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                                 'var', 'kbd', 'samp'
 404                         );
 405                         $tabletags = array( # Can only appear inside table, we will close them
 406                                 'td', 'th', 'tr',
 407                         );
 408                         $htmllist = array( # Tags used by list
 409                                 'ul', 'ol',
 410                         );
 411                         $listtags = array( # Tags that can appear in a list
 412                                 'li',
 413                         );
 414
 415                         if ( $wgAllowImageTag ) {
 416                                 $htmlsingle[] = 'img';
 417                                 $htmlsingleonly[] = 'img';
 418                         }
 419
 420                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 421                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 422
 423                         # Convert them all to hashtables for faster lookup
 424                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 425                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 426                         foreach ( $vars as $var ) {
 427                                 $$var = array_flip( $$var );
 428                         }
 429                         $staticInitialised = $globalContext;
 430                 }
 431                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 432                 $extratags = array_flip( $extratags );
 433                 $removetags = array_flip( $removetags );
 434                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 435                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
 436
 437                 # Remove HTML comments
 438                 $text = Sanitizer::removeHTMLcomments( $text );
 439                 $bits = explode( '<', $text );
 440                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 441                 if ( !$wgUseTidy ) {
 442                         $tagstack = $tablestack = array();
 443                         foreach ( $bits as $x ) {
 444                                 $regs = array();
 445                                 # $slash: Does the current element start with a '/'?
 446                                 # $t: Current element name
 447                                 # $params: String between element name and >
 448                                 # $brace: Ending '>' or '/>'
 449                                 # $rest: Everything until the next element of $bits
 450                                 if ( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 451                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 452                                 } else {
 453                                         $slash = $t = $params = $brace = $rest = null;
 454                                 }
 455
 456                                 $badtag = false;
 457                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 458                                         # Check our stack
 459                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 460                                                 $badtag = true;
 461                                         } elseif ( $slash ) {
 462                                                 # Closing a tag... is it the one we just opened?
 463                                                 $ot = @array_pop( $tagstack );
 464                                                 if ( $ot != $t ) {
 465                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 466                                                                 # Pop all elements with an optional close tag
 467                                                                 # and see if we find a match below them
 468                                                                 $optstack = array();
 469                                                                 array_push( $optstack, $ot );
 470                                                                 wfSuppressWarnings();
 471                                                                 $ot = array_pop( $tagstack );
 472                                                                 wfRestoreWarnings();
 473                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 474                                                                         array_push( $optstack, $ot );
 475                                                                         wfSuppressWarnings();
 476                                                                         $ot = array_pop( $tagstack );
 477                                                                         wfRestoreWarnings();
 478                                                                 }
 479                                                                 if ( $t != $ot ) {
 480                                                                         # No match. Push the optional elements back again
 481                                                                         $badtag = true;
 482                                                                         wfSuppressWarnings();
 483                                                                         $ot = array_pop( $optstack );
 484                                                                         wfRestoreWarnings();
 485                                                                         while ( $ot ) {
 486                                                                                 array_push( $tagstack, $ot );
 487                                                                                 wfSuppressWarnings();
 488                                                                                 $ot = array_pop( $optstack );
 489                                                                                 wfRestoreWarnings();
 490                                                                         }
 491                                                                 }
 492                                                         } else {
 493                                                                 @array_push( $tagstack, $ot );
 494                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 495                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 496                                                                         $badtag = true;
 497                                                                 }
 498                                                         }
 499                                                 } else {
 500                                                         if ( $t == 'table' ) {
 501                                                                 $tagstack = array_pop( $tablestack );
 502                                                         }
 503                                                 }
 504                                                 $newparams = '';
 505                                         } else {
 506                                                 # Keep track for later
 507                                                 if ( isset( $tabletags[$t] ) &&
 508                                                 !in_array( 'table', $tagstack ) ) {
 509                                                         $badtag = true;
 510                                                 } elseif ( in_array( $t, $tagstack ) &&
 511                                                 !isset( $htmlnest[$t] ) ) {
 512                                                         $badtag = true;
 513                                                 # Is it a self closed htmlpair ? (bug 5487)
 514                                                 } elseif ( $brace == '/>' &&
 515                                                 isset( $htmlpairs[$t] ) ) {
 516                                                         $badtag = true;
 517                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 518                                                         # Hack to force empty tag for unclosable elements
 519                                                         $brace = '/>';
 520                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 521                                                         # Hack to not close $htmlsingle tags
 522                                                         $brace = null;
 523                                                         # Still need to push this optionally-closed tag to
 524                                                         # the tag stack so that we can match end tags
 525                                                         # instead of marking them as bad.
 526                                                         array_push( $tagstack, $t );
 527                                                 } elseif ( isset( $tabletags[$t] )
 528                                                 && in_array( $t, $tagstack ) ) {
 529                                                         // New table tag but forgot to close the previous one
 530                                                         $text .= "</$t>";
 531                                                 } else {
 532                                                         if ( $t == 'table' ) {
 533                                                                 array_push( $tablestack, $tagstack );
 534                                                                 $tagstack = array();
 535                                                         }
 536                                                         array_push( $tagstack, $t );
 537                                                 }
 538
 539                                                 # Replace any variables or template parameters with
 540                                                 # plaintext results.
 541                                                 if ( is_callable( $processCallback ) ) {
 542                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 543                                                 }
 544
 545                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 546                                                         $badtag = true;
 547                                                 }
 548
 549                                                 # Strip non-approved attributes from the tag
 550                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 551                                         }
 552                                         if ( !$badtag ) {
 553                                                 $rest = str_replace( '>', '&gt;', $rest );
 554                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 555                                                 $text .= "<$slash$t$newparams$close>$rest";
 556                                                 continue;
 557                                         }
 558                                 }
 559                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 560                         }
 561                         # Close off any remaining tags
 562                         while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
 563                                 $text .= "</$t>\n";
 564                                 if ( $t == 'table' ) {
 565                                         $tagstack = array_pop( $tablestack );
 566                                 }
 567                         }
 568                 } else {
 569                         # this might be possible using tidy itself
 570                         foreach ( $bits as $x ) {
 571                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 572                                 $x, $regs );
 573                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 574                                 $badtag = false;
 575                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 576                                         if ( is_callable( $processCallback ) ) {
 577                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 578                                         }
 579
 580                                         if ( !Sanitizer::validateTag( $params, $t ) ) {
 581                                                 $badtag = true;
 582                                         }
 583
 584                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 585                                         if ( !$badtag ) {
 586                                                 $rest = str_replace( '>', '&gt;', $rest );
 587                                                 $text .= "<$slash$t$newparams$brace$rest";
 588                                                 continue;
 589                                         }
 590                                 }
 591                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 592                         }
 593                 }
 594                 wfProfileOut( __METHOD__ );
 595                 return $text;
 596         }
 597
 598         /**
 599          * Remove '<!--', '-->', and everything between.
 600          * To avoid leaving blank lines, when a comment is both preceded
 601          * and followed by a newline (ignoring spaces), trim leading and
 602          * trailing spaces and one of the newlines.
 603          *
 604          * @private
 605          * @param $text String
 606          * @return string
 607          */
 608         static function removeHTMLcomments( $text ) {
 609                 wfProfileIn( __METHOD__ );
 610                 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
 611                         $end = strpos( $text, '-->', $start + 4 );
 612                         if ( $end === false ) {
 613                                 # Unterminated comment; bail out
 614                                 break;
 615                         }
 616
 617                         $end += 3;
 618
 619                         # Trim space and newline if the comment is both
 620                         # preceded and followed by a newline
 621                         $spaceStart = max( $start - 1, 0 );
 622                         $spaceLen = $end - $spaceStart;
 623                         while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
 624                                 $spaceStart--;
 625                                 $spaceLen++;
 626                         }
 627                         while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
 628                                 $spaceLen++;
 629                         }
 630                         if ( substr( $text, $spaceStart, 1 ) === "\n" and substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
 631                                 # Remove the comment, leading and trailing
 632                                 # spaces, and leave only one newline.
 633                                 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
 634                         }
 635                         else {
 636                                 # Remove just the comment.
 637                                 $text = substr_replace( $text, '', $start, $end - $start );
 638                         }
 639                 }
 640                 wfProfileOut( __METHOD__ );
 641                 return $text;
 642         }
 643
 644         /**
 645          * Takes attribute names and values for a tag and the tag name and
 646          * validates that the tag is allowed to be present.
 647          * This DOES NOT validate the attributes, nor does it validate the
 648          * tags themselves. This method only handles the special circumstances
 649          * where we may want to allow a tag within content but ONLY when it has
 650          * specific attributes set.
 651          *
 652          * @param $params
 653          * @param $element
 654          * @return bool
 655          */
 656         static function validateTag( $params, $element ) {
 657                 $params = Sanitizer::decodeTagAttributes( $params );
 658
 659                 if ( $element == 'meta' || $element == 'link' ) {
 660                         if ( !isset( $params['itemprop'] ) ) {
 661                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 662                                 return false;
 663                         }
 664                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 665                                 // <meta> must have a content="" for the itemprop
 666                                 return false;
 667                         }
 668                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 669                                 // <link> must have an associated href=""
 670                                 return false;
 671                         }
 672                 }
 673
 674                 return true;
 675         }
 676
 677         /**
 678          * Take an array of attribute names and values and normalize or discard
 679          * illegal values for the given element type.
 680          *
 681          * - Discards attributes not on a whitelist for the given element
 682          * - Unsafe style attributes are discarded
 683          * - Invalid id attributes are re-encoded
 684          *
 685          * @param $attribs Array
 686          * @param $element String
 687          * @return Array
 688          *
 689          * @todo Check for legal values where the DTD limits things.
 690          * @todo Check for unique id attribute :P
 691          */
 692         static function validateTagAttributes( $attribs, $element ) {
 693                 return Sanitizer::validateAttributes( $attribs,
 694                         Sanitizer::attributeWhitelist( $element ) );
 695         }
 696
 697         /**
 698          * Take an array of attribute names and values and normalize or discard
 699          * illegal values for the given whitelist.
 700          *
 701          * - Discards attributes not the given whitelist
 702          * - Unsafe style attributes are discarded
 703          * - Invalid id attributes are re-encoded
 704          *
 705          * @param $attribs Array
 706          * @param array $whitelist list of allowed attribute names
 707          * @return Array
 708          *
 709          * @todo Check for legal values where the DTD limits things.
 710          * @todo Check for unique id attribute :P
 711          */
 712         static function validateAttributes( $attribs, $whitelist ) {
 713                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
 714
 715                 $whitelist = array_flip( $whitelist );
 716                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 717
 718                 $out = array();
 719                 foreach ( $attribs as $attribute => $value ) {
 720                         #allow XML namespace declaration if RDFa is enabled
 721                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 722                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 723                                         $out[$attribute] = $value;
 724                                 }
 725
 726                                 continue;
 727                         }
 728
 729                         # Allow any attribute beginning with "data-", if in HTML5 mode
 730                         if ( !( $wgHtml5 && preg_match( '/^data-/i', $attribute ) ) && !isset( $whitelist[$attribute] ) ) {
 731                                 continue;
 732                         }
 733
 734                         # Strip javascript "expression" from stylesheets.
 735                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 736                         if ( $attribute == 'style' ) {
 737                                 $value = Sanitizer::checkCss( $value );
 738                         }
 739
 740                         if ( $attribute === 'id' ) {
 741                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 742                         }
 743
 744                         # WAI-ARIA
 745                         # http://www.w3.org/TR/wai-aria/
 746                         # http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#wai-aria
 747                         # For now we only support role="presentation" until we work out what roles should be
 748                         # usable by content and we ensure that our code explicitly rejects patterns that
 749                         # violate HTML5's ARIA restrictions.
 750                         if ( $attribute === 'role' && $value !== 'presentation' ) {
 751                                 continue;
 752                         }
 753
 754                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 755                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 756                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 757                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 758                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 759                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 760
 761                                 //Paranoia. Allow "simple" values but suppress javascript
 762                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 763                                         continue;
 764                                 }
 765                         }
 766
 767                         # NOTE: even though elements using href/src are not allowed directly, supply
 768                         #       validation code that can be used by tag hook handlers, etc
 769                         if ( $attribute === 'href' || $attribute === 'src' ) {
 770                                 if ( !preg_match( $hrefExp, $value ) ) {
 771                                         continue; //drop any href or src attributes not using an allowed protocol.
 772                                                   //NOTE: this also drops all relative URLs
 773                                 }
 774                         }
 775
 776                         // If this attribute was previously set, override it.
 777                         // Output should only have one attribute of each name.
 778                         $out[$attribute] = $value;
 779                 }
 780
 781                 if ( $wgAllowMicrodataAttributes ) {
 782                         # itemtype, itemid, itemref don't make sense without itemscope
 783                         if ( !array_key_exists( 'itemscope', $out ) ) {
 784                                 unset( $out['itemtype'] );
 785                                 unset( $out['itemid'] );
 786                                 unset( $out['itemref'] );
 787                         }
 788                         # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 789                 }
 790                 return $out;
 791         }
 792
 793         /**
 794          * Merge two sets of HTML attributes.  Conflicting items in the second set
 795          * will override those in the first, except for 'class' attributes which
 796          * will be combined (if they're both strings).
 797          *
 798          * @todo implement merging for other attributes such as style
 799          * @param $a Array
 800          * @param $b Array
 801          * @return array
 802          */
 803         static function mergeAttributes( $a, $b ) {
 804                 $out = array_merge( $a, $b );
 805                 if ( isset( $a['class'] ) && isset( $b['class'] )
 806                         && is_string( $a['class'] ) && is_string( $b['class'] )
 807                         && $a['class'] !== $b['class']
 808                 ) {
 809                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 810                                 -1, PREG_SPLIT_NO_EMPTY );
 811                         $out['class'] = implode( ' ', array_unique( $classes ) );
 812                 }
 813                 return $out;
 814         }
 815
 816         /**
 817          * Pick apart some CSS and check it for forbidden or unsafe structures.
 818          * Returns a sanitized string. This sanitized string will have
 819          * character references and escape sequences decoded, and comments
 820          * stripped. If the input is just too evil, only a comment complaining
 821          * about evilness will be returned.
 822          *
 823          * Currently URL references, 'expression', 'tps' are forbidden.
 824          *
 825          * NOTE: Despite the fact that character references are decoded, the
 826          * returned string may contain character references given certain
 827          * clever input strings. These character references must
 828          * be escaped before the return value is embedded in HTML.
 829          *
 830          * @param $value String
 831          * @return String
 832          */
 833         static function checkCss( $value ) {
 834                 // Decode character references like &#123;
 835                 $value = Sanitizer::decodeCharReferences( $value );
 836
 837                 // Decode escape sequences and line continuation
 838                 // See the grammar in the CSS 2 spec, appendix D.
 839                 // This has to be done AFTER decoding character references.
 840                 // This means it isn't possible for this function to return
 841                 // unsanitized escape sequences. It is possible to manufacture
 842                 // input that contains character references that decode to
 843                 // escape sequences that decode to character references, but
 844                 // it's OK for the return value to contain character references
 845                 // because the caller is supposed to escape those anyway.
 846                 static $decodeRegex;
 847                 if ( !$decodeRegex ) {
 848                         $space = '[\\x20\\t\\r\\n\\f]';
 849                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 850                         $backslash = '\\\\';
 851                         $decodeRegex = "/ $backslash
 852                                 (?:
 853                                         ($nl) |  # 1. Line continuation
 854                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 855                                         (.) | # 3. backslash cancelling special meaning
 856                                         () | # 4. backslash at end of string
 857                                 )/xu";
 858                 }
 859                 $value = preg_replace_callback( $decodeRegex,
 860                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 861
 862                 // Remove any comments; IE gets token splitting wrong
 863                 // This must be done AFTER decoding character references and
 864                 // escape sequences, because those steps can introduce comments
 865                 // This step cannot introduce character references or escape
 866                 // sequences, because it replaces comments with spaces rather
 867                 // than removing them completely.
 868                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 869
 870                 // Remove anything after a comment-start token, to guard against
 871                 // incorrect client implementations.
 872                 $commentPos = strpos( $value, '/*' );
 873                 if ( $commentPos !== false ) {
 874                         $value = substr( $value, 0, $commentPos );
 875                 }
 876
 877                 // Reject problematic keywords and control characters
 878                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 879                         return '/* invalid control char */';
 880                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( | image\s*\( | image-set\s*\( !ix', $value ) ) {
 881                         return '/* insecure input */';
 882                 }
 883                 return $value;
 884         }
 885
 886         /**
 887          * @param $matches array
 888          * @return String
 889          */
 890         static function cssDecodeCallback( $matches ) {
 891                 if ( $matches[1] !== '' ) {
 892                         // Line continuation
 893                         return '';
 894                 } elseif ( $matches[2] !== '' ) {
 895                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 896                 } elseif ( $matches[3] !== '' ) {
 897                         $char = $matches[3];
 898                 } else {
 899                         $char = '\\';
 900                 }
 901                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 902                         // These characters need to be escaped in strings
 903                         // Clean up the escape sequence to avoid parsing errors by clients
 904                         return '\\' . dechex( ord( $char ) ) . ' ';
 905                 } else {
 906                         // Decode unnecessary escape
 907                         return $char;
 908                 }
 909         }
 910
 911         /**
 912          * Take a tag soup fragment listing an HTML element's attributes
 913          * and normalize it to well-formed XML, discarding unwanted attributes.
 914          * Output is safe for further wikitext processing, with escaping of
 915          * values that could trigger problems.
 916          *
 917          * - Normalizes attribute names to lowercase
 918          * - Discards attributes not on a whitelist for the given element
 919          * - Turns broken or invalid entities into plaintext
 920          * - Double-quotes all attribute values
 921          * - Attributes without values are given the name as attribute
 922          * - Double attributes are discarded
 923          * - Unsafe style attributes are discarded
 924          * - Prepends space if there are attributes.
 925          *
 926          * @param $text String
 927          * @param $element String
 928          * @return String
 929          */
 930         static function fixTagAttributes( $text, $element ) {
 931                 if ( trim( $text ) == '' ) {
 932                         return '';
 933                 }
 934
 935                 $decoded = Sanitizer::decodeTagAttributes( $text );
 936                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
 937
 938                 $attribs = array();
 939                 foreach ( $stripped as $attribute => $value ) {
 940                         $encAttribute = htmlspecialchars( $attribute );
 941                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 942
 943                         $attribs[] = "$encAttribute=\"$encValue\"";
 944                 }
 945                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 946         }
 947
 948         /**
 949          * Encode an attribute value for HTML output.
 950          * @param $text String
 951          * @return HTML-encoded text fragment
 952          */
 953         static function encodeAttribute( $text ) {
 954                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 955
 956                 // Whitespace is normalized during attribute decoding,
 957                 // so if we've been passed non-spaces we must encode them
 958                 // ahead of time or they won't be preserved.
 959                 $encValue = strtr( $encValue, array(
 960                         "\n" => '&#10;',
 961                         "\r" => '&#13;',
 962                         "\t" => '&#9;',
 963                 ) );
 964
 965                 return $encValue;
 966         }
 967
 968         /**
 969          * Encode an attribute value for HTML tags, with extra armoring
 970          * against further wiki processing.
 971          * @param $text String
 972          * @return HTML-encoded text fragment
 973          */
 974         static function safeEncodeAttribute( $text ) {
 975                 $encValue = Sanitizer::encodeAttribute( $text );
 976
 977                 # Templates and links may be expanded in later parsing,
 978                 # creating invalid or dangerous output. Suppress this.
 979                 $encValue = strtr( $encValue, array(
 980                         '<'    => '&lt;',   // This should never happen,
 981                         '>'    => '&gt;',   // we've received invalid input
 982                         '"'    => '&quot;', // which should have been escaped.
 983                         '{'    => '&#123;',
 984                         '['    => '&#91;',
 985                         "''"   => '&#39;&#39;',
 986                         'ISBN' => '&#73;SBN',
 987                         'RFC'  => '&#82;FC',
 988                         'PMID' => '&#80;MID',
 989                         '|'    => '&#124;',
 990                         '__'   => '&#95;_',
 991                 ) );
 992
 993                 # Stupid hack
 994                 $encValue = preg_replace_callback(
 995                         '/((?i)' . wfUrlProtocols() . ')/',
 996                         array( 'Sanitizer', 'armorLinksCallback' ),
 997                         $encValue );
 998                 return $encValue;
 999         }
1000
1001         /**
1002          * Given a value, escape it so that it can be used in an id attribute and
1003          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1004          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1005          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1006          * escaped with lots of dots.
1007          *
1008          * To ensure we don't have to bother escaping anything, we also strip ', ",
1009          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1010          * We also strip # because it upsets IE, and % because it could be
1011          * ambiguous if it's part of something that looks like a percent escape
1012          * (which don't work reliably in fragments cross-browser).
1013          *
1014          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1015          *                                                          in the id and
1016          *                                                          name attributes
1017          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1018          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1019          *   HTML5 definition of id attribute
1020          *
1021          * @param string $id id to escape
1022          * @param $options Mixed: string or array of strings (default is array()):
1023          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1024          *       so don't pay attention if the first character isn't valid at the
1025          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1026          *       false.
1027          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1028          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1029          *       anchors and links won't break.
1030          * @return String
1031          */
1032         static function escapeId( $id, $options = array() ) {
1033                 global $wgHtml5, $wgExperimentalHtmlIds;
1034                 $options = (array)$options;
1035
1036                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1037                         $id = Sanitizer::decodeCharReferences( $id );
1038                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1039                         $id = trim( $id, '_' );
1040                         if ( $id === '' ) {
1041                                 # Must have been all whitespace to start with.
1042                                 return '_';
1043                         } else {
1044                                 return $id;
1045                         }
1046                 }
1047
1048                 # HTML4-style escaping
1049                 static $replace = array(
1050                         '%3A' => ':',
1051                         '%' => '.'
1052                 );
1053
1054                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1055                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1056
1057                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1058                 && !in_array( 'noninitial', $options ) ) {
1059                         // Initial character must be a letter!
1060                         $id = "x$id";
1061                 }
1062                 return $id;
1063         }
1064
1065         /**
1066          * Given a value, escape it so that it can be used as a CSS class and
1067          * return it.
1068          *
1069          * @todo For extra validity, input should be validated UTF-8.
1070          *
1071          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1072          *
1073          * @param $class String
1074          * @return String
1075          */
1076         static function escapeClass( $class ) {
1077                 // Convert ugly stuff to underscores and kill underscores in ugly places
1078                 return rtrim( preg_replace(
1079                         array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1080                         '_',
1081                         $class ), '_' );
1082         }
1083
1084         /**
1085          * Given HTML input, escape with htmlspecialchars but un-escape entities.
1086          * This allows (generally harmless) entities like &#160; to survive.
1087          *
1088          * @param string $html to escape
1089          * @return String: escaped input
1090          */
1091         static function escapeHtmlAllowEntities( $html ) {
1092                 $html = Sanitizer::decodeCharReferences( $html );
1093                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1094                 # hurt.
1095                 $html = htmlspecialchars( $html, ENT_QUOTES );
1096                 return $html;
1097         }
1098
1099         /**
1100          * Regex replace callback for armoring links against further processing.
1101          * @param $matches Array
1102          * @return string
1103          */
1104         private static function armorLinksCallback( $matches ) {
1105                 return str_replace( ':', '&#58;', $matches[1] );
1106         }
1107
1108         /**
1109          * Return an associative array of attribute names and values from
1110          * a partial tag string. Attribute names are forces to lowercase,
1111          * character references are decoded to UTF-8 text.
1112          *
1113          * @param $text String
1114          * @return Array
1115          */
1116         public static function decodeTagAttributes( $text ) {
1117                 if ( trim( $text ) == '' ) {
1118                         return array();
1119                 }
1120
1121                 $attribs = array();
1122                 $pairs = array();
1123                 if ( !preg_match_all(
1124                         self::getAttribsRegex(),
1125                         $text,
1126                         $pairs,
1127                         PREG_SET_ORDER ) ) {
1128                         return $attribs;
1129                 }
1130
1131                 foreach ( $pairs as $set ) {
1132                         $attribute = strtolower( $set[1] );
1133                         $value = Sanitizer::getTagAttributeCallback( $set );
1134
1135                         // Normalize whitespace
1136                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1137                         $value = trim( $value );
1138
1139                         // Decode character references
1140                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1141                 }
1142                 return $attribs;
1143         }
1144
1145         /**
1146          * Pick the appropriate attribute value from a match set from the
1147          * attribs regex matches.
1148          *
1149          * @param $set Array
1150          * @throws MWException
1151          * @return String
1152          */
1153         private static function getTagAttributeCallback( $set ) {
1154                 if ( isset( $set[6] ) ) {
1155                         # Illegal #XXXXXX color with no quotes.
1156                         return $set[6];
1157                 } elseif ( isset( $set[5] ) ) {
1158                         # No quotes.
1159                         return $set[5];
1160                 } elseif ( isset( $set[4] ) ) {
1161                         # Single-quoted
1162                         return $set[4];
1163                 } elseif ( isset( $set[3] ) ) {
1164                         # Double-quoted
1165                         return $set[3];
1166                 } elseif ( !isset( $set[2] ) ) {
1167                         # In XHTML, attributes must have a value.
1168                         # For 'reduced' form, return explicitly the attribute name here.
1169                         return $set[1];
1170                 } else {
1171                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1172                 }
1173         }
1174
1175         /**
1176          * Normalize whitespace and character references in an XML source-
1177          * encoded text for an attribute value.
1178          *
1179          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1180          * but note that we're not returning the value, but are returning
1181          * XML source fragments that will be slapped into output.
1182          *
1183          * @param $text String
1184          * @return String
1185          */
1186         private static function normalizeAttributeValue( $text ) {
1187                 return str_replace( '"', '&quot;',
1188                         self::normalizeWhitespace(
1189                                 Sanitizer::normalizeCharReferences( $text ) ) );
1190         }
1191
1192         /**
1193          * @param $text string
1194          * @return mixed
1195          */
1196         private static function normalizeWhitespace( $text ) {
1197                 return preg_replace(
1198                         '/\r\n|[\x20\x0d\x0a\x09]/',
1199                         ' ',
1200                         $text );
1201         }
1202
1203         /**
1204          * Normalizes whitespace in a section name, such as might be returned
1205          * by Parser::stripSectionName(), for use in the id's that are used for
1206          * section links.
1207          *
1208          * @param $section String
1209          * @return String
1210          */
1211         static function normalizeSectionNameWhitespace( $section ) {
1212                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1213         }
1214
1215         /**
1216          * Ensure that any entities and character references are legal
1217          * for XML and XHTML specifically. Any stray bits will be
1218          * &amp;-escaped to result in a valid text fragment.
1219          *
1220          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1221          *   numericized (this way we're well-formed even without a DTD)
1222          * b. any numeric char refs must be legal chars, not invalid or forbidden
1223          * c. use lower cased "&#x", not "&#X"
1224          * d. fix or reject non-valid attributes
1225          *
1226          * @param $text String
1227          * @return String
1228          * @private
1229          */
1230         static function normalizeCharReferences( $text ) {
1231                 return preg_replace_callback(
1232                         self::CHAR_REFS_REGEX,
1233                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1234                         $text );
1235         }
1236         /**
1237          * @param $matches String
1238          * @return String
1239          */
1240         static function normalizeCharReferencesCallback( $matches ) {
1241                 $ret = null;
1242                 if ( $matches[1] != '' ) {
1243                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1244                 } elseif ( $matches[2] != '' ) {
1245                         $ret = Sanitizer::decCharReference( $matches[2] );
1246                 } elseif ( $matches[3] != '' ) {
1247                         $ret = Sanitizer::hexCharReference( $matches[3] );
1248                 }
1249                 if ( is_null( $ret ) ) {
1250                         return htmlspecialchars( $matches[0] );
1251                 } else {
1252                         return $ret;
1253                 }
1254         }
1255
1256         /**
1257          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1258          * return the equivalent numeric entity reference (except for the core &lt;
1259          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1260          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1261          * pseudo-entity source (eg &amp;foo;)
1262          *
1263          * @param $name String
1264          * @return String
1265          */
1266         static function normalizeEntity( $name ) {
1267                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1268                         return '&' . self::$htmlEntityAliases[$name] . ';';
1269                 } elseif ( in_array( $name,
1270                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1271                         return "&$name;";
1272                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1273                         return '&#' . self::$htmlEntities[$name] . ';';
1274                 } else {
1275                         return "&amp;$name;";
1276                 }
1277         }
1278
1279         /**
1280          * @param $codepoint
1281          * @return null|string
1282          */
1283         static function decCharReference( $codepoint ) {
1284                 $point = intval( $codepoint );
1285                 if ( Sanitizer::validateCodepoint( $point ) ) {
1286                         return sprintf( '&#%d;', $point );
1287                 } else {
1288                         return null;
1289                 }
1290         }
1291
1292         /**
1293          * @param $codepoint
1294          * @return null|string
1295          */
1296         static function hexCharReference( $codepoint ) {
1297                 $point = hexdec( $codepoint );
1298                 if ( Sanitizer::validateCodepoint( $point ) ) {
1299                         return sprintf( '&#x%x;', $point );
1300                 } else {
1301                         return null;
1302                 }
1303         }
1304
1305         /**
1306          * Returns true if a given Unicode codepoint is a valid character in XML.
1307          * @param $codepoint Integer
1308          * @return Boolean
1309          */
1310         private static function validateCodepoint( $codepoint ) {
1311                 return ($codepoint ==    0x09)
1312                         || ($codepoint ==    0x0a)
1313                         || ($codepoint ==    0x0d)
1314                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1315                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1316                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1317         }
1318
1319         /**
1320          * Decode any character references, numeric or named entities,
1321          * in the text and return a UTF-8 string.
1322          *
1323          * @param $text String
1324          * @return String
1325          */
1326         public static function decodeCharReferences( $text ) {
1327                 return preg_replace_callback(
1328                         self::CHAR_REFS_REGEX,
1329                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1330                         $text );
1331         }
1332
1333         /**
1334          * Decode any character references, numeric or named entities,
1335          * in the next and normalize the resulting string. (bug 14952)
1336          *
1337          * This is useful for page titles, not for text to be displayed,
1338          * MediaWiki allows HTML entities to escape normalization as a feature.
1339          *
1340          * @param string $text (already normalized, containing entities)
1341          * @return String (still normalized, without entities)
1342          */
1343         public static function decodeCharReferencesAndNormalize( $text ) {
1344                 global $wgContLang;
1345                 $text = preg_replace_callback(
1346                         self::CHAR_REFS_REGEX,
1347                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1348                         $text, /* limit */ -1, $count );
1349
1350                 if ( $count ) {
1351                         return $wgContLang->normalize( $text );
1352                 } else {
1353                         return $text;
1354                 }
1355         }
1356
1357         /**
1358          * @param $matches String
1359          * @return String
1360          */
1361         static function decodeCharReferencesCallback( $matches ) {
1362                 if ( $matches[1] != '' ) {
1363                         return Sanitizer::decodeEntity( $matches[1] );
1364                 } elseif ( $matches[2] != '' ) {
1365                         return Sanitizer::decodeChar( intval( $matches[2] ) );
1366                 } elseif ( $matches[3] != '' ) {
1367                         return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1368                 }
1369                 # Last case should be an ampersand by itself
1370                 return $matches[0];
1371         }
1372
1373         /**
1374          * Return UTF-8 string for a codepoint if that is a valid
1375          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1376          * @param $codepoint Integer
1377          * @return String
1378          * @private
1379          */
1380         static function decodeChar( $codepoint ) {
1381                 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1382                         return codepointToUtf8( $codepoint );
1383                 } else {
1384                         return UTF8_REPLACEMENT;
1385                 }
1386         }
1387
1388         /**
1389          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1390          * return the UTF-8 encoding of that character. Otherwise, returns
1391          * pseudo-entity source (eg "&foo;")
1392          *
1393          * @param $name String
1394          * @return String
1395          */
1396         static function decodeEntity( $name ) {
1397                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1398                         $name = self::$htmlEntityAliases[$name];
1399                 }
1400                 if ( isset( self::$htmlEntities[$name] ) ) {
1401                         return codepointToUtf8( self::$htmlEntities[$name] );
1402                 } else {
1403                         return "&$name;";
1404                 }
1405         }
1406
1407         /**
1408          * Fetch the whitelist of acceptable attributes for a given element name.
1409          *
1410          * @param $element String
1411          * @return Array
1412          */
1413         static function attributeWhitelist( $element ) {
1414                 $list = Sanitizer::setupAttributeWhitelist();
1415                 return isset( $list[$element] )
1416                         ? $list[$element]
1417                         : array();
1418         }
1419
1420         /**
1421          * Foreach array key (an allowed HTML element), return an array
1422          * of allowed attributes
1423          * @return Array
1424          */
1425         static function setupAttributeWhitelist() {
1426                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1427
1428                 static $whitelist, $staticInitialised;
1429                 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgHtml5', 'wgAllowMicrodataAttributes' ) );
1430
1431                 if ( isset( $whitelist ) && $staticInitialised == $globalContext ) {
1432                         return $whitelist;
1433                 }
1434
1435                 $common = array(
1436                         # HTML
1437                         'id',
1438                         'class',
1439                         'style',
1440                         'lang',
1441                         'dir',
1442                         'title',
1443
1444                         # WAI-ARIA
1445                         'role',
1446                 );
1447
1448                 if ( $wgAllowRdfaAttributes ) {
1449                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1450                         $common = array_merge( $common, array(
1451                                 'about', 'property', 'resource', 'datatype', 'typeof',
1452                         ) );
1453                 }
1454
1455                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1456                         # add HTML5 microdata tags as specified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1457                         $common = array_merge( $common, array(
1458                                 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1459                         ) );
1460                 }
1461
1462                 $block = array_merge( $common, array( 'align' ) );
1463                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1464                 $tablecell = array(
1465                         'abbr',
1466                         'axis',
1467                         'headers',
1468                         'scope',
1469                         'rowspan',
1470                         'colspan',
1471                         'nowrap', # deprecated
1472                         'width', # deprecated
1473                         'height', # deprecated
1474                         'bgcolor', # deprecated
1475                 );
1476
1477                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1478                 # See: http://www.w3.org/TR/html4/
1479                 $whitelist = array(
1480                         # 7.5.4
1481                         'div'        => $block,
1482                         'center'     => $common, # deprecated
1483                         'span'       => $block, # ??
1484
1485                         # 7.5.5
1486                         'h1'         => $block,
1487                         'h2'         => $block,
1488                         'h3'         => $block,
1489                         'h4'         => $block,
1490                         'h5'         => $block,
1491                         'h6'         => $block,
1492
1493                         # 7.5.6
1494                         # address
1495
1496                         # 8.2.4
1497                         # bdo
1498
1499                         # 9.2.1
1500                         'em'         => $common,
1501                         'strong'     => $common,
1502                         'cite'       => $common,
1503                         'dfn'        => $common,
1504                         'code'       => $common,
1505                         'samp'       => $common,
1506                         'kbd'        => $common,
1507                         'var'        => $common,
1508                         'abbr'       => $common,
1509                         # acronym
1510
1511                         # 9.2.2
1512                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1513                         # q
1514
1515                         # 9.2.3
1516                         'sub'        => $common,
1517                         'sup'        => $common,
1518
1519                         # 9.3.1
1520                         'p'          => $block,
1521
1522                         # 9.3.2
1523                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1524
1525                         # 9.3.4
1526                         'pre'        => array_merge( $common, array( 'width' ) ),
1527
1528                         # 9.4
1529                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1530                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1531
1532                         # 10.2
1533                         'ul'         => array_merge( $common, array( 'type' ) ),
1534                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1535                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1536
1537                         # 10.3
1538                         'dl'         => $common,
1539                         'dd'         => $common,
1540                         'dt'         => $common,
1541
1542                         # 11.2.1
1543                         'table'      => array_merge( $common,
1544                                                                 array( 'summary', 'width', 'border', 'frame',
1545                                                                                 'rules', 'cellspacing', 'cellpadding',
1546                                                                                 'align', 'bgcolor',
1547                                                                 ) ),
1548
1549                         # 11.2.2
1550                         'caption'    => array_merge( $common, array( 'align' ) ),
1551
1552                         # 11.2.3
1553                         'thead'      => array_merge( $common, $tablealign ),
1554                         'tfoot'      => array_merge( $common, $tablealign ),
1555                         'tbody'      => array_merge( $common, $tablealign ),
1556
1557                         # 11.2.4
1558                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1559                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1560
1561                         # 11.2.5
1562                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1563
1564                         # 11.2.6
1565                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1566                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1567
1568                         # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1569                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1570
1571                         # 13.2
1572                         # Not usually allowed, but may be used for extension-style hooks
1573                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1574                         # true
1575                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1576
1577                         # 15.2.1
1578                         'tt'         => $common,
1579                         'b'          => $common,
1580                         'i'          => $common,
1581                         'big'        => $common,
1582                         'small'      => $common,
1583                         'strike'     => $common,
1584                         's'          => $common,
1585                         'u'          => $common,
1586
1587                         # 15.2.2
1588                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1589                         # basefont
1590
1591                         # 15.3
1592                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1593
1594                         # XHTML Ruby annotation text module, simple ruby only.
1595                         # http://www.w3c.org/TR/ruby/
1596                         'ruby'       => $common,
1597                         # rbc
1598                         # rtc
1599                         'rb'         => $common,
1600                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1601                         'rp'         => $common,
1602
1603                         # MathML root element, where used for extensions
1604                         # 'title' may not be 100% valid here; it's XHTML
1605                         # http://www.w3.org/TR/REC-MathML/
1606                         'math'       => array( 'class', 'style', 'id', 'title' ),
1607
1608                         # HTML 5 section 4.6
1609                         'bdi' => $common,
1610
1611                 );
1612
1613                 if ( $wgHtml5 ) {
1614                         # HTML5 elements, defined by:
1615                         # http://www.whatwg.org/specs/web-apps/current-work/multipage/
1616                         $whitelist += array(
1617                                 'data' => array_merge( $common, array( 'value' ) ),
1618                                 'time' => array_merge( $common, array( 'datetime' ) ),
1619                                 'mark' => $common,
1620
1621                                 // meta and link are only permitted by removeHTMLtags when Microdata
1622                                 // is enabled so we don't bother adding a conditional to hide these
1623                                 // Also meta and link are only valid in WikiText as Microdata elements
1624                                 // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1625                                 // So we don't bother including $common attributes that have no purpose.
1626                                 'meta' => array( 'itemprop', 'content' ),
1627                                 'link' => array( 'itemprop', 'href' ),
1628                         );
1629                 }
1630
1631                 $staticInitialised = $globalContext;
1632
1633                 return $whitelist;
1634         }
1635
1636         /**
1637          * Take a fragment of (potentially invalid) HTML and return
1638          * a version with any tags removed, encoded as plain text.
1639          *
1640          * Warning: this return value must be further escaped for literal
1641          * inclusion in HTML output as of 1.10!
1642          *
1643          * @param string $text HTML fragment
1644          * @return String
1645          */
1646         static function stripAllTags( $text ) {
1647                 # Actual <tags>
1648                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1649
1650                 # Normalize &entities and whitespace
1651                 $text = self::decodeCharReferences( $text );
1652                 $text = self::normalizeWhitespace( $text );
1653
1654                 return $text;
1655         }
1656
1657         /**
1658          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1659          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1660          * PHP 5.1 doesn't.
1661          *
1662          * Use for passing XHTML fragments to PHP's XML parsing functions
1663          *
1664          * @return String
1665          */
1666         static function hackDocType() {
1667                 $out = "<!DOCTYPE html [\n";
1668                 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1669                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1670                 }
1671                 $out .= "]>\n";
1672                 return $out;
1673         }
1674
1675         /**
1676          * @param $url string
1677          * @return mixed|string
1678          */
1679         static function cleanUrl( $url ) {
1680                 # Normalize any HTML entities in input. They will be
1681                 # re-escaped by makeExternalLink().
1682                 $url = Sanitizer::decodeCharReferences( $url );
1683
1684                 # Escape any control characters introduced by the above step
1685                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1686                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1687
1688                 # Validate hostname portion
1689                 $matches = array();
1690                 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1691                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1692
1693                         // Characters that will be ignored in IDNs.
1694                         // http://tools.ietf.org/html/3454#section-3.1
1695                         // Strip them before further processing so blacklists and such work.
1696                         $strip = "/
1697                                 \\s|          # general whitespace
1698                                 \xc2\xad|     # 00ad SOFT HYPHEN
1699                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1700                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1701                                 \xe2\x81\xa0| # 2060 WORD JOINER
1702                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1703                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1704                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1705                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1706                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1707                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1708                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1709                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1710                                 /xuD";
1711
1712                         $host = preg_replace( $strip, '', $host );
1713
1714                         // @todo FIXME: Validate hostnames here
1715
1716                         return $protocol . $host . $rest;
1717                 } else {
1718                         return $url;
1719                 }
1720         }
1721
1722         /**
1723          * @param $matches array
1724          * @return string
1725          */
1726         static function cleanUrlCallback( $matches ) {
1727                 return urlencode( $matches[0] );
1728         }
1729
1730         /**
1731          * Does a string look like an e-mail address?
1732          *
1733          * This validates an email address using an HTML5 specification found at:
1734          * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1735          * Which as of 2011-01-24 says:
1736          *
1737          *   A valid e-mail address is a string that matches the ABNF production
1738          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1739          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1740          *   3.5.
1741          *
1742          * This function is an implementation of the specification as requested in
1743          * bug 22449.
1744          *
1745          * Client-side forms will use the same standard validation rules via JS or
1746          * HTML 5 validation; additional restrictions can be enforced server-side
1747          * by extensions via the 'isValidEmailAddr' hook.
1748          *
1749          * Note that this validation doesn't 100% match RFC 2822, but is believed
1750          * to be liberal enough for wide use. Some invalid addresses will still
1751          * pass validation here.
1752          *
1753          * @since 1.18
1754          *
1755          * @param string $addr E-mail address
1756          * @return Bool
1757          */
1758         public static function validateEmail( $addr ) {
1759                 $result = null;
1760                 if ( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1761                         return $result;
1762                 }
1763
1764                 // Please note strings below are enclosed in brackets [], this make the
1765                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1766                 // See bug 26948
1767                 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1768                 $rfc1034_ldh_str = "a-z0-9\\-";
1769
1770                 $HTML5_email_regexp = "/
1771                 ^                      # start of string
1772                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1773                 @                      # 'apostrophe'
1774                 [$rfc1034_ldh_str]+       # First domain part
1775                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1776                 $                      # End of string
1777                 /ix"; // case Insensitive, eXtended
1778
1779                 return (bool) preg_match( $HTML5_email_regexp, $addr );
1780         }
1781 }