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 = [
  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 = [
 324                 'רלמ' => 'rlm',
 325                 'رلم' => 'rlm',
 326         ];
 327
 328         /**
 329          * Lazy-initialised attributes regex, see getAttribsRegex()
 330          */
 331         private static $attribsRegex;
 332
 333         /**
 334          * Regular expression to match HTML/XML attribute pairs within a tag.
 335          * Allows some... latitude. Based on,
 336          * http://www.w3.org/TR/html5/syntax.html#before-attribute-value-state
 337          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 338          * @return string
 339          */
 340         static function getAttribsRegex() {
 341                 if ( self::$attribsRegex === null ) {
 342                         $attribFirst = '[:A-Z_a-z0-9]';
 343                         $attrib = '[:A-Z_a-z-.0-9]';
 344                         $space = '[\x09\x0a\x0c\x0d\x20]';
 345                         self::$attribsRegex =
 346                                 "/(?:^|$space)({$attribFirst}{$attrib}*)
 347                                   ($space*=$space*
 348                                         (?:
 349                                          # The attribute value: quoted or alone
 350                                           \"([^\"]*)(?:\"|\$)
 351                                          | '([^']*)(?:'|\$)
 352                                          |  (((?!$space|>).)*)
 353                                         )
 354                                 )?(?=$space|\$)/sx";
 355                 }
 356                 return self::$attribsRegex;
 357         }
 358
 359         /**
 360          * Return the various lists of recognized tags
 361          * @param array $extratags For any extra tags to include
 362          * @param array $removetags For any tags (default or extra) to exclude
 363          * @return array
 364          */
 365         public static function getRecognizedTagData( $extratags = [], $removetags = [] ) {
 366                 global $wgAllowImageTag;
 367
 368                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 369                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 370
 371                 // Base our staticInitialised variable off of the global config state so that if the globals
 372                 // are changed (like in the screwed up test system) we will re-initialise the settings.
 373                 $globalContext = $wgAllowImageTag;
 374                 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
 375                         $htmlpairsStatic = [ # Tags that must be closed
 376                                 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 377                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 378                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 379                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 380                                 'ruby', 'rb', 'rp', 'rt', 'rtc', 'p', 'span', 'abbr', 'dfn',
 381                                 'kbd', 'samp', 'data', 'time', 'mark'
 382                         ];
 383                         $htmlsingle = [
 384                                 'br', 'wbr', 'hr', 'li', 'dt', 'dd'
 385                         ];
 386                         $htmlsingleonly = [ # Elements that cannot have close tags
 387                                 'br', 'wbr', 'hr'
 388                         ];
 389
 390                         $htmlsingle[] = $htmlsingleonly[] = 'meta';
 391                         $htmlsingle[] = $htmlsingleonly[] = 'link';
 392
 393                         $htmlnest = [ # Tags that can be nested--??
 394                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 395                                 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
 396                                 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
 397                         ];
 398                         $tabletags = [ # Can only appear inside table, we will close them
 399                                 'td', 'th', 'tr',
 400                         ];
 401                         $htmllist = [ # Tags used by list
 402                                 'ul', 'ol',
 403                         ];
 404                         $listtags = [ # Tags that can appear in a list
 405                                 'li',
 406                         ];
 407
 408                         if ( $wgAllowImageTag ) {
 409                                 $htmlsingle[] = 'img';
 410                                 $htmlsingleonly[] = 'img';
 411                         }
 412
 413                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 414                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 415
 416                         # Convert them all to hashtables for faster lookup
 417                         $vars = [ 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 418                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' ];
 419                         foreach ( $vars as $var ) {
 420                                 $$var = array_flip( $$var );
 421                         }
 422                         $staticInitialised = $globalContext;
 423                 }
 424
 425                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 426                 $extratags = array_flip( $extratags );
 427                 $removetags = array_flip( $removetags );
 428                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 429                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
 430
 431                 return [
 432                         'htmlpairs' => $htmlpairs,
 433                         'htmlsingle' => $htmlsingle,
 434                         'htmlsingleonly' => $htmlsingleonly,
 435                         'htmlnest' => $htmlnest,
 436                         'tabletags' => $tabletags,
 437                         'htmllist' => $htmllist,
 438                         'listtags' => $listtags,
 439                         'htmlsingleallowed' => $htmlsingleallowed,
 440                         'htmlelements' => $htmlelements,
 441                 ];
 442         }
 443
 444         /**
 445          * Cleans up HTML, removes dangerous tags and attributes, and
 446          * removes HTML comments
 447          * @param string $text
 448          * @param callable $processCallback Callback to do any variable or parameter
 449          *   replacements in HTML attribute values
 450          * @param array|bool $args Arguments for the processing callback
 451          * @param array $extratags For any extra tags to include
 452          * @param array $removetags For any tags (default or extra) to exclude
 453          * @return string
 454          */
 455         public static function removeHTMLtags( $text, $processCallback = null,
 456                 $args = [], $extratags = [], $removetags = []
 457         ) {
 458                 extract( self::getRecognizedTagData( $extratags, $removetags ) );
 459
 460                 # Remove HTML comments
 461                 $text = Sanitizer::removeHTMLcomments( $text );
 462                 $bits = explode( '<', $text );
 463                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 464                 if ( !MWTidy::isEnabled() ) {
 465                         $tagstack = $tablestack = [];
 466                         foreach ( $bits as $x ) {
 467                                 $regs = [];
 468                                 # $slash: Does the current element start with a '/'?
 469                                 # $t: Current element name
 470                                 # $params: String between element name and >
 471                                 # $brace: Ending '>' or '/>'
 472                                 # $rest: Everything until the next element of $bits
 473                                 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 474                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 475                                 } else {
 476                                         $slash = $t = $params = $brace = $rest = null;
 477                                 }
 478
 479                                 $badtag = false;
 480                                 $t = strtolower( $t );
 481                                 if ( isset( $htmlelements[$t] ) ) {
 482                                         # Check our stack
 483                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 484                                                 $badtag = true;
 485                                         } elseif ( $slash ) {
 486                                                 # Closing a tag... is it the one we just opened?
 487                                                 MediaWiki\suppressWarnings();
 488                                                 $ot = array_pop( $tagstack );
 489                                                 MediaWiki\restoreWarnings();
 490
 491                                                 if ( $ot != $t ) {
 492                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 493                                                                 # Pop all elements with an optional close tag
 494                                                                 # and see if we find a match below them
 495                                                                 $optstack = [];
 496                                                                 array_push( $optstack, $ot );
 497                                                                 MediaWiki\suppressWarnings();
 498                                                                 $ot = array_pop( $tagstack );
 499                                                                 MediaWiki\restoreWarnings();
 500                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 501                                                                         array_push( $optstack, $ot );
 502                                                                         MediaWiki\suppressWarnings();
 503                                                                         $ot = array_pop( $tagstack );
 504                                                                         MediaWiki\restoreWarnings();
 505                                                                 }
 506                                                                 if ( $t != $ot ) {
 507                                                                         # No match. Push the optional elements back again
 508                                                                         $badtag = true;
 509                                                                         MediaWiki\suppressWarnings();
 510                                                                         $ot = array_pop( $optstack );
 511                                                                         MediaWiki\restoreWarnings();
 512                                                                         while ( $ot ) {
 513                                                                                 array_push( $tagstack, $ot );
 514                                                                                 MediaWiki\suppressWarnings();
 515                                                                                 $ot = array_pop( $optstack );
 516                                                                                 MediaWiki\restoreWarnings();
 517                                                                         }
 518                                                                 }
 519                                                         } else {
 520                                                                 MediaWiki\suppressWarnings();
 521                                                                 array_push( $tagstack, $ot );
 522                                                                 MediaWiki\restoreWarnings();
 523
 524                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 525                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 526                                                                         $badtag = true;
 527                                                                 }
 528                                                         }
 529                                                 } else {
 530                                                         if ( $t == 'table' ) {
 531                                                                 $tagstack = array_pop( $tablestack );
 532                                                         }
 533                                                 }
 534                                                 $newparams = '';
 535                                         } else {
 536                                                 # Keep track for later
 537                                                 if ( isset( $tabletags[$t] ) && !in_array( 'table', $tagstack ) ) {
 538                                                         $badtag = true;
 539                                                 } elseif ( in_array( $t, $tagstack ) && !isset( $htmlnest[$t] ) ) {
 540                                                         $badtag = true;
 541                                                 #  Is it a self closed htmlpair ? (bug 5487)
 542                                                 } elseif ( $brace == '/>' && isset( $htmlpairs[$t] ) ) {
 543                                                         $badtag = true;
 544                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 545                                                         # Hack to force empty tag for unclosable elements
 546                                                         $brace = '/>';
 547                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 548                                                         # Hack to not close $htmlsingle tags
 549                                                         $brace = null;
 550                                                         # Still need to push this optionally-closed tag to
 551                                                         # the tag stack so that we can match end tags
 552                                                         # instead of marking them as bad.
 553                                                         array_push( $tagstack, $t );
 554                                                 } elseif ( isset( $tabletags[$t] ) && in_array( $t, $tagstack ) ) {
 555                                                         // New table tag but forgot to close the previous one
 556                                                         $text .= "</$t>";
 557                                                 } else {
 558                                                         if ( $t == 'table' ) {
 559                                                                 array_push( $tablestack, $tagstack );
 560                                                                 $tagstack = [];
 561                                                         }
 562                                                         array_push( $tagstack, $t );
 563                                                 }
 564
 565                                                 # Replace any variables or template parameters with
 566                                                 # plaintext results.
 567                                                 if ( is_callable( $processCallback ) ) {
 568                                                         call_user_func_array( $processCallback, [ &$params, $args ] );
 569                                                 }
 570
 571                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 572                                                         $badtag = true;
 573                                                 }
 574
 575                                                 # Strip non-approved attributes from the tag
 576                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 577                                         }
 578                                         if ( !$badtag ) {
 579                                                 $rest = str_replace( '>', '&gt;', $rest );
 580                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 581                                                 $text .= "<$slash$t$newparams$close>$rest";
 582                                                 continue;
 583                                         }
 584                                 }
 585                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 586                         }
 587                         # Close off any remaining tags
 588                         while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
 589                                 $text .= "</$t>\n";
 590                                 if ( $t == 'table' ) {
 591                                         $tagstack = array_pop( $tablestack );
 592                                 }
 593                         }
 594                 } else {
 595                         # this might be possible using tidy itself
 596                         foreach ( $bits as $x ) {
 597                                 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 598                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 599
 600                                         $badtag = false;
 601                                         $t = strtolower( $t );
 602                                         if ( isset( $htmlelements[$t] ) ) {
 603                                                 if ( is_callable( $processCallback ) ) {
 604                                                         call_user_func_array( $processCallback, [ &$params, $args ] );
 605                                                 }
 606
 607                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 608                                                         $badtag = true;
 609                                                 }
 610
 611                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 612                                                 if ( !$badtag ) {
 613                                                         $rest = str_replace( '>', '&gt;', $rest );
 614                                                         $text .= "<$slash$t$newparams$brace$rest";
 615                                                         continue;
 616                                                 }
 617                                         }
 618                                 }
 619                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 620                         }
 621                 }
 622                 return $text;
 623         }
 624
 625         /**
 626          * Remove '<!--', '-->', and everything between.
 627          * To avoid leaving blank lines, when a comment is both preceded
 628          * and followed by a newline (ignoring spaces), trim leading and
 629          * trailing spaces and one of the newlines.
 630          *
 631          * @param string $text
 632          * @return string
 633          */
 634         public static function removeHTMLcomments( $text ) {
 635                 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
 636                         $end = strpos( $text, '-->', $start + 4 );
 637                         if ( $end === false ) {
 638                                 # Unterminated comment; bail out
 639                                 break;
 640                         }
 641
 642                         $end += 3;
 643
 644                         # Trim space and newline if the comment is both
 645                         # preceded and followed by a newline
 646                         $spaceStart = max( $start - 1, 0 );
 647                         $spaceLen = $end - $spaceStart;
 648                         while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
 649                                 $spaceStart--;
 650                                 $spaceLen++;
 651                         }
 652                         while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
 653                                 $spaceLen++;
 654                         }
 655                         if ( substr( $text, $spaceStart, 1 ) === "\n"
 656                                 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
 657                                 # Remove the comment, leading and trailing
 658                                 # spaces, and leave only one newline.
 659                                 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
 660                         } else {
 661                                 # Remove just the comment.
 662                                 $text = substr_replace( $text, '', $start, $end - $start );
 663                         }
 664                 }
 665                 return $text;
 666         }
 667
 668         /**
 669          * Takes attribute names and values for a tag and the tag name and
 670          * validates that the tag is allowed to be present.
 671          * This DOES NOT validate the attributes, nor does it validate the
 672          * tags themselves. This method only handles the special circumstances
 673          * where we may want to allow a tag within content but ONLY when it has
 674          * specific attributes set.
 675          *
 676          * @param string $params
 677          * @param string $element
 678          * @return bool
 679          */
 680         static function validateTag( $params, $element ) {
 681                 $params = Sanitizer::decodeTagAttributes( $params );
 682
 683                 if ( $element == 'meta' || $element == 'link' ) {
 684                         if ( !isset( $params['itemprop'] ) ) {
 685                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 686                                 return false;
 687                         }
 688                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 689                                 // <meta> must have a content="" for the itemprop
 690                                 return false;
 691                         }
 692                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 693                                 // <link> must have an associated href=""
 694                                 return false;
 695                         }
 696                 }
 697
 698                 return true;
 699         }
 700
 701         /**
 702          * Take an array of attribute names and values and normalize or discard
 703          * illegal values for the given element type.
 704          *
 705          * - Discards attributes not on a whitelist for the given element
 706          * - Unsafe style attributes are discarded
 707          * - Invalid id attributes are re-encoded
 708          *
 709          * @param array $attribs
 710          * @param string $element
 711          * @return array
 712          *
 713          * @todo Check for legal values where the DTD limits things.
 714          * @todo Check for unique id attribute :P
 715          */
 716         static function validateTagAttributes( $attribs, $element ) {
 717                 return Sanitizer::validateAttributes( $attribs,
 718                         Sanitizer::attributeWhitelist( $element ) );
 719         }
 720
 721         /**
 722          * Take an array of attribute names and values and normalize or discard
 723          * illegal values for the given whitelist.
 724          *
 725          * - Discards attributes not on the given whitelist
 726          * - Unsafe style attributes are discarded
 727          * - Invalid id attributes are re-encoded
 728          *
 729          * @param array $attribs
 730          * @param array $whitelist List of allowed attribute names
 731          * @return array
 732          *
 733          * @todo Check for legal values where the DTD limits things.
 734          * @todo Check for unique id attribute :P
 735          */
 736         static function validateAttributes( $attribs, $whitelist ) {
 737                 $whitelist = array_flip( $whitelist );
 738                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 739
 740                 $out = [];
 741                 foreach ( $attribs as $attribute => $value ) {
 742                         # Allow XML namespace declaration to allow RDFa
 743                         if ( preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 744                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 745                                         $out[$attribute] = $value;
 746                                 }
 747
 748                                 continue;
 749                         }
 750
 751                         # Allow any attribute beginning with "data-"
 752                         # However:
 753                         # * data-ooui is reserved for ooui
 754                         # * data-mw and data-parsoid are reserved for parsoid
 755                         # * data-mw-<name here> is reserved for extensions (or core) if
 756                         #   they need to communicate some data to the client and want to be
 757                         #   sure that it isn't coming from an untrusted user.
 758                         # * Ensure that the attribute is not namespaced by banning
 759                         #   colons.
 760                         if ( !preg_match( '/^data-(?!ooui|mw|parsoid)[^:]*$/i', $attribute )
 761                                 && !isset( $whitelist[$attribute] )
 762                         ) {
 763                                 continue;
 764                         }
 765
 766                         # Strip javascript "expression" from stylesheets.
 767                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 768                         if ( $attribute == 'style' ) {
 769                                 $value = Sanitizer::checkCss( $value );
 770                         }
 771
 772                         # Escape HTML id attributes
 773                         if ( $attribute === 'id' ) {
 774                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 775                         }
 776
 777                         # Escape HTML id reference lists
 778                         if ( $attribute === 'aria-describedby'
 779                                 || $attribute === 'aria-flowto'
 780                                 || $attribute === 'aria-labelledby'
 781                                 || $attribute === 'aria-owns'
 782                         ) {
 783                                 $value = Sanitizer::escapeIdReferenceList( $value, 'noninitial' );
 784                         }
 785
 786                         // RDFa and microdata properties allow URLs, URIs and/or CURIs.
 787                         // Check them for sanity.
 788                         if ( $attribute === 'rel' || $attribute === 'rev'
 789                                 # RDFa
 790                                 || $attribute === 'about' || $attribute === 'property'
 791                                 || $attribute === 'resource' || $attribute === 'datatype'
 792                                 || $attribute === 'typeof'
 793                                 # HTML5 microdata
 794                                 || $attribute === 'itemid' || $attribute === 'itemprop'
 795                                 || $attribute === 'itemref' || $attribute === 'itemscope'
 796                                 || $attribute === 'itemtype'
 797                         ) {
 798                                 // Paranoia. Allow "simple" values but suppress javascript
 799                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 800                                         continue;
 801                                 }
 802                         }
 803
 804                         # NOTE: even though elements using href/src are not allowed directly, supply
 805                         #       validation code that can be used by tag hook handlers, etc
 806                         if ( $attribute === 'href' || $attribute === 'src' ) {
 807                                 if ( !preg_match( $hrefExp, $value ) ) {
 808                                         continue; // drop any href or src attributes not using an allowed protocol.
 809                                         // NOTE: this also drops all relative URLs
 810                                 }
 811                         }
 812
 813                         // If this attribute was previously set, override it.
 814                         // Output should only have one attribute of each name.
 815                         $out[$attribute] = $value;
 816                 }
 817
 818                 # itemtype, itemid, itemref don't make sense without itemscope
 819                 if ( !array_key_exists( 'itemscope', $out ) ) {
 820                         unset( $out['itemtype'] );
 821                         unset( $out['itemid'] );
 822                         unset( $out['itemref'] );
 823                 }
 824                 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 825
 826                 return $out;
 827         }
 828
 829         /**
 830          * Merge two sets of HTML attributes.  Conflicting items in the second set
 831          * will override those in the first, except for 'class' attributes which
 832          * will be combined (if they're both strings).
 833          *
 834          * @todo implement merging for other attributes such as style
 835          * @param array $a
 836          * @param array $b
 837          * @return array
 838          */
 839         static function mergeAttributes( $a, $b ) {
 840                 $out = array_merge( $a, $b );
 841                 if ( isset( $a['class'] ) && isset( $b['class'] )
 842                         && is_string( $a['class'] ) && is_string( $b['class'] )
 843                         && $a['class'] !== $b['class']
 844                 ) {
 845                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 846                                 -1, PREG_SPLIT_NO_EMPTY );
 847                         $out['class'] = implode( ' ', array_unique( $classes ) );
 848                 }
 849                 return $out;
 850         }
 851
 852         /**
 853          * Normalize CSS into a format we can easily search for hostile input
 854          *  - decode character references
 855          *  - decode escape sequences
 856          *  - convert characters that IE6 interprets into ascii
 857          *  - remove comments, unless the entire value is one single comment
 858          * @param string $value the css string
 859          * @return string normalized css
 860          */
 861         public static function normalizeCss( $value ) {
 862
 863                 // Decode character references like &#123;
 864                 $value = Sanitizer::decodeCharReferences( $value );
 865
 866                 // Decode escape sequences and line continuation
 867                 // See the grammar in the CSS 2 spec, appendix D.
 868                 // This has to be done AFTER decoding character references.
 869                 // This means it isn't possible for this function to return
 870                 // unsanitized escape sequences. It is possible to manufacture
 871                 // input that contains character references that decode to
 872                 // escape sequences that decode to character references, but
 873                 // it's OK for the return value to contain character references
 874                 // because the caller is supposed to escape those anyway.
 875                 static $decodeRegex;
 876                 if ( !$decodeRegex ) {
 877                         $space = '[\\x20\\t\\r\\n\\f]';
 878                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 879                         $backslash = '\\\\';
 880                         $decodeRegex = "/ $backslash
 881                                 (?:
 882                                         ($nl) |  # 1. Line continuation
 883                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 884                                         (.) | # 3. backslash cancelling special meaning
 885                                         () | # 4. backslash at end of string
 886                                 )/xu";
 887                 }
 888                 $value = preg_replace_callback( $decodeRegex,
 889                         [ __CLASS__, 'cssDecodeCallback' ], $value );
 890
 891                 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
 892                 $value = preg_replace_callback(
 893                         '/[！-［］-ｚ]/u', // U+FF01 to U+FF5A, excluding U+FF3C (bug 58088)
 894                         function ( $matches ) {
 895                                 $cp = UtfNormal\Utils::utf8ToCodepoint( $matches[0] );
 896                                 if ( $cp === false ) {
 897                                         return '';
 898                                 }
 899                                 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
 900                         },
 901                         $value
 902                 );
 903
 904                 // Convert more characters IE6 might treat as ascii
 905                 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
 906                 $value = str_replace(
 907                         [ 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ],
 908                         [ 'r', 'n', 'n', 'l', 'i', '(', '(' ],
 909                         $value
 910                 );
 911
 912                 // Let the value through if it's nothing but a single comment, to
 913                 // allow other functions which may reject it to pass some error
 914                 // message through.
 915                 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
 916                         // Remove any comments; IE gets token splitting wrong
 917                         // This must be done AFTER decoding character references and
 918                         // escape sequences, because those steps can introduce comments
 919                         // This step cannot introduce character references or escape
 920                         // sequences, because it replaces comments with spaces rather
 921                         // than removing them completely.
 922                         $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 923
 924                         // Remove anything after a comment-start token, to guard against
 925                         // incorrect client implementations.
 926                         $commentPos = strpos( $value, '/*' );
 927                         if ( $commentPos !== false ) {
 928                                 $value = substr( $value, 0, $commentPos );
 929                         }
 930                 }
 931
 932                 // S followed by repeat, iteration, or prolonged sound marks,
 933                 // which IE will treat as "ss"
 934                 $value = preg_replace(
 935                         '/s(?:
 936                                 \xE3\x80\xB1 | # U+3031
 937                                 \xE3\x82\x9D | # U+309D
 938                                 \xE3\x83\xBC | # U+30FC
 939                                 \xE3\x83\xBD | # U+30FD
 940                                 \xEF\xB9\xBC | # U+FE7C
 941                                 \xEF\xB9\xBD | # U+FE7D
 942                                 \xEF\xBD\xB0   # U+FF70
 943                         )/ix',
 944                         'ss',
 945                         $value
 946                 );
 947
 948                 return $value;
 949         }
 950
 951         /**
 952          * Pick apart some CSS and check it for forbidden or unsafe structures.
 953          * Returns a sanitized string. This sanitized string will have
 954          * character references and escape sequences decoded and comments
 955          * stripped (unless it is itself one valid comment, in which case the value
 956          * will be passed through). If the input is just too evil, only a comment
 957          * complaining about evilness will be returned.
 958          *
 959          * Currently URL references, 'expression', 'tps' are forbidden.
 960          *
 961          * NOTE: Despite the fact that character references are decoded, the
 962          * returned string may contain character references given certain
 963          * clever input strings. These character references must
 964          * be escaped before the return value is embedded in HTML.
 965          *
 966          * @param string $value
 967          * @return string
 968          */
 969         static function checkCss( $value ) {
 970                 $value = self::normalizeCss( $value );
 971
 972                 // Reject problematic keywords and control characters
 973                 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ||
 974                         strpos( $value, UtfNormal\Constants::UTF8_REPLACEMENT ) !== false ) {
 975                         return '/* invalid control char */';
 976                 } elseif ( preg_match(
 977                         '! expression
 978                                 | filter\s*:
 979                                 | accelerator\s*:
 980                                 | -o-link\s*:
 981                                 | -o-link-source\s*:
 982                                 | -o-replace\s*:
 983                                 | url\s*\(
 984                                 | image\s*\(
 985                                 | image-set\s*\(
 986                         !ix', $value ) ) {
 987                         return '/* insecure input */';
 988                 }
 989                 return $value;
 990         }
 991
 992         /**
 993          * @param array $matches
 994          * @return string
 995          */
 996         static function cssDecodeCallback( $matches ) {
 997                 if ( $matches[1] !== '' ) {
 998                         // Line continuation
 999                         return '';
1000                 } elseif ( $matches[2] !== '' ) {
1001                         $char = UtfNormal\Utils::codepointToUtf8( hexdec( $matches[2] ) );
1002                 } elseif ( $matches[3] !== '' ) {
1003                         $char = $matches[3];
1004                 } else {
1005                         $char = '\\';
1006                 }
1007                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
1008                         // These characters need to be escaped in strings
1009                         // Clean up the escape sequence to avoid parsing errors by clients
1010                         return '\\' . dechex( ord( $char ) ) . ' ';
1011                 } else {
1012                         // Decode unnecessary escape
1013                         return $char;
1014                 }
1015         }
1016
1017         /**
1018          * Take a tag soup fragment listing an HTML element's attributes
1019          * and normalize it to well-formed XML, discarding unwanted attributes.
1020          * Output is safe for further wikitext processing, with escaping of
1021          * values that could trigger problems.
1022          *
1023          * - Normalizes attribute names to lowercase
1024          * - Discards attributes not on a whitelist for the given element
1025          * - Turns broken or invalid entities into plaintext
1026          * - Double-quotes all attribute values
1027          * - Attributes without values are given the name as attribute
1028          * - Double attributes are discarded
1029          * - Unsafe style attributes are discarded
1030          * - Prepends space if there are attributes.
1031          *
1032          * @param string $text
1033          * @param string $element
1034          * @return string
1035          */
1036         static function fixTagAttributes( $text, $element ) {
1037                 if ( trim( $text ) == '' ) {
1038                         return '';
1039                 }
1040
1041                 $decoded = Sanitizer::decodeTagAttributes( $text );
1042                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1043
1044                 return Sanitizer::safeEncodeTagAttributes( $stripped );
1045         }
1046
1047         /**
1048          * Encode an attribute value for HTML output.
1049          * @param string $text
1050          * @return string HTML-encoded text fragment
1051          */
1052         static function encodeAttribute( $text ) {
1053                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1054
1055                 // Whitespace is normalized during attribute decoding,
1056                 // so if we've been passed non-spaces we must encode them
1057                 // ahead of time or they won't be preserved.
1058                 $encValue = strtr( $encValue, [
1059                         "\n" => '&#10;',
1060                         "\r" => '&#13;',
1061                         "\t" => '&#9;',
1062                 ] );
1063
1064                 return $encValue;
1065         }
1066
1067         /**
1068          * Encode an attribute value for HTML tags, with extra armoring
1069          * against further wiki processing.
1070          * @param string $text
1071          * @return string HTML-encoded text fragment
1072          */
1073         static function safeEncodeAttribute( $text ) {
1074                 $encValue = Sanitizer::encodeAttribute( $text );
1075
1076                 # Templates and links may be expanded in later parsing,
1077                 # creating invalid or dangerous output. Suppress this.
1078                 $encValue = strtr( $encValue, [
1079                         '<'    => '&lt;',   // This should never happen,
1080                         '>'    => '&gt;',   // we've received invalid input
1081                         '"'    => '&quot;', // which should have been escaped.
1082                         '{'    => '&#123;',
1083                         '['    => '&#91;',
1084                         "''"   => '&#39;&#39;',
1085                         'ISBN' => '&#73;SBN',
1086                         'RFC'  => '&#82;FC',
1087                         'PMID' => '&#80;MID',
1088                         '|'    => '&#124;',
1089                         '__'   => '&#95;_',
1090                 ] );
1091
1092                 # Stupid hack
1093                 $encValue = preg_replace_callback(
1094                         '/((?i)' . wfUrlProtocols() . ')/',
1095                         [ 'Sanitizer', 'armorLinksCallback' ],
1096                         $encValue );
1097                 return $encValue;
1098         }
1099
1100         /**
1101          * Given a value, escape it so that it can be used in an id attribute and
1102          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1103          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1104          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1105          * escaped with lots of dots.
1106          *
1107          * To ensure we don't have to bother escaping anything, we also strip ', ",
1108          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1109          * We also strip # because it upsets IE, and % because it could be
1110          * ambiguous if it's part of something that looks like a percent escape
1111          * (which don't work reliably in fragments cross-browser).
1112          *
1113          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1114          *   in the id and name attributes
1115          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1116          *   the id attribute
1117          * @see http://www.whatwg.org/html/elements.html#the-id-attribute
1118          *   HTML5 definition of id attribute
1119          *
1120          * @param string $id Id to escape
1121          * @param string|array $options String or array of strings (default is array()):
1122          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1123          *       so don't pay attention if the first character isn't valid at the
1124          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1125          *       false.
1126          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1127          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1128          *       anchors and links won't break.
1129          * @return string
1130          */
1131         static function escapeId( $id, $options = [] ) {
1132                 global $wgExperimentalHtmlIds;
1133                 $options = (array)$options;
1134
1135                 $id = Sanitizer::decodeCharReferences( $id );
1136
1137                 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1138                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1139                         $id = trim( $id, '_' );
1140                         if ( $id === '' ) {
1141                                 // Must have been all whitespace to start with.
1142                                 return '_';
1143                         } else {
1144                                 return $id;
1145                         }
1146                 }
1147
1148                 // HTML4-style escaping
1149                 static $replace = [
1150                         '%3A' => ':',
1151                         '%' => '.'
1152                 ];
1153
1154                 $id = urlencode( strtr( $id, ' ', '_' ) );
1155                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1156
1157                 if ( !preg_match( '/^[a-zA-Z]/', $id ) && !in_array( 'noninitial', $options ) ) {
1158                         // Initial character must be a letter!
1159                         $id = "x$id";
1160                 }
1161                 return $id;
1162         }
1163
1164         /**
1165          * Given a string containing a space delimited list of ids, escape each id
1166          * to match ids escaped by the escapeId() function.
1167          *
1168          * @since 1.27
1169          *
1170          * @param string $referenceString Space delimited list of ids
1171          * @param string|array $options String or array of strings (default is array()):
1172          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1173          *       so don't pay attention if the first character isn't valid at the
1174          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1175          *       false.
1176          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1177          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1178          *       anchors and links won't break.
1179          * @return string
1180          */
1181         static function escapeIdReferenceList( $referenceString, $options = [] ) {
1182                 # Explode the space delimited list string into an array of tokens
1183                 $references = preg_split( '/\s+/', "{$referenceString}", -1, PREG_SPLIT_NO_EMPTY );
1184
1185                 # Escape each token as an id
1186                 foreach ( $references as &$ref ) {
1187                         $ref = Sanitizer::escapeId( $ref, $options );
1188                 }
1189
1190                 # Merge the array back to a space delimited list string
1191                 # If the array is empty, the result will be an empty string ('')
1192                 $referenceString = implode( ' ', $references );
1193
1194                 return $referenceString;
1195         }
1196
1197         /**
1198          * Given a value, escape it so that it can be used as a CSS class and
1199          * return it.
1200          *
1201          * @todo For extra validity, input should be validated UTF-8.
1202          *
1203          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1204          *
1205          * @param string $class
1206          * @return string
1207          */
1208         static function escapeClass( $class ) {
1209                 // Convert ugly stuff to underscores and kill underscores in ugly places
1210                 return rtrim( preg_replace(
1211                         [ '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ],
1212                         '_',
1213                         $class ), '_' );
1214         }
1215
1216         /**
1217          * Given HTML input, escape with htmlspecialchars but un-escape entities.
1218          * This allows (generally harmless) entities like &#160; to survive.
1219          *
1220          * @param string $html HTML to escape
1221          * @return string Escaped input
1222          */
1223         static function escapeHtmlAllowEntities( $html ) {
1224                 $html = Sanitizer::decodeCharReferences( $html );
1225                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1226                 # hurt.
1227                 $html = htmlspecialchars( $html, ENT_QUOTES );
1228                 return $html;
1229         }
1230
1231         /**
1232          * Regex replace callback for armoring links against further processing.
1233          * @param array $matches
1234          * @return string
1235          */
1236         private static function armorLinksCallback( $matches ) {
1237                 return str_replace( ':', '&#58;', $matches[1] );
1238         }
1239
1240         /**
1241          * Return an associative array of attribute names and values from
1242          * a partial tag string. Attribute names are forced to lowercase,
1243          * character references are decoded to UTF-8 text.
1244          *
1245          * @param string $text
1246          * @return array
1247          */
1248         public static function decodeTagAttributes( $text ) {
1249                 if ( trim( $text ) == '' ) {
1250                         return [];
1251                 }
1252
1253                 $attribs = [];
1254                 $pairs = [];
1255                 if ( !preg_match_all(
1256                         self::getAttribsRegex(),
1257                         $text,
1258                         $pairs,
1259                         PREG_SET_ORDER ) ) {
1260                         return $attribs;
1261                 }
1262
1263                 foreach ( $pairs as $set ) {
1264                         $attribute = strtolower( $set[1] );
1265                         $value = Sanitizer::getTagAttributeCallback( $set );
1266
1267                         // Normalize whitespace
1268                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1269                         $value = trim( $value );
1270
1271                         // Decode character references
1272                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1273                 }
1274                 return $attribs;
1275         }
1276
1277         /**
1278          * Build a partial tag string from an associative array of attribute
1279          * names and values as returned by decodeTagAttributes.
1280          *
1281          * @param array $assoc_array
1282          * @return string
1283          */
1284         public static function safeEncodeTagAttributes( $assoc_array ) {
1285                 $attribs = [];
1286                 foreach ( $assoc_array as $attribute => $value ) {
1287                         $encAttribute = htmlspecialchars( $attribute );
1288                         $encValue = Sanitizer::safeEncodeAttribute( $value );
1289
1290                         $attribs[] = "$encAttribute=\"$encValue\"";
1291                 }
1292                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1293         }
1294
1295         /**
1296          * Pick the appropriate attribute value from a match set from the
1297          * attribs regex matches.
1298          *
1299          * @param array $set
1300          * @throws MWException When tag conditions are not met.
1301          * @return string
1302          */
1303         private static function getTagAttributeCallback( $set ) {
1304                 if ( isset( $set[5] ) ) {
1305                         # No quotes.
1306                         return $set[5];
1307                 } elseif ( isset( $set[4] ) ) {
1308                         # Single-quoted
1309                         return $set[4];
1310                 } elseif ( isset( $set[3] ) ) {
1311                         # Double-quoted
1312                         return $set[3];
1313                 } elseif ( !isset( $set[2] ) ) {
1314                         # In XHTML, attributes must have a value so return an empty string.
1315                         # See "Empty attribute syntax",
1316                         # http://www.w3.org/TR/html5/syntax.html#syntax-attribute-name
1317                         return "";
1318                 } else {
1319                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1320                 }
1321         }
1322
1323         /**
1324          * @param string $text
1325          * @return string
1326          */
1327         private static function normalizeWhitespace( $text ) {
1328                 return preg_replace(
1329                         '/\r\n|[\x20\x0d\x0a\x09]/',
1330                         ' ',
1331                         $text );
1332         }
1333
1334         /**
1335          * Normalizes whitespace in a section name, such as might be returned
1336          * by Parser::stripSectionName(), for use in the id's that are used for
1337          * section links.
1338          *
1339          * @param string $section
1340          * @return string
1341          */
1342         static function normalizeSectionNameWhitespace( $section ) {
1343                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1344         }
1345
1346         /**
1347          * Ensure that any entities and character references are legal
1348          * for XML and XHTML specifically. Any stray bits will be
1349          * &amp;-escaped to result in a valid text fragment.
1350          *
1351          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1352          *   numericized (this way we're well-formed even without a DTD)
1353          * b. any numeric char refs must be legal chars, not invalid or forbidden
1354          * c. use lower cased "&#x", not "&#X"
1355          * d. fix or reject non-valid attributes
1356          *
1357          * @param string $text
1358          * @return string
1359          * @private
1360          */
1361         static function normalizeCharReferences( $text ) {
1362                 return preg_replace_callback(
1363                         self::CHAR_REFS_REGEX,
1364                         [ 'Sanitizer', 'normalizeCharReferencesCallback' ],
1365                         $text );
1366         }
1367
1368         /**
1369          * @param string $matches
1370          * @return string
1371          */
1372         static function normalizeCharReferencesCallback( $matches ) {
1373                 $ret = null;
1374                 if ( $matches[1] != '' ) {
1375                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1376                 } elseif ( $matches[2] != '' ) {
1377                         $ret = Sanitizer::decCharReference( $matches[2] );
1378                 } elseif ( $matches[3] != '' ) {
1379                         $ret = Sanitizer::hexCharReference( $matches[3] );
1380                 }
1381                 if ( is_null( $ret ) ) {
1382                         return htmlspecialchars( $matches[0] );
1383                 } else {
1384                         return $ret;
1385                 }
1386         }
1387
1388         /**
1389          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1390          * return the equivalent numeric entity reference (except for the core &lt;
1391          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1392          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1393          * pseudo-entity source (eg &amp;foo;)
1394          *
1395          * @param string $name
1396          * @return string
1397          */
1398         static function normalizeEntity( $name ) {
1399                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1400                         return '&' . self::$htmlEntityAliases[$name] . ';';
1401                 } elseif ( in_array( $name, [ 'lt', 'gt', 'amp', 'quot' ] ) ) {
1402                         return "&$name;";
1403                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1404                         return '&#' . self::$htmlEntities[$name] . ';';
1405                 } else {
1406                         return "&amp;$name;";
1407                 }
1408         }
1409
1410         /**
1411          * @param int $codepoint
1412          * @return null|string
1413          */
1414         static function decCharReference( $codepoint ) {
1415                 $point = intval( $codepoint );
1416                 if ( Sanitizer::validateCodepoint( $point ) ) {
1417                         return sprintf( '&#%d;', $point );
1418                 } else {
1419                         return null;
1420                 }
1421         }
1422
1423         /**
1424          * @param int $codepoint
1425          * @return null|string
1426          */
1427         static function hexCharReference( $codepoint ) {
1428                 $point = hexdec( $codepoint );
1429                 if ( Sanitizer::validateCodepoint( $point ) ) {
1430                         return sprintf( '&#x%x;', $point );
1431                 } else {
1432                         return null;
1433                 }
1434         }
1435
1436         /**
1437          * Returns true if a given Unicode codepoint is a valid character in
1438          * both HTML5 and XML.
1439          * @param int $codepoint
1440          * @return bool
1441          */
1442         private static function validateCodepoint( $codepoint ) {
1443                 # U+000C is valid in HTML5 but not allowed in XML.
1444                 # U+000D is valid in XML but not allowed in HTML5.
1445                 # U+007F - U+009F are disallowed in HTML5 (control characters).
1446                 return $codepoint == 0x09
1447                         || $codepoint == 0x0a
1448                         || ( $codepoint >= 0x20 && $codepoint <= 0x7e )
1449                         || ( $codepoint >= 0xa0 && $codepoint <= 0xd7ff )
1450                         || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1451                         || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1452         }
1453
1454         /**
1455          * Decode any character references, numeric or named entities,
1456          * in the text and return a UTF-8 string.
1457          *
1458          * @param string $text
1459          * @return string
1460          */
1461         public static function decodeCharReferences( $text ) {
1462                 return preg_replace_callback(
1463                         self::CHAR_REFS_REGEX,
1464                         [ 'Sanitizer', 'decodeCharReferencesCallback' ],
1465                         $text );
1466         }
1467
1468         /**
1469          * Decode any character references, numeric or named entities,
1470          * in the next and normalize the resulting string. (bug 14952)
1471          *
1472          * This is useful for page titles, not for text to be displayed,
1473          * MediaWiki allows HTML entities to escape normalization as a feature.
1474          *
1475          * @param string $text Already normalized, containing entities
1476          * @return string Still normalized, without entities
1477          */
1478         public static function decodeCharReferencesAndNormalize( $text ) {
1479                 global $wgContLang;
1480                 $text = preg_replace_callback(
1481                         self::CHAR_REFS_REGEX,
1482                         [ 'Sanitizer', 'decodeCharReferencesCallback' ],
1483                         $text, /* limit */ -1, $count );
1484
1485                 if ( $count ) {
1486                         return $wgContLang->normalize( $text );
1487                 } else {
1488                         return $text;
1489                 }
1490         }
1491
1492         /**
1493          * @param string $matches
1494          * @return string
1495          */
1496         static function decodeCharReferencesCallback( $matches ) {
1497                 if ( $matches[1] != '' ) {
1498                         return Sanitizer::decodeEntity( $matches[1] );
1499                 } elseif ( $matches[2] != '' ) {
1500                         return Sanitizer::decodeChar( intval( $matches[2] ) );
1501                 } elseif ( $matches[3] != '' ) {
1502                         return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1503                 }
1504                 # Last case should be an ampersand by itself
1505                 return $matches[0];
1506         }
1507
1508         /**
1509          * Return UTF-8 string for a codepoint if that is a valid
1510          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1511          * @param int $codepoint
1512          * @return string
1513          * @private
1514          */
1515         static function decodeChar( $codepoint ) {
1516                 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1517                         return UtfNormal\Utils::codepointToUtf8( $codepoint );
1518                 } else {
1519                         return UtfNormal\Constants::UTF8_REPLACEMENT;
1520                 }
1521         }
1522
1523         /**
1524          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1525          * return the UTF-8 encoding of that character. Otherwise, returns
1526          * pseudo-entity source (eg "&foo;")
1527          *
1528          * @param string $name
1529          * @return string
1530          */
1531         static function decodeEntity( $name ) {
1532                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1533                         $name = self::$htmlEntityAliases[$name];
1534                 }
1535                 if ( isset( self::$htmlEntities[$name] ) ) {
1536                         return UtfNormal\Utils::codepointToUtf8( self::$htmlEntities[$name] );
1537                 } else {
1538                         return "&$name;";
1539                 }
1540         }
1541
1542         /**
1543          * Fetch the whitelist of acceptable attributes for a given element name.
1544          *
1545          * @param string $element
1546          * @return array
1547          */
1548         static function attributeWhitelist( $element ) {
1549                 $list = Sanitizer::setupAttributeWhitelist();
1550                 return isset( $list[$element] )
1551                         ? $list[$element]
1552                         : [];
1553         }
1554
1555         /**
1556          * Foreach array key (an allowed HTML element), return an array
1557          * of allowed attributes
1558          * @return array
1559          */
1560         static function setupAttributeWhitelist() {
1561                 static $whitelist;
1562
1563                 if ( $whitelist !== null ) {
1564                         return $whitelist;
1565                 }
1566
1567                 $common = [
1568                         # HTML
1569                         'id',
1570                         'class',
1571                         'style',
1572                         'lang',
1573                         'dir',
1574                         'title',
1575
1576                         # WAI-ARIA
1577                         'aria-describedby',
1578                         'aria-flowto',
1579                         'aria-label',
1580                         'aria-labelledby',
1581                         'aria-owns',
1582                         'role',
1583
1584                         # RDFa
1585                         # These attributes are specified in section 9 of
1586                         # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1587                         'about',
1588                         'property',
1589                         'resource',
1590                         'datatype',
1591                         'typeof',
1592
1593                         # Microdata. These are specified by
1594                         # http://www.whatwg.org/html/microdata.html#the-microdata-model
1595                         'itemid',
1596                         'itemprop',
1597                         'itemref',
1598                         'itemscope',
1599                         'itemtype',
1600                 ];
1601
1602                 $block = array_merge( $common, [ 'align' ] );
1603                 $tablealign = [ 'align', 'valign' ];
1604                 $tablecell = [
1605                         'abbr',
1606                         'axis',
1607                         'headers',
1608                         'scope',
1609                         'rowspan',
1610                         'colspan',
1611                         'nowrap', # deprecated
1612                         'width', # deprecated
1613                         'height', # deprecated
1614                         'bgcolor', # deprecated
1615                 ];
1616
1617                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1618                 # See: http://www.w3.org/TR/html4/
1619                 $whitelist = [
1620                         # 7.5.4
1621                         'div'        => $block,
1622                         'center'     => $common, # deprecated
1623                         'span'       => $common,
1624
1625                         # 7.5.5
1626                         'h1'         => $block,
1627                         'h2'         => $block,
1628                         'h3'         => $block,
1629                         'h4'         => $block,
1630                         'h5'         => $block,
1631                         'h6'         => $block,
1632
1633                         # 7.5.6
1634                         # address
1635
1636                         # 8.2.4
1637                         'bdo'        => $common,
1638
1639                         # 9.2.1
1640                         'em'         => $common,
1641                         'strong'     => $common,
1642                         'cite'       => $common,
1643                         'dfn'        => $common,
1644                         'code'       => $common,
1645                         'samp'       => $common,
1646                         'kbd'        => $common,
1647                         'var'        => $common,
1648                         'abbr'       => $common,
1649                         # acronym
1650
1651                         # 9.2.2
1652                         'blockquote' => array_merge( $common, [ 'cite' ] ),
1653                         'q'          => array_merge( $common, [ 'cite' ] ),
1654
1655                         # 9.2.3
1656                         'sub'        => $common,
1657                         'sup'        => $common,
1658
1659                         # 9.3.1
1660                         'p'          => $block,
1661
1662                         # 9.3.2
1663                         'br'         => array_merge( $common, [ 'clear' ] ),
1664
1665                         # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1666                         'wbr'        => $common,
1667
1668                         # 9.3.4
1669                         'pre'        => array_merge( $common, [ 'width' ] ),
1670
1671                         # 9.4
1672                         'ins'        => array_merge( $common, [ 'cite', 'datetime' ] ),
1673                         'del'        => array_merge( $common, [ 'cite', 'datetime' ] ),
1674
1675                         # 10.2
1676                         'ul'         => array_merge( $common, [ 'type' ] ),
1677                         'ol'         => array_merge( $common, [ 'type', 'start', 'reversed' ] ),
1678                         'li'         => array_merge( $common, [ 'type', 'value' ] ),
1679
1680                         # 10.3
1681                         'dl'         => $common,
1682                         'dd'         => $common,
1683                         'dt'         => $common,
1684
1685                         # 11.2.1
1686                         'table'      => array_merge( $common,
1687                                                                 [ 'summary', 'width', 'border', 'frame',
1688                                                                                 'rules', 'cellspacing', 'cellpadding',
1689                                                                                 'align', 'bgcolor',
1690                                                                 ] ),
1691
1692                         # 11.2.2
1693                         'caption'    => $block,
1694
1695                         # 11.2.3
1696                         'thead'      => $common,
1697                         'tfoot'      => $common,
1698                         'tbody'      => $common,
1699
1700                         # 11.2.4
1701                         'colgroup'   => array_merge( $common, [ 'span' ] ),
1702                         'col'        => array_merge( $common, [ 'span' ] ),
1703
1704                         # 11.2.5
1705                         'tr'         => array_merge( $common, [ 'bgcolor' ], $tablealign ),
1706
1707                         # 11.2.6
1708                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1709                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1710
1711                         # 12.2
1712                         # NOTE: <a> is not allowed directly, but the attrib
1713                         # whitelist is used from the Parser object
1714                         'a'          => array_merge( $common, [ 'href', 'rel', 'rev' ] ), # rel/rev esp. for RDFa
1715
1716                         # 13.2
1717                         # Not usually allowed, but may be used for extension-style hooks
1718                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1719                         # true
1720                         'img'        => array_merge( $common, [ 'alt', 'src', 'width', 'height' ] ),
1721
1722                         # 15.2.1
1723                         'tt'         => $common,
1724                         'b'          => $common,
1725                         'i'          => $common,
1726                         'big'        => $common,
1727                         'small'      => $common,
1728                         'strike'     => $common,
1729                         's'          => $common,
1730                         'u'          => $common,
1731
1732                         # 15.2.2
1733                         'font'       => array_merge( $common, [ 'size', 'color', 'face' ] ),
1734                         # basefont
1735
1736                         # 15.3
1737                         'hr'         => array_merge( $common, [ 'width' ] ),
1738
1739                         # HTML Ruby annotation text module, simple ruby only.
1740                         # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
1741                         'ruby'       => $common,
1742                         # rbc
1743                         'rb'         => $common,
1744                         'rp'         => $common,
1745                         'rt'         => $common, # array_merge( $common, array( 'rbspan' ) ),
1746                         'rtc'         => $common,
1747
1748                         # MathML root element, where used for extensions
1749                         # 'title' may not be 100% valid here; it's XHTML
1750                         # http://www.w3.org/TR/REC-MathML/
1751                         'math'       => [ 'class', 'style', 'id', 'title' ],
1752
1753                         # HTML 5 section 4.6
1754                         'bdi' => $common,
1755
1756                         # HTML5 elements, defined by:
1757                         # http://www.whatwg.org/html/
1758                         'data' => array_merge( $common, [ 'value' ] ),
1759                         'time' => array_merge( $common, [ 'datetime' ] ),
1760                         'mark' => $common,
1761
1762                         // meta and link are only permitted by removeHTMLtags when Microdata
1763                         // is enabled so we don't bother adding a conditional to hide these
1764                         // Also meta and link are only valid in WikiText as Microdata elements
1765                         // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1766                         // So we don't bother including $common attributes that have no purpose.
1767                         'meta' => [ 'itemprop', 'content' ],
1768                         'link' => [ 'itemprop', 'href' ],
1769                 ];
1770
1771                 return $whitelist;
1772         }
1773
1774         /**
1775          * Take a fragment of (potentially invalid) HTML and return
1776          * a version with any tags removed, encoded as plain text.
1777          *
1778          * Warning: this return value must be further escaped for literal
1779          * inclusion in HTML output as of 1.10!
1780          *
1781          * @param string $text HTML fragment
1782          * @return string
1783          */
1784         static function stripAllTags( $text ) {
1785                 # Actual <tags>
1786                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1787
1788                 # Normalize &entities and whitespace
1789                 $text = self::decodeCharReferences( $text );
1790                 $text = self::normalizeWhitespace( $text );
1791
1792                 return $text;
1793         }
1794
1795         /**
1796          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1797          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1798          * PHP 5.1 doesn't.
1799          *
1800          * Use for passing XHTML fragments to PHP's XML parsing functions
1801          *
1802          * @return string
1803          */
1804         static function hackDocType() {
1805                 $out = "<!DOCTYPE html [\n";
1806                 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1807                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1808                 }
1809                 $out .= "]>\n";
1810                 return $out;
1811         }
1812
1813         /**
1814          * @param string $url
1815          * @return mixed|string
1816          */
1817         static function cleanUrl( $url ) {
1818                 # Normalize any HTML entities in input. They will be
1819                 # re-escaped by makeExternalLink().
1820                 $url = Sanitizer::decodeCharReferences( $url );
1821
1822                 # Escape any control characters introduced by the above step
1823                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1824                         [ __CLASS__, 'cleanUrlCallback' ], $url );
1825
1826                 # Validate hostname portion
1827                 $matches = [];
1828                 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1829                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1830
1831                         // Characters that will be ignored in IDNs.
1832                         // http://tools.ietf.org/html/3454#section-3.1
1833                         // Strip them before further processing so blacklists and such work.
1834                         $strip = "/
1835                                 \\s|          # general whitespace
1836                                 \xc2\xad|     # 00ad SOFT HYPHEN
1837                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1838                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1839                                 \xe2\x81\xa0| # 2060 WORD JOINER
1840                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1841                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1842                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1843                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1844                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1845                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1846                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1847                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1848                                 /xuD";
1849
1850                         $host = preg_replace( $strip, '', $host );
1851
1852                         // IPv6 host names are bracketed with [].  Url-decode these.
1853                         if ( substr_compare( "//%5B", $host, 0, 5 ) === 0 &&
1854                                 preg_match( '!^//%5B([0-9A-Fa-f:.]+)%5D((:\d+)?)$!', $host, $matches )
1855                         ) {
1856                                 $host = '//[' . $matches[1] . ']' . $matches[2];
1857                         }
1858
1859                         // @todo FIXME: Validate hostnames here
1860
1861                         return $protocol . $host . $rest;
1862                 } else {
1863                         return $url;
1864                 }
1865         }
1866
1867         /**
1868          * @param array $matches
1869          * @return string
1870          */
1871         static function cleanUrlCallback( $matches ) {
1872                 return urlencode( $matches[0] );
1873         }
1874
1875         /**
1876          * Does a string look like an e-mail address?
1877          *
1878          * This validates an email address using an HTML5 specification found at:
1879          * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1880          * Which as of 2011-01-24 says:
1881          *
1882          *   A valid e-mail address is a string that matches the ABNF production
1883          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1884          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1885          *   3.5.
1886          *
1887          * This function is an implementation of the specification as requested in
1888          * bug 22449.
1889          *
1890          * Client-side forms will use the same standard validation rules via JS or
1891          * HTML 5 validation; additional restrictions can be enforced server-side
1892          * by extensions via the 'isValidEmailAddr' hook.
1893          *
1894          * Note that this validation doesn't 100% match RFC 2822, but is believed
1895          * to be liberal enough for wide use. Some invalid addresses will still
1896          * pass validation here.
1897          *
1898          * @since 1.18
1899          *
1900          * @param string $addr E-mail address
1901          * @return bool
1902          */
1903         public static function validateEmail( $addr ) {
1904                 $result = null;
1905                 if ( !Hooks::run( 'isValidEmailAddr', [ $addr, &$result ] ) ) {
1906                         return $result;
1907                 }
1908
1909                 // Please note strings below are enclosed in brackets [], this make the
1910                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1911                 // See bug 26948
1912                 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1913                 $rfc1034_ldh_str = "a-z0-9\\-";
1914
1915                 $html5_email_regexp = "/
1916                 ^                      # start of string
1917                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1918                 @                      # 'apostrophe'
1919                 [$rfc1034_ldh_str]+       # First domain part
1920                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1921                 $                      # End of string
1922                 /ix"; // case Insensitive, eXtended
1923
1924                 return (bool)preg_match( $html5_email_regexp, $addr );
1925         }
1926 }