includes/Sanitizer.php

   1 <?php
   2 /**
   3  * HTML sanitizer for %MediaWiki.
   4  *
   5  * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
   6  * https://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          * Acceptable tag name charset from HTML5 parsing spec
  44          * http://www.w3.org/TR/html5/syntax.html#tag-open-state
  45          */
  46         const ELEMENT_BITS_REGEX = '!^(/?)([A-Za-z][^\t\n\v />\0]*+)([^>]*?)(/?>)([^<]*)$!';
  47
  48         /**
  49          * Blacklist for evil uris like javascript:
  50          * WARNING: DO NOT use this in any place that actually requires blacklisting
  51          * for security reasons. There are NUMEROUS[1] ways to bypass blacklisting, the
  52          * only way to be secure from javascript: uri based xss vectors is to whitelist
  53          * things that you know are safe and deny everything else.
  54          * [1]: http://ha.ckers.org/xss.html
  55          */
  56         const EVIL_URI_PATTERN = '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
  57         const XMLNS_ATTRIBUTE_PATTERN = "/^xmlns:[:A-Z_a-z-.0-9]+$/";
  58
  59         /**
  60          * List of all named character entities defined in HTML 4.01
  61          * http://www.w3.org/TR/html4/sgml/entities.html
  62          * As well as &apos; which is only defined starting in XHTML1.
  63          */
  64         private static $htmlEntities = array(
  65                 'Aacute'   => 193,
  66                 'aacute'   => 225,
  67                 'Acirc'    => 194,
  68                 'acirc'    => 226,
  69                 'acute'    => 180,
  70                 'AElig'    => 198,
  71                 'aelig'    => 230,
  72                 'Agrave'   => 192,
  73                 'agrave'   => 224,
  74                 'alefsym'  => 8501,
  75                 'Alpha'    => 913,
  76                 'alpha'    => 945,
  77                 'amp'      => 38,
  78                 'and'      => 8743,
  79                 'ang'      => 8736,
  80                 'apos'     => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
  81                 'Aring'    => 197,
  82                 'aring'    => 229,
  83                 'asymp'    => 8776,
  84                 'Atilde'   => 195,
  85                 'atilde'   => 227,
  86                 'Auml'     => 196,
  87                 'auml'     => 228,
  88                 'bdquo'    => 8222,
  89                 'Beta'     => 914,
  90                 'beta'     => 946,
  91                 'brvbar'   => 166,
  92                 'bull'     => 8226,
  93                 'cap'      => 8745,
  94                 'Ccedil'   => 199,
  95                 'ccedil'   => 231,
  96                 'cedil'    => 184,
  97                 'cent'     => 162,
  98                 'Chi'      => 935,
  99                 'chi'      => 967,
 100                 'circ'     => 710,
 101                 'clubs'    => 9827,
 102                 'cong'     => 8773,
 103                 'copy'     => 169,
 104                 'crarr'    => 8629,
 105                 'cup'      => 8746,
 106                 'curren'   => 164,
 107                 'dagger'   => 8224,
 108                 'Dagger'   => 8225,
 109                 'darr'     => 8595,
 110                 'dArr'     => 8659,
 111                 'deg'      => 176,
 112                 'Delta'    => 916,
 113                 'delta'    => 948,
 114                 'diams'    => 9830,
 115                 'divide'   => 247,
 116                 'Eacute'   => 201,
 117                 'eacute'   => 233,
 118                 'Ecirc'    => 202,
 119                 'ecirc'    => 234,
 120                 'Egrave'   => 200,
 121                 'egrave'   => 232,
 122                 'empty'    => 8709,
 123                 'emsp'     => 8195,
 124                 'ensp'     => 8194,
 125                 'Epsilon'  => 917,
 126                 'epsilon'  => 949,
 127                 'equiv'    => 8801,
 128                 'Eta'      => 919,
 129                 'eta'      => 951,
 130                 'ETH'      => 208,
 131                 'eth'      => 240,
 132                 'Euml'     => 203,
 133                 'euml'     => 235,
 134                 'euro'     => 8364,
 135                 'exist'    => 8707,
 136                 'fnof'     => 402,
 137                 'forall'   => 8704,
 138                 'frac12'   => 189,
 139                 'frac14'   => 188,
 140                 'frac34'   => 190,
 141                 'frasl'    => 8260,
 142                 'Gamma'    => 915,
 143                 'gamma'    => 947,
 144                 'ge'       => 8805,
 145                 'gt'       => 62,
 146                 'harr'     => 8596,
 147                 'hArr'     => 8660,
 148                 'hearts'   => 9829,
 149                 'hellip'   => 8230,
 150                 'Iacute'   => 205,
 151                 'iacute'   => 237,
 152                 'Icirc'    => 206,
 153                 'icirc'    => 238,
 154                 'iexcl'    => 161,
 155                 'Igrave'   => 204,
 156                 'igrave'   => 236,
 157                 'image'    => 8465,
 158                 'infin'    => 8734,
 159                 'int'      => 8747,
 160                 'Iota'     => 921,
 161                 'iota'     => 953,
 162                 'iquest'   => 191,
 163                 'isin'     => 8712,
 164                 'Iuml'     => 207,
 165                 'iuml'     => 239,
 166                 'Kappa'    => 922,
 167                 'kappa'    => 954,
 168                 'Lambda'   => 923,
 169                 'lambda'   => 955,
 170                 'lang'     => 9001,
 171                 'laquo'    => 171,
 172                 'larr'     => 8592,
 173                 'lArr'     => 8656,
 174                 'lceil'    => 8968,
 175                 'ldquo'    => 8220,
 176                 'le'       => 8804,
 177                 'lfloor'   => 8970,
 178                 'lowast'   => 8727,
 179                 'loz'      => 9674,
 180                 'lrm'      => 8206,
 181                 'lsaquo'   => 8249,
 182                 'lsquo'    => 8216,
 183                 'lt'       => 60,
 184                 'macr'     => 175,
 185                 'mdash'    => 8212,
 186                 'micro'    => 181,
 187                 'middot'   => 183,
 188                 'minus'    => 8722,
 189                 'Mu'       => 924,
 190                 'mu'       => 956,
 191                 'nabla'    => 8711,
 192                 'nbsp'     => 160,
 193                 'ndash'    => 8211,
 194                 'ne'       => 8800,
 195                 'ni'       => 8715,
 196                 'not'      => 172,
 197                 'notin'    => 8713,
 198                 'nsub'     => 8836,
 199                 'Ntilde'   => 209,
 200                 'ntilde'   => 241,
 201                 'Nu'       => 925,
 202                 'nu'       => 957,
 203                 'Oacute'   => 211,
 204                 'oacute'   => 243,
 205                 'Ocirc'    => 212,
 206                 'ocirc'    => 244,
 207                 'OElig'    => 338,
 208                 'oelig'    => 339,
 209                 'Ograve'   => 210,
 210                 'ograve'   => 242,
 211                 'oline'    => 8254,
 212                 'Omega'    => 937,
 213                 'omega'    => 969,
 214                 'Omicron'  => 927,
 215                 'omicron'  => 959,
 216                 'oplus'    => 8853,
 217                 'or'       => 8744,
 218                 'ordf'     => 170,
 219                 'ordm'     => 186,
 220                 'Oslash'   => 216,
 221                 'oslash'   => 248,
 222                 'Otilde'   => 213,
 223                 'otilde'   => 245,
 224                 'otimes'   => 8855,
 225                 'Ouml'     => 214,
 226                 'ouml'     => 246,
 227                 'para'     => 182,
 228                 'part'     => 8706,
 229                 'permil'   => 8240,
 230                 'perp'     => 8869,
 231                 'Phi'      => 934,
 232                 'phi'      => 966,
 233                 'Pi'       => 928,
 234                 'pi'       => 960,
 235                 'piv'      => 982,
 236                 'plusmn'   => 177,
 237                 'pound'    => 163,
 238                 'prime'    => 8242,
 239                 'Prime'    => 8243,
 240                 'prod'     => 8719,
 241                 'prop'     => 8733,
 242                 'Psi'      => 936,
 243                 'psi'      => 968,
 244                 'quot'     => 34,
 245                 'radic'    => 8730,
 246                 'rang'     => 9002,
 247                 'raquo'    => 187,
 248                 'rarr'     => 8594,
 249                 'rArr'     => 8658,
 250                 'rceil'    => 8969,
 251                 'rdquo'    => 8221,
 252                 'real'     => 8476,
 253                 'reg'      => 174,
 254                 'rfloor'   => 8971,
 255                 'Rho'      => 929,
 256                 'rho'      => 961,
 257                 'rlm'      => 8207,
 258                 'rsaquo'   => 8250,
 259                 'rsquo'    => 8217,
 260                 'sbquo'    => 8218,
 261                 'Scaron'   => 352,
 262                 'scaron'   => 353,
 263                 'sdot'     => 8901,
 264                 'sect'     => 167,
 265                 'shy'      => 173,
 266                 'Sigma'    => 931,
 267                 'sigma'    => 963,
 268                 'sigmaf'   => 962,
 269                 'sim'      => 8764,
 270                 'spades'   => 9824,
 271                 'sub'      => 8834,
 272                 'sube'     => 8838,
 273                 'sum'      => 8721,
 274                 'sup'      => 8835,
 275                 'sup1'     => 185,
 276                 'sup2'     => 178,
 277                 'sup3'     => 179,
 278                 'supe'     => 8839,
 279                 'szlig'    => 223,
 280                 'Tau'      => 932,
 281                 'tau'      => 964,
 282                 'there4'   => 8756,
 283                 'Theta'    => 920,
 284                 'theta'    => 952,
 285                 'thetasym' => 977,
 286                 'thinsp'   => 8201,
 287                 'THORN'    => 222,
 288                 'thorn'    => 254,
 289                 'tilde'    => 732,
 290                 'times'    => 215,
 291                 'trade'    => 8482,
 292                 'Uacute'   => 218,
 293                 'uacute'   => 250,
 294                 'uarr'     => 8593,
 295                 'uArr'     => 8657,
 296                 'Ucirc'    => 219,
 297                 'ucirc'    => 251,
 298                 'Ugrave'   => 217,
 299                 'ugrave'   => 249,
 300                 'uml'      => 168,
 301                 'upsih'    => 978,
 302                 'Upsilon'  => 933,
 303                 'upsilon'  => 965,
 304                 'Uuml'     => 220,
 305                 'uuml'     => 252,
 306                 'weierp'   => 8472,
 307                 'Xi'       => 926,
 308                 'xi'       => 958,
 309                 'Yacute'   => 221,
 310                 'yacute'   => 253,
 311                 'yen'      => 165,
 312                 'Yuml'     => 376,
 313                 'yuml'     => 255,
 314                 'Zeta'     => 918,
 315                 'zeta'     => 950,
 316                 'zwj'      => 8205,
 317                 'zwnj'     => 8204
 318         );
 319
 320         /**
 321          * Character entity aliases accepted by MediaWiki
 322          */
 323         private static $htmlEntityAliases = array(
 324                 'רלמ' => 'rlm',
 325                 'رلم' => 'rlm',
 326         );
 327
 328         /**
 329          * Lazy-initialised attributes regex, see getAttribsRegex()
 330          */
 331         private static $attribsRegex;
 332
 333         /**
 334          * Regular expression to match HTML/XML attribute pairs within a tag.
 335          * Allows some... latitude.
 336          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 337          * @return string
 338          */
 339         static function getAttribsRegex() {
 340                 if ( self::$attribsRegex === null ) {
 341                         $attribFirst = '[:A-Z_a-z0-9]';
 342                         $attrib = '[:A-Z_a-z-.0-9]';
 343                         $space = '[\x09\x0a\x0d\x20]';
 344                         self::$attribsRegex =
 345                                 "/(?:^|$space)({$attribFirst}{$attrib}*)
 346                                   ($space*=$space*
 347                                         (?:
 348                                          # The attribute value: quoted or alone
 349                                           \"([^<\"]*)(?:\"|\$)
 350                                          | '([^<']*)(?:'|\$)
 351                                          |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
 352                                         )
 353                                 )?(?=$space|\$)/sx";
 354                 }
 355                 return self::$attribsRegex;
 356         }
 357
 358         /**
 359          * Return the various lists of recognized tags
 360          * @param array $extratags For any extra tags to include
 361          * @param array $removetags For any tags (default or extra) to exclude
 362          * @return array
 363          */
 364         public static function getRecognizedTagData( $extratags = array(), $removetags = array() ) {
 365                 global $wgAllowMicrodataAttributes, $wgAllowImageTag;
 366
 367                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 368                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 369
 370                 // Base our staticInitialised variable off of the global config state so that if the globals
 371                 // are changed (like in the screwed up test system) we will re-initialise the settings.
 372                 $globalContext = implode( '-', compact( 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
 373                 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
 374                         $htmlpairsStatic = array( # Tags that must be closed
 375                                 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 376                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 377                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 378                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 379                                 'ruby', 'rb', 'rp', 'rt', 'rtc', 'p', 'span', 'abbr', 'dfn',
 380                                 'kbd', 'samp', 'data', 'time', 'mark'
 381                         );
 382                         $htmlsingle = array(
 383                                 'br', 'wbr', 'hr', 'li', 'dt', 'dd'
 384                         );
 385                         $htmlsingleonly = array( # Elements that cannot have close tags
 386                                 'br', 'wbr', 'hr'
 387                         );
 388                         if ( $wgAllowMicrodataAttributes ) {
 389                                 $htmlsingle[] = $htmlsingleonly[] = 'meta';
 390                                 $htmlsingle[] = $htmlsingleonly[] = 'link';
 391                         }
 392                         $htmlnest = array( # Tags that can be nested--??
 393                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 394                                 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
 395                                 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
 396                         );
 397                         $tabletags = array( # Can only appear inside table, we will close them
 398                                 'td', 'th', 'tr',
 399                         );
 400                         $htmllist = array( # Tags used by list
 401                                 'ul', 'ol',
 402                         );
 403                         $listtags = array( # Tags that can appear in a list
 404                                 'li',
 405                         );
 406
 407                         if ( $wgAllowImageTag ) {
 408                                 $htmlsingle[] = 'img';
 409                                 $htmlsingleonly[] = 'img';
 410                         }
 411
 412                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 413                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 414
 415                         # Convert them all to hashtables for faster lookup
 416                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 417                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 418                         foreach ( $vars as $var ) {
 419                                 $$var = array_flip( $$var );
 420                         }
 421                         $staticInitialised = $globalContext;
 422                 }
 423
 424                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 425                 $extratags = array_flip( $extratags );
 426                 $removetags = array_flip( $removetags );
 427                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 428                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
 429
 430                 return array(
 431                         'htmlpairs' => $htmlpairs,
 432                         'htmlsingle' => $htmlsingle,
 433                         'htmlsingleonly' => $htmlsingleonly,
 434                         'htmlnest' => $htmlnest,
 435                         'tabletags' => $tabletags,
 436                         'htmllist' => $htmllist,
 437                         'listtags' => $listtags,
 438                         'htmlsingleallowed' => $htmlsingleallowed,
 439                         'htmlelements' => $htmlelements,
 440                 );
 441         }
 442
 443         /**
 444          * Cleans up HTML, removes dangerous tags and attributes, and
 445          * removes HTML comments
 446          * @param string $text
 447          * @param callable $processCallback Callback to do any variable or parameter
 448          *   replacements in HTML attribute values
 449          * @param array|bool $args Arguments for the processing callback
 450          * @param array $extratags For any extra tags to include
 451          * @param array $removetags For any tags (default or extra) to exclude
 452          * @return string
 453          */
 454         public static function removeHTMLtags( $text, $processCallback = null,
 455                 $args = array(), $extratags = array(), $removetags = array()
 456         ) {
 457                 extract( self::getRecognizedTagData( $extratags, $removetags ) );
 458
 459                 # Remove HTML comments
 460                 $text = Sanitizer::removeHTMLcomments( $text );
 461                 $bits = explode( '<', $text );
 462                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 463                 if ( !MWTidy::isEnabled() ) {
 464                         $tagstack = $tablestack = array();
 465                         foreach ( $bits as $x ) {
 466                                 $regs = array();
 467                                 # $slash: Does the current element start with a '/'?
 468                                 # $t: Current element name
 469                                 # $params: String between element name and >
 470                                 # $brace: Ending '>' or '/>'
 471                                 # $rest: Everything until the next element of $bits
 472                                 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 473                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 474                                 } else {
 475                                         $slash = $t = $params = $brace = $rest = null;
 476                                 }
 477
 478                                 $badtag = false;
 479                                 $t = strtolower( $t );
 480                                 if ( isset( $htmlelements[$t] ) ) {
 481                                         # Check our stack
 482                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 483                                                 $badtag = true;
 484                                         } elseif ( $slash ) {
 485                                                 # Closing a tag... is it the one we just opened?
 486                                                 MediaWiki\suppressWarnings();
 487                                                 $ot = array_pop( $tagstack );
 488                                                 MediaWiki\restoreWarnings();
 489
 490                                                 if ( $ot != $t ) {
 491                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 492                                                                 # Pop all elements with an optional close tag
 493                                                                 # and see if we find a match below them
 494                                                                 $optstack = array();
 495                                                                 array_push( $optstack, $ot );
 496                                                                 MediaWiki\suppressWarnings();
 497                                                                 $ot = array_pop( $tagstack );
 498                                                                 MediaWiki\restoreWarnings();
 499                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 500                                                                         array_push( $optstack, $ot );
 501                                                                         MediaWiki\suppressWarnings();
 502                                                                         $ot = array_pop( $tagstack );
 503                                                                         MediaWiki\restoreWarnings();
 504                                                                 }
 505                                                                 if ( $t != $ot ) {
 506                                                                         # No match. Push the optional elements back again
 507                                                                         $badtag = true;
 508                                                                         MediaWiki\suppressWarnings();
 509                                                                         $ot = array_pop( $optstack );
 510                                                                         MediaWiki\restoreWarnings();
 511                                                                         while ( $ot ) {
 512                                                                                 array_push( $tagstack, $ot );
 513                                                                                 MediaWiki\suppressWarnings();
 514                                                                                 $ot = array_pop( $optstack );
 515                                                                                 MediaWiki\restoreWarnings();
 516                                                                         }
 517                                                                 }
 518                                                         } else {
 519                                                                 MediaWiki\suppressWarnings();
 520                                                                 array_push( $tagstack, $ot );
 521                                                                 MediaWiki\restoreWarnings();
 522
 523                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 524                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 525                                                                         $badtag = true;
 526                                                                 }
 527                                                         }
 528                                                 } else {
 529                                                         if ( $t == 'table' ) {
 530                                                                 $tagstack = array_pop( $tablestack );
 531                                                         }
 532                                                 }
 533                                                 $newparams = '';
 534                                         } else {
 535                                                 # Keep track for later
 536                                                 if ( isset( $tabletags[$t] ) && !in_array( 'table', $tagstack ) ) {
 537                                                         $badtag = true;
 538                                                 } elseif ( in_array( $t, $tagstack ) && !isset( $htmlnest[$t] ) ) {
 539                                                         $badtag = true;
 540                                                 #  Is it a self closed htmlpair ? (bug 5487)
 541                                                 } elseif ( $brace == '/>' && isset( $htmlpairs[$t] ) ) {
 542                                                         $badtag = true;
 543                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 544                                                         # Hack to force empty tag for unclosable elements
 545                                                         $brace = '/>';
 546                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 547                                                         # Hack to not close $htmlsingle tags
 548                                                         $brace = null;
 549                                                         # Still need to push this optionally-closed tag to
 550                                                         # the tag stack so that we can match end tags
 551                                                         # instead of marking them as bad.
 552                                                         array_push( $tagstack, $t );
 553                                                 } elseif ( isset( $tabletags[$t] ) && in_array( $t, $tagstack ) ) {
 554                                                         // New table tag but forgot to close the previous one
 555                                                         $text .= "</$t>";
 556                                                 } else {
 557                                                         if ( $t == 'table' ) {
 558                                                                 array_push( $tablestack, $tagstack );
 559                                                                 $tagstack = array();
 560                                                         }
 561                                                         array_push( $tagstack, $t );
 562                                                 }
 563
 564                                                 # Replace any variables or template parameters with
 565                                                 # plaintext results.
 566                                                 if ( is_callable( $processCallback ) ) {
 567                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 568                                                 }
 569
 570                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 571                                                         $badtag = true;
 572                                                 }
 573
 574                                                 # Strip non-approved attributes from the tag
 575                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 576                                         }
 577                                         if ( !$badtag ) {
 578                                                 $rest = str_replace( '>', '&gt;', $rest );
 579                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 580                                                 $text .= "<$slash$t$newparams$close>$rest";
 581                                                 continue;
 582                                         }
 583                                 }
 584                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 585                         }
 586                         # Close off any remaining tags
 587                         while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
 588                                 $text .= "</$t>\n";
 589                                 if ( $t == 'table' ) {
 590                                         $tagstack = array_pop( $tablestack );
 591                                 }
 592                         }
 593                 } else {
 594                         # this might be possible using tidy itself
 595                         foreach ( $bits as $x ) {
 596                                 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 597                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 598
 599                                         $badtag = false;
 600                                         $t = strtolower( $t );
 601                                         if ( isset( $htmlelements[$t] ) ) {
 602                                                 if ( is_callable( $processCallback ) ) {
 603                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 604                                                 }
 605
 606                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 607                                                         $badtag = true;
 608                                                 }
 609
 610                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 611                                                 if ( !$badtag ) {
 612                                                         $rest = str_replace( '>', '&gt;', $rest );
 613                                                         $text .= "<$slash$t$newparams$brace$rest";
 614                                                         continue;
 615                                                 }
 616                                         }
 617                                 }
 618                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 619                         }
 620                 }
 621                 return $text;
 622         }
 623
 624         /**
 625          * Remove '<!--', '-->', and everything between.
 626          * To avoid leaving blank lines, when a comment is both preceded
 627          * and followed by a newline (ignoring spaces), trim leading and
 628          * trailing spaces and one of the newlines.
 629          *
 630          * @param string $text
 631          * @return string
 632          */
 633         public static function removeHTMLcomments( $text ) {
 634                 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
 635                         $end = strpos( $text, '-->', $start + 4 );
 636                         if ( $end === false ) {
 637                                 # Unterminated comment; bail out
 638                                 break;
 639                         }
 640
 641                         $end += 3;
 642
 643                         # Trim space and newline if the comment is both
 644                         # preceded and followed by a newline
 645                         $spaceStart = max( $start - 1, 0 );
 646                         $spaceLen = $end - $spaceStart;
 647                         while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
 648                                 $spaceStart--;
 649                                 $spaceLen++;
 650                         }
 651                         while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
 652                                 $spaceLen++;
 653                         }
 654                         if ( substr( $text, $spaceStart, 1 ) === "\n"
 655                                 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
 656                                 # Remove the comment, leading and trailing
 657                                 # spaces, and leave only one newline.
 658                                 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
 659                         } else {
 660                                 # Remove just the comment.
 661                                 $text = substr_replace( $text, '', $start, $end - $start );
 662                         }
 663                 }
 664                 return $text;
 665         }
 666
 667         /**
 668          * Takes attribute names and values for a tag and the tag name and
 669          * validates that the tag is allowed to be present.
 670          * This DOES NOT validate the attributes, nor does it validate the
 671          * tags themselves. This method only handles the special circumstances
 672          * where we may want to allow a tag within content but ONLY when it has
 673          * specific attributes set.
 674          *
 675          * @param string $params
 676          * @param string $element
 677          * @return bool
 678          */
 679         static function validateTag( $params, $element ) {
 680                 $params = Sanitizer::decodeTagAttributes( $params );
 681
 682                 if ( $element == 'meta' || $element == 'link' ) {
 683                         if ( !isset( $params['itemprop'] ) ) {
 684                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 685                                 return false;
 686                         }
 687                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 688                                 // <meta> must have a content="" for the itemprop
 689                                 return false;
 690                         }
 691                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 692                                 // <link> must have an associated href=""
 693                                 return false;
 694                         }
 695                 }
 696
 697                 return true;
 698         }
 699
 700         /**
 701          * Take an array of attribute names and values and normalize or discard
 702          * illegal values for the given element type.
 703          *
 704          * - Discards attributes not on a whitelist for the given element
 705          * - Unsafe style attributes are discarded
 706          * - Invalid id attributes are re-encoded
 707          *
 708          * @param array $attribs
 709          * @param string $element
 710          * @return array
 711          *
 712          * @todo Check for legal values where the DTD limits things.
 713          * @todo Check for unique id attribute :P
 714          */
 715         static function validateTagAttributes( $attribs, $element ) {
 716                 return Sanitizer::validateAttributes( $attribs,
 717                         Sanitizer::attributeWhitelist( $element ) );
 718         }
 719
 720         /**
 721          * Take an array of attribute names and values and normalize or discard
 722          * illegal values for the given whitelist.
 723          *
 724          * - Discards attributes not the given whitelist
 725          * - Unsafe style attributes are discarded
 726          * - Invalid id attributes are re-encoded
 727          *
 728          * @param array $attribs
 729          * @param array $whitelist List of allowed attribute names
 730          * @return array
 731          *
 732          * @todo Check for legal values where the DTD limits things.
 733          * @todo Check for unique id attribute :P
 734          */
 735         static function validateAttributes( $attribs, $whitelist ) {
 736                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
 737
 738                 $whitelist = array_flip( $whitelist );
 739                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 740
 741                 $out = array();
 742                 foreach ( $attribs as $attribute => $value ) {
 743                         # allow XML namespace declaration if RDFa is enabled
 744                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 745                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 746                                         $out[$attribute] = $value;
 747                                 }
 748
 749                                 continue;
 750                         }
 751
 752                         # Allow any attribute beginning with "data-"
 753                         # However:
 754                         # * data-ooui is reserved for ooui
 755                         # * data-mw and data-parsoid are reserved for parsoid
 756                         # * data-mw-<ext name here> is reserved for extensions (or core) if
 757                         #   they need to communicate some data to the client and want to be
 758                         #   sure that it isn't coming from an untrusted user.
 759                         if ( !preg_match( '/^data-(?!ooui|mw|parsoid)/i', $attribute )
 760                                 && !isset( $whitelist[$attribute] )
 761                         ) {
 762                                 continue;
 763                         }
 764
 765                         # Strip javascript "expression" from stylesheets.
 766                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 767                         if ( $attribute == 'style' ) {
 768                                 $value = Sanitizer::checkCss( $value );
 769                         }
 770
 771                         if ( $attribute === 'id' ) {
 772                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 773                         }
 774
 775                         # WAI-ARIA
 776                         # http://www.w3.org/TR/wai-aria/
 777                         # http://www.whatwg.org/html/elements.html#wai-aria
 778                         # For now we only support role="presentation" until we work out what roles should be
 779                         # usable by content and we ensure that our code explicitly rejects patterns that
 780                         # violate HTML5's ARIA restrictions.
 781                         if ( $attribute === 'role' && $value !== 'presentation' ) {
 782                                 continue;
 783                         }
 784
 785                         // RDFa and microdata properties allow URLs, URIs and/or CURIs.
 786                         // Check them for sanity.
 787                         if ( $attribute === 'rel' || $attribute === 'rev'
 788                                 # RDFa
 789                                 || $attribute === 'about' || $attribute === 'property'
 790                                 || $attribute === 'resource' || $attribute === 'datatype'
 791                                 || $attribute === 'typeof'
 792                                 # HTML5 microdata
 793                                 || $attribute === 'itemid' || $attribute === 'itemprop'
 794                                 || $attribute === 'itemref' || $attribute === 'itemscope'
 795                                 || $attribute === 'itemtype'
 796                         ) {
 797                                 // Paranoia. Allow "simple" values but suppress javascript
 798                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 799                                         continue;
 800                                 }
 801                         }
 802
 803                         # NOTE: even though elements using href/src are not allowed directly, supply
 804                         #       validation code that can be used by tag hook handlers, etc
 805                         if ( $attribute === 'href' || $attribute === 'src' ) {
 806                                 if ( !preg_match( $hrefExp, $value ) ) {
 807                                         continue; // drop any href or src attributes not using an allowed protocol.
 808                                         // NOTE: this also drops all relative URLs
 809                                 }
 810                         }
 811
 812                         // If this attribute was previously set, override it.
 813                         // Output should only have one attribute of each name.
 814                         $out[$attribute] = $value;
 815                 }
 816
 817                 if ( $wgAllowMicrodataAttributes ) {
 818                         # itemtype, itemid, itemref don't make sense without itemscope
 819                         if ( !array_key_exists( 'itemscope', $out ) ) {
 820                                 unset( $out['itemtype'] );
 821                                 unset( $out['itemid'] );
 822                                 unset( $out['itemref'] );
 823                         }
 824                         # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 825                 }
 826                 return $out;
 827         }
 828
 829         /**
 830          * Merge two sets of HTML attributes.  Conflicting items in the second set
 831          * will override those in the first, except for 'class' attributes which
 832          * will be combined (if they're both strings).
 833          *
 834          * @todo implement merging for other attributes such as style
 835          * @param array $a
 836          * @param array $b
 837          * @return array
 838          */
 839         static function mergeAttributes( $a, $b ) {
 840                 $out = array_merge( $a, $b );
 841                 if ( isset( $a['class'] ) && isset( $b['class'] )
 842                         && is_string( $a['class'] ) && is_string( $b['class'] )
 843                         && $a['class'] !== $b['class']
 844                 ) {
 845                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 846                                 -1, PREG_SPLIT_NO_EMPTY );
 847                         $out['class'] = implode( ' ', array_unique( $classes ) );
 848                 }
 849                 return $out;
 850         }
 851
 852         /**
 853          * Normalize CSS into a format we can easily search for hostile input
 854          *  - decode character references
 855          *  - decode escape sequences
 856          *  - convert characters that IE6 interprets into ascii
 857          *  - remove comments, unless the entire value is one single comment
 858          * @param string $value the css string
 859          * @return string normalized css
 860          */
 861         public static function normalizeCss( $value ) {
 862
 863                 // Decode character references like &#123;
 864                 $value = Sanitizer::decodeCharReferences( $value );
 865
 866                 // Decode escape sequences and line continuation
 867                 // See the grammar in the CSS 2 spec, appendix D.
 868                 // This has to be done AFTER decoding character references.
 869                 // This means it isn't possible for this function to return
 870                 // unsanitized escape sequences. It is possible to manufacture
 871                 // input that contains character references that decode to
 872                 // escape sequences that decode to character references, but
 873                 // it's OK for the return value to contain character references
 874                 // because the caller is supposed to escape those anyway.
 875                 static $decodeRegex;
 876                 if ( !$decodeRegex ) {
 877                         $space = '[\\x20\\t\\r\\n\\f]';
 878                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 879                         $backslash = '\\\\';
 880                         $decodeRegex = "/ $backslash
 881                                 (?:
 882                                         ($nl) |  # 1. Line continuation
 883                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 884                                         (.) | # 3. backslash cancelling special meaning
 885                                         () | # 4. backslash at end of string
 886                                 )/xu";
 887                 }
 888                 $value = preg_replace_callback( $decodeRegex,
 889                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 890
 891                 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
 892                 $value = preg_replace_callback(
 893                         '/[！-［］-ｚ]/u', // U+FF01 to U+FF5A, excluding U+FF3C (bug 58088)
 894                         function ( $matches ) {
 895                                 $cp = UtfNormal\Utils::utf8ToCodepoint( $matches[0] );
 896                                 if ( $cp === false ) {
 897                                         return '';
 898                                 }
 899                                 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
 900                         },
 901                         $value
 902                 );
 903
 904                 // Convert more characters IE6 might treat as ascii
 905                 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
 906                 $value = str_replace(
 907                         array( 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ),
 908                         array( 'r', 'n', 'n', 'l', 'i', '(', '(' ),
 909                         $value
 910                 );
 911
 912                 // Let the value through if it's nothing but a single comment, to
 913                 // allow other functions which may reject it to pass some error
 914                 // message through.
 915                 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
 916                         // Remove any comments; IE gets token splitting wrong
 917                         // This must be done AFTER decoding character references and
 918                         // escape sequences, because those steps can introduce comments
 919                         // This step cannot introduce character references or escape
 920                         // sequences, because it replaces comments with spaces rather
 921                         // than removing them completely.
 922                         $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 923
 924                         // Remove anything after a comment-start token, to guard against
 925                         // incorrect client implementations.
 926                         $commentPos = strpos( $value, '/*' );
 927                         if ( $commentPos !== false ) {
 928                                 $value = substr( $value, 0, $commentPos );
 929                         }
 930                 }
 931
 932                 // S followed by repeat, iteration, or prolonged sound marks,
 933                 // which IE will treat as "ss"
 934                 $value = preg_replace(
 935                         '/s(?:
 936                                 \xE3\x80\xB1 | # U+3031
 937                                 \xE3\x82\x9D | # U+309D
 938                                 \xE3\x83\xBC | # U+30FC
 939                                 \xE3\x83\xBD | # U+30FD
 940                                 \xEF\xB9\xBC | # U+FE7C
 941                                 \xEF\xB9\xBD | # U+FE7D
 942                                 \xEF\xBD\xB0   # U+FF70
 943                         )/ix',
 944                         'ss',
 945                         $value
 946                 );
 947
 948                 return $value;
 949         }
 950
 951
 952         /**
 953          * Pick apart some CSS and check it for forbidden or unsafe structures.
 954          * Returns a sanitized string. This sanitized string will have
 955          * character references and escape sequences decoded and comments
 956          * stripped (unless it is itself one valid comment, in which case the value
 957          * will be passed through). If the input is just too evil, only a comment
 958          * complaining about evilness will be returned.
 959          *
 960          * Currently URL references, 'expression', 'tps' are forbidden.
 961          *
 962          * NOTE: Despite the fact that character references are decoded, the
 963          * returned string may contain character references given certain
 964          * clever input strings. These character references must
 965          * be escaped before the return value is embedded in HTML.
 966          *
 967          * @param string $value
 968          * @return string
 969          */
 970         static function checkCss( $value ) {
 971                 $value = self::normalizeCss( $value );
 972
 973                 // Reject problematic keywords and control characters
 974                 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ||
 975                         strpos( $value, UtfNormal\Constants::UTF8_REPLACEMENT ) !== false ) {
 976                         return '/* invalid control char */';
 977                 } elseif ( preg_match(
 978                         '! expression
 979                                 | filter\s*:
 980                                 | accelerator\s*:
 981                                 | -o-link\s*:
 982                                 | -o-link-source\s*:
 983                                 | -o-replace\s*:
 984                                 | url\s*\(
 985                                 | image\s*\(
 986                                 | image-set\s*\(
 987                         !ix', $value ) ) {
 988                         return '/* insecure input */';
 989                 }
 990                 return $value;
 991         }
 992
 993         /**
 994          * @param array $matches
 995          * @return string
 996          */
 997         static function cssDecodeCallback( $matches ) {
 998                 if ( $matches[1] !== '' ) {
 999                         // Line continuation
1000                         return '';
1001                 } elseif ( $matches[2] !== '' ) {
1002                         $char = UtfNormal\Utils::codepointToUtf8( hexdec( $matches[2] ) );
1003                 } elseif ( $matches[3] !== '' ) {
1004                         $char = $matches[3];
1005                 } else {
1006                         $char = '\\';
1007                 }
1008                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
1009                         // These characters need to be escaped in strings
1010                         // Clean up the escape sequence to avoid parsing errors by clients
1011                         return '\\' . dechex( ord( $char ) ) . ' ';
1012                 } else {
1013                         // Decode unnecessary escape
1014                         return $char;
1015                 }
1016         }
1017
1018         /**
1019          * Take a tag soup fragment listing an HTML element's attributes
1020          * and normalize it to well-formed XML, discarding unwanted attributes.
1021          * Output is safe for further wikitext processing, with escaping of
1022          * values that could trigger problems.
1023          *
1024          * - Normalizes attribute names to lowercase
1025          * - Discards attributes not on a whitelist for the given element
1026          * - Turns broken or invalid entities into plaintext
1027          * - Double-quotes all attribute values
1028          * - Attributes without values are given the name as attribute
1029          * - Double attributes are discarded
1030          * - Unsafe style attributes are discarded
1031          * - Prepends space if there are attributes.
1032          *
1033          * @param string $text
1034          * @param string $element
1035          * @return string
1036          */
1037         static function fixTagAttributes( $text, $element ) {
1038                 if ( trim( $text ) == '' ) {
1039                         return '';
1040                 }
1041
1042                 $decoded = Sanitizer::decodeTagAttributes( $text );
1043                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1044
1045                 return Sanitizer::safeEncodeTagAttributes( $stripped );
1046         }
1047
1048         /**
1049          * Encode an attribute value for HTML output.
1050          * @param string $text
1051          * @return string HTML-encoded text fragment
1052          */
1053         static function encodeAttribute( $text ) {
1054                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1055
1056                 // Whitespace is normalized during attribute decoding,
1057                 // so if we've been passed non-spaces we must encode them
1058                 // ahead of time or they won't be preserved.
1059                 $encValue = strtr( $encValue, array(
1060                         "\n" => '&#10;',
1061                         "\r" => '&#13;',
1062                         "\t" => '&#9;',
1063                 ) );
1064
1065                 return $encValue;
1066         }
1067
1068         /**
1069          * Encode an attribute value for HTML tags, with extra armoring
1070          * against further wiki processing.
1071          * @param string $text
1072          * @return string HTML-encoded text fragment
1073          */
1074         static function safeEncodeAttribute( $text ) {
1075                 $encValue = Sanitizer::encodeAttribute( $text );
1076
1077                 # Templates and links may be expanded in later parsing,
1078                 # creating invalid or dangerous output. Suppress this.
1079                 $encValue = strtr( $encValue, array(
1080                         '<'    => '&lt;',   // This should never happen,
1081                         '>'    => '&gt;',   // we've received invalid input
1082                         '"'    => '&quot;', // which should have been escaped.
1083                         '{'    => '&#123;',
1084                         '['    => '&#91;',
1085                         "''"   => '&#39;&#39;',
1086                         'ISBN' => '&#73;SBN',
1087                         'RFC'  => '&#82;FC',
1088                         'PMID' => '&#80;MID',
1089                         '|'    => '&#124;',
1090                         '__'   => '&#95;_',
1091                 ) );
1092
1093                 # Stupid hack
1094                 $encValue = preg_replace_callback(
1095                         '/((?i)' . wfUrlProtocols() . ')/',
1096                         array( 'Sanitizer', 'armorLinksCallback' ),
1097                         $encValue );
1098                 return $encValue;
1099         }
1100
1101         /**
1102          * Given a value, escape it so that it can be used in an id attribute and
1103          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1104          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1105          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1106          * escaped with lots of dots.
1107          *
1108          * To ensure we don't have to bother escaping anything, we also strip ', ",
1109          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1110          * We also strip # because it upsets IE, and % because it could be
1111          * ambiguous if it's part of something that looks like a percent escape
1112          * (which don't work reliably in fragments cross-browser).
1113          *
1114          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1115          *   in the id and name attributes
1116          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1117          *   the id attribute
1118          * @see http://www.whatwg.org/html/elements.html#the-id-attribute
1119          *   HTML5 definition of id attribute
1120          *
1121          * @param string $id Id to escape
1122          * @param string|array $options String or array of strings (default is array()):
1123          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1124          *       so don't pay attention if the first character isn't valid at the
1125          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1126          *       false.
1127          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1128          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1129          *       anchors and links won't break.
1130          * @return string
1131          */
1132         static function escapeId( $id, $options = array() ) {
1133                 global $wgExperimentalHtmlIds;
1134                 $options = (array)$options;
1135
1136                 $id = Sanitizer::decodeCharReferences( $id );
1137
1138                 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1139                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1140                         $id = trim( $id, '_' );
1141                         if ( $id === '' ) {
1142                                 // Must have been all whitespace to start with.
1143                                 return '_';
1144                         } else {
1145                                 return $id;
1146                         }
1147                 }
1148
1149                 // HTML4-style escaping
1150                 static $replace = array(
1151                         '%3A' => ':',
1152                         '%' => '.'
1153                 );
1154
1155                 $id = urlencode( strtr( $id, ' ', '_' ) );
1156                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1157
1158                 if ( !preg_match( '/^[a-zA-Z]/', $id ) && !in_array( 'noninitial', $options ) ) {
1159                         // Initial character must be a letter!
1160                         $id = "x$id";
1161                 }
1162                 return $id;
1163         }
1164
1165         /**
1166          * Given a value, escape it so that it can be used as a CSS class and
1167          * return it.
1168          *
1169          * @todo For extra validity, input should be validated UTF-8.
1170          *
1171          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1172          *
1173          * @param string $class
1174          * @return string
1175          */
1176         static function escapeClass( $class ) {
1177                 // Convert ugly stuff to underscores and kill underscores in ugly places
1178                 return rtrim( preg_replace(
1179                         array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1180                         '_',
1181                         $class ), '_' );
1182         }
1183
1184         /**
1185          * Given HTML input, escape with htmlspecialchars but un-escape entities.
1186          * This allows (generally harmless) entities like &#160; to survive.
1187          *
1188          * @param string $html HTML to escape
1189          * @return string Escaped input
1190          */
1191         static function escapeHtmlAllowEntities( $html ) {
1192                 $html = Sanitizer::decodeCharReferences( $html );
1193                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1194                 # hurt.
1195                 $html = htmlspecialchars( $html, ENT_QUOTES );
1196                 return $html;
1197         }
1198
1199         /**
1200          * Regex replace callback for armoring links against further processing.
1201          * @param array $matches
1202          * @return string
1203          */
1204         private static function armorLinksCallback( $matches ) {
1205                 return str_replace( ':', '&#58;', $matches[1] );
1206         }
1207
1208         /**
1209          * Return an associative array of attribute names and values from
1210          * a partial tag string. Attribute names are forces to lowercase,
1211          * character references are decoded to UTF-8 text.
1212          *
1213          * @param string $text
1214          * @return array
1215          */
1216         public static function decodeTagAttributes( $text ) {
1217                 if ( trim( $text ) == '' ) {
1218                         return array();
1219                 }
1220
1221                 $attribs = array();
1222                 $pairs = array();
1223                 if ( !preg_match_all(
1224                         self::getAttribsRegex(),
1225                         $text,
1226                         $pairs,
1227                         PREG_SET_ORDER ) ) {
1228                         return $attribs;
1229                 }
1230
1231                 foreach ( $pairs as $set ) {
1232                         $attribute = strtolower( $set[1] );
1233                         $value = Sanitizer::getTagAttributeCallback( $set );
1234
1235                         // Normalize whitespace
1236                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1237                         $value = trim( $value );
1238
1239                         // Decode character references
1240                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1241                 }
1242                 return $attribs;
1243         }
1244
1245         /**
1246          * Build a partial tag string from an associative array of attribute
1247          * names and values as returned by decodeTagAttributes.
1248          *
1249          * @param array $assoc_array
1250          * @return string
1251          */
1252         public static function safeEncodeTagAttributes( $assoc_array ) {
1253                 $attribs = array();
1254                 foreach ( $assoc_array as $attribute => $value ) {
1255                         $encAttribute = htmlspecialchars( $attribute );
1256                         $encValue = Sanitizer::safeEncodeAttribute( $value );
1257
1258                         $attribs[] = "$encAttribute=\"$encValue\"";
1259                 }
1260                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1261         }
1262
1263         /**
1264          * Pick the appropriate attribute value from a match set from the
1265          * attribs regex matches.
1266          *
1267          * @param array $set
1268          * @throws MWException When tag conditions are not met.
1269          * @return string
1270          */
1271         private static function getTagAttributeCallback( $set ) {
1272                 if ( isset( $set[5] ) ) {
1273                         # No quotes.
1274                         return $set[5];
1275                 } elseif ( isset( $set[4] ) ) {
1276                         # Single-quoted
1277                         return $set[4];
1278                 } elseif ( isset( $set[3] ) ) {
1279                         # Double-quoted
1280                         return $set[3];
1281                 } elseif ( !isset( $set[2] ) ) {
1282                         # In XHTML, attributes must have a value so return an empty string.
1283                         # See "Empty attribute syntax",
1284                         # http://www.w3.org/TR/html5/syntax.html#syntax-attribute-name
1285                         return "";
1286                 } else {
1287                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1288                 }
1289         }
1290
1291         /**
1292          * @param string $text
1293          * @return string
1294          */
1295         private static function normalizeWhitespace( $text ) {
1296                 return preg_replace(
1297                         '/\r\n|[\x20\x0d\x0a\x09]/',
1298                         ' ',
1299                         $text );
1300         }
1301
1302         /**
1303          * Normalizes whitespace in a section name, such as might be returned
1304          * by Parser::stripSectionName(), for use in the id's that are used for
1305          * section links.
1306          *
1307          * @param string $section
1308          * @return string
1309          */
1310         static function normalizeSectionNameWhitespace( $section ) {
1311                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1312         }
1313
1314         /**
1315          * Ensure that any entities and character references are legal
1316          * for XML and XHTML specifically. Any stray bits will be
1317          * &amp;-escaped to result in a valid text fragment.
1318          *
1319          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1320          *   numericized (this way we're well-formed even without a DTD)
1321          * b. any numeric char refs must be legal chars, not invalid or forbidden
1322          * c. use lower cased "&#x", not "&#X"
1323          * d. fix or reject non-valid attributes
1324          *
1325          * @param string $text
1326          * @return string
1327          * @private
1328          */
1329         static function normalizeCharReferences( $text ) {
1330                 return preg_replace_callback(
1331                         self::CHAR_REFS_REGEX,
1332                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1333                         $text );
1334         }
1335
1336         /**
1337          * @param string $matches
1338          * @return string
1339          */
1340         static function normalizeCharReferencesCallback( $matches ) {
1341                 $ret = null;
1342                 if ( $matches[1] != '' ) {
1343                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1344                 } elseif ( $matches[2] != '' ) {
1345                         $ret = Sanitizer::decCharReference( $matches[2] );
1346                 } elseif ( $matches[3] != '' ) {
1347                         $ret = Sanitizer::hexCharReference( $matches[3] );
1348                 }
1349                 if ( is_null( $ret ) ) {
1350                         return htmlspecialchars( $matches[0] );
1351                 } else {
1352                         return $ret;
1353                 }
1354         }
1355
1356         /**
1357          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1358          * return the equivalent numeric entity reference (except for the core &lt;
1359          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1360          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1361          * pseudo-entity source (eg &amp;foo;)
1362          *
1363          * @param string $name
1364          * @return string
1365          */
1366         static function normalizeEntity( $name ) {
1367                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1368                         return '&' . self::$htmlEntityAliases[$name] . ';';
1369                 } elseif ( in_array( $name, array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1370                         return "&$name;";
1371                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1372                         return '&#' . self::$htmlEntities[$name] . ';';
1373                 } else {
1374                         return "&amp;$name;";
1375                 }
1376         }
1377
1378         /**
1379          * @param int $codepoint
1380          * @return null|string
1381          */
1382         static function decCharReference( $codepoint ) {
1383                 $point = intval( $codepoint );
1384                 if ( Sanitizer::validateCodepoint( $point ) ) {
1385                         return sprintf( '&#%d;', $point );
1386                 } else {
1387                         return null;
1388                 }
1389         }
1390
1391         /**
1392          * @param int $codepoint
1393          * @return null|string
1394          */
1395         static function hexCharReference( $codepoint ) {
1396                 $point = hexdec( $codepoint );
1397                 if ( Sanitizer::validateCodepoint( $point ) ) {
1398                         return sprintf( '&#x%x;', $point );
1399                 } else {
1400                         return null;
1401                 }
1402         }
1403
1404         /**
1405          * Returns true if a given Unicode codepoint is a valid character in
1406          * both HTML5 and XML.
1407          * @param int $codepoint
1408          * @return bool
1409          */
1410         private static function validateCodepoint( $codepoint ) {
1411                 # U+000C is valid in HTML5 but not allowed in XML.
1412                 # U+000D is valid in XML but not allowed in HTML5.
1413                 # U+007F - U+009F are disallowed in HTML5 (control characters).
1414                 return $codepoint == 0x09
1415                         || $codepoint == 0x0a
1416                         || ( $codepoint >= 0x20 && $codepoint <= 0x7e )
1417                         || ( $codepoint >= 0xa0 && $codepoint <= 0xd7ff )
1418                         || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1419                         || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1420         }
1421
1422         /**
1423          * Decode any character references, numeric or named entities,
1424          * in the text and return a UTF-8 string.
1425          *
1426          * @param string $text
1427          * @return string
1428          */
1429         public static function decodeCharReferences( $text ) {
1430                 return preg_replace_callback(
1431                         self::CHAR_REFS_REGEX,
1432                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1433                         $text );
1434         }
1435
1436         /**
1437          * Decode any character references, numeric or named entities,
1438          * in the next and normalize the resulting string. (bug 14952)
1439          *
1440          * This is useful for page titles, not for text to be displayed,
1441          * MediaWiki allows HTML entities to escape normalization as a feature.
1442          *
1443          * @param string $text Already normalized, containing entities
1444          * @return string Still normalized, without entities
1445          */
1446         public static function decodeCharReferencesAndNormalize( $text ) {
1447                 global $wgContLang;
1448                 $text = preg_replace_callback(
1449                         self::CHAR_REFS_REGEX,
1450                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1451                         $text, /* limit */ -1, $count );
1452
1453                 if ( $count ) {
1454                         return $wgContLang->normalize( $text );
1455                 } else {
1456                         return $text;
1457                 }
1458         }
1459
1460         /**
1461          * @param string $matches
1462          * @return string
1463          */
1464         static function decodeCharReferencesCallback( $matches ) {
1465                 if ( $matches[1] != '' ) {
1466                         return Sanitizer::decodeEntity( $matches[1] );
1467                 } elseif ( $matches[2] != '' ) {
1468                         return Sanitizer::decodeChar( intval( $matches[2] ) );
1469                 } elseif ( $matches[3] != '' ) {
1470                         return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1471                 }
1472                 # Last case should be an ampersand by itself
1473                 return $matches[0];
1474         }
1475
1476         /**
1477          * Return UTF-8 string for a codepoint if that is a valid
1478          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1479          * @param int $codepoint
1480          * @return string
1481          * @private
1482          */
1483         static function decodeChar( $codepoint ) {
1484                 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1485                         return UtfNormal\Utils::codepointToUtf8( $codepoint );
1486                 } else {
1487                         return UtfNormal\Constants::UTF8_REPLACEMENT;
1488                 }
1489         }
1490
1491         /**
1492          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1493          * return the UTF-8 encoding of that character. Otherwise, returns
1494          * pseudo-entity source (eg "&foo;")
1495          *
1496          * @param string $name
1497          * @return string
1498          */
1499         static function decodeEntity( $name ) {
1500                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1501                         $name = self::$htmlEntityAliases[$name];
1502                 }
1503                 if ( isset( self::$htmlEntities[$name] ) ) {
1504                         return UtfNormal\Utils::codepointToUtf8( self::$htmlEntities[$name] );
1505                 } else {
1506                         return "&$name;";
1507                 }
1508         }
1509
1510         /**
1511          * Fetch the whitelist of acceptable attributes for a given element name.
1512          *
1513          * @param string $element
1514          * @return array
1515          */
1516         static function attributeWhitelist( $element ) {
1517                 $list = Sanitizer::setupAttributeWhitelist();
1518                 return isset( $list[$element] )
1519                         ? $list[$element]
1520                         : array();
1521         }
1522
1523         /**
1524          * Foreach array key (an allowed HTML element), return an array
1525          * of allowed attributes
1526          * @return array
1527          */
1528         static function setupAttributeWhitelist() {
1529                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
1530                 static $whitelist, $staticInitialised;
1531
1532                 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgAllowMicrodataAttributes' ) );
1533
1534                 if ( $whitelist !== null && $staticInitialised == $globalContext ) {
1535                         return $whitelist;
1536                 }
1537
1538                 $common = array(
1539                         # HTML
1540                         'id',
1541                         'class',
1542                         'style',
1543                         'lang',
1544                         'dir',
1545                         'title',
1546
1547                         # WAI-ARIA
1548                         'role',
1549                 );
1550
1551                 if ( $wgAllowRdfaAttributes ) {
1552                         # RDFa attributes as specified in section 9 of
1553                         # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1554                         $common = array_merge( $common, array(
1555                                 'about', 'property', 'resource', 'datatype', 'typeof',
1556                         ) );
1557                 }
1558
1559                 if ( $wgAllowMicrodataAttributes ) {
1560                         # add HTML5 microdata tags as specified by
1561                         # http://www.whatwg.org/html/microdata.html#the-microdata-model
1562                         $common = array_merge( $common, array(
1563                                 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1564                         ) );
1565                 }
1566
1567                 $block = array_merge( $common, array( 'align' ) );
1568                 $tablealign = array( 'align', 'valign' );
1569                 $tablecell = array(
1570                         'abbr',
1571                         'axis',
1572                         'headers',
1573                         'scope',
1574                         'rowspan',
1575                         'colspan',
1576                         'nowrap', # deprecated
1577                         'width', # deprecated
1578                         'height', # deprecated
1579                         'bgcolor', # deprecated
1580                 );
1581
1582                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1583                 # See: http://www.w3.org/TR/html4/
1584                 $whitelist = array(
1585                         # 7.5.4
1586                         'div'        => $block,
1587                         'center'     => $common, # deprecated
1588                         'span'       => $common,
1589
1590                         # 7.5.5
1591                         'h1'         => $block,
1592                         'h2'         => $block,
1593                         'h3'         => $block,
1594                         'h4'         => $block,
1595                         'h5'         => $block,
1596                         'h6'         => $block,
1597
1598                         # 7.5.6
1599                         # address
1600
1601                         # 8.2.4
1602                         'bdo'        => $common,
1603
1604                         # 9.2.1
1605                         'em'         => $common,
1606                         'strong'     => $common,
1607                         'cite'       => $common,
1608                         'dfn'        => $common,
1609                         'code'       => $common,
1610                         'samp'       => $common,
1611                         'kbd'        => $common,
1612                         'var'        => $common,
1613                         'abbr'       => $common,
1614                         # acronym
1615
1616                         # 9.2.2
1617                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1618                         'q'          => array_merge( $common, array( 'cite' ) ),
1619
1620                         # 9.2.3
1621                         'sub'        => $common,
1622                         'sup'        => $common,
1623
1624                         # 9.3.1
1625                         'p'          => $block,
1626
1627                         # 9.3.2
1628                         'br'         => array_merge( $common, array( 'clear' ) ),
1629
1630                         # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1631                         'wbr'        => $common,
1632
1633                         # 9.3.4
1634                         'pre'        => array_merge( $common, array( 'width' ) ),
1635
1636                         # 9.4
1637                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1638                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1639
1640                         # 10.2
1641                         'ul'         => array_merge( $common, array( 'type' ) ),
1642                         'ol'         => array_merge( $common, array( 'type', 'start', 'reversed' ) ),
1643                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1644
1645                         # 10.3
1646                         'dl'         => $common,
1647                         'dd'         => $common,
1648                         'dt'         => $common,
1649
1650                         # 11.2.1
1651                         'table'      => array_merge( $common,
1652                                                                 array( 'summary', 'width', 'border', 'frame',
1653                                                                                 'rules', 'cellspacing', 'cellpadding',
1654                                                                                 'align', 'bgcolor',
1655                                                                 ) ),
1656
1657                         # 11.2.2
1658                         'caption'    => $block,
1659
1660                         # 11.2.3
1661                         'thead'      => $common,
1662                         'tfoot'      => $common,
1663                         'tbody'      => $common,
1664
1665                         # 11.2.4
1666                         'colgroup'   => array_merge( $common, array( 'span' ) ),
1667                         'col'        => array_merge( $common, array( 'span' ) ),
1668
1669                         # 11.2.5
1670                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1671
1672                         # 11.2.6
1673                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1674                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1675
1676                         # 12.2
1677                         # NOTE: <a> is not allowed directly, but the attrib
1678                         # whitelist is used from the Parser object
1679                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1680
1681                         # 13.2
1682                         # Not usually allowed, but may be used for extension-style hooks
1683                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1684                         # true
1685                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1686
1687                         # 15.2.1
1688                         'tt'         => $common,
1689                         'b'          => $common,
1690                         'i'          => $common,
1691                         'big'        => $common,
1692                         'small'      => $common,
1693                         'strike'     => $common,
1694                         's'          => $common,
1695                         'u'          => $common,
1696
1697                         # 15.2.2
1698                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1699                         # basefont
1700
1701                         # 15.3
1702                         'hr'         => array_merge( $common, array( 'width' ) ),
1703
1704                         # HTML Ruby annotation text module, simple ruby only.
1705                         # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
1706                         'ruby'       => $common,
1707                         # rbc
1708                         'rb'         => $common,
1709                         'rp'         => $common,
1710                         'rt'         => $common, # array_merge( $common, array( 'rbspan' ) ),
1711                         'rtc'         => $common,
1712
1713                         # MathML root element, where used for extensions
1714                         # 'title' may not be 100% valid here; it's XHTML
1715                         # http://www.w3.org/TR/REC-MathML/
1716                         'math'       => array( 'class', 'style', 'id', 'title' ),
1717
1718                         # HTML 5 section 4.6
1719                         'bdi' => $common,
1720
1721                         # HTML5 elements, defined by:
1722                         # http://www.whatwg.org/html/
1723                         'data' => array_merge( $common, array( 'value' ) ),
1724                         'time' => array_merge( $common, array( 'datetime' ) ),
1725                         'mark' => $common,
1726
1727                         // meta and link are only permitted by removeHTMLtags when Microdata
1728                         // is enabled so we don't bother adding a conditional to hide these
1729                         // Also meta and link are only valid in WikiText as Microdata elements
1730                         // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1731                         // So we don't bother including $common attributes that have no purpose.
1732                         'meta' => array( 'itemprop', 'content' ),
1733                         'link' => array( 'itemprop', 'href' ),
1734                 );
1735
1736                 $staticInitialised = $globalContext;
1737
1738                 return $whitelist;
1739         }
1740
1741         /**
1742          * Take a fragment of (potentially invalid) HTML and return
1743          * a version with any tags removed, encoded as plain text.
1744          *
1745          * Warning: this return value must be further escaped for literal
1746          * inclusion in HTML output as of 1.10!
1747          *
1748          * @param string $text HTML fragment
1749          * @return string
1750          */
1751         static function stripAllTags( $text ) {
1752                 # Actual <tags>
1753                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1754
1755                 # Normalize &entities and whitespace
1756                 $text = self::decodeCharReferences( $text );
1757                 $text = self::normalizeWhitespace( $text );
1758
1759                 return $text;
1760         }
1761
1762         /**
1763          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1764          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1765          * PHP 5.1 doesn't.
1766          *
1767          * Use for passing XHTML fragments to PHP's XML parsing functions
1768          *
1769          * @return string
1770          */
1771         static function hackDocType() {
1772                 $out = "<!DOCTYPE html [\n";
1773                 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1774                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1775                 }
1776                 $out .= "]>\n";
1777                 return $out;
1778         }
1779
1780         /**
1781          * @param string $url
1782          * @return mixed|string
1783          */
1784         static function cleanUrl( $url ) {
1785                 # Normalize any HTML entities in input. They will be
1786                 # re-escaped by makeExternalLink().
1787                 $url = Sanitizer::decodeCharReferences( $url );
1788
1789                 # Escape any control characters introduced by the above step
1790                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1791                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1792
1793                 # Validate hostname portion
1794                 $matches = array();
1795                 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1796                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1797
1798                         // Characters that will be ignored in IDNs.
1799                         // http://tools.ietf.org/html/3454#section-3.1
1800                         // Strip them before further processing so blacklists and such work.
1801                         $strip = "/
1802                                 \\s|          # general whitespace
1803                                 \xc2\xad|     # 00ad SOFT HYPHEN
1804                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1805                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1806                                 \xe2\x81\xa0| # 2060 WORD JOINER
1807                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1808                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1809                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1810                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1811                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1812                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1813                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1814                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1815                                 /xuD";
1816
1817                         $host = preg_replace( $strip, '', $host );
1818
1819                         // IPv6 host names are bracketed with [].  Url-decode these.
1820                         if ( substr_compare( "//%5B", $host, 0, 5 ) === 0 &&
1821                                 preg_match( '!^//%5B([0-9A-Fa-f:.]+)%5D((:\d+)?)$!', $host, $matches )
1822                         ) {
1823                                 $host = '//[' . $matches[1] . ']' . $matches[2];
1824                         }
1825
1826                         // @todo FIXME: Validate hostnames here
1827
1828                         return $protocol . $host . $rest;
1829                 } else {
1830                         return $url;
1831                 }
1832         }
1833
1834         /**
1835          * @param array $matches
1836          * @return string
1837          */
1838         static function cleanUrlCallback( $matches ) {
1839                 return urlencode( $matches[0] );
1840         }
1841
1842         /**
1843          * Does a string look like an e-mail address?
1844          *
1845          * This validates an email address using an HTML5 specification found at:
1846          * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1847          * Which as of 2011-01-24 says:
1848          *
1849          *   A valid e-mail address is a string that matches the ABNF production
1850          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1851          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1852          *   3.5.
1853          *
1854          * This function is an implementation of the specification as requested in
1855          * bug 22449.
1856          *
1857          * Client-side forms will use the same standard validation rules via JS or
1858          * HTML 5 validation; additional restrictions can be enforced server-side
1859          * by extensions via the 'isValidEmailAddr' hook.
1860          *
1861          * Note that this validation doesn't 100% match RFC 2822, but is believed
1862          * to be liberal enough for wide use. Some invalid addresses will still
1863          * pass validation here.
1864          *
1865          * @since 1.18
1866          *
1867          * @param string $addr E-mail address
1868          * @return bool
1869          */
1870         public static function validateEmail( $addr ) {
1871                 $result = null;
1872                 if ( !Hooks::run( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1873                         return $result;
1874                 }
1875
1876                 // Please note strings below are enclosed in brackets [], this make the
1877                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1878                 // See bug 26948
1879                 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1880                 $rfc1034_ldh_str = "a-z0-9\\-";
1881
1882                 $html5_email_regexp = "/
1883                 ^                      # start of string
1884                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1885                 @                      # 'apostrophe'
1886                 [$rfc1034_ldh_str]+       # First domain part
1887                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1888                 $                      # End of string
1889                 /ix"; // case Insensitive, eXtended
1890
1891                 return (bool)preg_match( $html5_email_regexp, $addr );
1892         }
1893 }