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         /**
  43          * Blacklist for evil uris like javascript:
  44          * WARNING: DO NOT use this in any place that actually requires blacklisting
  45          * for security reasons. There are NUMEROUS[1] ways to bypass blacklisting, the
  46          * only way to be secure from javascript: uri based xss vectors is to whitelist
  47          * things that you know are safe and deny everything else.
  48          * [1]: http://ha.ckers.org/xss.html
  49          */
  50         const EVIL_URI_PATTERN = '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
  51         const XMLNS_ATTRIBUTE_PATTERN = "/^xmlns:[:A-Z_a-z-.0-9]+$/";
  52
  53         /**
  54          * List of all named character entities defined in HTML 4.01
  55          * http://www.w3.org/TR/html4/sgml/entities.html
  56          * As well as &apos; which is only defined starting in XHTML1.
  57          * @private
  58          */
  59         static $htmlEntities = array(
  60                 'Aacute'   => 193,
  61                 'aacute'   => 225,
  62                 'Acirc'    => 194,
  63                 'acirc'    => 226,
  64                 'acute'    => 180,
  65                 'AElig'    => 198,
  66                 'aelig'    => 230,
  67                 'Agrave'   => 192,
  68                 'agrave'   => 224,
  69                 'alefsym'  => 8501,
  70                 'Alpha'    => 913,
  71                 'alpha'    => 945,
  72                 'amp'      => 38,
  73                 'and'      => 8743,
  74                 'ang'      => 8736,
  75                 'apos'     => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
  76                 'Aring'    => 197,
  77                 'aring'    => 229,
  78                 'asymp'    => 8776,
  79                 'Atilde'   => 195,
  80                 'atilde'   => 227,
  81                 'Auml'     => 196,
  82                 'auml'     => 228,
  83                 'bdquo'    => 8222,
  84                 'Beta'     => 914,
  85                 'beta'     => 946,
  86                 'brvbar'   => 166,
  87                 'bull'     => 8226,
  88                 'cap'      => 8745,
  89                 'Ccedil'   => 199,
  90                 'ccedil'   => 231,
  91                 'cedil'    => 184,
  92                 'cent'     => 162,
  93                 'Chi'      => 935,
  94                 'chi'      => 967,
  95                 'circ'     => 710,
  96                 'clubs'    => 9827,
  97                 'cong'     => 8773,
  98                 'copy'     => 169,
  99                 'crarr'    => 8629,
 100                 'cup'      => 8746,
 101                 'curren'   => 164,
 102                 'dagger'   => 8224,
 103                 'Dagger'   => 8225,
 104                 'darr'     => 8595,
 105                 'dArr'     => 8659,
 106                 'deg'      => 176,
 107                 'Delta'    => 916,
 108                 'delta'    => 948,
 109                 'diams'    => 9830,
 110                 'divide'   => 247,
 111                 'Eacute'   => 201,
 112                 'eacute'   => 233,
 113                 'Ecirc'    => 202,
 114                 'ecirc'    => 234,
 115                 'Egrave'   => 200,
 116                 'egrave'   => 232,
 117                 'empty'    => 8709,
 118                 'emsp'     => 8195,
 119                 'ensp'     => 8194,
 120                 'Epsilon'  => 917,
 121                 'epsilon'  => 949,
 122                 'equiv'    => 8801,
 123                 'Eta'      => 919,
 124                 'eta'      => 951,
 125                 'ETH'      => 208,
 126                 'eth'      => 240,
 127                 'Euml'     => 203,
 128                 'euml'     => 235,
 129                 'euro'     => 8364,
 130                 'exist'    => 8707,
 131                 'fnof'     => 402,
 132                 'forall'   => 8704,
 133                 'frac12'   => 189,
 134                 'frac14'   => 188,
 135                 'frac34'   => 190,
 136                 'frasl'    => 8260,
 137                 'Gamma'    => 915,
 138                 'gamma'    => 947,
 139                 'ge'       => 8805,
 140                 'gt'       => 62,
 141                 'harr'     => 8596,
 142                 'hArr'     => 8660,
 143                 'hearts'   => 9829,
 144                 'hellip'   => 8230,
 145                 'Iacute'   => 205,
 146                 'iacute'   => 237,
 147                 'Icirc'    => 206,
 148                 'icirc'    => 238,
 149                 'iexcl'    => 161,
 150                 'Igrave'   => 204,
 151                 'igrave'   => 236,
 152                 'image'    => 8465,
 153                 'infin'    => 8734,
 154                 'int'      => 8747,
 155                 'Iota'     => 921,
 156                 'iota'     => 953,
 157                 'iquest'   => 191,
 158                 'isin'     => 8712,
 159                 'Iuml'     => 207,
 160                 'iuml'     => 239,
 161                 'Kappa'    => 922,
 162                 'kappa'    => 954,
 163                 'Lambda'   => 923,
 164                 'lambda'   => 955,
 165                 'lang'     => 9001,
 166                 'laquo'    => 171,
 167                 'larr'     => 8592,
 168                 'lArr'     => 8656,
 169                 'lceil'    => 8968,
 170                 'ldquo'    => 8220,
 171                 'le'       => 8804,
 172                 'lfloor'   => 8970,
 173                 'lowast'   => 8727,
 174                 'loz'      => 9674,
 175                 'lrm'      => 8206,
 176                 'lsaquo'   => 8249,
 177                 'lsquo'    => 8216,
 178                 'lt'       => 60,
 179                 'macr'     => 175,
 180                 'mdash'    => 8212,
 181                 'micro'    => 181,
 182                 'middot'   => 183,
 183                 'minus'    => 8722,
 184                 'Mu'       => 924,
 185                 'mu'       => 956,
 186                 'nabla'    => 8711,
 187                 'nbsp'     => 160,
 188                 'ndash'    => 8211,
 189                 'ne'       => 8800,
 190                 'ni'       => 8715,
 191                 'not'      => 172,
 192                 'notin'    => 8713,
 193                 'nsub'     => 8836,
 194                 'Ntilde'   => 209,
 195                 'ntilde'   => 241,
 196                 'Nu'       => 925,
 197                 'nu'       => 957,
 198                 'Oacute'   => 211,
 199                 'oacute'   => 243,
 200                 'Ocirc'    => 212,
 201                 'ocirc'    => 244,
 202                 'OElig'    => 338,
 203                 'oelig'    => 339,
 204                 'Ograve'   => 210,
 205                 'ograve'   => 242,
 206                 'oline'    => 8254,
 207                 'Omega'    => 937,
 208                 'omega'    => 969,
 209                 'Omicron'  => 927,
 210                 'omicron'  => 959,
 211                 'oplus'    => 8853,
 212                 'or'       => 8744,
 213                 'ordf'     => 170,
 214                 'ordm'     => 186,
 215                 'Oslash'   => 216,
 216                 'oslash'   => 248,
 217                 'Otilde'   => 213,
 218                 'otilde'   => 245,
 219                 'otimes'   => 8855,
 220                 'Ouml'     => 214,
 221                 'ouml'     => 246,
 222                 'para'     => 182,
 223                 'part'     => 8706,
 224                 'permil'   => 8240,
 225                 'perp'     => 8869,
 226                 'Phi'      => 934,
 227                 'phi'      => 966,
 228                 'Pi'       => 928,
 229                 'pi'       => 960,
 230                 'piv'      => 982,
 231                 'plusmn'   => 177,
 232                 'pound'    => 163,
 233                 'prime'    => 8242,
 234                 'Prime'    => 8243,
 235                 'prod'     => 8719,
 236                 'prop'     => 8733,
 237                 'Psi'      => 936,
 238                 'psi'      => 968,
 239                 'quot'     => 34,
 240                 'radic'    => 8730,
 241                 'rang'     => 9002,
 242                 'raquo'    => 187,
 243                 'rarr'     => 8594,
 244                 'rArr'     => 8658,
 245                 'rceil'    => 8969,
 246                 'rdquo'    => 8221,
 247                 'real'     => 8476,
 248                 'reg'      => 174,
 249                 'rfloor'   => 8971,
 250                 'Rho'      => 929,
 251                 'rho'      => 961,
 252                 'rlm'      => 8207,
 253                 'rsaquo'   => 8250,
 254                 'rsquo'    => 8217,
 255                 'sbquo'    => 8218,
 256                 'Scaron'   => 352,
 257                 'scaron'   => 353,
 258                 'sdot'     => 8901,
 259                 'sect'     => 167,
 260                 'shy'      => 173,
 261                 'Sigma'    => 931,
 262                 'sigma'    => 963,
 263                 'sigmaf'   => 962,
 264                 'sim'      => 8764,
 265                 'spades'   => 9824,
 266                 'sub'      => 8834,
 267                 'sube'     => 8838,
 268                 'sum'      => 8721,
 269                 'sup'      => 8835,
 270                 'sup1'     => 185,
 271                 'sup2'     => 178,
 272                 'sup3'     => 179,
 273                 'supe'     => 8839,
 274                 'szlig'    => 223,
 275                 'Tau'      => 932,
 276                 'tau'      => 964,
 277                 'there4'   => 8756,
 278                 'Theta'    => 920,
 279                 'theta'    => 952,
 280                 'thetasym' => 977,
 281                 'thinsp'   => 8201,
 282                 'THORN'    => 222,
 283                 'thorn'    => 254,
 284                 'tilde'    => 732,
 285                 'times'    => 215,
 286                 'trade'    => 8482,
 287                 'Uacute'   => 218,
 288                 'uacute'   => 250,
 289                 'uarr'     => 8593,
 290                 'uArr'     => 8657,
 291                 'Ucirc'    => 219,
 292                 'ucirc'    => 251,
 293                 'Ugrave'   => 217,
 294                 'ugrave'   => 249,
 295                 'uml'      => 168,
 296                 'upsih'    => 978,
 297                 'Upsilon'  => 933,
 298                 'upsilon'  => 965,
 299                 'Uuml'     => 220,
 300                 'uuml'     => 252,
 301                 'weierp'   => 8472,
 302                 'Xi'       => 926,
 303                 'xi'       => 958,
 304                 'Yacute'   => 221,
 305                 'yacute'   => 253,
 306                 'yen'      => 165,
 307                 'Yuml'     => 376,
 308                 'yuml'     => 255,
 309                 'Zeta'     => 918,
 310                 'zeta'     => 950,
 311                 'zwj'      => 8205,
 312                 'zwnj'     => 8204
 313         );
 314
 315         /**
 316          * Character entity aliases accepted by MediaWiki
 317          */
 318         static $htmlEntityAliases = array(
 319                 'רלמ' => 'rlm',
 320                 'رلم' => 'rlm',
 321         );
 322
 323         /**
 324          * Lazy-initialised attributes regex, see getAttribsRegex()
 325          */
 326         static $attribsRegex;
 327
 328         /**
 329          * Regular expression to match HTML/XML attribute pairs within a tag.
 330          * Allows some... latitude.
 331          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 332          */
 333         static function getAttribsRegex() {
 334                 if ( self::$attribsRegex === null ) {
 335                         $attribFirst = '[:A-Z_a-z0-9]';
 336                         $attrib = '[:A-Z_a-z-.0-9]';
 337                         $space = '[\x09\x0a\x0d\x20]';
 338                         self::$attribsRegex =
 339                                 "/(?:^|$space)({$attribFirst}{$attrib}*)
 340                                   ($space*=$space*
 341                                         (?:
 342                                          # The attribute value: quoted or alone
 343                                           \"([^<\"]*)\"
 344                                          | '([^<']*)'
 345                                          |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
 346                                          |  (\#[0-9a-fA-F]+) # Technically wrong, but lots of
 347                                                                                  # colors are specified like this.
 348                                                                                  # We'll be normalizing it.
 349                                         )
 350                                 )?(?=$space|\$)/sx";
 351                 }
 352                 return self::$attribsRegex;
 353         }
 354
 355         /**
 356          * Cleans up HTML, removes dangerous tags and attributes, and
 357          * removes HTML comments
 358          * @private
 359          * @param $text String
 360          * @param $processCallback Callback to do any variable or parameter replacements in HTML attribute values
 361          * @param $args Array for the processing callback
 362          * @param $extratags Array for any extra tags to include
 363          * @param $removetags Array for any tags (default or extra) to exclude
 364          * @return string
 365          */
 366         static function removeHTMLtags( $text, $processCallback = null, $args = array(), $extratags = array(), $removetags = array() ) {
 367                 global $wgUseTidy;
 368
 369                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 370                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 371
 372                 wfProfileIn( __METHOD__ );
 373
 374                 if ( !$staticInitialised ) {
 375
 376                         $htmlpairsStatic = array( # Tags that must be closed
 377                                 'b', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 378                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 379                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 380                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 381                                 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'abbr', 'dfn',
 382                                 'kbd', 'samp'
 383                         );
 384                         $htmlsingle = array(
 385                                 'br', 'hr', 'li', 'dt', 'dd'
 386                         );
 387                         $htmlsingleonly = array( # Elements that cannot have close tags
 388                                 'br', 'hr'
 389                         );
 390                         $htmlnest = array( # Tags that can be nested--??
 391                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 392                                 'dl', 'font', 'big', 'small', 'sub', 'sup', 'span'
 393                         );
 394                         $tabletags = array( # Can only appear inside table, we will close them
 395                                 'td', 'th', 'tr',
 396                         );
 397                         $htmllist = array( # Tags used by list
 398                                 'ul','ol',
 399                         );
 400                         $listtags = array( # Tags that can appear in a list
 401                                 'li',
 402                         );
 403
 404                         global $wgAllowImageTag;
 405                         if ( $wgAllowImageTag ) {
 406                                 $htmlsingle[] = 'img';
 407                                 $htmlsingleonly[] = 'img';
 408                         }
 409
 410                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 411                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 412
 413                         # Convert them all to hashtables for faster lookup
 414                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 415                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 416                         foreach ( $vars as $var ) {
 417                                 $$var = array_flip( $$var );
 418                         }
 419                         $staticInitialised = true;
 420                 }
 421                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 422                 $extratags = array_flip( $extratags );
 423                 $removetags = array_flip( $removetags );
 424                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 425                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ) , $removetags );
 426
 427                 # Remove HTML comments
 428                 $text = Sanitizer::removeHTMLcomments( $text );
 429                 $bits = explode( '<', $text );
 430                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 431                 if ( !$wgUseTidy ) {
 432                         $tagstack = $tablestack = array();
 433                         foreach ( $bits as $x ) {
 434                                 $regs = array();
 435                                 # $slash: Does the current element start with a '/'?
 436                                 # $t: Current element name
 437                                 # $params: String between element name and >
 438                                 # $brace: Ending '>' or '/>'
 439                                 # $rest: Everything until the next element of $bits
 440                                 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 441                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 442                                 } else {
 443                                         $slash = $t = $params = $brace = $rest = null;
 444                                 }
 445
 446                                 $badtag = false;
 447                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 448                                         # Check our stack
 449                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 450                                                 $badtag = true;
 451                                         } elseif ( $slash ) {
 452                                                 # Closing a tag... is it the one we just opened?
 453                                                 $ot = @array_pop( $tagstack );
 454                                                 if ( $ot != $t ) {
 455                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 456                                                                 # Pop all elements with an optional close tag
 457                                                                 # and see if we find a match below them
 458                                                                 $optstack = array();
 459                                                                 array_push( $optstack, $ot );
 460                                                                 $ot = @array_pop( $tagstack );
 461                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 462                                                                         array_push( $optstack, $ot );
 463                                                                         $ot = @array_pop( $tagstack );
 464                                                                 }
 465                                                                 if ( $t != $ot ) {
 466                                                                         # No match. Push the optional elements back again
 467                                                                         $badtag = true;
 468                                                                         while ( $ot = @array_pop( $optstack ) ) {
 469                                                                                 array_push( $tagstack, $ot );
 470                                                                         }
 471                                                                 }
 472                                                         } else {
 473                                                                 @array_push( $tagstack, $ot );
 474                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 475                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 476                                                                         $badtag = true;
 477                                                                 }
 478                                                         }
 479                                                 } else {
 480                                                         if ( $t == 'table' ) {
 481                                                                 $tagstack = array_pop( $tablestack );
 482                                                         }
 483                                                 }
 484                                                 $newparams = '';
 485                                         } else {
 486                                                 # Keep track for later
 487                                                 if ( isset( $tabletags[$t] ) &&
 488                                                 !in_array( 'table', $tagstack ) ) {
 489                                                         $badtag = true;
 490                                                 } elseif ( in_array( $t, $tagstack ) &&
 491                                                 !isset( $htmlnest [$t ] ) ) {
 492                                                         $badtag = true;
 493                                                 # Is it a self closed htmlpair ? (bug 5487)
 494                                                 } elseif ( $brace == '/>' &&
 495                                                 isset( $htmlpairs[$t] ) ) {
 496                                                         $badtag = true;
 497                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 498                                                         # Hack to force empty tag for uncloseable elements
 499                                                         $brace = '/>';
 500                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 501                                                         # Hack to not close $htmlsingle tags
 502                                                         $brace = null;
 503                                                 } elseif ( isset( $tabletags[$t] )
 504                                                 && in_array( $t, $tagstack ) ) {
 505                                                         // New table tag but forgot to close the previous one
 506                                                         $text .= "</$t>";
 507                                                 } else {
 508                                                         if ( $t == 'table' ) {
 509                                                                 array_push( $tablestack, $tagstack );
 510                                                                 $tagstack = array();
 511                                                         }
 512                                                         array_push( $tagstack, $t );
 513                                                 }
 514
 515                                                 # Replace any variables or template parameters with
 516                                                 # plaintext results.
 517                                                 if( is_callable( $processCallback ) ) {
 518                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 519                                                 }
 520
 521                                                 # Strip non-approved attributes from the tag
 522                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 523                                         }
 524                                         if ( !$badtag ) {
 525                                                 $rest = str_replace( '>', '&gt;', $rest );
 526                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 527                                                 $text .= "<$slash$t$newparams$close>$rest";
 528                                                 continue;
 529                                         }
 530                                 }
 531                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 532                         }
 533                         # Close off any remaining tags
 534                         while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
 535                                 $text .= "</$t>\n";
 536                                 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
 537                         }
 538                 } else {
 539                         # this might be possible using tidy itself
 540                         foreach ( $bits as $x ) {
 541                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 542                                 $x, $regs );
 543                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 544                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 545                                         if( is_callable( $processCallback ) ) {
 546                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 547                                         }
 548                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 549                                         $rest = str_replace( '>', '&gt;', $rest );
 550                                         $text .= "<$slash$t$newparams$brace$rest";
 551                                 } else {
 552                                         $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 553                                 }
 554                         }
 555                 }
 556                 wfProfileOut( __METHOD__ );
 557                 return $text;
 558         }
 559
 560         /**
 561          * Remove '<!--', '-->', and everything between.
 562          * To avoid leaving blank lines, when a comment is both preceded
 563          * and followed by a newline (ignoring spaces), trim leading and
 564          * trailing spaces and one of the newlines.
 565          *
 566          * @private
 567          * @param $text String
 568          * @return string
 569          */
 570         static function removeHTMLcomments( $text ) {
 571                 wfProfileIn( __METHOD__ );
 572                 while (($start = strpos($text, '<!--')) !== false) {
 573                         $end = strpos($text, '-->', $start + 4);
 574                         if ($end === false) {
 575                                 # Unterminated comment; bail out
 576                                 break;
 577                         }
 578
 579                         $end += 3;
 580
 581                         # Trim space and newline if the comment is both
 582                         # preceded and followed by a newline
 583                         $spaceStart = max($start - 1, 0);
 584                         $spaceLen = $end - $spaceStart;
 585                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 586                                 $spaceStart--;
 587                                 $spaceLen++;
 588                         }
 589                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 590                                 $spaceLen++;
 591                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 592                                 # Remove the comment, leading and trailing
 593                                 # spaces, and leave only one newline.
 594                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 595                         }
 596                         else {
 597                                 # Remove just the comment.
 598                                 $text = substr_replace($text, '', $start, $end - $start);
 599                         }
 600                 }
 601                 wfProfileOut( __METHOD__ );
 602                 return $text;
 603         }
 604
 605         /**
 606          * Take an array of attribute names and values and fix some deprecated values
 607          * for the given element type.
 608          * This does not validate properties, so you should ensure that you call
 609          * validateTagAttributes AFTER this to ensure that the resulting style rule
 610          * this may add is safe.
 611          *
 612          * - Converts most presentational attributes like align into inline css
 613          *
 614          * @param $attribs Array
 615          * @param $element String
 616          * @return Array
 617          */
 618         static function fixDeprecatedAttributes( $attribs, $element ) {
 619                 global $wgHtml5, $wgCleanupPresentationalAttributes;
 620
 621                 // presentational attributes were removed from html5, we can leave them
 622                 // in when html5 is turned off
 623                 if ( !$wgHtml5 || !$wgCleanupPresentationalAttributes ) {
 624                         return $attribs;
 625                 }
 626
 627                 $table = array( 'table' );
 628                 $cells = array( 'td', 'th' );
 629                 $colls = array( 'col', 'colgroup' );
 630                 $tblocks = array( 'tbody', 'tfoot', 'thead' );
 631                 $h = array( 'h1', 'h2', 'h3', 'h4', 'h5', 'h6' );
 632
 633                 $presentationalAttribs = array(
 634                         'align' => array( 'text-align', array_merge( array( 'caption', 'hr', 'div', 'p', 'tr' ), $table, $cells, $colls, $tblocks, $h ) ),
 635                         'clear' => array( 'clear', array( 'br' ) ),
 636                         'height' => array( 'height', $cells ),
 637                         'nowrap' => array( 'white-space', $cells ),
 638                         'size' => array( 'height', array( 'hr' ) ),
 639                         'type' => array( 'list-style-type', array( 'li', 'ol', 'ul' ) ),
 640                         'valign' => array( 'vertical-align', array_merge( $cells, $colls, $tblocks ) ),
 641                         'width' => array( 'width', array_merge( array( 'hr', 'pre' ), $table, $cells, $colls ) ),
 642                 );
 643
 644                 $style = "";
 645                 foreach ( $presentationalAttribs as $attribute => $info ) {
 646                         list( $property, $elements ) = $info;
 647
 648                         // Skip if this attribute is not relevant to this element
 649                         if ( !in_array( $element, $elements ) ) {
 650                                 continue;
 651                         }
 652
 653                         // Skip if the attribute is not used
 654                         if ( !array_key_exists( $attribute, $attribs ) ) {
 655                                 continue;
 656                         }
 657
 658                         $value = $attribs[$attribute];
 659
 660                         // For nowrap the value should be nowrap instead of whatever text is in the value
 661                         if ( $attribute === 'nowrap' ) {
 662                                 $value = 'nowrap';
 663                         }
 664
 665                         // Size based properties should have px applied to them if they have no unit
 666                         if ( in_array( $attribute, array( 'height', 'width', 'size' ) ) ) {
 667                                 if ( preg_match( '/^[\d.]+$/', $value ) ) {
 668                                         $value = "{$value}px";
 669                                 }
 670                         }
 671
 672                         $style .= " $property: $value;";
 673
 674                         unset( $attribs[$attribute] );
 675                 }
 676
 677                 if ( !empty($style) ) {
 678                         // Prepend our style rules so that they can be overridden by user css
 679                         if ( isset($attribs['style']) ) {
 680                                 $style .= " " . $attribs['style'];
 681                         }
 682                         $attribs['style'] = trim($style);
 683                 }
 684
 685                 return $attribs;
 686         }
 687
 688         /**
 689          * Take an array of attribute names and values and normalize or discard
 690          * illegal values for the given element type.
 691          *
 692          * - Discards attributes not on a whitelist for the given element
 693          * - Unsafe style attributes are discarded
 694          * - Invalid id attributes are reencoded
 695          *
 696          * @param $attribs Array
 697          * @param $element String
 698          * @return Array
 699          *
 700          * @todo Check for legal values where the DTD limits things.
 701          * @todo Check for unique id attribute :P
 702          */
 703         static function validateTagAttributes( $attribs, $element ) {
 704                 return Sanitizer::validateAttributes( $attribs,
 705                         Sanitizer::attributeWhitelist( $element ) );
 706         }
 707
 708         /**
 709          * Take an array of attribute names and values and normalize or discard
 710          * illegal values for the given whitelist.
 711          *
 712          * - Discards attributes not the given whitelist
 713          * - Unsafe style attributes are discarded
 714          * - Invalid id attributes are reencoded
 715          *
 716          * @param $attribs Array
 717          * @param $whitelist Array: list of allowed attribute names
 718          * @return Array
 719          *
 720          * @todo Check for legal values where the DTD limits things.
 721          * @todo Check for unique id attribute :P
 722          */
 723         static function validateAttributes( $attribs, $whitelist ) {
 724                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
 725
 726                 $whitelist = array_flip( $whitelist );
 727                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 728
 729                 $out = array();
 730                 foreach( $attribs as $attribute => $value ) {
 731                         #allow XML namespace declaration if RDFa is enabled
 732                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 733                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 734                                         $out[$attribute] = $value;
 735                                 }
 736
 737                                 continue;
 738                         }
 739
 740                         # Allow any attribute beginning with "data-", if in HTML5 mode
 741                         if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
 742                                 continue;
 743                         }
 744
 745                         # Strip javascript "expression" from stylesheets.
 746                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 747                         if( $attribute == 'style' ) {
 748                                 $value = Sanitizer::checkCss( $value );
 749                         }
 750
 751                         if ( $attribute === 'id' ) {
 752                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 753                         }
 754
 755                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 756                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 757                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 758                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 759                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 760                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 761
 762                                 //Paranoia. Allow "simple" values but suppress javascript
 763                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 764                                         continue;
 765                                 }
 766                         }
 767
 768                         # NOTE: even though elements using href/src are not allowed directly, supply
 769                         #       validation code that can be used by tag hook handlers, etc
 770                         if ( $attribute === 'href' || $attribute === 'src' ) {
 771                                 if ( !preg_match( $hrefExp, $value ) ) {
 772                                         continue; //drop any href or src attributes not using an allowed protocol.
 773                                                   //NOTE: this also drops all relative URLs
 774                                 }
 775                         }
 776
 777                         // If this attribute was previously set, override it.
 778                         // Output should only have one attribute of each name.
 779                         $out[$attribute] = $value;
 780                 }
 781
 782                 if ( $wgAllowMicrodataAttributes ) {
 783                         # itemtype, itemid, itemref don't make sense without itemscope
 784                         if ( !array_key_exists( 'itemscope', $out ) ) {
 785                                 unset( $out['itemtype'] );
 786                                 unset( $out['itemid'] );
 787                                 unset( $out['itemref'] );
 788                         }
 789                         # TODO: Strip itemprop if we aren't descendants of an itemscope.
 790                 }
 791                 return $out;
 792         }
 793
 794         /**
 795          * Merge two sets of HTML attributes.  Conflicting items in the second set
 796          * will override those in the first, except for 'class' attributes which
 797          * will be combined (if they're both strings).
 798          *
 799          * @todo implement merging for other attributes such as style
 800          * @param $a Array
 801          * @param $b Array
 802          * @return array
 803          */
 804         static function mergeAttributes( $a, $b ) {
 805                 $out = array_merge( $a, $b );
 806                 if( isset( $a['class'] ) && isset( $b['class'] )
 807                 && is_string( $a['class'] ) && is_string( $b['class'] )
 808                 && $a['class'] !== $b['class'] ) {
 809                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 810                                 -1, PREG_SPLIT_NO_EMPTY );
 811                         $out['class'] = implode( ' ', array_unique( $classes ) );
 812                 }
 813                 return $out;
 814         }
 815
 816         /**
 817          * Pick apart some CSS and check it for forbidden or unsafe structures.
 818          * Returns a sanitized string. This sanitized string will have
 819          * character references and escape sequences decoded, and comments
 820          * stripped. If the input is just too evil, only a comment complaining
 821          * about evilness will be returned.
 822          *
 823          * Currently URL references, 'expression', 'tps' are forbidden.
 824          *
 825          * NOTE: Despite the fact that character references are decoded, the
 826          * returned string may contain character references given certain
 827          * clever input strings. These character references must
 828          * be escaped before the return value is embedded in HTML.
 829          *
 830          * @param $value String
 831          * @return String
 832          */
 833         static function checkCss( $value ) {
 834                 // Decode character references like &#123;
 835                 $value = Sanitizer::decodeCharReferences( $value );
 836
 837                 // Decode escape sequences and line continuation
 838                 // See the grammar in the CSS 2 spec, appendix D.
 839                 // This has to be done AFTER decoding character references.
 840                 // This means it isn't possible for this function to return
 841                 // unsanitized escape sequences. It is possible to manufacture
 842                 // input that contains character references that decode to
 843                 // escape sequences that decode to character references, but
 844                 // it's OK for the return value to contain character references
 845                 // because the caller is supposed to escape those anyway.
 846                 static $decodeRegex;
 847                 if ( !$decodeRegex ) {
 848                         $space = '[\\x20\\t\\r\\n\\f]';
 849                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 850                         $backslash = '\\\\';
 851                         $decodeRegex = "/ $backslash
 852                                 (?:
 853                                         ($nl) |  # 1. Line continuation
 854                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 855                                         (.) | # 3. backslash cancelling special meaning
 856                                         () | # 4. backslash at end of string
 857                                 )/xu";
 858                 }
 859                 $value = preg_replace_callback( $decodeRegex,
 860                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 861
 862                 // Remove any comments; IE gets token splitting wrong
 863                 // This must be done AFTER decoding character references and
 864                 // escape sequences, because those steps can introduce comments
 865                 // This step cannot introduce character references or escape
 866                 // sequences, because it replaces comments with spaces rather
 867                 // than removing them completely.
 868                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 869
 870                 // Remove anything after a comment-start token, to guard against
 871                 // incorrect client implementations.
 872                 $commentPos = strpos( $value, '/*' );
 873                 if ( $commentPos !== false ) {
 874                         $value = substr( $value, 0, $commentPos );
 875                 }
 876
 877                 // Reject problematic keywords and control characters
 878                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 879                         return '/* invalid control char */';
 880                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
 881                         return '/* insecure input */';
 882                 }
 883                 return $value;
 884         }
 885
 886         /**
 887          * @param $matches array
 888          * @return String
 889          */
 890         static function cssDecodeCallback( $matches ) {
 891                 if ( $matches[1] !== '' ) {
 892                         // Line continuation
 893                         return '';
 894                 } elseif ( $matches[2] !== '' ) {
 895                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 896                 } elseif ( $matches[3] !== '' ) {
 897                         $char = $matches[3];
 898                 } else {
 899                         $char = '\\';
 900                 }
 901                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 902                         // These characters need to be escaped in strings
 903                         // Clean up the escape sequence to avoid parsing errors by clients
 904                         return '\\' . dechex( ord( $char ) ) . ' ';
 905                 } else {
 906                         // Decode unnecessary escape
 907                         return $char;
 908                 }
 909         }
 910
 911         /**
 912          * Take a tag soup fragment listing an HTML element's attributes
 913          * and normalize it to well-formed XML, discarding unwanted attributes.
 914          * Output is safe for further wikitext processing, with escaping of
 915          * values that could trigger problems.
 916          *
 917          * - Normalizes attribute names to lowercase
 918          * - Discards attributes not on a whitelist for the given element
 919          * - Turns broken or invalid entities into plaintext
 920          * - Double-quotes all attribute values
 921          * - Attributes without values are given the name as attribute
 922          * - Double attributes are discarded
 923          * - Unsafe style attributes are discarded
 924          * - Prepends space if there are attributes.
 925          *
 926          * @param $text String
 927          * @param $element String
 928          * @return String
 929          */
 930         static function fixTagAttributes( $text, $element ) {
 931                 if( trim( $text ) == '' ) {
 932                         return '';
 933                 }
 934
 935                 $decoded = Sanitizer::decodeTagAttributes( $text );
 936                 $decoded = Sanitizer::fixDeprecatedAttributes( $decoded, $element );
 937                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
 938
 939                 $attribs = array();
 940                 foreach( $stripped as $attribute => $value ) {
 941                         $encAttribute = htmlspecialchars( $attribute );
 942                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 943
 944                         $attribs[] = "$encAttribute=\"$encValue\"";
 945                 }
 946                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 947         }
 948
 949         /**
 950          * Encode an attribute value for HTML output.
 951          * @param $text String
 952          * @return HTML-encoded text fragment
 953          */
 954         static function encodeAttribute( $text ) {
 955                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 956
 957                 // Whitespace is normalized during attribute decoding,
 958                 // so if we've been passed non-spaces we must encode them
 959                 // ahead of time or they won't be preserved.
 960                 $encValue = strtr( $encValue, array(
 961                         "\n" => '&#10;',
 962                         "\r" => '&#13;',
 963                         "\t" => '&#9;',
 964                 ) );
 965
 966                 return $encValue;
 967         }
 968
 969         /**
 970          * Encode an attribute value for HTML tags, with extra armoring
 971          * against further wiki processing.
 972          * @param $text String
 973          * @return HTML-encoded text fragment
 974          */
 975         static function safeEncodeAttribute( $text ) {
 976                 $encValue = Sanitizer::encodeAttribute( $text );
 977
 978                 # Templates and links may be expanded in later parsing,
 979                 # creating invalid or dangerous output. Suppress this.
 980                 $encValue = strtr( $encValue, array(
 981                         '<'    => '&lt;',   // This should never happen,
 982                         '>'    => '&gt;',   // we've received invalid input
 983                         '"'    => '&quot;', // which should have been escaped.
 984                         '{'    => '&#123;',
 985                         '['    => '&#91;',
 986                         "''"   => '&#39;&#39;',
 987                         'ISBN' => '&#73;SBN',
 988                         'RFC'  => '&#82;FC',
 989                         'PMID' => '&#80;MID',
 990                         '|'    => '&#124;',
 991                         '__'   => '&#95;_',
 992                 ) );
 993
 994                 # Stupid hack
 995                 $encValue = preg_replace_callback(
 996                         '/(' . wfUrlProtocols() . ')/',
 997                         array( 'Sanitizer', 'armorLinksCallback' ),
 998                         $encValue );
 999                 return $encValue;
1000         }
1001
1002         /**
1003          * Given a value, escape it so that it can be used in an id attribute and
1004          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1005          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1006          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1007          * escaped with lots of dots.
1008          *
1009          * To ensure we don't have to bother escaping anything, we also strip ', ",
1010          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1011          * We also strip # because it upsets IE, and % because it could be
1012          * ambiguous if it's part of something that looks like a percent escape
1013          * (which don't work reliably in fragments cross-browser).
1014          *
1015          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1016          *                                                          in the id and
1017          *                                                          name attributes
1018          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1019          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1020          *   HTML5 definition of id attribute
1021          *
1022          * @param $id String: id to escape
1023          * @param $options Mixed: string or array of strings (default is array()):
1024          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1025          *       so don't pay attention if the first character isn't valid at the
1026          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1027          *       false.
1028          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1029          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1030          *       anchors and links won't break.
1031          * @return String
1032          */
1033         static function escapeId( $id, $options = array() ) {
1034                 global $wgHtml5, $wgExperimentalHtmlIds;
1035                 $options = (array)$options;
1036
1037                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1038                         $id = Sanitizer::decodeCharReferences( $id );
1039                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1040                         $id = trim( $id, '_' );
1041                         if ( $id === '' ) {
1042                                 # Must have been all whitespace to start with.
1043                                 return '_';
1044                         } else {
1045                                 return $id;
1046                         }
1047                 }
1048
1049                 # HTML4-style escaping
1050                 static $replace = array(
1051                         '%3A' => ':',
1052                         '%' => '.'
1053                 );
1054
1055                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1056                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1057
1058                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1059                 && !in_array( 'noninitial', $options ) )  {
1060                         // Initial character must be a letter!
1061                         $id = "x$id";
1062                 }
1063                 return $id;
1064         }
1065
1066         /**
1067          * Given a value, escape it so that it can be used as a CSS class and
1068          * return it.
1069          *
1070          * @todo For extra validity, input should be validated UTF-8.
1071          *
1072          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1073          *
1074          * @param $class String
1075          * @return String
1076          */
1077         static function escapeClass( $class ) {
1078                 // Convert ugly stuff to underscores and kill underscores in ugly places
1079                 return rtrim(preg_replace(
1080                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
1081                         '_',
1082                         $class ), '_');
1083         }
1084
1085         /**
1086          * Given HTML input, escape with htmlspecialchars but un-escape entites.
1087          * This allows (generally harmless) entities like &#160; to survive.
1088          *
1089          * @param $html String to escape
1090          * @return String: escaped input
1091          */
1092         static function escapeHtmlAllowEntities( $html ) {
1093                 $html = Sanitizer::decodeCharReferences( $html );
1094                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1095                 # hurt.
1096                 $html = htmlspecialchars( $html, ENT_QUOTES );
1097                 return $html;
1098         }
1099
1100         /**
1101          * Regex replace callback for armoring links against further processing.
1102          * @param $matches Array
1103          * @return string
1104          */
1105         private static function armorLinksCallback( $matches ) {
1106                 return str_replace( ':', '&#58;', $matches[1] );
1107         }
1108
1109         /**
1110          * Return an associative array of attribute names and values from
1111          * a partial tag string. Attribute names are forces to lowercase,
1112          * character references are decoded to UTF-8 text.
1113          *
1114          * @param $text String
1115          * @return Array
1116          */
1117         public static function decodeTagAttributes( $text ) {
1118                 if( trim( $text ) == '' ) {
1119                         return array();
1120                 }
1121
1122                 $attribs = array();
1123                 $pairs = array();
1124                 if( !preg_match_all(
1125                         self::getAttribsRegex(),
1126                         $text,
1127                         $pairs,
1128                         PREG_SET_ORDER ) ) {
1129                         return $attribs;
1130                 }
1131
1132                 foreach( $pairs as $set ) {
1133                         $attribute = strtolower( $set[1] );
1134                         $value = Sanitizer::getTagAttributeCallback( $set );
1135
1136                         // Normalize whitespace
1137                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1138                         $value = trim( $value );
1139
1140                         // Decode character references
1141                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1142                 }
1143                 return $attribs;
1144         }
1145
1146         /**
1147          * Pick the appropriate attribute value from a match set from the
1148          * attribs regex matches.
1149          *
1150          * @param $set Array
1151          * @return String
1152          */
1153         private static function getTagAttributeCallback( $set ) {
1154                 if( isset( $set[6] ) ) {
1155                         # Illegal #XXXXXX color with no quotes.
1156                         return $set[6];
1157                 } elseif( isset( $set[5] ) ) {
1158                         # No quotes.
1159                         return $set[5];
1160                 } elseif( isset( $set[4] ) ) {
1161                         # Single-quoted
1162                         return $set[4];
1163                 } elseif( isset( $set[3] ) ) {
1164                         # Double-quoted
1165                         return $set[3];
1166                 } elseif( !isset( $set[2] ) ) {
1167                         # In XHTML, attributes must have a value.
1168                         # For 'reduced' form, return explicitly the attribute name here.
1169                         return $set[1];
1170                 } else {
1171                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1172                 }
1173         }
1174
1175         /**
1176          * Normalize whitespace and character references in an XML source-
1177          * encoded text for an attribute value.
1178          *
1179          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1180          * but note that we're not returning the value, but are returning
1181          * XML source fragments that will be slapped into output.
1182          *
1183          * @param $text String
1184          * @return String
1185          */
1186         private static function normalizeAttributeValue( $text ) {
1187                 return str_replace( '"', '&quot;',
1188                         self::normalizeWhitespace(
1189                                 Sanitizer::normalizeCharReferences( $text ) ) );
1190         }
1191
1192         /**
1193          * @param $text string
1194          * @return mixed
1195          */
1196         private static function normalizeWhitespace( $text ) {
1197                 return preg_replace(
1198                         '/\r\n|[\x20\x0d\x0a\x09]/',
1199                         ' ',
1200                         $text );
1201         }
1202
1203         /**
1204          * Normalizes whitespace in a section name, such as might be returned
1205          * by Parser::stripSectionName(), for use in the id's that are used for
1206          * section links.
1207          *
1208          * @param $section String
1209          * @return String
1210          */
1211         static function normalizeSectionNameWhitespace( $section ) {
1212                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1213         }
1214
1215         /**
1216          * Ensure that any entities and character references are legal
1217          * for XML and XHTML specifically. Any stray bits will be
1218          * &amp;-escaped to result in a valid text fragment.
1219          *
1220          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1221          *   numericized (this way we're well-formed even without a DTD)
1222          * b. any numeric char refs must be legal chars, not invalid or forbidden
1223          * c. use &#x, not &#X
1224          * d. fix or reject non-valid attributes
1225          *
1226          * @param $text String
1227          * @return String
1228          * @private
1229          */
1230         static function normalizeCharReferences( $text ) {
1231                 return preg_replace_callback(
1232                         self::CHAR_REFS_REGEX,
1233                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1234                         $text );
1235         }
1236         /**
1237          * @param $matches String
1238          * @return String
1239          */
1240         static function normalizeCharReferencesCallback( $matches ) {
1241                 $ret = null;
1242                 if( $matches[1] != '' ) {
1243                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1244                 } elseif( $matches[2] != '' ) {
1245                         $ret = Sanitizer::decCharReference( $matches[2] );
1246                 } elseif( $matches[3] != ''  ) {
1247                         $ret = Sanitizer::hexCharReference( $matches[3] );
1248                 }
1249                 if( is_null( $ret ) ) {
1250                         return htmlspecialchars( $matches[0] );
1251                 } else {
1252                         return $ret;
1253                 }
1254         }
1255
1256         /**
1257          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1258          * return the equivalent numeric entity reference (except for the core &lt;
1259          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1260          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1261          * pseudo-entity source (eg &amp;foo;)
1262          *
1263          * @param $name String
1264          * @return String
1265          */
1266         static function normalizeEntity( $name ) {
1267                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1268                         return '&' . self::$htmlEntityAliases[$name] . ';';
1269                 } elseif ( in_array( $name,
1270                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1271                         return "&$name;";
1272                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1273                         return '&#' . self::$htmlEntities[$name] . ';';
1274                 } else {
1275                         return "&amp;$name;";
1276                 }
1277         }
1278
1279         /**
1280          * @param $codepoint
1281          * @return null|string
1282          */
1283         static function decCharReference( $codepoint ) {
1284                 $point = intval( $codepoint );
1285                 if( Sanitizer::validateCodepoint( $point ) ) {
1286                         return sprintf( '&#%d;', $point );
1287                 } else {
1288                         return null;
1289                 }
1290         }
1291
1292         /**
1293          * @param $codepoint
1294          * @return null|string
1295          */
1296         static function hexCharReference( $codepoint ) {
1297                 $point = hexdec( $codepoint );
1298                 if( Sanitizer::validateCodepoint( $point ) ) {
1299                         return sprintf( '&#x%x;', $point );
1300                 } else {
1301                         return null;
1302                 }
1303         }
1304
1305         /**
1306          * Returns true if a given Unicode codepoint is a valid character in XML.
1307          * @param $codepoint Integer
1308          * @return Boolean
1309          */
1310         private static function validateCodepoint( $codepoint ) {
1311                 return ($codepoint ==    0x09)
1312                         || ($codepoint ==    0x0a)
1313                         || ($codepoint ==    0x0d)
1314                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1315                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1316                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1317         }
1318
1319         /**
1320          * Decode any character references, numeric or named entities,
1321          * in the text and return a UTF-8 string.
1322          *
1323          * @param $text String
1324          * @return String
1325          */
1326         public static function decodeCharReferences( $text ) {
1327                 return preg_replace_callback(
1328                         self::CHAR_REFS_REGEX,
1329                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1330                         $text );
1331         }
1332
1333         /**
1334          * Decode any character references, numeric or named entities,
1335          * in the next and normalize the resulting string. (bug 14952)
1336          *
1337          * This is useful for page titles, not for text to be displayed,
1338          * MediaWiki allows HTML entities to escape normalization as a feature.
1339          *
1340          * @param $text String (already normalized, containing entities)
1341          * @return String (still normalized, without entities)
1342          */
1343         public static function decodeCharReferencesAndNormalize( $text ) {
1344                 global $wgContLang;
1345                 $text = preg_replace_callback(
1346                         self::CHAR_REFS_REGEX,
1347                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1348                         $text, /* limit */ -1, $count );
1349
1350                 if ( $count ) {
1351                         return $wgContLang->normalize( $text );
1352                 } else {
1353                         return $text;
1354                 }
1355         }
1356
1357         /**
1358          * @param $matches String
1359          * @return String
1360          */
1361         static function decodeCharReferencesCallback( $matches ) {
1362                 if( $matches[1] != '' ) {
1363                         return Sanitizer::decodeEntity( $matches[1] );
1364                 } elseif( $matches[2] != '' ) {
1365                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1366                 } elseif( $matches[3] != ''  ) {
1367                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1368                 }
1369                 # Last case should be an ampersand by itself
1370                 return $matches[0];
1371         }
1372
1373         /**
1374          * Return UTF-8 string for a codepoint if that is a valid
1375          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1376          * @param $codepoint Integer
1377          * @return String
1378          * @private
1379          */
1380         static function decodeChar( $codepoint ) {
1381                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1382                         return codepointToUtf8( $codepoint );
1383                 } else {
1384                         return UTF8_REPLACEMENT;
1385                 }
1386         }
1387
1388         /**
1389          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1390          * return the UTF-8 encoding of that character. Otherwise, returns
1391          * pseudo-entity source (eg &foo;)
1392          *
1393          * @param $name String
1394          * @return String
1395          */
1396         static function decodeEntity( $name ) {
1397                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1398                         $name = self::$htmlEntityAliases[$name];
1399                 }
1400                 if( isset( self::$htmlEntities[$name] ) ) {
1401                         return codepointToUtf8( self::$htmlEntities[$name] );
1402                 } else {
1403                         return "&$name;";
1404                 }
1405         }
1406
1407         /**
1408          * Fetch the whitelist of acceptable attributes for a given element name.
1409          *
1410          * @param $element String
1411          * @return Array
1412          */
1413         static function attributeWhitelist( $element ) {
1414                 static $list;
1415                 if( !isset( $list ) ) {
1416                         $list = Sanitizer::setupAttributeWhitelist();
1417                 }
1418                 return isset( $list[$element] )
1419                         ? $list[$element]
1420                         : array();
1421         }
1422
1423         /**
1424          * Foreach array key (an allowed HTML element), return an array
1425          * of allowed attributes
1426          * @return Array
1427          */
1428         static function setupAttributeWhitelist() {
1429                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1430
1431                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1432
1433                 if ( $wgAllowRdfaAttributes ) {
1434                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1435                         $common = array_merge( $common, array(
1436                             'about', 'property', 'resource', 'datatype', 'typeof',
1437                         ) );
1438                 }
1439
1440                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1441                         # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1442                         $common = array_merge( $common, array(
1443                             'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1444                         ) );
1445                 }
1446
1447                 $block = array_merge( $common, array( 'align' ) );
1448                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1449                 $tablecell = array( 'abbr',
1450                                     'axis',
1451                                     'headers',
1452                                     'scope',
1453                                     'rowspan',
1454                                     'colspan',
1455                                     'nowrap', # deprecated
1456                                     'width',  # deprecated
1457                                     'height', # deprecated
1458                                     'bgcolor' # deprecated
1459                                     );
1460
1461                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1462                 # See: http://www.w3.org/TR/html4/
1463                 $whitelist = array (
1464                         # 7.5.4
1465                         'div'        => $block,
1466                         'center'     => $common, # deprecated
1467                         'span'       => $block, # ??
1468
1469                         # 7.5.5
1470                         'h1'         => $block,
1471                         'h2'         => $block,
1472                         'h3'         => $block,
1473                         'h4'         => $block,
1474                         'h5'         => $block,
1475                         'h6'         => $block,
1476
1477                         # 7.5.6
1478                         # address
1479
1480                         # 8.2.4
1481                         # bdo
1482
1483                         # 9.2.1
1484                         'em'         => $common,
1485                         'strong'     => $common,
1486                         'cite'       => $common,
1487                         'dfn'        => $common,
1488                         'code'       => $common,
1489                         'samp'       => $common,
1490                         'kbd'        => $common,
1491                         'var'        => $common,
1492                         'abbr'       => $common,
1493                         # acronym
1494
1495                         # 9.2.2
1496                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1497                         # q
1498
1499                         # 9.2.3
1500                         'sub'        => $common,
1501                         'sup'        => $common,
1502
1503                         # 9.3.1
1504                         'p'          => $block,
1505
1506                         # 9.3.2
1507                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1508
1509                         # 9.3.4
1510                         'pre'        => array_merge( $common, array( 'width' ) ),
1511
1512                         # 9.4
1513                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1514                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1515
1516                         # 10.2
1517                         'ul'         => array_merge( $common, array( 'type' ) ),
1518                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1519                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1520
1521                         # 10.3
1522                         'dl'         => $common,
1523                         'dd'         => $common,
1524                         'dt'         => $common,
1525
1526                         # 11.2.1
1527                         'table'      => array_merge( $common,
1528                                                                 array( 'summary', 'width', 'border', 'frame',
1529                                                                                 'rules', 'cellspacing', 'cellpadding',
1530                                                                                 'align', 'bgcolor',
1531                                                                 ) ),
1532
1533                         # 11.2.2
1534                         'caption'    => array_merge( $common, array( 'align' ) ),
1535
1536                         # 11.2.3
1537                         'thead'      => array_merge( $common, $tablealign ),
1538                         'tfoot'      => array_merge( $common, $tablealign ),
1539                         'tbody'      => array_merge( $common, $tablealign ),
1540
1541                         # 11.2.4
1542                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1543                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1544
1545                         # 11.2.5
1546                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1547
1548                         # 11.2.6
1549                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1550                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1551
1552                         # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1553                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1554
1555                         # 13.2
1556                         # Not usually allowed, but may be used for extension-style hooks
1557                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1558                         # true
1559                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1560
1561                         # 15.2.1
1562                         'tt'         => $common,
1563                         'b'          => $common,
1564                         'i'          => $common,
1565                         'big'        => $common,
1566                         'small'      => $common,
1567                         'strike'     => $common,
1568                         's'          => $common,
1569                         'u'          => $common,
1570
1571                         # 15.2.2
1572                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1573                         # basefont
1574
1575                         # 15.3
1576                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1577
1578                         # XHTML Ruby annotation text module, simple ruby only.
1579                         # http://www.w3c.org/TR/ruby/
1580                         'ruby'       => $common,
1581                         # rbc
1582                         # rtc
1583                         'rb'         => $common,
1584                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1585                         'rp'         => $common,
1586
1587                         # MathML root element, where used for extensions
1588                         # 'title' may not be 100% valid here; it's XHTML
1589                         # http://www.w3.org/TR/REC-MathML/
1590                         'math'       => array( 'class', 'style', 'id', 'title' ),
1591                         );
1592                 return $whitelist;
1593         }
1594
1595         /**
1596          * Take a fragment of (potentially invalid) HTML and return
1597          * a version with any tags removed, encoded as plain text.
1598          *
1599          * Warning: this return value must be further escaped for literal
1600          * inclusion in HTML output as of 1.10!
1601          *
1602          * @param $text String: HTML fragment
1603          * @return String
1604          */
1605         static function stripAllTags( $text ) {
1606                 # Actual <tags>
1607                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1608
1609                 # Normalize &entities and whitespace
1610                 $text = self::decodeCharReferences( $text );
1611                 $text = self::normalizeWhitespace( $text );
1612
1613                 return $text;
1614         }
1615
1616         /**
1617          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1618          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1619          * PHP 5.1 doesn't.
1620          *
1621          * Use for passing XHTML fragments to PHP's XML parsing functions
1622          *
1623          * @return String
1624          */
1625         static function hackDocType() {
1626                 $out = "<!DOCTYPE html [\n";
1627                 foreach( self::$htmlEntities as $entity => $codepoint ) {
1628                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1629                 }
1630                 $out .= "]>\n";
1631                 return $out;
1632         }
1633
1634         /**
1635          * @param $url string
1636          * @return mixed|string
1637          */
1638         static function cleanUrl( $url ) {
1639                 # Normalize any HTML entities in input. They will be
1640                 # re-escaped by makeExternalLink().
1641                 $url = Sanitizer::decodeCharReferences( $url );
1642
1643                 # Escape any control characters introduced by the above step
1644                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1645                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1646
1647                 # Validate hostname portion
1648                 $matches = array();
1649                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1650                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1651
1652                         // Characters that will be ignored in IDNs.
1653                         // http://tools.ietf.org/html/3454#section-3.1
1654                         // Strip them before further processing so blacklists and such work.
1655                         $strip = "/
1656                                 \\s|          # general whitespace
1657                                 \xc2\xad|     # 00ad SOFT HYPHEN
1658                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1659                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1660                                 \xe2\x81\xa0| # 2060 WORD JOINER
1661                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1662                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1663                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1664                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1665                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1666                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1667                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1668                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe00f VARIATION SELECTOR-1-16
1669                                 /xuD";
1670
1671                         $host = preg_replace( $strip, '', $host );
1672
1673                         // @todo FIXME: Validate hostnames here
1674
1675                         return $protocol . $host . $rest;
1676                 } else {
1677                         return $url;
1678                 }
1679         }
1680
1681         /**
1682          * @param $matches array
1683          * @return string
1684          */
1685         static function cleanUrlCallback( $matches ) {
1686                 return urlencode( $matches[0] );
1687         }
1688
1689         /**
1690          * Does a string look like an e-mail address?
1691          *
1692          * This validates an email address using an HTML5 specification found at:
1693          * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1694          * Which as of 2011-01-24 says:
1695          *
1696          *   A valid e-mail address is a string that matches the ABNF production
1697          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1698          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1699          *   3.5.
1700          *
1701          * This function is an implementation of the specification as requested in
1702          * bug 22449.
1703          *
1704          * Client-side forms will use the same standard validation rules via JS or
1705          * HTML 5 validation; additional restrictions can be enforced server-side
1706          * by extensions via the 'isValidEmailAddr' hook.
1707          *
1708          * Note that this validation doesn't 100% match RFC 2822, but is believed
1709          * to be liberal enough for wide use. Some invalid addresses will still
1710          * pass validation here.
1711          *
1712          * @since 1.18
1713          *
1714          * @param $addr String E-mail address
1715          * @return Bool
1716          */
1717         public static function validateEmail( $addr ) {
1718                 $result = null;
1719                 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1720                         return $result;
1721                 }
1722
1723                 // Please note strings below are enclosed in brackets [], this make the
1724                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1725                 // See bug 26948
1726                 $rfc5322_atext   = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1727                 $rfc1034_ldh_str = "a-z0-9\\-" ;
1728
1729                 $HTML5_email_regexp = "/
1730                 ^                      # start of string
1731                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1732                 @                      # 'apostrophe'
1733                 [$rfc1034_ldh_str]+       # First domain part
1734                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1735                 $                      # End of string
1736                 /ix" ; // case Insensitive, eXtended
1737
1738                 return (bool) preg_match( $HTML5_email_regexp, $addr );
1739         }
1740 }