @param $text String
1323          * @return String
1324          */
1325         public static function decodeCharReferences( $text ) {
1326                 return preg_replace_callback(
1327                         self::CHAR_REFS_REGEX,
1328                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1329                         $text );
1330         }
1331
1332         /**
1333          * Decode any character references, numeric or named entities,
1334          * in the next and normalize the resulting string. (bug 14952)
1335          *
1336          * This is useful for page titles, not for text to be displayed,
1337          * MediaWiki allows HTML entities to escape normalization as a feature.
1338          *
1339          * @param string $text (already normalized, containing entities)
1340          * @return String (still normalized, without entities)
1341          */
1342         public static function decodeCharReferencesAndNormalize( $text ) {
1343                 global $wgContLang;
1344                 $text = preg_replace_callback(
1345                         self::CHAR_REFS_REGEX,
1346                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1347                         $text, /* limit */ -1, $count );
1348
1349                 if ( $count ) {
1350                         return $wgContLang->normalize( $text );
1351                 } else {
1352                         return $text;
1353                 }
1354         }
1355
1356         /**
1357          * @param $matches String
1358          * @return String
1359          */
1360         static function decodeCharReferencesCallback( $matches ) {
1361                 if ( $matches[1] != '' ) {
1362                         return Sanitizer::decodeEntity( $matches[1] );
1363                 } elseif ( $matches[2] != '' ) {
1364                         return Sanitizer::decodeChar( intval( $matches[2] ) );
1365                 } elseif ( $matches[3] != '' ) {
1366                         return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1367                 }
1368                 # Last case should be an ampersand by itself
1369                 return $matches[0];
1370         }
1371
1372         /**
1373          * Return UTF-8 string for a codepoint if that is a valid
1374          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1375          * @param $codepoint Integer
1376          * @return String
1377          * @private
1378          */
1379         static function decodeChar( $codepoint ) {
1380                 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1381                         return codepointToUtf8( $codepoint );
1382                 } else {
1383                         return UTF8_REPLACEMENT;
1384                 }
1385         }
1386
1387         /**
1388          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1389          * return the UTF-8 encoding of that character. Otherwise, returns
1390          * pseudo-entity source (eg "&foo;")
1391          *
1392          * @param $name String
1393          * @return String
1394          */
1395         static function decodeEntity( $name ) {
1396                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1397                         $name = self::$htmlEntityAliases[$name];
1398                 }
1399                 if ( isset( self::$htmlEntities[$name] ) ) {
1400                         return codepointToUtf8( self::$htmlEntities[$name] );
1401                 } else {
1402                         return "&$name;";
1403                 }
1404         }
1405
1406         /**
1407          * Fetch the whitelist of acceptable attributes for a given element name.
1408          *
1409          * @param $element String
1410          * @return Array
1411          */
1412         static function attributeWhitelist( $element ) {
1413                 $list = Sanitizer::setupAttributeWhitelist();
1414                 return isset( $list[$element] )
1415                         ? $list[$element]
1416                         : array();
1417         }
1418
1419         /**
1420          * Foreach array key (an allowed HTML element), return an array
1421          * of allowed attributes
1422          * @return Array
1423          */
1424         static function setupAttributeWhitelist() {
1425                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1426
1427                 static $whitelist, $staticInitialised;
1428                 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgHtml5', 'wgAllowMicrodataAttributes' ) );
1429
1430                 if ( isset( $whitelist ) && $staticInitialised == $globalContext ) {
1431                         return $whitelist;
1432                 }
1433
1434                 $common = array(
1435                         # HTML
1436                         'id',
1437                         'class',
1438                         'style',
1439                         'lang',
1440                         'dir',
1441                         'title',
1442
1443                         # WAI-ARIA
1444                         'role',
1445                 );
1446
1447                 if ( $wgAllowRdfaAttributes ) {
1448                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1449                         $common = array_merge( $common, array(
1450                                 'about', 'property', 'resource', 'datatype', 'typeof',
1451                         ) );
1452                 }
1453
1454                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1455                         # add HTML5 microdata tags as specified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1456                         $common = array_merge( $common, array(
1457                                 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1458                         ) );
1459                 }
1460
1461                 $block = array_merge( $common, array( 'align' ) );
1462                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1463                 $tablecell = array(
1464                         'abbr',
1465                         'axis',
1466                         'headers',
1467                         'scope',
1468                         'rowspan',
1469                         'colspan',
1470                         'nowrap', # deprecated
1471                         'width', # deprecated
1472                         'height', # deprecated
1473                         'bgcolor', # deprecated
1474                 );
1475
1476                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1477                 # See: http://www.w3.org/TR/html4/
1478                 $whitelist = array(
1479                         # 7.5.4
1480                         'div'        => $block,
1481                         'center'     => $common, # deprecated
1482                         'span'       => $block, # ??
1483
1484                         # 7.5.5
1485                         'h1'         => $block,
1486                         'h2'         => $block,
1487                         'h3'         => $block,
1488                         'h4'         => $block,
1489                         'h5'         => $block,
1490                         'h6'         => $block,
1491
1492                         # 7.5.6
1493                         # address
1494
1495                         # 8.2.4
1496                         # bdo
1497
1498                         # 9.2.1
1499                         'em'         => $common,
1500                         'strong'     => $common,
1501                         'cite'       => $common,
1502                         'dfn'        => $common,
1503                         'code'       => $common,
1504                         'samp'       => $common,
1505                         'kbd'        => $common,
1506                         'var'        => $common,
1507                         'abbr'       => $common,
1508                         # acronym
1509
1510                         # 9.2.2
1511                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1512                         # q
1513
1514                         # 9.2.3
1515                         'sub'        => $common,
1516                         'sup'        => $common,
1517
1518                         # 9.3.1
1519                         'p'          => $block,
1520
1521                         # 9.3.2
1522                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1523
1524                         # 9.3.4
1525                         'pre'        => array_merge( $common, array( 'width' ) ),
1526
1527                         # 9.4
1528                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1529                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1530
1531                         # 10.2
1532                         'ul'         => array_merge( $common, array( 'type' ) ),
1533                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1534                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1535
1536                         # 10.3
1537                         'dl'         => $common,
1538                         'dd'         => $common,
1539                         'dt'         => $common,
1540
1541                         # 11.2.1
1542                         'table'      => array_merge( $common,
1543                                                                 array( 'summary', 'width', 'border', 'frame',
1544                                                                                 'rules', 'cellspacing', 'cellpadding',
1545                                                                                 'align', 'bgcolor',
1546                                                                 ) ),
1547
1548                         # 11.2.2
1549                         'caption'    => array_merge( $common, array( 'align' ) ),
1550
1551                         # 11.2.3
1552                         'thead'      => array_merge( $common, $tablealign ),
1553                         'tfoot'      => array_merge( $common, $tablealign ),
1554                         'tbody'      => array_merge( $common, $tablealign ),
1555
1556                         # 11.2.4
1557                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1558                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1559
1560                         # 11.2.5
1561                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1562
1563                         # 11.2.6
1564                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1565                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1566
1567                         # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1568                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1569
1570                         # 13.2
1571                         # Not usually allowed, but may be used for extension-style hooks
1572                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1573                         # true
1574                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1575
1576                         # 15.2.1
1577                         'tt'         => $common,
1578                         'b'          => $common,
1579                         'i'          => $common,
1580                         'big'        => $common,
1581                         'small'      => $common,
1582                         'strike'     => $common,
1583                         's'          => $common,
1584                         'u'          => $common,
1585
1586                         # 15.2.2
1587                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1588                         # basefont
1589
1590                         # 15.3
1591                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1592
1593                         # XHTML Ruby annotation text module, simple ruby only.
1594                         # http://www.w3c.org/TR/ruby/
1595                         'ruby'       => $common,
1596                         # rbc
1597                         # rtc
1598                         'rb'         => $common,
1599                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1600                         'rp'         => $common,
1601
1602                         # MathML root element, where used for extensions
1603                         # 'title' may not be 100% valid here; it's XHTML
1604                         # http://www.w3.org/TR/REC-MathML/
1605                         'math'       => array( 'class', 'style', 'id', 'title' ),
1606
1607                         # HTML 5 section 4.6
1608                         'bdi' => $common,
1609
1610                 );
1611
1612                 if ( $wgHtml5 ) {
1613                         # HTML5 elements, defined by:
1614                         # http://www.whatwg.org/specs/web-apps/current-work/multipage/
1615                         $whitelist += array(
1616                                 'data' => array_merge( $common, array( 'value' ) ),
1617                                 'time' => array_merge( $common, array( 'datetime' ) ),
1618                                 'mark' => $common,
1619
1620                                 // meta and link are only permitted by removeHTMLtags when Microdata
1621                                 // is enabled so we don't bother adding a conditional to hide these
1622                                 // Also meta and link are only valid in WikiText as Microdata elements
1623                                 // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1624                                 // So we don't bother including $common attributes that have no purpose.
1625                                 'meta' => array( 'itemprop', 'content' ),
1626                                 'link' => array( 'itemprop', 'href' ),
1627                         );
1628                 }
1629
1630                 $staticInitialised = $globalContext;
1631
1632                 return $whitelist;
1633         }
1634
1635         /**
1636          * Take a fragment of (potentially invalid) HTML and return
1637          * a version with any tags removed, encoded as plain text.
1638          *
1639          * Warning: this return value must be further escaped for literal
1640          * inclusion in HTML output as of 1.10!
1641          *
1642          * @param string $text HTML fragment
1643          * @return String
1644          */
1645         static function stripAllTags( $text ) {
1646                 # Actual <tags>
1647                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1648
1649                 # Normalize &entities and whitespace
1650                 $text = self::decodeCharReferences( $text );
1651                 $text = self::normalizeWhitespace( $text );
1652
1653                 return $text;
1654         }
1655
1656         /**
1657          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1658          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1659          * PHP 5.1 doesn't.
1660          *
1661          * Use for passing XHTML fragments to PHP's XML parsing functions
1662          *
1663          * @return String
1664          */
1665         static function hackDocType() {
1666                 $out = "<!DOCTYPE html [\n";
1667                 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1668                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1669                 }
1670                 $out .= "]>\n";
1671                 return $out;
1672         }
1673
1674         /**
1675          * @param $url string
1676          * @return mixed|string
1677          */
1678         static function cleanUrl( $url ) {
1679                 # Normalize any HTML entities in input. They will be
1680                 # re-escaped by makeExternalLink().
1681                 $url = Sanitizer::decodeCharReferences( $url );
1682
1683                 # Escape any control characters introduced by the above step
1684                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1685                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1686
1687                 # Validate hostname portion
1688                 $matches = array();
1689                 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1690                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1691
1692                         // Characters that will be ignored in IDNs.
1693                         // http://tools.ietf.org/html/3454#section-3.1
1694                         // Strip them before further processing so blacklists and such work.
1695                         $strip = "/
1696                                 \\s|          # general whitespace
1697                                 \xc2\xad|     # 00ad SOFT HYPHEN
1698                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1699                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1700                                 \xe2\x81\xa0| # 2060 WORD JOINER
1701                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1702                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1703                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1704                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1705                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1706                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1707                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1708                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1709                                 /xuD";
1710
1711                         $host = preg_replace( $strip, '', $host );
1712
1713                         // @todo FIXME: Validate hostnames here
1714
1715                         return $protocol . $host . $rest;
1716                 } else {
1717                         return $url;
1718                 }
1719         }
1720
1721         /**
1722          * @param $matches array
1723          * @return string
1724          */
1725         static function cleanUrlCallback( $matches ) {
1726                 return urlencode( $matches[0] );
1727         }
1728
1729         /**
1730          * Does a string look like an e-mail address?
1731          *
1732          * This validates an email address using an HTML5 specification found at:
1733          * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1734          * Which as of 2011-01-24 says:
1735          *
1736          *   A valid e-mail address is a string that matches the ABNF production
1737          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1738          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1739          *   3.5.
1740          *
1741          * This function is an implementation of the specification as requested in
1742          * bug 22449.
1743          *
1744          * Client-side forms will use the same standard validation rules via JS or
1745          * HTML 5 validation; additional restrictions can be enforced server-side
1746          * by extensions via the 'isValidEmailAddr' hook.
1747          *
1748          * Note that this validation doesn't 100% match RFC 2822, but is believed
1749          * to be liberal enough for wide use. Some invalid addresses will still
1750          * pass validation here.
1751          *
1752          * @since 1.18
1753          *
1754          * @param string $addr E-mail address
1755          * @return Bool
1756          */
1757         public static function validateEmail( $addr ) {
1758                 $result = null;
1759                 if ( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1760                         return $result;
1761                 }
1762
1763                 // Please note strings below are enclosed in brackets [], this make the
1764                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1765                 // See bug 26948
1766                 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1767                 $rfc1034_ldh_str = "a-z0-9\\-";
1768
1769                 $HTML5_email_regexp = "/
1770                 ^                      # start of string
1771                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1772                 @                      # 'apostrophe'
1773                 [$rfc1034_ldh_str]+       # First domain part
1774                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1775                 $                      # End of string
1776                 /ix"; // case Insensitive, eXtended
1777
1778                 return (bool) preg_match( $HTML5_email_regexp, $addr );
1779         }
1780 }