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