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  * Regular expression to match various types of character references in
  29  * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
  30  */
  31 define( 'MW_CHAR_REFS_REGEX',
  32         '/&([A-Za-z0-9\x80-\xff]+);
  33          |&\#([0-9]+);
  34          |&\#[xX]([0-9A-Fa-f]+);
  35          |(&)/x' );
  36
  37 /**
  38  * Regular expression to match HTML/XML attribute pairs within a tag.
  39  * Allows some... latitude.
  40  * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
  41  */
  42 $attribFirst = '[:A-Z_a-z]';
  43 $attrib = '[:A-Z_a-z-.0-9]';
  44 $space = '[\x09\x0a\x0d\x20]';
  45 define( 'MW_ATTRIBS_REGEX',
  46         "/(?:^|$space)({$attribFirst}{$attrib}*)
  47           ($space*=$space*
  48                 (?:
  49                  # The attribute value: quoted or alone
  50                   \"([^<\"]*)\"
  51                  | '([^<']*)'
  52                  |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
  53                  |  (\#[0-9a-fA-F]+) # Technically wrong, but lots of
  54                                                          # colors are specified like this.
  55                                                          # We'll be normalizing it.
  56                 )
  57            )?(?=$space|\$)/sx" );
  58
  59 /**
  60  * Regular expression to match URIs that could trigger script execution
  61  */
  62 define( 'MW_EVIL_URI_PATTERN', '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i' );
  63
  64 /**
  65  * Regular expression to match namespace attributes
  66  */
  67 define( 'MW_XMLNS_ATTRIBUTE_PATTRN', "/^xmlns:$attrib+$/" );
  68
  69 /**
  70  * List of all named character entities defined in HTML 4.01
  71  * http://www.w3.org/TR/html4/sgml/entities.html
  72  * @private
  73  */
  74 global $wgHtmlEntities;
  75 $wgHtmlEntities = array(
  76         'Aacute'   => 193,
  77         'aacute'   => 225,
  78         'Acirc'    => 194,
  79         'acirc'    => 226,
  80         'acute'    => 180,
  81         'AElig'    => 198,
  82         'aelig'    => 230,
  83         'Agrave'   => 192,
  84         'agrave'   => 224,
  85         'alefsym'  => 8501,
  86         'Alpha'    => 913,
  87         'alpha'    => 945,
  88         'amp'      => 38,
  89         'and'      => 8743,
  90         'ang'      => 8736,
  91         'Aring'    => 197,
  92         'aring'    => 229,
  93         'asymp'    => 8776,
  94         'Atilde'   => 195,
  95         'atilde'   => 227,
  96         'Auml'     => 196,
  97         'auml'     => 228,
  98         'bdquo'    => 8222,
  99         'Beta'     => 914,
 100         'beta'     => 946,
 101         'brvbar'   => 166,
 102         'bull'     => 8226,
 103         'cap'      => 8745,
 104         'Ccedil'   => 199,
 105         'ccedil'   => 231,
 106         'cedil'    => 184,
 107         'cent'     => 162,
 108         'Chi'      => 935,
 109         'chi'      => 967,
 110         'circ'     => 710,
 111         'clubs'    => 9827,
 112         'cong'     => 8773,
 113         'copy'     => 169,
 114         'crarr'    => 8629,
 115         'cup'      => 8746,
 116         'curren'   => 164,
 117         'dagger'   => 8224,
 118         'Dagger'   => 8225,
 119         'darr'     => 8595,
 120         'dArr'     => 8659,
 121         'deg'      => 176,
 122         'Delta'    => 916,
 123         'delta'    => 948,
 124         'diams'    => 9830,
 125         'divide'   => 247,
 126         'Eacute'   => 201,
 127         'eacute'   => 233,
 128         'Ecirc'    => 202,
 129         'ecirc'    => 234,
 130         'Egrave'   => 200,
 131         'egrave'   => 232,
 132         'empty'    => 8709,
 133         'emsp'     => 8195,
 134         'ensp'     => 8194,
 135         'Epsilon'  => 917,
 136         'epsilon'  => 949,
 137         'equiv'    => 8801,
 138         'Eta'      => 919,
 139         'eta'      => 951,
 140         'ETH'      => 208,
 141         'eth'      => 240,
 142         'Euml'     => 203,
 143         'euml'     => 235,
 144         'euro'     => 8364,
 145         'exist'    => 8707,
 146         'fnof'     => 402,
 147         'forall'   => 8704,
 148         'frac12'   => 189,
 149         'frac14'   => 188,
 150         'frac34'   => 190,
 151         'frasl'    => 8260,
 152         'Gamma'    => 915,
 153         'gamma'    => 947,
 154         'ge'       => 8805,
 155         'gt'       => 62,
 156         'harr'     => 8596,
 157         'hArr'     => 8660,
 158         'hearts'   => 9829,
 159         'hellip'   => 8230,
 160         'Iacute'   => 205,
 161         'iacute'   => 237,
 162         'Icirc'    => 206,
 163         'icirc'    => 238,
 164         'iexcl'    => 161,
 165         'Igrave'   => 204,
 166         'igrave'   => 236,
 167         'image'    => 8465,
 168         'infin'    => 8734,
 169         'int'      => 8747,
 170         'Iota'     => 921,
 171         'iota'     => 953,
 172         'iquest'   => 191,
 173         'isin'     => 8712,
 174         'Iuml'     => 207,
 175         'iuml'     => 239,
 176         'Kappa'    => 922,
 177         'kappa'    => 954,
 178         'Lambda'   => 923,
 179         'lambda'   => 955,
 180         'lang'     => 9001,
 181         'laquo'    => 171,
 182         'larr'     => 8592,
 183         'lArr'     => 8656,
 184         'lceil'    => 8968,
 185         'ldquo'    => 8220,
 186         'le'       => 8804,
 187         'lfloor'   => 8970,
 188         'lowast'   => 8727,
 189         'loz'      => 9674,
 190         'lrm'      => 8206,
 191         'lsaquo'   => 8249,
 192         'lsquo'    => 8216,
 193         'lt'       => 60,
 194         'macr'     => 175,
 195         'mdash'    => 8212,
 196         'micro'    => 181,
 197         'middot'   => 183,
 198         'minus'    => 8722,
 199         'Mu'       => 924,
 200         'mu'       => 956,
 201         'nabla'    => 8711,
 202         'nbsp'     => 160,
 203         'ndash'    => 8211,
 204         'ne'       => 8800,
 205         'ni'       => 8715,
 206         'not'      => 172,
 207         'notin'    => 8713,
 208         'nsub'     => 8836,
 209         'Ntilde'   => 209,
 210         'ntilde'   => 241,
 211         'Nu'       => 925,
 212         'nu'       => 957,
 213         'Oacute'   => 211,
 214         'oacute'   => 243,
 215         'Ocirc'    => 212,
 216         'ocirc'    => 244,
 217         'OElig'    => 338,
 218         'oelig'    => 339,
 219         'Ograve'   => 210,
 220         'ograve'   => 242,
 221         'oline'    => 8254,
 222         'Omega'    => 937,
 223         'omega'    => 969,
 224         'Omicron'  => 927,
 225         'omicron'  => 959,
 226         'oplus'    => 8853,
 227         'or'       => 8744,
 228         'ordf'     => 170,
 229         'ordm'     => 186,
 230         'Oslash'   => 216,
 231         'oslash'   => 248,
 232         'Otilde'   => 213,
 233         'otilde'   => 245,
 234         'otimes'   => 8855,
 235         'Ouml'     => 214,
 236         'ouml'     => 246,
 237         'para'     => 182,
 238         'part'     => 8706,
 239         'permil'   => 8240,
 240         'perp'     => 8869,
 241         'Phi'      => 934,
 242         'phi'      => 966,
 243         'Pi'       => 928,
 244         'pi'       => 960,
 245         'piv'      => 982,
 246         'plusmn'   => 177,
 247         'pound'    => 163,
 248         'prime'    => 8242,
 249         'Prime'    => 8243,
 250         'prod'     => 8719,
 251         'prop'     => 8733,
 252         'Psi'      => 936,
 253         'psi'      => 968,
 254         'quot'     => 34,
 255         'radic'    => 8730,
 256         'rang'     => 9002,
 257         'raquo'    => 187,
 258         'rarr'     => 8594,
 259         'rArr'     => 8658,
 260         'rceil'    => 8969,
 261         'rdquo'    => 8221,
 262         'real'     => 8476,
 263         'reg'      => 174,
 264         'rfloor'   => 8971,
 265         'Rho'      => 929,
 266         'rho'      => 961,
 267         'rlm'      => 8207,
 268         'rsaquo'   => 8250,
 269         'rsquo'    => 8217,
 270         'sbquo'    => 8218,
 271         'Scaron'   => 352,
 272         'scaron'   => 353,
 273         'sdot'     => 8901,
 274         'sect'     => 167,
 275         'shy'      => 173,
 276         'Sigma'    => 931,
 277         'sigma'    => 963,
 278         'sigmaf'   => 962,
 279         'sim'      => 8764,
 280         'spades'   => 9824,
 281         'sub'      => 8834,
 282         'sube'     => 8838,
 283         'sum'      => 8721,
 284         'sup'      => 8835,
 285         'sup1'     => 185,
 286         'sup2'     => 178,
 287         'sup3'     => 179,
 288         'supe'     => 8839,
 289         'szlig'    => 223,
 290         'Tau'      => 932,
 291         'tau'      => 964,
 292         'there4'   => 8756,
 293         'Theta'    => 920,
 294         'theta'    => 952,
 295         'thetasym' => 977,
 296         'thinsp'   => 8201,
 297         'THORN'    => 222,
 298         'thorn'    => 254,
 299         'tilde'    => 732,
 300         'times'    => 215,
 301         'trade'    => 8482,
 302         'Uacute'   => 218,
 303         'uacute'   => 250,
 304         'uarr'     => 8593,
 305         'uArr'     => 8657,
 306         'Ucirc'    => 219,
 307         'ucirc'    => 251,
 308         'Ugrave'   => 217,
 309         'ugrave'   => 249,
 310         'uml'      => 168,
 311         'upsih'    => 978,
 312         'Upsilon'  => 933,
 313         'upsilon'  => 965,
 314         'Uuml'     => 220,
 315         'uuml'     => 252,
 316         'weierp'   => 8472,
 317         'Xi'       => 926,
 318         'xi'       => 958,
 319         'Yacute'   => 221,
 320         'yacute'   => 253,
 321         'yen'      => 165,
 322         'Yuml'     => 376,
 323         'yuml'     => 255,
 324         'Zeta'     => 918,
 325         'zeta'     => 950,
 326         'zwj'      => 8205,
 327         'zwnj'     => 8204 );
 328
 329 /**
 330  * Character entity aliases accepted by MediaWiki
 331  */
 332 global $wgHtmlEntityAliases;
 333 $wgHtmlEntityAliases = array(
 334         'רלמ' => 'rlm',
 335         'رلم' => 'rlm',
 336 );
 337
 338
 339 /**
 340  * XHTML sanitizer for MediaWiki
 341  * @ingroup Parser
 342  */
 343 class Sanitizer {
 344         /**
 345          * Cleans up HTML, removes dangerous tags and attributes, and
 346          * removes HTML comments
 347          * @private
 348          * @param $text String
 349          * @param $processCallback Callback to do any variable or parameter replacements in HTML attribute values
 350          * @param $args Array for the processing callback
 351          * @param $extratags Array for any extra tags to include
 352          * @param $removetags Array for any tags (default or extra) to exclude
 353          * @return string
 354          */
 355         static function removeHTMLtags( $text, $processCallback = null, $args = array(), $extratags = array(), $removetags = array() ) {
 356                 global $wgUseTidy;
 357
 358                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 359                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 360
 361                 wfProfileIn( __METHOD__ );
 362
 363                 if ( !$staticInitialised ) {
 364
 365                         $htmlpairsStatic = array( # Tags that must be closed
 366                                 'b', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 367                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 368                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 369                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 370                                 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'abbr', 'dfn',
 371                                 'kbd', 'samp'
 372                         );
 373                         $htmlsingle = array(
 374                                 'br', 'hr', 'li', 'dt', 'dd'
 375                         );
 376                         $htmlsingleonly = array( # Elements that cannot have close tags
 377                                 'br', 'hr'
 378                         );
 379                         $htmlnest = array( # Tags that can be nested--??
 380                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 381                                 'dl', 'font', 'big', 'small', 'sub', 'sup', 'span'
 382                         );
 383                         $tabletags = array( # Can only appear inside table, we will close them
 384                                 'td', 'th', 'tr',
 385                         );
 386                         $htmllist = array( # Tags used by list
 387                                 'ul','ol',
 388                         );
 389                         $listtags = array( # Tags that can appear in a list
 390                                 'li',
 391                         );
 392
 393                         global $wgAllowImageTag;
 394                         if ( $wgAllowImageTag ) {
 395                                 $htmlsingle[] = 'img';
 396                                 $htmlsingleonly[] = 'img';
 397                         }
 398
 399                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 400                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 401
 402                         # Convert them all to hashtables for faster lookup
 403                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 404                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 405                         foreach ( $vars as $var ) {
 406                                 $$var = array_flip( $$var );
 407                         }
 408                         $staticInitialised = true;
 409                 }
 410                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 411                 $extratags = array_flip( $extratags );
 412                 $removetags = array_flip( $removetags );
 413                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 414                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ) , $removetags );
 415
 416                 # Remove HTML comments
 417                 $text = Sanitizer::removeHTMLcomments( $text );
 418                 $bits = explode( '<', $text );
 419                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 420                 if ( !$wgUseTidy ) {
 421                         $tagstack = $tablestack = array();
 422                         foreach ( $bits as $x ) {
 423                                 $regs = array();
 424                                 # $slash: Does the current element start with a '/'?
 425                                 # $t: Current element name
 426                                 # $params: String between element name and >
 427                                 # $brace: Ending '>' or '/>'
 428                                 # $rest: Everything until the next element of $bits
 429                                 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 430                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 431                                 } else {
 432                                         $slash = $t = $params = $brace = $rest = null;
 433                                 }
 434
 435                                 $badtag = false;
 436                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 437                                         # Check our stack
 438                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 439                                                 $badtag = true;
 440                                         } elseif ( $slash ) {
 441                                                 # Closing a tag... is it the one we just opened?
 442                                                 $ot = @array_pop( $tagstack );
 443                                                 if ( $ot != $t ) {
 444                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 445                                                                 # Pop all elements with an optional close tag
 446                                                                 # and see if we find a match below them
 447                                                                 $optstack = array();
 448                                                                 array_push( $optstack, $ot );
 449                                                                 $ot = @array_pop( $tagstack );
 450                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 451                                                                         array_push( $optstack, $ot );
 452                                                                         $ot = @array_pop( $tagstack );
 453                                                                 }
 454                                                                 if ( $t != $ot ) {
 455                                                                         # No match. Push the optional elements back again
 456                                                                         $badtag = true;
 457                                                                         while ( $ot = @array_pop( $optstack ) ) {
 458                                                                                 array_push( $tagstack, $ot );
 459                                                                         }
 460                                                                 }
 461                                                         } else {
 462                                                                 @array_push( $tagstack, $ot );
 463                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 464                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 465                                                                         $badtag = true;
 466                                                                 }
 467                                                         }
 468                                                 } else {
 469                                                         if ( $t == 'table' ) {
 470                                                                 $tagstack = array_pop( $tablestack );
 471                                                         }
 472                                                 }
 473                                                 $newparams = '';
 474                                         } else {
 475                                                 # Keep track for later
 476                                                 if ( isset( $tabletags[$t] ) &&
 477                                                 !in_array( 'table', $tagstack ) ) {
 478                                                         $badtag = true;
 479                                                 } elseif ( in_array( $t, $tagstack ) &&
 480                                                 !isset( $htmlnest [$t ] ) ) {
 481                                                         $badtag = true;
 482                                                 # Is it a self closed htmlpair ? (bug 5487)
 483                                                 } elseif ( $brace == '/>' &&
 484                                                 isset( $htmlpairs[$t] ) ) {
 485                                                         $badtag = true;
 486                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 487                                                         # Hack to force empty tag for uncloseable elements
 488                                                         $brace = '/>';
 489                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 490                                                         # Hack to not close $htmlsingle tags
 491                                                         $brace = null;
 492                                                 } elseif ( isset( $tabletags[$t] )
 493                                                 && in_array( $t, $tagstack ) ) {
 494                                                         // New table tag but forgot to close the previous one
 495                                                         $text .= "</$t>";
 496                                                 } else {
 497                                                         if ( $t == 'table' ) {
 498                                                                 array_push( $tablestack, $tagstack );
 499                                                                 $tagstack = array();
 500                                                         }
 501                                                         array_push( $tagstack, $t );
 502                                                 }
 503
 504                                                 # Replace any variables or template parameters with
 505                                                 # plaintext results.
 506                                                 if( is_callable( $processCallback ) ) {
 507                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 508                                                 }
 509
 510                                                 # Strip non-approved attributes from the tag
 511                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 512                                         }
 513                                         if ( !$badtag ) {
 514                                                 $rest = str_replace( '>', '&gt;', $rest );
 515                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 516                                                 $text .= "<$slash$t$newparams$close>$rest";
 517                                                 continue;
 518                                         }
 519                                 }
 520                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 521                         }
 522                         # Close off any remaining tags
 523                         while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
 524                                 $text .= "</$t>\n";
 525                                 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
 526                         }
 527                 } else {
 528                         # this might be possible using tidy itself
 529                         foreach ( $bits as $x ) {
 530                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 531                                 $x, $regs );
 532                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 533                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 534                                         if( is_callable( $processCallback ) ) {
 535                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 536                                         }
 537                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 538                                         $rest = str_replace( '>', '&gt;', $rest );
 539                                         $text .= "<$slash$t$newparams$brace$rest";
 540                                 } else {
 541                                         $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 542                                 }
 543                         }
 544                 }
 545                 wfProfileOut( __METHOD__ );
 546                 return $text;
 547         }
 548
 549         /**
 550          * Remove '<!--', '-->', and everything between.
 551          * To avoid leaving blank lines, when a comment is both preceded
 552          * and followed by a newline (ignoring spaces), trim leading and
 553          * trailing spaces and one of the newlines.
 554          *
 555          * @private
 556          * @param $text String
 557          * @return string
 558          */
 559         static function removeHTMLcomments( $text ) {
 560                 wfProfileIn( __METHOD__ );
 561                 while (($start = strpos($text, '<!--')) !== false) {
 562                         $end = strpos($text, '-->', $start + 4);
 563                         if ($end === false) {
 564                                 # Unterminated comment; bail out
 565                                 break;
 566                         }
 567
 568                         $end += 3;
 569
 570                         # Trim space and newline if the comment is both
 571                         # preceded and followed by a newline
 572                         $spaceStart = max($start - 1, 0);
 573                         $spaceLen = $end - $spaceStart;
 574                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 575                                 $spaceStart--;
 576                                 $spaceLen++;
 577                         }
 578                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 579                                 $spaceLen++;
 580                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 581                                 # Remove the comment, leading and trailing
 582                                 # spaces, and leave only one newline.
 583                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 584                         }
 585                         else {
 586                                 # Remove just the comment.
 587                                 $text = substr_replace($text, '', $start, $end - $start);
 588                         }
 589                 }
 590                 wfProfileOut( __METHOD__ );
 591                 return $text;
 592         }
 593
 594         /**
 595          * Take an array of attribute names and values and normalize or discard
 596          * illegal values for the given element type.
 597          *
 598          * - Discards attributes not on a whitelist for the given element
 599          * - Unsafe style attributes are discarded
 600          * - Invalid id attributes are reencoded
 601          *
 602          * @param $attribs Array
 603          * @param $element String
 604          * @return Array
 605          *
 606          * @todo Check for legal values where the DTD limits things.
 607          * @todo Check for unique id attribute :P
 608          */
 609         static function validateTagAttributes( $attribs, $element ) {
 610                 return Sanitizer::validateAttributes( $attribs,
 611                         Sanitizer::attributeWhitelist( $element ) );
 612         }
 613
 614         /**
 615          * Take an array of attribute names and values and normalize or discard
 616          * illegal values for the given whitelist.
 617          *
 618          * - Discards attributes not the given whitelist
 619          * - Unsafe style attributes are discarded
 620          * - Invalid id attributes are reencoded
 621          *
 622          * @param $attribs Array
 623          * @param $whitelist Array: list of allowed attribute names
 624          * @return Array
 625          *
 626          * @todo Check for legal values where the DTD limits things.
 627          * @todo Check for unique id attribute :P
 628          */
 629         static function validateAttributes( $attribs, $whitelist ) {
 630                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
 631
 632                 $whitelist = array_flip( $whitelist );
 633                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 634
 635                 $out = array();
 636                 foreach( $attribs as $attribute => $value ) {
 637                         #allow XML namespace declaration if RDFa is enabled
 638                         if ( $wgAllowRdfaAttributes && preg_match( MW_XMLNS_ATTRIBUTE_PATTRN, $attribute ) ) {
 639                                 if ( !preg_match( MW_EVIL_URI_PATTERN, $value ) ) {
 640                                         $out[$attribute] = $value;
 641                                 }
 642
 643                                 continue;
 644                         }
 645
 646                         # Allow any attribute beginning with "data-", if in HTML5 mode
 647                         if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
 648                                 continue;
 649                         }
 650
 651                         # Strip javascript "expression" from stylesheets.
 652                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 653                         if( $attribute == 'style' ) {
 654                                 $value = Sanitizer::checkCss( $value );
 655                         }
 656
 657                         if ( $attribute === 'id' ) {
 658                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 659                         }
 660
 661                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 662                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 663                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 664                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 665                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 666                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 667
 668                                 //Paranoia. Allow "simple" values but suppress javascript
 669                                 if ( preg_match( MW_EVIL_URI_PATTERN, $value ) ) {
 670                                         continue;
 671                                 }
 672                         }
 673
 674                         # NOTE: even though elements using href/src are not allowed directly, supply
 675                         #       validation code that can be used by tag hook handlers, etc
 676                         if ( $attribute === 'href' || $attribute === 'src' ) {
 677                                 if ( !preg_match( $hrefExp, $value ) ) {
 678                                         continue; //drop any href or src attributes not using an allowed protocol.
 679                                                   //NOTE: this also drops all relative URLs
 680                                 }
 681                         }
 682
 683                         // If this attribute was previously set, override it.
 684                         // Output should only have one attribute of each name.
 685                         $out[$attribute] = $value;
 686                 }
 687
 688                 if ( $wgAllowMicrodataAttributes ) {
 689                         # There are some complicated validity constraints we need to
 690                         # enforce here.  First of all, we don't want to allow non-standard
 691                         # itemtypes.
 692                         $allowedTypes = array(
 693                                 'http://microformats.org/profile/hcard',
 694                                 'http://microformats.org/profile/hcalendar#vevent',
 695                                 'http://n.whatwg.org/work',
 696                         );
 697                         if ( isset( $out['itemtype'] ) && !in_array( $out['itemtype'],
 698                         $allowedTypes ) ) {
 699                                 # Kill everything
 700                                 unset( $out['itemscope'] );
 701                         }
 702                         # itemtype, itemid, itemref don't make sense without itemscope
 703                         if ( !array_key_exists( 'itemscope', $out ) ) {
 704                                 unset( $out['itemtype'] );
 705                                 unset( $out['itemid'] );
 706                                 unset( $out['itemref'] );
 707                         }
 708                         # TODO: Strip itemprop if we aren't descendants of an itemscope.
 709                 }
 710                 return $out;
 711         }
 712
 713         /**
 714          * Merge two sets of HTML attributes.  Conflicting items in the second set
 715          * will override those in the first, except for 'class' attributes which
 716          * will be combined (if they're both strings).
 717          *
 718          * @todo implement merging for other attributes such as style
 719          * @param $a Array
 720          * @param $b Array
 721          * @return array
 722          */
 723         static function mergeAttributes( $a, $b ) {
 724                 $out = array_merge( $a, $b );
 725                 if( isset( $a['class'] ) && isset( $b['class'] )
 726                 && is_string( $a['class'] ) && is_string( $b['class'] )
 727                 && $a['class'] !== $b['class'] ) {
 728                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 729                                 -1, PREG_SPLIT_NO_EMPTY );
 730                         $out['class'] = implode( ' ', array_unique( $classes ) );
 731                 }
 732                 return $out;
 733         }
 734
 735         /**
 736          * Pick apart some CSS and check it for forbidden or unsafe structures.
 737          * Returns a sanitized string, or false if it was just too evil.
 738          *
 739          * Currently URL references, 'expression', 'tps' are forbidden.
 740          *
 741          * @param $value String
 742          * @return Mixed
 743          */
 744         static function checkCss( $value ) {
 745                 $value = Sanitizer::decodeCharReferences( $value );
 746
 747                 // Remove any comments; IE gets token splitting wrong
 748                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 749
 750                 // Decode escape sequences and line continuation
 751                 // See the grammar in the CSS 2 spec, appendix D.
 752                 static $decodeRegex;
 753                 if ( !$decodeRegex ) {
 754                         $space = '[\\x20\\t\\r\\n\\f]';
 755                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 756                         $backslash = '\\\\';
 757                         $decodeRegex = "/ $backslash
 758                                 (?:
 759                                         ($nl) |  # 1. Line continuation
 760                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 761                                         (.) | # 3. backslash cancelling special meaning
 762                                         () | # 4. backslash at end of string
 763                                 )/xu";
 764                 }
 765                 $value = preg_replace_callback( $decodeRegex,
 766                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 767
 768                 // Reject problematic keywords and control characters
 769                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 770                         return '/* invalid control char */';
 771                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
 772                         return '/* insecure input */';
 773                 }
 774                 return $value;
 775         }
 776
 777         static function cssDecodeCallback( $matches ) {
 778                 if ( $matches[1] !== '' ) {
 779                         // Line continuation
 780                         return '';
 781                 } elseif ( $matches[2] !== '' ) {
 782                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 783                 } elseif ( $matches[3] !== '' ) {
 784                         $char = $matches[3];
 785                 } else {
 786                         $char = '\\';
 787                 }
 788                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 789                         // These characters need to be escaped in strings
 790                         // Clean up the escape sequence to avoid parsing errors by clients
 791                         return '\\' . dechex( ord( $char ) ) . ' ';
 792                 } else {
 793                         // Decode unnecessary escape
 794                         return $char;
 795                 }
 796         }
 797
 798         /**
 799          * Take a tag soup fragment listing an HTML element's attributes
 800          * and normalize it to well-formed XML, discarding unwanted attributes.
 801          * Output is safe for further wikitext processing, with escaping of
 802          * values that could trigger problems.
 803          *
 804          * - Normalizes attribute names to lowercase
 805          * - Discards attributes not on a whitelist for the given element
 806          * - Turns broken or invalid entities into plaintext
 807          * - Double-quotes all attribute values
 808          * - Attributes without values are given the name as attribute
 809          * - Double attributes are discarded
 810          * - Unsafe style attributes are discarded
 811          * - Prepends space if there are attributes.
 812          *
 813          * @param $text String
 814          * @param $element String
 815          * @return String
 816          */
 817         static function fixTagAttributes( $text, $element ) {
 818                 if( trim( $text ) == '' ) {
 819                         return '';
 820                 }
 821
 822                 $stripped = Sanitizer::validateTagAttributes(
 823                         Sanitizer::decodeTagAttributes( $text ), $element );
 824
 825                 $attribs = array();
 826                 foreach( $stripped as $attribute => $value ) {
 827                         $encAttribute = htmlspecialchars( $attribute );
 828                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 829
 830                         $attribs[] = "$encAttribute=\"$encValue\"";
 831                 }
 832                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 833         }
 834
 835         /**
 836          * Encode an attribute value for HTML output.
 837          * @param $text String
 838          * @return HTML-encoded text fragment
 839          */
 840         static function encodeAttribute( $text ) {
 841                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 842
 843                 // Whitespace is normalized during attribute decoding,
 844                 // so if we've been passed non-spaces we must encode them
 845                 // ahead of time or they won't be preserved.
 846                 $encValue = strtr( $encValue, array(
 847                         "\n" => '&#10;',
 848                         "\r" => '&#13;',
 849                         "\t" => '&#9;',
 850                 ) );
 851
 852                 return $encValue;
 853         }
 854
 855         /**
 856          * Encode an attribute value for HTML tags, with extra armoring
 857          * against further wiki processing.
 858          * @param $text String
 859          * @return HTML-encoded text fragment
 860          */
 861         static function safeEncodeAttribute( $text ) {
 862                 $encValue = Sanitizer::encodeAttribute( $text );
 863
 864                 # Templates and links may be expanded in later parsing,
 865                 # creating invalid or dangerous output. Suppress this.
 866                 $encValue = strtr( $encValue, array(
 867                         '<'    => '&lt;',   // This should never happen,
 868                         '>'    => '&gt;',   // we've received invalid input
 869                         '"'    => '&quot;', // which should have been escaped.
 870                         '{'    => '&#123;',
 871                         '['    => '&#91;',
 872                         "''"   => '&#39;&#39;',
 873                         'ISBN' => '&#73;SBN',
 874                         'RFC'  => '&#82;FC',
 875                         'PMID' => '&#80;MID',
 876                         '|'    => '&#124;',
 877                         '__'   => '&#95;_',
 878                 ) );
 879
 880                 # Stupid hack
 881                 $encValue = preg_replace_callback(
 882                         '/(' . wfUrlProtocols() . ')/',
 883                         array( 'Sanitizer', 'armorLinksCallback' ),
 884                         $encValue );
 885                 return $encValue;
 886         }
 887
 888         /**
 889          * Given a value, escape it so that it can be used in an id attribute and
 890          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
 891          * true, allowing anything but ASCII whitespace.  Otherwise it will use
 892          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
 893          * escaped with lots of dots.
 894          *
 895          * To ensure we don't have to bother escaping anything, we also strip ', ",
 896          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
 897          * We also strip # because it upsets IE, and % because it could be
 898          * ambiguous if it's part of something that looks like a percent escape
 899          * (which don't work reliably in fragments cross-browser).
 900          *
 901          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
 902          *                                                          in the id and
 903          *                                                          name attributes
 904          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
 905          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
 906          *   HTML5 definition of id attribute
 907          *
 908          * @param $id String: id to escape
 909          * @param $options Mixed: string or array of strings (default is array()):
 910          *   'noninitial': This is a non-initial fragment of an id, not a full id,
 911          *       so don't pay attention if the first character isn't valid at the
 912          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
 913          *       false.
 914          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
 915          *       if $wgExperimentalHtmlIds is used, so we can generate extra
 916          *       anchors and links won't break.
 917          * @return String
 918          */
 919         static function escapeId( $id, $options = array() ) {
 920                 global $wgHtml5, $wgExperimentalHtmlIds;
 921                 $options = (array)$options;
 922
 923                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
 924                         $id = Sanitizer::decodeCharReferences( $id );
 925                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
 926                         $id = trim( $id, '_' );
 927                         if ( $id === '' ) {
 928                                 # Must have been all whitespace to start with.
 929                                 return '_';
 930                         } else {
 931                                 return $id;
 932                         }
 933                 }
 934
 935                 # HTML4-style escaping
 936                 static $replace = array(
 937                         '%3A' => ':',
 938                         '%' => '.'
 939                 );
 940
 941                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
 942                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
 943
 944                 if ( !preg_match( '/^[a-zA-Z]/', $id )
 945                 && !in_array( 'noninitial', $options ) )  {
 946                         // Initial character must be a letter!
 947                         $id = "x$id";
 948                 }
 949                 return $id;
 950         }
 951
 952         /**
 953          * Given a value, escape it so that it can be used as a CSS class and
 954          * return it.
 955          *
 956          * @todo For extra validity, input should be validated UTF-8.
 957          *
 958          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
 959          *
 960          * @param $class String
 961          * @return String
 962          */
 963         static function escapeClass( $class ) {
 964                 // Convert ugly stuff to underscores and kill underscores in ugly places
 965                 return rtrim(preg_replace(
 966                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
 967                         '_',
 968                         $class ), '_');
 969         }
 970
 971         /**
 972          * Given HTML input, escape with htmlspecialchars but un-escape entites.
 973          * This allows (generally harmless) entities like &#160; to survive.
 974          *
 975          * @param $html String to escape
 976          * @return String: escaped input
 977          */
 978         static function escapeHtmlAllowEntities( $html ) {
 979                 $html = Sanitizer::decodeCharReferences( $html );
 980                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
 981                 # hurt.
 982                 $html = htmlspecialchars( $html, ENT_QUOTES );
 983                 return $html;
 984         }
 985
 986         /**
 987          * Regex replace callback for armoring links against further processing.
 988          * @param $matches Array
 989          * @return string
 990          */
 991         private static function armorLinksCallback( $matches ) {
 992                 return str_replace( ':', '&#58;', $matches[1] );
 993         }
 994
 995         /**
 996          * Return an associative array of attribute names and values from
 997          * a partial tag string. Attribute names are forces to lowercase,
 998          * character references are decoded to UTF-8 text.
 999          *
1000          * @param $text String
1001          * @return Array
1002          */
1003         public static function decodeTagAttributes( $text ) {
1004                 if( trim( $text ) == '' ) {
1005                         return array();
1006                 }
1007
1008                 $attribs = array();
1009                 $pairs = array();
1010                 if( !preg_match_all(
1011                         MW_ATTRIBS_REGEX,
1012                         $text,
1013                         $pairs,
1014                         PREG_SET_ORDER ) ) {
1015                         return $attribs;
1016                 }
1017
1018                 foreach( $pairs as $set ) {
1019                         $attribute = strtolower( $set[1] );
1020                         $value = Sanitizer::getTagAttributeCallback( $set );
1021
1022                         // Normalize whitespace
1023                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1024                         $value = trim( $value );
1025
1026                         // Decode character references
1027                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1028                 }
1029                 return $attribs;
1030         }
1031
1032         /**
1033          * Pick the appropriate attribute value from a match set from the
1034          * MW_ATTRIBS_REGEX matches.
1035          *
1036          * @param $set Array
1037          * @return String
1038          */
1039         private static function getTagAttributeCallback( $set ) {
1040                 if( isset( $set[6] ) ) {
1041                         # Illegal #XXXXXX color with no quotes.
1042                         return $set[6];
1043                 } elseif( isset( $set[5] ) ) {
1044                         # No quotes.
1045                         return $set[5];
1046                 } elseif( isset( $set[4] ) ) {
1047                         # Single-quoted
1048                         return $set[4];
1049                 } elseif( isset( $set[3] ) ) {
1050                         # Double-quoted
1051                         return $set[3];
1052                 } elseif( !isset( $set[2] ) ) {
1053                         # In XHTML, attributes must have a value.
1054                         # For 'reduced' form, return explicitly the attribute name here.
1055                         return $set[1];
1056                 } else {
1057                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1058                 }
1059         }
1060
1061         /**
1062          * Normalize whitespace and character references in an XML source-
1063          * encoded text for an attribute value.
1064          *
1065          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1066          * but note that we're not returning the value, but are returning
1067          * XML source fragments that will be slapped into output.
1068          *
1069          * @param $text String
1070          * @return String
1071          */
1072         private static function normalizeAttributeValue( $text ) {
1073                 return str_replace( '"', '&quot;',
1074                         self::normalizeWhitespace(
1075                                 Sanitizer::normalizeCharReferences( $text ) ) );
1076         }
1077
1078         private static function normalizeWhitespace( $text ) {
1079                 return preg_replace(
1080                         '/\r\n|[\x20\x0d\x0a\x09]/',
1081                         ' ',
1082                         $text );
1083         }
1084
1085         /**
1086          * Normalizes whitespace in a section name, such as might be returned
1087          * by Parser::stripSectionName(), for use in the id's that are used for
1088          * section links.
1089          *
1090          * @param $section String
1091          * @return String
1092          */
1093         static function normalizeSectionNameWhitespace( $section ) {
1094                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1095         }
1096
1097         /**
1098          * Ensure that any entities and character references are legal
1099          * for XML and XHTML specifically. Any stray bits will be
1100          * &amp;-escaped to result in a valid text fragment.
1101          *
1102          * a. any named char refs must be known in XHTML
1103          * b. any numeric char refs must be legal chars, not invalid or forbidden
1104          * c. use &#x, not &#X
1105          * d. fix or reject non-valid attributes
1106          *
1107          * @param $text String
1108          * @return String
1109          * @private
1110          */
1111         static function normalizeCharReferences( $text ) {
1112                 return preg_replace_callback(
1113                         MW_CHAR_REFS_REGEX,
1114                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1115                         $text );
1116         }
1117         /**
1118          * @param $matches String
1119          * @return String
1120          */
1121         static function normalizeCharReferencesCallback( $matches ) {
1122                 $ret = null;
1123                 if( $matches[1] != '' ) {
1124                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1125                 } elseif( $matches[2] != '' ) {
1126                         $ret = Sanitizer::decCharReference( $matches[2] );
1127                 } elseif( $matches[3] != ''  ) {
1128                         $ret = Sanitizer::hexCharReference( $matches[3] );
1129                 }
1130                 if( is_null( $ret ) ) {
1131                         return htmlspecialchars( $matches[0] );
1132                 } else {
1133                         return $ret;
1134                 }
1135         }
1136
1137         /**
1138          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1139          * return the named entity reference as is. If the entity is a
1140          * MediaWiki-specific alias, returns the HTML equivalent. Otherwise,
1141          * returns HTML-escaped text of pseudo-entity source (eg &amp;foo;)
1142          *
1143          * @param $name String
1144          * @return String
1145          */
1146         static function normalizeEntity( $name ) {
1147                 global $wgHtmlEntities, $wgHtmlEntityAliases;
1148                 if ( isset( $wgHtmlEntityAliases[$name] ) ) {
1149                         return "&{$wgHtmlEntityAliases[$name]};";
1150                 } elseif( isset( $wgHtmlEntities[$name] ) ) {
1151                         return "&$name;";
1152                 } else {
1153                         return "&amp;$name;";
1154                 }
1155         }
1156
1157         static function decCharReference( $codepoint ) {
1158                 $point = intval( $codepoint );
1159                 if( Sanitizer::validateCodepoint( $point ) ) {
1160                         return sprintf( '&#%d;', $point );
1161                 } else {
1162                         return null;
1163                 }
1164         }
1165
1166         static function hexCharReference( $codepoint ) {
1167                 $point = hexdec( $codepoint );
1168                 if( Sanitizer::validateCodepoint( $point ) ) {
1169                         return sprintf( '&#x%x;', $point );
1170                 } else {
1171                         return null;
1172                 }
1173         }
1174
1175         /**
1176          * Returns true if a given Unicode codepoint is a valid character in XML.
1177          * @param $codepoint Integer
1178          * @return Boolean
1179          */
1180         private static function validateCodepoint( $codepoint ) {
1181                 return ($codepoint ==    0x09)
1182                         || ($codepoint ==    0x0a)
1183                         || ($codepoint ==    0x0d)
1184                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1185                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1186                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1187         }
1188
1189         /**
1190          * Decode any character references, numeric or named entities,
1191          * in the text and return a UTF-8 string.
1192          *
1193          * @param $text String
1194          * @return String
1195          */
1196         public static function decodeCharReferences( $text ) {
1197                 return preg_replace_callback(
1198                         MW_CHAR_REFS_REGEX,
1199                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1200                         $text );
1201         }
1202
1203         /**
1204          * Decode any character references, numeric or named entities,
1205          * in the next and normalize the resulting string. (bug 14952)
1206          *
1207          * This is useful for page titles, not for text to be displayed,
1208          * MediaWiki allows HTML entities to escape normalization as a feature.
1209          *
1210          * @param $text String (already normalized, containing entities)
1211          * @return String (still normalized, without entities)
1212          */
1213         public static function decodeCharReferencesAndNormalize( $text ) {
1214                 global $wgContLang;
1215                 $text = preg_replace_callback(
1216                         MW_CHAR_REFS_REGEX,
1217                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1218                         $text, /* limit */ -1, $count );
1219
1220                 if ( $count ) {
1221                         return $wgContLang->normalize( $text );
1222                 } else {
1223                         return $text;
1224                 }
1225         }
1226
1227         /**
1228          * @param $matches String
1229          * @return String
1230          */
1231         static function decodeCharReferencesCallback( $matches ) {
1232                 if( $matches[1] != '' ) {
1233                         return Sanitizer::decodeEntity( $matches[1] );
1234                 } elseif( $matches[2] != '' ) {
1235                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1236                 } elseif( $matches[3] != ''  ) {
1237                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1238                 }
1239                 # Last case should be an ampersand by itself
1240                 return $matches[0];
1241         }
1242
1243         /**
1244          * Return UTF-8 string for a codepoint if that is a valid
1245          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1246          * @param $codepoint Integer
1247          * @return String
1248          * @private
1249          */
1250         static function decodeChar( $codepoint ) {
1251                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1252                         return codepointToUtf8( $codepoint );
1253                 } else {
1254                         return UTF8_REPLACEMENT;
1255                 }
1256         }
1257
1258         /**
1259          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1260          * return the UTF-8 encoding of that character. Otherwise, returns
1261          * pseudo-entity source (eg &foo;)
1262          *
1263          * @param $name Strings
1264          * @return String
1265          */
1266         static function decodeEntity( $name ) {
1267                 global $wgHtmlEntities, $wgHtmlEntityAliases;
1268                 if ( isset( $wgHtmlEntityAliases[$name] ) ) {
1269                         $name = $wgHtmlEntityAliases[$name];
1270                 }
1271                 if( isset( $wgHtmlEntities[$name] ) ) {
1272                         return codepointToUtf8( $wgHtmlEntities[$name] );
1273                 } else {
1274                         return "&$name;";
1275                 }
1276         }
1277
1278         /**
1279          * Fetch the whitelist of acceptable attributes for a given element name.
1280          *
1281          * @param $element String
1282          * @return Array
1283          */
1284         static function attributeWhitelist( $element ) {
1285                 static $list;
1286                 if( !isset( $list ) ) {
1287                         $list = Sanitizer::setupAttributeWhitelist();
1288                 }
1289                 return isset( $list[$element] )
1290                         ? $list[$element]
1291                         : array();
1292         }
1293
1294         /**
1295          * Foreach array key (an allowed HTML element), return an array
1296          * of allowed attributes
1297          * @return Array
1298          */
1299         static function setupAttributeWhitelist() {
1300                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1301
1302                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1303
1304                 if ( $wgAllowRdfaAttributes ) {
1305                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1306                         $common = array_merge( $common, array(
1307                             'about', 'property', 'resource', 'datatype', 'typeof',
1308                         ) );
1309                 }
1310
1311                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1312                         # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1313                         $common = array_merge( $common, array(
1314                             'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1315                         ) );
1316                 }
1317
1318                 $block = array_merge( $common, array( 'align' ) );
1319                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1320                 $tablecell = array( 'abbr',
1321                                     'axis',
1322                                     'headers',
1323                                     'scope',
1324                                     'rowspan',
1325                                     'colspan',
1326                                     'nowrap', # deprecated
1327                                     'width',  # deprecated
1328                                     'height', # deprecated
1329                                     'bgcolor' # deprecated
1330                                     );
1331
1332                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1333                 # See: http://www.w3.org/TR/html4/
1334                 $whitelist = array (
1335                         # 7.5.4
1336                         'div'        => $block,
1337                         'center'     => $common, # deprecated
1338                         'span'       => $block, # ??
1339
1340                         # 7.5.5
1341                         'h1'         => $block,
1342                         'h2'         => $block,
1343                         'h3'         => $block,
1344                         'h4'         => $block,
1345                         'h5'         => $block,
1346                         'h6'         => $block,
1347
1348                         # 7.5.6
1349                         # address
1350
1351                         # 8.2.4
1352                         # bdo
1353
1354                         # 9.2.1
1355                         'em'         => $common,
1356                         'strong'     => $common,
1357                         'cite'       => $common,
1358                         'dfn'        => $common,
1359                         'code'       => $common,
1360                         'samp'       => $common,
1361                         'kbd'        => $common,
1362                         'var'        => $common,
1363                         'abbr'       => $common,
1364                         # acronym
1365
1366                         # 9.2.2
1367                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1368                         # q
1369
1370                         # 9.2.3
1371                         'sub'        => $common,
1372                         'sup'        => $common,
1373
1374                         # 9.3.1
1375                         'p'          => $block,
1376
1377                         # 9.3.2
1378                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1379
1380                         # 9.3.4
1381                         'pre'        => array_merge( $common, array( 'width' ) ),
1382
1383                         # 9.4
1384                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1385                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1386
1387                         # 10.2
1388                         'ul'         => array_merge( $common, array( 'type' ) ),
1389                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1390                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1391
1392                         # 10.3
1393                         'dl'         => $common,
1394                         'dd'         => $common,
1395                         'dt'         => $common,
1396
1397                         # 11.2.1
1398                         'table'      => array_merge( $common,
1399                                                                 array( 'summary', 'width', 'border', 'frame',
1400                                                                                 'rules', 'cellspacing', 'cellpadding',
1401                                                                                 'align', 'bgcolor',
1402                                                                 ) ),
1403
1404                         # 11.2.2
1405                         'caption'    => array_merge( $common, array( 'align' ) ),
1406
1407                         # 11.2.3
1408                         'thead'      => array_merge( $common, $tablealign ),
1409                         'tfoot'      => array_merge( $common, $tablealign ),
1410                         'tbody'      => array_merge( $common, $tablealign ),
1411
1412                         # 11.2.4
1413                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1414                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1415
1416                         # 11.2.5
1417                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1418
1419                         # 11.2.6
1420                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1421                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1422
1423                         # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1424                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1425
1426                         # 13.2
1427                         # Not usually allowed, but may be used for extension-style hooks
1428                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1429                         # true
1430                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1431
1432                         # 15.2.1
1433                         'tt'         => $common,
1434                         'b'          => $common,
1435                         'i'          => $common,
1436                         'big'        => $common,
1437                         'small'      => $common,
1438                         'strike'     => $common,
1439                         's'          => $common,
1440                         'u'          => $common,
1441
1442                         # 15.2.2
1443                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1444                         # basefont
1445
1446                         # 15.3
1447                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1448
1449                         # XHTML Ruby annotation text module, simple ruby only.
1450                         # http://www.w3c.org/TR/ruby/
1451                         'ruby'       => $common,
1452                         # rbc
1453                         # rtc
1454                         'rb'         => $common,
1455                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1456                         'rp'         => $common,
1457
1458                         # MathML root element, where used for extensions
1459                         # 'title' may not be 100% valid here; it's XHTML
1460                         # http://www.w3.org/TR/REC-MathML/
1461                         'math'       => array( 'class', 'style', 'id', 'title' ),
1462                         );
1463                 return $whitelist;
1464         }
1465
1466         /**
1467          * Take a fragment of (potentially invalid) HTML and return
1468          * a version with any tags removed, encoded as plain text.
1469          *
1470          * Warning: this return value must be further escaped for literal
1471          * inclusion in HTML output as of 1.10!
1472          *
1473          * @param $text String: HTML fragment
1474          * @return String
1475          */
1476         static function stripAllTags( $text ) {
1477                 # Actual <tags>
1478                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1479
1480                 # Normalize &entities and whitespace
1481                 $text = self::decodeCharReferences( $text );
1482                 $text = self::normalizeWhitespace( $text );
1483
1484                 return $text;
1485         }
1486
1487         /**
1488          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1489          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1490          * PHP 5.1 doesn't.
1491          *
1492          * Use for passing XHTML fragments to PHP's XML parsing functions
1493          *
1494          * @return String
1495          */
1496         static function hackDocType() {
1497                 global $wgHtmlEntities;
1498                 $out = "<!DOCTYPE html [\n";
1499                 foreach( $wgHtmlEntities as $entity => $codepoint ) {
1500                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1501                 }
1502                 $out .= "]>\n";
1503                 return $out;
1504         }
1505
1506         static function cleanUrl( $url ) {
1507                 # Normalize any HTML entities in input. They will be
1508                 # re-escaped by makeExternalLink().
1509                 $url = Sanitizer::decodeCharReferences( $url );
1510
1511                 # Escape any control characters introduced by the above step
1512                 $url = preg_replace( '/[\][<>"\\x00-\\x20\\x7F\|]/e', "urlencode('\\0')", $url );
1513
1514                 # Validate hostname portion
1515                 $matches = array();
1516                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1517                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1518
1519                         // Characters that will be ignored in IDNs.
1520                         // http://tools.ietf.org/html/3454#section-3.1
1521                         // Strip them before further processing so blacklists and such work.
1522                         $strip = "/
1523                                 \\s|          # general whitespace
1524                                 \xc2\xad|     # 00ad SOFT HYPHEN
1525                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1526                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1527                                 \xe2\x81\xa0| # 2060 WORD JOINER
1528                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1529                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1530                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1531                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1532                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1533                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1534                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1535                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe00f VARIATION SELECTOR-1-16
1536                                 /xuD";
1537
1538                         $host = preg_replace( $strip, '', $host );
1539
1540                         // @todo Fixme: validate hostnames here
1541
1542                         return $protocol . $host . $rest;
1543                 } else {
1544                         return $url;
1545                 }
1546         }
1547
1548 }