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  * @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          |&\#x([0-9A-Za-z]+);
  35          |&\#X([0-9A-Za-z]+);
  36          |(&)/x' );
  37
  38 /**
  39  * Regular expression to match HTML/XML attribute pairs within a tag.
  40  * Allows some... latitude.
  41  * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
  42  */
  43 $attrib = '[A-Za-z0-9]';
  44 $space = '[\x09\x0a\x0d\x20]';
  45 define( 'MW_ATTRIBS_REGEX',
  46         "/(?:^|$space)($attrib+)
  47           ($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  * List of all named character entities defined in HTML 4.01
  61  * http://www.w3.org/TR/html4/sgml/entities.html
  62  * @private
  63  */
  64 global $wgHtmlEntities;
  65 $wgHtmlEntities = array(
  66         'Aacute'   => 193,
  67         'aacute'   => 225,
  68         'Acirc'    => 194,
  69         'acirc'    => 226,
  70         'acute'    => 180,
  71         'AElig'    => 198,
  72         'aelig'    => 230,
  73         'Agrave'   => 192,
  74         'agrave'   => 224,
  75         'alefsym'  => 8501,
  76         'Alpha'    => 913,
  77         'alpha'    => 945,
  78         'amp'      => 38,
  79         'and'      => 8743,
  80         'ang'      => 8736,
  81         'Aring'    => 197,
  82         'aring'    => 229,
  83         'asymp'    => 8776,
  84         'Atilde'   => 195,
  85         'atilde'   => 227,
  86         'Auml'     => 196,
  87         'auml'     => 228,
  88         'bdquo'    => 8222,
  89         'Beta'     => 914,
  90         'beta'     => 946,
  91         'brvbar'   => 166,
  92         'bull'     => 8226,
  93         'cap'      => 8745,
  94         'Ccedil'   => 199,
  95         'ccedil'   => 231,
  96         'cedil'    => 184,
  97         'cent'     => 162,
  98         'Chi'      => 935,
  99         'chi'      => 967,
 100         'circ'     => 710,
 101         'clubs'    => 9827,
 102         'cong'     => 8773,
 103         'copy'     => 169,
 104         'crarr'    => 8629,
 105         'cup'      => 8746,
 106         'curren'   => 164,
 107         'dagger'   => 8224,
 108         'Dagger'   => 8225,
 109         'darr'     => 8595,
 110         'dArr'     => 8659,
 111         'deg'      => 176,
 112         'Delta'    => 916,
 113         'delta'    => 948,
 114         'diams'    => 9830,
 115         'divide'   => 247,
 116         'Eacute'   => 201,
 117         'eacute'   => 233,
 118         'Ecirc'    => 202,
 119         'ecirc'    => 234,
 120         'Egrave'   => 200,
 121         'egrave'   => 232,
 122         'empty'    => 8709,
 123         'emsp'     => 8195,
 124         'ensp'     => 8194,
 125         'Epsilon'  => 917,
 126         'epsilon'  => 949,
 127         'equiv'    => 8801,
 128         'Eta'      => 919,
 129         'eta'      => 951,
 130         'ETH'      => 208,
 131         'eth'      => 240,
 132         'Euml'     => 203,
 133         'euml'     => 235,
 134         'euro'     => 8364,
 135         'exist'    => 8707,
 136         'fnof'     => 402,
 137         'forall'   => 8704,
 138         'frac12'   => 189,
 139         'frac14'   => 188,
 140         'frac34'   => 190,
 141         'frasl'    => 8260,
 142         'Gamma'    => 915,
 143         'gamma'    => 947,
 144         'ge'       => 8805,
 145         'gt'       => 62,
 146         'harr'     => 8596,
 147         'hArr'     => 8660,
 148         'hearts'   => 9829,
 149         'hellip'   => 8230,
 150         'Iacute'   => 205,
 151         'iacute'   => 237,
 152         'Icirc'    => 206,
 153         'icirc'    => 238,
 154         'iexcl'    => 161,
 155         'Igrave'   => 204,
 156         'igrave'   => 236,
 157         'image'    => 8465,
 158         'infin'    => 8734,
 159         'int'      => 8747,
 160         'Iota'     => 921,
 161         'iota'     => 953,
 162         'iquest'   => 191,
 163         'isin'     => 8712,
 164         'Iuml'     => 207,
 165         'iuml'     => 239,
 166         'Kappa'    => 922,
 167         'kappa'    => 954,
 168         'Lambda'   => 923,
 169         'lambda'   => 955,
 170         'lang'     => 9001,
 171         'laquo'    => 171,
 172         'larr'     => 8592,
 173         'lArr'     => 8656,
 174         'lceil'    => 8968,
 175         'ldquo'    => 8220,
 176         'le'       => 8804,
 177         'lfloor'   => 8970,
 178         'lowast'   => 8727,
 179         'loz'      => 9674,
 180         'lrm'      => 8206,
 181         'lsaquo'   => 8249,
 182         'lsquo'    => 8216,
 183         'lt'       => 60,
 184         'macr'     => 175,
 185         'mdash'    => 8212,
 186         'micro'    => 181,
 187         'middot'   => 183,
 188         'minus'    => 8722,
 189         'Mu'       => 924,
 190         'mu'       => 956,
 191         'nabla'    => 8711,
 192         'nbsp'     => 160,
 193         'ndash'    => 8211,
 194         'ne'       => 8800,
 195         'ni'       => 8715,
 196         'not'      => 172,
 197         'notin'    => 8713,
 198         'nsub'     => 8836,
 199         'Ntilde'   => 209,
 200         'ntilde'   => 241,
 201         'Nu'       => 925,
 202         'nu'       => 957,
 203         'Oacute'   => 211,
 204         'oacute'   => 243,
 205         'Ocirc'    => 212,
 206         'ocirc'    => 244,
 207         'OElig'    => 338,
 208         'oelig'    => 339,
 209         'Ograve'   => 210,
 210         'ograve'   => 242,
 211         'oline'    => 8254,
 212         'Omega'    => 937,
 213         'omega'    => 969,
 214         'Omicron'  => 927,
 215         'omicron'  => 959,
 216         'oplus'    => 8853,
 217         'or'       => 8744,
 218         'ordf'     => 170,
 219         'ordm'     => 186,
 220         'Oslash'   => 216,
 221         'oslash'   => 248,
 222         'Otilde'   => 213,
 223         'otilde'   => 245,
 224         'otimes'   => 8855,
 225         'Ouml'     => 214,
 226         'ouml'     => 246,
 227         'para'     => 182,
 228         'part'     => 8706,
 229         'permil'   => 8240,
 230         'perp'     => 8869,
 231         'Phi'      => 934,
 232         'phi'      => 966,
 233         'Pi'       => 928,
 234         'pi'       => 960,
 235         'piv'      => 982,
 236         'plusmn'   => 177,
 237         'pound'    => 163,
 238         'prime'    => 8242,
 239         'Prime'    => 8243,
 240         'prod'     => 8719,
 241         'prop'     => 8733,
 242         'Psi'      => 936,
 243         'psi'      => 968,
 244         'quot'     => 34,
 245         'radic'    => 8730,
 246         'rang'     => 9002,
 247         'raquo'    => 187,
 248         'rarr'     => 8594,
 249         'rArr'     => 8658,
 250         'rceil'    => 8969,
 251         'rdquo'    => 8221,
 252         'real'     => 8476,
 253         'reg'      => 174,
 254         'rfloor'   => 8971,
 255         'Rho'      => 929,
 256         'rho'      => 961,
 257         'rlm'      => 8207,
 258         'rsaquo'   => 8250,
 259         'rsquo'    => 8217,
 260         'sbquo'    => 8218,
 261         'Scaron'   => 352,
 262         'scaron'   => 353,
 263         'sdot'     => 8901,
 264         'sect'     => 167,
 265         'shy'      => 173,
 266         'Sigma'    => 931,
 267         'sigma'    => 963,
 268         'sigmaf'   => 962,
 269         'sim'      => 8764,
 270         'spades'   => 9824,
 271         'sub'      => 8834,
 272         'sube'     => 8838,
 273         'sum'      => 8721,
 274         'sup'      => 8835,
 275         'sup1'     => 185,
 276         'sup2'     => 178,
 277         'sup3'     => 179,
 278         'supe'     => 8839,
 279         'szlig'    => 223,
 280         'Tau'      => 932,
 281         'tau'      => 964,
 282         'there4'   => 8756,
 283         'Theta'    => 920,
 284         'theta'    => 952,
 285         'thetasym' => 977,
 286         'thinsp'   => 8201,
 287         'THORN'    => 222,
 288         'thorn'    => 254,
 289         'tilde'    => 732,
 290         'times'    => 215,
 291         'trade'    => 8482,
 292         'Uacute'   => 218,
 293         'uacute'   => 250,
 294         'uarr'     => 8593,
 295         'uArr'     => 8657,
 296         'Ucirc'    => 219,
 297         'ucirc'    => 251,
 298         'Ugrave'   => 217,
 299         'ugrave'   => 249,
 300         'uml'      => 168,
 301         'upsih'    => 978,
 302         'Upsilon'  => 933,
 303         'upsilon'  => 965,
 304         'Uuml'     => 220,
 305         'uuml'     => 252,
 306         'weierp'   => 8472,
 307         'Xi'       => 926,
 308         'xi'       => 958,
 309         'Yacute'   => 221,
 310         'yacute'   => 253,
 311         'yen'      => 165,
 312         'Yuml'     => 376,
 313         'yuml'     => 255,
 314         'Zeta'     => 918,
 315         'zeta'     => 950,
 316         'zwj'      => 8205,
 317         'zwnj'     => 8204 );
 318
 319 /**
 320  * Character entity aliases accepted by MediaWiki
 321  */
 322 global $wgHtmlEntityAliases;
 323 $wgHtmlEntityAliases = array(
 324         'רלמ' => 'rlm',
 325         'رلم' => 'rlm',
 326 );
 327
 328
 329 /**
 330  * XHTML sanitizer for MediaWiki
 331  * @ingroup Parser
 332  */
 333 class Sanitizer {
 334         const NONE = 0;
 335         const INITIAL_NONLETTER = 1;
 336
 337         /**
 338          * Cleans up HTML, removes dangerous tags and attributes, and
 339          * removes HTML comments
 340          * @private
 341          * @param string $text
 342          * @param callback $processCallback to do any variable or parameter replacements in HTML attribute values
 343          * @param array $args for the processing callback
 344          * @return string
 345          */
 346         static function removeHTMLtags( $text, $processCallback = null, $args = array(), $extratags = array() ) {
 347                 global $wgUseTidy;
 348
 349                 static $htmlpairs, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 350                         $htmllist, $listtags, $htmlsingleallowed, $htmlelements, $staticInitialised;
 351
 352                 wfProfileIn( __METHOD__ );
 353
 354                 if ( !$staticInitialised ) {
 355
 356                         $htmlpairs = array_merge( $extratags, array( # Tags that must be closed
 357                                 'b', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 358                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 359                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 360                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 361                                 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'u'
 362                         ) );
 363                         $htmlsingle = array(
 364                                 'br', 'hr', 'li', 'dt', 'dd'
 365                         );
 366                         $htmlsingleonly = array( # Elements that cannot have close tags
 367                                 'br', 'hr'
 368                         );
 369                         $htmlnest = array( # Tags that can be nested--??
 370                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 371                                 'dl', 'font', 'big', 'small', 'sub', 'sup', 'span'
 372                         );
 373                         $tabletags = array( # Can only appear inside table, we will close them
 374                                 'td', 'th', 'tr',
 375                         );
 376                         $htmllist = array( # Tags used by list
 377                                 'ul','ol',
 378                         );
 379                         $listtags = array( # Tags that can appear in a list
 380                                 'li',
 381                         );
 382
 383                         $htmlsingleallowed = array_merge( $htmlsingle, $tabletags );
 384                         $htmlelements = array_merge( $htmlsingle, $htmlpairs, $htmlnest );
 385
 386                         # Convert them all to hashtables for faster lookup
 387                         $vars = array( 'htmlpairs', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 388                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelements' );
 389                         foreach ( $vars as $var ) {
 390                                 $$var = array_flip( $$var );
 391                         }
 392                         $staticInitialised = true;
 393                 }
 394
 395                 # Remove HTML comments
 396                 $text = Sanitizer::removeHTMLcomments( $text );
 397                 $bits = explode( '<', $text );
 398                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 399                 if(!$wgUseTidy) {
 400                         $tagstack = $tablestack = array();
 401                         foreach ( $bits as $x ) {
 402                                 $regs = array();
 403                                 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 404                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 405                                 } else {
 406                                         $slash = $t = $params = $brace = $rest = null;
 407                                 }
 408
 409                                 $badtag = 0 ;
 410                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 411                                         # Check our stack
 412                                         if ( $slash ) {
 413                                                 # Closing a tag...
 414                                                 if( isset( $htmlsingleonly[$t] ) ) {
 415                                                         $badtag = 1;
 416                                                 } elseif ( ( $ot = @array_pop( $tagstack ) ) != $t ) {
 417                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 418                                                                 # Pop all elements with an optional close tag
 419                                                                 # and see if we find a match below them
 420                                                                 $optstack = array();
 421                                                                 array_push ($optstack, $ot);
 422                                                                 while ( ( ( $ot = @array_pop( $tagstack ) ) != $t ) &&
 423                                                                                 isset( $htmlsingleallowed[$ot] ) )
 424                                                                 {
 425                                                                         array_push ($optstack, $ot);
 426                                                                 }
 427                                                                 if ( $t != $ot ) {
 428                                                                         # No match. Push the optinal elements back again
 429                                                                         $badtag = 1;
 430                                                                         while ( $ot = @array_pop( $optstack ) ) {
 431                                                                                 array_push( $tagstack, $ot );
 432                                                                         }
 433                                                                 }
 434                                                         } else {
 435                                                                 @array_push( $tagstack, $ot );
 436                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 437                                                                 if(!(isset( $htmllist[$ot] ) && isset( $listtags[$t] ) )) {
 438                                                                         $badtag = 1;
 439                                                                 }
 440                                                         }
 441                                                 } else {
 442                                                         if ( $t == 'table' ) {
 443                                                                 $tagstack = array_pop( $tablestack );
 444                                                         }
 445                                                 }
 446                                                 $newparams = '';
 447                                         } else {
 448                                                 # Keep track for later
 449                                                 if ( isset( $tabletags[$t] ) &&
 450                                                 ! in_array( 'table', $tagstack ) ) {
 451                                                         $badtag = 1;
 452                                                 } else if ( in_array( $t, $tagstack ) &&
 453                                                 ! isset( $htmlnest [$t ] ) ) {
 454                                                         $badtag = 1 ;
 455                                                 # Is it a self closed htmlpair ? (bug 5487)
 456                                                 } else if( $brace == '/>' &&
 457                                                 isset( $htmlpairs[$t] ) ) {
 458                                                         $badtag = 1;
 459                                                 } elseif( isset( $htmlsingleonly[$t] ) ) {
 460                                                         # Hack to force empty tag for uncloseable elements
 461                                                         $brace = '/>';
 462                                                 } else if( isset( $htmlsingle[$t] ) ) {
 463                                                         # Hack to not close $htmlsingle tags
 464                                                         $brace = NULL;
 465                                                 } else if( isset( $tabletags[$t] )
 466                                                 &&  in_array($t ,$tagstack) ) {
 467                                                         // New table tag but forgot to close the previous one
 468                                                         $text .= "</$t>";
 469                                                 } else {
 470                                                         if ( $t == 'table' ) {
 471                                                                 array_push( $tablestack, $tagstack );
 472                                                                 $tagstack = array();
 473                                                         }
 474                                                         array_push( $tagstack, $t );
 475                                                 }
 476
 477                                                 # Replace any variables or template parameters with
 478                                                 # plaintext results.
 479                                                 if( is_callable( $processCallback ) ) {
 480                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 481                                                 }
 482
 483                                                 # Strip non-approved attributes from the tag
 484                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 485                                         }
 486                                         if ( ! $badtag ) {
 487                                                 $rest = str_replace( '>', '&gt;', $rest );
 488                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 489                                                 $text .= "<$slash$t$newparams$close>$rest";
 490                                                 continue;
 491                                         }
 492                                 }
 493                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 494                         }
 495                         # Close off any remaining tags
 496                         while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
 497                                 $text .= "</$t>\n";
 498                                 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
 499                         }
 500                 } else {
 501                         # this might be possible using tidy itself
 502                         foreach ( $bits as $x ) {
 503                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 504                                 $x, $regs );
 505                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 506                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 507                                         if( is_callable( $processCallback ) ) {
 508                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 509                                         }
 510                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 511                                         $rest = str_replace( '>', '&gt;', $rest );
 512                                         $text .= "<$slash$t$newparams$brace$rest";
 513                                 } else {
 514                                         $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 515                                 }
 516                         }
 517                 }
 518                 wfProfileOut( __METHOD__ );
 519                 return $text;
 520         }
 521
 522         /**
 523          * Remove '<!--', '-->', and everything between.
 524          * To avoid leaving blank lines, when a comment is both preceded
 525          * and followed by a newline (ignoring spaces), trim leading and
 526          * trailing spaces and one of the newlines.
 527          *
 528          * @private
 529          * @param string $text
 530          * @return string
 531          */
 532         static function removeHTMLcomments( $text ) {
 533                 wfProfileIn( __METHOD__ );
 534                 while (($start = strpos($text, '<!--')) !== false) {
 535                         $end = strpos($text, '-->', $start + 4);
 536                         if ($end === false) {
 537                                 # Unterminated comment; bail out
 538                                 break;
 539                         }
 540
 541                         $end += 3;
 542
 543                         # Trim space and newline if the comment is both
 544                         # preceded and followed by a newline
 545                         $spaceStart = max($start - 1, 0);
 546                         $spaceLen = $end - $spaceStart;
 547                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 548                                 $spaceStart--;
 549                                 $spaceLen++;
 550                         }
 551                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 552                                 $spaceLen++;
 553                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 554                                 # Remove the comment, leading and trailing
 555                                 # spaces, and leave only one newline.
 556                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 557                         }
 558                         else {
 559                                 # Remove just the comment.
 560                                 $text = substr_replace($text, '', $start, $end - $start);
 561                         }
 562                 }
 563                 wfProfileOut( __METHOD__ );
 564                 return $text;
 565         }
 566
 567         /**
 568          * Take an array of attribute names and values and normalize or discard
 569          * illegal values for the given element type.
 570          *
 571          * - Discards attributes not on a whitelist for the given element
 572          * - Unsafe style attributes are discarded
 573          * - Invalid id attributes are reencoded
 574          *
 575          * @param array $attribs
 576          * @param string $element
 577          * @return array
 578          *
 579          * @todo Check for legal values where the DTD limits things.
 580          * @todo Check for unique id attribute :P
 581          */
 582         static function validateTagAttributes( $attribs, $element ) {
 583                 return Sanitizer::validateAttributes( $attribs,
 584                         Sanitizer::attributeWhitelist( $element ) );
 585         }
 586
 587         /**
 588          * Take an array of attribute names and values and normalize or discard
 589          * illegal values for the given whitelist.
 590          *
 591          * - Discards attributes not the given whitelist
 592          * - Unsafe style attributes are discarded
 593          * - Invalid id attributes are reencoded
 594          *
 595          * @param array $attribs
 596          * @param array $whitelist list of allowed attribute names
 597          * @return array
 598          *
 599          * @todo Check for legal values where the DTD limits things.
 600          * @todo Check for unique id attribute :P
 601          */
 602         static function validateAttributes( $attribs, $whitelist ) {
 603                 $whitelist = array_flip( $whitelist );
 604                 $out = array();
 605                 foreach( $attribs as $attribute => $value ) {
 606                         if( !isset( $whitelist[$attribute] ) ) {
 607                                 continue;
 608                         }
 609                         # Strip javascript "expression" from stylesheets.
 610                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 611                         if( $attribute == 'style' ) {
 612                                 $value = Sanitizer::checkCss( $value );
 613                                 if( $value === false ) {
 614                                         # haxx0r
 615                                         continue;
 616                                 }
 617                         }
 618
 619                         if ( $attribute === 'id' )
 620                                 $value = Sanitizer::escapeId( $value );
 621
 622                         // If this attribute was previously set, override it.
 623                         // Output should only have one attribute of each name.
 624                         $out[$attribute] = $value;
 625                 }
 626                 return $out;
 627         }
 628
 629         /**
 630          * Merge two sets of HTML attributes.  Conflicting items in the second set
 631          * will override those in the first, except for 'class' attributes which
 632          * will be combined (if they're both strings).
 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'] ) && isset( $b['class'] )
 642                 && is_string( $a['class'] ) && is_string( $b['class'] )
 643                 && $a['class'] !== $b['class'] ) {
 644                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 645                                 -1, PREG_SPLIT_NO_EMPTY );
 646                         $out['class'] = implode( ' ', array_unique( $classes ) );
 647                 }
 648                 return $out;
 649         }
 650
 651         /**
 652          * Pick apart some CSS and check it for forbidden or unsafe structures.
 653          * Returns a sanitized string, or false if it was just too evil.
 654          *
 655          * Currently URL references, 'expression', 'tps' are forbidden.
 656          *
 657          * @param string $value
 658          * @return mixed
 659          */
 660         static function checkCss( $value ) {
 661                 $stripped = Sanitizer::decodeCharReferences( $value );
 662
 663                 // Remove any comments; IE gets token splitting wrong
 664                 $stripped = StringUtils::delimiterReplace( '/*', '*/', ' ', $stripped );
 665
 666                 $value = $stripped;
 667
 668                 // ... and continue checks
 669                 $stripped = preg_replace( '!\\\\([0-9A-Fa-f]{1,6})[ \\n\\r\\t\\f]?!e',
 670                         'codepointToUtf8(hexdec("$1"))', $stripped );
 671                 $stripped = str_replace( '\\', '', $stripped );
 672                 if( preg_match( '/(?:expression|tps*:\/\/|url\\s*\().*/is',
 673                                 $stripped ) ) {
 674                         # haxx0r
 675                         return false;
 676                 }
 677
 678                 return $value;
 679         }
 680
 681         /**
 682          * Take a tag soup fragment listing an HTML element's attributes
 683          * and normalize it to well-formed XML, discarding unwanted attributes.
 684          * Output is safe for further wikitext processing, with escaping of
 685          * values that could trigger problems.
 686          *
 687          * - Normalizes attribute names to lowercase
 688          * - Discards attributes not on a whitelist for the given element
 689          * - Turns broken or invalid entities into plaintext
 690          * - Double-quotes all attribute values
 691          * - Attributes without values are given the name as attribute
 692          * - Double attributes are discarded
 693          * - Unsafe style attributes are discarded
 694          * - Prepends space if there are attributes.
 695          *
 696          * @param string $text
 697          * @param string $element
 698          * @return string
 699          */
 700         static function fixTagAttributes( $text, $element ) {
 701                 if( trim( $text ) == '' ) {
 702                         return '';
 703                 }
 704
 705                 $stripped = Sanitizer::validateTagAttributes(
 706                         Sanitizer::decodeTagAttributes( $text ), $element );
 707
 708                 $attribs = array();
 709                 foreach( $stripped as $attribute => $value ) {
 710                         $encAttribute = htmlspecialchars( $attribute );
 711                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 712
 713                         $attribs[] = "$encAttribute=\"$encValue\"";
 714                 }
 715                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 716         }
 717
 718         /**
 719          * Encode an attribute value for HTML output.
 720          * @param $text
 721          * @return HTML-encoded text fragment
 722          */
 723         static function encodeAttribute( $text ) {
 724                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 725
 726                 // Whitespace is normalized during attribute decoding,
 727                 // so if we've been passed non-spaces we must encode them
 728                 // ahead of time or they won't be preserved.
 729                 $encValue = strtr( $encValue, array(
 730                         "\n" => '&#10;',
 731                         "\r" => '&#13;',
 732                         "\t" => '&#9;',
 733                 ) );
 734
 735                 return $encValue;
 736         }
 737
 738         /**
 739          * Encode an attribute value for HTML tags, with extra armoring
 740          * against further wiki processing.
 741          * @param $text
 742          * @return HTML-encoded text fragment
 743          */
 744         static function safeEncodeAttribute( $text ) {
 745                 $encValue = Sanitizer::encodeAttribute( $text );
 746
 747                 # Templates and links may be expanded in later parsing,
 748                 # creating invalid or dangerous output. Suppress this.
 749                 $encValue = strtr( $encValue, array(
 750                         '<'    => '&lt;',   // This should never happen,
 751                         '>'    => '&gt;',   // we've received invalid input
 752                         '"'    => '&quot;', // which should have been escaped.
 753                         '{'    => '&#123;',
 754                         '['    => '&#91;',
 755                         "''"   => '&#39;&#39;',
 756                         'ISBN' => '&#73;SBN',
 757                         'RFC'  => '&#82;FC',
 758                         'PMID' => '&#80;MID',
 759                         '|'    => '&#124;',
 760                         '__'   => '&#95;_',
 761                 ) );
 762
 763                 # Stupid hack
 764                 $encValue = preg_replace_callback(
 765                         '/(' . wfUrlProtocols() . ')/',
 766                         array( 'Sanitizer', 'armorLinksCallback' ),
 767                         $encValue );
 768                 return $encValue;
 769         }
 770
 771         /**
 772          * Given a value escape it so that it can be used in an id attribute and
 773          * return it, this does not validate the value however (see first link)
 774          *
 775          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
 776          *                                                          in the id and
 777          *                                                          name attributes
 778          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
 779          *
 780          * @param string $id    Id to validate
 781          * @param int    $flags Currently only two values: Sanitizer::INITIAL_NONLETTER
 782          *                      (default) permits initial non-letter characters,
 783          *                      such as if you're adding a prefix to them.
 784          *                      Sanitizer::NONE will prepend an 'x' if the id
 785          *                      would otherwise start with a nonletter.
 786          * @return string
 787          */
 788         static function escapeId( $id, $flags = Sanitizer::INITIAL_NONLETTER ) {
 789                 static $replace = array(
 790                         '%3A' => ':',
 791                         '%' => '.'
 792                 );
 793
 794                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
 795                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
 796
 797                 if( ~$flags & Sanitizer::INITIAL_NONLETTER
 798                 && !preg_match( '/[a-zA-Z]/', $id[0] ) ) {
 799                         // Initial character must be a letter!
 800                         $id = "x$id";
 801                 }
 802                 return $id;
 803         }
 804
 805         /**
 806          * Given a value, escape it so that it can be used as a CSS class and
 807          * return it.
 808          *
 809          * @todo For extra validity, input should be validated UTF-8.
 810          *
 811          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
 812          *
 813          * @param string $class
 814          * @return string
 815          */
 816         static function escapeClass( $class ) {
 817                 // Convert ugly stuff to underscores and kill underscores in ugly places
 818                 return rtrim(preg_replace(
 819                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
 820                         '_',
 821                         $class ), '_');
 822         }
 823
 824         /**
 825          * Regex replace callback for armoring links against further processing.
 826          * @param array $matches
 827          * @return string
 828          * @private
 829          */
 830         private static function armorLinksCallback( $matches ) {
 831                 return str_replace( ':', '&#58;', $matches[1] );
 832         }
 833
 834         /**
 835          * Return an associative array of attribute names and values from
 836          * a partial tag string. Attribute names are forces to lowercase,
 837          * character references are decoded to UTF-8 text.
 838          *
 839          * @param string
 840          * @return array
 841          */
 842         static function decodeTagAttributes( $text ) {
 843                 $attribs = array();
 844
 845                 if( trim( $text ) == '' ) {
 846                         return $attribs;
 847                 }
 848
 849                 $pairs = array();
 850                 if( !preg_match_all(
 851                         MW_ATTRIBS_REGEX,
 852                         $text,
 853                         $pairs,
 854                         PREG_SET_ORDER ) ) {
 855                         return $attribs;
 856                 }
 857
 858                 foreach( $pairs as $set ) {
 859                         $attribute = strtolower( $set[1] );
 860                         $value = Sanitizer::getTagAttributeCallback( $set );
 861
 862                         // Normalize whitespace
 863                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
 864                         $value = trim( $value );
 865
 866                         // Decode character references
 867                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
 868                 }
 869                 return $attribs;
 870         }
 871
 872         /**
 873          * Pick the appropriate attribute value from a match set from the
 874          * MW_ATTRIBS_REGEX matches.
 875          *
 876          * @param array $set
 877          * @return string
 878          * @private
 879          */
 880         private static function getTagAttributeCallback( $set ) {
 881                 if( isset( $set[6] ) ) {
 882                         # Illegal #XXXXXX color with no quotes.
 883                         return $set[6];
 884                 } elseif( isset( $set[5] ) ) {
 885                         # No quotes.
 886                         return $set[5];
 887                 } elseif( isset( $set[4] ) ) {
 888                         # Single-quoted
 889                         return $set[4];
 890                 } elseif( isset( $set[3] ) ) {
 891                         # Double-quoted
 892                         return $set[3];
 893                 } elseif( !isset( $set[2] ) ) {
 894                         # In XHTML, attributes must have a value.
 895                         # For 'reduced' form, return explicitly the attribute name here.
 896                         return $set[1];
 897                 } else {
 898                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
 899                 }
 900         }
 901
 902         /**
 903          * Normalize whitespace and character references in an XML source-
 904          * encoded text for an attribute value.
 905          *
 906          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
 907          * but note that we're not returning the value, but are returning
 908          * XML source fragments that will be slapped into output.
 909          *
 910          * @param string $text
 911          * @return string
 912          * @private
 913          */
 914         private static function normalizeAttributeValue( $text ) {
 915                 return str_replace( '"', '&quot;',
 916                         self::normalizeWhitespace(
 917                                 Sanitizer::normalizeCharReferences( $text ) ) );
 918         }
 919
 920         private static function normalizeWhitespace( $text ) {
 921                 return preg_replace(
 922                         '/\r\n|[\x20\x0d\x0a\x09]/',
 923                         ' ',
 924                         $text );
 925         }
 926
 927         /**
 928          * Ensure that any entities and character references are legal
 929          * for XML and XHTML specifically. Any stray bits will be
 930          * &amp;-escaped to result in a valid text fragment.
 931          *
 932          * a. any named char refs must be known in XHTML
 933          * b. any numeric char refs must be legal chars, not invalid or forbidden
 934          * c. use &#x, not &#X
 935          * d. fix or reject non-valid attributes
 936          *
 937          * @param string $text
 938          * @return string
 939          * @private
 940          */
 941         static function normalizeCharReferences( $text ) {
 942                 return preg_replace_callback(
 943                         MW_CHAR_REFS_REGEX,
 944                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
 945                         $text );
 946         }
 947         /**
 948          * @param string $matches
 949          * @return string
 950          */
 951         static function normalizeCharReferencesCallback( $matches ) {
 952                 $ret = null;
 953                 if( $matches[1] != '' ) {
 954                         $ret = Sanitizer::normalizeEntity( $matches[1] );
 955                 } elseif( $matches[2] != '' ) {
 956                         $ret = Sanitizer::decCharReference( $matches[2] );
 957                 } elseif( $matches[3] != ''  ) {
 958                         $ret = Sanitizer::hexCharReference( $matches[3] );
 959                 } elseif( $matches[4] != '' ) {
 960                         $ret = Sanitizer::hexCharReference( $matches[4] );
 961                 }
 962                 if( is_null( $ret ) ) {
 963                         return htmlspecialchars( $matches[0] );
 964                 } else {
 965                         return $ret;
 966                 }
 967         }
 968
 969         /**
 970          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
 971          * return the named entity reference as is. If the entity is a
 972          * MediaWiki-specific alias, returns the HTML equivalent. Otherwise,
 973          * returns HTML-escaped text of pseudo-entity source (eg &amp;foo;)
 974          *
 975          * @param string $name
 976          * @return string
 977          * @static
 978          */
 979         static function normalizeEntity( $name ) {
 980                 global $wgHtmlEntities, $wgHtmlEntityAliases;
 981                 if ( isset( $wgHtmlEntityAliases[$name] ) ) {
 982                         return "&{$wgHtmlEntityAliases[$name]};";
 983                 } elseif( isset( $wgHtmlEntities[$name] ) ) {
 984                         return "&$name;";
 985                 } else {
 986                         return "&amp;$name;";
 987                 }
 988         }
 989
 990         static function decCharReference( $codepoint ) {
 991                 $point = intval( $codepoint );
 992                 if( Sanitizer::validateCodepoint( $point ) ) {
 993                         return sprintf( '&#%d;', $point );
 994                 } else {
 995                         return null;
 996                 }
 997         }
 998
 999         static function hexCharReference( $codepoint ) {
1000                 $point = hexdec( $codepoint );
1001                 if( Sanitizer::validateCodepoint( $point ) ) {
1002                         return sprintf( '&#x%x;', $point );
1003                 } else {
1004                         return null;
1005                 }
1006         }
1007
1008         /**
1009          * Returns true if a given Unicode codepoint is a valid character in XML.
1010          * @param int $codepoint
1011          * @return bool
1012          */
1013         private static function validateCodepoint( $codepoint ) {
1014                 return ($codepoint ==    0x09)
1015                         || ($codepoint ==    0x0a)
1016                         || ($codepoint ==    0x0d)
1017                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1018                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1019                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1020         }
1021
1022         /**
1023          * Decode any character references, numeric or named entities,
1024          * in the text and return a UTF-8 string.
1025          *
1026          * @param string $text
1027          * @return string
1028          * @public
1029          * @static
1030          */
1031         public static function decodeCharReferences( $text ) {
1032                 return preg_replace_callback(
1033                         MW_CHAR_REFS_REGEX,
1034                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1035                         $text );
1036         }
1037
1038         /**
1039          * @param string $matches
1040          * @return string
1041          */
1042         static function decodeCharReferencesCallback( $matches ) {
1043                 if( $matches[1] != '' ) {
1044                         return Sanitizer::decodeEntity( $matches[1] );
1045                 } elseif( $matches[2] != '' ) {
1046                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1047                 } elseif( $matches[3] != ''  ) {
1048                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1049                 } elseif( $matches[4] != '' ) {
1050                         return  Sanitizer::decodeChar( hexdec( $matches[4] ) );
1051                 }
1052                 # Last case should be an ampersand by itself
1053                 return $matches[0];
1054         }
1055
1056         /**
1057          * Return UTF-8 string for a codepoint if that is a valid
1058          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1059          * @param int $codepoint
1060          * @return string
1061          * @private
1062          */
1063         static function decodeChar( $codepoint ) {
1064                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1065                         return codepointToUtf8( $codepoint );
1066                 } else {
1067                         return UTF8_REPLACEMENT;
1068                 }
1069         }
1070
1071         /**
1072          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1073          * return the UTF-8 encoding of that character. Otherwise, returns
1074          * pseudo-entity source (eg &foo;)
1075          *
1076          * @param string $name
1077          * @return string
1078          */
1079         static function decodeEntity( $name ) {
1080                 global $wgHtmlEntities, $wgHtmlEntityAliases;
1081                 if ( isset( $wgHtmlEntityAliases[$name] ) ) {
1082                         $name = $wgHtmlEntityAliases[$name];
1083                 }
1084                 if( isset( $wgHtmlEntities[$name] ) ) {
1085                         return codepointToUtf8( $wgHtmlEntities[$name] );
1086                 } else {
1087                         return "&$name;";
1088                 }
1089         }
1090
1091         /**
1092          * Fetch the whitelist of acceptable attributes for a given
1093          * element name.
1094          *
1095          * @param string $element
1096          * @return array
1097          */
1098         static function attributeWhitelist( $element ) {
1099                 static $list;
1100                 if( !isset( $list ) ) {
1101                         $list = Sanitizer::setupAttributeWhitelist();
1102                 }
1103                 return isset( $list[$element] )
1104                         ? $list[$element]
1105                         : array();
1106         }
1107
1108         /**
1109          * @todo Document it a bit
1110          * @return array
1111          */
1112         static function setupAttributeWhitelist() {
1113                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1114                 $block = array_merge( $common, array( 'align' ) );
1115                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1116                 $tablecell = array( 'abbr',
1117                                     'axis',
1118                                     'headers',
1119                                     'scope',
1120                                     'rowspan',
1121                                     'colspan',
1122                                     'nowrap', # deprecated
1123                                     'width',  # deprecated
1124                                     'height', # deprecated
1125                                     'bgcolor' # deprecated
1126                                     );
1127
1128                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1129                 # See: http://www.w3.org/TR/html4/
1130                 $whitelist = array (
1131                         # 7.5.4
1132                         'div'        => $block,
1133                         'center'     => $common, # deprecated
1134                         'span'       => $block, # ??
1135
1136                         # 7.5.5
1137                         'h1'         => $block,
1138                         'h2'         => $block,
1139                         'h3'         => $block,
1140                         'h4'         => $block,
1141                         'h5'         => $block,
1142                         'h6'         => $block,
1143
1144                         # 7.5.6
1145                         # address
1146
1147                         # 8.2.4
1148                         # bdo
1149
1150                         # 9.2.1
1151                         'em'         => $common,
1152                         'strong'     => $common,
1153                         'cite'       => $common,
1154                         # dfn
1155                         'code'       => $common,
1156                         # samp
1157                         # kbd
1158                         'var'        => $common,
1159                         # abbr
1160                         # acronym
1161
1162                         # 9.2.2
1163                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1164                         # q
1165
1166                         # 9.2.3
1167                         'sub'        => $common,
1168                         'sup'        => $common,
1169
1170                         # 9.3.1
1171                         'p'          => $block,
1172
1173                         # 9.3.2
1174                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1175
1176                         # 9.3.4
1177                         'pre'        => array_merge( $common, array( 'width' ) ),
1178
1179                         # 9.4
1180                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1181                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1182
1183                         # 10.2
1184                         'ul'         => array_merge( $common, array( 'type' ) ),
1185                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1186                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1187
1188                         # 10.3
1189                         'dl'         => $common,
1190                         'dd'         => $common,
1191                         'dt'         => $common,
1192
1193                         # 11.2.1
1194                         'table'      => array_merge( $common,
1195                                                                 array( 'summary', 'width', 'border', 'frame',
1196                                                                                 'rules', 'cellspacing', 'cellpadding',
1197                                                                                 'align', 'bgcolor',
1198                                                                 ) ),
1199
1200                         # 11.2.2
1201                         'caption'    => array_merge( $common, array( 'align' ) ),
1202
1203                         # 11.2.3
1204                         'thead'      => array_merge( $common, $tablealign ),
1205                         'tfoot'      => array_merge( $common, $tablealign ),
1206                         'tbody'      => array_merge( $common, $tablealign ),
1207
1208                         # 11.2.4
1209                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1210                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1211
1212                         # 11.2.5
1213                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1214
1215                         # 11.2.6
1216                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1217                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1218
1219                         # 13.2
1220                         # Not usually allowed, but may be used for extension-style hooks
1221                         # such as <math> when it is rasterized
1222                         'img'        => array_merge( $common, array( 'alt' ) ),
1223
1224                         # 15.2.1
1225                         'tt'         => $common,
1226                         'b'          => $common,
1227                         'i'          => $common,
1228                         'big'        => $common,
1229                         'small'      => $common,
1230                         'strike'     => $common,
1231                         's'          => $common,
1232                         'u'          => $common,
1233
1234                         # 15.2.2
1235                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1236                         # basefont
1237
1238                         # 15.3
1239                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1240
1241                         # XHTML Ruby annotation text module, simple ruby only.
1242                         # http://www.w3c.org/TR/ruby/
1243                         'ruby'       => $common,
1244                         # rbc
1245                         # rtc
1246                         'rb'         => $common,
1247                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1248                         'rp'         => $common,
1249
1250                         # MathML root element, where used for extensions
1251                         # 'title' may not be 100% valid here; it's XHTML
1252                         # http://www.w3.org/TR/REC-MathML/
1253                         'math'       => array( 'class', 'style', 'id', 'title' ),
1254                         );
1255                 return $whitelist;
1256         }
1257
1258         /**
1259          * Take a fragment of (potentially invalid) HTML and return
1260          * a version with any tags removed, encoded as plain text.
1261          *
1262          * Warning: this return value must be further escaped for literal
1263          * inclusion in HTML output as of 1.10!
1264          *
1265          * @param string $text HTML fragment
1266          * @return string
1267          */
1268         static function stripAllTags( $text ) {
1269                 # Actual <tags>
1270                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1271
1272                 # Normalize &entities and whitespace
1273                 $text = self::decodeCharReferences( $text );
1274                 $text = self::normalizeWhitespace( $text );
1275
1276                 return $text;
1277         }
1278
1279         /**
1280          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1281          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1282          * PHP 5.1 doesn't.
1283          *
1284          * Use for passing XHTML fragments to PHP's XML parsing functions
1285          *
1286          * @return string
1287          * @static
1288          */
1289         static function hackDocType() {
1290                 global $wgHtmlEntities;
1291                 $out = "<!DOCTYPE html [\n";
1292                 foreach( $wgHtmlEntities as $entity => $codepoint ) {
1293                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1294                 }
1295                 $out .= "]>\n";
1296                 return $out;
1297         }
1298
1299         static function cleanUrl( $url ) {
1300                 # Normalize any HTML entities in input. They will be
1301                 # re-escaped by makeExternalLink().
1302                 $url = Sanitizer::decodeCharReferences( $url );
1303
1304                 # Escape any control characters introduced by the above step
1305                 $url = preg_replace( '/[\][<>"\\x00-\\x20\\x7F]/e', "urlencode('\\0')", $url );
1306
1307                 # Validate hostname portion
1308                 $matches = array();
1309                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1310                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1311
1312                         // Characters that will be ignored in IDNs.
1313                         // http://tools.ietf.org/html/3454#section-3.1
1314                         // Strip them before further processing so blacklists and such work.
1315                         $strip = "/
1316                                 \\s|          # general whitespace
1317                                 \xc2\xad|     # 00ad SOFT HYPHEN
1318                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1319                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1320                                 \xe2\x81\xa0| # 2060 WORD JOINER
1321                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1322                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1323                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1324                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1325                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1326                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1327                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1328                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe00f VARIATION SELECTOR-1-16
1329                                 /xuD";
1330
1331                         $host = preg_replace( $strip, '', $host );
1332
1333                         // @fixme: validate hostnames here
1334
1335                         return $protocol . $host . $rest;
1336                 } else {
1337                         return $url;
1338                 }
1339         }
1340
1341 }