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)((?:xml:|xmlns:)?$attrib+)
  47           ($space*=$space*
  48                 (?:
  49                  # The attribute value: quoted or alone
  50                   \"([^<\"]*)\"
  51                  | '([^<']*)'
  52                  |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
  53                  |  (\#[0-9a-fA-F]+) # Technically wrong, but lots of
  54                                                          # colors are specified like this.
  55                                                          # We'll be normalizing it.
  56                 )
  57            )?(?=$space|\$)/sx" );
  58
  59 /**
  60  * Regular expression to match URIs that could trigger script execution
  61  */
  62 define( 'MW_EVIL_URI_PATTERN', '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i' );
  63
  64 /**
  65  * Regular expression to match namespace attributes
  66  */
  67 define( 'MW_XMLNS_ATTRIBUTE_PATTRN', "/^xmlns:$attrib+$/" );
  68
  69 /**
  70  * List of all named character entities defined in HTML 4.01
  71  * http://www.w3.org/TR/html4/sgml/entities.html
  72  * @private
  73  */
  74 global $wgHtmlEntities;
  75 $wgHtmlEntities = array(
  76         'Aacute'   => 193,
  77         'aacute'   => 225,
  78         'Acirc'    => 194,
  79         'acirc'    => 226,
  80         'acute'    => 180,
  81         'AElig'    => 198,
  82         'aelig'    => 230,
  83         'Agrave'   => 192,
  84         'agrave'   => 224,
  85         'alefsym'  => 8501,
  86         'Alpha'    => 913,
  87         'alpha'    => 945,
  88         'amp'      => 38,
  89         'and'      => 8743,
  90         'ang'      => 8736,
  91         'Aring'    => 197,
  92         'aring'    => 229,
  93         'asymp'    => 8776,
  94         'Atilde'   => 195,
  95         'atilde'   => 227,
  96         'Auml'     => 196,
  97         'auml'     => 228,
  98         'bdquo'    => 8222,
  99         'Beta'     => 914,
 100         'beta'     => 946,
 101         'brvbar'   => 166,
 102         'bull'     => 8226,
 103         'cap'      => 8745,
 104         'Ccedil'   => 199,
 105         'ccedil'   => 231,
 106         'cedil'    => 184,
 107         'cent'     => 162,
 108         'Chi'      => 935,
 109         'chi'      => 967,
 110         'circ'     => 710,
 111         'clubs'    => 9827,
 112         'cong'     => 8773,
 113         'copy'     => 169,
 114         'crarr'    => 8629,
 115         'cup'      => 8746,
 116         'curren'   => 164,
 117         'dagger'   => 8224,
 118         'Dagger'   => 8225,
 119         'darr'     => 8595,
 120         'dArr'     => 8659,
 121         'deg'      => 176,
 122         'Delta'    => 916,
 123         'delta'    => 948,
 124         'diams'    => 9830,
 125         'divide'   => 247,
 126         'Eacute'   => 201,
 127         'eacute'   => 233,
 128         'Ecirc'    => 202,
 129         'ecirc'    => 234,
 130         'Egrave'   => 200,
 131         'egrave'   => 232,
 132         'empty'    => 8709,
 133         'emsp'     => 8195,
 134         'ensp'     => 8194,
 135         'Epsilon'  => 917,
 136         'epsilon'  => 949,
 137         'equiv'    => 8801,
 138         'Eta'      => 919,
 139         'eta'      => 951,
 140         'ETH'      => 208,
 141         'eth'      => 240,
 142         'Euml'     => 203,
 143         'euml'     => 235,
 144         'euro'     => 8364,
 145         'exist'    => 8707,
 146         'fnof'     => 402,
 147         'forall'   => 8704,
 148         'frac12'   => 189,
 149         'frac14'   => 188,
 150         'frac34'   => 190,
 151         'frasl'    => 8260,
 152         'Gamma'    => 915,
 153         'gamma'    => 947,
 154         'ge'       => 8805,
 155         'gt'       => 62,
 156         'harr'     => 8596,
 157         'hArr'     => 8660,
 158         'hearts'   => 9829,
 159         'hellip'   => 8230,
 160         'Iacute'   => 205,
 161         'iacute'   => 237,
 162         'Icirc'    => 206,
 163         'icirc'    => 238,
 164         'iexcl'    => 161,
 165         'Igrave'   => 204,
 166         'igrave'   => 236,
 167         'image'    => 8465,
 168         'infin'    => 8734,
 169         'int'      => 8747,
 170         'Iota'     => 921,
 171         'iota'     => 953,
 172         'iquest'   => 191,
 173         'isin'     => 8712,
 174         'Iuml'     => 207,
 175         'iuml'     => 239,
 176         'Kappa'    => 922,
 177         'kappa'    => 954,
 178         'Lambda'   => 923,
 179         'lambda'   => 955,
 180         'lang'     => 9001,
 181         'laquo'    => 171,
 182         'larr'     => 8592,
 183         'lArr'     => 8656,
 184         'lceil'    => 8968,
 185         'ldquo'    => 8220,
 186         'le'       => 8804,
 187         'lfloor'   => 8970,
 188         'lowast'   => 8727,
 189         'loz'      => 9674,
 190         'lrm'      => 8206,
 191         'lsaquo'   => 8249,
 192         'lsquo'    => 8216,
 193         'lt'       => 60,
 194         'macr'     => 175,
 195         'mdash'    => 8212,
 196         'micro'    => 181,
 197         'middot'   => 183,
 198         'minus'    => 8722,
 199         'Mu'       => 924,
 200         'mu'       => 956,
 201         'nabla'    => 8711,
 202         'nbsp'     => 160,
 203         'ndash'    => 8211,
 204         'ne'       => 8800,
 205         'ni'       => 8715,
 206         'not'      => 172,
 207         'notin'    => 8713,
 208         'nsub'     => 8836,
 209         'Ntilde'   => 209,
 210         'ntilde'   => 241,
 211         'Nu'       => 925,
 212         'nu'       => 957,
 213         'Oacute'   => 211,
 214         'oacute'   => 243,
 215         'Ocirc'    => 212,
 216         'ocirc'    => 244,
 217         'OElig'    => 338,
 218         'oelig'    => 339,
 219         'Ograve'   => 210,
 220         'ograve'   => 242,
 221         'oline'    => 8254,
 222         'Omega'    => 937,
 223         'omega'    => 969,
 224         'Omicron'  => 927,
 225         'omicron'  => 959,
 226         'oplus'    => 8853,
 227         'or'       => 8744,
 228         'ordf'     => 170,
 229         'ordm'     => 186,
 230         'Oslash'   => 216,
 231         'oslash'   => 248,
 232         'Otilde'   => 213,
 233         'otilde'   => 245,
 234         'otimes'   => 8855,
 235         'Ouml'     => 214,
 236         'ouml'     => 246,
 237         'para'     => 182,
 238         'part'     => 8706,
 239         'permil'   => 8240,
 240         'perp'     => 8869,
 241         'Phi'      => 934,
 242         'phi'      => 966,
 243         'Pi'       => 928,
 244         'pi'       => 960,
 245         'piv'      => 982,
 246         'plusmn'   => 177,
 247         'pound'    => 163,
 248         'prime'    => 8242,
 249         'Prime'    => 8243,
 250         'prod'     => 8719,
 251         'prop'     => 8733,
 252         'Psi'      => 936,
 253         'psi'      => 968,
 254         'quot'     => 34,
 255         'radic'    => 8730,
 256         'rang'     => 9002,
 257         'raquo'    => 187,
 258         'rarr'     => 8594,
 259         'rArr'     => 8658,
 260         'rceil'    => 8969,
 261         'rdquo'    => 8221,
 262         'real'     => 8476,
 263         'reg'      => 174,
 264         'rfloor'   => 8971,
 265         'Rho'      => 929,
 266         'rho'      => 961,
 267         'rlm'      => 8207,
 268         'rsaquo'   => 8250,
 269         'rsquo'    => 8217,
 270         'sbquo'    => 8218,
 271         'Scaron'   => 352,
 272         'scaron'   => 353,
 273         'sdot'     => 8901,
 274         'sect'     => 167,
 275         'shy'      => 173,
 276         'Sigma'    => 931,
 277         'sigma'    => 963,
 278         'sigmaf'   => 962,
 279         'sim'      => 8764,
 280         'spades'   => 9824,
 281         'sub'      => 8834,
 282         'sube'     => 8838,
 283         'sum'      => 8721,
 284         'sup'      => 8835,
 285         'sup1'     => 185,
 286         'sup2'     => 178,
 287         'sup3'     => 179,
 288         'supe'     => 8839,
 289         'szlig'    => 223,
 290         'Tau'      => 932,
 291         'tau'      => 964,
 292         'there4'   => 8756,
 293         'Theta'    => 920,
 294         'theta'    => 952,
 295         'thetasym' => 977,
 296         'thinsp'   => 8201,
 297         'THORN'    => 222,
 298         'thorn'    => 254,
 299         'tilde'    => 732,
 300         'times'    => 215,
 301         'trade'    => 8482,
 302         'Uacute'   => 218,
 303         'uacute'   => 250,
 304         'uarr'     => 8593,
 305         'uArr'     => 8657,
 306         'Ucirc'    => 219,
 307         'ucirc'    => 251,
 308         'Ugrave'   => 217,
 309         'ugrave'   => 249,
 310         'uml'      => 168,
 311         'upsih'    => 978,
 312         'Upsilon'  => 933,
 313         'upsilon'  => 965,
 314         'Uuml'     => 220,
 315         'uuml'     => 252,
 316         'weierp'   => 8472,
 317         'Xi'       => 926,
 318         'xi'       => 958,
 319         'Yacute'   => 221,
 320         'yacute'   => 253,
 321         'yen'      => 165,
 322         'Yuml'     => 376,
 323         'yuml'     => 255,
 324         'Zeta'     => 918,
 325         'zeta'     => 950,
 326         'zwj'      => 8205,
 327         'zwnj'     => 8204 );
 328
 329 /**
 330  * Character entity aliases accepted by MediaWiki
 331  */
 332 global $wgHtmlEntityAliases;
 333 $wgHtmlEntityAliases = array(
 334         'רלמ' => 'rlm',
 335         'رلم' => 'rlm',
 336 );
 337
 338
 339 /**
 340  * XHTML sanitizer for MediaWiki
 341  * @ingroup Parser
 342  */
 343 class Sanitizer {
 344         /**
 345          * Cleans up HTML, removes dangerous tags and attributes, and
 346          * removes HTML comments
 347          * @private
 348          * @param $text String
 349          * @param $processCallback Callback to do any variable or parameter replacements in HTML attribute values
 350          * @param $args Array for the processing callback
 351          * @param $extratags Array for any extra tags to include
 352          * @param $removetags Array for any tags (default or extra) to exclude
 353          * @return string
 354          */
 355         static function removeHTMLtags( $text, $processCallback = null, $args = array(), $extratags = array(), $removetags = array() ) {
 356                 global $wgUseTidy;
 357
 358                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 359                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 360
 361                 wfProfileIn( __METHOD__ );
 362
 363                 if ( !$staticInitialised ) {
 364
 365                         $htmlpairsStatic = array( # Tags that must be closed
 366                                 'b', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 367                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 368                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 369                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 370                                 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'u', 'abbr'
 371                         );
 372                         $htmlsingle = array(
 373                                 'br', 'hr', 'li', 'dt', 'dd'
 374                         );
 375                         $htmlsingleonly = array( # Elements that cannot have close tags
 376                                 'br', 'hr'
 377                         );
 378                         $htmlnest = array( # Tags that can be nested--??
 379                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 380                                 'dl', 'font', 'big', 'small', 'sub', 'sup', 'span'
 381                         );
 382                         $tabletags = array( # Can only appear inside table, we will close them
 383                                 'td', 'th', 'tr',
 384                         );
 385                         $htmllist = array( # Tags used by list
 386                                 'ul','ol',
 387                         );
 388                         $listtags = array( # Tags that can appear in a list
 389                                 'li',
 390                         );
 391
 392                         global $wgAllowImageTag;
 393                         if ( $wgAllowImageTag ) {
 394                                 $htmlsingle[] = 'img';
 395                                 $htmlsingleonly[] = 'img';
 396                         }
 397
 398                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 399                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 400
 401                         # Convert them all to hashtables for faster lookup
 402                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 403                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 404                         foreach ( $vars as $var ) {
 405                                 $$var = array_flip( $$var );
 406                         }
 407                         $staticInitialised = true;
 408                 }
 409                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 410                 $extratags = array_flip( $extratags );
 411                 $removetags = array_flip( $removetags );
 412                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 413                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ) , $removetags );
 414
 415                 # Remove HTML comments
 416                 $text = Sanitizer::removeHTMLcomments( $text );
 417                 $bits = explode( '<', $text );
 418                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 419                 if ( !$wgUseTidy ) {
 420                         $tagstack = $tablestack = array();
 421                         foreach ( $bits as $x ) {
 422                                 $regs = array();
 423                                 # $slash: Does the current element start with a '/'?
 424                                 # $t: Current element name
 425                                 # $params: String between element name and >
 426                                 # $brace: Ending '>' or '/>'
 427                                 # $rest: Everything until the next element of $bits
 428                                 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 429                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 430                                 } else {
 431                                         $slash = $t = $params = $brace = $rest = null;
 432                                 }
 433
 434                                 $badtag = false;
 435                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 436                                         # Check our stack
 437                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 438                                                 $badtag = true;
 439                                         } elseif ( $slash ) {
 440                                                 # Closing a tag... is it the one we just opened?
 441                                                 $ot = @array_pop( $tagstack );
 442                                                 if ( $ot != $t ) {
 443                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 444                                                                 # Pop all elements with an optional close tag
 445                                                                 # and see if we find a match below them
 446                                                                 $optstack = array();
 447                                                                 array_push( $optstack, $ot );
 448                                                                 $ot = @array_pop( $tagstack );
 449                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 450                                                                         array_push( $optstack, $ot );
 451                                                                         $ot = @array_pop( $tagstack );
 452                                                                 }
 453                                                                 if ( $t != $ot ) {
 454                                                                         # No match. Push the optional elements back again
 455                                                                         $badtag = true;
 456                                                                         while ( $ot = @array_pop( $optstack ) ) {
 457                                                                                 array_push( $tagstack, $ot );
 458                                                                         }
 459                                                                 }
 460                                                         } else {
 461                                                                 @array_push( $tagstack, $ot );
 462                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 463                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 464                                                                         $badtag = true;
 465                                                                 }
 466                                                         }
 467                                                 } else {
 468                                                         if ( $t == 'table' ) {
 469                                                                 $tagstack = array_pop( $tablestack );
 470                                                         }
 471                                                 }
 472                                                 $newparams = '';
 473                                         } else {
 474                                                 # Keep track for later
 475                                                 if ( isset( $tabletags[$t] ) &&
 476                                                 !in_array( 'table', $tagstack ) ) {
 477                                                         $badtag = true;
 478                                                 } elseif ( in_array( $t, $tagstack ) &&
 479                                                 !isset( $htmlnest [$t ] ) ) {
 480                                                         $badtag = true;
 481                                                 # Is it a self closed htmlpair ? (bug 5487)
 482                                                 } elseif ( $brace == '/>' &&
 483                                                 isset( $htmlpairs[$t] ) ) {
 484                                                         $badtag = true;
 485                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 486                                                         # Hack to force empty tag for uncloseable elements
 487                                                         $brace = '/>';
 488                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 489                                                         # Hack to not close $htmlsingle tags
 490                                                         $brace = null;
 491                                                 } elseif ( isset( $tabletags[$t] )
 492                                                 && in_array( $t, $tagstack ) ) {
 493                                                         // New table tag but forgot to close the previous one
 494                                                         $text .= "</$t>";
 495                                                 } else {
 496                                                         if ( $t == 'table' ) {
 497                                                                 array_push( $tablestack, $tagstack );
 498                                                                 $tagstack = array();
 499                                                         }
 500                                                         array_push( $tagstack, $t );
 501                                                 }
 502
 503                                                 # Replace any variables or template parameters with
 504                                                 # plaintext results.
 505                                                 if( is_callable( $processCallback ) ) {
 506                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 507                                                 }
 508
 509                                                 # Strip non-approved attributes from the tag
 510                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 511                                         }
 512                                         if ( !$badtag ) {
 513                                                 $rest = str_replace( '>', '&gt;', $rest );
 514                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 515                                                 $text .= "<$slash$t$newparams$close>$rest";
 516                                                 continue;
 517                                         }
 518                                 }
 519                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 520                         }
 521                         # Close off any remaining tags
 522                         while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
 523                                 $text .= "</$t>\n";
 524                                 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
 525                         }
 526                 } else {
 527                         # this might be possible using tidy itself
 528                         foreach ( $bits as $x ) {
 529                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 530                                 $x, $regs );
 531                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 532                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 533                                         if( is_callable( $processCallback ) ) {
 534                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 535                                         }
 536                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 537                                         $rest = str_replace( '>', '&gt;', $rest );
 538                                         $text .= "<$slash$t$newparams$brace$rest";
 539                                 } else {
 540                                         $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 541                                 }
 542                         }
 543                 }
 544                 wfProfileOut( __METHOD__ );
 545                 return $text;
 546         }
 547
 548         /**
 549          * Remove '<!--', '-->', and everything between.
 550          * To avoid leaving blank lines, when a comment is both preceded
 551          * and followed by a newline (ignoring spaces), trim leading and
 552          * trailing spaces and one of the newlines.
 553          *
 554          * @private
 555          * @param $text String
 556          * @return string
 557          */
 558         static function removeHTMLcomments( $text ) {
 559                 wfProfileIn( __METHOD__ );
 560                 while (($start = strpos($text, '<!--')) !== false) {
 561                         $end = strpos($text, '-->', $start + 4);
 562                         if ($end === false) {
 563                                 # Unterminated comment; bail out
 564                                 break;
 565                         }
 566
 567                         $end += 3;
 568
 569                         # Trim space and newline if the comment is both
 570                         # preceded and followed by a newline
 571                         $spaceStart = max($start - 1, 0);
 572                         $spaceLen = $end - $spaceStart;
 573                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 574                                 $spaceStart--;
 575                                 $spaceLen++;
 576                         }
 577                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 578                                 $spaceLen++;
 579                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 580                                 # Remove the comment, leading and trailing
 581                                 # spaces, and leave only one newline.
 582                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 583                         }
 584                         else {
 585                                 # Remove just the comment.
 586                                 $text = substr_replace($text, '', $start, $end - $start);
 587                         }
 588                 }
 589                 wfProfileOut( __METHOD__ );
 590                 return $text;
 591         }
 592
 593         /**
 594          * Take an array of attribute names and values and normalize or discard
 595          * illegal values for the given element type.
 596          *
 597          * - Discards attributes not on a whitelist for the given element
 598          * - Unsafe style attributes are discarded
 599          * - Invalid id attributes are reencoded
 600          *
 601          * @param $attribs Array
 602          * @param $element String
 603          * @return Array
 604          *
 605          * @todo Check for legal values where the DTD limits things.
 606          * @todo Check for unique id attribute :P
 607          */
 608         static function validateTagAttributes( $attribs, $element ) {
 609                 return Sanitizer::validateAttributes( $attribs,
 610                         Sanitizer::attributeWhitelist( $element ) );
 611         }
 612
 613         /**
 614          * Take an array of attribute names and values and normalize or discard
 615          * illegal values for the given whitelist.
 616          *
 617          * - Discards attributes not the given whitelist
 618          * - Unsafe style attributes are discarded
 619          * - Invalid id attributes are reencoded
 620          *
 621          * @param $attribs Array
 622          * @param $whitelist Array: list of allowed attribute names
 623          * @return Array
 624          *
 625          * @todo Check for legal values where the DTD limits things.
 626          * @todo Check for unique id attribute :P
 627          */
 628         static function validateAttributes( $attribs, $whitelist ) {
 629                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
 630
 631                 $whitelist = array_flip( $whitelist );
 632                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 633
 634                 $out = array();
 635                 foreach( $attribs as $attribute => $value ) {
 636                         #allow XML namespace declaration if RDFa is enabled
 637                         if ( $wgAllowRdfaAttributes && preg_match( MW_XMLNS_ATTRIBUTE_PATTRN, $attribute ) ) {
 638                                 if ( !preg_match( MW_EVIL_URI_PATTERN, $value ) ) {
 639                                         $out[$attribute] = $value;
 640                                 }
 641
 642                                 continue;
 643                         }
 644
 645                         if( !isset( $whitelist[$attribute] ) ) {
 646                                 continue;
 647                         }
 648
 649                         # Strip javascript "expression" from stylesheets.
 650                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 651                         if( $attribute == 'style' ) {
 652                                 $value = Sanitizer::checkCss( $value );
 653                         }
 654
 655                         if ( $attribute === 'id' ) {
 656                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 657                         }
 658
 659                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 660                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 661                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 662                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 663                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 664                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 665
 666                                 //Paranoia. Allow "simple" values but suppress javascript
 667                                 if ( preg_match( MW_EVIL_URI_PATTERN, $value ) ) {
 668                                         continue;
 669                                 }
 670                         }
 671
 672                         # NOTE: even though elements using href/src are not allowed directly, supply
 673                         #       validation code that can be used by tag hook handlers, etc
 674                         if ( $attribute === 'href' || $attribute === 'src' ) {
 675                                 if ( !preg_match( $hrefExp, $value ) ) {
 676                                         continue; //drop any href or src attributes not using an allowed protocol.
 677                                                   //NOTE: this also drops all relative URLs
 678                                 }
 679                         }
 680
 681                         // If this attribute was previously set, override it.
 682                         // Output should only have one attribute of each name.
 683                         $out[$attribute] = $value;
 684                 }
 685
 686                 if ( $wgAllowMicrodataAttributes ) {
 687                         # There are some complicated validity constraints we need to
 688                         # enforce here.  First of all, we don't want to allow non-standard
 689                         # itemtypes.
 690                         $allowedTypes = array(
 691                                 'http://microformats.org/profile/hcard',
 692                                 'http://microformats.org/profile/hcalendar#vevent',
 693                                 'http://n.whatwg.org/work',
 694                         );
 695                         if ( isset( $out['itemtype'] ) && !in_array( $out['itemtype'],
 696                         $allowedTypes ) ) {
 697                                 # Kill everything
 698                                 unset( $out['itemscope'] );
 699                         }
 700                         # itemtype, itemid, itemref don't make sense without itemscope
 701                         if ( !array_key_exists( 'itemscope', $out ) ) {
 702                                 unset( $out['itemtype'] );
 703                                 unset( $out['itemid'] );
 704                                 unset( $out['itemref'] );
 705                         }
 706                         # TODO: Strip itemprop if we aren't descendants of an itemscope.
 707                 }
 708                 return $out;
 709         }
 710
 711         /**
 712          * Merge two sets of HTML attributes.  Conflicting items in the second set
 713          * will override those in the first, except for 'class' attributes which
 714          * will be combined (if they're both strings).
 715          *
 716          * @todo implement merging for other attributes such as style
 717          * @param $a Array
 718          * @param $b Array
 719          * @return array
 720          */
 721         static function mergeAttributes( $a, $b ) {
 722                 $out = array_merge( $a, $b );
 723                 if( isset( $a['class'] ) && isset( $b['class'] )
 724                 && is_string( $a['class'] ) && is_string( $b['class'] )
 725                 && $a['class'] !== $b['class'] ) {
 726                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 727                                 -1, PREG_SPLIT_NO_EMPTY );
 728                         $out['class'] = implode( ' ', array_unique( $classes ) );
 729                 }
 730                 return $out;
 731         }
 732
 733         /**
 734          * Pick apart some CSS and check it for forbidden or unsafe structures.
 735          * Returns a sanitized string, or false if it was just too evil.
 736          *
 737          * Currently URL references, 'expression', 'tps' are forbidden.
 738          *
 739          * @param $value String
 740          * @return Mixed
 741          */
 742         static function checkCss( $value ) {
 743                 $value = Sanitizer::decodeCharReferences( $value );
 744
 745                 // Remove any comments; IE gets token splitting wrong
 746                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 747
 748                 // Decode escape sequences and line continuation
 749                 // See the grammar in the CSS 2 spec, appendix D.
 750                 static $decodeRegex, $reencodeTable;
 751                 if ( !$decodeRegex ) {
 752                         $space = '[\\x20\\t\\r\\n\\f]';
 753                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 754                         $backslash = '\\\\';
 755                         $decodeRegex = "/ $backslash
 756                                 (?:
 757                                         ($nl) |  # 1. Line continuation
 758                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 759                                         (.) | # 3. backslash cancelling special meaning
 760                                         () | # 4. backslash at end of string
 761                                 )/xu";
 762                 }
 763                 $value = preg_replace_callback( $decodeRegex,
 764                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 765
 766                 // Reject problematic keywords and control characters
 767                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 768                         return '/* invalid control char */';
 769                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
 770                         return '/* insecure input */';
 771                 }
 772                 return $value;
 773         }
 774
 775         static function cssDecodeCallback( $matches ) {
 776                 if ( $matches[1] !== '' ) {
 777                         // Line continuation
 778                         return '';
 779                 } elseif ( $matches[2] !== '' ) {
 780                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 781                 } elseif ( $matches[3] !== '' ) {
 782                         $char = $matches[3];
 783                 } else {
 784                         $char = '\\';
 785                 }
 786                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 787                         // These characters need to be escaped in strings
 788                         // Clean up the escape sequence to avoid parsing errors by clients
 789                         return '\\' . dechex( ord( $char ) ) . ' ';
 790                 } else {
 791                         // Decode unnecessary escape
 792                         return $char;
 793                 }
 794         }
 795
 796         /**
 797          * Take a tag soup fragment listing an HTML element's attributes
 798          * and normalize it to well-formed XML, discarding unwanted attributes.
 799          * Output is safe for further wikitext processing, with escaping of
 800          * values that could trigger problems.
 801          *
 802          * - Normalizes attribute names to lowercase
 803          * - Discards attributes not on a whitelist for the given element
 804          * - Turns broken or invalid entities into plaintext
 805          * - Double-quotes all attribute values
 806          * - Attributes without values are given the name as attribute
 807          * - Double attributes are discarded
 808          * - Unsafe style attributes are discarded
 809          * - Prepends space if there are attributes.
 810          *
 811          * @param $text String
 812          * @param $element String
 813          * @return String
 814          */
 815         static function fixTagAttributes( $text, $element ) {
 816                 if( trim( $text ) == '' ) {
 817                         return '';
 818                 }
 819
 820                 $stripped = Sanitizer::validateTagAttributes(
 821                         Sanitizer::decodeTagAttributes( $text ), $element );
 822
 823                 $attribs = array();
 824                 foreach( $stripped as $attribute => $value ) {
 825                         $encAttribute = htmlspecialchars( $attribute );
 826                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 827
 828                         $attribs[] = "$encAttribute=\"$encValue\"";
 829                 }
 830                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 831         }
 832
 833         /**
 834          * Encode an attribute value for HTML output.
 835          * @param $text String
 836          * @return HTML-encoded text fragment
 837          */
 838         static function encodeAttribute( $text ) {
 839                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 840
 841                 // Whitespace is normalized during attribute decoding,
 842                 // so if we've been passed non-spaces we must encode them
 843                 // ahead of time or they won't be preserved.
 844                 $encValue = strtr( $encValue, array(
 845                         "\n" => '&#10;',
 846                         "\r" => '&#13;',
 847                         "\t" => '&#9;',
 848                 ) );
 849
 850                 return $encValue;
 851         }
 852
 853         /**
 854          * Encode an attribute value for HTML tags, with extra armoring
 855          * against further wiki processing.
 856          * @param $text String
 857          * @return HTML-encoded text fragment
 858          */
 859         static function safeEncodeAttribute( $text ) {
 860                 $encValue = Sanitizer::encodeAttribute( $text );
 861
 862                 # Templates and links may be expanded in later parsing,
 863                 # creating invalid or dangerous output. Suppress this.
 864                 $encValue = strtr( $encValue, array(
 865                         '<'    => '&lt;',   // This should never happen,
 866                         '>'    => '&gt;',   // we've received invalid input
 867                         '"'    => '&quot;', // which should have been escaped.
 868                         '{'    => '&#123;',
 869                         '['    => '&#91;',
 870                         "''"   => '&#39;&#39;',
 871                         'ISBN' => '&#73;SBN',
 872                         'RFC'  => '&#82;FC',
 873                         'PMID' => '&#80;MID',
 874                         '|'    => '&#124;',
 875                         '__'   => '&#95;_',
 876                 ) );
 877
 878                 # Stupid hack
 879                 $encValue = preg_replace_callback(
 880                         '/(' . wfUrlProtocols() . ')/',
 881                         array( 'Sanitizer', 'armorLinksCallback' ),
 882                         $encValue );
 883                 return $encValue;
 884         }
 885
 886         /**
 887          * Given a value, escape it so that it can be used in an id attribute and
 888          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
 889          * true, allowing anything but ASCII whitespace.  Otherwise it will use
 890          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
 891          * escaped with lots of dots.
 892          *
 893          * To ensure we don't have to bother escaping anything, we also strip ', ",
 894          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
 895          * We also strip # because it upsets IE6.
 896          *
 897          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
 898          *                                                          in the id and
 899          *                                                          name attributes
 900          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
 901          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
 902          *   HTML5 definition of id attribute
 903          *
 904          * @param $id String: id to escape
 905          * @param $options Mixed: string or array of strings (default is array()):
 906          *   'noninitial': This is a non-initial fragment of an id, not a full id,
 907          *       so don't pay attention if the first character isn't valid at the
 908          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
 909          *       false.
 910          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
 911          *       if $wgExperimentalHtmlIds is used, so we can generate extra
 912          *       anchors and links won't break.
 913          * @return String
 914          */
 915         static function escapeId( $id, $options = array() ) {
 916                 global $wgHtml5, $wgExperimentalHtmlIds;
 917                 $options = (array)$options;
 918
 919                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
 920                         $id = Sanitizer::decodeCharReferences( $id );
 921                         $id = preg_replace( '/[ \t\n\r\f_\'"&#]+/', '_', $id );
 922                         $id = trim( $id, '_' );
 923                         if ( $id === '' ) {
 924                                 # Must have been all whitespace to start with.
 925                                 return '_';
 926                         } else {
 927                                 return $id;
 928                         }
 929                 }
 930
 931                 # HTML4-style escaping
 932                 static $replace = array(
 933                         '%3A' => ':',
 934                         '%' => '.'
 935                 );
 936
 937                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
 938                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
 939
 940                 if ( !preg_match( '/^[a-zA-Z]/', $id )
 941                 && !in_array( 'noninitial', $options ) )  {
 942                         // Initial character must be a letter!
 943                         $id = "x$id";
 944                 }
 945                 return $id;
 946         }
 947
 948         /**
 949          * Given a value, escape it so that it can be used as a CSS class and
 950          * return it.
 951          *
 952          * @todo For extra validity, input should be validated UTF-8.
 953          *
 954          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
 955          *
 956          * @param $class String
 957          * @return String
 958          */
 959         static function escapeClass( $class ) {
 960                 // Convert ugly stuff to underscores and kill underscores in ugly places
 961                 return rtrim(preg_replace(
 962                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
 963                         '_',
 964                         $class ), '_');
 965         }
 966
 967         /**
 968          * Given HTML input, escape with htmlspecialchars but un-escape entites.
 969          * This allows (generally harmless) entities like &nbsp; to survive.
 970          *
 971          * @param $html String to escape
 972          * @return String: escaped input
 973          */
 974         static function escapeHtmlAllowEntities( $html ) {
 975                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
 976                 # hurt.
 977                 $html = htmlspecialchars( $html, ENT_QUOTES );
 978                 $html = str_replace( '&amp;', '&', $html );
 979                 $html = Sanitizer::normalizeCharReferences( $html );
 980                 return $html;
 981         }
 982
 983         /**
 984          * Regex replace callback for armoring links against further processing.
 985          * @param $matches Array
 986          * @return string
 987          */
 988         private static function armorLinksCallback( $matches ) {
 989                 return str_replace( ':', '&#58;', $matches[1] );
 990         }
 991
 992         /**
 993          * Return an associative array of attribute names and values from
 994          * a partial tag string. Attribute names are forces to lowercase,
 995          * character references are decoded to UTF-8 text.
 996          *
 997          * @param $text String
 998          * @return Array
 999          */
1000         public static function decodeTagAttributes( $text ) {
1001                 if( trim( $text ) == '' ) {
1002                         return array();
1003                 }
1004
1005                 $attribs = array();
1006                 $pairs = array();
1007                 if( !preg_match_all(
1008                         MW_ATTRIBS_REGEX,
1009                         $text,
1010                         $pairs,
1011                         PREG_SET_ORDER ) ) {
1012                         return $attribs;
1013                 }
1014
1015                 foreach( $pairs as $set ) {
1016                         $attribute = strtolower( $set[1] );
1017                         $value = Sanitizer::getTagAttributeCallback( $set );
1018
1019                         // Normalize whitespace
1020                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1021                         $value = trim( $value );
1022
1023                         // Decode character references
1024                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1025                 }
1026                 return $attribs;
1027         }
1028
1029         /**
1030          * Pick the appropriate attribute value from a match set from the
1031          * MW_ATTRIBS_REGEX matches.
1032          *
1033          * @param $set Array
1034          * @return String
1035          */
1036         private static function getTagAttributeCallback( $set ) {
1037                 if( isset( $set[6] ) ) {
1038                         # Illegal #XXXXXX color with no quotes.
1039                         return $set[6];
1040                 } elseif( isset( $set[5] ) ) {
1041                         # No quotes.
1042                         return $set[5];
1043                 } elseif( isset( $set[4] ) ) {
1044                         # Single-quoted
1045                         return $set[4];
1046                 } elseif( isset( $set[3] ) ) {
1047                         # Double-quoted
1048                         return $set[3];
1049                 } elseif( !isset( $set[2] ) ) {
1050                         # In XHTML, attributes must have a value.
1051                         # For 'reduced' form, return explicitly the attribute name here.
1052                         return $set[1];
1053                 } else {
1054                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1055                 }
1056         }
1057
1058         /**
1059          * Normalize whitespace and character references in an XML source-
1060          * encoded text for an attribute value.
1061          *
1062          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1063          * but note that we're not returning the value, but are returning
1064          * XML source fragments that will be slapped into output.
1065          *
1066          * @param $text String
1067          * @return String
1068          */
1069         private static function normalizeAttributeValue( $text ) {
1070                 return str_replace( '"', '&quot;',
1071                         self::normalizeWhitespace(
1072                                 Sanitizer::normalizeCharReferences( $text ) ) );
1073         }
1074
1075         private static function normalizeWhitespace( $text ) {
1076                 return preg_replace(
1077                         '/\r\n|[\x20\x0d\x0a\x09]/',
1078                         ' ',
1079                         $text );
1080         }
1081
1082         /**
1083          * Ensure that any entities and character references are legal
1084          * for XML and XHTML specifically. Any stray bits will be
1085          * &amp;-escaped to result in a valid text fragment.
1086          *
1087          * a. any named char refs must be known in XHTML
1088          * b. any numeric char refs must be legal chars, not invalid or forbidden
1089          * c. use &#x, not &#X
1090          * d. fix or reject non-valid attributes
1091          *
1092          * @param $text String
1093          * @return String
1094          * @private
1095          */
1096         static function normalizeCharReferences( $text ) {
1097                 return preg_replace_callback(
1098                         MW_CHAR_REFS_REGEX,
1099                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1100                         $text );
1101         }
1102         /**
1103          * @param $matches String
1104          * @return String
1105          */
1106         static function normalizeCharReferencesCallback( $matches ) {
1107                 $ret = null;
1108                 if( $matches[1] != '' ) {
1109                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1110                 } elseif( $matches[2] != '' ) {
1111                         $ret = Sanitizer::decCharReference( $matches[2] );
1112                 } elseif( $matches[3] != ''  ) {
1113                         $ret = Sanitizer::hexCharReference( $matches[3] );
1114                 } elseif( $matches[4] != '' ) {
1115                         $ret = Sanitizer::hexCharReference( $matches[4] );
1116                 }
1117                 if( is_null( $ret ) ) {
1118                         return htmlspecialchars( $matches[0] );
1119                 } else {
1120                         return $ret;
1121                 }
1122         }
1123
1124         /**
1125          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1126          * return the named entity reference as is. If the entity is a
1127          * MediaWiki-specific alias, returns the HTML equivalent. Otherwise,
1128          * returns HTML-escaped text of pseudo-entity source (eg &amp;foo;)
1129          *
1130          * @param $name String
1131          * @return String
1132          */
1133         static function normalizeEntity( $name ) {
1134                 global $wgHtmlEntities, $wgHtmlEntityAliases;
1135                 if ( isset( $wgHtmlEntityAliases[$name] ) ) {
1136                         return "&{$wgHtmlEntityAliases[$name]};";
1137                 } elseif( isset( $wgHtmlEntities[$name] ) ) {
1138                         return "&$name;";
1139                 } else {
1140                         return "&amp;$name;";
1141                 }
1142         }
1143
1144         static function decCharReference( $codepoint ) {
1145                 $point = intval( $codepoint );
1146                 if( Sanitizer::validateCodepoint( $point ) ) {
1147                         return sprintf( '&#%d;', $point );
1148                 } else {
1149                         return null;
1150                 }
1151         }
1152
1153         static function hexCharReference( $codepoint ) {
1154                 $point = hexdec( $codepoint );
1155                 if( Sanitizer::validateCodepoint( $point ) ) {
1156                         return sprintf( '&#x%x;', $point );
1157                 } else {
1158                         return null;
1159                 }
1160         }
1161
1162         /**
1163          * Returns true if a given Unicode codepoint is a valid character in XML.
1164          * @param $codepoint Integer
1165          * @return Boolean
1166          */
1167         private static function validateCodepoint( $codepoint ) {
1168                 return ($codepoint ==    0x09)
1169                         || ($codepoint ==    0x0a)
1170                         || ($codepoint ==    0x0d)
1171                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1172                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1173                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1174         }
1175
1176         /**
1177          * Decode any character references, numeric or named entities,
1178          * in the text and return a UTF-8 string.
1179          *
1180          * @param $text String
1181          * @return String
1182          */
1183         public static function decodeCharReferences( $text ) {
1184                 return preg_replace_callback(
1185                         MW_CHAR_REFS_REGEX,
1186                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1187                         $text );
1188         }
1189
1190         /**
1191          * Decode any character references, numeric or named entities,
1192          * in the next and normalize the resulting string. (bug 14952)
1193          *
1194          * This is useful for page titles, not for text to be displayed,
1195          * MediaWiki allows HTML entities to escape normalization as a feature.
1196          *
1197          * @param $text String (already normalized, containing entities)
1198          * @return String (still normalized, without entities)
1199          */
1200         public static function decodeCharReferencesAndNormalize( $text ) {
1201                 global $wgContLang;
1202                 $text = preg_replace_callback(
1203                         MW_CHAR_REFS_REGEX,
1204                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1205                         $text, /* limit */ -1, $count );
1206
1207                 if ( $count ) {
1208                         return $wgContLang->normalize( $text );
1209                 } else {
1210                         return $text;
1211                 }
1212         }
1213
1214         /**
1215          * @param $matches String
1216          * @return String
1217          */
1218         static function decodeCharReferencesCallback( $matches ) {
1219                 if( $matches[1] != '' ) {
1220                         return Sanitizer::decodeEntity( $matches[1] );
1221                 } elseif( $matches[2] != '' ) {
1222                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1223                 } elseif( $matches[3] != ''  ) {
1224                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1225                 } elseif( $matches[4] != '' ) {
1226                         return  Sanitizer::decodeChar( hexdec( $matches[4] ) );
1227                 }
1228                 # Last case should be an ampersand by itself
1229                 return $matches[0];
1230         }
1231
1232         /**
1233          * Return UTF-8 string for a codepoint if that is a valid
1234          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1235          * @param $codepoint Integer
1236          * @return String
1237          * @private
1238          */
1239         static function decodeChar( $codepoint ) {
1240                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1241                         return codepointToUtf8( $codepoint );
1242                 } else {
1243                         return UTF8_REPLACEMENT;
1244                 }
1245         }
1246
1247         /**
1248          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1249          * return the UTF-8 encoding of that character. Otherwise, returns
1250          * pseudo-entity source (eg &foo;)
1251          *
1252          * @param $name Strings
1253          * @return String
1254          */
1255         static function decodeEntity( $name ) {
1256                 global $wgHtmlEntities, $wgHtmlEntityAliases;
1257                 if ( isset( $wgHtmlEntityAliases[$name] ) ) {
1258                         $name = $wgHtmlEntityAliases[$name];
1259                 }
1260                 if( isset( $wgHtmlEntities[$name] ) ) {
1261                         return codepointToUtf8( $wgHtmlEntities[$name] );
1262                 } else {
1263                         return "&$name;";
1264                 }
1265         }
1266
1267         /**
1268          * Fetch the whitelist of acceptable attributes for a given element name.
1269          *
1270          * @param $element String
1271          * @return Array
1272          */
1273         static function attributeWhitelist( $element ) {
1274                 static $list;
1275                 if( !isset( $list ) ) {
1276                         $list = Sanitizer::setupAttributeWhitelist();
1277                 }
1278                 return isset( $list[$element] )
1279                         ? $list[$element]
1280                         : array();
1281         }
1282
1283         /**
1284          * Foreach array key (an allowed HTML element), return an array
1285          * of allowed attributes
1286          * @return Array
1287          */
1288         static function setupAttributeWhitelist() {
1289                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1290
1291                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1292
1293                 if ( $wgAllowRdfaAttributes ) {
1294                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1295                         $common = array_merge( $common, array(
1296                             'about', 'property', 'resource', 'datatype', 'typeof',
1297                         ) );
1298                 }
1299
1300                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1301                         # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1302                         $common = array_merge( $common, array(
1303                             'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1304                         ) );
1305                 }
1306
1307                 $block = array_merge( $common, array( 'align' ) );
1308                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1309                 $tablecell = array( 'abbr',
1310                                     'axis',
1311                                     'headers',
1312                                     'scope',
1313                                     'rowspan',
1314                                     'colspan',
1315                                     'nowrap', # deprecated
1316                                     'width',  # deprecated
1317                                     'height', # deprecated
1318                                     'bgcolor' # deprecated
1319                                     );
1320
1321                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1322                 # See: http://www.w3.org/TR/html4/
1323                 $whitelist = array (
1324                         # 7.5.4
1325                         'div'        => $block,
1326                         'center'     => $common, # deprecated
1327                         'span'       => $block, # ??
1328
1329                         # 7.5.5
1330                         'h1'         => $block,
1331                         'h2'         => $block,
1332                         'h3'         => $block,
1333                         'h4'         => $block,
1334                         'h5'         => $block,
1335                         'h6'         => $block,
1336
1337                         # 7.5.6
1338                         # address
1339
1340                         # 8.2.4
1341                         # bdo
1342
1343                         # 9.2.1
1344                         'em'         => $common,
1345                         'strong'     => $common,
1346                         'cite'       => $common,
1347                         # dfn
1348                         'code'       => $common,
1349                         # samp
1350                         # kbd
1351                         'var'        => $common,
1352                         'abbr'       => $common,
1353                         # acronym
1354
1355                         # 9.2.2
1356                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1357                         # q
1358
1359                         # 9.2.3
1360                         'sub'        => $common,
1361                         'sup'        => $common,
1362
1363                         # 9.3.1
1364                         'p'          => $block,
1365
1366                         # 9.3.2
1367                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1368
1369                         # 9.3.4
1370                         'pre'        => array_merge( $common, array( 'width' ) ),
1371
1372                         # 9.4
1373                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1374                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1375
1376                         # 10.2
1377                         'ul'         => array_merge( $common, array( 'type' ) ),
1378                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1379                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1380
1381                         # 10.3
1382                         'dl'         => $common,
1383                         'dd'         => $common,
1384                         'dt'         => $common,
1385
1386                         # 11.2.1
1387                         'table'      => array_merge( $common,
1388                                                                 array( 'summary', 'width', 'border', 'frame',
1389                                                                                 'rules', 'cellspacing', 'cellpadding',
1390                                                                                 'align', 'bgcolor',
1391                                                                 ) ),
1392
1393                         # 11.2.2
1394                         'caption'    => array_merge( $common, array( 'align' ) ),
1395
1396                         # 11.2.3
1397                         'thead'      => array_merge( $common, $tablealign ),
1398                         'tfoot'      => array_merge( $common, $tablealign ),
1399                         'tbody'      => array_merge( $common, $tablealign ),
1400
1401                         # 11.2.4
1402                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1403                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1404
1405                         # 11.2.5
1406                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1407
1408                         # 11.2.6
1409                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1410                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1411
1412                         # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1413                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1414
1415                         # 13.2
1416                         # Not usually allowed, but may be used for extension-style hooks
1417                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1418                         # true
1419                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1420
1421                         # 15.2.1
1422                         'tt'         => $common,
1423                         'b'          => $common,
1424                         'i'          => $common,
1425                         'big'        => $common,
1426                         'small'      => $common,
1427                         'strike'     => $common,
1428                         's'          => $common,
1429                         'u'          => $common,
1430
1431                         # 15.2.2
1432                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1433                         # basefont
1434
1435                         # 15.3
1436                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1437
1438                         # XHTML Ruby annotation text module, simple ruby only.
1439                         # http://www.w3c.org/TR/ruby/
1440                         'ruby'       => $common,
1441                         # rbc
1442                         # rtc
1443                         'rb'         => $common,
1444                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1445                         'rp'         => $common,
1446
1447                         # MathML root element, where used for extensions
1448                         # 'title' may not be 100% valid here; it's XHTML
1449                         # http://www.w3.org/TR/REC-MathML/
1450                         'math'       => array( 'class', 'style', 'id', 'title' ),
1451                         );
1452                 return $whitelist;
1453         }
1454
1455         /**
1456          * Take a fragment of (potentially invalid) HTML and return
1457          * a version with any tags removed, encoded as plain text.
1458          *
1459          * Warning: this return value must be further escaped for literal
1460          * inclusion in HTML output as of 1.10!
1461          *
1462          * @param $text String: HTML fragment
1463          * @return String
1464          */
1465         static function stripAllTags( $text ) {
1466                 # Actual <tags>
1467                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1468
1469                 # Normalize &entities and whitespace
1470                 $text = self::decodeCharReferences( $text );
1471                 $text = self::normalizeWhitespace( $text );
1472
1473                 return $text;
1474         }
1475
1476         /**
1477          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1478          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1479          * PHP 5.1 doesn't.
1480          *
1481          * Use for passing XHTML fragments to PHP's XML parsing functions
1482          *
1483          * @return String
1484          */
1485         static function hackDocType() {
1486                 global $wgHtmlEntities;
1487                 $out = "<!DOCTYPE html [\n";
1488                 foreach( $wgHtmlEntities as $entity => $codepoint ) {
1489                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1490                 }
1491                 $out .= "]>\n";
1492                 return $out;
1493         }
1494
1495         static function cleanUrl( $url ) {
1496                 # Normalize any HTML entities in input. They will be
1497                 # re-escaped by makeExternalLink().
1498                 $url = Sanitizer::decodeCharReferences( $url );
1499
1500                 # Escape any control characters introduced by the above step
1501                 $url = preg_replace( '/[\][<>"\\x00-\\x20\\x7F\|]/e', "urlencode('\\0')", $url );
1502
1503                 # Validate hostname portion
1504                 $matches = array();
1505                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1506                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1507
1508                         // Characters that will be ignored in IDNs.
1509                         // http://tools.ietf.org/html/3454#section-3.1
1510                         // Strip them before further processing so blacklists and such work.
1511                         $strip = "/
1512                                 \\s|          # general whitespace
1513                                 \xc2\xad|     # 00ad SOFT HYPHEN
1514                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1515                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1516                                 \xe2\x81\xa0| # 2060 WORD JOINER
1517                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1518                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1519                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1520                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1521                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1522                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1523                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1524                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe00f VARIATION SELECTOR-1-16
1525                                 /xuD";
1526
1527                         $host = preg_replace( $strip, '', $host );
1528
1529                         // @todo Fixme: validate hostnames here
1530
1531                         return $protocol . $host . $rest;
1532                 } else {
1533                         return $url;
1534                 }
1535         }
1536
1537 }