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                                                                         wfSuppressWarnings();
 469                                                                         $ot = array_pop( $optstack );
 470                                                                         wfRestoreWarnings();
 471                                                                         while ( $ot ) {
 472                                                                                 array_push( $tagstack, $ot );
 473                                                                                 wfSuppressWarnings();
 474                                                                                 $ot = array_pop( $optstack );
 475                                                                                 wfRestoreWarnings();
 476                                                                         }
 477                                                                 }
 478                                                         } else {
 479                                                                 @array_push( $tagstack, $ot );
 480                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 481                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 482                                                                         $badtag = true;
 483                                                                 }
 484                                                         }
 485                                                 } else {
 486                                                         if ( $t == 'table' ) {
 487                                                                 $tagstack = array_pop( $tablestack );
 488                                                         }
 489                                                 }
 490                                                 $newparams = '';
 491                                         } else {
 492                                                 # Keep track for later
 493                                                 if ( isset( $tabletags[$t] ) &&
 494                                                 !in_array( 'table', $tagstack ) ) {
 495                                                         $badtag = true;
 496                                                 } elseif ( in_array( $t, $tagstack ) &&
 497                                                 !isset( $htmlnest [$t ] ) ) {
 498                                                         $badtag = true;
 499                                                 # Is it a self closed htmlpair ? (bug 5487)
 500                                                 } elseif ( $brace == '/>' &&
 501                                                 isset( $htmlpairs[$t] ) ) {
 502                                                         $badtag = true;
 503                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 504                                                         # Hack to force empty tag for uncloseable elements
 505                                                         $brace = '/>';
 506                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 507                                                         # Hack to not close $htmlsingle tags
 508                                                         $brace = null;
 509                                                 } elseif ( isset( $tabletags[$t] )
 510                                                 && in_array( $t, $tagstack ) ) {
 511                                                         // New table tag but forgot to close the previous one
 512                                                         $text .= "</$t>";
 513                                                 } else {
 514                                                         if ( $t == 'table' ) {
 515                                                                 array_push( $tablestack, $tagstack );
 516                                                                 $tagstack = array();
 517                                                         }
 518                                                         array_push( $tagstack, $t );
 519                                                 }
 520
 521                                                 # Replace any variables or template parameters with
 522                                                 # plaintext results.
 523                                                 if( is_callable( $processCallback ) ) {
 524                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 525                                                 }
 526
 527                                                 # Strip non-approved attributes from the tag
 528                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 529                                         }
 530                                         if ( !$badtag ) {
 531                                                 $rest = str_replace( '>', '&gt;', $rest );
 532                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 533                                                 $text .= "<$slash$t$newparams$close>$rest";
 534                                                 continue;
 535                                         }
 536                                 }
 537                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 538                         }
 539                         # Close off any remaining tags
 540                         while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
 541                                 $text .= "</$t>\n";
 542                                 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
 543                         }
 544                 } else {
 545                         # this might be possible using tidy itself
 546                         foreach ( $bits as $x ) {
 547                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 548                                 $x, $regs );
 549                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 550                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 551                                         if( is_callable( $processCallback ) ) {
 552                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 553                                         }
 554                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 555                                         $rest = str_replace( '>', '&gt;', $rest );
 556                                         $text .= "<$slash$t$newparams$brace$rest";
 557                                 } else {
 558                                         $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 559                                 }
 560                         }
 561                 }
 562                 wfProfileOut( __METHOD__ );
 563                 return $text;
 564         }
 565
 566         /**
 567          * Remove '<!--', '-->', and everything between.
 568          * To avoid leaving blank lines, when a comment is both preceded
 569          * and followed by a newline (ignoring spaces), trim leading and
 570          * trailing spaces and one of the newlines.
 571          *
 572          * @private
 573          * @param $text String
 574          * @return string
 575          */
 576         static function removeHTMLcomments( $text ) {
 577                 wfProfileIn( __METHOD__ );
 578                 while (($start = strpos($text, '<!--')) !== false) {
 579                         $end = strpos($text, '-->', $start + 4);
 580                         if ($end === false) {
 581                                 # Unterminated comment; bail out
 582                                 break;
 583                         }
 584
 585                         $end += 3;
 586
 587                         # Trim space and newline if the comment is both
 588                         # preceded and followed by a newline
 589                         $spaceStart = max($start - 1, 0);
 590                         $spaceLen = $end - $spaceStart;
 591                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 592                                 $spaceStart--;
 593                                 $spaceLen++;
 594                         }
 595                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 596                                 $spaceLen++;
 597                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 598                                 # Remove the comment, leading and trailing
 599                                 # spaces, and leave only one newline.
 600                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 601                         }
 602                         else {
 603                                 # Remove just the comment.
 604                                 $text = substr_replace($text, '', $start, $end - $start);
 605                         }
 606                 }
 607                 wfProfileOut( __METHOD__ );
 608                 return $text;
 609         }
 610
 611         /**
 612          * Take an array of attribute names and values and fix some deprecated values
 613          * for the given element type.
 614          * This does not validate properties, so you should ensure that you call
 615          * validateTagAttributes AFTER this to ensure that the resulting style rule
 616          * this may add is safe.
 617          *
 618          * - Converts most presentational attributes like align into inline css
 619          *
 620          * @param $attribs Array
 621          * @param $element String
 622          * @return Array
 623          */
 624         static function fixDeprecatedAttributes( $attribs, $element ) {
 625                 global $wgHtml5, $wgCleanupPresentationalAttributes;
 626
 627                 // presentational attributes were removed from html5, we can leave them
 628                 // in when html5 is turned off
 629                 if ( !$wgHtml5 || !$wgCleanupPresentationalAttributes ) {
 630                         return $attribs;
 631                 }
 632
 633                 $table = array( 'table' );
 634                 $cells = array( 'td', 'th' );
 635                 $colls = array( 'col', 'colgroup' );
 636                 $tblocks = array( 'tbody', 'tfoot', 'thead' );
 637                 $h = array( 'h1', 'h2', 'h3', 'h4', 'h5', 'h6' );
 638
 639                 $presentationalAttribs = array(
 640                         'align' => array( 'text-align', array_merge( array( 'caption', 'hr', 'div', 'p', 'tr' ), $table, $cells, $colls, $tblocks, $h ) ),
 641                         'clear' => array( 'clear', array( 'br' ) ),
 642                         'height' => array( 'height', $cells ),
 643                         'nowrap' => array( 'white-space', $cells ),
 644                         'size' => array( 'height', array( 'hr' ) ),
 645                         'type' => array( 'list-style-type', array( 'li', 'ol', 'ul' ) ),
 646                         'valign' => array( 'vertical-align', array_merge( $cells, $colls, $tblocks ) ),
 647                         'width' => array( 'width', array_merge( array( 'hr', 'pre' ), $table, $cells, $colls ) ),
 648                 );
 649
 650                 // Ensure that any upper case or mixed case attributes are converted to lowercase
 651                 foreach ( $attribs as $attribute => $value ) {
 652                         if ( $attribute !== strtolower( $attribute ) && array_key_exists( strtolower( $attribute ), $presentationalAttribs ) ) {
 653                                 $attribs[strtolower( $attribute )] = $value;
 654                                 unset( $attribs[$attribute] );
 655                         }
 656                 }
 657
 658                 $style = "";
 659                 foreach ( $presentationalAttribs as $attribute => $info ) {
 660                         list( $property, $elements ) = $info;
 661
 662                         // Skip if this attribute is not relevant to this element
 663                         if ( !in_array( $element, $elements ) ) {
 664                                 continue;
 665                         }
 666
 667                         // Skip if the attribute is not used
 668                         if ( !array_key_exists( $attribute, $attribs ) ) {
 669                                 continue;
 670                         }
 671
 672                         $value = $attribs[$attribute];
 673
 674                         // For nowrap the value should be nowrap instead of whatever text is in the value
 675                         if ( $attribute === 'nowrap' ) {
 676                                 $value = 'nowrap';
 677                         }
 678
 679                         // clear="all" is clear: both; in css
 680                         if ( $attribute === 'clear' && strtolower( $value ) === 'all' ) {
 681                                 $value = 'both';
 682                         }
 683
 684                         // Size based properties should have px applied to them if they have no unit
 685                         if ( in_array( $attribute, array( 'height', 'width', 'size' ) ) ) {
 686                                 if ( preg_match( '/^[\d.]+$/', $value ) ) {
 687                                         $value = "{$value}px";
 688                                 }
 689                         }
 690
 691                         $style .= " $property: $value;";
 692
 693                         unset( $attribs[$attribute] );
 694                 }
 695
 696                 if ( $style ) {
 697                         // Prepend our style rules so that they can be overridden by user css
 698                         if ( isset($attribs['style']) ) {
 699                                 $style .= " " . $attribs['style'];
 700                         }
 701                         $attribs['style'] = trim($style);
 702                 }
 703
 704                 return $attribs;
 705         }
 706
 707         /**
 708          * Take an array of attribute names and values and normalize or discard
 709          * illegal values for the given element type.
 710          *
 711          * - Discards attributes not on a whitelist for the given element
 712          * - Unsafe style attributes are discarded
 713          * - Invalid id attributes are reencoded
 714          *
 715          * @param $attribs Array
 716          * @param $element String
 717          * @return Array
 718          *
 719          * @todo Check for legal values where the DTD limits things.
 720          * @todo Check for unique id attribute :P
 721          */
 722         static function validateTagAttributes( $attribs, $element ) {
 723                 return Sanitizer::validateAttributes( $attribs,
 724                         Sanitizer::attributeWhitelist( $element ) );
 725         }
 726
 727         /**
 728          * Take an array of attribute names and values and normalize or discard
 729          * illegal values for the given whitelist.
 730          *
 731          * - Discards attributes not the given whitelist
 732          * - Unsafe style attributes are discarded
 733          * - Invalid id attributes are reencoded
 734          *
 735          * @param $attribs Array
 736          * @param $whitelist Array: list of allowed attribute names
 737          * @return Array
 738          *
 739          * @todo Check for legal values where the DTD limits things.
 740          * @todo Check for unique id attribute :P
 741          */
 742         static function validateAttributes( $attribs, $whitelist ) {
 743                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
 744
 745                 $whitelist = array_flip( $whitelist );
 746                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 747
 748                 $out = array();
 749                 foreach( $attribs as $attribute => $value ) {
 750                         #allow XML namespace declaration if RDFa is enabled
 751                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 752                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 753                                         $out[$attribute] = $value;
 754                                 }
 755
 756                                 continue;
 757                         }
 758
 759                         # Allow any attribute beginning with "data-", if in HTML5 mode
 760                         if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
 761                                 continue;
 762                         }
 763
 764                         # Strip javascript "expression" from stylesheets.
 765                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 766                         if( $attribute == 'style' ) {
 767                                 $value = Sanitizer::checkCss( $value );
 768                         }
 769
 770                         if ( $attribute === 'id' ) {
 771                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 772                         }
 773
 774                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 775                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 776                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 777                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 778                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 779                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 780
 781                                 //Paranoia. Allow "simple" values but suppress javascript
 782                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 783                                         continue;
 784                                 }
 785                         }
 786
 787                         # NOTE: even though elements using href/src are not allowed directly, supply
 788                         #       validation code that can be used by tag hook handlers, etc
 789                         if ( $attribute === 'href' || $attribute === 'src' ) {
 790                                 if ( !preg_match( $hrefExp, $value ) ) {
 791                                         continue; //drop any href or src attributes not using an allowed protocol.
 792                                                   //NOTE: this also drops all relative URLs
 793                                 }
 794                         }
 795
 796                         // If this attribute was previously set, override it.
 797                         // Output should only have one attribute of each name.
 798                         $out[$attribute] = $value;
 799                 }
 800
 801                 if ( $wgAllowMicrodataAttributes ) {
 802                         # itemtype, itemid, itemref don't make sense without itemscope
 803                         if ( !array_key_exists( 'itemscope', $out ) ) {
 804                                 unset( $out['itemtype'] );
 805                                 unset( $out['itemid'] );
 806                                 unset( $out['itemref'] );
 807                         }
 808                         # TODO: Strip itemprop if we aren't descendants of an itemscope.
 809                 }
 810                 return $out;
 811         }
 812
 813         /**
 814          * Merge two sets of HTML attributes.  Conflicting items in the second set
 815          * will override those in the first, except for 'class' attributes which
 816          * will be combined (if they're both strings).
 817          *
 818          * @todo implement merging for other attributes such as style
 819          * @param $a Array
 820          * @param $b Array
 821          * @return array
 822          */
 823         static function mergeAttributes( $a, $b ) {
 824                 $out = array_merge( $a, $b );
 825                 if( isset( $a['class'] ) && isset( $b['class'] )
 826                 && is_string( $a['class'] ) && is_string( $b['class'] )
 827                 && $a['class'] !== $b['class'] ) {
 828                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 829                                 -1, PREG_SPLIT_NO_EMPTY );
 830                         $out['class'] = implode( ' ', array_unique( $classes ) );
 831                 }
 832                 return $out;
 833         }
 834
 835         /**
 836          * Pick apart some CSS and check it for forbidden or unsafe structures.
 837          * Returns a sanitized string. This sanitized string will have
 838          * character references and escape sequences decoded, and comments
 839          * stripped. If the input is just too evil, only a comment complaining
 840          * about evilness will be returned.
 841          *
 842          * Currently URL references, 'expression', 'tps' are forbidden.
 843          *
 844          * NOTE: Despite the fact that character references are decoded, the
 845          * returned string may contain character references given certain
 846          * clever input strings. These character references must
 847          * be escaped before the return value is embedded in HTML.
 848          *
 849          * @param $value String
 850          * @return String
 851          */
 852         static function checkCss( $value ) {
 853                 // Decode character references like &#123;
 854                 $value = Sanitizer::decodeCharReferences( $value );
 855
 856                 // Decode escape sequences and line continuation
 857                 // See the grammar in the CSS 2 spec, appendix D.
 858                 // This has to be done AFTER decoding character references.
 859                 // This means it isn't possible for this function to return
 860                 // unsanitized escape sequences. It is possible to manufacture
 861                 // input that contains character references that decode to
 862                 // escape sequences that decode to character references, but
 863                 // it's OK for the return value to contain character references
 864                 // because the caller is supposed to escape those anyway.
 865                 static $decodeRegex;
 866                 if ( !$decodeRegex ) {
 867                         $space = '[\\x20\\t\\r\\n\\f]';
 868                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 869                         $backslash = '\\\\';
 870                         $decodeRegex = "/ $backslash
 871                                 (?:
 872                                         ($nl) |  # 1. Line continuation
 873                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 874                                         (.) | # 3. backslash cancelling special meaning
 875                                         () | # 4. backslash at end of string
 876                                 )/xu";
 877                 }
 878                 $value = preg_replace_callback( $decodeRegex,
 879                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 880
 881                 // Remove any comments; IE gets token splitting wrong
 882                 // This must be done AFTER decoding character references and
 883                 // escape sequences, because those steps can introduce comments
 884                 // This step cannot introduce character references or escape
 885                 // sequences, because it replaces comments with spaces rather
 886                 // than removing them completely.
 887                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 888
 889                 // Remove anything after a comment-start token, to guard against
 890                 // incorrect client implementations.
 891                 $commentPos = strpos( $value, '/*' );
 892                 if ( $commentPos !== false ) {
 893                         $value = substr( $value, 0, $commentPos );
 894                 }
 895
 896                 // Reject problematic keywords and control characters
 897                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 898                         return '/* invalid control char */';
 899                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
 900                         return '/* insecure input */';
 901                 }
 902                 return $value;
 903         }
 904
 905         /**
 906          * @param $matches array
 907          * @return String
 908          */
 909         static function cssDecodeCallback( $matches ) {
 910                 if ( $matches[1] !== '' ) {
 911                         // Line continuation
 912                         return '';
 913                 } elseif ( $matches[2] !== '' ) {
 914                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 915                 } elseif ( $matches[3] !== '' ) {
 916                         $char = $matches[3];
 917                 } else {
 918                         $char = '\\';
 919                 }
 920                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 921                         // These characters need to be escaped in strings
 922                         // Clean up the escape sequence to avoid parsing errors by clients
 923                         return '\\' . dechex( ord( $char ) ) . ' ';
 924                 } else {
 925                         // Decode unnecessary escape
 926                         return $char;
 927                 }
 928         }
 929
 930         /**
 931          * Take a tag soup fragment listing an HTML element's attributes
 932          * and normalize it to well-formed XML, discarding unwanted attributes.
 933          * Output is safe for further wikitext processing, with escaping of
 934          * values that could trigger problems.
 935          *
 936          * - Normalizes attribute names to lowercase
 937          * - Discards attributes not on a whitelist for the given element
 938          * - Turns broken or invalid entities into plaintext
 939          * - Double-quotes all attribute values
 940          * - Attributes without values are given the name as attribute
 941          * - Double attributes are discarded
 942          * - Unsafe style attributes are discarded
 943          * - Prepends space if there are attributes.
 944          *
 945          * @param $text String
 946          * @param $element String
 947          * @return String
 948          */
 949         static function fixTagAttributes( $text, $element ) {
 950                 if( trim( $text ) == '' ) {
 951                         return '';
 952                 }
 953
 954                 $decoded = Sanitizer::decodeTagAttributes( $text );
 955                 $decoded = Sanitizer::fixDeprecatedAttributes( $decoded, $element );
 956                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
 957
 958                 $attribs = array();
 959                 foreach( $stripped as $attribute => $value ) {
 960                         $encAttribute = htmlspecialchars( $attribute );
 961                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 962
 963                         $attribs[] = "$encAttribute=\"$encValue\"";
 964                 }
 965                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 966         }
 967
 968         /**
 969          * Encode an attribute value for HTML output.
 970          * @param $text String
 971          * @return HTML-encoded text fragment
 972          */
 973         static function encodeAttribute( $text ) {
 974                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 975
 976                 // Whitespace is normalized during attribute decoding,
 977                 // so if we've been passed non-spaces we must encode them
 978                 // ahead of time or they won't be preserved.
 979                 $encValue = strtr( $encValue, array(
 980                         "\n" => '&#10;',
 981                         "\r" => '&#13;',
 982                         "\t" => '&#9;',
 983                 ) );
 984
 985                 return $encValue;
 986         }
 987
 988         /**
 989          * Encode an attribute value for HTML tags, with extra armoring
 990          * against further wiki processing.
 991          * @param $text String
 992          * @return HTML-encoded text fragment
 993          */
 994         static function safeEncodeAttribute( $text ) {
 995                 $encValue = Sanitizer::encodeAttribute( $text );
 996
 997                 # Templates and links may be expanded in later parsing,
 998                 # creating invalid or dangerous output. Suppress this.
 999                 $encValue = strtr( $encValue, array(
1000                         '<'    => '&lt;',   // This should never happen,
1001                         '>'    => '&gt;',   // we've received invalid input
1002                         '"'    => '&quot;', // which should have been escaped.
1003                         '{'    => '&#123;',
1004                         '['    => '&#91;',
1005                         "''"   => '&#39;&#39;',
1006                         'ISBN' => '&#73;SBN',
1007                         'RFC'  => '&#82;FC',
1008                         'PMID' => '&#80;MID',
1009                         '|'    => '&#124;',
1010                         '__'   => '&#95;_',
1011                 ) );
1012
1013                 # Stupid hack
1014                 $encValue = preg_replace_callback(
1015                         '/(' . wfUrlProtocols() . ')/',
1016                         array( 'Sanitizer', 'armorLinksCallback' ),
1017                         $encValue );
1018                 return $encValue;
1019         }
1020
1021         /**
1022          * Given a value, escape it so that it can be used in an id attribute and
1023          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1024          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1025          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1026          * escaped with lots of dots.
1027          *
1028          * To ensure we don't have to bother escaping anything, we also strip ', ",
1029          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1030          * We also strip # because it upsets IE, and % because it could be
1031          * ambiguous if it's part of something that looks like a percent escape
1032          * (which don't work reliably in fragments cross-browser).
1033          *
1034          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1035          *                                                          in the id and
1036          *                                                          name attributes
1037          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1038          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1039          *   HTML5 definition of id attribute
1040          *
1041          * @param $id String: id to escape
1042          * @param $options Mixed: string or array of strings (default is array()):
1043          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1044          *       so don't pay attention if the first character isn't valid at the
1045          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1046          *       false.
1047          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1048          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1049          *       anchors and links won't break.
1050          * @return String
1051          */
1052         static function escapeId( $id, $options = array() ) {
1053                 global $wgHtml5, $wgExperimentalHtmlIds;
1054                 $options = (array)$options;
1055
1056                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1057                         $id = Sanitizer::decodeCharReferences( $id );
1058                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1059                         $id = trim( $id, '_' );
1060                         if ( $id === '' ) {
1061                                 # Must have been all whitespace to start with.
1062                                 return '_';
1063                         } else {
1064                                 return $id;
1065                         }
1066                 }
1067
1068                 # HTML4-style escaping
1069                 static $replace = array(
1070                         '%3A' => ':',
1071                         '%' => '.'
1072                 );
1073
1074                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1075                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1076
1077                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1078                 && !in_array( 'noninitial', $options ) )  {
1079                         // Initial character must be a letter!
1080                         $id = "x$id";
1081                 }
1082                 return $id;
1083         }
1084
1085         /**
1086          * Given a value, escape it so that it can be used as a CSS class and
1087          * return it.
1088          *
1089          * @todo For extra validity, input should be validated UTF-8.
1090          *
1091          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1092          *
1093          * @param $class String
1094          * @return String
1095          */
1096         static function escapeClass( $class ) {
1097                 // Convert ugly stuff to underscores and kill underscores in ugly places
1098                 return rtrim(preg_replace(
1099                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
1100                         '_',
1101                         $class ), '_');
1102         }
1103
1104         /**
1105          * Given HTML input, escape with htmlspecialchars but un-escape entites.
1106          * This allows (generally harmless) entities like &#160; to survive.
1107          *
1108          * @param $html String to escape
1109          * @return String: escaped input
1110          */
1111         static function escapeHtmlAllowEntities( $html ) {
1112                 $html = Sanitizer::decodeCharReferences( $html );
1113                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1114                 # hurt.
1115                 $html = htmlspecialchars( $html, ENT_QUOTES );
1116                 return $html;
1117         }
1118
1119         /**
1120          * Regex replace callback for armoring links against further processing.
1121          * @param $matches Array
1122          * @return string
1123          */
1124         private static function armorLinksCallback( $matches ) {
1125                 return str_replace( ':', '&#58;', $matches[1] );
1126         }
1127
1128         /**
1129          * Return an associative array of attribute names and values from
1130          * a partial tag string. Attribute names are forces to lowercase,
1131          * character references are decoded to UTF-8 text.
1132          *
1133          * @param $text String
1134          * @return Array
1135          */
1136         public static function decodeTagAttributes( $text ) {
1137                 if( trim( $text ) == '' ) {
1138                         return array();
1139                 }
1140
1141                 $attribs = array();
1142                 $pairs = array();
1143                 if( !preg_match_all(
1144                         self::getAttribsRegex(),
1145                         $text,
1146                         $pairs,
1147                         PREG_SET_ORDER ) ) {
1148                         return $attribs;
1149                 }
1150
1151                 foreach( $pairs as $set ) {
1152                         $attribute = strtolower( $set[1] );
1153                         $value = Sanitizer::getTagAttributeCallback( $set );
1154
1155                         // Normalize whitespace
1156                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1157                         $value = trim( $value );
1158
1159                         // Decode character references
1160                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1161                 }
1162                 return $attribs;
1163         }
1164
1165         /**
1166          * Pick the appropriate attribute value from a match set from the
1167          * attribs regex matches.
1168          *
1169          * @param $set Array
1170          * @return String
1171          */
1172         private static function getTagAttributeCallback( $set ) {
1173                 if( isset( $set[6] ) ) {
1174                         # Illegal #XXXXXX color with no quotes.
1175                         return $set[6];
1176                 } elseif( isset( $set[5] ) ) {
1177                         # No quotes.
1178                         return $set[5];
1179                 } elseif( isset( $set[4] ) ) {
1180                         # Single-quoted
1181                         return $set[4];
1182                 } elseif( isset( $set[3] ) ) {
1183                         # Double-quoted
1184                         return $set[3];
1185                 } elseif( !isset( $set[2] ) ) {
1186                         # In XHTML, attributes must have a value.
1187                         # For 'reduced' form, return explicitly the attribute name here.
1188                         return $set[1];
1189                 } else {
1190                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1191                 }
1192         }
1193
1194         /**
1195          * Normalize whitespace and character references in an XML source-
1196          * encoded text for an attribute value.
1197          *
1198          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1199          * but note that we're not returning the value, but are returning
1200          * XML source fragments that will be slapped into output.
1201          *
1202          * @param $text String
1203          * @return String
1204          */
1205         private static function normalizeAttributeValue( $text ) {
1206                 return str_replace( '"', '&quot;',
1207                         self::normalizeWhitespace(
1208                                 Sanitizer::normalizeCharReferences( $text ) ) );
1209         }
1210
1211         /**
1212          * @param $text string
1213          * @return mixed
1214          */
1215         private static function normalizeWhitespace( $text ) {
1216                 return preg_replace(
1217                         '/\r\n|[\x20\x0d\x0a\x09]/',
1218                         ' ',
1219                         $text );
1220         }
1221
1222         /**
1223          * Normalizes whitespace in a section name, such as might be returned
1224          * by Parser::stripSectionName(), for use in the id's that are used for
1225          * section links.
1226          *
1227          * @param $section String
1228          * @return String
1229          */
1230         static function normalizeSectionNameWhitespace( $section ) {
1231                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1232         }
1233
1234         /**
1235          * Ensure that any entities and character references are legal
1236          * for XML and XHTML specifically. Any stray bits will be
1237          * &amp;-escaped to result in a valid text fragment.
1238          *
1239          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1240          *   numericized (this way we're well-formed even without a DTD)
1241          * b. any numeric char refs must be legal chars, not invalid or forbidden
1242          * c. use &#x, not &#X
1243          * d. fix or reject non-valid attributes
1244          *
1245          * @param $text String
1246          * @return String
1247          * @private
1248          */
1249         static function normalizeCharReferences( $text ) {
1250                 return preg_replace_callback(
1251                         self::CHAR_REFS_REGEX,
1252                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1253                         $text );
1254         }
1255         /**
1256          * @param $matches String
1257          * @return String
1258          */
1259         static function normalizeCharReferencesCallback( $matches ) {
1260                 $ret = null;
1261                 if( $matches[1] != '' ) {
1262                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1263                 } elseif( $matches[2] != '' ) {
1264                         $ret = Sanitizer::decCharReference( $matches[2] );
1265                 } elseif( $matches[3] != ''  ) {
1266                         $ret = Sanitizer::hexCharReference( $matches[3] );
1267                 }
1268                 if( is_null( $ret ) ) {
1269                         return htmlspecialchars( $matches[0] );
1270                 } else {
1271                         return $ret;
1272                 }
1273         }
1274
1275         /**
1276          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1277          * return the equivalent numeric entity reference (except for the core &lt;
1278          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1279          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1280          * pseudo-entity source (eg &amp;foo;)
1281          *
1282          * @param $name String
1283          * @return String
1284          */
1285         static function normalizeEntity( $name ) {
1286                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1287                         return '&' . self::$htmlEntityAliases[$name] . ';';
1288                 } elseif ( in_array( $name,
1289                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1290                         return "&$name;";
1291                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1292                         return '&#' . self::$htmlEntities[$name] . ';';
1293                 } else {
1294                         return "&amp;$name;";
1295                 }
1296         }
1297
1298         /**
1299          * @param $codepoint
1300          * @return null|string
1301          */
1302         static function decCharReference( $codepoint ) {
1303                 $point = intval( $codepoint );
1304                 if( Sanitizer::validateCodepoint( $point ) ) {
1305                         return sprintf( '&#%d;', $point );
1306                 } else {
1307                         return null;
1308                 }
1309         }
1310
1311         /**
1312          * @param $codepoint
1313          * @return null|string
1314          */
1315         static function hexCharReference( $codepoint ) {
1316                 $point = hexdec( $codepoint );
1317                 if( Sanitizer::validateCodepoint( $point ) ) {
1318                         return sprintf( '&#x%x;', $point );
1319                 } else {
1320                         return null;
1321                 }
1322         }
1323
1324         /**
1325          * Returns true if a given Unicode codepoint is a valid character in XML.
1326          * @param $codepoint Integer
1327          * @return Boolean
1328          */
1329         private static function validateCodepoint( $codepoint ) {
1330                 return ($codepoint ==    0x09)
1331                         || ($codepoint ==    0x0a)
1332                         || ($codepoint ==    0x0d)
1333                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1334                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1335                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1336         }
1337
1338         /**
1339          * Decode any character references, numeric or named entities,
1340          * in the text and return a UTF-8 string.
1341          *
1342          * @param $text String
1343          * @return String
1344          */
1345         public static function decodeCharReferences( $text ) {
1346                 return preg_replace_callback(
1347                         self::CHAR_REFS_REGEX,
1348                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1349                         $text );
1350         }
1351
1352         /**
1353          * Decode any character references, numeric or named entities,
1354          * in the next and normalize the resulting string. (bug 14952)
1355          *
1356          * This is useful for page titles, not for text to be displayed,
1357          * MediaWiki allows HTML entities to escape normalization as a feature.
1358          *
1359          * @param $text String (already normalized, containing entities)
1360          * @return String (still normalized, without entities)
1361          */
1362         public static function decodeCharReferencesAndNormalize( $text ) {
1363                 global $wgContLang;
1364                 $text = preg_replace_callback(
1365                         self::CHAR_REFS_REGEX,
1366                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1367                         $text, /* limit */ -1, $count );
1368
1369                 if ( $count ) {
1370                         return $wgContLang->normalize( $text );
1371                 } else {
1372                         return $text;
1373                 }
1374         }
1375
1376         /**
1377          * @param $matches String
1378          * @return String
1379          */
1380         static function decodeCharReferencesCallback( $matches ) {
1381                 if( $matches[1] != '' ) {
1382                         return Sanitizer::decodeEntity( $matches[1] );
1383                 } elseif( $matches[2] != '' ) {
1384                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1385                 } elseif( $matches[3] != ''  ) {
1386                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1387                 }
1388                 # Last case should be an ampersand by itself
1389                 return $matches[0];
1390         }
1391
1392         /**
1393          * Return UTF-8 string for a codepoint if that is a valid
1394          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1395          * @param $codepoint Integer
1396          * @return String
1397          * @private
1398          */
1399         static function decodeChar( $codepoint ) {
1400                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1401                         return codepointToUtf8( $codepoint );
1402                 } else {
1403                         return UTF8_REPLACEMENT;
1404                 }
1405         }
1406
1407         /**
1408          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1409          * return the UTF-8 encoding of that character. Otherwise, returns
1410          * pseudo-entity source (eg &foo;)
1411          *
1412          * @param $name String
1413          * @return String
1414          */
1415         static function decodeEntity( $name ) {
1416                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1417                         $name = self::$htmlEntityAliases[$name];
1418                 }
1419                 if( isset( self::$htmlEntities[$name] ) ) {
1420                         return codepointToUtf8( self::$htmlEntities[$name] );
1421                 } else {
1422                         return "&$name;";
1423                 }
1424         }
1425
1426         /**
1427          * Fetch the whitelist of acceptable attributes for a given element name.
1428          *
1429          * @param $element String
1430          * @return Array
1431          */
1432         static function attributeWhitelist( $element ) {
1433                 static $list;
1434                 if( !isset( $list ) ) {
1435                         $list = Sanitizer::setupAttributeWhitelist();
1436                 }
1437                 return isset( $list[$element] )
1438                         ? $list[$element]
1439                         : array();
1440         }
1441
1442         /**
1443          * Foreach array key (an allowed HTML element), return an array
1444          * of allowed attributes
1445          * @return Array
1446          */
1447         static function setupAttributeWhitelist() {
1448                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1449
1450                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1451
1452                 if ( $wgAllowRdfaAttributes ) {
1453                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1454                         $common = array_merge( $common, array(
1455                             'about', 'property', 'resource', 'datatype', 'typeof',
1456                         ) );
1457                 }
1458
1459                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1460                         # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1461                         $common = array_merge( $common, array(
1462                             'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1463                         ) );
1464                 }
1465
1466                 $block = array_merge( $common, array( 'align' ) );
1467                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1468                 $tablecell = array( 'abbr',
1469                                     'axis',
1470                                     'headers',
1471                                     'scope',
1472                                     'rowspan',
1473                                     'colspan',
1474                                     'nowrap', # deprecated
1475                                     'width',  # deprecated
1476                                     'height', # deprecated
1477                                     'bgcolor' # deprecated
1478                                     );
1479
1480                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1481                 # See: http://www.w3.org/TR/html4/
1482                 $whitelist = array (
1483                         # 7.5.4
1484                         'div'        => $block,
1485                         'center'     => $common, # deprecated
1486                         'span'       => $block, # ??
1487
1488                         # 7.5.5
1489                         'h1'         => $block,
1490                         'h2'         => $block,
1491                         'h3'         => $block,
1492                         'h4'         => $block,
1493                         'h5'         => $block,
1494                         'h6'         => $block,
1495
1496                         # 7.5.6
1497                         # address
1498
1499                         # 8.2.4
1500                         # bdo
1501
1502                         # 9.2.1
1503                         'em'         => $common,
1504                         'strong'     => $common,
1505                         'cite'       => $common,
1506                         'dfn'        => $common,
1507                         'code'       => $common,
1508                         'samp'       => $common,
1509                         'kbd'        => $common,
1510                         'var'        => $common,
1511                         'abbr'       => $common,
1512                         # acronym
1513
1514                         # 9.2.2
1515                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1516                         # q
1517
1518                         # 9.2.3
1519                         'sub'        => $common,
1520                         'sup'        => $common,
1521
1522                         # 9.3.1
1523                         'p'          => $block,
1524
1525                         # 9.3.2
1526                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1527
1528                         # 9.3.4
1529                         'pre'        => array_merge( $common, array( 'width' ) ),
1530
1531                         # 9.4
1532                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1533                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1534
1535                         # 10.2
1536                         'ul'         => array_merge( $common, array( 'type' ) ),
1537                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1538                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1539
1540                         # 10.3
1541                         'dl'         => $common,
1542                         'dd'         => $common,
1543                         'dt'         => $common,
1544
1545                         # 11.2.1
1546                         'table'      => array_merge( $common,
1547                                                                 array( 'summary', 'width', 'border', 'frame',
1548                                                                                 'rules', 'cellspacing', 'cellpadding',
1549                                                                                 'align', 'bgcolor',
1550                                                                 ) ),
1551
1552                         # 11.2.2
1553                         'caption'    => array_merge( $common, array( 'align' ) ),
1554
1555                         # 11.2.3
1556                         'thead'      => array_merge( $common, $tablealign ),
1557                         'tfoot'      => array_merge( $common, $tablealign ),
1558                         'tbody'      => array_merge( $common, $tablealign ),
1559
1560                         # 11.2.4
1561                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1562                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1563
1564                         # 11.2.5
1565                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1566
1567                         # 11.2.6
1568                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1569                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1570
1571                         # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1572                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1573
1574                         # 13.2
1575                         # Not usually allowed, but may be used for extension-style hooks
1576                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1577                         # true
1578                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1579
1580                         # 15.2.1
1581                         'tt'         => $common,
1582                         'b'          => $common,
1583                         'i'          => $common,
1584                         'big'        => $common,
1585                         'small'      => $common,
1586                         'strike'     => $common,
1587                         's'          => $common,
1588                         'u'          => $common,
1589
1590                         # 15.2.2
1591                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1592                         # basefont
1593
1594                         # 15.3
1595                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1596
1597                         # XHTML Ruby annotation text module, simple ruby only.
1598                         # http://www.w3c.org/TR/ruby/
1599                         'ruby'       => $common,
1600                         # rbc
1601                         # rtc
1602                         'rb'         => $common,
1603                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1604                         'rp'         => $common,
1605
1606                         # MathML root element, where used for extensions
1607                         # 'title' may not be 100% valid here; it's XHTML
1608                         # http://www.w3.org/TR/REC-MathML/
1609                         'math'       => array( 'class', 'style', 'id', 'title' ),
1610                         );
1611                 return $whitelist;
1612         }
1613
1614         /**
1615          * Take a fragment of (potentially invalid) HTML and return
1616          * a version with any tags removed, encoded as plain text.
1617          *
1618          * Warning: this return value must be further escaped for literal
1619          * inclusion in HTML output as of 1.10!
1620          *
1621          * @param $text String: HTML fragment
1622          * @return String
1623          */
1624         static function stripAllTags( $text ) {
1625                 # Actual <tags>
1626                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1627
1628                 # Normalize &entities and whitespace
1629                 $text = self::decodeCharReferences( $text );
1630                 $text = self::normalizeWhitespace( $text );
1631
1632                 return $text;
1633         }
1634
1635         /**
1636          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1637          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1638          * PHP 5.1 doesn't.
1639          *
1640          * Use for passing XHTML fragments to PHP's XML parsing functions
1641          *
1642          * @return String
1643          */
1644         static function hackDocType() {
1645                 $out = "<!DOCTYPE html [\n";
1646                 foreach( self::$htmlEntities as $entity => $codepoint ) {
1647                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1648                 }
1649                 $out .= "]>\n";
1650                 return $out;
1651         }
1652
1653         /**
1654          * @param $url string
1655          * @return mixed|string
1656          */
1657         static function cleanUrl( $url ) {
1658                 # Normalize any HTML entities in input. They will be
1659                 # re-escaped by makeExternalLink().
1660                 $url = Sanitizer::decodeCharReferences( $url );
1661
1662                 # Escape any control characters introduced by the above step
1663                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1664                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1665
1666                 # Validate hostname portion
1667                 $matches = array();
1668                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1669                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1670
1671                         // Characters that will be ignored in IDNs.
1672                         // http://tools.ietf.org/html/3454#section-3.1
1673                         // Strip them before further processing so blacklists and such work.
1674                         $strip = "/
1675                                 \\s|          # general whitespace
1676                                 \xc2\xad|     # 00ad SOFT HYPHEN
1677                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1678                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1679                                 \xe2\x81\xa0| # 2060 WORD JOINER
1680                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1681                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1682                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1683                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1684                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1685                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1686                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1687                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe00f VARIATION SELECTOR-1-16
1688                                 /xuD";
1689
1690                         $host = preg_replace( $strip, '', $host );
1691
1692                         // @todo FIXME: Validate hostnames here
1693
1694                         return $protocol . $host . $rest;
1695                 } else {
1696                         return $url;
1697                 }
1698         }
1699
1700         /**
1701          * @param $matches array
1702          * @return string
1703          */
1704         static function cleanUrlCallback( $matches ) {
1705                 return urlencode( $matches[0] );
1706         }
1707
1708         /**
1709          * Does a string look like an e-mail address?
1710          *
1711          * This validates an email address using an HTML5 specification found at:
1712          * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1713          * Which as of 2011-01-24 says:
1714          *
1715          *   A valid e-mail address is a string that matches the ABNF production
1716          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1717          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1718          *   3.5.
1719          *
1720          * This function is an implementation of the specification as requested in
1721          * bug 22449.
1722          *
1723          * Client-side forms will use the same standard validation rules via JS or
1724          * HTML 5 validation; additional restrictions can be enforced server-side
1725          * by extensions via the 'isValidEmailAddr' hook.
1726          *
1727          * Note that this validation doesn't 100% match RFC 2822, but is believed
1728          * to be liberal enough for wide use. Some invalid addresses will still
1729          * pass validation here.
1730          *
1731          * @since 1.18
1732          *
1733          * @param $addr String E-mail address
1734          * @return Bool
1735          */
1736         public static function validateEmail( $addr ) {
1737                 $result = null;
1738                 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1739                         return $result;
1740                 }
1741
1742                 // Please note strings below are enclosed in brackets [], this make the
1743                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1744                 // See bug 26948
1745                 $rfc5322_atext   = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1746                 $rfc1034_ldh_str = "a-z0-9\\-" ;
1747
1748                 $HTML5_email_regexp = "/
1749                 ^                      # start of string
1750                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1751                 @                      # 'apostrophe'
1752                 [$rfc1034_ldh_str]+       # First domain part
1753                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1754                 $                      # End of string
1755                 /ix" ; // case Insensitive, eXtended
1756
1757                 return (bool) preg_match( $HTML5_email_regexp, $addr );
1758         }
1759 }