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          * Given HTML input, escape with htmlspecialchars but un-escape entites.
 826          * This allows (generally harmless) entities like &nbsp; to survive.
 827          *
 828          * @param  string $html String to escape
 829          * @return string Escaped input
 830          */
 831         static function escapeHtmlAllowEntities( $html ) {
 832                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
 833                 # hurt.
 834                 $html = htmlspecialchars( $html, ENT_QUOTES );
 835                 $html = str_replace( '&amp;', '&', $html );
 836                 $html = Sanitizer::normalizeCharReferences( $html );
 837                 return $html;
 838         }
 839
 840         /**
 841          * Regex replace callback for armoring links against further processing.
 842          * @param array $matches
 843          * @return string
 844          * @private
 845          */
 846         private static function armorLinksCallback( $matches ) {
 847                 return str_replace( ':', '&#58;', $matches[1] );
 848         }
 849
 850         /**
 851          * Return an associative array of attribute names and values from
 852          * a partial tag string. Attribute names are forces to lowercase,
 853          * character references are decoded to UTF-8 text.
 854          *
 855          * @param string
 856          * @return array
 857          */
 858         public static function decodeTagAttributes( $text ) {
 859                 $attribs = array();
 860
 861                 if( trim( $text ) == '' ) {
 862                         return $attribs;
 863                 }
 864
 865                 $pairs = array();
 866                 if( !preg_match_all(
 867                         MW_ATTRIBS_REGEX,
 868                         $text,
 869                         $pairs,
 870                         PREG_SET_ORDER ) ) {
 871                         return $attribs;
 872                 }
 873
 874                 foreach( $pairs as $set ) {
 875                         $attribute = strtolower( $set[1] );
 876                         $value = Sanitizer::getTagAttributeCallback( $set );
 877
 878                         // Normalize whitespace
 879                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
 880                         $value = trim( $value );
 881
 882                         // Decode character references
 883                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
 884                 }
 885                 return $attribs;
 886         }
 887
 888         /**
 889          * Pick the appropriate attribute value from a match set from the
 890          * MW_ATTRIBS_REGEX matches.
 891          *
 892          * @param array $set
 893          * @return string
 894          * @private
 895          */
 896         private static function getTagAttributeCallback( $set ) {
 897                 if( isset( $set[6] ) ) {
 898                         # Illegal #XXXXXX color with no quotes.
 899                         return $set[6];
 900                 } elseif( isset( $set[5] ) ) {
 901                         # No quotes.
 902                         return $set[5];
 903                 } elseif( isset( $set[4] ) ) {
 904                         # Single-quoted
 905                         return $set[4];
 906                 } elseif( isset( $set[3] ) ) {
 907                         # Double-quoted
 908                         return $set[3];
 909                 } elseif( !isset( $set[2] ) ) {
 910                         # In XHTML, attributes must have a value.
 911                         # For 'reduced' form, return explicitly the attribute name here.
 912                         return $set[1];
 913                 } else {
 914                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
 915                 }
 916         }
 917
 918         /**
 919          * Normalize whitespace and character references in an XML source-
 920          * encoded text for an attribute value.
 921          *
 922          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
 923          * but note that we're not returning the value, but are returning
 924          * XML source fragments that will be slapped into output.
 925          *
 926          * @param string $text
 927          * @return string
 928          * @private
 929          */
 930         private static function normalizeAttributeValue( $text ) {
 931                 return str_replace( '"', '&quot;',
 932                         self::normalizeWhitespace(
 933                                 Sanitizer::normalizeCharReferences( $text ) ) );
 934         }
 935
 936         private static function normalizeWhitespace( $text ) {
 937                 return preg_replace(
 938                         '/\r\n|[\x20\x0d\x0a\x09]/',
 939                         ' ',
 940                         $text );
 941         }
 942
 943         /**
 944          * Ensure that any entities and character references are legal
 945          * for XML and XHTML specifically. Any stray bits will be
 946          * &amp;-escaped to result in a valid text fragment.
 947          *
 948          * a. any named char refs must be known in XHTML
 949          * b. any numeric char refs must be legal chars, not invalid or forbidden
 950          * c. use &#x, not &#X
 951          * d. fix or reject non-valid attributes
 952          *
 953          * @param string $text
 954          * @return string
 955          * @private
 956          */
 957         static function normalizeCharReferences( $text ) {
 958                 return preg_replace_callback(
 959                         MW_CHAR_REFS_REGEX,
 960                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
 961                         $text );
 962         }
 963         /**
 964          * @param string $matches
 965          * @return string
 966          */
 967         static function normalizeCharReferencesCallback( $matches ) {
 968                 $ret = null;
 969                 if( $matches[1] != '' ) {
 970                         $ret = Sanitizer::normalizeEntity( $matches[1] );
 971                 } elseif( $matches[2] != '' ) {
 972                         $ret = Sanitizer::decCharReference( $matches[2] );
 973                 } elseif( $matches[3] != ''  ) {
 974                         $ret = Sanitizer::hexCharReference( $matches[3] );
 975                 } elseif( $matches[4] != '' ) {
 976                         $ret = Sanitizer::hexCharReference( $matches[4] );
 977                 }
 978                 if( is_null( $ret ) ) {
 979                         return htmlspecialchars( $matches[0] );
 980                 } else {
 981                         return $ret;
 982                 }
 983         }
 984
 985         /**
 986          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
 987          * return the named entity reference as is. If the entity is a
 988          * MediaWiki-specific alias, returns the HTML equivalent. Otherwise,
 989          * returns HTML-escaped text of pseudo-entity source (eg &amp;foo;)
 990          *
 991          * @param string $name
 992          * @return string
 993          * @static
 994          */
 995         static function normalizeEntity( $name ) {
 996                 global $wgHtmlEntities, $wgHtmlEntityAliases;
 997                 if ( isset( $wgHtmlEntityAliases[$name] ) ) {
 998                         return "&{$wgHtmlEntityAliases[$name]};";
 999                 } elseif( isset( $wgHtmlEntities[$name] ) ) {
1000                         return "&$name;";
1001                 } else {
1002                         return "&amp;$name;";
1003                 }
1004         }
1005
1006         static function decCharReference( $codepoint ) {
1007                 $point = intval( $codepoint );
1008                 if( Sanitizer::validateCodepoint( $point ) ) {
1009                         return sprintf( '&#%d;', $point );
1010                 } else {
1011                         return null;
1012                 }
1013         }
1014
1015         static function hexCharReference( $codepoint ) {
1016                 $point = hexdec( $codepoint );
1017                 if( Sanitizer::validateCodepoint( $point ) ) {
1018                         return sprintf( '&#x%x;', $point );
1019                 } else {
1020                         return null;
1021                 }
1022         }
1023
1024         /**
1025          * Returns true if a given Unicode codepoint is a valid character in XML.
1026          * @param int $codepoint
1027          * @return bool
1028          */
1029         private static function validateCodepoint( $codepoint ) {
1030                 return ($codepoint ==    0x09)
1031                         || ($codepoint ==    0x0a)
1032                         || ($codepoint ==    0x0d)
1033                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1034                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1035                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1036         }
1037
1038         /**
1039          * Decode any character references, numeric or named entities,
1040          * in the text and return a UTF-8 string.
1041          *
1042          * @param string $text
1043          * @return string
1044          * @public
1045          * @static
1046          */
1047         public static function decodeCharReferences( $text ) {
1048                 return preg_replace_callback(
1049                         MW_CHAR_REFS_REGEX,
1050                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1051                         $text );
1052         }
1053
1054         /**
1055          * @param string $matches
1056          * @return string
1057          */
1058         static function decodeCharReferencesCallback( $matches ) {
1059                 if( $matches[1] != '' ) {
1060                         return Sanitizer::decodeEntity( $matches[1] );
1061                 } elseif( $matches[2] != '' ) {
1062                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1063                 } elseif( $matches[3] != ''  ) {
1064                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1065                 } elseif( $matches[4] != '' ) {
1066                         return  Sanitizer::decodeChar( hexdec( $matches[4] ) );
1067                 }
1068                 # Last case should be an ampersand by itself
1069                 return $matches[0];
1070         }
1071
1072         /**
1073          * Return UTF-8 string for a codepoint if that is a valid
1074          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1075          * @param int $codepoint
1076          * @return string
1077          * @private
1078          */
1079         static function decodeChar( $codepoint ) {
1080                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1081                         return codepointToUtf8( $codepoint );
1082                 } else {
1083                         return UTF8_REPLACEMENT;
1084                 }
1085         }
1086
1087         /**
1088          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1089          * return the UTF-8 encoding of that character. Otherwise, returns
1090          * pseudo-entity source (eg &foo;)
1091          *
1092          * @param string $name
1093          * @return string
1094          */
1095         static function decodeEntity( $name ) {
1096                 global $wgHtmlEntities, $wgHtmlEntityAliases;
1097                 if ( isset( $wgHtmlEntityAliases[$name] ) ) {
1098                         $name = $wgHtmlEntityAliases[$name];
1099                 }
1100                 if( isset( $wgHtmlEntities[$name] ) ) {
1101                         return codepointToUtf8( $wgHtmlEntities[$name] );
1102                 } else {
1103                         return "&$name;";
1104                 }
1105         }
1106
1107         /**
1108          * Fetch the whitelist of acceptable attributes for a given
1109          * element name.
1110          *
1111          * @param string $element
1112          * @return array
1113          */
1114         static function attributeWhitelist( $element ) {
1115                 static $list;
1116                 if( !isset( $list ) ) {
1117                         $list = Sanitizer::setupAttributeWhitelist();
1118                 }
1119                 return isset( $list[$element] )
1120                         ? $list[$element]
1121                         : array();
1122         }
1123
1124         /**
1125          * @todo Document it a bit
1126          * @return array
1127          */
1128         static function setupAttributeWhitelist() {
1129                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1130                 $block = array_merge( $common, array( 'align' ) );
1131                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1132                 $tablecell = array( 'abbr',
1133                                     'axis',
1134                                     'headers',
1135                                     'scope',
1136                                     'rowspan',
1137                                     'colspan',
1138                                     'nowrap', # deprecated
1139                                     'width',  # deprecated
1140                                     'height', # deprecated
1141                                     'bgcolor' # deprecated
1142                                     );
1143
1144                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1145                 # See: http://www.w3.org/TR/html4/
1146                 $whitelist = array (
1147                         # 7.5.4
1148                         'div'        => $block,
1149                         'center'     => $common, # deprecated
1150                         'span'       => $block, # ??
1151
1152                         # 7.5.5
1153                         'h1'         => $block,
1154                         'h2'         => $block,
1155                         'h3'         => $block,
1156                         'h4'         => $block,
1157                         'h5'         => $block,
1158                         'h6'         => $block,
1159
1160                         # 7.5.6
1161                         # address
1162
1163                         # 8.2.4
1164                         # bdo
1165
1166                         # 9.2.1
1167                         'em'         => $common,
1168                         'strong'     => $common,
1169                         'cite'       => $common,
1170                         # dfn
1171                         'code'       => $common,
1172                         # samp
1173                         # kbd
1174                         'var'        => $common,
1175                         # abbr
1176                         # acronym
1177
1178                         # 9.2.2
1179                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1180                         # q
1181
1182                         # 9.2.3
1183                         'sub'        => $common,
1184                         'sup'        => $common,
1185
1186                         # 9.3.1
1187                         'p'          => $block,
1188
1189                         # 9.3.2
1190                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1191
1192                         # 9.3.4
1193                         'pre'        => array_merge( $common, array( 'width' ) ),
1194
1195                         # 9.4
1196                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1197                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1198
1199                         # 10.2
1200                         'ul'         => array_merge( $common, array( 'type' ) ),
1201                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1202                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1203
1204                         # 10.3
1205                         'dl'         => $common,
1206                         'dd'         => $common,
1207                         'dt'         => $common,
1208
1209                         # 11.2.1
1210                         'table'      => array_merge( $common,
1211                                                                 array( 'summary', 'width', 'border', 'frame',
1212                                                                                 'rules', 'cellspacing', 'cellpadding',
1213                                                                                 'align', 'bgcolor',
1214                                                                 ) ),
1215
1216                         # 11.2.2
1217                         'caption'    => array_merge( $common, array( 'align' ) ),
1218
1219                         # 11.2.3
1220                         'thead'      => array_merge( $common, $tablealign ),
1221                         'tfoot'      => array_merge( $common, $tablealign ),
1222                         'tbody'      => array_merge( $common, $tablealign ),
1223
1224                         # 11.2.4
1225                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1226                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1227
1228                         # 11.2.5
1229                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1230
1231                         # 11.2.6
1232                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1233                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1234
1235                         # 13.2
1236                         # Not usually allowed, but may be used for extension-style hooks
1237                         # such as <math> when it is rasterized
1238                         'img'        => array_merge( $common, array( 'alt' ) ),
1239
1240                         # 15.2.1
1241                         'tt'         => $common,
1242                         'b'          => $common,
1243                         'i'          => $common,
1244                         'big'        => $common,
1245                         'small'      => $common,
1246                         'strike'     => $common,
1247                         's'          => $common,
1248                         'u'          => $common,
1249
1250                         # 15.2.2
1251                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1252                         # basefont
1253
1254                         # 15.3
1255                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1256
1257                         # XHTML Ruby annotation text module, simple ruby only.
1258                         # http://www.w3c.org/TR/ruby/
1259                         'ruby'       => $common,
1260                         # rbc
1261                         # rtc
1262                         'rb'         => $common,
1263                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1264                         'rp'         => $common,
1265
1266                         # MathML root element, where used for extensions
1267                         # 'title' may not be 100% valid here; it's XHTML
1268                         # http://www.w3.org/TR/REC-MathML/
1269                         'math'       => array( 'class', 'style', 'id', 'title' ),
1270                         );
1271                 return $whitelist;
1272         }
1273
1274         /**
1275          * Take a fragment of (potentially invalid) HTML and return
1276          * a version with any tags removed, encoded as plain text.
1277          *
1278          * Warning: this return value must be further escaped for literal
1279          * inclusion in HTML output as of 1.10!
1280          *
1281          * @param string $text HTML fragment
1282          * @return string
1283          */
1284         static function stripAllTags( $text ) {
1285                 # Actual <tags>
1286                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1287
1288                 # Normalize &entities and whitespace
1289                 $text = self::decodeCharReferences( $text );
1290                 $text = self::normalizeWhitespace( $text );
1291
1292                 return $text;
1293         }
1294
1295         /**
1296          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1297          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1298          * PHP 5.1 doesn't.
1299          *
1300          * Use for passing XHTML fragments to PHP's XML parsing functions
1301          *
1302          * @return string
1303          * @static
1304          */
1305         static function hackDocType() {
1306                 global $wgHtmlEntities;
1307                 $out = "<!DOCTYPE html [\n";
1308                 foreach( $wgHtmlEntities as $entity => $codepoint ) {
1309                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1310                 }
1311                 $out .= "]>\n";
1312                 return $out;
1313         }
1314
1315         static function cleanUrl( $url ) {
1316                 # Normalize any HTML entities in input. They will be
1317                 # re-escaped by makeExternalLink().
1318                 $url = Sanitizer::decodeCharReferences( $url );
1319
1320                 # Escape any control characters introduced by the above step
1321                 $url = preg_replace( '/[\][<>"\\x00-\\x20\\x7F]/e', "urlencode('\\0')", $url );
1322
1323                 # Validate hostname portion
1324                 $matches = array();
1325                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1326                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1327
1328                         // Characters that will be ignored in IDNs.
1329                         // http://tools.ietf.org/html/3454#section-3.1
1330                         // Strip them before further processing so blacklists and such work.
1331                         $strip = "/
1332                                 \\s|          # general whitespace
1333                                 \xc2\xad|     # 00ad SOFT HYPHEN
1334                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1335                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1336                                 \xe2\x81\xa0| # 2060 WORD JOINER
1337                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1338                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1339                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1340                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1341                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1342                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1343                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1344                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe00f VARIATION SELECTOR-1-16
1345                                 /xuD";
1346
1347                         $host = preg_replace( $strip, '', $host );
1348
1349                         // @fixme: validate hostnames here
1350
1351                         return $protocol . $host . $rest;
1352                 } else {
1353                         return $url;
1354                 }
1355         }
1356
1357 }