includes/Sanitizer.php

   1 <?php
   2 /**
   3  * XHTML sanitizer for %MediaWiki.
   4  *
   5  * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
   6  * http://www.mediawiki.org/
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License as published by
  10  * the Free Software Foundation; either version 2 of the License, or
  11  * (at your option) any later version.
  12  *
  13  * This program is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  16  * GNU General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU General Public License along
  19  * with this program; if not, write to the Free Software Foundation, Inc.,
  20  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  21  * http://www.gnu.org/copyleft/gpl.html
  22  *
  23  * @file
  24  * @ingroup Parser
  25  */
  26
  27 /**
  28  * XHTML sanitizer for MediaWiki
  29  * @ingroup Parser
  30  */
  31 class Sanitizer {
  32         /**
  33          * Regular expression to match various types of character references in
  34          * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
  35          */
  36         const CHAR_REFS_REGEX =
  37                 '/&([A-Za-z0-9\x80-\xff]+);
  38                  |&\#([0-9]+);
  39                  |&\#[xX]([0-9A-Fa-f]+);
  40                  |(&)/x';
  41
  42         /**
  43          * Blacklist for evil uris like javascript:
  44          * WARNING: DO NOT use this in any place that actually requires blacklisting
  45          * for security reasons. There are NUMEROUS[1] ways to bypass blacklisting, the
  46          * only way to be secure from javascript: uri based xss vectors is to whitelist
  47          * things that you know are safe and deny everything else.
  48          * [1]: http://ha.ckers.org/xss.html
  49          */
  50         const EVIL_URI_PATTERN = '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
  51         const XMLNS_ATTRIBUTE_PATTERN = "/^xmlns:[:A-Z_a-z-.0-9]+$/";
  52
  53         /**
  54          * List of all named character entities defined in HTML 4.01
  55          * http://www.w3.org/TR/html4/sgml/entities.html
  56          * As well as &apos; which is only defined starting in XHTML1.
  57          * @private
  58          */
  59         static $htmlEntities = array(
  60                 'Aacute'   => 193,
  61                 'aacute'   => 225,
  62                 'Acirc'    => 194,
  63                 'acirc'    => 226,
  64                 'acute'    => 180,
  65                 'AElig'    => 198,
  66                 'aelig'    => 230,
  67                 'Agrave'   => 192,
  68                 'agrave'   => 224,
  69                 'alefsym'  => 8501,
  70                 'Alpha'    => 913,
  71                 'alpha'    => 945,
  72                 'amp'      => 38,
  73                 'and'      => 8743,
  74                 'ang'      => 8736,
  75                 'apos'     => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
  76                 'Aring'    => 197,
  77                 'aring'    => 229,
  78                 'asymp'    => 8776,
  79                 'Atilde'   => 195,
  80                 'atilde'   => 227,
  81                 'Auml'     => 196,
  82                 'auml'     => 228,
  83                 'bdquo'    => 8222,
  84                 'Beta'     => 914,
  85                 'beta'     => 946,
  86                 'brvbar'   => 166,
  87                 'bull'     => 8226,
  88                 'cap'      => 8745,
  89                 'Ccedil'   => 199,
  90                 'ccedil'   => 231,
  91                 'cedil'    => 184,
  92                 'cent'     => 162,
  93                 'Chi'      => 935,
  94                 'chi'      => 967,
  95                 'circ'     => 710,
  96                 'clubs'    => 9827,
  97                 'cong'     => 8773,
  98                 'copy'     => 169,
  99                 'crarr'    => 8629,
 100                 'cup'      => 8746,
 101                 'curren'   => 164,
 102                 'dagger'   => 8224,
 103                 'Dagger'   => 8225,
 104                 'darr'     => 8595,
 105                 'dArr'     => 8659,
 106                 'deg'      => 176,
 107                 'Delta'    => 916,
 108                 'delta'    => 948,
 109                 'diams'    => 9830,
 110                 'divide'   => 247,
 111                 'Eacute'   => 201,
 112                 'eacute'   => 233,
 113                 'Ecirc'    => 202,
 114                 'ecirc'    => 234,
 115                 'Egrave'   => 200,
 116                 'egrave'   => 232,
 117                 'empty'    => 8709,
 118                 'emsp'     => 8195,
 119                 'ensp'     => 8194,
 120                 'Epsilon'  => 917,
 121                 'epsilon'  => 949,
 122                 'equiv'    => 8801,
 123                 'Eta'      => 919,
 124                 'eta'      => 951,
 125                 'ETH'      => 208,
 126                 'eth'      => 240,
 127                 'Euml'     => 203,
 128                 'euml'     => 235,
 129                 'euro'     => 8364,
 130                 'exist'    => 8707,
 131                 'fnof'     => 402,
 132                 'forall'   => 8704,
 133                 'frac12'   => 189,
 134                 'frac14'   => 188,
 135                 'frac34'   => 190,
 136                 'frasl'    => 8260,
 137                 'Gamma'    => 915,
 138                 'gamma'    => 947,
 139                 'ge'       => 8805,
 140                 'gt'       => 62,
 141                 'harr'     => 8596,
 142                 'hArr'     => 8660,
 143                 'hearts'   => 9829,
 144                 'hellip'   => 8230,
 145                 'Iacute'   => 205,
 146                 'iacute'   => 237,
 147                 'Icirc'    => 206,
 148                 'icirc'    => 238,
 149                 'iexcl'    => 161,
 150                 'Igrave'   => 204,
 151                 'igrave'   => 236,
 152                 'image'    => 8465,
 153                 'infin'    => 8734,
 154                 'int'      => 8747,
 155                 'Iota'     => 921,
 156                 'iota'     => 953,
 157                 'iquest'   => 191,
 158                 'isin'     => 8712,
 159                 'Iuml'     => 207,
 160                 'iuml'     => 239,
 161                 'Kappa'    => 922,
 162                 'kappa'    => 954,
 163                 'Lambda'   => 923,
 164                 'lambda'   => 955,
 165                 'lang'     => 9001,
 166                 'laquo'    => 171,
 167                 'larr'     => 8592,
 168                 'lArr'     => 8656,
 169                 'lceil'    => 8968,
 170                 'ldquo'    => 8220,
 171                 'le'       => 8804,
 172                 'lfloor'   => 8970,
 173                 'lowast'   => 8727,
 174                 'loz'      => 9674,
 175                 'lrm'      => 8206,
 176                 'lsaquo'   => 8249,
 177                 'lsquo'    => 8216,
 178                 'lt'       => 60,
 179                 'macr'     => 175,
 180                 'mdash'    => 8212,
 181                 'micro'    => 181,
 182                 'middot'   => 183,
 183                 'minus'    => 8722,
 184                 'Mu'       => 924,
 185                 'mu'       => 956,
 186                 'nabla'    => 8711,
 187                 'nbsp'     => 160,
 188                 'ndash'    => 8211,
 189                 'ne'       => 8800,
 190                 'ni'       => 8715,
 191                 'not'      => 172,
 192                 'notin'    => 8713,
 193                 'nsub'     => 8836,
 194                 'Ntilde'   => 209,
 195                 'ntilde'   => 241,
 196                 'Nu'       => 925,
 197                 'nu'       => 957,
 198                 'Oacute'   => 211,
 199                 'oacute'   => 243,
 200                 'Ocirc'    => 212,
 201                 'ocirc'    => 244,
 202                 'OElig'    => 338,
 203                 'oelig'    => 339,
 204                 'Ograve'   => 210,
 205                 'ograve'   => 242,
 206                 'oline'    => 8254,
 207                 'Omega'    => 937,
 208                 'omega'    => 969,
 209                 'Omicron'  => 927,
 210                 'omicron'  => 959,
 211                 'oplus'    => 8853,
 212                 'or'       => 8744,
 213                 'ordf'     => 170,
 214                 'ordm'     => 186,
 215                 'Oslash'   => 216,
 216                 'oslash'   => 248,
 217                 'Otilde'   => 213,
 218                 'otilde'   => 245,
 219                 'otimes'   => 8855,
 220                 'Ouml'     => 214,
 221                 'ouml'     => 246,
 222                 'para'     => 182,
 223                 'part'     => 8706,
 224                 'permil'   => 8240,
 225                 'perp'     => 8869,
 226                 'Phi'      => 934,
 227                 'phi'      => 966,
 228                 'Pi'       => 928,
 229                 'pi'       => 960,
 230                 'piv'      => 982,
 231                 'plusmn'   => 177,
 232                 'pound'    => 163,
 233                 'prime'    => 8242,
 234                 'Prime'    => 8243,
 235                 'prod'     => 8719,
 236                 'prop'     => 8733,
 237                 'Psi'      => 936,
 238                 'psi'      => 968,
 239                 'quot'     => 34,
 240                 'radic'    => 8730,
 241                 'rang'     => 9002,
 242                 'raquo'    => 187,
 243                 'rarr'     => 8594,
 244                 'rArr'     => 8658,
 245                 'rceil'    => 8969,
 246                 'rdquo'    => 8221,
 247                 'real'     => 8476,
 248                 'reg'      => 174,
 249                 'rfloor'   => 8971,
 250                 'Rho'      => 929,
 251                 'rho'      => 961,
 252                 'rlm'      => 8207,
 253                 'rsaquo'   => 8250,
 254                 'rsquo'    => 8217,
 255                 'sbquo'    => 8218,
 256                 'Scaron'   => 352,
 257                 'scaron'   => 353,
 258                 'sdot'     => 8901,
 259                 'sect'     => 167,
 260                 'shy'      => 173,
 261                 'Sigma'    => 931,
 262                 'sigma'    => 963,
 263                 'sigmaf'   => 962,
 264                 'sim'      => 8764,
 265                 'spades'   => 9824,
 266                 'sub'      => 8834,
 267                 'sube'     => 8838,
 268                 'sum'      => 8721,
 269                 'sup'      => 8835,
 270                 'sup1'     => 185,
 271                 'sup2'     => 178,
 272                 'sup3'     => 179,
 273                 'supe'     => 8839,
 274                 'szlig'    => 223,
 275                 'Tau'      => 932,
 276                 'tau'      => 964,
 277                 'there4'   => 8756,
 278                 'Theta'    => 920,
 279                 'theta'    => 952,
 280                 'thetasym' => 977,
 281                 'thinsp'   => 8201,
 282                 'THORN'    => 222,
 283                 'thorn'    => 254,
 284                 'tilde'    => 732,
 285                 'times'    => 215,
 286                 'trade'    => 8482,
 287                 'Uacute'   => 218,
 288                 'uacute'   => 250,
 289                 'uarr'     => 8593,
 290                 'uArr'     => 8657,
 291                 'Ucirc'    => 219,
 292                 'ucirc'    => 251,
 293                 'Ugrave'   => 217,
 294                 'ugrave'   => 249,
 295                 'uml'      => 168,
 296                 'upsih'    => 978,
 297                 'Upsilon'  => 933,
 298                 'upsilon'  => 965,
 299                 'Uuml'     => 220,
 300                 'uuml'     => 252,
 301                 'weierp'   => 8472,
 302                 'Xi'       => 926,
 303                 'xi'       => 958,
 304                 'Yacute'   => 221,
 305                 'yacute'   => 253,
 306                 'yen'      => 165,
 307                 'Yuml'     => 376,
 308                 'yuml'     => 255,
 309                 'Zeta'     => 918,
 310                 'zeta'     => 950,
 311                 'zwj'      => 8205,
 312                 'zwnj'     => 8204
 313         );
 314
 315         /**
 316          * Character entity aliases accepted by MediaWiki
 317          */
 318         static $htmlEntityAliases = array(
 319                 'רלמ' => 'rlm',
 320                 'رلم' => 'rlm',
 321         );
 322
 323         /**
 324          * Lazy-initialised attributes regex, see getAttribsRegex()
 325          */
 326         static $attribsRegex;
 327
 328         /**
 329          * Regular expression to match HTML/XML attribute pairs within a tag.
 330          * Allows some... latitude.
 331          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 332          */
 333         static function getAttribsRegex() {
 334                 if ( self::$attribsRegex === null ) {
 335                         $attribFirst = '[:A-Z_a-z0-9]';
 336                         $attrib = '[:A-Z_a-z-.0-9]';
 337                         $space = '[\x09\x0a\x0d\x20]';
 338                         self::$attribsRegex =
 339                                 "/(?:^|$space)({$attribFirst}{$attrib}*)
 340                                   ($space*=$space*
 341                                         (?:
 342                                          # The attribute value: quoted or alone
 343                                           \"([^<\"]*)\"
 344                                          | '([^<']*)'
 345                                          |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
 346                                          |  (\#[0-9a-fA-F]+) # Technically wrong, but lots of
 347                                                                                  # colors are specified like this.
 348                                                                                  # We'll be normalizing it.
 349                                         )
 350                                 )?(?=$space|\$)/sx";
 351                 }
 352                 return self::$attribsRegex;
 353         }
 354
 355         /**
 356          * Cleans up HTML, removes dangerous tags and attributes, and
 357          * removes HTML comments
 358          * @private
 359          * @param $text String
 360          * @param $processCallback Callback to do any variable or parameter replacements in HTML attribute values
 361          * @param $args Array for the processing callback
 362          * @param $extratags Array for any extra tags to include
 363          * @param $removetags Array for any tags (default or extra) to exclude
 364          * @return string
 365          */
 366         static function removeHTMLtags( $text, $processCallback = null, $args = array(), $extratags = array(), $removetags = array() ) {
 367                 global $wgUseTidy, $wgHtml5, $wgAllowMicrodataAttributes, $wgAllowImageTag;
 368
 369                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 370                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 371
 372                 wfProfileIn( __METHOD__ );
 373
 374                 // Base our staticInitialised variable off of the global config state so that if the globals
 375                 // are changed (like in the secrewed up test system) we will re-initialise the settings.
 376                 $globalContext = implode( '-', compact( 'wgHtml5', 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
 377                 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
 378
 379                         $htmlpairsStatic = array( # Tags that must be closed
 380                                 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 381                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 382                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 383                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 384                                 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'abbr', 'dfn',
 385                                 'kbd', 'samp'
 386                         );
 387                         if ( $wgHtml5 ) {
 388                                 $htmlpairsStatic = array_merge( $htmlpairsStatic, array( 'data', 'time', '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 uncloseable elements
 518                                                         $brace = '/>';
 519                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 520                                                         # Hack to not close $htmlsingle tags
 521                                                         $brace = null;
 522                                                         # Still need to push this optionally-closed tag to
 523                                                         # the tag stack so that we can match end tags
 524                                                         # instead of marking them as bad.
 525                                                         array_push( $tagstack, $t );
 526                                                 } elseif ( isset( $tabletags[$t] )
 527                                                 && in_array( $t, $tagstack ) ) {
 528                                                         // New table tag but forgot to close the previous one
 529                                                         $text .= "</$t>";
 530                                                 } else {
 531                                                         if ( $t == 'table' ) {
 532                                                                 array_push( $tablestack, $tagstack );
 533                                                                 $tagstack = array();
 534                                                         }
 535                                                         array_push( $tagstack, $t );
 536                                                 }
 537
 538                                                 # Replace any variables or template parameters with
 539                                                 # plaintext results.
 540                                                 if( is_callable( $processCallback ) ) {
 541                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 542                                                 }
 543
 544                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 545                                                         $badtag = true;
 546                                                 }
 547
 548                                                 # Strip non-approved attributes from the tag
 549                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 550                                         }
 551                                         if ( !$badtag ) {
 552                                                 $rest = str_replace( '>', '&gt;', $rest );
 553                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 554                                                 $text .= "<$slash$t$newparams$close>$rest";
 555                                                 continue;
 556                                         }
 557                                 }
 558                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 559                         }
 560                         # Close off any remaining tags
 561                         while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
 562                                 $text .= "</$t>\n";
 563                                 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
 564                         }
 565                 } else {
 566                         # this might be possible using tidy itself
 567                         foreach ( $bits as $x ) {
 568                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 569                                 $x, $regs );
 570                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 571                                 $badtag = false;
 572                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 573                                         if( is_callable( $processCallback ) ) {
 574                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 575                                         }
 576
 577                                         if ( !Sanitizer::validateTag( $params, $t ) ) {
 578                                                 $badtag = true;
 579                                         }
 580
 581                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 582                                         if ( !$badtag ) {
 583                                                 $rest = str_replace( '>', '&gt;', $rest );
 584                                                 $text .= "<$slash$t$newparams$brace$rest";
 585                                                 continue;
 586                                         }
 587                                 }
 588                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 589                         }
 590                 }
 591                 wfProfileOut( __METHOD__ );
 592                 return $text;
 593         }
 594
 595         /**
 596          * Remove '<!--', '-->', and everything between.
 597          * To avoid leaving blank lines, when a comment is both preceded
 598          * and followed by a newline (ignoring spaces), trim leading and
 599          * trailing spaces and one of the newlines.
 600          *
 601          * @private
 602          * @param $text String
 603          * @return string
 604          */
 605         static function removeHTMLcomments( $text ) {
 606                 wfProfileIn( __METHOD__ );
 607                 while (($start = strpos($text, '<!--')) !== false) {
 608                         $end = strpos($text, '-->', $start + 4);
 609                         if ($end === false) {
 610                                 # Unterminated comment; bail out
 611                                 break;
 612                         }
 613
 614                         $end += 3;
 615
 616                         # Trim space and newline if the comment is both
 617                         # preceded and followed by a newline
 618                         $spaceStart = max($start - 1, 0);
 619                         $spaceLen = $end - $spaceStart;
 620                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 621                                 $spaceStart--;
 622                                 $spaceLen++;
 623                         }
 624                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 625                                 $spaceLen++;
 626                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 627                                 # Remove the comment, leading and trailing
 628                                 # spaces, and leave only one newline.
 629                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 630                         }
 631                         else {
 632                                 # Remove just the comment.
 633                                 $text = substr_replace($text, '', $start, $end - $start);
 634                         }
 635                 }
 636                 wfProfileOut( __METHOD__ );
 637                 return $text;
 638         }
 639
 640         /**
 641          * Takes attribute names and values for a tag and the tag name and
 642          * validates that the tag is allowed to be present.
 643          * This DOES NOT validate the attributes, nor does it validate the
 644          * tags themselves. This method only handles the special circumstances
 645          * where we may want to allow a tag within content but ONLY when it has
 646          * specific attributes set.
 647          *
 648          * @param $params
 649          * @param $element
 650          */
 651         static function validateTag( $params, $element ) {
 652                 $params = Sanitizer::decodeTagAttributes( $params );
 653
 654                 if ( $element == 'meta' || $element == 'link' ) {
 655                         if ( !isset( $params['itemprop'] ) ) {
 656                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 657                                 return false;
 658                         }
 659                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 660                                 // <meta> must have a content="" for the itemprop
 661                                 return false;
 662                         }
 663                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 664                                 // <link> must have an associated href=""
 665                                 return false;
 666                         }
 667                 }
 668
 669                 return true;
 670         }
 671
 672         /**
 673          * Take an array of attribute names and values and normalize or discard
 674          * illegal values for the given element type.
 675          *
 676          * - Discards attributes not on a whitelist for the given element
 677          * - Unsafe style attributes are discarded
 678          * - Invalid id attributes are reencoded
 679          *
 680          * @param $attribs Array
 681          * @param $element String
 682          * @return Array
 683          *
 684          * @todo Check for legal values where the DTD limits things.
 685          * @todo Check for unique id attribute :P
 686          */
 687         static function validateTagAttributes( $attribs, $element ) {
 688                 return Sanitizer::validateAttributes( $attribs,
 689                         Sanitizer::attributeWhitelist( $element ) );
 690         }
 691
 692         /**
 693          * Take an array of attribute names and values and normalize or discard
 694          * illegal values for the given whitelist.
 695          *
 696          * - Discards attributes not the given whitelist
 697          * - Unsafe style attributes are discarded
 698          * - Invalid id attributes are reencoded
 699          *
 700          * @param $attribs Array
 701          * @param $whitelist Array: list of allowed attribute names
 702          * @return Array
 703          *
 704          * @todo Check for legal values where the DTD limits things.
 705          * @todo Check for unique id attribute :P
 706          */
 707         static function validateAttributes( $attribs, $whitelist ) {
 708                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
 709
 710                 $whitelist = array_flip( $whitelist );
 711                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 712
 713                 $out = array();
 714                 foreach( $attribs as $attribute => $value ) {
 715                         #allow XML namespace declaration if RDFa is enabled
 716                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 717                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 718                                         $out[$attribute] = $value;
 719                                 }
 720
 721                                 continue;
 722                         }
 723
 724                         # Allow any attribute beginning with "data-", if in HTML5 mode
 725                         if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
 726                                 continue;
 727                         }
 728
 729                         # Strip javascript "expression" from stylesheets.
 730                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 731                         if( $attribute == 'style' ) {
 732                                 $value = Sanitizer::checkCss( $value );
 733                         }
 734
 735                         if ( $attribute === 'id' ) {
 736                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 737                         }
 738
 739                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 740                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 741                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 742                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 743                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 744                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 745
 746                                 //Paranoia. Allow "simple" values but suppress javascript
 747                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 748                                         continue;
 749                                 }
 750                         }
 751
 752                         # NOTE: even though elements using href/src are not allowed directly, supply
 753                         #       validation code that can be used by tag hook handlers, etc
 754                         if ( $attribute === 'href' || $attribute === 'src' ) {
 755                                 if ( !preg_match( $hrefExp, $value ) ) {
 756                                         continue; //drop any href or src attributes not using an allowed protocol.
 757                                                   //NOTE: this also drops all relative URLs
 758                                 }
 759                         }
 760
 761                         // If this attribute was previously set, override it.
 762                         // Output should only have one attribute of each name.
 763                         $out[$attribute] = $value;
 764                 }
 765
 766                 if ( $wgAllowMicrodataAttributes ) {
 767                         # itemtype, itemid, itemref don't make sense without itemscope
 768                         if ( !array_key_exists( 'itemscope', $out ) ) {
 769                                 unset( $out['itemtype'] );
 770                                 unset( $out['itemid'] );
 771                                 unset( $out['itemref'] );
 772                         }
 773                         # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 774                 }
 775                 return $out;
 776         }
 777
 778         /**
 779          * Merge two sets of HTML attributes.  Conflicting items in the second set
 780          * will override those in the first, except for 'class' attributes which
 781          * will be combined (if they're both strings).
 782          *
 783          * @todo implement merging for other attributes such as style
 784          * @param $a Array
 785          * @param $b Array
 786          * @return array
 787          */
 788         static function mergeAttributes( $a, $b ) {
 789                 $out = array_merge( $a, $b );
 790                 if( isset( $a['class'] ) && isset( $b['class'] )
 791                 && is_string( $a['class'] ) && is_string( $b['class'] )
 792                 && $a['class'] !== $b['class'] ) {
 793                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 794                                 -1, PREG_SPLIT_NO_EMPTY );
 795                         $out['class'] = implode( ' ', array_unique( $classes ) );
 796                 }
 797                 return $out;
 798         }
 799
 800         /**
 801          * Pick apart some CSS and check it for forbidden or unsafe structures.
 802          * Returns a sanitized string. This sanitized string will have
 803          * character references and escape sequences decoded, and comments
 804          * stripped. If the input is just too evil, only a comment complaining
 805          * about evilness will be returned.
 806          *
 807          * Currently URL references, 'expression', 'tps' are forbidden.
 808          *
 809          * NOTE: Despite the fact that character references are decoded, the
 810          * returned string may contain character references given certain
 811          * clever input strings. These character references must
 812          * be escaped before the return value is embedded in HTML.
 813          *
 814          * @param $value String
 815          * @return String
 816          */
 817         static function checkCss( $value ) {
 818                 // Decode character references like &#123;
 819                 $value = Sanitizer::decodeCharReferences( $value );
 820
 821                 // Decode escape sequences and line continuation
 822                 // See the grammar in the CSS 2 spec, appendix D.
 823                 // This has to be done AFTER decoding character references.
 824                 // This means it isn't possible for this function to return
 825                 // unsanitized escape sequences. It is possible to manufacture
 826                 // input that contains character references that decode to
 827                 // escape sequences that decode to character references, but
 828                 // it's OK for the return value to contain character references
 829                 // because the caller is supposed to escape those anyway.
 830                 static $decodeRegex;
 831                 if ( !$decodeRegex ) {
 832                         $space = '[\\x20\\t\\r\\n\\f]';
 833                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 834                         $backslash = '\\\\';
 835                         $decodeRegex = "/ $backslash
 836                                 (?:
 837                                         ($nl) |  # 1. Line continuation
 838                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 839                                         (.) | # 3. backslash cancelling special meaning
 840                                         () | # 4. backslash at end of string
 841                                 )/xu";
 842                 }
 843                 $value = preg_replace_callback( $decodeRegex,
 844                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 845
 846                 // Remove any comments; IE gets token splitting wrong
 847                 // This must be done AFTER decoding character references and
 848                 // escape sequences, because those steps can introduce comments
 849                 // This step cannot introduce character references or escape
 850                 // sequences, because it replaces comments with spaces rather
 851                 // than removing them completely.
 852                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 853
 854                 // Remove anything after a comment-start token, to guard against
 855                 // incorrect client implementations.
 856                 $commentPos = strpos( $value, '/*' );
 857                 if ( $commentPos !== false ) {
 858                         $value = substr( $value, 0, $commentPos );
 859                 }
 860
 861                 // Reject problematic keywords and control characters
 862                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 863                         return '/* invalid control char */';
 864                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( | image\s*\( | image-set\s*\( !ix', $value ) ) {
 865                         return '/* insecure input */';
 866                 }
 867                 return $value;
 868         }
 869
 870         /**
 871          * @param $matches array
 872          * @return String
 873          */
 874         static function cssDecodeCallback( $matches ) {
 875                 if ( $matches[1] !== '' ) {
 876                         // Line continuation
 877                         return '';
 878                 } elseif ( $matches[2] !== '' ) {
 879                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 880                 } elseif ( $matches[3] !== '' ) {
 881                         $char = $matches[3];
 882                 } else {
 883                         $char = '\\';
 884                 }
 885                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 886                         // These characters need to be escaped in strings
 887                         // Clean up the escape sequence to avoid parsing errors by clients
 888                         return '\\' . dechex( ord( $char ) ) . ' ';
 889                 } else {
 890                         // Decode unnecessary escape
 891                         return $char;
 892                 }
 893         }
 894
 895         /**
 896          * Take a tag soup fragment listing an HTML element's attributes
 897          * and normalize it to well-formed XML, discarding unwanted attributes.
 898          * Output is safe for further wikitext processing, with escaping of
 899          * values that could trigger problems.
 900          *
 901          * - Normalizes attribute names to lowercase
 902          * - Discards attributes not on a whitelist for the given element
 903          * - Turns broken or invalid entities into plaintext
 904          * - Double-quotes all attribute values
 905          * - Attributes without values are given the name as attribute
 906          * - Double attributes are discarded
 907          * - Unsafe style attributes are discarded
 908          * - Prepends space if there are attributes.
 909          *
 910          * @param $text String
 911          * @param $element String
 912          * @return String
 913          */
 914         static function fixTagAttributes( $text, $element ) {
 915                 if( trim( $text ) == '' ) {
 916                         return '';
 917                 }
 918
 919                 $decoded = Sanitizer::decodeTagAttributes( $text );
 920                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
 921
 922                 $attribs = array();
 923                 foreach( $stripped as $attribute => $value ) {
 924                         $encAttribute = htmlspecialchars( $attribute );
 925                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 926
 927                         $attribs[] = "$encAttribute=\"$encValue\"";
 928                 }
 929                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 930         }
 931
 932         /**
 933          * Encode an attribute value for HTML output.
 934          * @param $text String
 935          * @return HTML-encoded text fragment
 936          */
 937         static function encodeAttribute( $text ) {
 938                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 939
 940                 // Whitespace is normalized during attribute decoding,
 941                 // so if we've been passed non-spaces we must encode them
 942                 // ahead of time or they won't be preserved.
 943                 $encValue = strtr( $encValue, array(
 944                         "\n" => '&#10;',
 945                         "\r" => '&#13;',
 946                         "\t" => '&#9;',
 947                 ) );
 948
 949                 return $encValue;
 950         }
 951
 952         /**
 953          * Encode an attribute value for HTML tags, with extra armoring
 954          * against further wiki processing.
 955          * @param $text String
 956          * @return HTML-encoded text fragment
 957          */
 958         static function safeEncodeAttribute( $text ) {
 959                 $encValue = Sanitizer::encodeAttribute( $text );
 960
 961                 # Templates and links may be expanded in later parsing,
 962                 # creating invalid or dangerous output. Suppress this.
 963                 $encValue = strtr( $encValue, array(
 964                         '<'    => '&lt;',   // This should never happen,
 965                         '>'    => '&gt;',   // we've received invalid input
 966                         '"'    => '&quot;', // which should have been escaped.
 967                         '{'    => '&#123;',
 968                         '['    => '&#91;',
 969                         "''"   => '&#39;&#39;',
 970                         'ISBN' => '&#73;SBN',
 971                         'RFC'  => '&#82;FC',
 972                         'PMID' => '&#80;MID',
 973                         '|'    => '&#124;',
 974                         '__'   => '&#95;_',
 975                 ) );
 976
 977                 # Stupid hack
 978                 $encValue = preg_replace_callback(
 979                         '/((?i)' . wfUrlProtocols() . ')/',
 980                         array( 'Sanitizer', 'armorLinksCallback' ),
 981                         $encValue );
 982                 return $encValue;
 983         }
 984
 985         /**
 986          * Given a value, escape it so that it can be used in an id attribute and
 987          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
 988          * true, allowing anything but ASCII whitespace.  Otherwise it will use
 989          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
 990          * escaped with lots of dots.
 991          *
 992          * To ensure we don't have to bother escaping anything, we also strip ', ",
 993          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
 994          * We also strip # because it upsets IE, and % because it could be
 995          * ambiguous if it's part of something that looks like a percent escape
 996          * (which don't work reliably in fragments cross-browser).
 997          *
 998          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
 999          *                                                          in the id and
1000          *                                                          name attributes
1001          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1002          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1003          *   HTML5 definition of id attribute
1004          *
1005          * @param $id String: id to escape
1006          * @param $options Mixed: string or array of strings (default is array()):
1007          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1008          *       so don't pay attention if the first character isn't valid at the
1009          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1010          *       false.
1011          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1012          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1013          *       anchors and links won't break.
1014          * @return String
1015          */
1016         static function escapeId( $id, $options = array() ) {
1017                 global $wgHtml5, $wgExperimentalHtmlIds;
1018                 $options = (array)$options;
1019
1020                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1021                         $id = Sanitizer::decodeCharReferences( $id );
1022                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1023                         $id = trim( $id, '_' );
1024                         if ( $id === '' ) {
1025                                 # Must have been all whitespace to start with.
1026                                 return '_';
1027                         } else {
1028                                 return $id;
1029                         }
1030                 }
1031
1032                 # HTML4-style escaping
1033                 static $replace = array(
1034                         '%3A' => ':',
1035                         '%' => '.'
1036                 );
1037
1038                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1039                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1040
1041                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1042                 && !in_array( 'noninitial', $options ) )  {
1043                         // Initial character must be a letter!
1044                         $id = "x$id";
1045                 }
1046                 return $id;
1047         }
1048
1049         /**
1050          * Given a value, escape it so that it can be used as a CSS class and
1051          * return it.
1052          *
1053          * @todo For extra validity, input should be validated UTF-8.
1054          *
1055          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1056          *
1057          * @param $class String
1058          * @return String
1059          */
1060         static function escapeClass( $class ) {
1061                 // Convert ugly stuff to underscores and kill underscores in ugly places
1062                 return rtrim(preg_replace(
1063                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
1064                         '_',
1065                         $class ), '_');
1066         }
1067
1068         /**
1069          * Given HTML input, escape with htmlspecialchars but un-escape entites.
1070          * This allows (generally harmless) entities like &#160; to survive.
1071          *
1072          * @param $html String to escape
1073          * @return String: escaped input
1074          */
1075         static function escapeHtmlAllowEntities( $html ) {
1076                 $html = Sanitizer::decodeCharReferences( $html );
1077                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1078                 # hurt.
1079                 $html = htmlspecialchars( $html, ENT_QUOTES );
1080                 return $html;
1081         }
1082
1083         /**
1084          * Regex replace callback for armoring links against further processing.
1085          * @param $matches Array
1086          * @return string
1087          */
1088         private static function armorLinksCallback( $matches ) {
1089                 return str_replace( ':', '&#58;', $matches[1] );
1090         }
1091
1092         /**
1093          * Return an associative array of attribute names and values from
1094          * a partial tag string. Attribute names are forces to lowercase,
1095          * character references are decoded to UTF-8 text.
1096          *
1097          * @param $text String
1098          * @return Array
1099          */
1100         public static function decodeTagAttributes( $text ) {
1101                 if( trim( $text ) == '' ) {
1102                         return array();
1103                 }
1104
1105                 $attribs = array();
1106                 $pairs = array();
1107                 if( !preg_match_all(
1108                         self::getAttribsRegex(),
1109                         $text,
1110                         $pairs,
1111                         PREG_SET_ORDER ) ) {
1112                         return $attribs;
1113                 }
1114
1115                 foreach( $pairs as $set ) {
1116                         $attribute = strtolower( $set[1] );
1117                         $value = Sanitizer::getTagAttributeCallback( $set );
1118
1119                         // Normalize whitespace
1120                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1121                         $value = trim( $value );
1122
1123                         // Decode character references
1124                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1125                 }
1126                 return $attribs;
1127         }
1128
1129         /**
1130          * Pick the appropriate attribute value from a match set from the
1131          * attribs regex matches.
1132          *
1133          * @param $set Array
1134          * @throws MWException
1135          * @return String
1136          */
1137         private static function getTagAttributeCallback( $set ) {
1138                 if( isset( $set[6] ) ) {
1139                         # Illegal #XXXXXX color with no quotes.
1140                         return $set[6];
1141                 } elseif( isset( $set[5] ) ) {
1142                         # No quotes.
1143                         return $set[5];
1144                 } elseif( isset( $set[4] ) ) {
1145                         # Single-quoted
1146                         return $set[4];
1147                 } elseif( isset( $set[3] ) ) {
1148                         # Double-quoted
1149                         return $set[3];
1150                 } elseif( !isset( $set[2] ) ) {
1151                         # In XHTML, attributes must have a value.
1152                         # For 'reduced' form, return explicitly the attribute name here.
1153                         return $set[1];
1154                 } else {
1155                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1156                 }
1157         }
1158
1159         /**
1160          * Normalize whitespace and character references in an XML source-
1161          * encoded text for an attribute value.
1162          *
1163          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1164          * but note that we're not returning the value, but are returning
1165          * XML source fragments that will be slapped into output.
1166          *
1167          * @param $text String
1168          * @return String
1169          */
1170         private static function normalizeAttributeValue( $text ) {
1171                 return str_replace( '"', '&quot;',
1172                         self::normalizeWhitespace(
1173                                 Sanitizer::normalizeCharReferences( $text ) ) );
1174         }
1175
1176         /**
1177          * @param $text string
1178          * @return mixed
1179          */
1180         private static function normalizeWhitespace( $text ) {
1181                 return preg_replace(
1182                         '/\r\n|[\x20\x0d\x0a\x09]/',
1183                         ' ',
1184                         $text );
1185         }
1186
1187         /**
1188          * Normalizes whitespace in a section name, such as might be returned
1189          * by Parser::stripSectionName(), for use in the id's that are used for
1190          * section links.
1191          *
1192          * @param $section String
1193          * @return String
1194          */
1195         static function normalizeSectionNameWhitespace( $section ) {
1196                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1197         }
1198
1199         /**
1200          * Ensure that any entities and character references are legal
1201          * for XML and XHTML specifically. Any stray bits will be
1202          * &amp;-escaped to result in a valid text fragment.
1203          *
1204          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1205          *   numericized (this way we're well-formed even without a DTD)
1206          * b. any numeric char refs must be legal chars, not invalid or forbidden
1207          * c. use lower cased "&#x", not "&#X"
1208          * d. fix or reject non-valid attributes
1209          *
1210          * @param $text String
1211          * @return String
1212          * @private
1213          */
1214         static function normalizeCharReferences( $text ) {
1215                 return preg_replace_callback(
1216                         self::CHAR_REFS_REGEX,
1217                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1218                         $text );
1219         }
1220         /**
1221          * @param $matches String
1222          * @return String
1223          */
1224         static function normalizeCharReferencesCallback( $matches ) {
1225                 $ret = null;
1226                 if( $matches[1] != '' ) {
1227                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1228                 } elseif( $matches[2] != '' ) {
1229                         $ret = Sanitizer::decCharReference( $matches[2] );
1230                 } elseif( $matches[3] != ''  ) {
1231                         $ret = Sanitizer::hexCharReference( $matches[3] );
1232                 }
1233                 if( is_null( $ret ) ) {
1234                         return htmlspecialchars( $matches[0] );
1235                 } else {
1236                         return $ret;
1237                 }
1238         }
1239
1240         /**
1241          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1242          * return the equivalent numeric entity reference (except for the core &lt;
1243          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1244          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1245          * pseudo-entity source (eg &amp;foo;)
1246          *
1247          * @param $name String
1248          * @return String
1249          */
1250         static function normalizeEntity( $name ) {
1251                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1252                         return '&' . self::$htmlEntityAliases[$name] . ';';
1253                 } elseif ( in_array( $name,
1254                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1255                         return "&$name;";
1256                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1257                         return '&#' . self::$htmlEntities[$name] . ';';
1258                 } else {
1259                         return "&amp;$name;";
1260                 }
1261         }
1262
1263         /**
1264          * @param $codepoint
1265          * @return null|string
1266          */
1267         static function decCharReference( $codepoint ) {
1268                 $point = intval( $codepoint );
1269                 if( Sanitizer::validateCodepoint( $point ) ) {
1270                         return sprintf( '&#%d;', $point );
1271                 } else {
1272                         return null;
1273                 }
1274         }
1275
1276         /**
1277          * @param $codepoint
1278          * @return null|string
1279          */
1280         static function hexCharReference( $codepoint ) {
1281                 $point = hexdec( $codepoint );
1282                 if( Sanitizer::validateCodepoint( $point ) ) {
1283                         return sprintf( '&#x%x;', $point );
1284                 } else {
1285                         return null;
1286                 }
1287         }
1288
1289         /**
1290          * Returns true if a given Unicode codepoint is a valid character in XML.
1291          * @param $codepoint Integer
1292          * @return Boolean
1293          */
1294         private static function validateCodepoint( $codepoint ) {
1295                 return ($codepoint ==    0x09)
1296                         || ($codepoint ==    0x0a)
1297                         || ($codepoint ==    0x0d)
1298                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1299                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1300                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1301         }
1302
1303         /**
1304          * Decode any character references, numeric or named entities,
1305          * in the text and return a UTF-8 string.
1306          *
1307          * @param $text String
1308          * @return String
1309          */
1310         public static function decodeCharReferences( $text ) {
1311                 return preg_replace_callback(
1312                         self::CHAR_REFS_REGEX,
1313                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1314                         $text );
1315         }
1316
1317         /**
1318          * Decode any character references, numeric or named entities,
1319          * in the next and normalize the resulting string. (bug 14952)
1320          *
1321          * This is useful for page titles, not for text to be displayed,
1322          * MediaWiki allows HTML entities to escape normalization as a feature.
1323          *
1324          * @param $text String (already normalized, containing entities)
1325          * @return String (still normalized, without entities)
1326          */
1327         public static function decodeCharReferencesAndNormalize( $text ) {
1328                 global $wgContLang;
1329                 $text = preg_replace_callback(
1330                         self::CHAR_REFS_REGEX,
1331                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1332                         $text, /* limit */ -1, $count );
1333
1334                 if ( $count ) {
1335                         return $wgContLang->normalize( $text );
1336                 } else {
1337                         return $text;
1338                 }
1339         }
1340
1341         /**
1342          * @param $matches String
1343          * @return String
1344          */
1345         static function decodeCharReferencesCallback( $matches ) {
1346                 if( $matches[1] != '' ) {
1347                         return Sanitizer::decodeEntity( $matches[1] );
1348                 } elseif( $matches[2] != '' ) {
1349                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1350                 } elseif( $matches[3] != ''  ) {
1351                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1352                 }
1353                 # Last case should be an ampersand by itself
1354                 return $matches[0];
1355         }
1356
1357         /**
1358          * Return UTF-8 string for a codepoint if that is a valid
1359          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1360          * @param $codepoint Integer
1361          * @return String
1362          * @private
1363          */
1364         static function decodeChar( $codepoint ) {
1365                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1366                         return codepointToUtf8( $codepoint );
1367                 } else {
1368                         return UTF8_REPLACEMENT;
1369                 }
1370         }
1371
1372         /**
1373          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1374          * return the UTF-8 encoding of that character. Otherwise, returns
1375          * pseudo-entity source (eg "&foo;")
1376          *
1377          * @param $name String
1378          * @return String
1379          */
1380         static function decodeEntity( $name ) {
1381                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1382                         $name = self::$htmlEntityAliases[$name];
1383                 }
1384                 if( isset( self::$htmlEntities[$name] ) ) {
1385                         return codepointToUtf8( self::$htmlEntities[$name] );
1386                 } else {
1387                         return "&$name;";
1388                 }
1389         }
1390
1391         /**
1392          * Fetch the whitelist of acceptable attributes for a given element name.
1393          *
1394          * @param $element String
1395          * @return Array
1396          */
1397         static function attributeWhitelist( $element ) {
1398                 $list = Sanitizer::setupAttributeWhitelist();
1399                 return isset( $list[$element] )
1400                         ? $list[$element]
1401                         : array();
1402         }
1403
1404         /**
1405          * Foreach array key (an allowed HTML element), return an array
1406          * of allowed attributes
1407          * @return Array
1408          */
1409         static function setupAttributeWhitelist() {
1410                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1411
1412                 static $whitelist, $staticInitialised;
1413                 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgHtml5', 'wgAllowMicrodataAttributes' ) );
1414
1415                 if ( isset( $whitelist ) && $staticInitialised == $globalContext ) {
1416                         return $whitelist;
1417                 }
1418
1419                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1420
1421                 if ( $wgAllowRdfaAttributes ) {
1422                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1423                         $common = array_merge( $common, array(
1424                             'about', 'property', 'resource', 'datatype', 'typeof',
1425                         ) );
1426                 }
1427
1428                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1429                         # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1430                         $common = array_merge( $common, array(
1431                             'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1432                         ) );
1433                 }
1434
1435                 $block = array_merge( $common, array( 'align' ) );
1436                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1437                 $tablecell = array( 'abbr',
1438                                     'axis',
1439                                     'headers',
1440                                     'scope',
1441                                     'rowspan',
1442                                     'colspan',
1443                                     'nowrap', # deprecated
1444                                     'width',  # deprecated
1445                                     'height', # deprecated
1446                                     'bgcolor' # deprecated
1447                                     );
1448
1449                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1450                 # See: http://www.w3.org/TR/html4/
1451                 $whitelist = array(
1452                         # 7.5.4
1453                         'div'        => $block,
1454                         'center'     => $common, # deprecated
1455                         'span'       => $block, # ??
1456
1457                         # 7.5.5
1458                         'h1'         => $block,
1459                         'h2'         => $block,
1460                         'h3'         => $block,
1461                         'h4'         => $block,
1462                         'h5'         => $block,
1463                         'h6'         => $block,
1464
1465                         # 7.5.6
1466                         # address
1467
1468                         # 8.2.4
1469                         # bdo
1470
1471                         # 9.2.1
1472                         'em'         => $common,
1473                         'strong'     => $common,
1474                         'cite'       => $common,
1475                         'dfn'        => $common,
1476                         'code'       => $common,
1477                         'samp'       => $common,
1478                         'kbd'        => $common,
1479                         'var'        => $common,
1480                         'abbr'       => $common,
1481                         # acronym
1482
1483                         # 9.2.2
1484                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1485                         # q
1486
1487                         # 9.2.3
1488                         'sub'        => $common,
1489                         'sup'        => $common,
1490
1491                         # 9.3.1
1492                         'p'          => $block,
1493
1494                         # 9.3.2
1495                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1496
1497                         # 9.3.4
1498                         'pre'        => array_merge( $common, array( 'width' ) ),
1499
1500                         # 9.4
1501                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1502                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1503
1504                         # 10.2
1505                         'ul'         => array_merge( $common, array( 'type' ) ),
1506                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1507                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1508
1509                         # 10.3
1510                         'dl'         => $common,
1511                         'dd'         => $common,
1512                         'dt'         => $common,
1513
1514                         # 11.2.1
1515                         'table'      => array_merge( $common,
1516                                                                 array( 'summary', 'width', 'border', 'frame',
1517                                                                                 'rules', 'cellspacing', 'cellpadding',
1518                                                                                 'align', 'bgcolor',
1519                                                                 ) ),
1520
1521                         # 11.2.2
1522                         'caption'    => array_merge( $common, array( 'align' ) ),
1523
1524                         # 11.2.3
1525                         'thead'      => array_merge( $common, $tablealign ),
1526                         'tfoot'      => array_merge( $common, $tablealign ),
1527                         'tbody'      => array_merge( $common, $tablealign ),
1528
1529                         # 11.2.4
1530                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1531                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1532
1533                         # 11.2.5
1534                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1535
1536                         # 11.2.6
1537                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1538                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1539
1540                         # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1541                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1542
1543                         # 13.2
1544                         # Not usually allowed, but may be used for extension-style hooks
1545                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1546                         # true
1547                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1548
1549                         # 15.2.1
1550                         'tt'         => $common,
1551                         'b'          => $common,
1552                         'i'          => $common,
1553                         'big'        => $common,
1554                         'small'      => $common,
1555                         'strike'     => $common,
1556                         's'          => $common,
1557                         'u'          => $common,
1558
1559                         # 15.2.2
1560                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1561                         # basefont
1562
1563                         # 15.3
1564                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1565
1566                         # XHTML Ruby annotation text module, simple ruby only.
1567                         # http://www.w3c.org/TR/ruby/
1568                         'ruby'       => $common,
1569                         # rbc
1570                         # rtc
1571                         'rb'         => $common,
1572                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1573                         'rp'         => $common,
1574
1575                         # MathML root element, where used for extensions
1576                         # 'title' may not be 100% valid here; it's XHTML
1577                         # http://www.w3.org/TR/REC-MathML/
1578                         'math'       => array( 'class', 'style', 'id', 'title' ),
1579
1580                         # HTML 5 section 4.6
1581                         'bdi' => $common,
1582
1583                 );
1584
1585                 if ( $wgHtml5 ) {
1586                         # HTML5 elements, defined by:
1587                         # http://www.whatwg.org/specs/web-apps/current-work/multipage/
1588                         $whitelist += array(
1589                                 'data' => array_merge( $common, array( 'value' ) ),
1590                                 'time' => array_merge( $common, array( 'datetime' ) ),
1591                                 'mark' => $common,
1592
1593                                 // meta and link are only permitted by removeHTMLtags when Microdata
1594                                 // is enabled so we don't bother adding a conditional to hide these
1595                                 // Also meta and link are only valid in WikiText as Microdata elements
1596                                 // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1597                                 // So we don't bother including $common attributes that have no purpose.
1598                                 'meta' => array( 'itemprop', 'content' ),
1599                                 'link' => array( 'itemprop', 'href' ),
1600                         );
1601                 }
1602
1603                 $staticInitialised = $globalContext;
1604
1605                 return $whitelist;
1606         }
1607
1608         /**
1609          * Take a fragment of (potentially invalid) HTML and return
1610          * a version with any tags removed, encoded as plain text.
1611          *
1612          * Warning: this return value must be further escaped for literal
1613          * inclusion in HTML output as of 1.10!
1614          *
1615          * @param $text String: HTML fragment
1616          * @return String
1617          */
1618         static function stripAllTags( $text ) {
1619                 # Actual <tags>
1620                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1621
1622                 # Normalize &entities and whitespace
1623                 $text = self::decodeCharReferences( $text );
1624                 $text = self::normalizeWhitespace( $text );
1625
1626                 return $text;
1627         }
1628
1629         /**
1630          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1631          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1632          * PHP 5.1 doesn't.
1633          *
1634          * Use for passing XHTML fragments to PHP's XML parsing functions
1635          *
1636          * @return String
1637          */
1638         static function hackDocType() {
1639                 $out = "<!DOCTYPE html [\n";
1640                 foreach( self::$htmlEntities as $entity => $codepoint ) {
1641                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1642                 }
1643                 $out .= "]>\n";
1644                 return $out;
1645         }
1646
1647         /**
1648          * @param $url string
1649          * @return mixed|string
1650          */
1651         static function cleanUrl( $url ) {
1652                 # Normalize any HTML entities in input. They will be
1653                 # re-escaped by makeExternalLink().
1654                 $url = Sanitizer::decodeCharReferences( $url );
1655
1656                 # Escape any control characters introduced by the above step
1657                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1658                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1659
1660                 # Validate hostname portion
1661                 $matches = array();
1662                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1663                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1664
1665                         // Characters that will be ignored in IDNs.
1666                         // http://tools.ietf.org/html/3454#section-3.1
1667                         // Strip them before further processing so blacklists and such work.
1668                         $strip = "/
1669                                 \\s|          # general whitespace
1670                                 \xc2\xad|     # 00ad SOFT HYPHEN
1671                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1672                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1673                                 \xe2\x81\xa0| # 2060 WORD JOINER
1674                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1675                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1676                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1677                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1678                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1679                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1680                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1681                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1682                                 /xuD";
1683
1684                         $host = preg_replace( $strip, '', $host );
1685
1686                         // @todo FIXME: Validate hostnames here
1687
1688                         return $protocol . $host . $rest;
1689                 } else {
1690                         return $url;
1691                 }
1692         }
1693
1694         /**
1695          * @param $matches array
1696          * @return string
1697          */
1698         static function cleanUrlCallback( $matches ) {
1699                 return urlencode( $matches[0] );
1700         }
1701
1702         /**
1703          * Does a string look like an e-mail address?
1704          *
1705          * This validates an email address using an HTML5 specification found at:
1706          * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1707          * Which as of 2011-01-24 says:
1708          *
1709          *   A valid e-mail address is a string that matches the ABNF production
1710          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1711          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1712          *   3.5.
1713          *
1714          * This function is an implementation of the specification as requested in
1715          * bug 22449.
1716          *
1717          * Client-side forms will use the same standard validation rules via JS or
1718          * HTML 5 validation; additional restrictions can be enforced server-side
1719          * by extensions via the 'isValidEmailAddr' hook.
1720          *
1721          * Note that this validation doesn't 100% match RFC 2822, but is believed
1722          * to be liberal enough for wide use. Some invalid addresses will still
1723          * pass validation here.
1724          *
1725          * @since 1.18
1726          *
1727          * @param $addr String E-mail address
1728          * @return Bool
1729          */
1730         public static function validateEmail( $addr ) {
1731                 $result = null;
1732                 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1733                         return $result;
1734                 }
1735
1736                 // Please note strings below are enclosed in brackets [], this make the
1737                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1738                 // See bug 26948
1739                 $rfc5322_atext   = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1740                 $rfc1034_ldh_str = "a-z0-9\\-" ;
1741
1742                 $HTML5_email_regexp = "/
1743                 ^                      # start of string
1744                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1745                 @                      # 'apostrophe'
1746                 [$rfc1034_ldh_str]+       # First domain part
1747                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1748                 $                      # End of string
1749                 /ix" ; // case Insensitive, eXtended
1750
1751                 return (bool) preg_match( $HTML5_email_regexp, $addr );
1752         }
1753 }