includes/Sanitizer.php

   1 <?php
   2 /**
   3  * XHTML sanitizer for MediaWiki
   4  *
   5  * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
   6  * http://www.mediawiki.org/
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License as published by
  10  * the Free Software Foundation; either version 2 of the License, or
  11  * (at your option) any later version.
  12  *
  13  * This program is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  16  * GNU General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU General Public License along
  19  * with this program; if not, write to the Free Software Foundation, Inc.,
  20  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  21  * http://www.gnu.org/copyleft/gpl.html
  22  *
  23  * @file
  24  * @ingroup Parser
  25  */
  26
  27 /**
  28  * XHTML sanitizer for MediaWiki
  29  * @ingroup Parser
  30  */
  31 class Sanitizer {
  32         /**
  33          * Regular expression to match various types of character references in
  34          * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
  35          */
  36         const CHAR_REFS_REGEX =
  37                 '/&([A-Za-z0-9\x80-\xff]+);
  38                  |&\#([0-9]+);
  39                  |&\#[xX]([0-9A-Fa-f]+);
  40                  |(&)/x';
  41
  42         const EVIL_URI_PATTERN = '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
  43         const XMLNS_ATTRIBUTE_PATTERN = "/^xmlns:[:A-Z_a-z-.0-9]+$/";
  44
  45         /**
  46          * List of all named character entities defined in HTML 4.01
  47          * http://www.w3.org/TR/html4/sgml/entities.html
  48          * As well as &apos; which is only defined starting in XHTML1.
  49          * @private
  50          */
  51         static $htmlEntities = array(
  52                 'Aacute'   => 193,
  53                 'aacute'   => 225,
  54                 'Acirc'    => 194,
  55                 'acirc'    => 226,
  56                 'acute'    => 180,
  57                 'AElig'    => 198,
  58                 'aelig'    => 230,
  59                 'Agrave'   => 192,
  60                 'agrave'   => 224,
  61                 'alefsym'  => 8501,
  62                 'Alpha'    => 913,
  63                 'alpha'    => 945,
  64                 'amp'      => 38,
  65                 'and'      => 8743,
  66                 'ang'      => 8736,
  67                 'apos'     => 39,
  68                 'Aring'    => 197,
  69                 'aring'    => 229,
  70                 'asymp'    => 8776,
  71                 'Atilde'   => 195,
  72                 'atilde'   => 227,
  73                 'Auml'     => 196,
  74                 'auml'     => 228,
  75                 'bdquo'    => 8222,
  76                 'Beta'     => 914,
  77                 'beta'     => 946,
  78                 'brvbar'   => 166,
  79                 'bull'     => 8226,
  80                 'cap'      => 8745,
  81                 'Ccedil'   => 199,
  82                 'ccedil'   => 231,
  83                 'cedil'    => 184,
  84                 'cent'     => 162,
  85                 'Chi'      => 935,
  86                 'chi'      => 967,
  87                 'circ'     => 710,
  88                 'clubs'    => 9827,
  89                 'cong'     => 8773,
  90                 'copy'     => 169,
  91                 'crarr'    => 8629,
  92                 'cup'      => 8746,
  93                 'curren'   => 164,
  94                 'dagger'   => 8224,
  95                 'Dagger'   => 8225,
  96                 'darr'     => 8595,
  97                 'dArr'     => 8659,
  98                 'deg'      => 176,
  99                 'Delta'    => 916,
 100                 'delta'    => 948,
 101                 'diams'    => 9830,
 102                 'divide'   => 247,
 103                 'Eacute'   => 201,
 104                 'eacute'   => 233,
 105                 'Ecirc'    => 202,
 106                 'ecirc'    => 234,
 107                 'Egrave'   => 200,
 108                 'egrave'   => 232,
 109                 'empty'    => 8709,
 110                 'emsp'     => 8195,
 111                 'ensp'     => 8194,
 112                 'Epsilon'  => 917,
 113                 'epsilon'  => 949,
 114                 'equiv'    => 8801,
 115                 'Eta'      => 919,
 116                 'eta'      => 951,
 117                 'ETH'      => 208,
 118                 'eth'      => 240,
 119                 'Euml'     => 203,
 120                 'euml'     => 235,
 121                 'euro'     => 8364,
 122                 'exist'    => 8707,
 123                 'fnof'     => 402,
 124                 'forall'   => 8704,
 125                 'frac12'   => 189,
 126                 'frac14'   => 188,
 127                 'frac34'   => 190,
 128                 'frasl'    => 8260,
 129                 'Gamma'    => 915,
 130                 'gamma'    => 947,
 131                 'ge'       => 8805,
 132                 'gt'       => 62,
 133                 'harr'     => 8596,
 134                 'hArr'     => 8660,
 135                 'hearts'   => 9829,
 136                 'hellip'   => 8230,
 137                 'Iacute'   => 205,
 138                 'iacute'   => 237,
 139                 'Icirc'    => 206,
 140                 'icirc'    => 238,
 141                 'iexcl'    => 161,
 142                 'Igrave'   => 204,
 143                 'igrave'   => 236,
 144                 'image'    => 8465,
 145                 'infin'    => 8734,
 146                 'int'      => 8747,
 147                 'Iota'     => 921,
 148                 'iota'     => 953,
 149                 'iquest'   => 191,
 150                 'isin'     => 8712,
 151                 'Iuml'     => 207,
 152                 'iuml'     => 239,
 153                 'Kappa'    => 922,
 154                 'kappa'    => 954,
 155                 'Lambda'   => 923,
 156                 'lambda'   => 955,
 157                 'lang'     => 9001,
 158                 'laquo'    => 171,
 159                 'larr'     => 8592,
 160                 'lArr'     => 8656,
 161                 'lceil'    => 8968,
 162                 'ldquo'    => 8220,
 163                 'le'       => 8804,
 164                 'lfloor'   => 8970,
 165                 'lowast'   => 8727,
 166                 'loz'      => 9674,
 167                 'lrm'      => 8206,
 168                 'lsaquo'   => 8249,
 169                 'lsquo'    => 8216,
 170                 'lt'       => 60,
 171                 'macr'     => 175,
 172                 'mdash'    => 8212,
 173                 'micro'    => 181,
 174                 'middot'   => 183,
 175                 'minus'    => 8722,
 176                 'Mu'       => 924,
 177                 'mu'       => 956,
 178                 'nabla'    => 8711,
 179                 'nbsp'     => 160,
 180                 'ndash'    => 8211,
 181                 'ne'       => 8800,
 182                 'ni'       => 8715,
 183                 'not'      => 172,
 184                 'notin'    => 8713,
 185                 'nsub'     => 8836,
 186                 'Ntilde'   => 209,
 187                 'ntilde'   => 241,
 188                 'Nu'       => 925,
 189                 'nu'       => 957,
 190                 'Oacute'   => 211,
 191                 'oacute'   => 243,
 192                 'Ocirc'    => 212,
 193                 'ocirc'    => 244,
 194                 'OElig'    => 338,
 195                 'oelig'    => 339,
 196                 'Ograve'   => 210,
 197                 'ograve'   => 242,
 198                 'oline'    => 8254,
 199                 'Omega'    => 937,
 200                 'omega'    => 969,
 201                 'Omicron'  => 927,
 202                 'omicron'  => 959,
 203                 'oplus'    => 8853,
 204                 'or'       => 8744,
 205                 'ordf'     => 170,
 206                 'ordm'     => 186,
 207                 'Oslash'   => 216,
 208                 'oslash'   => 248,
 209                 'Otilde'   => 213,
 210                 'otilde'   => 245,
 211                 'otimes'   => 8855,
 212                 'Ouml'     => 214,
 213                 'ouml'     => 246,
 214                 'para'     => 182,
 215                 'part'     => 8706,
 216                 'permil'   => 8240,
 217                 'perp'     => 8869,
 218                 'Phi'      => 934,
 219                 'phi'      => 966,
 220                 'Pi'       => 928,
 221                 'pi'       => 960,
 222                 'piv'      => 982,
 223                 'plusmn'   => 177,
 224                 'pound'    => 163,
 225                 'prime'    => 8242,
 226                 'Prime'    => 8243,
 227                 'prod'     => 8719,
 228                 'prop'     => 8733,
 229                 'Psi'      => 936,
 230                 'psi'      => 968,
 231                 'quot'     => 34,
 232                 'radic'    => 8730,
 233                 'rang'     => 9002,
 234                 'raquo'    => 187,
 235                 'rarr'     => 8594,
 236                 'rArr'     => 8658,
 237                 'rceil'    => 8969,
 238                 'rdquo'    => 8221,
 239                 'real'     => 8476,
 240                 'reg'      => 174,
 241                 'rfloor'   => 8971,
 242                 'Rho'      => 929,
 243                 'rho'      => 961,
 244                 'rlm'      => 8207,
 245                 'rsaquo'   => 8250,
 246                 'rsquo'    => 8217,
 247                 'sbquo'    => 8218,
 248                 'Scaron'   => 352,
 249                 'scaron'   => 353,
 250                 'sdot'     => 8901,
 251                 'sect'     => 167,
 252                 'shy'      => 173,
 253                 'Sigma'    => 931,
 254                 'sigma'    => 963,
 255                 'sigmaf'   => 962,
 256                 'sim'      => 8764,
 257                 'spades'   => 9824,
 258                 'sub'      => 8834,
 259                 'sube'     => 8838,
 260                 'sum'      => 8721,
 261                 'sup'      => 8835,
 262                 'sup1'     => 185,
 263                 'sup2'     => 178,
 264                 'sup3'     => 179,
 265                 'supe'     => 8839,
 266                 'szlig'    => 223,
 267                 'Tau'      => 932,
 268                 'tau'      => 964,
 269                 'there4'   => 8756,
 270                 'Theta'    => 920,
 271                 'theta'    => 952,
 272                 'thetasym' => 977,
 273                 'thinsp'   => 8201,
 274                 'THORN'    => 222,
 275                 'thorn'    => 254,
 276                 'tilde'    => 732,
 277                 'times'    => 215,
 278                 'trade'    => 8482,
 279                 'Uacute'   => 218,
 280                 'uacute'   => 250,
 281                 'uarr'     => 8593,
 282                 'uArr'     => 8657,
 283                 'Ucirc'    => 219,
 284                 'ucirc'    => 251,
 285                 'Ugrave'   => 217,
 286                 'ugrave'   => 249,
 287                 'uml'      => 168,
 288                 'upsih'    => 978,
 289                 'Upsilon'  => 933,
 290                 'upsilon'  => 965,
 291                 'Uuml'     => 220,
 292                 'uuml'     => 252,
 293                 'weierp'   => 8472,
 294                 'Xi'       => 926,
 295                 'xi'       => 958,
 296                 'Yacute'   => 221,
 297                 'yacute'   => 253,
 298                 'yen'      => 165,
 299                 'Yuml'     => 376,
 300                 'yuml'     => 255,
 301                 'Zeta'     => 918,
 302                 'zeta'     => 950,
 303                 'zwj'      => 8205,
 304                 'zwnj'     => 8204
 305         );
 306
 307         /**
 308          * Character entity aliases accepted by MediaWiki
 309          */
 310         static $htmlEntityAliases = array(
 311                 'רלמ' => 'rlm',
 312                 'رلم' => 'rlm',
 313         );
 314
 315         /**
 316          * Lazy-initialised attributes regex, see getAttribsRegex()
 317          */
 318         static $attribsRegex;
 319
 320         /**
 321          * Regular expression to match HTML/XML attribute pairs within a tag.
 322          * Allows some... latitude.
 323          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 324          */
 325         static function getAttribsRegex() {
 326                 if ( self::$attribsRegex === null ) {
 327                         $attribFirst = '[:A-Z_a-z0-9]';
 328                         $attrib = '[:A-Z_a-z-.0-9]';
 329                         $space = '[\x09\x0a\x0d\x20]';
 330                         self::$attribsRegex =
 331                                 "/(?:^|$space)({$attribFirst}{$attrib}*)
 332                                   ($space*=$space*
 333                                         (?:
 334                                          # The attribute value: quoted or alone
 335                                           \"([^<\"]*)\"
 336                                          | '([^<']*)'
 337                                          |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
 338                                          |  (\#[0-9a-fA-F]+) # Technically wrong, but lots of
 339                                                                                  # colors are specified like this.
 340                                                                                  # We'll be normalizing it.
 341                                         )
 342                                 )?(?=$space|\$)/sx";
 343                 }
 344                 return self::$attribsRegex;
 345         }
 346
 347         /**
 348          * Cleans up HTML, removes dangerous tags and attributes, and
 349          * removes HTML comments
 350          * @private
 351          * @param $text String
 352          * @param $processCallback Callback to do any variable or parameter replacements in HTML attribute values
 353          * @param $args Array for the processing callback
 354          * @param $extratags Array for any extra tags to include
 355          * @param $removetags Array for any tags (default or extra) to exclude
 356          * @return string
 357          */
 358         static function removeHTMLtags( $text, $processCallback = null, $args = array(), $extratags = array(), $removetags = array() ) {
 359                 global $wgUseTidy;
 360
 361                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 362                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 363
 364                 wfProfileIn( __METHOD__ );
 365
 366                 if ( !$staticInitialised ) {
 367
 368                         $htmlpairsStatic = array( # Tags that must be closed
 369                                 'b', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 370                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 371                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 372                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 373                                 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'abbr', 'dfn',
 374                                 'kbd', 'samp', 'thead', 'tbody', 'tfoot'
 375                         );
 376                         $htmlsingle = array(
 377                                 'br', 'hr', 'li', 'dt', 'dd'
 378                         );
 379                         $htmlsingleonly = array( # Elements that cannot have close tags
 380                                 'br', 'hr'
 381                         );
 382                         $htmlnest = array( # Tags that can be nested--??
 383                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 384                                 'dl', 'font', 'big', 'small', 'sub', 'sup', 'span'
 385                         );
 386                         $tabletags = array( # Can only appear inside table, we will close them
 387                                 'td', 'th', 'tr',
 388                         );
 389                         $htmllist = array( # Tags used by list
 390                                 'ul','ol',
 391                         );
 392                         $listtags = array( # Tags that can appear in a list
 393                                 'li',
 394                         );
 395
 396                         global $wgAllowImageTag;
 397                         if ( $wgAllowImageTag ) {
 398                                 $htmlsingle[] = 'img';
 399                                 $htmlsingleonly[] = 'img';
 400                         }
 401
 402                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 403                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 404
 405                         # Convert them all to hashtables for faster lookup
 406                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 407                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 408                         foreach ( $vars as $var ) {
 409                                 $$var = array_flip( $$var );
 410                         }
 411                         $staticInitialised = true;
 412                 }
 413                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 414                 $extratags = array_flip( $extratags );
 415                 $removetags = array_flip( $removetags );
 416                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 417                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ) , $removetags );
 418
 419                 # Remove HTML comments
 420                 $text = Sanitizer::removeHTMLcomments( $text );
 421                 $bits = explode( '<', $text );
 422                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 423                 if ( !$wgUseTidy ) {
 424                         $tagstack = $tablestack = array();
 425                         foreach ( $bits as $x ) {
 426                                 $regs = array();
 427                                 # $slash: Does the current element start with a '/'?
 428                                 # $t: Current element name
 429                                 # $params: String between element name and >
 430                                 # $brace: Ending '>' or '/>'
 431                                 # $rest: Everything until the next element of $bits
 432                                 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 433                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 434                                 } else {
 435                                         $slash = $t = $params = $brace = $rest = null;
 436                                 }
 437
 438                                 $badtag = false;
 439                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 440                                         # Check our stack
 441                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 442                                                 $badtag = true;
 443                                         } elseif ( $slash ) {
 444                                                 # Closing a tag... is it the one we just opened?
 445                                                 $ot = @array_pop( $tagstack );
 446                                                 if ( $ot != $t ) {
 447                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 448                                                                 # Pop all elements with an optional close tag
 449                                                                 # and see if we find a match below them
 450                                                                 $optstack = array();
 451                                                                 array_push( $optstack, $ot );
 452                                                                 $ot = @array_pop( $tagstack );
 453                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 454                                                                         array_push( $optstack, $ot );
 455                                                                         $ot = @array_pop( $tagstack );
 456                                                                 }
 457                                                                 if ( $t != $ot ) {
 458                                                                         # No match. Push the optional elements back again
 459                                                                         $badtag = true;
 460                                                                         while ( $ot = @array_pop( $optstack ) ) {
 461                                                                                 array_push( $tagstack, $ot );
 462                                                                         }
 463                                                                 }
 464                                                         } else {
 465                                                                 @array_push( $tagstack, $ot );
 466                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 467                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 468                                                                         $badtag = true;
 469                                                                 }
 470                                                         }
 471                                                 } else {
 472                                                         if ( $t == 'table' ) {
 473                                                                 $tagstack = array_pop( $tablestack );
 474                                                         }
 475                                                 }
 476                                                 $newparams = '';
 477                                         } else {
 478                                                 # Keep track for later
 479                                                 if ( isset( $tabletags[$t] ) &&
 480                                                 !in_array( 'table', $tagstack ) ) {
 481                                                         $badtag = true;
 482                                                 } elseif ( in_array( $t, $tagstack ) &&
 483                                                 !isset( $htmlnest [$t ] ) ) {
 484                                                         $badtag = true;
 485                                                 # Is it a self closed htmlpair ? (bug 5487)
 486                                                 } elseif ( $brace == '/>' &&
 487                                                 isset( $htmlpairs[$t] ) ) {
 488                                                         $badtag = true;
 489                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 490                                                         # Hack to force empty tag for uncloseable elements
 491                                                         $brace = '/>';
 492                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 493                                                         # Hack to not close $htmlsingle tags
 494                                                         $brace = null;
 495                                                 } elseif ( isset( $tabletags[$t] )
 496                                                 && in_array( $t, $tagstack ) ) {
 497                                                         // New table tag but forgot to close the previous one
 498                                                         $text .= "</$t>";
 499                                                 } else {
 500                                                         if ( $t == 'table' ) {
 501                                                                 array_push( $tablestack, $tagstack );
 502                                                                 $tagstack = array();
 503                                                         }
 504                                                         array_push( $tagstack, $t );
 505                                                 }
 506
 507                                                 # Replace any variables or template parameters with
 508                                                 # plaintext results.
 509                                                 if( is_callable( $processCallback ) ) {
 510                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 511                                                 }
 512
 513                                                 # Strip non-approved attributes from the tag
 514                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 515                                         }
 516                                         if ( !$badtag ) {
 517                                                 $rest = str_replace( '>', '&gt;', $rest );
 518                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 519                                                 $text .= "<$slash$t$newparams$close>$rest";
 520                                                 continue;
 521                                         }
 522                                 }
 523                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 524                         }
 525                         # Close off any remaining tags
 526                         while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
 527                                 $text .= "</$t>\n";
 528                                 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
 529                         }
 530                 } else {
 531                         # this might be possible using tidy itself
 532                         foreach ( $bits as $x ) {
 533                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 534                                 $x, $regs );
 535                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 536                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 537                                         if( is_callable( $processCallback ) ) {
 538                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 539                                         }
 540                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 541                                         $rest = str_replace( '>', '&gt;', $rest );
 542                                         $text .= "<$slash$t$newparams$brace$rest";
 543                                 } else {
 544                                         $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 545                                 }
 546                         }
 547                 }
 548                 wfProfileOut( __METHOD__ );
 549                 return $text;
 550         }
 551
 552         /**
 553          * Remove '<!--', '-->', and everything between.
 554          * To avoid leaving blank lines, when a comment is both preceded
 555          * and followed by a newline (ignoring spaces), trim leading and
 556          * trailing spaces and one of the newlines.
 557          *
 558          * @private
 559          * @param $text String
 560          * @return string
 561          */
 562         static function removeHTMLcomments( $text ) {
 563                 wfProfileIn( __METHOD__ );
 564                 while (($start = strpos($text, '<!--')) !== false) {
 565                         $end = strpos($text, '-->', $start + 4);
 566                         if ($end === false) {
 567                                 # Unterminated comment; bail out
 568                                 break;
 569                         }
 570
 571                         $end += 3;
 572
 573                         # Trim space and newline if the comment is both
 574                         # preceded and followed by a newline
 575                         $spaceStart = max($start - 1, 0);
 576                         $spaceLen = $end - $spaceStart;
 577                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 578                                 $spaceStart--;
 579                                 $spaceLen++;
 580                         }
 581                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 582                                 $spaceLen++;
 583                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 584                                 # Remove the comment, leading and trailing
 585                                 # spaces, and leave only one newline.
 586                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 587                         }
 588                         else {
 589                                 # Remove just the comment.
 590                                 $text = substr_replace($text, '', $start, $end - $start);
 591                         }
 592                 }
 593                 wfProfileOut( __METHOD__ );
 594                 return $text;
 595         }
 596
 597         /**
 598          * Take an array of attribute names and values and normalize or discard
 599          * illegal values for the given element type.
 600          *
 601          * - Discards attributes not on a whitelist for the given element
 602          * - Unsafe style attributes are discarded
 603          * - Invalid id attributes are reencoded
 604          *
 605          * @param $attribs Array
 606          * @param $element String
 607          * @return Array
 608          *
 609          * @todo Check for legal values where the DTD limits things.
 610          * @todo Check for unique id attribute :P
 611          */
 612         static function validateTagAttributes( $attribs, $element ) {
 613                 return Sanitizer::validateAttributes( $attribs,
 614                         Sanitizer::attributeWhitelist( $element ) );
 615         }
 616
 617         /**
 618          * Take an array of attribute names and values and normalize or discard
 619          * illegal values for the given whitelist.
 620          *
 621          * - Discards attributes not the given whitelist
 622          * - Unsafe style attributes are discarded
 623          * - Invalid id attributes are reencoded
 624          *
 625          * @param $attribs Array
 626          * @param $whitelist Array: list of allowed attribute names
 627          * @return Array
 628          *
 629          * @todo Check for legal values where the DTD limits things.
 630          * @todo Check for unique id attribute :P
 631          */
 632         static function validateAttributes( $attribs, $whitelist ) {
 633                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
 634
 635                 $whitelist = array_flip( $whitelist );
 636                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 637
 638                 $out = array();
 639                 foreach( $attribs as $attribute => $value ) {
 640                         #allow XML namespace declaration if RDFa is enabled
 641                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 642                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 643                                         $out[$attribute] = $value;
 644                                 }
 645
 646                                 continue;
 647                         }
 648
 649                         # Allow any attribute beginning with "data-", if in HTML5 mode
 650                         if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
 651                                 continue;
 652                         }
 653
 654                         # Strip javascript "expression" from stylesheets.
 655                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 656                         if( $attribute == 'style' ) {
 657                                 $value = Sanitizer::checkCss( $value );
 658                         }
 659
 660                         if ( $attribute === 'id' ) {
 661                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 662                         }
 663
 664                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 665                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 666                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 667                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 668                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 669                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 670
 671                                 //Paranoia. Allow "simple" values but suppress javascript
 672                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 673                                         continue;
 674                                 }
 675                         }
 676
 677                         # NOTE: even though elements using href/src are not allowed directly, supply
 678                         #       validation code that can be used by tag hook handlers, etc
 679                         if ( $attribute === 'href' || $attribute === 'src' ) {
 680                                 if ( !preg_match( $hrefExp, $value ) ) {
 681                                         continue; //drop any href or src attributes not using an allowed protocol.
 682                                                   //NOTE: this also drops all relative URLs
 683                                 }
 684                         }
 685
 686                         // If this attribute was previously set, override it.
 687                         // Output should only have one attribute of each name.
 688                         $out[$attribute] = $value;
 689                 }
 690
 691                 if ( $wgAllowMicrodataAttributes ) {
 692                         # itemtype, itemid, itemref don't make sense without itemscope
 693                         if ( !array_key_exists( 'itemscope', $out ) ) {
 694                                 unset( $out['itemtype'] );
 695                                 unset( $out['itemid'] );
 696                                 unset( $out['itemref'] );
 697                         }
 698                         # TODO: Strip itemprop if we aren't descendants of an itemscope.
 699                 }
 700                 return $out;
 701         }
 702
 703         /**
 704          * Merge two sets of HTML attributes.  Conflicting items in the second set
 705          * will override those in the first, except for 'class' attributes which
 706          * will be combined (if they're both strings).
 707          *
 708          * @todo implement merging for other attributes such as style
 709          * @param $a Array
 710          * @param $b Array
 711          * @return array
 712          */
 713         static function mergeAttributes( $a, $b ) {
 714                 $out = array_merge( $a, $b );
 715                 if( isset( $a['class'] ) && isset( $b['class'] )
 716                 && is_string( $a['class'] ) && is_string( $b['class'] )
 717                 && $a['class'] !== $b['class'] ) {
 718                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 719                                 -1, PREG_SPLIT_NO_EMPTY );
 720                         $out['class'] = implode( ' ', array_unique( $classes ) );
 721                 }
 722                 return $out;
 723         }
 724
 725         /**
 726          * Pick apart some CSS and check it for forbidden or unsafe structures.
 727          * Returns a sanitized string. This sanitized string will have
 728          * character references and escape sequences decoded, and comments
 729          * stripped. If the input is just too evil, only a comment complaining
 730          * about evilness will be returned.
 731          *
 732          * Currently URL references, 'expression', 'tps' are forbidden.
 733          *
 734          * NOTE: Despite the fact that character references are decoded, the
 735          * returned string may contain character references given certain
 736          * clever input strings. These character references must
 737          * be escaped before the return value is embedded in HTML.
 738          *
 739          * @param $value String
 740          * @return String
 741          */
 742         static function checkCss( $value ) {
 743                 // Decode character references like &#123;
 744                 $value = Sanitizer::decodeCharReferences( $value );
 745
 746                 // Decode escape sequences and line continuation
 747                 // See the grammar in the CSS 2 spec, appendix D.
 748                 // This has to be done AFTER decoding character references.
 749                 // This means it isn't possible for this function to return
 750                 // unsanitized escape sequences. It is possible to manufacture
 751                 // input that contains character references that decode to
 752                 // escape sequences that decode to character references, but
 753                 // it's OK for the return value to contain character references
 754                 // because the caller is supposed to escape those anyway.
 755                 static $decodeRegex;
 756                 if ( !$decodeRegex ) {
 757                         $space = '[\\x20\\t\\r\\n\\f]';
 758                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 759                         $backslash = '\\\\';
 760                         $decodeRegex = "/ $backslash
 761                                 (?:
 762                                         ($nl) |  # 1. Line continuation
 763                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 764                                         (.) | # 3. backslash cancelling special meaning
 765                                         () | # 4. backslash at end of string
 766                                 )/xu";
 767                 }
 768                 $value = preg_replace_callback( $decodeRegex,
 769                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 770
 771                 // Remove any comments; IE gets token splitting wrong
 772                 // This must be done AFTER decoding character references and
 773                 // escape sequences, because those steps can introduce comments
 774                 // This step cannot introduce character references or escape
 775                 // sequences, because it replaces comments with spaces rather
 776                 // than removing them completely.
 777                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 778
 779                 // Remove anything after a comment-start token, to guard against
 780                 // incorrect client implementations.
 781                 $commentPos = strpos( $value, '/*' );
 782                 if ( $commentPos !== false ) {
 783                         $value = substr( $value, 0, $commentPos );
 784                 }
 785
 786                 // Reject problematic keywords and control characters
 787                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 788                         return '/* invalid control char */';
 789                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
 790                         return '/* insecure input */';
 791                 }
 792                 return $value;
 793         }
 794
 795         static function cssDecodeCallback( $matches ) {
 796                 if ( $matches[1] !== '' ) {
 797                         // Line continuation
 798                         return '';
 799                 } elseif ( $matches[2] !== '' ) {
 800                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 801                 } elseif ( $matches[3] !== '' ) {
 802                         $char = $matches[3];
 803                 } else {
 804                         $char = '\\';
 805                 }
 806                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 807                         // These characters need to be escaped in strings
 808                         // Clean up the escape sequence to avoid parsing errors by clients
 809                         return '\\' . dechex( ord( $char ) ) . ' ';
 810                 } else {
 811                         // Decode unnecessary escape
 812                         return $char;
 813                 }
 814         }
 815
 816         /**
 817          * Take a tag soup fragment listing an HTML element's attributes
 818          * and normalize it to well-formed XML, discarding unwanted attributes.
 819          * Output is safe for further wikitext processing, with escaping of
 820          * values that could trigger problems.
 821          *
 822          * - Normalizes attribute names to lowercase
 823          * - Discards attributes not on a whitelist for the given element
 824          * - Turns broken or invalid entities into plaintext
 825          * - Double-quotes all attribute values
 826          * - Attributes without values are given the name as attribute
 827          * - Double attributes are discarded
 828          * - Unsafe style attributes are discarded
 829          * - Prepends space if there are attributes.
 830          *
 831          * @param $text String
 832          * @param $element String
 833          * @return String
 834          */
 835         static function fixTagAttributes( $text, $element ) {
 836                 if( trim( $text ) == '' ) {
 837                         return '';
 838                 }
 839
 840                 $stripped = Sanitizer::validateTagAttributes(
 841                         Sanitizer::decodeTagAttributes( $text ), $element );
 842
 843                 $attribs = array();
 844                 foreach( $stripped as $attribute => $value ) {
 845                         $encAttribute = htmlspecialchars( $attribute );
 846                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 847
 848                         $attribs[] = "$encAttribute=\"$encValue\"";
 849                 }
 850                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 851         }
 852
 853         /**
 854          * Encode an attribute value for HTML output.
 855          * @param $text String
 856          * @return HTML-encoded text fragment
 857          */
 858         static function encodeAttribute( $text ) {
 859                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 860
 861                 // Whitespace is normalized during attribute decoding,
 862                 // so if we've been passed non-spaces we must encode them
 863                 // ahead of time or they won't be preserved.
 864                 $encValue = strtr( $encValue, array(
 865                         "\n" => '&#10;',
 866                         "\r" => '&#13;',
 867                         "\t" => '&#9;',
 868                 ) );
 869
 870                 return $encValue;
 871         }
 872
 873         /**
 874          * Encode an attribute value for HTML tags, with extra armoring
 875          * against further wiki processing.
 876          * @param $text String
 877          * @return HTML-encoded text fragment
 878          */
 879         static function safeEncodeAttribute( $text ) {
 880                 $encValue = Sanitizer::encodeAttribute( $text );
 881
 882                 # Templates and links may be expanded in later parsing,
 883                 # creating invalid or dangerous output. Suppress this.
 884                 $encValue = strtr( $encValue, array(
 885                         '<'    => '&lt;',   // This should never happen,
 886                         '>'    => '&gt;',   // we've received invalid input
 887                         '"'    => '&quot;', // which should have been escaped.
 888                         '{'    => '&#123;',
 889                         '['    => '&#91;',
 890                         "''"   => '&#39;&#39;',
 891                         'ISBN' => '&#73;SBN',
 892                         'RFC'  => '&#82;FC',
 893                         'PMID' => '&#80;MID',
 894                         '|'    => '&#124;',
 895                         '__'   => '&#95;_',
 896                 ) );
 897
 898                 # Stupid hack
 899                 $encValue = preg_replace_callback(
 900                         '/(' . wfUrlProtocols() . ')/',
 901                         array( 'Sanitizer', 'armorLinksCallback' ),
 902                         $encValue );
 903                 return $encValue;
 904         }
 905
 906         /**
 907          * Given a value, escape it so that it can be used in an id attribute and
 908          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
 909          * true, allowing anything but ASCII whitespace.  Otherwise it will use
 910          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
 911          * escaped with lots of dots.
 912          *
 913          * To ensure we don't have to bother escaping anything, we also strip ', ",
 914          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
 915          * We also strip # because it upsets IE, and % because it could be
 916          * ambiguous if it's part of something that looks like a percent escape
 917          * (which don't work reliably in fragments cross-browser).
 918          *
 919          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
 920          *                                                          in the id and
 921          *                                                          name attributes
 922          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
 923          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
 924          *   HTML5 definition of id attribute
 925          *
 926          * @param $id String: id to escape
 927          * @param $options Mixed: string or array of strings (default is array()):
 928          *   'noninitial': This is a non-initial fragment of an id, not a full id,
 929          *       so don't pay attention if the first character isn't valid at the
 930          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
 931          *       false.
 932          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
 933          *       if $wgExperimentalHtmlIds is used, so we can generate extra
 934          *       anchors and links won't break.
 935          * @return String
 936          */
 937         static function escapeId( $id, $options = array() ) {
 938                 global $wgHtml5, $wgExperimentalHtmlIds;
 939                 $options = (array)$options;
 940
 941                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
 942                         $id = Sanitizer::decodeCharReferences( $id );
 943                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
 944                         $id = trim( $id, '_' );
 945                         if ( $id === '' ) {
 946                                 # Must have been all whitespace to start with.
 947                                 return '_';
 948                         } else {
 949                                 return $id;
 950                         }
 951                 }
 952
 953                 # HTML4-style escaping
 954                 static $replace = array(
 955                         '%3A' => ':',
 956                         '%' => '.'
 957                 );
 958
 959                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
 960                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
 961
 962                 if ( !preg_match( '/^[a-zA-Z]/', $id )
 963                 && !in_array( 'noninitial', $options ) )  {
 964                         // Initial character must be a letter!
 965                         $id = "x$id";
 966                 }
 967                 return $id;
 968         }
 969
 970         /**
 971          * Given a value, escape it so that it can be used as a CSS class and
 972          * return it.
 973          *
 974          * @todo For extra validity, input should be validated UTF-8.
 975          *
 976          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
 977          *
 978          * @param $class String
 979          * @return String
 980          */
 981         static function escapeClass( $class ) {
 982                 // Convert ugly stuff to underscores and kill underscores in ugly places
 983                 return rtrim(preg_replace(
 984                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
 985                         '_',
 986                         $class ), '_');
 987         }
 988
 989         /**
 990          * Given HTML input, escape with htmlspecialchars but un-escape entites.
 991          * This allows (generally harmless) entities like &#160; to survive.
 992          *
 993          * @param $html String to escape
 994          * @return String: escaped input
 995          */
 996         static function escapeHtmlAllowEntities( $html ) {
 997                 $html = Sanitizer::decodeCharReferences( $html );
 998                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
 999                 # hurt.
1000                 $html = htmlspecialchars( $html, ENT_QUOTES );
1001                 return $html;
1002         }
1003
1004         /**
1005          * Regex replace callback for armoring links against further processing.
1006          * @param $matches Array
1007          * @return string
1008          */
1009         private static function armorLinksCallback( $matches ) {
1010                 return str_replace( ':', '&#58;', $matches[1] );
1011         }
1012
1013         /**
1014          * Return an associative array of attribute names and values from
1015          * a partial tag string. Attribute names are forces to lowercase,
1016          * character references are decoded to UTF-8 text.
1017          *
1018          * @param $text String
1019          * @return Array
1020          */
1021         public static function decodeTagAttributes( $text ) {
1022                 if( trim( $text ) == '' ) {
1023                         return array();
1024                 }
1025
1026                 $attribs = array();
1027                 $pairs = array();
1028                 if( !preg_match_all(
1029                         self::getAttribsRegex(),
1030                         $text,
1031                         $pairs,
1032                         PREG_SET_ORDER ) ) {
1033                         return $attribs;
1034                 }
1035
1036                 foreach( $pairs as $set ) {
1037                         $attribute = strtolower( $set[1] );
1038                         $value = Sanitizer::getTagAttributeCallback( $set );
1039
1040                         // Normalize whitespace
1041                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1042                         $value = trim( $value );
1043
1044                         // Decode character references
1045                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1046                 }
1047                 return $attribs;
1048         }
1049
1050         /**
1051          * Pick the appropriate attribute value from a match set from the
1052          * attribs regex matches.
1053          *
1054          * @param $set Array
1055          * @return String
1056          */
1057         private static function getTagAttributeCallback( $set ) {
1058                 if( isset( $set[6] ) ) {
1059                         # Illegal #XXXXXX color with no quotes.
1060                         return $set[6];
1061                 } elseif( isset( $set[5] ) ) {
1062                         # No quotes.
1063                         return $set[5];
1064                 } elseif( isset( $set[4] ) ) {
1065                         # Single-quoted
1066                         return $set[4];
1067                 } elseif( isset( $set[3] ) ) {
1068                         # Double-quoted
1069                         return $set[3];
1070                 } elseif( !isset( $set[2] ) ) {
1071                         # In XHTML, attributes must have a value.
1072                         # For 'reduced' form, return explicitly the attribute name here.
1073                         return $set[1];
1074                 } else {
1075                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1076                 }
1077         }
1078
1079         /**
1080          * Normalize whitespace and character references in an XML source-
1081          * encoded text for an attribute value.
1082          *
1083          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1084          * but note that we're not returning the value, but are returning
1085          * XML source fragments that will be slapped into output.
1086          *
1087          * @param $text String
1088          * @return String
1089          */
1090         private static function normalizeAttributeValue( $text ) {
1091                 return str_replace( '"', '&quot;',
1092                         self::normalizeWhitespace(
1093                                 Sanitizer::normalizeCharReferences( $text ) ) );
1094         }
1095
1096         private static function normalizeWhitespace( $text ) {
1097                 return preg_replace(
1098                         '/\r\n|[\x20\x0d\x0a\x09]/',
1099                         ' ',
1100                         $text );
1101         }
1102
1103         /**
1104          * Normalizes whitespace in a section name, such as might be returned
1105          * by Parser::stripSectionName(), for use in the id's that are used for
1106          * section links.
1107          *
1108          * @param $section String
1109          * @return String
1110          */
1111         static function normalizeSectionNameWhitespace( $section ) {
1112                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1113         }
1114
1115         /**
1116          * Ensure that any entities and character references are legal
1117          * for XML and XHTML specifically. Any stray bits will be
1118          * &amp;-escaped to result in a valid text fragment.
1119          *
1120          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1121          *   numericized (this way we're well-formed even without a DTD)
1122          * b. any numeric char refs must be legal chars, not invalid or forbidden
1123          * c. use &#x, not &#X
1124          * d. fix or reject non-valid attributes
1125          *
1126          * @param $text String
1127          * @return String
1128          * @private
1129          */
1130         static function normalizeCharReferences( $text ) {
1131                 return preg_replace_callback(
1132                         self::CHAR_REFS_REGEX,
1133                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1134                         $text );
1135         }
1136         /**
1137          * @param $matches String
1138          * @return String
1139          */
1140         static function normalizeCharReferencesCallback( $matches ) {
1141                 $ret = null;
1142                 if( $matches[1] != '' ) {
1143                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1144                 } elseif( $matches[2] != '' ) {
1145                         $ret = Sanitizer::decCharReference( $matches[2] );
1146                 } elseif( $matches[3] != ''  ) {
1147                         $ret = Sanitizer::hexCharReference( $matches[3] );
1148                 }
1149                 if( is_null( $ret ) ) {
1150                         return htmlspecialchars( $matches[0] );
1151                 } else {
1152                         return $ret;
1153                 }
1154         }
1155
1156         /**
1157          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1158          * return the equivalent numeric entity reference (except for the core &lt;
1159          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1160          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1161          * pseudo-entity source (eg &amp;foo;)
1162          *
1163          * @param $name String
1164          * @return String
1165          */
1166         static function normalizeEntity( $name ) {
1167                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1168                         return '&' . self::$htmlEntityAliases[$name] . ';';
1169                 } elseif ( in_array( $name,
1170                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1171                         return "&$name;";
1172                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1173                         return '&#' . self::$htmlEntities[$name] . ';';
1174                 } else {
1175                         return "&amp;$name;";
1176                 }
1177         }
1178
1179         static function decCharReference( $codepoint ) {
1180                 $point = intval( $codepoint );
1181                 if( Sanitizer::validateCodepoint( $point ) ) {
1182                         return sprintf( '&#%d;', $point );
1183                 } else {
1184                         return null;
1185                 }
1186         }
1187
1188         static function hexCharReference( $codepoint ) {
1189                 $point = hexdec( $codepoint );
1190                 if( Sanitizer::validateCodepoint( $point ) ) {
1191                         return sprintf( '&#x%x;', $point );
1192                 } else {
1193                         return null;
1194                 }
1195         }
1196
1197         /**
1198          * Returns true if a given Unicode codepoint is a valid character in XML.
1199          * @param $codepoint Integer
1200          * @return Boolean
1201          */
1202         private static function validateCodepoint( $codepoint ) {
1203                 return ($codepoint ==    0x09)
1204                         || ($codepoint ==    0x0a)
1205                         || ($codepoint ==    0x0d)
1206                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1207                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1208                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1209         }
1210
1211         /**
1212          * Decode any character references, numeric or named entities,
1213          * in the text and return a UTF-8 string.
1214          *
1215          * @param $text String
1216          * @return String
1217          */
1218         public static function decodeCharReferences( $text ) {
1219                 return preg_replace_callback(
1220                         self::CHAR_REFS_REGEX,
1221                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1222                         $text );
1223         }
1224
1225         /**
1226          * Decode any character references, numeric or named entities,
1227          * in the next and normalize the resulting string. (bug 14952)
1228          *
1229          * This is useful for page titles, not for text to be displayed,
1230          * MediaWiki allows HTML entities to escape normalization as a feature.
1231          *
1232          * @param $text String (already normalized, containing entities)
1233          * @return String (still normalized, without entities)
1234          */
1235         public static function decodeCharReferencesAndNormalize( $text ) {
1236                 global $wgContLang;
1237                 $text = preg_replace_callback(
1238                         self::CHAR_REFS_REGEX,
1239                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1240                         $text, /* limit */ -1, $count );
1241
1242                 if ( $count ) {
1243                         return $wgContLang->normalize( $text );
1244                 } else {
1245                         return $text;
1246                 }
1247         }
1248
1249         /**
1250          * @param $matches String
1251          * @return String
1252          */
1253         static function decodeCharReferencesCallback( $matches ) {
1254                 if( $matches[1] != '' ) {
1255                         return Sanitizer::decodeEntity( $matches[1] );
1256                 } elseif( $matches[2] != '' ) {
1257                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1258                 } elseif( $matches[3] != ''  ) {
1259                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1260                 }
1261                 # Last case should be an ampersand by itself
1262                 return $matches[0];
1263         }
1264
1265         /**
1266          * Return UTF-8 string for a codepoint if that is a valid
1267          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1268          * @param $codepoint Integer
1269          * @return String
1270          * @private
1271          */
1272         static function decodeChar( $codepoint ) {
1273                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1274                         return codepointToUtf8( $codepoint );
1275                 } else {
1276                         return UTF8_REPLACEMENT;
1277                 }
1278         }
1279
1280         /**
1281          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1282          * return the UTF-8 encoding of that character. Otherwise, returns
1283          * pseudo-entity source (eg &foo;)
1284          *
1285          * @param $name Strings
1286          * @return String
1287          */
1288         static function decodeEntity( $name ) {
1289                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1290                         $name = self::$htmlEntityAliases[$name];
1291                 }
1292                 if( isset( self::$htmlEntities[$name] ) ) {
1293                         return codepointToUtf8( self::$htmlEntities[$name] );
1294                 } else {
1295                         return "&$name;";
1296                 }
1297         }
1298
1299         /**
1300          * Fetch the whitelist of acceptable attributes for a given element name.
1301          *
1302          * @param $element String
1303          * @return Array
1304          */
1305         static function attributeWhitelist( $element ) {
1306                 static $list;
1307                 if( !isset( $list ) ) {
1308                         $list = Sanitizer::setupAttributeWhitelist();
1309                 }
1310                 return isset( $list[$element] )
1311                         ? $list[$element]
1312                         : array();
1313         }
1314
1315         /**
1316          * Foreach array key (an allowed HTML element), return an array
1317          * of allowed attributes
1318          * @return Array
1319          */
1320         static function setupAttributeWhitelist() {
1321                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1322
1323                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1324
1325                 if ( $wgAllowRdfaAttributes ) {
1326                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1327                         $common = array_merge( $common, array(
1328                             'about', 'property', 'resource', 'datatype', 'typeof',
1329                         ) );
1330                 }
1331
1332                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1333                         # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1334                         $common = array_merge( $common, array(
1335                             'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1336                         ) );
1337                 }
1338
1339                 $block = array_merge( $common, array( 'align' ) );
1340                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1341                 $tablecell = array( 'abbr',
1342                                     'axis',
1343                                     'headers',
1344                                     'scope',
1345                                     'rowspan',
1346                                     'colspan',
1347                                     'nowrap', # deprecated
1348                                     'width',  # deprecated
1349                                     'height', # deprecated
1350                                     'bgcolor' # deprecated
1351                                     );
1352
1353                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1354                 # See: http://www.w3.org/TR/html4/
1355                 $whitelist = array (
1356                         # 7.5.4
1357                         'div'        => $block,
1358                         'center'     => $common, # deprecated
1359                         'span'       => $block, # ??
1360
1361                         # 7.5.5
1362                         'h1'         => $block,
1363                         'h2'         => $block,
1364                         'h3'         => $block,
1365                         'h4'         => $block,
1366                         'h5'         => $block,
1367                         'h6'         => $block,
1368
1369                         # 7.5.6
1370                         # address
1371
1372                         # 8.2.4
1373                         # bdo
1374
1375                         # 9.2.1
1376                         'em'         => $common,
1377                         'strong'     => $common,
1378                         'cite'       => $common,
1379                         'dfn'        => $common,
1380                         'code'       => $common,
1381                         'samp'       => $common,
1382                         'kbd'        => $common,
1383                         'var'        => $common,
1384                         'abbr'       => $common,
1385                         # acronym
1386
1387                         # 9.2.2
1388                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1389                         # q
1390
1391                         # 9.2.3
1392                         'sub'        => $common,
1393                         'sup'        => $common,
1394
1395                         # 9.3.1
1396                         'p'          => $block,
1397
1398                         # 9.3.2
1399                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1400
1401                         # 9.3.4
1402                         'pre'        => array_merge( $common, array( 'width' ) ),
1403
1404                         # 9.4
1405                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1406                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1407
1408                         # 10.2
1409                         'ul'         => array_merge( $common, array( 'type' ) ),
1410                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1411                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1412
1413                         # 10.3
1414                         'dl'         => $common,
1415                         'dd'         => $common,
1416                         'dt'         => $common,
1417
1418                         # 11.2.1
1419                         'table'      => array_merge( $common,
1420                                                                 array( 'summary', 'width', 'border', 'frame',
1421                                                                                 'rules', 'cellspacing', 'cellpadding',
1422                                                                                 'align', 'bgcolor',
1423                                                                 ) ),
1424
1425                         # 11.2.2
1426                         'caption'    => array_merge( $common, array( 'align' ) ),
1427
1428                         # 11.2.3
1429                         'thead'      => array_merge( $common, $tablealign ),
1430                         'tfoot'      => array_merge( $common, $tablealign ),
1431                         'tbody'      => array_merge( $common, $tablealign ),
1432
1433                         # 11.2.4
1434                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1435                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1436
1437                         # 11.2.5
1438                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1439
1440                         # 11.2.6
1441                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1442                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1443
1444                         # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1445                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1446
1447                         # 13.2
1448                         # Not usually allowed, but may be used for extension-style hooks
1449                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1450                         # true
1451                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1452
1453                         # 15.2.1
1454                         'tt'         => $common,
1455                         'b'          => $common,
1456                         'i'          => $common,
1457                         'big'        => $common,
1458                         'small'      => $common,
1459                         'strike'     => $common,
1460                         's'          => $common,
1461                         'u'          => $common,
1462
1463                         # 15.2.2
1464                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1465                         # basefont
1466
1467                         # 15.3
1468                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1469
1470                         # XHTML Ruby annotation text module, simple ruby only.
1471                         # http://www.w3c.org/TR/ruby/
1472                         'ruby'       => $common,
1473                         # rbc
1474                         # rtc
1475                         'rb'         => $common,
1476                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1477                         'rp'         => $common,
1478
1479                         # MathML root element, where used for extensions
1480                         # 'title' may not be 100% valid here; it's XHTML
1481                         # http://www.w3.org/TR/REC-MathML/
1482                         'math'       => array( 'class', 'style', 'id', 'title' ),
1483                         );
1484                 return $whitelist;
1485         }
1486
1487         /**
1488          * Take a fragment of (potentially invalid) HTML and return
1489          * a version with any tags removed, encoded as plain text.
1490          *
1491          * Warning: this return value must be further escaped for literal
1492          * inclusion in HTML output as of 1.10!
1493          *
1494          * @param $text String: HTML fragment
1495          * @return String
1496          */
1497         static function stripAllTags( $text ) {
1498                 # Actual <tags>
1499                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1500
1501                 # Normalize &entities and whitespace
1502                 $text = self::decodeCharReferences( $text );
1503                 $text = self::normalizeWhitespace( $text );
1504
1505                 return $text;
1506         }
1507
1508         /**
1509          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1510          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1511          * PHP 5.1 doesn't.
1512          *
1513          * Use for passing XHTML fragments to PHP's XML parsing functions
1514          *
1515          * @return String
1516          */
1517         static function hackDocType() {
1518                 $out = "<!DOCTYPE html [\n";
1519                 foreach( self::$htmlEntities as $entity => $codepoint ) {
1520                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1521                 }
1522                 $out .= "]>\n";
1523                 return $out;
1524         }
1525
1526         static function cleanUrl( $url ) {
1527                 # Normalize any HTML entities in input. They will be
1528                 # re-escaped by makeExternalLink().
1529                 $url = Sanitizer::decodeCharReferences( $url );
1530
1531                 # Escape any control characters introduced by the above step
1532                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1533                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1534
1535                 # Validate hostname portion
1536                 $matches = array();
1537                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1538                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1539
1540                         // Characters that will be ignored in IDNs.
1541                         // http://tools.ietf.org/html/3454#section-3.1
1542                         // Strip them before further processing so blacklists and such work.
1543                         $strip = "/
1544                                 \\s|          # general whitespace
1545                                 \xc2\xad|     # 00ad SOFT HYPHEN
1546                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1547                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1548                                 \xe2\x81\xa0| # 2060 WORD JOINER
1549                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1550                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1551                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1552                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1553                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1554                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1555                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1556                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe00f VARIATION SELECTOR-1-16
1557                                 /xuD";
1558
1559                         $host = preg_replace( $strip, '', $host );
1560
1561                         // @todo Fixme: validate hostnames here
1562
1563                         return $protocol . $host . $rest;
1564                 } else {
1565                         return $url;
1566                 }
1567         }
1568
1569         static function cleanUrlCallback( $matches ) {
1570                 return urlencode( $matches[0] );
1571         }
1572
1573         /**
1574          * Does a string look like an e-mail address?
1575          *
1576          * This validates an email address using an HTML5 specification found at:
1577          * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1578          * Which as of 2011-01-24 says:
1579          *
1580          *   A valid e-mail address is a string that matches the ABNF production
1581          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1582          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1583          *   3.5.
1584          *
1585          * This function is an implementation of the specification as requested in
1586          * bug 22449.
1587          *
1588          * Client-side forms will use the same standard validation rules via JS or
1589          * HTML 5 validation; additional restrictions can be enforced server-side
1590          * by extensions via the 'isValidEmailAddr' hook.
1591          *
1592          * Note that this validation doesn't 100% match RFC 2822, but is believed
1593          * to be liberal enough for wide use. Some invalid addresses will still
1594          * pass validation here.
1595          *
1596          * @param $addr String E-mail address
1597          * @return Bool
1598          */
1599         public static function validateEmail( $addr ) {
1600                 $result = null;
1601                 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1602                         return $result;
1603                 }
1604
1605                 // Please note strings below are enclosed in brackets [], this make the
1606                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1607                 // See bug 26948
1608                 $rfc5322_atext   = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1609                 $rfc1034_ldh_str = "a-z0-9\\-" ;
1610
1611                 $HTML5_email_regexp = "/
1612                 ^                      # start of string
1613                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1614                 @                      # 'apostrophe'
1615                 [$rfc1034_ldh_str]+       # First domain part
1616                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1617                 $                      # End of string
1618                 /ix" ; // case Insensitive, eXtended
1619
1620                 return (bool) preg_match( $HTML5_email_regexp, $addr );
1621         }
1622 }