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