includes/Sanitizer.php

   1 <?php
   2 /**
   3  * XHTML sanitizer for MediaWiki
   4  *
   5  * Copyright (C) 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  * @addtogroup Parser
  24  */
  25
  26 /**
  27  * Regular expression to match various types of character references in
  28  * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
  29  */
  30 define( 'MW_CHAR_REFS_REGEX',
  31         '/&([A-Za-z0-9\x80-\xff]+);
  32          |&\#([0-9]+);
  33          |&\#x([0-9A-Za-z]+);
  34          |&\#X([0-9A-Za-z]+);
  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 $attrib = '[A-Za-z0-9]';
  43 $space = '[\x09\x0a\x0d\x20]';
  44 define( 'MW_ATTRIBS_REGEX',
  45         "/(?:^|$space)($attrib+)
  46           ($space*=$space*
  47                 (?:
  48                  # The attribute value: quoted or alone
  49                   \"([^<\"]*)\"
  50                  | '([^<']*)'
  51                  |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
  52                  |  (\#[0-9a-fA-F]+) # Technically wrong, but lots of
  53                                                          # colors are specified like this.
  54                                                          # We'll be normalizing it.
  55                 )
  56            )?(?=$space|\$)/sx" );
  57
  58 /**
  59  * List of all named character entities defined in HTML 4.01
  60  * http://www.w3.org/TR/html4/sgml/entities.html
  61  * @private
  62  */
  63 global $wgHtmlEntities;
  64 $wgHtmlEntities = array(
  65         'Aacute'   => 193,
  66         'aacute'   => 225,
  67         'Acirc'    => 194,
  68         'acirc'    => 226,
  69         'acute'    => 180,
  70         'AElig'    => 198,
  71         'aelig'    => 230,
  72         'Agrave'   => 192,
  73         'agrave'   => 224,
  74         'alefsym'  => 8501,
  75         'Alpha'    => 913,
  76         'alpha'    => 945,
  77         'amp'      => 38,
  78         'and'      => 8743,
  79         'ang'      => 8736,
  80         'Aring'    => 197,
  81         'aring'    => 229,
  82         'asymp'    => 8776,
  83         'Atilde'   => 195,
  84         'atilde'   => 227,
  85         'Auml'     => 196,
  86         'auml'     => 228,
  87         'bdquo'    => 8222,
  88         'Beta'     => 914,
  89         'beta'     => 946,
  90         'brvbar'   => 166,
  91         'bull'     => 8226,
  92         'cap'      => 8745,
  93         'Ccedil'   => 199,
  94         'ccedil'   => 231,
  95         'cedil'    => 184,
  96         'cent'     => 162,
  97         'Chi'      => 935,
  98         'chi'      => 967,
  99         'circ'     => 710,
 100         'clubs'    => 9827,
 101         'cong'     => 8773,
 102         'copy'     => 169,
 103         'crarr'    => 8629,
 104         'cup'      => 8746,
 105         'curren'   => 164,
 106         'dagger'   => 8224,
 107         'Dagger'   => 8225,
 108         'darr'     => 8595,
 109         'dArr'     => 8659,
 110         'deg'      => 176,
 111         'Delta'    => 916,
 112         'delta'    => 948,
 113         'diams'    => 9830,
 114         'divide'   => 247,
 115         'Eacute'   => 201,
 116         'eacute'   => 233,
 117         'Ecirc'    => 202,
 118         'ecirc'    => 234,
 119         'Egrave'   => 200,
 120         'egrave'   => 232,
 121         'empty'    => 8709,
 122         'emsp'     => 8195,
 123         'ensp'     => 8194,
 124         'Epsilon'  => 917,
 125         'epsilon'  => 949,
 126         'equiv'    => 8801,
 127         'Eta'      => 919,
 128         'eta'      => 951,
 129         'ETH'      => 208,
 130         'eth'      => 240,
 131         'Euml'     => 203,
 132         'euml'     => 235,
 133         'euro'     => 8364,
 134         'exist'    => 8707,
 135         'fnof'     => 402,
 136         'forall'   => 8704,
 137         'frac12'   => 189,
 138         'frac14'   => 188,
 139         'frac34'   => 190,
 140         'frasl'    => 8260,
 141         'Gamma'    => 915,
 142         'gamma'    => 947,
 143         'ge'       => 8805,
 144         'gt'       => 62,
 145         'harr'     => 8596,
 146         'hArr'     => 8660,
 147         'hearts'   => 9829,
 148         'hellip'   => 8230,
 149         'Iacute'   => 205,
 150         'iacute'   => 237,
 151         'Icirc'    => 206,
 152         'icirc'    => 238,
 153         'iexcl'    => 161,
 154         'Igrave'   => 204,
 155         'igrave'   => 236,
 156         'image'    => 8465,
 157         'infin'    => 8734,
 158         'int'      => 8747,
 159         'Iota'     => 921,
 160         'iota'     => 953,
 161         'iquest'   => 191,
 162         'isin'     => 8712,
 163         'Iuml'     => 207,
 164         'iuml'     => 239,
 165         'Kappa'    => 922,
 166         'kappa'    => 954,
 167         'Lambda'   => 923,
 168         'lambda'   => 955,
 169         'lang'     => 9001,
 170         'laquo'    => 171,
 171         'larr'     => 8592,
 172         'lArr'     => 8656,
 173         'lceil'    => 8968,
 174         'ldquo'    => 8220,
 175         'le'       => 8804,
 176         'lfloor'   => 8970,
 177         'lowast'   => 8727,
 178         'loz'      => 9674,
 179         'lrm'      => 8206,
 180         'lsaquo'   => 8249,
 181         'lsquo'    => 8216,
 182         'lt'       => 60,
 183         'macr'     => 175,
 184         'mdash'    => 8212,
 185         'micro'    => 181,
 186         'middot'   => 183,
 187         'minus'    => 8722,
 188         'Mu'       => 924,
 189         'mu'       => 956,
 190         'nabla'    => 8711,
 191         'nbsp'     => 160,
 192         'ndash'    => 8211,
 193         'ne'       => 8800,
 194         'ni'       => 8715,
 195         'not'      => 172,
 196         'notin'    => 8713,
 197         'nsub'     => 8836,
 198         'Ntilde'   => 209,
 199         'ntilde'   => 241,
 200         'Nu'       => 925,
 201         'nu'       => 957,
 202         'Oacute'   => 211,
 203         'oacute'   => 243,
 204         'Ocirc'    => 212,
 205         'ocirc'    => 244,
 206         'OElig'    => 338,
 207         'oelig'    => 339,
 208         'Ograve'   => 210,
 209         'ograve'   => 242,
 210         'oline'    => 8254,
 211         'Omega'    => 937,
 212         'omega'    => 969,
 213         'Omicron'  => 927,
 214         'omicron'  => 959,
 215         'oplus'    => 8853,
 216         'or'       => 8744,
 217         'ordf'     => 170,
 218         'ordm'     => 186,
 219         'Oslash'   => 216,
 220         'oslash'   => 248,
 221         'Otilde'   => 213,
 222         'otilde'   => 245,
 223         'otimes'   => 8855,
 224         'Ouml'     => 214,
 225         'ouml'     => 246,
 226         'para'     => 182,
 227         'part'     => 8706,
 228         'permil'   => 8240,
 229         'perp'     => 8869,
 230         'Phi'      => 934,
 231         'phi'      => 966,
 232         'Pi'       => 928,
 233         'pi'       => 960,
 234         'piv'      => 982,
 235         'plusmn'   => 177,
 236         'pound'    => 163,
 237         'prime'    => 8242,
 238         'Prime'    => 8243,
 239         'prod'     => 8719,
 240         'prop'     => 8733,
 241         'Psi'      => 936,
 242         'psi'      => 968,
 243         'quot'     => 34,
 244         'radic'    => 8730,
 245         'rang'     => 9002,
 246         'raquo'    => 187,
 247         'rarr'     => 8594,
 248         'rArr'     => 8658,
 249         'rceil'    => 8969,
 250         'rdquo'    => 8221,
 251         'real'     => 8476,
 252         'reg'      => 174,
 253         'rfloor'   => 8971,
 254         'Rho'      => 929,
 255         'rho'      => 961,
 256         'rlm'      => 8207,
 257         'rsaquo'   => 8250,
 258         'rsquo'    => 8217,
 259         'sbquo'    => 8218,
 260         'Scaron'   => 352,
 261         'scaron'   => 353,
 262         'sdot'     => 8901,
 263         'sect'     => 167,
 264         'shy'      => 173,
 265         'Sigma'    => 931,
 266         'sigma'    => 963,
 267         'sigmaf'   => 962,
 268         'sim'      => 8764,
 269         'spades'   => 9824,
 270         'sub'      => 8834,
 271         'sube'     => 8838,
 272         'sum'      => 8721,
 273         'sup'      => 8835,
 274         'sup1'     => 185,
 275         'sup2'     => 178,
 276         'sup3'     => 179,
 277         'supe'     => 8839,
 278         'szlig'    => 223,
 279         'Tau'      => 932,
 280         'tau'      => 964,
 281         'there4'   => 8756,
 282         'Theta'    => 920,
 283         'theta'    => 952,
 284         'thetasym' => 977,
 285         'thinsp'   => 8201,
 286         'THORN'    => 222,
 287         'thorn'    => 254,
 288         'tilde'    => 732,
 289         'times'    => 215,
 290         'trade'    => 8482,
 291         'Uacute'   => 218,
 292         'uacute'   => 250,
 293         'uarr'     => 8593,
 294         'uArr'     => 8657,
 295         'Ucirc'    => 219,
 296         'ucirc'    => 251,
 297         'Ugrave'   => 217,
 298         'ugrave'   => 249,
 299         'uml'      => 168,
 300         'upsih'    => 978,
 301         'Upsilon'  => 933,
 302         'upsilon'  => 965,
 303         'Uuml'     => 220,
 304         'uuml'     => 252,
 305         'weierp'   => 8472,
 306         'Xi'       => 926,
 307         'xi'       => 958,
 308         'Yacute'   => 221,
 309         'yacute'   => 253,
 310         'yen'      => 165,
 311         'Yuml'     => 376,
 312         'yuml'     => 255,
 313         'Zeta'     => 918,
 314         'zeta'     => 950,
 315         'zwj'      => 8205,
 316         'zwnj'     => 8204 );
 317
 318 /**
 319  * Character entity aliases accepted by MediaWiki
 320  */
 321 global $wgHtmlEntityAliases;
 322 $wgHtmlEntityAliases = array(
 323         'רלמ' => 'rlm',
 324         'رلم' => 'rlm',
 325 );
 326
 327
 328 /**
 329  * XHTML sanitizer for MediaWiki
 330  * @addtogroup Parser
 331  */
 332 class Sanitizer {
 333         const NONE = 0;
 334         const INITIAL_NONLETTER = 1;
 335
 336         /**
 337          * Cleans up HTML, removes dangerous tags and attributes, and
 338          * removes HTML comments
 339          * @private
 340          * @param string $text
 341          * @param callback $processCallback to do any variable or parameter replacements in HTML attribute values
 342          * @param array $args for the processing callback
 343          * @return string
 344          */
 345         static function removeHTMLtags( $text, $processCallback = null, $args = array(), $extratags = array() ) {
 346                 global $wgUseTidy;
 347
 348                 static $htmlpairs, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 349                         $htmllist, $listtags, $htmlsingleallowed, $htmlelements, $staticInitialised;
 350
 351                 wfProfileIn( __METHOD__ );
 352
 353                 if ( !$staticInitialised ) {
 354
 355                         $htmlpairs = array_merge( $extratags, array( # Tags that must be closed
 356                                 'b', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 357                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 358                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 359                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 360                                 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'u'
 361                         ) );
 362                         $htmlsingle = array(
 363                                 'br', 'hr', 'li', 'dt', 'dd'
 364                         );
 365                         $htmlsingleonly = array( # Elements that cannot have close tags
 366                                 'br', 'hr'
 367                         );
 368                         $htmlnest = array( # Tags that can be nested--??
 369                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 370                                 'dl', 'font', 'big', 'small', 'sub', 'sup', 'span'
 371                         );
 372                         $tabletags = array( # Can only appear inside table, we will close them
 373                                 'td', 'th', 'tr',
 374                         );
 375                         $htmllist = array( # Tags used by list
 376                                 'ul','ol',
 377                         );
 378                         $listtags = array( # Tags that can appear in a list
 379                                 'li',
 380                         );
 381
 382                         $htmlsingleallowed = array_merge( $htmlsingle, $tabletags );
 383                         $htmlelements = array_merge( $htmlsingle, $htmlpairs, $htmlnest );
 384
 385                         # Convert them all to hashtables for faster lookup
 386                         $vars = array( 'htmlpairs', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 387                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelements' );
 388                         foreach ( $vars as $var ) {
 389                                 $$var = array_flip( $$var );
 390                         }
 391                         $staticInitialised = true;
 392                 }
 393
 394                 # Remove HTML comments
 395                 $text = Sanitizer::removeHTMLcomments( $text );
 396                 $bits = explode( '<', $text );
 397                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 398                 if(!$wgUseTidy) {
 399                         $tagstack = $tablestack = array();
 400                         foreach ( $bits as $x ) {
 401                                 $regs = array();
 402                                 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 403                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 404                                 } else {
 405                                         $slash = $t = $params = $brace = $rest = null;
 406                                 }
 407
 408                                 $badtag = 0 ;
 409                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 410                                         # Check our stack
 411                                         if ( $slash ) {
 412                                                 # Closing a tag...
 413                                                 if( isset( $htmlsingleonly[$t] ) ) {
 414                                                         $badtag = 1;
 415                                                 } elseif ( ( $ot = @array_pop( $tagstack ) ) != $t ) {
 416                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 417                                                                 # Pop all elements with an optional close tag
 418                                                                 # and see if we find a match below them
 419                                                                 $optstack = array();
 420                                                                 array_push ($optstack, $ot);
 421                                                                 while ( ( ( $ot = @array_pop( $tagstack ) ) != $t ) &&
 422                                                                                 isset( $htmlsingleallowed[$ot] ) )
 423                                                                 {
 424                                                                         array_push ($optstack, $ot);
 425                                                                 }
 426                                                                 if ( $t != $ot ) {
 427                                                                         # No match. Push the optinal elements back again
 428                                                                         $badtag = 1;
 429                                                                         while ( $ot = @array_pop( $optstack ) ) {
 430                                                                                 array_push( $tagstack, $ot );
 431                                                                         }
 432                                                                 }
 433                                                         } else {
 434                                                                 @array_push( $tagstack, $ot );
 435                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 436                                                                 if(!(isset( $htmllist[$ot] ) && isset( $listtags[$t] ) )) {
 437                                                                         $badtag = 1;
 438                                                                 }
 439                                                         }
 440                                                 } else {
 441                                                         if ( $t == 'table' ) {
 442                                                                 $tagstack = array_pop( $tablestack );
 443                                                         }
 444                                                 }
 445                                                 $newparams = '';
 446                                         } else {
 447                                                 # Keep track for later
 448                                                 if ( isset( $tabletags[$t] ) &&
 449                                                 ! in_array( 'table', $tagstack ) ) {
 450                                                         $badtag = 1;
 451                                                 } else if ( in_array( $t, $tagstack ) &&
 452                                                 ! isset( $htmlnest [$t ] ) ) {
 453                                                         $badtag = 1 ;
 454                                                 # Is it a self closed htmlpair ? (bug 5487)
 455                                                 } else if( $brace == '/>' &&
 456                                                 isset( $htmlpairs[$t] ) ) {
 457                                                         $badtag = 1;
 458                                                 } elseif( isset( $htmlsingleonly[$t] ) ) {
 459                                                         # Hack to force empty tag for uncloseable elements
 460                                                         $brace = '/>';
 461                                                 } else if( isset( $htmlsingle[$t] ) ) {
 462                                                         # Hack to not close $htmlsingle tags
 463                                                         $brace = NULL;
 464                                                 } else if( isset( $tabletags[$t] )
 465                                                 &&  in_array($t ,$tagstack) ) {
 466                                                         // New table tag but forgot to close the previous one
 467                                                         $text .= "</$t>";
 468                                                 } else {
 469                                                         if ( $t == 'table' ) {
 470                                                                 array_push( $tablestack, $tagstack );
 471                                                                 $tagstack = array();
 472                                                         }
 473                                                         array_push( $tagstack, $t );
 474                                                 }
 475
 476                                                 # Replace any variables or template parameters with
 477                                                 # plaintext results.
 478                                                 if( is_callable( $processCallback ) ) {
 479                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 480                                                 }
 481
 482                                                 # Strip non-approved attributes from the tag
 483                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 484                                         }
 485                                         if ( ! $badtag ) {
 486                                                 $rest = str_replace( '>', '&gt;', $rest );
 487                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 488                                                 $text .= "<$slash$t$newparams$close>$rest";
 489                                                 continue;
 490                                         }
 491                                 }
 492                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 493                         }
 494                         # Close off any remaining tags
 495                         while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
 496                                 $text .= "</$t>\n";
 497                                 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
 498                         }
 499                 } else {
 500                         # this might be possible using tidy itself
 501                         foreach ( $bits as $x ) {
 502                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 503                                 $x, $regs );
 504                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 505                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 506                                         if( is_callable( $processCallback ) ) {
 507                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 508                                         }
 509                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 510                                         $rest = str_replace( '>', '&gt;', $rest );
 511                                         $text .= "<$slash$t$newparams$brace$rest";
 512                                 } else {
 513                                         $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 514                                 }
 515                         }
 516                 }
 517                 wfProfileOut( __METHOD__ );
 518                 return $text;
 519         }
 520
 521         /**
 522          * Remove '<!--', '-->', and everything between.
 523          * To avoid leaving blank lines, when a comment is both preceded
 524          * and followed by a newline (ignoring spaces), trim leading and
 525          * trailing spaces and one of the newlines.
 526          *
 527          * @private
 528          * @param string $text
 529          * @return string
 530          */
 531         static function removeHTMLcomments( $text ) {
 532                 wfProfileIn( __METHOD__ );
 533                 while (($start = strpos($text, '<!--')) !== false) {
 534                         $end = strpos($text, '-->', $start + 4);
 535                         if ($end === false) {
 536                                 # Unterminated comment; bail out
 537                                 break;
 538                         }
 539
 540                         $end += 3;
 541
 542                         # Trim space and newline if the comment is both
 543                         # preceded and followed by a newline
 544                         $spaceStart = max($start - 1, 0);
 545                         $spaceLen = $end - $spaceStart;
 546                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 547                                 $spaceStart--;
 548                                 $spaceLen++;
 549                         }
 550                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 551                                 $spaceLen++;
 552                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 553                                 # Remove the comment, leading and trailing
 554                                 # spaces, and leave only one newline.
 555                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 556                         }
 557                         else {
 558                                 # Remove just the comment.
 559                                 $text = substr_replace($text, '', $start, $end - $start);
 560                         }
 561                 }
 562                 wfProfileOut( __METHOD__ );
 563                 return $text;
 564         }
 565
 566         /**
 567          * Take an array of attribute names and values and normalize or discard
 568          * illegal values for the given element type.
 569          *
 570          * - Discards attributes not on a whitelist for the given element
 571          * - Unsafe style attributes are discarded
 572          * - Invalid id attributes are reencoded
 573          *
 574          * @param array $attribs
 575          * @param string $element
 576          * @return array
 577          *
 578          * @todo Check for legal values where the DTD limits things.
 579          * @todo Check for unique id attribute :P
 580          */
 581         static function validateTagAttributes( $attribs, $element ) {
 582                 return Sanitizer::validateAttributes( $attribs,
 583                         Sanitizer::attributeWhitelist( $element ) );
 584         }
 585
 586         /**
 587          * Take an array of attribute names and values and normalize or discard
 588          * illegal values for the given whitelist.
 589          *
 590          * - Discards attributes not the given whitelist
 591          * - Unsafe style attributes are discarded
 592          * - Invalid id attributes are reencoded
 593          *
 594          * @param array $attribs
 595          * @param array $whitelist list of allowed attribute names
 596          * @return array
 597          *
 598          * @todo Check for legal values where the DTD limits things.
 599          * @todo Check for unique id attribute :P
 600          */
 601         static function validateAttributes( $attribs, $whitelist ) {
 602                 $whitelist = array_flip( $whitelist );
 603                 $out = array();
 604                 foreach( $attribs as $attribute => $value ) {
 605                         if( !isset( $whitelist[$attribute] ) ) {
 606                                 continue;
 607                         }
 608                         # Strip javascript "expression" from stylesheets.
 609                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 610                         if( $attribute == 'style' ) {
 611                                 $value = Sanitizer::checkCss( $value );
 612                                 if( $value === false ) {
 613                                         # haxx0r
 614                                         continue;
 615                                 }
 616                         }
 617
 618                         if ( $attribute === 'id' )
 619                                 $value = Sanitizer::escapeId( $value );
 620
 621                         // If this attribute was previously set, override it.
 622                         // Output should only have one attribute of each name.
 623                         $out[$attribute] = $value;
 624                 }
 625                 return $out;
 626         }
 627
 628         /**
 629          * Merge two sets of HTML attributes.
 630          * Conflicting items in the second set will override those
 631          * in the first, except for 'class' attributes which will be
 632          * combined.
 633          *
 634          * @todo implement merging for other attributes such as style
 635          * @param array $a
 636          * @param array $b
 637          * @return array
 638          */
 639         static function mergeAttributes( $a, $b ) {
 640                 $out = array_merge( $a, $b );
 641                 if( isset( $a['class'] )
 642                         && isset( $b['class'] )
 643                         && $a['class'] !== $b['class'] ) {
 644
 645                         $out['class'] = implode( ' ',
 646                                 array_unique(
 647                                         preg_split( '/\s+/',
 648                                                 $a['class'] . ' ' . $b['class'],
 649                                                 -1,
 650                                                 PREG_SPLIT_NO_EMPTY ) ) );
 651                 }
 652                 return $out;
 653         }
 654
 655         /**
 656          * Pick apart some CSS and check it for forbidden or unsafe structures.
 657          * Returns a sanitized string, or false if it was just too evil.
 658          *
 659          * Currently URL references, 'expression', 'tps' are forbidden.
 660          *
 661          * @param string $value
 662          * @return mixed
 663          */
 664         static function checkCss( $value ) {
 665                 $stripped = Sanitizer::decodeCharReferences( $value );
 666
 667                 // Remove any comments; IE gets token splitting wrong
 668                 $stripped = StringUtils::delimiterReplace( '/*', '*/', ' ', $stripped );
 669
 670                 $value = $stripped;
 671
 672                 // ... and continue checks
 673                 $stripped = preg_replace( '!\\\\([0-9A-Fa-f]{1,6})[ \\n\\r\\t\\f]?!e',
 674                         'codepointToUtf8(hexdec("$1"))', $stripped );
 675                 $stripped = str_replace( '\\', '', $stripped );
 676                 if( preg_match( '/(?:expression|tps*:\/\/|url\\s*\().*/is',
 677                                 $stripped ) ) {
 678                         # haxx0r
 679                         return false;
 680                 }
 681
 682                 return $value;
 683         }
 684
 685         /**
 686          * Take a tag soup fragment listing an HTML element's attributes
 687          * and normalize it to well-formed XML, discarding unwanted attributes.
 688          * Output is safe for further wikitext processing, with escaping of
 689          * values that could trigger problems.
 690          *
 691          * - Normalizes attribute names to lowercase
 692          * - Discards attributes not on a whitelist for the given element
 693          * - Turns broken or invalid entities into plaintext
 694          * - Double-quotes all attribute values
 695          * - Attributes without values are given the name as attribute
 696          * - Double attributes are discarded
 697          * - Unsafe style attributes are discarded
 698          * - Prepends space if there are attributes.
 699          *
 700          * @param string $text
 701          * @param string $element
 702          * @return string
 703          */
 704         static function fixTagAttributes( $text, $element ) {
 705                 if( trim( $text ) == '' ) {
 706                         return '';
 707                 }
 708
 709                 $stripped = Sanitizer::validateTagAttributes(
 710                         Sanitizer::decodeTagAttributes( $text ), $element );
 711
 712                 $attribs = array();
 713                 foreach( $stripped as $attribute => $value ) {
 714                         $encAttribute = htmlspecialchars( $attribute );
 715                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 716
 717                         $attribs[] = "$encAttribute=\"$encValue\"";
 718                 }
 719                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 720         }
 721
 722         /**
 723          * Encode an attribute value for HTML output.
 724          * @param $text
 725          * @return HTML-encoded text fragment
 726          */
 727         static function encodeAttribute( $text ) {
 728                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 729
 730                 // Whitespace is normalized during attribute decoding,
 731                 // so if we've been passed non-spaces we must encode them
 732                 // ahead of time or they won't be preserved.
 733                 $encValue = strtr( $encValue, array(
 734                         "\n" => '&#10;',
 735                         "\r" => '&#13;',
 736                         "\t" => '&#9;',
 737                 ) );
 738
 739                 return $encValue;
 740         }
 741
 742         /**
 743          * Encode an attribute value for HTML tags, with extra armoring
 744          * against further wiki processing.
 745          * @param $text
 746          * @return HTML-encoded text fragment
 747          */
 748         static function safeEncodeAttribute( $text ) {
 749                 $encValue = Sanitizer::encodeAttribute( $text );
 750
 751                 # Templates and links may be expanded in later parsing,
 752                 # creating invalid or dangerous output. Suppress this.
 753                 $encValue = strtr( $encValue, array(
 754                         '<'    => '&lt;',   // This should never happen,
 755                         '>'    => '&gt;',   // we've received invalid input
 756                         '"'    => '&quot;', // which should have been escaped.
 757                         '{'    => '&#123;',
 758                         '['    => '&#91;',
 759                         "''"   => '&#39;&#39;',
 760                         'ISBN' => '&#73;SBN',
 761                         'RFC'  => '&#82;FC',
 762                         'PMID' => '&#80;MID',
 763                         '|'    => '&#124;',
 764                         '__'   => '&#95;_',
 765                 ) );
 766
 767                 # Stupid hack
 768                 $encValue = preg_replace_callback(
 769                         '/(' . wfUrlProtocols() . ')/',
 770                         array( 'Sanitizer', 'armorLinksCallback' ),
 771                         $encValue );
 772                 return $encValue;
 773         }
 774
 775         /**
 776          * Given a value escape it so that it can be used in an id attribute and
 777          * return it, this does not validate the value however (see first link)
 778          *
 779          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
 780          *                                                          in the id and
 781          *                                                          name attributes
 782          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
 783          *
 784          * @param string $id    Id to validate
 785          * @param int    $flags Currently only two values: Sanitizer::INITIAL_NONLETTER
 786          *                      (default) permits initial non-letter characters,
 787          *                      such as if you're adding a prefix to them.
 788          *                      Sanitizer::NONE will prepend an 'x' if the id
 789          *                      would otherwise start with a nonletter.
 790          * @return string
 791          */
 792         static function escapeId( $id, $flags = Sanitizer::INITIAL_NONLETTER ) {
 793                 static $replace = array(
 794                         '%3A' => ':',
 795                         '%' => '.'
 796                 );
 797
 798                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
 799                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
 800
 801                 if( ~$flags & Sanitizer::INITIAL_NONLETTER
 802                 && !preg_match( '/[a-zA-Z]/', $id[0] ) ) {
 803                         // Initial character must be a letter!
 804                         $id = "x$id";
 805                 }
 806                 return $id;
 807         }
 808
 809         /**
 810          * Given a value, escape it so that it can be used as a CSS class and
 811          * return it.
 812          *
 813          * @todo For extra validity, input should be validated UTF-8.
 814          *
 815          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
 816          *
 817          * @param string $class
 818          * @return string
 819          */
 820         static function escapeClass( $class ) {
 821                 // Convert ugly stuff to underscores and kill underscores in ugly places
 822                 return rtrim(preg_replace(
 823                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
 824                         '_',
 825                         $class ), '_');
 826         }
 827
 828         /**
 829          * Regex replace callback for armoring links against further processing.
 830          * @param array $matches
 831          * @return string
 832          * @private
 833          */
 834         private static function armorLinksCallback( $matches ) {
 835                 return str_replace( ':', '&#58;', $matches[1] );
 836         }
 837
 838         /**
 839          * Return an associative array of attribute names and values from
 840          * a partial tag string. Attribute names are forces to lowercase,
 841          * character references are decoded to UTF-8 text.
 842          *
 843          * @param string
 844          * @return array
 845          */
 846         static function decodeTagAttributes( $text ) {
 847                 $attribs = array();
 848
 849                 if( trim( $text ) == '' ) {
 850                         return $attribs;
 851                 }
 852
 853                 $pairs = array();
 854                 if( !preg_match_all(
 855                         MW_ATTRIBS_REGEX,
 856                         $text,
 857                         $pairs,
 858                         PREG_SET_ORDER ) ) {
 859                         return $attribs;
 860                 }
 861
 862                 foreach( $pairs as $set ) {
 863                         $attribute = strtolower( $set[1] );
 864                         $value = Sanitizer::getTagAttributeCallback( $set );
 865
 866                         // Normalize whitespace
 867                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
 868                         $value = trim( $value );
 869
 870                         // Decode character references
 871                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
 872                 }
 873                 return $attribs;
 874         }
 875
 876         /**
 877          * Pick the appropriate attribute value from a match set from the
 878          * MW_ATTRIBS_REGEX matches.
 879          *
 880          * @param array $set
 881          * @return string
 882          * @private
 883          */
 884         private static function getTagAttributeCallback( $set ) {
 885                 if( isset( $set[6] ) ) {
 886                         # Illegal #XXXXXX color with no quotes.
 887                         return $set[6];
 888                 } elseif( isset( $set[5] ) ) {
 889                         # No quotes.
 890                         return $set[5];
 891                 } elseif( isset( $set[4] ) ) {
 892                         # Single-quoted
 893                         return $set[4];
 894                 } elseif( isset( $set[3] ) ) {
 895                         # Double-quoted
 896                         return $set[3];
 897                 } elseif( !isset( $set[2] ) ) {
 898                         # In XHTML, attributes must have a value.
 899                         # For 'reduced' form, return explicitly the attribute name here.
 900                         return $set[1];
 901                 } else {
 902                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
 903                 }
 904         }
 905
 906         /**
 907          * Normalize whitespace and character references in an XML source-
 908          * encoded text for an attribute value.
 909          *
 910          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
 911          * but note that we're not returning the value, but are returning
 912          * XML source fragments that will be slapped into output.
 913          *
 914          * @param string $text
 915          * @return string
 916          * @private
 917          */
 918         private static function normalizeAttributeValue( $text ) {
 919                 return str_replace( '"', '&quot;',
 920                         self::normalizeWhitespace(
 921                                 Sanitizer::normalizeCharReferences( $text ) ) );
 922         }
 923
 924         private static function normalizeWhitespace( $text ) {
 925                 return preg_replace(
 926                         '/\r\n|[\x20\x0d\x0a\x09]/',
 927                         ' ',
 928                         $text );
 929         }
 930
 931         /**
 932          * Ensure that any entities and character references are legal
 933          * for XML and XHTML specifically. Any stray bits will be
 934          * &amp;-escaped to result in a valid text fragment.
 935          *
 936          * a. any named char refs must be known in XHTML
 937          * b. any numeric char refs must be legal chars, not invalid or forbidden
 938          * c. use &#x, not &#X
 939          * d. fix or reject non-valid attributes
 940          *
 941          * @param string $text
 942          * @return string
 943          * @private
 944          */
 945         static function normalizeCharReferences( $text ) {
 946                 return preg_replace_callback(
 947                         MW_CHAR_REFS_REGEX,
 948                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
 949                         $text );
 950         }
 951         /**
 952          * @param string $matches
 953          * @return string
 954          */
 955         static function normalizeCharReferencesCallback( $matches ) {
 956                 $ret = null;
 957                 if( $matches[1] != '' ) {
 958                         $ret = Sanitizer::normalizeEntity( $matches[1] );
 959                 } elseif( $matches[2] != '' ) {
 960                         $ret = Sanitizer::decCharReference( $matches[2] );
 961                 } elseif( $matches[3] != ''  ) {
 962                         $ret = Sanitizer::hexCharReference( $matches[3] );
 963                 } elseif( $matches[4] != '' ) {
 964                         $ret = Sanitizer::hexCharReference( $matches[4] );
 965                 }
 966                 if( is_null( $ret ) ) {
 967                         return htmlspecialchars( $matches[0] );
 968                 } else {
 969                         return $ret;
 970                 }
 971         }
 972
 973         /**
 974          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
 975          * return the named entity reference as is. If the entity is a
 976          * MediaWiki-specific alias, returns the HTML equivalent. Otherwise,
 977          * returns HTML-escaped text of pseudo-entity source (eg &amp;foo;)
 978          *
 979          * @param string $name
 980          * @return string
 981          * @static
 982          */
 983         static function normalizeEntity( $name ) {
 984                 global $wgHtmlEntities, $wgHtmlEntityAliases;
 985                 if ( isset( $wgHtmlEntityAliases[$name] ) ) {
 986                         return "&{$wgHtmlEntityAliases[$name]};";
 987                 } elseif( isset( $wgHtmlEntities[$name] ) ) {
 988                         return "&$name;";
 989                 } else {
 990                         return "&amp;$name;";
 991                 }
 992         }
 993
 994         static function decCharReference( $codepoint ) {
 995                 $point = intval( $codepoint );
 996                 if( Sanitizer::validateCodepoint( $point ) ) {
 997                         return sprintf( '&#%d;', $point );
 998                 } else {
 999                         return null;
1000                 }
1001         }
1002
1003         static function hexCharReference( $codepoint ) {
1004                 $point = hexdec( $codepoint );
1005                 if( Sanitizer::validateCodepoint( $point ) ) {
1006                         return sprintf( '&#x%x;', $point );
1007                 } else {
1008                         return null;
1009                 }
1010         }
1011
1012         /**
1013          * Returns true if a given Unicode codepoint is a valid character in XML.
1014          * @param int $codepoint
1015          * @return bool
1016          */
1017         private static function validateCodepoint( $codepoint ) {
1018                 return ($codepoint ==    0x09)
1019                         || ($codepoint ==    0x0a)
1020                         || ($codepoint ==    0x0d)
1021                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1022                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1023                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1024         }
1025
1026         /**
1027          * Decode any character references, numeric or named entities,
1028          * in the text and return a UTF-8 string.
1029          *
1030          * @param string $text
1031          * @return string
1032          * @public
1033          * @static
1034          */
1035         public static function decodeCharReferences( $text ) {
1036                 return preg_replace_callback(
1037                         MW_CHAR_REFS_REGEX,
1038                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1039                         $text );
1040         }
1041
1042         /**
1043          * @param string $matches
1044          * @return string
1045          */
1046         static function decodeCharReferencesCallback( $matches ) {
1047                 if( $matches[1] != '' ) {
1048                         return Sanitizer::decodeEntity( $matches[1] );
1049                 } elseif( $matches[2] != '' ) {
1050                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1051                 } elseif( $matches[3] != ''  ) {
1052                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1053                 } elseif( $matches[4] != '' ) {
1054                         return  Sanitizer::decodeChar( hexdec( $matches[4] ) );
1055                 }
1056                 # Last case should be an ampersand by itself
1057                 return $matches[0];
1058         }
1059
1060         /**
1061          * Return UTF-8 string for a codepoint if that is a valid
1062          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1063          * @param int $codepoint
1064          * @return string
1065          * @private
1066          */
1067         static function decodeChar( $codepoint ) {
1068                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1069                         return codepointToUtf8( $codepoint );
1070                 } else {
1071                         return UTF8_REPLACEMENT;
1072                 }
1073         }
1074
1075         /**
1076          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1077          * return the UTF-8 encoding of that character. Otherwise, returns
1078          * pseudo-entity source (eg &foo;)
1079          *
1080          * @param string $name
1081          * @return string
1082          */
1083         static function decodeEntity( $name ) {
1084                 global $wgHtmlEntities, $wgHtmlEntityAliases;
1085                 if ( isset( $wgHtmlEntityAliases[$name] ) ) {
1086                         $name = $wgHtmlEntityAliases[$name];
1087                 }
1088                 if( isset( $wgHtmlEntities[$name] ) ) {
1089                         return codepointToUtf8( $wgHtmlEntities[$name] );
1090                 } else {
1091                         return "&$name;";
1092                 }
1093         }
1094
1095         /**
1096          * Fetch the whitelist of acceptable attributes for a given
1097          * element name.
1098          *
1099          * @param string $element
1100          * @return array
1101          */
1102         static function attributeWhitelist( $element ) {
1103                 static $list;
1104                 if( !isset( $list ) ) {
1105                         $list = Sanitizer::setupAttributeWhitelist();
1106                 }
1107                 return isset( $list[$element] )
1108                         ? $list[$element]
1109                         : array();
1110         }
1111
1112         /**
1113          * @todo Document it a bit
1114          * @return array
1115          */
1116         static function setupAttributeWhitelist() {
1117                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1118                 $block = array_merge( $common, array( 'align' ) );
1119                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1120                 $tablecell = array( 'abbr',
1121                                     'axis',
1122                                     'headers',
1123                                     'scope',
1124                                     'rowspan',
1125                                     'colspan',
1126                                     'nowrap', # deprecated
1127                                     'width',  # deprecated
1128                                     'height', # deprecated
1129                                     'bgcolor' # deprecated
1130                                     );
1131
1132                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1133                 # See: http://www.w3.org/TR/html4/
1134                 $whitelist = array (
1135                         # 7.5.4
1136                         'div'        => $block,
1137                         'center'     => $common, # deprecated
1138                         'span'       => $block, # ??
1139
1140                         # 7.5.5
1141                         'h1'         => $block,
1142                         'h2'         => $block,
1143                         'h3'         => $block,
1144                         'h4'         => $block,
1145                         'h5'         => $block,
1146                         'h6'         => $block,
1147
1148                         # 7.5.6
1149                         # address
1150
1151                         # 8.2.4
1152                         # bdo
1153
1154                         # 9.2.1
1155                         'em'         => $common,
1156                         'strong'     => $common,
1157                         'cite'       => $common,
1158                         # dfn
1159                         'code'       => $common,
1160                         # samp
1161                         # kbd
1162                         'var'        => $common,
1163                         # abbr
1164                         # acronym
1165
1166                         # 9.2.2
1167                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1168                         # q
1169
1170                         # 9.2.3
1171                         'sub'        => $common,
1172                         'sup'        => $common,
1173
1174                         # 9.3.1
1175                         'p'          => $block,
1176
1177                         # 9.3.2
1178                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1179
1180                         # 9.3.4
1181                         'pre'        => array_merge( $common, array( 'width' ) ),
1182
1183                         # 9.4
1184                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1185                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1186
1187                         # 10.2
1188                         'ul'         => array_merge( $common, array( 'type' ) ),
1189                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1190                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1191
1192                         # 10.3
1193                         'dl'         => $common,
1194                         'dd'         => $common,
1195                         'dt'         => $common,
1196
1197                         # 11.2.1
1198                         'table'      => array_merge( $common,
1199                                                                 array( 'summary', 'width', 'border', 'frame',
1200                                                                                 'rules', 'cellspacing', 'cellpadding',
1201                                                                                 'align', 'bgcolor',
1202                                                                 ) ),
1203
1204                         # 11.2.2
1205                         'caption'    => array_merge( $common, array( 'align' ) ),
1206
1207                         # 11.2.3
1208                         'thead'      => array_merge( $common, $tablealign ),
1209                         'tfoot'      => array_merge( $common, $tablealign ),
1210                         'tbody'      => array_merge( $common, $tablealign ),
1211
1212                         # 11.2.4
1213                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1214                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1215
1216                         # 11.2.5
1217                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1218
1219                         # 11.2.6
1220                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1221                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1222
1223                         # 13.2
1224                         # Not usually allowed, but may be used for extension-style hooks
1225                         # such as <math> when it is rasterized
1226                         'img'        => array_merge( $common, array( 'alt' ) ),
1227
1228                         # 15.2.1
1229                         'tt'         => $common,
1230                         'b'          => $common,
1231                         'i'          => $common,
1232                         'big'        => $common,
1233                         'small'      => $common,
1234                         'strike'     => $common,
1235                         's'          => $common,
1236                         'u'          => $common,
1237
1238                         # 15.2.2
1239                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1240                         # basefont
1241
1242                         # 15.3
1243                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1244
1245                         # XHTML Ruby annotation text module, simple ruby only.
1246                         # http://www.w3c.org/TR/ruby/
1247                         'ruby'       => $common,
1248                         # rbc
1249                         # rtc
1250                         'rb'         => $common,
1251                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1252                         'rp'         => $common,
1253
1254                         # MathML root element, where used for extensions
1255                         # 'title' may not be 100% valid here; it's XHTML
1256                         # http://www.w3.org/TR/REC-MathML/
1257                         'math'       => array( 'class', 'style', 'id', 'title' ),
1258                         );
1259                 return $whitelist;
1260         }
1261
1262         /**
1263          * Take a fragment of (potentially invalid) HTML and return
1264          * a version with any tags removed, encoded as plain text.
1265          *
1266          * Warning: this return value must be further escaped for literal
1267          * inclusion in HTML output as of 1.10!
1268          *
1269          * @param string $text HTML fragment
1270          * @return string
1271          */
1272         static function stripAllTags( $text ) {
1273                 # Actual <tags>
1274                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1275
1276                 # Normalize &entities and whitespace
1277                 $text = self::decodeCharReferences( $text );
1278                 $text = self::normalizeWhitespace( $text );
1279
1280                 return $text;
1281         }
1282
1283         /**
1284          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1285          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1286          * PHP 5.1 doesn't.
1287          *
1288          * Use for passing XHTML fragments to PHP's XML parsing functions
1289          *
1290          * @return string
1291          * @static
1292          */
1293         static function hackDocType() {
1294                 global $wgHtmlEntities;
1295                 $out = "<!DOCTYPE html [\n";
1296                 foreach( $wgHtmlEntities as $entity => $codepoint ) {
1297                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1298                 }
1299                 $out .= "]>\n";
1300                 return $out;
1301         }
1302
1303         static function cleanUrl( $url, $hostname=true ) {
1304                 # Normalize any HTML entities in input. They will be
1305                 # re-escaped by makeExternalLink().
1306                 $url = Sanitizer::decodeCharReferences( $url );
1307
1308                 # Escape any control characters introduced by the above step
1309                 $url = preg_replace( '/[\][<>"\\x00-\\x20\\x7F]/e', "urlencode('\\0')", $url );
1310
1311                 # Validate hostname portion
1312                 $matches = array();
1313                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1314                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1315
1316                         // Characters that will be ignored in IDNs.
1317                         // http://tools.ietf.org/html/3454#section-3.1
1318                         // Strip them before further processing so blacklists and such work.
1319                         $strip = "/
1320                                 \\s|          # general whitespace
1321                                 \xc2\xad|     # 00ad SOFT HYPHEN
1322                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1323                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1324                                 \xe2\x81\xa0| # 2060 WORD JOINER
1325                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1326                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1327                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1328                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1329                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1330                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1331                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1332                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe00f VARIATION SELECTOR-1-16
1333                                 /xuD";
1334
1335                         $host = preg_replace( $strip, '', $host );
1336
1337                         // @fixme: validate hostnames here
1338
1339                         return $protocol . $host . $rest;
1340                 } else {
1341                         return $url;
1342                 }
1343         }
1344
1345 }
1346
1347