includes/Sanitizer.php

   1 <?php
   2 /**
   3  * HTML 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  * HTML 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          */
  58         private static $htmlEntities = array(
  59                 'Aacute'   => 193,
  60                 'aacute'   => 225,
  61                 'Acirc'    => 194,
  62                 'acirc'    => 226,
  63                 'acute'    => 180,
  64                 'AElig'    => 198,
  65                 'aelig'    => 230,
  66                 'Agrave'   => 192,
  67                 'agrave'   => 224,
  68                 'alefsym'  => 8501,
  69                 'Alpha'    => 913,
  70                 'alpha'    => 945,
  71                 'amp'      => 38,
  72                 'and'      => 8743,
  73                 'ang'      => 8736,
  74                 'apos'     => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
  75                 'Aring'    => 197,
  76                 'aring'    => 229,
  77                 'asymp'    => 8776,
  78                 'Atilde'   => 195,
  79                 'atilde'   => 227,
  80                 'Auml'     => 196,
  81                 'auml'     => 228,
  82                 'bdquo'    => 8222,
  83                 'Beta'     => 914,
  84                 'beta'     => 946,
  85                 'brvbar'   => 166,
  86                 'bull'     => 8226,
  87                 'cap'      => 8745,
  88                 'Ccedil'   => 199,
  89                 'ccedil'   => 231,
  90                 'cedil'    => 184,
  91                 'cent'     => 162,
  92                 'Chi'      => 935,
  93                 'chi'      => 967,
  94                 'circ'     => 710,
  95                 'clubs'    => 9827,
  96                 'cong'     => 8773,
  97                 'copy'     => 169,
  98                 'crarr'    => 8629,
  99                 'cup'      => 8746,
 100                 'curren'   => 164,
 101                 'dagger'   => 8224,
 102                 'Dagger'   => 8225,
 103                 'darr'     => 8595,
 104                 'dArr'     => 8659,
 105                 'deg'      => 176,
 106                 'Delta'    => 916,
 107                 'delta'    => 948,
 108                 'diams'    => 9830,
 109                 'divide'   => 247,
 110                 'Eacute'   => 201,
 111                 'eacute'   => 233,
 112                 'Ecirc'    => 202,
 113                 'ecirc'    => 234,
 114                 'Egrave'   => 200,
 115                 'egrave'   => 232,
 116                 'empty'    => 8709,
 117                 'emsp'     => 8195,
 118                 'ensp'     => 8194,
 119                 'Epsilon'  => 917,
 120                 'epsilon'  => 949,
 121                 'equiv'    => 8801,
 122                 'Eta'      => 919,
 123                 'eta'      => 951,
 124                 'ETH'      => 208,
 125                 'eth'      => 240,
 126                 'Euml'     => 203,
 127                 'euml'     => 235,
 128                 'euro'     => 8364,
 129                 'exist'    => 8707,
 130                 'fnof'     => 402,
 131                 'forall'   => 8704,
 132                 'frac12'   => 189,
 133                 'frac14'   => 188,
 134                 'frac34'   => 190,
 135                 'frasl'    => 8260,
 136                 'Gamma'    => 915,
 137                 'gamma'    => 947,
 138                 'ge'       => 8805,
 139                 'gt'       => 62,
 140                 'harr'     => 8596,
 141                 'hArr'     => 8660,
 142                 'hearts'   => 9829,
 143                 'hellip'   => 8230,
 144                 'Iacute'   => 205,
 145                 'iacute'   => 237,
 146                 'Icirc'    => 206,
 147                 'icirc'    => 238,
 148                 'iexcl'    => 161,
 149                 'Igrave'   => 204,
 150                 'igrave'   => 236,
 151                 'image'    => 8465,
 152                 'infin'    => 8734,
 153                 'int'      => 8747,
 154                 'Iota'     => 921,
 155                 'iota'     => 953,
 156                 'iquest'   => 191,
 157                 'isin'     => 8712,
 158                 'Iuml'     => 207,
 159                 'iuml'     => 239,
 160                 'Kappa'    => 922,
 161                 'kappa'    => 954,
 162                 'Lambda'   => 923,
 163                 'lambda'   => 955,
 164                 'lang'     => 9001,
 165                 'laquo'    => 171,
 166                 'larr'     => 8592,
 167                 'lArr'     => 8656,
 168                 'lceil'    => 8968,
 169                 'ldquo'    => 8220,
 170                 'le'       => 8804,
 171                 'lfloor'   => 8970,
 172                 'lowast'   => 8727,
 173                 'loz'      => 9674,
 174                 'lrm'      => 8206,
 175                 'lsaquo'   => 8249,
 176                 'lsquo'    => 8216,
 177                 'lt'       => 60,
 178                 'macr'     => 175,
 179                 'mdash'    => 8212,
 180                 'micro'    => 181,
 181                 'middot'   => 183,
 182                 'minus'    => 8722,
 183                 'Mu'       => 924,
 184                 'mu'       => 956,
 185                 'nabla'    => 8711,
 186                 'nbsp'     => 160,
 187                 'ndash'    => 8211,
 188                 'ne'       => 8800,
 189                 'ni'       => 8715,
 190                 'not'      => 172,
 191                 'notin'    => 8713,
 192                 'nsub'     => 8836,
 193                 'Ntilde'   => 209,
 194                 'ntilde'   => 241,
 195                 'Nu'       => 925,
 196                 'nu'       => 957,
 197                 'Oacute'   => 211,
 198                 'oacute'   => 243,
 199                 'Ocirc'    => 212,
 200                 'ocirc'    => 244,
 201                 'OElig'    => 338,
 202                 'oelig'    => 339,
 203                 'Ograve'   => 210,
 204                 'ograve'   => 242,
 205                 'oline'    => 8254,
 206                 'Omega'    => 937,
 207                 'omega'    => 969,
 208                 'Omicron'  => 927,
 209                 'omicron'  => 959,
 210                 'oplus'    => 8853,
 211                 'or'       => 8744,
 212                 'ordf'     => 170,
 213                 'ordm'     => 186,
 214                 'Oslash'   => 216,
 215                 'oslash'   => 248,
 216                 'Otilde'   => 213,
 217                 'otilde'   => 245,
 218                 'otimes'   => 8855,
 219                 'Ouml'     => 214,
 220                 'ouml'     => 246,
 221                 'para'     => 182,
 222                 'part'     => 8706,
 223                 'permil'   => 8240,
 224                 'perp'     => 8869,
 225                 'Phi'      => 934,
 226                 'phi'      => 966,
 227                 'Pi'       => 928,
 228                 'pi'       => 960,
 229                 'piv'      => 982,
 230                 'plusmn'   => 177,
 231                 'pound'    => 163,
 232                 'prime'    => 8242,
 233                 'Prime'    => 8243,
 234                 'prod'     => 8719,
 235                 'prop'     => 8733,
 236                 'Psi'      => 936,
 237                 'psi'      => 968,
 238                 'quot'     => 34,
 239                 'radic'    => 8730,
 240                 'rang'     => 9002,
 241                 'raquo'    => 187,
 242                 'rarr'     => 8594,
 243                 'rArr'     => 8658,
 244                 'rceil'    => 8969,
 245                 'rdquo'    => 8221,
 246                 'real'     => 8476,
 247                 'reg'      => 174,
 248                 'rfloor'   => 8971,
 249                 'Rho'      => 929,
 250                 'rho'      => 961,
 251                 'rlm'      => 8207,
 252                 'rsaquo'   => 8250,
 253                 'rsquo'    => 8217,
 254                 'sbquo'    => 8218,
 255                 'Scaron'   => 352,
 256                 'scaron'   => 353,
 257                 'sdot'     => 8901,
 258                 'sect'     => 167,
 259                 'shy'      => 173,
 260                 'Sigma'    => 931,
 261                 'sigma'    => 963,
 262                 'sigmaf'   => 962,
 263                 'sim'      => 8764,
 264                 'spades'   => 9824,
 265                 'sub'      => 8834,
 266                 'sube'     => 8838,
 267                 'sum'      => 8721,
 268                 'sup'      => 8835,
 269                 'sup1'     => 185,
 270                 'sup2'     => 178,
 271                 'sup3'     => 179,
 272                 'supe'     => 8839,
 273                 'szlig'    => 223,
 274                 'Tau'      => 932,
 275                 'tau'      => 964,
 276                 'there4'   => 8756,
 277                 'Theta'    => 920,
 278                 'theta'    => 952,
 279                 'thetasym' => 977,
 280                 'thinsp'   => 8201,
 281                 'THORN'    => 222,
 282                 'thorn'    => 254,
 283                 'tilde'    => 732,
 284                 'times'    => 215,
 285                 'trade'    => 8482,
 286                 'Uacute'   => 218,
 287                 'uacute'   => 250,
 288                 'uarr'     => 8593,
 289                 'uArr'     => 8657,
 290                 'Ucirc'    => 219,
 291                 'ucirc'    => 251,
 292                 'Ugrave'   => 217,
 293                 'ugrave'   => 249,
 294                 'uml'      => 168,
 295                 'upsih'    => 978,
 296                 'Upsilon'  => 933,
 297                 'upsilon'  => 965,
 298                 'Uuml'     => 220,
 299                 'uuml'     => 252,
 300                 'weierp'   => 8472,
 301                 'Xi'       => 926,
 302                 'xi'       => 958,
 303                 'Yacute'   => 221,
 304                 'yacute'   => 253,
 305                 'yen'      => 165,
 306                 'Yuml'     => 376,
 307                 'yuml'     => 255,
 308                 'Zeta'     => 918,
 309                 'zeta'     => 950,
 310                 'zwj'      => 8205,
 311                 'zwnj'     => 8204
 312         );
 313
 314         /**
 315          * Character entity aliases accepted by MediaWiki
 316          */
 317         private static $htmlEntityAliases = array(
 318                 'רלמ' => 'rlm',
 319                 'رلم' => 'rlm',
 320         );
 321
 322         /**
 323          * Lazy-initialised attributes regex, see getAttribsRegex()
 324          */
 325         private static $attribsRegex;
 326
 327         /**
 328          * Regular expression to match HTML/XML attribute pairs within a tag.
 329          * Allows some... latitude.
 330          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 331          */
 332         static function getAttribsRegex() {
 333                 if ( self::$attribsRegex === null ) {
 334                         $attribFirst = '[:A-Z_a-z0-9]';
 335                         $attrib = '[:A-Z_a-z-.0-9]';
 336                         $space = '[\x09\x0a\x0d\x20]';
 337                         self::$attribsRegex =
 338                                 "/(?:^|$space)({$attribFirst}{$attrib}*)
 339                                   ($space*=$space*
 340                                         (?:
 341                                          # The attribute value: quoted or alone
 342                                           \"([^<\"]*)\"
 343                                          | '([^<']*)'
 344                                          |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
 345                                          |  (\#[0-9a-fA-F]+) # Technically wrong, but lots of
 346                                                                                  # colors are specified like this.
 347                                                                                  # We'll be normalizing it.
 348                                         )
 349                                 )?(?=$space|\$)/sx";
 350                 }
 351                 return self::$attribsRegex;
 352         }
 353
 354         /**
 355          * Cleans up HTML, removes dangerous tags and attributes, and
 356          * removes HTML comments
 357          * @private
 358          * @param $text String
 359          * @param $processCallback Callback to do any variable or parameter
 360          *        replacements in HTML attribute values
 361          * @param array $args for the processing callback
 362          * @param array $extratags for any extra tags to include
 363          * @param array $removetags for any tags (default or extra) to exclude
 364          * @return string
 365          */
 366         static function removeHTMLtags( $text, $processCallback = null,
 367                 $args = array(), $extratags = array(), $removetags = array()
 368         ) {
 369                 global $wgUseTidy, $wgAllowMicrodataAttributes, $wgAllowImageTag;
 370
 371                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 372                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 373
 374                 wfProfileIn( __METHOD__ );
 375
 376                 // Base our staticInitialised variable off of the global config state so that if the globals
 377                 // are changed (like in the screwed up test system) we will re-initialise the settings.
 378                 $globalContext = implode( '-', compact( 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
 379                 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
 380
 381                         $htmlpairsStatic = array( # Tags that must be closed
 382                                 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 383                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 384                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 385                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 386                                 'ruby', 'rt', 'rb', 'rp', 'p', 'span', 'abbr', 'dfn',
 387                                 'kbd', 'samp', 'data', 'time', 'mark'
 388                         );
 389                         $htmlsingle = array(
 390                                 'br', 'wbr', 'hr', 'li', 'dt', 'dd'
 391                         );
 392                         $htmlsingleonly = array( # Elements that cannot have close tags
 393                                 'br', 'wbr', 'hr'
 394                         );
 395                         if ( $wgAllowMicrodataAttributes ) {
 396                                 $htmlsingle[] = $htmlsingleonly[] = 'meta';
 397                                 $htmlsingle[] = $htmlsingleonly[] = 'link';
 398                         }
 399                         $htmlnest = array( # Tags that can be nested--??
 400                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 401                                 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
 402                                 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
 403                         );
 404                         $tabletags = array( # Can only appear inside table, we will close them
 405                                 'td', 'th', 'tr',
 406                         );
 407                         $htmllist = array( # Tags used by list
 408                                 'ul', 'ol',
 409                         );
 410                         $listtags = array( # Tags that can appear in a list
 411                                 'li',
 412                         );
 413
 414                         if ( $wgAllowImageTag ) {
 415                                 $htmlsingle[] = 'img';
 416                                 $htmlsingleonly[] = 'img';
 417                         }
 418
 419                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 420                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 421
 422                         # Convert them all to hashtables for faster lookup
 423                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 424                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 425                         foreach ( $vars as $var ) {
 426                                 $$var = array_flip( $$var );
 427                         }
 428                         $staticInitialised = $globalContext;
 429                 }
 430                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 431                 $extratags = array_flip( $extratags );
 432                 $removetags = array_flip( $removetags );
 433                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 434                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
 435
 436                 # Remove HTML comments
 437                 $text = Sanitizer::removeHTMLcomments( $text );
 438                 $bits = explode( '<', $text );
 439                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 440                 if ( !$wgUseTidy ) {
 441                         $tagstack = $tablestack = array();
 442                         foreach ( $bits as $x ) {
 443                                 $regs = array();
 444                                 # $slash: Does the current element start with a '/'?
 445                                 # $t: Current element name
 446                                 # $params: String between element name and >
 447                                 # $brace: Ending '>' or '/>'
 448                                 # $rest: Everything until the next element of $bits
 449                                 if ( preg_match( '!^(/?)([^\\s/>]+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 450                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 451                                 } else {
 452                                         $slash = $t = $params = $brace = $rest = null;
 453                                 }
 454
 455                                 $badtag = false;
 456                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 457                                         # Check our stack
 458                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 459                                                 $badtag = true;
 460                                         } elseif ( $slash ) {
 461                                                 # Closing a tag... is it the one we just opened?
 462                                                 $ot = @array_pop( $tagstack );
 463                                                 if ( $ot != $t ) {
 464                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 465                                                                 # Pop all elements with an optional close tag
 466                                                                 # and see if we find a match below them
 467                                                                 $optstack = array();
 468                                                                 array_push( $optstack, $ot );
 469                                                                 wfSuppressWarnings();
 470                                                                 $ot = array_pop( $tagstack );
 471                                                                 wfRestoreWarnings();
 472                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 473                                                                         array_push( $optstack, $ot );
 474                                                                         wfSuppressWarnings();
 475                                                                         $ot = array_pop( $tagstack );
 476                                                                         wfRestoreWarnings();
 477                                                                 }
 478                                                                 if ( $t != $ot ) {
 479                                                                         # No match. Push the optional elements back again
 480                                                                         $badtag = true;
 481                                                                         wfSuppressWarnings();
 482                                                                         $ot = array_pop( $optstack );
 483                                                                         wfRestoreWarnings();
 484                                                                         while ( $ot ) {
 485                                                                                 array_push( $tagstack, $ot );
 486                                                                                 wfSuppressWarnings();
 487                                                                                 $ot = array_pop( $optstack );
 488                                                                                 wfRestoreWarnings();
 489                                                                         }
 490                                                                 }
 491                                                         } else {
 492                                                                 @array_push( $tagstack, $ot );
 493                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 494                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 495                                                                         $badtag = true;
 496                                                                 }
 497                                                         }
 498                                                 } else {
 499                                                         if ( $t == 'table' ) {
 500                                                                 $tagstack = array_pop( $tablestack );
 501                                                         }
 502                                                 }
 503                                                 $newparams = '';
 504                                         } else {
 505                                                 # Keep track for later
 506                                                 if ( isset( $tabletags[$t] ) &&
 507                                                 !in_array( 'table', $tagstack ) ) {
 508                                                         $badtag = true;
 509                                                 } elseif ( in_array( $t, $tagstack ) &&
 510                                                 !isset( $htmlnest[$t] ) ) {
 511                                                         $badtag = true;
 512                                                 # Is it a self closed htmlpair ? (bug 5487)
 513                                                 } elseif ( $brace == '/>' &&
 514                                                 isset( $htmlpairs[$t] ) ) {
 515                                                         $badtag = true;
 516                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 517                                                         # Hack to force empty tag for unclosable elements
 518                                                         $brace = '/>';
 519                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 520                                                         # Hack to not close $htmlsingle tags
 521                                                         $brace = null;
 522                                                         # Still need to push this optionally-closed tag to
 523                                                         # the tag stack so that we can match end tags
 524                                                         # instead of marking them as bad.
 525                                                         array_push( $tagstack, $t );
 526                                                 } elseif ( isset( $tabletags[$t] )
 527                                                 && in_array( $t, $tagstack ) ) {
 528                                                         // New table tag but forgot to close the previous one
 529                                                         $text .= "</$t>";
 530                                                 } else {
 531                                                         if ( $t == 'table' ) {
 532                                                                 array_push( $tablestack, $tagstack );
 533                                                                 $tagstack = array();
 534                                                         }
 535                                                         array_push( $tagstack, $t );
 536                                                 }
 537
 538                                                 # Replace any variables or template parameters with
 539                                                 # plaintext results.
 540                                                 if ( is_callable( $processCallback ) ) {
 541                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 542                                                 }
 543
 544                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 545                                                         $badtag = true;
 546                                                 }
 547
 548                                                 # Strip non-approved attributes from the tag
 549                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 550                                         }
 551                                         if ( !$badtag ) {
 552                                                 $rest = str_replace( '>', '&gt;', $rest );
 553                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 554                                                 $text .= "<$slash$t$newparams$close>$rest";
 555                                                 continue;
 556                                         }
 557                                 }
 558                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 559                         }
 560                         # Close off any remaining tags
 561                         while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
 562                                 $text .= "</$t>\n";
 563                                 if ( $t == 'table' ) {
 564                                         $tagstack = array_pop( $tablestack );
 565                                 }
 566                         }
 567                 } else {
 568                         # this might be possible using tidy itself
 569                         foreach ( $bits as $x ) {
 570                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 571                                 $x, $regs );
 572                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 573                                 $badtag = false;
 574                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 575                                         if ( is_callable( $processCallback ) ) {
 576                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 577                                         }
 578
 579                                         if ( !Sanitizer::validateTag( $params, $t ) ) {
 580                                                 $badtag = true;
 581                                         }
 582
 583                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 584                                         if ( !$badtag ) {
 585                                                 $rest = str_replace( '>', '&gt;', $rest );
 586                                                 $text .= "<$slash$t$newparams$brace$rest";
 587                                                 continue;
 588                                         }
 589                                 }
 590                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 591                         }
 592                 }
 593                 wfProfileOut( __METHOD__ );
 594                 return $text;
 595         }
 596
 597         /**
 598          * Remove '<!--', '-->', and everything between.
 599          * To avoid leaving blank lines, when a comment is both preceded
 600          * and followed by a newline (ignoring spaces), trim leading and
 601          * trailing spaces and one of the newlines.
 602          *
 603          * @private
 604          * @param $text String
 605          * @return string
 606          */
 607         static function removeHTMLcomments( $text ) {
 608                 wfProfileIn( __METHOD__ );
 609                 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
 610                         $end = strpos( $text, '-->', $start + 4 );
 611                         if ( $end === false ) {
 612                                 # Unterminated comment; bail out
 613                                 break;
 614                         }
 615
 616                         $end += 3;
 617
 618                         # Trim space and newline if the comment is both
 619                         # preceded and followed by a newline
 620                         $spaceStart = max( $start - 1, 0 );
 621                         $spaceLen = $end - $spaceStart;
 622                         while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
 623                                 $spaceStart--;
 624                                 $spaceLen++;
 625                         }
 626                         while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
 627                                 $spaceLen++;
 628                         }
 629                         if ( substr( $text, $spaceStart, 1 ) === "\n"
 630                                 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
 631                                 # Remove the comment, leading and trailing
 632                                 # spaces, and leave only one newline.
 633                                 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
 634                         }
 635                         else {
 636                                 # Remove just the comment.
 637                                 $text = substr_replace( $text, '', $start, $end - $start );
 638                         }
 639                 }
 640                 wfProfileOut( __METHOD__ );
 641                 return $text;
 642         }
 643
 644         /**
 645          * Takes attribute names and values for a tag and the tag name and
 646          * validates that the tag is allowed to be present.
 647          * This DOES NOT validate the attributes, nor does it validate the
 648          * tags themselves. This method only handles the special circumstances
 649          * where we may want to allow a tag within content but ONLY when it has
 650          * specific attributes set.
 651          *
 652          * @param $params
 653          * @param $element
 654          * @return bool
 655          */
 656         static function validateTag( $params, $element ) {
 657                 $params = Sanitizer::decodeTagAttributes( $params );
 658
 659                 if ( $element == 'meta' || $element == 'link' ) {
 660                         if ( !isset( $params['itemprop'] ) ) {
 661                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 662                                 return false;
 663                         }
 664                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 665                                 // <meta> must have a content="" for the itemprop
 666                                 return false;
 667                         }
 668                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 669                                 // <link> must have an associated href=""
 670                                 return false;
 671                         }
 672                 }
 673
 674                 return true;
 675         }
 676
 677         /**
 678          * Take an array of attribute names and values and normalize or discard
 679          * illegal values for the given element type.
 680          *
 681          * - Discards attributes not on a whitelist for the given element
 682          * - Unsafe style attributes are discarded
 683          * - Invalid id attributes are re-encoded
 684          *
 685          * @param $attribs Array
 686          * @param $element String
 687          * @return Array
 688          *
 689          * @todo Check for legal values where the DTD limits things.
 690          * @todo Check for unique id attribute :P
 691          */
 692         static function validateTagAttributes( $attribs, $element ) {
 693                 return Sanitizer::validateAttributes( $attribs,
 694                         Sanitizer::attributeWhitelist( $element ) );
 695         }
 696
 697         /**
 698          * Take an array of attribute names and values and normalize or discard
 699          * illegal values for the given whitelist.
 700          *
 701          * - Discards attributes not the given whitelist
 702          * - Unsafe style attributes are discarded
 703          * - Invalid id attributes are re-encoded
 704          *
 705          * @param $attribs Array
 706          * @param array $whitelist list of allowed attribute names
 707          * @return Array
 708          *
 709          * @todo Check for legal values where the DTD limits things.
 710          * @todo Check for unique id attribute :P
 711          */
 712         static function validateAttributes( $attribs, $whitelist ) {
 713                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
 714
 715                 $whitelist = array_flip( $whitelist );
 716                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 717
 718                 $out = array();
 719                 foreach ( $attribs as $attribute => $value ) {
 720                         #allow XML namespace declaration if RDFa is enabled
 721                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 722                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 723                                         $out[$attribute] = $value;
 724                                 }
 725
 726                                 continue;
 727                         }
 728
 729                         # Allow any attribute beginning with "data-"
 730                         if ( !preg_match( '/^data-/i', $attribute ) && !isset( $whitelist[$attribute] ) ) {
 731                                 continue;
 732                         }
 733
 734                         # Strip javascript "expression" from stylesheets.
 735                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 736                         if ( $attribute == 'style' ) {
 737                                 $value = Sanitizer::checkCss( $value );
 738                         }
 739
 740                         if ( $attribute === 'id' ) {
 741                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 742                         }
 743
 744                         # WAI-ARIA
 745                         # http://www.w3.org/TR/wai-aria/
 746                         # http://www.whatwg.org/html/elements.html#wai-aria
 747                         # For now we only support role="presentation" until we work out what roles should be
 748                         # usable by content and we ensure that our code explicitly rejects patterns that
 749                         # violate HTML5's ARIA restrictions.
 750                         if ( $attribute === 'role' && $value !== 'presentation' ) {
 751                                 continue;
 752                         }
 753
 754                         // RDFa and microdata properties allow URLs, URIs and/or CURIs.
 755                         // Check them for sanity.
 756                         if ( $attribute === 'rel' || $attribute === 'rev'
 757                                 # RDFa
 758                                 || $attribute === 'about' || $attribute === 'property'
 759                                 || $attribute === 'resource' || $attribute === 'datatype'
 760                                 || $attribute === 'typeof'
 761                                 # HTML5 microdata
 762                                 || $attribute === 'itemid' || $attribute === 'itemprop'
 763                                 || $attribute === 'itemref' || $attribute === 'itemscope'
 764                                 || $attribute === 'itemtype'
 765                         ) {
 766                                 //Paranoia. Allow "simple" values but suppress javascript
 767                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 768                                         continue;
 769                                 }
 770                         }
 771
 772                         # NOTE: even though elements using href/src are not allowed directly, supply
 773                         #       validation code that can be used by tag hook handlers, etc
 774                         if ( $attribute === 'href' || $attribute === 'src' ) {
 775                                 if ( !preg_match( $hrefExp, $value ) ) {
 776                                         continue; //drop any href or src attributes not using an allowed protocol.
 777                                         // NOTE: this also drops all relative URLs
 778                                 }
 779                         }
 780
 781                         // If this attribute was previously set, override it.
 782                         // Output should only have one attribute of each name.
 783                         $out[$attribute] = $value;
 784                 }
 785
 786                 if ( $wgAllowMicrodataAttributes ) {
 787                         # itemtype, itemid, itemref don't make sense without itemscope
 788                         if ( !array_key_exists( 'itemscope', $out ) ) {
 789                                 unset( $out['itemtype'] );
 790                                 unset( $out['itemid'] );
 791                                 unset( $out['itemref'] );
 792                         }
 793                         # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 794                 }
 795                 return $out;
 796         }
 797
 798         /**
 799          * Merge two sets of HTML attributes.  Conflicting items in the second set
 800          * will override those in the first, except for 'class' attributes which
 801          * will be combined (if they're both strings).
 802          *
 803          * @todo implement merging for other attributes such as style
 804          * @param $a Array
 805          * @param $b Array
 806          * @return array
 807          */
 808         static function mergeAttributes( $a, $b ) {
 809                 $out = array_merge( $a, $b );
 810                 if ( isset( $a['class'] ) && isset( $b['class'] )
 811                         && is_string( $a['class'] ) && is_string( $b['class'] )
 812                         && $a['class'] !== $b['class']
 813                 ) {
 814                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 815                                 -1, PREG_SPLIT_NO_EMPTY );
 816                         $out['class'] = implode( ' ', array_unique( $classes ) );
 817                 }
 818                 return $out;
 819         }
 820
 821         /**
 822          * Pick apart some CSS and check it for forbidden or unsafe structures.
 823          * Returns a sanitized string. This sanitized string will have
 824          * character references and escape sequences decoded and comments
 825          * stripped (unless it is itself one valid comment, in which case the value
 826          * will be passed through). If the input is just too evil, only a comment
 827          * complaining about evilness will be returned.
 828          *
 829          * Currently URL references, 'expression', 'tps' are forbidden.
 830          *
 831          * NOTE: Despite the fact that character references are decoded, the
 832          * returned string may contain character references given certain
 833          * clever input strings. These character references must
 834          * be escaped before the return value is embedded in HTML.
 835          *
 836          * @param $value String
 837          * @return String
 838          */
 839         static function checkCss( $value ) {
 840                 // Decode character references like &#123;
 841                 $value = Sanitizer::decodeCharReferences( $value );
 842
 843                 // Decode escape sequences and line continuation
 844                 // See the grammar in the CSS 2 spec, appendix D.
 845                 // This has to be done AFTER decoding character references.
 846                 // This means it isn't possible for this function to return
 847                 // unsanitized escape sequences. It is possible to manufacture
 848                 // input that contains character references that decode to
 849                 // escape sequences that decode to character references, but
 850                 // it's OK for the return value to contain character references
 851                 // because the caller is supposed to escape those anyway.
 852                 static $decodeRegex;
 853                 if ( !$decodeRegex ) {
 854                         $space = '[\\x20\\t\\r\\n\\f]';
 855                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 856                         $backslash = '\\\\';
 857                         $decodeRegex = "/ $backslash
 858                                 (?:
 859                                         ($nl) |  # 1. Line continuation
 860                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 861                                         (.) | # 3. backslash cancelling special meaning
 862                                         () | # 4. backslash at end of string
 863                                 )/xu";
 864                 }
 865                 $value = preg_replace_callback( $decodeRegex,
 866                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 867
 868                 // Let the value through if it's nothing but a single comment, to
 869                 // allow other functions which may reject it to pass some error
 870                 // message through.
 871                 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
 872                         // Remove any comments; IE gets token splitting wrong
 873                         // This must be done AFTER decoding character references and
 874                         // escape sequences, because those steps can introduce comments
 875                         // This step cannot introduce character references or escape
 876                         // sequences, because it replaces comments with spaces rather
 877                         // than removing them completely.
 878                         $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 879
 880                         // Remove anything after a comment-start token, to guard against
 881                         // incorrect client implementations.
 882                         $commentPos = strpos( $value, '/*' );
 883                         if ( $commentPos !== false ) {
 884                                 $value = substr( $value, 0, $commentPos );
 885                         }
 886                 }
 887
 888                 // Reject problematic keywords and control characters
 889                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 890                         return '/* invalid control char */';
 891                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( | image\s*\( | image-set\s*\( !ix', $value ) ) {
 892                         return '/* insecure input */';
 893                 }
 894                 return $value;
 895         }
 896
 897         /**
 898          * @param $matches array
 899          * @return String
 900          */
 901         static function cssDecodeCallback( $matches ) {
 902                 if ( $matches[1] !== '' ) {
 903                         // Line continuation
 904                         return '';
 905                 } elseif ( $matches[2] !== '' ) {
 906                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 907                 } elseif ( $matches[3] !== '' ) {
 908                         $char = $matches[3];
 909                 } else {
 910                         $char = '\\';
 911                 }
 912                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 913                         // These characters need to be escaped in strings
 914                         // Clean up the escape sequence to avoid parsing errors by clients
 915                         return '\\' . dechex( ord( $char ) ) . ' ';
 916                 } else {
 917                         // Decode unnecessary escape
 918                         return $char;
 919                 }
 920         }
 921
 922         /**
 923          * Take a tag soup fragment listing an HTML element's attributes
 924          * and normalize it to well-formed XML, discarding unwanted attributes.
 925          * Output is safe for further wikitext processing, with escaping of
 926          * values that could trigger problems.
 927          *
 928          * - Normalizes attribute names to lowercase
 929          * - Discards attributes not on a whitelist for the given element
 930          * - Turns broken or invalid entities into plaintext
 931          * - Double-quotes all attribute values
 932          * - Attributes without values are given the name as attribute
 933          * - Double attributes are discarded
 934          * - Unsafe style attributes are discarded
 935          * - Prepends space if there are attributes.
 936          *
 937          * @param $text String
 938          * @param $element String
 939          * @return String
 940          */
 941         static function fixTagAttributes( $text, $element ) {
 942                 if ( trim( $text ) == '' ) {
 943                         return '';
 944                 }
 945
 946                 $decoded = Sanitizer::decodeTagAttributes( $text );
 947                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
 948
 949                 return Sanitizer::safeEncodeTagAttributes( $stripped );
 950         }
 951
 952         /**
 953          * Encode an attribute value for HTML output.
 954          * @param $text String
 955          * @return HTML-encoded text fragment
 956          */
 957         static function encodeAttribute( $text ) {
 958                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 959
 960                 // Whitespace is normalized during attribute decoding,
 961                 // so if we've been passed non-spaces we must encode them
 962                 // ahead of time or they won't be preserved.
 963                 $encValue = strtr( $encValue, array(
 964                         "\n" => '&#10;',
 965                         "\r" => '&#13;',
 966                         "\t" => '&#9;',
 967                 ) );
 968
 969                 return $encValue;
 970         }
 971
 972         /**
 973          * Encode an attribute value for HTML tags, with extra armoring
 974          * against further wiki processing.
 975          * @param $text String
 976          * @return HTML-encoded text fragment
 977          */
 978         static function safeEncodeAttribute( $text ) {
 979                 $encValue = Sanitizer::encodeAttribute( $text );
 980
 981                 # Templates and links may be expanded in later parsing,
 982                 # creating invalid or dangerous output. Suppress this.
 983                 $encValue = strtr( $encValue, array(
 984                         '<'    => '&lt;',   // This should never happen,
 985                         '>'    => '&gt;',   // we've received invalid input
 986                         '"'    => '&quot;', // which should have been escaped.
 987                         '{'    => '&#123;',
 988                         '['    => '&#91;',
 989                         "''"   => '&#39;&#39;',
 990                         'ISBN' => '&#73;SBN',
 991                         'RFC'  => '&#82;FC',
 992                         'PMID' => '&#80;MID',
 993                         '|'    => '&#124;',
 994                         '__'   => '&#95;_',
 995                 ) );
 996
 997                 # Stupid hack
 998                 $encValue = preg_replace_callback(
 999                         '/((?i)' . wfUrlProtocols() . ')/',
1000                         array( 'Sanitizer', 'armorLinksCallback' ),
1001                         $encValue );
1002                 return $encValue;
1003         }
1004
1005         /**
1006          * Given a value, escape it so that it can be used in an id attribute and
1007          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1008          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1009          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1010          * escaped with lots of dots.
1011          *
1012          * To ensure we don't have to bother escaping anything, we also strip ', ",
1013          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1014          * We also strip # because it upsets IE, and % because it could be
1015          * ambiguous if it's part of something that looks like a percent escape
1016          * (which don't work reliably in fragments cross-browser).
1017          *
1018          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1019          *                                                          in the id and
1020          *                                                          name attributes
1021          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1022          * @see http://www.whatwg.org/html/elements.html#the-id-attribute
1023          *   HTML5 definition of id attribute
1024          *
1025          * @param string $id id to escape
1026          * @param $options Mixed: string or array of strings (default is array()):
1027          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1028          *       so don't pay attention if the first character isn't valid at the
1029          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1030          *       false.
1031          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1032          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1033          *       anchors and links won't break.
1034          * @return String
1035          */
1036         static function escapeId( $id, $options = array() ) {
1037                 global $wgExperimentalHtmlIds;
1038                 $options = (array)$options;
1039
1040                 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1041                         $id = Sanitizer::decodeCharReferences( $id );
1042                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1043                         $id = trim( $id, '_' );
1044                         if ( $id === '' ) {
1045                                 # Must have been all whitespace to start with.
1046                                 return '_';
1047                         } else {
1048                                 return $id;
1049                         }
1050                 }
1051
1052                 # HTML4-style escaping
1053                 static $replace = array(
1054                         '%3A' => ':',
1055                         '%' => '.'
1056                 );
1057
1058                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1059                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1060
1061                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1062                 && !in_array( 'noninitial', $options ) ) {
1063                         // Initial character must be a letter!
1064                         $id = "x$id";
1065                 }
1066                 return $id;
1067         }
1068
1069         /**
1070          * Given a value, escape it so that it can be used as a CSS class and
1071          * return it.
1072          *
1073          * @todo For extra validity, input should be validated UTF-8.
1074          *
1075          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1076          *
1077          * @param $class String
1078          * @return String
1079          */
1080         static function escapeClass( $class ) {
1081                 // Convert ugly stuff to underscores and kill underscores in ugly places
1082                 return rtrim( preg_replace(
1083                         array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1084                         '_',
1085                         $class ), '_' );
1086         }
1087
1088         /**
1089          * Given HTML input, escape with htmlspecialchars but un-escape entities.
1090          * This allows (generally harmless) entities like &#160; to survive.
1091          *
1092          * @param string $html to escape
1093          * @return String: escaped input
1094          */
1095         static function escapeHtmlAllowEntities( $html ) {
1096                 $html = Sanitizer::decodeCharReferences( $html );
1097                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1098                 # hurt.
1099                 $html = htmlspecialchars( $html, ENT_QUOTES );
1100                 return $html;
1101         }
1102
1103         /**
1104          * Regex replace callback for armoring links against further processing.
1105          * @param $matches Array
1106          * @return string
1107          */
1108         private static function armorLinksCallback( $matches ) {
1109                 return str_replace( ':', '&#58;', $matches[1] );
1110         }
1111
1112         /**
1113          * Return an associative array of attribute names and values from
1114          * a partial tag string. Attribute names are forces to lowercase,
1115          * character references are decoded to UTF-8 text.
1116          *
1117          * @param $text String
1118          * @return Array
1119          */
1120         public static function decodeTagAttributes( $text ) {
1121                 if ( trim( $text ) == '' ) {
1122                         return array();
1123                 }
1124
1125                 $attribs = array();
1126                 $pairs = array();
1127                 if ( !preg_match_all(
1128                         self::getAttribsRegex(),
1129                         $text,
1130                         $pairs,
1131                         PREG_SET_ORDER ) ) {
1132                         return $attribs;
1133                 }
1134
1135                 foreach ( $pairs as $set ) {
1136                         $attribute = strtolower( $set[1] );
1137                         $value = Sanitizer::getTagAttributeCallback( $set );
1138
1139                         // Normalize whitespace
1140                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1141                         $value = trim( $value );
1142
1143                         // Decode character references
1144                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1145                 }
1146                 return $attribs;
1147         }
1148
1149         /**
1150          * Build a partial tag string from an associative array of attribute
1151          * names and values as returned by decodeTagAttributes.
1152          *
1153          * @param $assoc_array Array
1154          * @return String
1155          */
1156         public static function safeEncodeTagAttributes( $assoc_array ) {
1157                 $attribs = array();
1158                 foreach ( $assoc_array as $attribute => $value ) {
1159                         $encAttribute = htmlspecialchars( $attribute );
1160                         $encValue = Sanitizer::safeEncodeAttribute( $value );
1161
1162                         $attribs[] = "$encAttribute=\"$encValue\"";
1163                 }
1164                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1165         }
1166
1167         /**
1168          * Pick the appropriate attribute value from a match set from the
1169          * attribs regex matches.
1170          *
1171          * @param $set Array
1172          * @throws MWException
1173          * @return String
1174          */
1175         private static function getTagAttributeCallback( $set ) {
1176                 if ( isset( $set[6] ) ) {
1177                         # Illegal #XXXXXX color with no quotes.
1178                         return $set[6];
1179                 } elseif ( isset( $set[5] ) ) {
1180                         # No quotes.
1181                         return $set[5];
1182                 } elseif ( isset( $set[4] ) ) {
1183                         # Single-quoted
1184                         return $set[4];
1185                 } elseif ( isset( $set[3] ) ) {
1186                         # Double-quoted
1187                         return $set[3];
1188                 } elseif ( !isset( $set[2] ) ) {
1189                         # In XHTML, attributes must have a value.
1190                         # For 'reduced' form, return explicitly the attribute name here.
1191                         return $set[1];
1192                 } else {
1193                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1194                 }
1195         }
1196
1197         /**
1198          * Normalize whitespace and character references in an XML source-
1199          * encoded text for an attribute value.
1200          *
1201          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1202          * but note that we're not returning the value, but are returning
1203          * XML source fragments that will be slapped into output.
1204          *
1205          * @param $text String
1206          * @return String
1207          */
1208         private static function normalizeAttributeValue( $text ) {
1209                 return str_replace( '"', '&quot;',
1210                         self::normalizeWhitespace(
1211                                 Sanitizer::normalizeCharReferences( $text ) ) );
1212         }
1213
1214         /**
1215          * @param $text string
1216          * @return mixed
1217          */
1218         private static function normalizeWhitespace( $text ) {
1219                 return preg_replace(
1220                         '/\r\n|[\x20\x0d\x0a\x09]/',
1221                         ' ',
1222                         $text );
1223         }
1224
1225         /**
1226          * Normalizes whitespace in a section name, such as might be returned
1227          * by Parser::stripSectionName(), for use in the id's that are used for
1228          * section links.
1229          *
1230          * @param $section String
1231          * @return String
1232          */
1233         static function normalizeSectionNameWhitespace( $section ) {
1234                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1235         }
1236
1237         /**
1238          * Ensure that any entities and character references are legal
1239          * for XML and XHTML specifically. Any stray bits will be
1240          * &amp;-escaped to result in a valid text fragment.
1241          *
1242          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1243          *   numericized (this way we're well-formed even without a DTD)
1244          * b. any numeric char refs must be legal chars, not invalid or forbidden
1245          * c. use lower cased "&#x", not "&#X"
1246          * d. fix or reject non-valid attributes
1247          *
1248          * @param $text String
1249          * @return String
1250          * @private
1251          */
1252         static function normalizeCharReferences( $text ) {
1253                 return preg_replace_callback(
1254                         self::CHAR_REFS_REGEX,
1255                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1256                         $text );
1257         }
1258         /**
1259          * @param $matches String
1260          * @return String
1261          */
1262         static function normalizeCharReferencesCallback( $matches ) {
1263                 $ret = null;
1264                 if ( $matches[1] != '' ) {
1265                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1266                 } elseif ( $matches[2] != '' ) {
1267                         $ret = Sanitizer::decCharReference( $matches[2] );
1268                 } elseif ( $matches[3] != '' ) {
1269                         $ret = Sanitizer::hexCharReference( $matches[3] );
1270                 }
1271                 if ( is_null( $ret ) ) {
1272                         return htmlspecialchars( $matches[0] );
1273                 } else {
1274                         return $ret;
1275                 }
1276         }
1277
1278         /**
1279          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1280          * return the equivalent numeric entity reference (except for the core &lt;
1281          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1282          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1283          * pseudo-entity source (eg &amp;foo;)
1284          *
1285          * @param $name String
1286          * @return String
1287          */
1288         static function normalizeEntity( $name ) {
1289                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1290                         return '&' . self::$htmlEntityAliases[$name] . ';';
1291                 } elseif ( in_array( $name,
1292                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1293                         return "&$name;";
1294                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1295                         return '&#' . self::$htmlEntities[$name] . ';';
1296                 } else {
1297                         return "&amp;$name;";
1298                 }
1299         }
1300
1301         /**
1302          * @param $codepoint
1303          * @return null|string
1304          */
1305         static function decCharReference( $codepoint ) {
1306                 $point = intval( $codepoint );
1307                 if ( Sanitizer::validateCodepoint( $point ) ) {
1308                         return sprintf( '&#%d;', $point );
1309                 } else {
1310                         return null;
1311                 }
1312         }
1313
1314         /**
1315          * @param $codepoint
1316          * @return null|string
1317          */
1318         static function hexCharReference( $codepoint ) {
1319                 $point = hexdec( $codepoint );
1320                 if ( Sanitizer::validateCodepoint( $point ) ) {
1321                         return sprintf( '&#x%x;', $point );
1322                 } else {
1323                         return null;
1324                 }
1325         }
1326
1327         /**
1328          * Returns true if a given Unicode codepoint is a valid character in XML.
1329          * @param $codepoint Integer
1330          * @return Boolean
1331          */
1332         private static function validateCodepoint( $codepoint ) {
1333                 return $codepoint == 0x09
1334                         || $codepoint == 0x0a
1335                         || $codepoint == 0x0d
1336                         || ( $codepoint >= 0x20 && $codepoint <= 0xd7ff )
1337                         || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1338                         || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1339         }
1340
1341         /**
1342          * Decode any character references, numeric or named entities,
1343          * in the text and return a UTF-8 string.
1344          *
1345          * @param $text String
1346          * @return String
1347          */
1348         public static function decodeCharReferences( $text ) {
1349                 return preg_replace_callback(
1350                         self::CHAR_REFS_REGEX,
1351                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1352                         $text );
1353         }
1354
1355         /**
1356          * Decode any character references, numeric or named entities,
1357          * in the next and normalize the resulting string. (bug 14952)
1358          *
1359          * This is useful for page titles, not for text to be displayed,
1360          * MediaWiki allows HTML entities to escape normalization as a feature.
1361          *
1362          * @param string $text (already normalized, containing entities)
1363          * @return String (still normalized, without entities)
1364          */
1365         public static function decodeCharReferencesAndNormalize( $text ) {
1366                 global $wgContLang;
1367                 $text = preg_replace_callback(
1368                         self::CHAR_REFS_REGEX,
1369                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1370                         $text, /* limit */ -1, $count );
1371
1372                 if ( $count ) {
1373                         return $wgContLang->normalize( $text );
1374                 } else {
1375                         return $text;
1376                 }
1377         }
1378
1379         /**
1380          * @param $matches String
1381          * @return String
1382          */
1383         static function decodeCharReferencesCallback( $matches ) {
1384                 if ( $matches[1] != '' ) {
1385                         return Sanitizer::decodeEntity( $matches[1] );
1386                 } elseif ( $matches[2] != '' ) {
1387                         return Sanitizer::decodeChar( intval( $matches[2] ) );
1388                 } elseif ( $matches[3] != '' ) {
1389                         return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1390                 }
1391                 # Last case should be an ampersand by itself
1392                 return $matches[0];
1393         }
1394
1395         /**
1396          * Return UTF-8 string for a codepoint if that is a valid
1397          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1398          * @param $codepoint Integer
1399          * @return String
1400          * @private
1401          */
1402         static function decodeChar( $codepoint ) {
1403                 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1404                         return codepointToUtf8( $codepoint );
1405                 } else {
1406                         return UTF8_REPLACEMENT;
1407                 }
1408         }
1409
1410         /**
1411          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1412          * return the UTF-8 encoding of that character. Otherwise, returns
1413          * pseudo-entity source (eg "&foo;")
1414          *
1415          * @param $name String
1416          * @return String
1417          */
1418         static function decodeEntity( $name ) {
1419                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1420                         $name = self::$htmlEntityAliases[$name];
1421                 }
1422                 if ( isset( self::$htmlEntities[$name] ) ) {
1423                         return codepointToUtf8( self::$htmlEntities[$name] );
1424                 } else {
1425                         return "&$name;";
1426                 }
1427         }
1428
1429         /**
1430          * Fetch the whitelist of acceptable attributes for a given element name.
1431          *
1432          * @param $element String
1433          * @return Array
1434          */
1435         static function attributeWhitelist( $element ) {
1436                 $list = Sanitizer::setupAttributeWhitelist();
1437                 return isset( $list[$element] )
1438                         ? $list[$element]
1439                         : array();
1440         }
1441
1442         /**
1443          * Foreach array key (an allowed HTML element), return an array
1444          * of allowed attributes
1445          * @return Array
1446          */
1447         static function setupAttributeWhitelist() {
1448                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
1449
1450                 static $whitelist, $staticInitialised;
1451                 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgAllowMicrodataAttributes' ) );
1452
1453                 if ( isset( $whitelist ) && $staticInitialised == $globalContext ) {
1454                         return $whitelist;
1455                 }
1456
1457                 $common = array(
1458                         # HTML
1459                         'id',
1460                         'class',
1461                         'style',
1462                         'lang',
1463                         'dir',
1464                         'title',
1465
1466                         # WAI-ARIA
1467                         'role',
1468                 );
1469
1470                 if ( $wgAllowRdfaAttributes ) {
1471                         # RDFa attributes as specified in section 9 of
1472                         # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1473                         $common = array_merge( $common, array(
1474                                 'about', 'property', 'resource', 'datatype', 'typeof',
1475                         ) );
1476                 }
1477
1478                 if ( $wgAllowMicrodataAttributes ) {
1479                         # add HTML5 microdata tags as specified by
1480                         # http://www.whatwg.org/html/microdata.html#the-microdata-model
1481                         $common = array_merge( $common, array(
1482                                 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1483                         ) );
1484                 }
1485
1486                 $block = array_merge( $common, array( 'align' ) );
1487                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1488                 $tablecell = array(
1489                         'abbr',
1490                         'axis',
1491                         'headers',
1492                         'scope',
1493                         'rowspan',
1494                         'colspan',
1495                         'nowrap', # deprecated
1496                         'width', # deprecated
1497                         'height', # deprecated
1498                         'bgcolor', # deprecated
1499                 );
1500
1501                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1502                 # See: http://www.w3.org/TR/html4/
1503                 $whitelist = array(
1504                         # 7.5.4
1505                         'div'        => $block,
1506                         'center'     => $common, # deprecated
1507                         'span'       => $block, # ??
1508
1509                         # 7.5.5
1510                         'h1'         => $block,
1511                         'h2'         => $block,
1512                         'h3'         => $block,
1513                         'h4'         => $block,
1514                         'h5'         => $block,
1515                         'h6'         => $block,
1516
1517                         # 7.5.6
1518                         # address
1519
1520                         # 8.2.4
1521                         # bdo
1522
1523                         # 9.2.1
1524                         'em'         => $common,
1525                         'strong'     => $common,
1526                         'cite'       => $common,
1527                         'dfn'        => $common,
1528                         'code'       => $common,
1529                         'samp'       => $common,
1530                         'kbd'        => $common,
1531                         'var'        => $common,
1532                         'abbr'       => $common,
1533                         # acronym
1534
1535                         # 9.2.2
1536                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1537                         # q
1538
1539                         # 9.2.3
1540                         'sub'        => $common,
1541                         'sup'        => $common,
1542
1543                         # 9.3.1
1544                         'p'          => $block,
1545
1546                         # 9.3.2
1547                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1548
1549                         # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1550                         'wbr'        => array( 'id', 'class', 'title', 'style' ),
1551
1552                         # 9.3.4
1553                         'pre'        => array_merge( $common, array( 'width' ) ),
1554
1555                         # 9.4
1556                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1557                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1558
1559                         # 10.2
1560                         'ul'         => array_merge( $common, array( 'type' ) ),
1561                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1562                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1563
1564                         # 10.3
1565                         'dl'         => $common,
1566                         'dd'         => $common,
1567                         'dt'         => $common,
1568
1569                         # 11.2.1
1570                         'table'      => array_merge( $common,
1571                                                                 array( 'summary', 'width', 'border', 'frame',
1572                                                                                 'rules', 'cellspacing', 'cellpadding',
1573                                                                                 'align', 'bgcolor',
1574                                                                 ) ),
1575
1576                         # 11.2.2
1577                         'caption'    => array_merge( $common, array( 'align' ) ),
1578
1579                         # 11.2.3
1580                         'thead'      => array_merge( $common, $tablealign ),
1581                         'tfoot'      => array_merge( $common, $tablealign ),
1582                         'tbody'      => array_merge( $common, $tablealign ),
1583
1584                         # 11.2.4
1585                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1586                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1587
1588                         # 11.2.5
1589                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1590
1591                         # 11.2.6
1592                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1593                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1594
1595                         # 12.2
1596                         # NOTE: <a> is not allowed directly, but the attrib
1597                         # whitelist is used from the Parser object
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                         # HTML Ruby annotation text module, simple ruby only.
1624                         # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
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                         # HTML 5 section 4.6
1638                         'bdi' => $common,
1639
1640                         # HTML5 elements, defined by:
1641                         # http://www.whatwg.org/html/
1642                         'data' => array_merge( $common, array( 'value' ) ),
1643                         'time' => array_merge( $common, array( 'datetime' ) ),
1644                         'mark' => $common,
1645
1646                         // meta and link are only permitted by removeHTMLtags when Microdata
1647                         // is enabled so we don't bother adding a conditional to hide these
1648                         // Also meta and link are only valid in WikiText as Microdata elements
1649                         // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1650                         // So we don't bother including $common attributes that have no purpose.
1651                         'meta' => array( 'itemprop', 'content' ),
1652                         'link' => array( 'itemprop', 'href' ),
1653                 );
1654
1655                 $staticInitialised = $globalContext;
1656
1657                 return $whitelist;
1658         }
1659
1660         /**
1661          * Take a fragment of (potentially invalid) HTML and return
1662          * a version with any tags removed, encoded as plain text.
1663          *
1664          * Warning: this return value must be further escaped for literal
1665          * inclusion in HTML output as of 1.10!
1666          *
1667          * @param string $text HTML fragment
1668          * @return String
1669          */
1670         static function stripAllTags( $text ) {
1671                 # Actual <tags>
1672                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1673
1674                 # Normalize &entities and whitespace
1675                 $text = self::decodeCharReferences( $text );
1676                 $text = self::normalizeWhitespace( $text );
1677
1678                 return $text;
1679         }
1680
1681         /**
1682          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1683          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1684          * PHP 5.1 doesn't.
1685          *
1686          * Use for passing XHTML fragments to PHP's XML parsing functions
1687          *
1688          * @return String
1689          */
1690         static function hackDocType() {
1691                 $out = "<!DOCTYPE html [\n";
1692                 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1693                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1694                 }
1695                 $out .= "]>\n";
1696                 return $out;
1697         }
1698
1699         /**
1700          * @param $url string
1701          * @return mixed|string
1702          */
1703         static function cleanUrl( $url ) {
1704                 # Normalize any HTML entities in input. They will be
1705                 # re-escaped by makeExternalLink().
1706                 $url = Sanitizer::decodeCharReferences( $url );
1707
1708                 # Escape any control characters introduced by the above step
1709                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1710                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1711
1712                 # Validate hostname portion
1713                 $matches = array();
1714                 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1715                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1716
1717                         // Characters that will be ignored in IDNs.
1718                         // http://tools.ietf.org/html/3454#section-3.1
1719                         // Strip them before further processing so blacklists and such work.
1720                         $strip = "/
1721                                 \\s|          # general whitespace
1722                                 \xc2\xad|     # 00ad SOFT HYPHEN
1723                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1724                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1725                                 \xe2\x81\xa0| # 2060 WORD JOINER
1726                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1727                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1728                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1729                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1730                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1731                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1732                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1733                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1734                                 /xuD";
1735
1736                         $host = preg_replace( $strip, '', $host );
1737
1738                         // @todo FIXME: Validate hostnames here
1739
1740                         return $protocol . $host . $rest;
1741                 } else {
1742                         return $url;
1743                 }
1744         }
1745
1746         /**
1747          * @param $matches array
1748          * @return string
1749          */
1750         static function cleanUrlCallback( $matches ) {
1751                 return urlencode( $matches[0] );
1752         }
1753
1754         /**
1755          * Does a string look like an e-mail address?
1756          *
1757          * This validates an email address using an HTML5 specification found at:
1758          * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1759          * Which as of 2011-01-24 says:
1760          *
1761          *   A valid e-mail address is a string that matches the ABNF production
1762          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1763          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1764          *   3.5.
1765          *
1766          * This function is an implementation of the specification as requested in
1767          * bug 22449.
1768          *
1769          * Client-side forms will use the same standard validation rules via JS or
1770          * HTML 5 validation; additional restrictions can be enforced server-side
1771          * by extensions via the 'isValidEmailAddr' hook.
1772          *
1773          * Note that this validation doesn't 100% match RFC 2822, but is believed
1774          * to be liberal enough for wide use. Some invalid addresses will still
1775          * pass validation here.
1776          *
1777          * @since 1.18
1778          *
1779          * @param string $addr E-mail address
1780          * @return Bool
1781          */
1782         public static function validateEmail( $addr ) {
1783                 $result = null;
1784                 if ( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1785                         return $result;
1786                 }
1787
1788                 // Please note strings below are enclosed in brackets [], this make the
1789                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1790                 // See bug 26948
1791                 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1792                 $rfc1034_ldh_str = "a-z0-9\\-";
1793
1794                 $HTML5_email_regexp = "/
1795                 ^                      # start of string
1796                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1797                 @                      # 'apostrophe'
1798                 [$rfc1034_ldh_str]+       # First domain part
1799                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1800                 $                      # End of string
1801                 /ix"; // case Insensitive, eXtended
1802
1803                 return (bool)preg_match( $HTML5_email_regexp, $addr );
1804         }
1805 }