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