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