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                         global $wgAllowATag;
 411                         if ( $wgAllowATag ) {
 412                                 $htmlpairsStatic[] = 'a';
 413                         }
 414
 415                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 416                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 417
 418                         # Convert them all to hashtables for faster lookup
 419                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 420                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 421                         foreach ( $vars as $var ) {
 422                                 $$var = array_flip( $$var );
 423                         }
 424                         $staticInitialised = true;
 425                 }
 426                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 427                 $extratags = array_flip( $extratags );
 428                 $removetags = array_flip( $removetags );
 429                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 430                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ) , $removetags );
 431
 432                 # Remove HTML comments
 433                 $text = Sanitizer::removeHTMLcomments( $text );
 434                 $bits = explode( '<', $text );
 435                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 436                 if ( !$wgUseTidy ) {
 437                         $tagstack = $tablestack = array();
 438                         foreach ( $bits as $x ) {
 439                                 $regs = array();
 440                                 # $slash: Does the current element start with a '/'?
 441                                 # $t: Current element name
 442                                 # $params: String between element name and >
 443                                 # $brace: Ending '>' or '/>'
 444                                 # $rest: Everything until the next element of $bits
 445                                 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 446                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 447                                 } else {
 448                                         $slash = $t = $params = $brace = $rest = null;
 449                                 }
 450
 451                                 $badtag = false;
 452                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 453                                         # Check our stack
 454                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 455                                                 $badtag = true;
 456                                         } elseif ( $slash ) {
 457                                                 # Closing a tag... is it the one we just opened?
 458                                                 $ot = @array_pop( $tagstack );
 459                                                 if ( $ot != $t ) {
 460                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 461                                                                 # Pop all elements with an optional close tag
 462                                                                 # and see if we find a match below them
 463                                                                 $optstack = array();
 464                                                                 array_push( $optstack, $ot );
 465                                                                 wfSuppressWarnings();
 466                                                                 $ot = array_pop( $tagstack );
 467                                                                 wfRestoreWarnings();
 468                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 469                                                                         array_push( $optstack, $ot );
 470                                                                         wfSuppressWarnings();
 471                                                                         $ot = array_pop( $tagstack );
 472                                                                         wfRestoreWarnings();
 473                                                                 }
 474                                                                 if ( $t != $ot ) {
 475                                                                         # No match. Push the optional elements back again
 476                                                                         $badtag = true;
 477                                                                         wfSuppressWarnings();
 478                                                                         $ot = array_pop( $optstack );
 479                                                                         wfRestoreWarnings();
 480                                                                         while ( $ot ) {
 481                                                                                 array_push( $tagstack, $ot );
 482                                                                                 wfSuppressWarnings();
 483                                                                                 $ot = array_pop( $optstack );
 484                                                                                 wfRestoreWarnings();
 485                                                                         }
 486                                                                 }
 487                                                         } else {
 488                                                                 @array_push( $tagstack, $ot );
 489                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 490                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 491                                                                         $badtag = true;
 492                                                                 }
 493                                                         }
 494                                                 } else {
 495                                                         if ( $t == 'table' ) {
 496                                                                 $tagstack = array_pop( $tablestack );
 497                                                         }
 498                                                 }
 499                                                 $newparams = '';
 500                                         } else {
 501                                                 # Keep track for later
 502                                                 if ( isset( $tabletags[$t] ) &&
 503                                                 !in_array( 'table', $tagstack ) ) {
 504                                                         $badtag = true;
 505                                                 } elseif ( in_array( $t, $tagstack ) &&
 506                                                 !isset( $htmlnest [$t ] ) ) {
 507                                                         $badtag = true;
 508                                                 # Is it a self closed htmlpair ? (bug 5487)
 509                                                 } elseif ( $brace == '/>' &&
 510                                                 isset( $htmlpairs[$t] ) ) {
 511                                                         $badtag = true;
 512                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 513                                                         # Hack to force empty tag for uncloseable elements
 514                                                         $brace = '/>';
 515                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 516                                                         # Hack to not close $htmlsingle tags
 517                                                         $brace = null;
 518                                                 } elseif ( isset( $tabletags[$t] )
 519                                                 && in_array( $t, $tagstack ) ) {
 520                                                         // New table tag but forgot to close the previous one
 521                                                         $text .= "</$t>";
 522                                                 } else {
 523                                                         if ( $t == 'table' ) {
 524                                                                 array_push( $tablestack, $tagstack );
 525                                                                 $tagstack = array();
 526                                                         }
 527                                                         array_push( $tagstack, $t );
 528                                                 }
 529
 530                                                 # Replace any variables or template parameters with
 531                                                 # plaintext results.
 532                                                 if( is_callable( $processCallback ) ) {
 533                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 534                                                 }
 535
 536                                                 # Strip non-approved attributes from the tag
 537                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 538                                         }
 539                                         if ( !$badtag ) {
 540                                                 $rest = str_replace( '>', '&gt;', $rest );
 541                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 542                                                 $text .= "<$slash$t$newparams$close>$rest";
 543                                                 continue;
 544                                         }
 545                                 }
 546                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 547                         }
 548                         # Close off any remaining tags
 549                         while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
 550                                 $text .= "</$t>\n";
 551                                 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
 552                         }
 553                 } else {
 554                         # this might be possible using tidy itself
 555                         foreach ( $bits as $x ) {
 556                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 557                                 $x, $regs );
 558                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 559                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 560                                         if( is_callable( $processCallback ) ) {
 561                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 562                                         }
 563                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 564                                         $rest = str_replace( '>', '&gt;', $rest );
 565                                         $text .= "<$slash$t$newparams$brace$rest";
 566                                 } else {
 567                                         $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 568                                 }
 569                         }
 570                 }
 571                 wfProfileOut( __METHOD__ );
 572                 return $text;
 573         }
 574
 575         /**
 576          * Remove '<!--', '-->', and everything between.
 577          * To avoid leaving blank lines, when a comment is both preceded
 578          * and followed by a newline (ignoring spaces), trim leading and
 579          * trailing spaces and one of the newlines.
 580          *
 581          * @private
 582          * @param $text String
 583          * @return string
 584          */
 585         static function removeHTMLcomments( $text ) {
 586                 wfProfileIn( __METHOD__ );
 587                 while (($start = strpos($text, '<!--')) !== false) {
 588                         $end = strpos($text, '-->', $start + 4);
 589                         if ($end === false) {
 590                                 # Unterminated comment; bail out
 591                                 break;
 592                         }
 593
 594                         $end += 3;
 595
 596                         # Trim space and newline if the comment is both
 597                         # preceded and followed by a newline
 598                         $spaceStart = max($start - 1, 0);
 599                         $spaceLen = $end - $spaceStart;
 600                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 601                                 $spaceStart--;
 602                                 $spaceLen++;
 603                         }
 604                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 605                                 $spaceLen++;
 606                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 607                                 # Remove the comment, leading and trailing
 608                                 # spaces, and leave only one newline.
 609                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 610                         }
 611                         else {
 612                                 # Remove just the comment.
 613                                 $text = substr_replace($text, '', $start, $end - $start);
 614                         }
 615                 }
 616                 wfProfileOut( __METHOD__ );
 617                 return $text;
 618         }
 619
 620         /**
 621          * Take an array of attribute names and values and fix some deprecated values
 622          * for the given element type.
 623          * This does not validate properties, so you should ensure that you call
 624          * validateTagAttributes AFTER this to ensure that the resulting style rule
 625          * this may add is safe.
 626          *
 627          * - Converts most presentational attributes like align into inline css
 628          *
 629          * @param $attribs Array
 630          * @param $element String
 631          * @return Array
 632          */
 633         static function fixDeprecatedAttributes( $attribs, $element ) {
 634                 global $wgHtml5, $wgCleanupPresentationalAttributes;
 635
 636                 // presentational attributes were removed from html5, we can leave them
 637                 // in when html5 is turned off
 638                 if ( !$wgHtml5 || !$wgCleanupPresentationalAttributes ) {
 639                         return $attribs;
 640                 }
 641
 642                 $table = array( 'table' );
 643                 $cells = array( 'td', 'th' );
 644                 $colls = array( 'col', 'colgroup' );
 645                 $tblocks = array( 'tbody', 'tfoot', 'thead' );
 646                 $h = array( 'h1', 'h2', 'h3', 'h4', 'h5', 'h6' );
 647
 648                 $presentationalAttribs = array(
 649                         'align' => array( 'text-align', array_merge( array( 'caption', 'hr', 'div', 'p', 'tr' ), $table, $cells, $colls, $tblocks, $h ) ),
 650                         'clear' => array( 'clear', array( 'br' ) ),
 651                         'height' => array( 'height', $cells ),
 652                         'nowrap' => array( 'white-space', $cells ),
 653                         'size' => array( 'height', array( 'hr' ) ),
 654                         'type' => array( 'list-style-type', array( 'li', 'ol', 'ul' ) ),
 655                         'valign' => array( 'vertical-align', array_merge( $cells, $colls, $tblocks ) ),
 656                         'width' => array( 'width', array_merge( array( 'hr', 'pre' ), $table, $cells, $colls ) ),
 657                 );
 658
 659                 // Ensure that any upper case or mixed case attributes are converted to lowercase
 660                 foreach ( $attribs as $attribute => $value ) {
 661                         if ( $attribute !== strtolower( $attribute ) && array_key_exists( strtolower( $attribute ), $presentationalAttribs ) ) {
 662                                 $attribs[strtolower( $attribute )] = $value;
 663                                 unset( $attribs[$attribute] );
 664                         }
 665                 }
 666
 667                 $style = "";
 668                 foreach ( $presentationalAttribs as $attribute => $info ) {
 669                         list( $property, $elements ) = $info;
 670
 671                         // Skip if this attribute is not relevant to this element
 672                         if ( !in_array( $element, $elements ) ) {
 673                                 continue;
 674                         }
 675
 676                         // Skip if the attribute is not used
 677                         if ( !array_key_exists( $attribute, $attribs ) ) {
 678                                 continue;
 679                         }
 680
 681                         $value = $attribs[$attribute];
 682
 683                         // For nowrap the value should be nowrap instead of whatever text is in the value
 684                         if ( $attribute === 'nowrap' ) {
 685                                 $value = 'nowrap';
 686                         }
 687
 688                         // clear="all" is clear: both; in css
 689                         if ( $attribute === 'clear' && strtolower( $value ) === 'all' ) {
 690                                 $value = 'both';
 691                         }
 692
 693                         // Size based properties should have px applied to them if they have no unit
 694                         if ( in_array( $attribute, array( 'height', 'width', 'size' ) ) ) {
 695                                 if ( preg_match( '/^[\d.]+$/', $value ) ) {
 696                                         $value = "{$value}px";
 697                                 }
 698                         }
 699
 700                         $style .= " $property: $value;";
 701
 702                         unset( $attribs[$attribute] );
 703                 }
 704
 705                 if ( $style ) {
 706                         // Prepend our style rules so that they can be overridden by user css
 707                         if ( isset($attribs['style']) ) {
 708                                 $style .= " " . $attribs['style'];
 709                         }
 710                         $attribs['style'] = trim($style);
 711                 }
 712
 713                 return $attribs;
 714         }
 715
 716         /**
 717          * Take an array of attribute names and values and normalize or discard
 718          * illegal values for the given element type.
 719          *
 720          * - Discards attributes not on a whitelist for the given element
 721          * - Unsafe style attributes are discarded
 722          * - Invalid id attributes are reencoded
 723          *
 724          * @param $attribs Array
 725          * @param $element String
 726          * @return Array
 727          *
 728          * @todo Check for legal values where the DTD limits things.
 729          * @todo Check for unique id attribute :P
 730          */
 731         static function validateTagAttributes( $attribs, $element ) {
 732                 return Sanitizer::validateAttributes( $attribs,
 733                         Sanitizer::attributeWhitelist( $element ) );
 734         }
 735
 736         /**
 737          * Take an array of attribute names and values and normalize or discard
 738          * illegal values for the given whitelist.
 739          *
 740          * - Discards attributes not the given whitelist
 741          * - Unsafe style attributes are discarded
 742          * - Invalid id attributes are reencoded
 743          *
 744          * @param $attribs Array
 745          * @param $whitelist Array: list of allowed attribute names
 746          * @return Array
 747          *
 748          * @todo Check for legal values where the DTD limits things.
 749          * @todo Check for unique id attribute :P
 750          */
 751         static function validateAttributes( $attribs, $whitelist ) {
 752                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
 753
 754                 $whitelist = array_flip( $whitelist );
 755                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 756
 757                 $out = array();
 758                 foreach( $attribs as $attribute => $value ) {
 759                         #allow XML namespace declaration if RDFa is enabled
 760                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 761                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 762                                         $out[$attribute] = $value;
 763                                 }
 764
 765                                 continue;
 766                         }
 767
 768                         # Allow any attribute beginning with "data-", if in HTML5 mode
 769                         if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
 770                                 continue;
 771                         }
 772
 773                         # Strip javascript "expression" from stylesheets.
 774                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 775                         if( $attribute == 'style' ) {
 776                                 $value = Sanitizer::checkCss( $value );
 777                         }
 778
 779                         if ( $attribute === 'id' ) {
 780                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 781                         }
 782
 783                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 784                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 785                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 786                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 787                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 788                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 789
 790                                 //Paranoia. Allow "simple" values but suppress javascript
 791                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 792                                         continue;
 793                                 }
 794                         }
 795
 796                         # NOTE: even though elements using href/src are not allowed directly, supply
 797                         #       validation code that can be used by tag hook handlers, etc
 798                         if ( $attribute === 'href' || $attribute === 'src' ) {
 799                                 if ( !preg_match( $hrefExp, $value ) ) {
 800                                         continue; //drop any href or src attributes not using an allowed protocol.
 801                                                   //NOTE: this also drops all relative URLs
 802                                 }
 803                         }
 804
 805                         if ( $attribute === 'href' || $attribute === 'src' ) {
 806                                 if ( !preg_match( $hrefExp, $value ) ) {
 807                                         continue; //drop any href or src attributes not using an allowed protocol.
 808                                                   //NOTE: this also drops all relative URLs
 809                                 }
 810                         }
 811
 812                         //RDFa properties allow URIs. check them
 813                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 814                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' ||
 815                                 $attribute === 'datatype' || $attribute === 'typeof' ) {
 816                                 //Paranoia. Allow "simple" values but suppress javascript
 817                                 if ( preg_match( '/(^|\s)javascript\s*:/i', $value ) ) {
 818                                         continue;
 819                                 }
 820                         }
 821
 822                         // If this attribute was previously set, override it.
 823                         // Output should only have one attribute of each name.
 824                         $out[$attribute] = $value;
 825                 }
 826
 827                 if ( $wgAllowMicrodataAttributes ) {
 828                         # itemtype, itemid, itemref don't make sense without itemscope
 829                         if ( !array_key_exists( 'itemscope', $out ) ) {
 830                                 unset( $out['itemtype'] );
 831                                 unset( $out['itemid'] );
 832                                 unset( $out['itemref'] );
 833                         }
 834                         # TODO: Strip itemprop if we aren't descendants of an itemscope.
 835                 }
 836                 return $out;
 837         }
 838
 839         /**
 840          * Merge two sets of HTML attributes.  Conflicting items in the second set
 841          * will override those in the first, except for 'class' attributes which
 842          * will be combined (if they're both strings).
 843          *
 844          * @todo implement merging for other attributes such as style
 845          * @param $a Array
 846          * @param $b Array
 847          * @return array
 848          */
 849         static function mergeAttributes( $a, $b ) {
 850                 $out = array_merge( $a, $b );
 851                 if( isset( $a['class'] ) && isset( $b['class'] )
 852                 && is_string( $a['class'] ) && is_string( $b['class'] )
 853                 && $a['class'] !== $b['class'] ) {
 854                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 855                                 -1, PREG_SPLIT_NO_EMPTY );
 856                         $out['class'] = implode( ' ', array_unique( $classes ) );
 857                 }
 858                 return $out;
 859         }
 860
 861         /**
 862          * Pick apart some CSS and check it for forbidden or unsafe structures.
 863          * Returns a sanitized string. This sanitized string will have
 864          * character references and escape sequences decoded, and comments
 865          * stripped. If the input is just too evil, only a comment complaining
 866          * about evilness will be returned.
 867          *
 868          * Currently URL references, 'expression', 'tps' are forbidden.
 869          *
 870          * NOTE: Despite the fact that character references are decoded, the
 871          * returned string may contain character references given certain
 872          * clever input strings. These character references must
 873          * be escaped before the return value is embedded in HTML.
 874          *
 875          * @param $value String
 876          * @return String
 877          */
 878         static function checkCss( $value ) {
 879                 // Decode character references like &#123;
 880                 $value = Sanitizer::decodeCharReferences( $value );
 881
 882                 // Decode escape sequences and line continuation
 883                 // See the grammar in the CSS 2 spec, appendix D.
 884                 // This has to be done AFTER decoding character references.
 885                 // This means it isn't possible for this function to return
 886                 // unsanitized escape sequences. It is possible to manufacture
 887                 // input that contains character references that decode to
 888                 // escape sequences that decode to character references, but
 889                 // it's OK for the return value to contain character references
 890                 // because the caller is supposed to escape those anyway.
 891                 static $decodeRegex;
 892                 if ( !$decodeRegex ) {
 893                         $space = '[\\x20\\t\\r\\n\\f]';
 894                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 895                         $backslash = '\\\\';
 896                         $decodeRegex = "/ $backslash
 897                                 (?:
 898                                         ($nl) |  # 1. Line continuation
 899                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 900                                         (.) | # 3. backslash cancelling special meaning
 901                                         () | # 4. backslash at end of string
 902                                 )/xu";
 903                 }
 904                 $value = preg_replace_callback( $decodeRegex,
 905                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 906
 907                 // Remove any comments; IE gets token splitting wrong
 908                 // This must be done AFTER decoding character references and
 909                 // escape sequences, because those steps can introduce comments
 910                 // This step cannot introduce character references or escape
 911                 // sequences, because it replaces comments with spaces rather
 912                 // than removing them completely.
 913                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 914
 915                 // Remove anything after a comment-start token, to guard against
 916                 // incorrect client implementations.
 917                 $commentPos = strpos( $value, '/*' );
 918                 if ( $commentPos !== false ) {
 919                         $value = substr( $value, 0, $commentPos );
 920                 }
 921
 922                 // Reject problematic keywords and control characters
 923                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 924                         return '/* invalid control char */';
 925                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
 926                         return '/* insecure input */';
 927                 }
 928                 return $value;
 929         }
 930
 931         /**
 932          * @param $matches array
 933          * @return String
 934          */
 935         static function cssDecodeCallback( $matches ) {
 936                 if ( $matches[1] !== '' ) {
 937                         // Line continuation
 938                         return '';
 939                 } elseif ( $matches[2] !== '' ) {
 940                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 941                 } elseif ( $matches[3] !== '' ) {
 942                         $char = $matches[3];
 943                 } else {
 944                         $char = '\\';
 945                 }
 946                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 947                         // These characters need to be escaped in strings
 948                         // Clean up the escape sequence to avoid parsing errors by clients
 949                         return '\\' . dechex( ord( $char ) ) . ' ';
 950                 } else {
 951                         // Decode unnecessary escape
 952                         return $char;
 953                 }
 954         }
 955
 956         /**
 957          * Take a tag soup fragment listing an HTML element's attributes
 958          * and normalize it to well-formed XML, discarding unwanted attributes.
 959          * Output is safe for further wikitext processing, with escaping of
 960          * values that could trigger problems.
 961          *
 962          * - Normalizes attribute names to lowercase
 963          * - Discards attributes not on a whitelist for the given element
 964          * - Turns broken or invalid entities into plaintext
 965          * - Double-quotes all attribute values
 966          * - Attributes without values are given the name as attribute
 967          * - Double attributes are discarded
 968          * - Unsafe style attributes are discarded
 969          * - Prepends space if there are attributes.
 970          *
 971          * @param $text String
 972          * @param $element String
 973          * @return String
 974          */
 975         static function fixTagAttributes( $text, $element ) {
 976                 if( trim( $text ) == '' ) {
 977                         return '';
 978                 }
 979
 980                 $decoded = Sanitizer::decodeTagAttributes( $text );
 981                 $decoded = Sanitizer::fixDeprecatedAttributes( $decoded, $element );
 982                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
 983
 984                 $attribs = array();
 985                 foreach( $stripped as $attribute => $value ) {
 986                         $encAttribute = htmlspecialchars( $attribute );
 987                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 988
 989                         $attribs[] = "$encAttribute=\"$encValue\"";
 990                 }
 991                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 992         }
 993
 994         /**
 995          * Encode an attribute value for HTML output.
 996          * @param $text String
 997          * @return HTML-encoded text fragment
 998          */
 999         static function encodeAttribute( $text ) {
1000                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1001
1002                 // Whitespace is normalized during attribute decoding,
1003                 // so if we've been passed non-spaces we must encode them
1004                 // ahead of time or they won't be preserved.
1005                 $encValue = strtr( $encValue, array(
1006                         "\n" => '&#10;',
1007                         "\r" => '&#13;',
1008                         "\t" => '&#9;',
1009                 ) );
1010
1011                 return $encValue;
1012         }
1013
1014         /**
1015          * Encode an attribute value for HTML tags, with extra armoring
1016          * against further wiki processing.
1017          * @param $text String
1018          * @return HTML-encoded text fragment
1019          */
1020         static function safeEncodeAttribute( $text ) {
1021                 $encValue = Sanitizer::encodeAttribute( $text );
1022
1023                 # Templates and links may be expanded in later parsing,
1024                 # creating invalid or dangerous output. Suppress this.
1025                 $encValue = strtr( $encValue, array(
1026                         '<'    => '&lt;',   // This should never happen,
1027                         '>'    => '&gt;',   // we've received invalid input
1028                         '"'    => '&quot;', // which should have been escaped.
1029                         '{'    => '&#123;',
1030                         '['    => '&#91;',
1031                         "''"   => '&#39;&#39;',
1032                         'ISBN' => '&#73;SBN',
1033                         'RFC'  => '&#82;FC',
1034                         'PMID' => '&#80;MID',
1035                         '|'    => '&#124;',
1036                         '__'   => '&#95;_',
1037                 ) );
1038
1039                 # Stupid hack
1040                 $encValue = preg_replace_callback(
1041                         '/(' . wfUrlProtocols() . ')/',
1042                         array( 'Sanitizer', 'armorLinksCallback' ),
1043                         $encValue );
1044                 return $encValue;
1045         }
1046
1047         /**
1048          * Given a value, escape it so that it can be used in an id attribute and
1049          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1050          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1051          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1052          * escaped with lots of dots.
1053          *
1054          * To ensure we don't have to bother escaping anything, we also strip ', ",
1055          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1056          * We also strip # because it upsets IE, and % because it could be
1057          * ambiguous if it's part of something that looks like a percent escape
1058          * (which don't work reliably in fragments cross-browser).
1059          *
1060          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1061          *                                                          in the id and
1062          *                                                          name attributes
1063          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1064          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1065          *   HTML5 definition of id attribute
1066          *
1067          * @param $id String: id to escape
1068          * @param $options Mixed: string or array of strings (default is array()):
1069          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1070          *       so don't pay attention if the first character isn't valid at the
1071          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1072          *       false.
1073          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1074          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1075          *       anchors and links won't break.
1076          * @return String
1077          */
1078         static function escapeId( $id, $options = array() ) {
1079                 global $wgHtml5, $wgExperimentalHtmlIds;
1080                 $options = (array)$options;
1081
1082                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1083                         $id = Sanitizer::decodeCharReferences( $id );
1084                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1085                         $id = trim( $id, '_' );
1086                         if ( $id === '' ) {
1087                                 # Must have been all whitespace to start with.
1088                                 return '_';
1089                         } else {
1090                                 return $id;
1091                         }
1092                 }
1093
1094                 # HTML4-style escaping
1095                 static $replace = array(
1096                         '%3A' => ':',
1097                         '%' => '.'
1098                 );
1099
1100                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1101                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1102
1103                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1104                 && !in_array( 'noninitial', $options ) )  {
1105                         // Initial character must be a letter!
1106                         $id = "x$id";
1107                 }
1108                 return $id;
1109         }
1110
1111         /**
1112          * Given a value, escape it so that it can be used as a CSS class and
1113          * return it.
1114          *
1115          * @todo For extra validity, input should be validated UTF-8.
1116          *
1117          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1118          *
1119          * @param $class String
1120          * @return String
1121          */
1122         static function escapeClass( $class ) {
1123                 // Convert ugly stuff to underscores and kill underscores in ugly places
1124                 return rtrim(preg_replace(
1125                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
1126                         '_',
1127                         $class ), '_');
1128         }
1129
1130         /**
1131          * Given HTML input, escape with htmlspecialchars but un-escape entites.
1132          * This allows (generally harmless) entities like &#160; to survive.
1133          *
1134          * @param $html String to escape
1135          * @return String: escaped input
1136          */
1137         static function escapeHtmlAllowEntities( $html ) {
1138                 $html = Sanitizer::decodeCharReferences( $html );
1139                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1140                 # hurt.
1141                 $html = htmlspecialchars( $html, ENT_QUOTES );
1142                 return $html;
1143         }
1144
1145         /**
1146          * Regex replace callback for armoring links against further processing.
1147          * @param $matches Array
1148          * @return string
1149          */
1150         private static function armorLinksCallback( $matches ) {
1151                 return str_replace( ':', '&#58;', $matches[1] );
1152         }
1153
1154         /**
1155          * Return an associative array of attribute names and values from
1156          * a partial tag string. Attribute names are forces to lowercase,
1157          * character references are decoded to UTF-8 text.
1158          *
1159          * @param $text String
1160          * @return Array
1161          */
1162         public static function decodeTagAttributes( $text ) {
1163                 if( trim( $text ) == '' ) {
1164                         return array();
1165                 }
1166
1167                 $attribs = array();
1168                 $pairs = array();
1169                 if( !preg_match_all(
1170                         self::getAttribsRegex(),
1171                         $text,
1172                         $pairs,
1173                         PREG_SET_ORDER ) ) {
1174                         return $attribs;
1175                 }
1176
1177                 foreach( $pairs as $set ) {
1178                         $attribute = strtolower( $set[1] );
1179                         $value = Sanitizer::getTagAttributeCallback( $set );
1180
1181                         // Normalize whitespace
1182                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1183                         $value = trim( $value );
1184
1185                         // Decode character references
1186                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1187                 }
1188                 return $attribs;
1189         }
1190
1191         /**
1192          * Pick the appropriate attribute value from a match set from the
1193          * attribs regex matches.
1194          *
1195          * @param $set Array
1196          * @return String
1197          */
1198         private static function getTagAttributeCallback( $set ) {
1199                 if( isset( $set[6] ) ) {
1200                         # Illegal #XXXXXX color with no quotes.
1201                         return $set[6];
1202                 } elseif( isset( $set[5] ) ) {
1203                         # No quotes.
1204                         return $set[5];
1205                 } elseif( isset( $set[4] ) ) {
1206                         # Single-quoted
1207                         return $set[4];
1208                 } elseif( isset( $set[3] ) ) {
1209                         # Double-quoted
1210                         return $set[3];
1211                 } elseif( !isset( $set[2] ) ) {
1212                         # In XHTML, attributes must have a value.
1213                         # For 'reduced' form, return explicitly the attribute name here.
1214                         return $set[1];
1215                 } else {
1216                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1217                 }
1218         }
1219
1220         /**
1221          * Normalize whitespace and character references in an XML source-
1222          * encoded text for an attribute value.
1223          *
1224          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1225          * but note that we're not returning the value, but are returning
1226          * XML source fragments that will be slapped into output.
1227          *
1228          * @param $text String
1229          * @return String
1230          */
1231         private static function normalizeAttributeValue( $text ) {
1232                 return str_replace( '"', '&quot;',
1233                         self::normalizeWhitespace(
1234                                 Sanitizer::normalizeCharReferences( $text ) ) );
1235         }
1236
1237         /**
1238          * @param $text string
1239          * @return mixed
1240          */
1241         private static function normalizeWhitespace( $text ) {
1242                 return preg_replace(
1243                         '/\r\n|[\x20\x0d\x0a\x09]/',
1244                         ' ',
1245                         $text );
1246         }
1247
1248         /**
1249          * Normalizes whitespace in a section name, such as might be returned
1250          * by Parser::stripSectionName(), for use in the id's that are used for
1251          * section links.
1252          *
1253          * @param $section String
1254          * @return String
1255          */
1256         static function normalizeSectionNameWhitespace( $section ) {
1257                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1258         }
1259
1260         /**
1261          * Ensure that any entities and character references are legal
1262          * for XML and XHTML specifically. Any stray bits will be
1263          * &amp;-escaped to result in a valid text fragment.
1264          *
1265          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1266          *   numericized (this way we're well-formed even without a DTD)
1267          * b. any numeric char refs must be legal chars, not invalid or forbidden
1268          * c. use &#x, not &#X
1269          * d. fix or reject non-valid attributes
1270          *
1271          * @param $text String
1272          * @return String
1273          * @private
1274          */
1275         static function normalizeCharReferences( $text ) {
1276                 return preg_replace_callback(
1277                         self::CHAR_REFS_REGEX,
1278                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1279                         $text );
1280         }
1281         /**
1282          * @param $matches String
1283          * @return String
1284          */
1285         static function normalizeCharReferencesCallback( $matches ) {
1286                 $ret = null;
1287                 if( $matches[1] != '' ) {
1288                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1289                 } elseif( $matches[2] != '' ) {
1290                         $ret = Sanitizer::decCharReference( $matches[2] );
1291                 } elseif( $matches[3] != ''  ) {
1292                         $ret = Sanitizer::hexCharReference( $matches[3] );
1293                 }
1294                 if( is_null( $ret ) ) {
1295                         return htmlspecialchars( $matches[0] );
1296                 } else {
1297                         return $ret;
1298                 }
1299         }
1300
1301         /**
1302          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1303          * return the equivalent numeric entity reference (except for the core &lt;
1304          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1305          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1306          * pseudo-entity source (eg &amp;foo;)
1307          *
1308          * @param $name String
1309          * @return String
1310          */
1311         static function normalizeEntity( $name ) {
1312                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1313                         return '&' . self::$htmlEntityAliases[$name] . ';';
1314                 } elseif ( in_array( $name,
1315                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1316                         return "&$name;";
1317                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1318                         return '&#' . self::$htmlEntities[$name] . ';';
1319                 } else {
1320                         return "&amp;$name;";
1321                 }
1322         }
1323
1324         /**
1325          * @param $codepoint
1326          * @return null|string
1327          */
1328         static function decCharReference( $codepoint ) {
1329                 $point = intval( $codepoint );
1330                 if( Sanitizer::validateCodepoint( $point ) ) {
1331                         return sprintf( '&#%d;', $point );
1332                 } else {
1333                         return null;
1334                 }
1335         }
1336
1337         /**
1338          * @param $codepoint
1339          * @return null|string
1340          */
1341         static function hexCharReference( $codepoint ) {
1342                 $point = hexdec( $codepoint );
1343                 if( Sanitizer::validateCodepoint( $point ) ) {
1344                         return sprintf( '&#x%x;', $point );
1345                 } else {
1346                         return null;
1347                 }
1348         }
1349
1350         /**
1351          * Returns true if a given Unicode codepoint is a valid character in XML.
1352          * @param $codepoint Integer
1353          * @return Boolean
1354          */
1355         private static function validateCodepoint( $codepoint ) {
1356                 return ($codepoint ==    0x09)
1357                         || ($codepoint ==    0x0a)
1358                         || ($codepoint ==    0x0d)
1359                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1360                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1361                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1362         }
1363
1364         /**
1365          * Decode any character references, numeric or named entities,
1366          * in the text and return a UTF-8 string.
1367          *
1368          * @param $text String
1369          * @return String
1370          */
1371         public static function decodeCharReferences( $text ) {
1372                 return preg_replace_callback(
1373                         self::CHAR_REFS_REGEX,
1374                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1375                         $text );
1376         }
1377
1378         /**
1379          * Decode any character references, numeric or named entities,
1380          * in the next and normalize the resulting string. (bug 14952)
1381          *
1382          * This is useful for page titles, not for text to be displayed,
1383          * MediaWiki allows HTML entities to escape normalization as a feature.
1384          *
1385          * @param $text String (already normalized, containing entities)
1386          * @return String (still normalized, without entities)
1387          */
1388         public static function decodeCharReferencesAndNormalize( $text ) {
1389                 global $wgContLang;
1390                 $text = preg_replace_callback(
1391                         self::CHAR_REFS_REGEX,
1392                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1393                         $text, /* limit */ -1, $count );
1394
1395                 if ( $count ) {
1396                         return $wgContLang->normalize( $text );
1397                 } else {
1398                         return $text;
1399                 }
1400         }
1401
1402         /**
1403          * @param $matches String
1404          * @return String
1405          */
1406         static function decodeCharReferencesCallback( $matches ) {
1407                 if( $matches[1] != '' ) {
1408                         return Sanitizer::decodeEntity( $matches[1] );
1409                 } elseif( $matches[2] != '' ) {
1410                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1411                 } elseif( $matches[3] != ''  ) {
1412                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1413                 }
1414                 # Last case should be an ampersand by itself
1415                 return $matches[0];
1416         }
1417
1418         /**
1419          * Return UTF-8 string for a codepoint if that is a valid
1420          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1421          * @param $codepoint Integer
1422          * @return String
1423          * @private
1424          */
1425         static function decodeChar( $codepoint ) {
1426                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1427                         return codepointToUtf8( $codepoint );
1428                 } else {
1429                         return UTF8_REPLACEMENT;
1430                 }
1431         }
1432
1433         /**
1434          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1435          * return the UTF-8 encoding of that character. Otherwise, returns
1436          * pseudo-entity source (eg &foo;)
1437          *
1438          * @param $name String
1439          * @return String
1440          */
1441         static function decodeEntity( $name ) {
1442                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1443                         $name = self::$htmlEntityAliases[$name];
1444                 }
1445                 if( isset( self::$htmlEntities[$name] ) ) {
1446                         return codepointToUtf8( self::$htmlEntities[$name] );
1447                 } else {
1448                         return "&$name;";
1449                 }
1450         }
1451
1452         /**
1453          * Fetch the whitelist of acceptable attributes for a given element name.
1454          *
1455          * @param $element String
1456          * @return Array
1457          */
1458         static function attributeWhitelist( $element ) {
1459                 static $list;
1460                 if( !isset( $list ) ) {
1461                         $list = Sanitizer::setupAttributeWhitelist();
1462                 }
1463                 return isset( $list[$element] )
1464                         ? $list[$element]
1465                         : array();
1466         }
1467
1468         /**
1469          * Foreach array key (an allowed HTML element), return an array
1470          * of allowed attributes
1471          * @return Array
1472          */
1473         static function setupAttributeWhitelist() {
1474                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1475
1476                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1477
1478                 if ( $wgAllowRdfaAttributes ) {
1479                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1480                         $common = array_merge( $common, array(
1481                             'about', 'property', 'resource', 'datatype', 'typeof',
1482                         ) );
1483                 }
1484
1485                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1486                         # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1487                         $common = array_merge( $common, array(
1488                             'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1489                         ) );
1490                 }
1491
1492                 $block = array_merge( $common, array( 'align' ) );
1493                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1494                 $tablecell = array( 'abbr',
1495                                     'axis',
1496                                     'headers',
1497                                     'scope',
1498                                     'rowspan',
1499                                     'colspan',
1500                                     'nowrap', # deprecated
1501                                     'width',  # deprecated
1502                                     'height', # deprecated
1503                                     'bgcolor' # deprecated
1504                                     );
1505
1506                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1507                 # See: http://www.w3.org/TR/html4/
1508                 $whitelist = array (
1509                         # 7.5.4
1510                         'div'        => $block,
1511                         'center'     => $common, # deprecated
1512                         'span'       => $block, # ??
1513
1514                         # 7.5.5
1515                         'h1'         => $block,
1516                         'h2'         => $block,
1517                         'h3'         => $block,
1518                         'h4'         => $block,
1519                         'h5'         => $block,
1520                         'h6'         => $block,
1521
1522                         # 7.5.6
1523                         # address
1524
1525                         # 8.2.4
1526                         # bdo
1527
1528                         # 9.2.1
1529                         'em'         => $common,
1530                         'strong'     => $common,
1531                         'cite'       => $common,
1532                         'dfn'        => $common,
1533                         'code'       => $common,
1534                         'samp'       => $common,
1535                         'kbd'        => $common,
1536                         'var'        => $common,
1537                         'abbr'       => $common,
1538                         # acronym
1539
1540                         # 9.2.2
1541                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1542                         # q
1543
1544                         # 9.2.3
1545                         'sub'        => $common,
1546                         'sup'        => $common,
1547
1548                         # 9.3.1
1549                         'p'          => $block,
1550
1551                         # 9.3.2
1552                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1553
1554                         # 9.3.4
1555                         'pre'        => array_merge( $common, array( 'width' ) ),
1556
1557                         # 9.4
1558                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1559                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1560
1561                         # 10.2
1562                         'ul'         => array_merge( $common, array( 'type' ) ),
1563                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1564                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1565
1566                         # 10.3
1567                         'dl'         => $common,
1568                         'dd'         => $common,
1569                         'dt'         => $common,
1570
1571                         # 11.2.1
1572                         'table'      => array_merge( $common,
1573                                                                 array( 'summary', 'width', 'border', 'frame',
1574                                                                                 'rules', 'cellspacing', 'cellpadding',
1575                                                                                 'align', 'bgcolor',
1576                                                                 ) ),
1577
1578                         # 11.2.2
1579                         'caption'    => array_merge( $common, array( 'align' ) ),
1580
1581                         # 11.2.3
1582                         'thead'      => array_merge( $common, $tablealign ),
1583                         'tfoot'      => array_merge( $common, $tablealign ),
1584                         'tbody'      => array_merge( $common, $tablealign ),
1585
1586                         # 11.2.4
1587                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1588                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1589
1590                         # 11.2.5
1591                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1592
1593                         # 11.2.6
1594                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1595                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1596
1597                         # 12.2
1598                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1599
1600                         # 13.2
1601                         # Not usually allowed, but may be used for extension-style hooks
1602                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1603                         # true
1604                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1605
1606                         # 15.2.1
1607                         'tt'         => $common,
1608                         'b'          => $common,
1609                         'i'          => $common,
1610                         'big'        => $common,
1611                         'small'      => $common,
1612                         'strike'     => $common,
1613                         's'          => $common,
1614                         'u'          => $common,
1615
1616                         # 15.2.2
1617                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1618                         # basefont
1619
1620                         # 15.3
1621                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1622
1623                         # XHTML Ruby annotation text module, simple ruby only.
1624                         # http://www.w3c.org/TR/ruby/
1625                         'ruby'       => $common,
1626                         # rbc
1627                         # rtc
1628                         'rb'         => $common,
1629                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1630                         'rp'         => $common,
1631
1632                         # MathML root element, where used for extensions
1633                         # 'title' may not be 100% valid here; it's XHTML
1634                         # http://www.w3.org/TR/REC-MathML/
1635                         'math'       => array( 'class', 'style', 'id', 'title' ),
1636                         );
1637                 return $whitelist;
1638         }
1639
1640         /**
1641          * Take a fragment of (potentially invalid) HTML and return
1642          * a version with any tags removed, encoded as plain text.
1643          *
1644          * Warning: this return value must be further escaped for literal
1645          * inclusion in HTML output as of 1.10!
1646          *
1647          * @param $text String: HTML fragment
1648          * @return String
1649          */
1650         static function stripAllTags( $text ) {
1651                 # Actual <tags>
1652                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1653
1654                 # Normalize &entities and whitespace
1655                 $text = self::decodeCharReferences( $text );
1656                 $text = self::normalizeWhitespace( $text );
1657
1658                 return $text;
1659         }
1660
1661         /**
1662          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1663          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1664          * PHP 5.1 doesn't.
1665          *
1666          * Use for passing XHTML fragments to PHP's XML parsing functions
1667          *
1668          * @return String
1669          */
1670         static function hackDocType() {
1671                 $out = "<!DOCTYPE html [\n";
1672                 foreach( self::$htmlEntities as $entity => $codepoint ) {
1673                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1674                 }
1675                 $out .= "]>\n";
1676                 return $out;
1677         }
1678
1679         /**
1680          * @param $url string
1681          * @return mixed|string
1682          */
1683         static function cleanUrl( $url ) {
1684                 # Normalize any HTML entities in input. They will be
1685                 # re-escaped by makeExternalLink().
1686                 $url = Sanitizer::decodeCharReferences( $url );
1687
1688                 # Escape any control characters introduced by the above step
1689                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1690                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1691
1692                 # Validate hostname portion
1693                 $matches = array();
1694                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1695                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1696
1697                         // Characters that will be ignored in IDNs.
1698                         // http://tools.ietf.org/html/3454#section-3.1
1699                         // Strip them before further processing so blacklists and such work.
1700                         $strip = "/
1701                                 \\s|          # general whitespace
1702                                 \xc2\xad|     # 00ad SOFT HYPHEN
1703                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1704                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1705                                 \xe2\x81\xa0| # 2060 WORD JOINER
1706                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1707                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1708                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1709                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1710                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1711                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1712                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1713                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1714                                 /xuD";
1715
1716                         $host = preg_replace( $strip, '', $host );
1717
1718                         // @todo FIXME: Validate hostnames here
1719
1720                         return $protocol . $host . $rest;
1721                 } else {
1722                         return $url;
1723                 }
1724         }
1725
1726         /**
1727          * @param $matches array
1728          * @return string
1729          */
1730         static function cleanUrlCallback( $matches ) {
1731                 return urlencode( $matches[0] );
1732         }
1733
1734         /**
1735          * Does a string look like an e-mail address?
1736          *
1737          * This validates an email address using an HTML5 specification found at:
1738          * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1739          * Which as of 2011-01-24 says:
1740          *
1741          *   A valid e-mail address is a string that matches the ABNF production
1742          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1743          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1744          *   3.5.
1745          *
1746          * This function is an implementation of the specification as requested in
1747          * bug 22449.
1748          *
1749          * Client-side forms will use the same standard validation rules via JS or
1750          * HTML 5 validation; additional restrictions can be enforced server-side
1751          * by extensions via the 'isValidEmailAddr' hook.
1752          *
1753          * Note that this validation doesn't 100% match RFC 2822, but is believed
1754          * to be liberal enough for wide use. Some invalid addresses will still
1755          * pass validation here.
1756          *
1757          * @since 1.18
1758          *
1759          * @param $addr String E-mail address
1760          * @return Bool
1761          */
1762         public static function validateEmail( $addr ) {
1763                 $result = null;
1764                 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1765                         return $result;
1766                 }
1767
1768                 // Please note strings below are enclosed in brackets [], this make the
1769                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1770                 // See bug 26948
1771                 $rfc5322_atext   = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1772                 $rfc1034_ldh_str = "a-z0-9\\-" ;
1773
1774                 $HTML5_email_regexp = "/
1775                 ^                      # start of string
1776                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1777                 @                      # 'apostrophe'
1778                 [$rfc1034_ldh_str]+       # First domain part
1779                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1780                 $                      # End of string
1781                 /ix" ; // case Insensitive, eXtended
1782
1783                 return (bool) preg_match( $HTML5_email_regexp, $addr );
1784         }
1785 }