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          * https://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          * https://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 $attribsRegex;
 332
 333         /**
 334          * Regular expression to match HTML/XML attribute pairs within a tag.
 335          * Allows some... latitude. Based on,
 336          * https://www.w3.org/TR/html5/syntax.html#before-attribute-value-state
 337          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 338          * @return string
 339          */
 340         static function getAttribsRegex() {
 341                 if ( self::$attribsRegex === null ) {
 342                         $attribFirst = "[:_\p{L}\p{N}]";
 343                         $attrib = "[:_\.\-\p{L}\p{N}]";
 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                                 )?(?=$space|\$)/sxu";
 355                 }
 356                 return self::$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', 'meta', 'link'
 385                         ];
 386
 387                         # Elements that cannot have close tags. This is (not coincidentally)
 388                         # also the list of tags for which the HTML 5 parsing algorithm
 389                         # requires you to "acknowledge the token's self-closing flag", i.e.
 390                         # a self-closing tag like <br/> is not an HTML 5 parse error only
 391                         # for this list.
 392                         $htmlsingleonly = [
 393                                 'br', 'wbr', 'hr', 'meta', 'link'
 394                         ];
 395
 396                         $htmlnest = [ # Tags that can be nested--??
 397                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 398                                 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
 399                                 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
 400                         ];
 401                         $tabletags = [ # Can only appear inside table, we will close them
 402                                 'td', 'th', 'tr',
 403                         ];
 404                         $htmllist = [ # Tags used by list
 405                                 'ul', 'ol',
 406                         ];
 407                         $listtags = [ # Tags that can appear in a list
 408                                 'li',
 409                         ];
 410
 411                         if ( $wgAllowImageTag ) {
 412                                 $htmlsingle[] = 'img';
 413                                 $htmlsingleonly[] = 'img';
 414                         }
 415
 416                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 417                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 418
 419                         # Convert them all to hashtables for faster lookup
 420                         $vars = [ 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 421                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' ];
 422                         foreach ( $vars as $var ) {
 423                                 $$var = array_flip( $$var );
 424                         }
 425                         $staticInitialised = $globalContext;
 426                 }
 427
 428                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 429                 $extratags = array_flip( $extratags );
 430                 $removetags = array_flip( $removetags );
 431                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 432                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
 433
 434                 return [
 435                         'htmlpairs' => $htmlpairs,
 436                         'htmlsingle' => $htmlsingle,
 437                         'htmlsingleonly' => $htmlsingleonly,
 438                         'htmlnest' => $htmlnest,
 439                         'tabletags' => $tabletags,
 440                         'htmllist' => $htmllist,
 441                         'listtags' => $listtags,
 442                         'htmlsingleallowed' => $htmlsingleallowed,
 443                         'htmlelements' => $htmlelements,
 444                 ];
 445         }
 446
 447         /**
 448          * Cleans up HTML, removes dangerous tags and attributes, and
 449          * removes HTML comments
 450          * @param string $text
 451          * @param callable $processCallback Callback to do any variable or parameter
 452          *   replacements in HTML attribute values
 453          * @param array|bool $args Arguments for the processing callback
 454          * @param array $extratags For any extra tags to include
 455          * @param array $removetags For any tags (default or extra) to exclude
 456          * @param callable $warnCallback (Deprecated) Callback allowing the
 457          *   addition of a tracking category when bad input is encountered.
 458          *   DO NOT ADD NEW PARAMETERS AFTER $warnCallback, since it will be
 459          *   removed shortly.
 460          * @return string
 461          */
 462         public static function removeHTMLtags( $text, $processCallback = null,
 463                 $args = [], $extratags = [], $removetags = [], $warnCallback = null
 464         ) {
 465                 extract( self::getRecognizedTagData( $extratags, $removetags ) );
 466
 467                 # Remove HTML comments
 468                 $text = Sanitizer::removeHTMLcomments( $text );
 469                 $bits = explode( '<', $text );
 470                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 471                 if ( !MWTidy::isEnabled() ) {
 472                         $tagstack = $tablestack = [];
 473                         foreach ( $bits as $x ) {
 474                                 $regs = [];
 475                                 # $slash: Does the current element start with a '/'?
 476                                 # $t: Current element name
 477                                 # $params: String between element name and >
 478                                 # $brace: Ending '>' or '/>'
 479                                 # $rest: Everything until the next element of $bits
 480                                 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 481                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 482                                 } else {
 483                                         $slash = $t = $params = $brace = $rest = null;
 484                                 }
 485
 486                                 $badtag = false;
 487                                 $t = strtolower( $t );
 488                                 if ( isset( $htmlelements[$t] ) ) {
 489                                         # Check our stack
 490                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 491                                                 $badtag = true;
 492                                         } elseif ( $slash ) {
 493                                                 # Closing a tag... is it the one we just opened?
 494                                                 MediaWiki\suppressWarnings();
 495                                                 $ot = array_pop( $tagstack );
 496                                                 MediaWiki\restoreWarnings();
 497
 498                                                 if ( $ot != $t ) {
 499                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 500                                                                 # Pop all elements with an optional close tag
 501                                                                 # and see if we find a match below them
 502                                                                 $optstack = [];
 503                                                                 array_push( $optstack, $ot );
 504                                                                 MediaWiki\suppressWarnings();
 505                                                                 $ot = array_pop( $tagstack );
 506                                                                 MediaWiki\restoreWarnings();
 507                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 508                                                                         array_push( $optstack, $ot );
 509                                                                         MediaWiki\suppressWarnings();
 510                                                                         $ot = array_pop( $tagstack );
 511                                                                         MediaWiki\restoreWarnings();
 512                                                                 }
 513                                                                 if ( $t != $ot ) {
 514                                                                         # No match. Push the optional elements back again
 515                                                                         $badtag = true;
 516                                                                         MediaWiki\suppressWarnings();
 517                                                                         $ot = array_pop( $optstack );
 518                                                                         MediaWiki\restoreWarnings();
 519                                                                         while ( $ot ) {
 520                                                                                 array_push( $tagstack, $ot );
 521                                                                                 MediaWiki\suppressWarnings();
 522                                                                                 $ot = array_pop( $optstack );
 523                                                                                 MediaWiki\restoreWarnings();
 524                                                                         }
 525                                                                 }
 526                                                         } else {
 527                                                                 MediaWiki\suppressWarnings();
 528                                                                 array_push( $tagstack, $ot );
 529                                                                 MediaWiki\restoreWarnings();
 530
 531                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 532                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 533                                                                         $badtag = true;
 534                                                                 }
 535                                                         }
 536                                                 } else {
 537                                                         if ( $t == 'table' ) {
 538                                                                 $tagstack = array_pop( $tablestack );
 539                                                         }
 540                                                 }
 541                                                 $newparams = '';
 542                                         } else {
 543                                                 # Keep track for later
 544                                                 if ( isset( $tabletags[$t] ) && !in_array( 'table', $tagstack ) ) {
 545                                                         $badtag = true;
 546                                                 } elseif ( in_array( $t, $tagstack ) && !isset( $htmlnest[$t] ) ) {
 547                                                         $badtag = true;
 548                                                 #  Is it a self closed htmlpair ? (T7487)
 549                                                 } elseif ( $brace == '/>' && isset( $htmlpairs[$t] ) ) {
 550                                                         // Eventually we'll just remove the self-closing
 551                                                         // slash, in order to be consistent with HTML5
 552                                                         // semantics.
 553                                                         // $brace = '>';
 554                                                         // For now, let's just warn authors to clean up.
 555                                                         if ( is_callable( $warnCallback ) ) {
 556                                                                 call_user_func_array( $warnCallback, [ 'deprecated-self-close-category' ] );
 557                                                         }
 558                                                         $badtag = true;
 559                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 560                                                         # Hack to force empty tag for unclosable elements
 561                                                         $brace = '/>';
 562                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 563                                                         # Hack to not close $htmlsingle tags
 564                                                         $brace = null;
 565                                                         # Still need to push this optionally-closed tag to
 566                                                         # the tag stack so that we can match end tags
 567                                                         # instead of marking them as bad.
 568                                                         array_push( $tagstack, $t );
 569                                                 } elseif ( isset( $tabletags[$t] ) && in_array( $t, $tagstack ) ) {
 570                                                         // New table tag but forgot to close the previous one
 571                                                         $text .= "</$t>";
 572                                                 } else {
 573                                                         if ( $t == 'table' ) {
 574                                                                 array_push( $tablestack, $tagstack );
 575                                                                 $tagstack = [];
 576                                                         }
 577                                                         array_push( $tagstack, $t );
 578                                                 }
 579
 580                                                 # Replace any variables or template parameters with
 581                                                 # plaintext results.
 582                                                 if ( is_callable( $processCallback ) ) {
 583                                                         call_user_func_array( $processCallback, [ &$params, $args ] );
 584                                                 }
 585
 586                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 587                                                         $badtag = true;
 588                                                 }
 589
 590                                                 # Strip non-approved attributes from the tag
 591                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 592                                         }
 593                                         if ( !$badtag ) {
 594                                                 $rest = str_replace( '>', '&gt;', $rest );
 595                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 596                                                 $text .= "<$slash$t$newparams$close>$rest";
 597                                                 continue;
 598                                         }
 599                                 }
 600                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 601                         }
 602                         # Close off any remaining tags
 603                         while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
 604                                 $text .= "</$t>\n";
 605                                 if ( $t == 'table' ) {
 606                                         $tagstack = array_pop( $tablestack );
 607                                 }
 608                         }
 609                 } else {
 610                         # this might be possible using tidy itself
 611                         foreach ( $bits as $x ) {
 612                                 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 613                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 614
 615                                         $badtag = false;
 616                                         $t = strtolower( $t );
 617                                         if ( isset( $htmlelements[$t] ) ) {
 618                                                 if ( is_callable( $processCallback ) ) {
 619                                                         call_user_func_array( $processCallback, [ &$params, $args ] );
 620                                                 }
 621
 622                                                 if ( $brace == '/>' && !( isset( $htmlsingle[$t] ) || isset( $htmlsingleonly[$t] ) ) ) {
 623                                                         // Eventually we'll just remove the self-closing
 624                                                         // slash, in order to be consistent with HTML5
 625                                                         // semantics.
 626                                                         // $brace = '>';
 627                                                         // For now, let's just warn authors to clean up.
 628                                                         if ( is_callable( $warnCallback ) ) {
 629                                                                 call_user_func_array( $warnCallback, [ 'deprecated-self-close-category' ] );
 630                                                         }
 631                                                 }
 632                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 633                                                         $badtag = true;
 634                                                 }
 635
 636                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 637                                                 if ( !$badtag ) {
 638                                                         if ( $brace === '/>' && !isset( $htmlsingleonly[$t] ) ) {
 639                                                                 # Interpret self-closing tags as empty tags even when
 640                                                                 # HTML 5 would interpret them as start tags. Such input
 641                                                                 # is commonly seen on Wikimedia wikis with this intention.
 642                                                                 $brace = "></$t>";
 643                                                         }
 644
 645                                                         $rest = str_replace( '>', '&gt;', $rest );
 646                                                         $text .= "<$slash$t$newparams$brace$rest";
 647                                                         continue;
 648                                                 }
 649                                         }
 650                                 }
 651                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 652                         }
 653                 }
 654                 return $text;
 655         }
 656
 657         /**
 658          * Remove '<!--', '-->', and everything between.
 659          * To avoid leaving blank lines, when a comment is both preceded
 660          * and followed by a newline (ignoring spaces), trim leading and
 661          * trailing spaces and one of the newlines.
 662          *
 663          * @param string $text
 664          * @return string
 665          */
 666         public static function removeHTMLcomments( $text ) {
 667                 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
 668                         $end = strpos( $text, '-->', $start + 4 );
 669                         if ( $end === false ) {
 670                                 # Unterminated comment; bail out
 671                                 break;
 672                         }
 673
 674                         $end += 3;
 675
 676                         # Trim space and newline if the comment is both
 677                         # preceded and followed by a newline
 678                         $spaceStart = max( $start - 1, 0 );
 679                         $spaceLen = $end - $spaceStart;
 680                         while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
 681                                 $spaceStart--;
 682                                 $spaceLen++;
 683                         }
 684                         while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
 685                                 $spaceLen++;
 686                         }
 687                         if ( substr( $text, $spaceStart, 1 ) === "\n"
 688                                 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
 689                                 # Remove the comment, leading and trailing
 690                                 # spaces, and leave only one newline.
 691                                 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
 692                         } else {
 693                                 # Remove just the comment.
 694                                 $text = substr_replace( $text, '', $start, $end - $start );
 695                         }
 696                 }
 697                 return $text;
 698         }
 699
 700         /**
 701          * Takes attribute names and values for a tag and the tag name and
 702          * validates that the tag is allowed to be present.
 703          * This DOES NOT validate the attributes, nor does it validate the
 704          * tags themselves. This method only handles the special circumstances
 705          * where we may want to allow a tag within content but ONLY when it has
 706          * specific attributes set.
 707          *
 708          * @param string $params
 709          * @param string $element
 710          * @return bool
 711          */
 712         static function validateTag( $params, $element ) {
 713                 $params = Sanitizer::decodeTagAttributes( $params );
 714
 715                 if ( $element == 'meta' || $element == 'link' ) {
 716                         if ( !isset( $params['itemprop'] ) ) {
 717                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 718                                 return false;
 719                         }
 720                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 721                                 // <meta> must have a content="" for the itemprop
 722                                 return false;
 723                         }
 724                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 725                                 // <link> must have an associated href=""
 726                                 return false;
 727                         }
 728                 }
 729
 730                 return true;
 731         }
 732
 733         /**
 734          * Take an array of attribute names and values and normalize or discard
 735          * illegal values for the given element type.
 736          *
 737          * - Discards attributes not on a whitelist for the given element
 738          * - Unsafe style attributes are discarded
 739          * - Invalid id attributes are re-encoded
 740          *
 741          * @param array $attribs
 742          * @param string $element
 743          * @return array
 744          *
 745          * @todo Check for legal values where the DTD limits things.
 746          * @todo Check for unique id attribute :P
 747          */
 748         static function validateTagAttributes( $attribs, $element ) {
 749                 return Sanitizer::validateAttributes( $attribs,
 750                         Sanitizer::attributeWhitelist( $element ) );
 751         }
 752
 753         /**
 754          * Take an array of attribute names and values and normalize or discard
 755          * illegal values for the given whitelist.
 756          *
 757          * - Discards attributes not on the given whitelist
 758          * - Unsafe style attributes are discarded
 759          * - Invalid id attributes are re-encoded
 760          *
 761          * @param array $attribs
 762          * @param array $whitelist List of allowed attribute names
 763          * @return array
 764          *
 765          * @todo Check for legal values where the DTD limits things.
 766          * @todo Check for unique id attribute :P
 767          */
 768         static function validateAttributes( $attribs, $whitelist ) {
 769                 $whitelist = array_flip( $whitelist );
 770                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 771
 772                 $out = [];
 773                 foreach ( $attribs as $attribute => $value ) {
 774                         # Allow XML namespace declaration to allow RDFa
 775                         if ( preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 776                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 777                                         $out[$attribute] = $value;
 778                                 }
 779
 780                                 continue;
 781                         }
 782
 783                         # Allow any attribute beginning with "data-"
 784                         # However:
 785                         # * Disallow data attributes used by MediaWiki code
 786                         # * Ensure that the attribute is not namespaced by banning
 787                         #   colons.
 788                         if ( !preg_match( '/^data-[^:]*$/i', $attribute )
 789                                 && !isset( $whitelist[$attribute] )
 790                                 || self::isReservedDataAttribute( $attribute )
 791                         ) {
 792                                 continue;
 793                         }
 794
 795                         # Strip javascript "expression" from stylesheets.
 796                         # https://msdn.microsoft.com/en-us/library/ms537634.aspx
 797                         if ( $attribute == 'style' ) {
 798                                 $value = Sanitizer::checkCss( $value );
 799                         }
 800
 801                         # Escape HTML id attributes
 802                         if ( $attribute === 'id' ) {
 803                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 804                         }
 805
 806                         # Escape HTML id reference lists
 807                         if ( $attribute === 'aria-describedby'
 808                                 || $attribute === 'aria-flowto'
 809                                 || $attribute === 'aria-labelledby'
 810                                 || $attribute === 'aria-owns'
 811                         ) {
 812                                 $value = Sanitizer::escapeIdReferenceList( $value, 'noninitial' );
 813                         }
 814
 815                         // RDFa and microdata properties allow URLs, URIs and/or CURIs.
 816                         // Check them for sanity.
 817                         if ( $attribute === 'rel' || $attribute === 'rev'
 818                                 # RDFa
 819                                 || $attribute === 'about' || $attribute === 'property'
 820                                 || $attribute === 'resource' || $attribute === 'datatype'
 821                                 || $attribute === 'typeof'
 822                                 # HTML5 microdata
 823                                 || $attribute === 'itemid' || $attribute === 'itemprop'
 824                                 || $attribute === 'itemref' || $attribute === 'itemscope'
 825                                 || $attribute === 'itemtype'
 826                         ) {
 827                                 // Paranoia. Allow "simple" values but suppress javascript
 828                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 829                                         continue;
 830                                 }
 831                         }
 832
 833                         # NOTE: even though elements using href/src are not allowed directly, supply
 834                         #       validation code that can be used by tag hook handlers, etc
 835                         if ( $attribute === 'href' || $attribute === 'src' || $attribute === 'poster' ) {
 836                                 if ( !preg_match( $hrefExp, $value ) ) {
 837                                         continue; // drop any href or src attributes not using an allowed protocol.
 838                                         // NOTE: this also drops all relative URLs
 839                                 }
 840                         }
 841
 842                         // If this attribute was previously set, override it.
 843                         // Output should only have one attribute of each name.
 844                         $out[$attribute] = $value;
 845                 }
 846
 847                 # itemtype, itemid, itemref don't make sense without itemscope
 848                 if ( !array_key_exists( 'itemscope', $out ) ) {
 849                         unset( $out['itemtype'] );
 850                         unset( $out['itemid'] );
 851                         unset( $out['itemref'] );
 852                 }
 853                 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 854
 855                 return $out;
 856         }
 857
 858         /**
 859          * Given an attribute name, checks whether it is a reserved data attribute
 860          * (such as data-mw-foo) which is unavailable to user-generated HTML so MediaWiki
 861          * core and extension code can safely use it to communicate with frontend code.
 862          * @param string $attr Attribute name.
 863          * @return bool
 864          */
 865         public static function isReservedDataAttribute( $attr ) {
 866                 // data-ooui is reserved for ooui.
 867                 // data-mw and data-parsoid are reserved for parsoid.
 868                 // data-mw-<name here> is reserved for extensions (or core) if
 869                 // they need to communicate some data to the client and want to be
 870                 // sure that it isn't coming from an untrusted user.
 871                 // We ignore the possibility of namespaces since user-generated HTML
 872                 // can't use them anymore.
 873                 return (bool)preg_match( '/^data-(ooui|mw|parsoid)/i', $attr );
 874         }
 875
 876         /**
 877          * Merge two sets of HTML attributes.  Conflicting items in the second set
 878          * will override those in the first, except for 'class' attributes which
 879          * will be combined (if they're both strings).
 880          *
 881          * @todo implement merging for other attributes such as style
 882          * @param array $a
 883          * @param array $b
 884          * @return array
 885          */
 886         static function mergeAttributes( $a, $b ) {
 887                 $out = array_merge( $a, $b );
 888                 if ( isset( $a['class'] ) && isset( $b['class'] )
 889                         && is_string( $a['class'] ) && is_string( $b['class'] )
 890                         && $a['class'] !== $b['class']
 891                 ) {
 892                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 893                                 -1, PREG_SPLIT_NO_EMPTY );
 894                         $out['class'] = implode( ' ', array_unique( $classes ) );
 895                 }
 896                 return $out;
 897         }
 898
 899         /**
 900          * Normalize CSS into a format we can easily search for hostile input
 901          *  - decode character references
 902          *  - decode escape sequences
 903          *  - convert characters that IE6 interprets into ascii
 904          *  - remove comments, unless the entire value is one single comment
 905          * @param string $value the css string
 906          * @return string normalized css
 907          */
 908         public static function normalizeCss( $value ) {
 909                 // Decode character references like &#123;
 910                 $value = Sanitizer::decodeCharReferences( $value );
 911
 912                 // Decode escape sequences and line continuation
 913                 // See the grammar in the CSS 2 spec, appendix D.
 914                 // This has to be done AFTER decoding character references.
 915                 // This means it isn't possible for this function to return
 916                 // unsanitized escape sequences. It is possible to manufacture
 917                 // input that contains character references that decode to
 918                 // escape sequences that decode to character references, but
 919                 // it's OK for the return value to contain character references
 920                 // because the caller is supposed to escape those anyway.
 921                 static $decodeRegex;
 922                 if ( !$decodeRegex ) {
 923                         $space = '[\\x20\\t\\r\\n\\f]';
 924                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 925                         $backslash = '\\\\';
 926                         $decodeRegex = "/ $backslash
 927                                 (?:
 928                                         ($nl) |  # 1. Line continuation
 929                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 930                                         (.) | # 3. backslash cancelling special meaning
 931                                         () | # 4. backslash at end of string
 932                                 )/xu";
 933                 }
 934                 $value = preg_replace_callback( $decodeRegex,
 935                         [ __CLASS__, 'cssDecodeCallback' ], $value );
 936
 937                 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
 938                 $value = preg_replace_callback(
 939                         '/[！-［］-ｚ]/u', // U+FF01 to U+FF5A, excluding U+FF3C (T60088)
 940                         function ( $matches ) {
 941                                 $cp = UtfNormal\Utils::utf8ToCodepoint( $matches[0] );
 942                                 if ( $cp === false ) {
 943                                         return '';
 944                                 }
 945                                 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
 946                         },
 947                         $value
 948                 );
 949
 950                 // Convert more characters IE6 might treat as ascii
 951                 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
 952                 $value = str_replace(
 953                         [ 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ],
 954                         [ 'r', 'n', 'n', 'l', 'i', '(', '(' ],
 955                         $value
 956                 );
 957
 958                 // Let the value through if it's nothing but a single comment, to
 959                 // allow other functions which may reject it to pass some error
 960                 // message through.
 961                 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
 962                         // Remove any comments; IE gets token splitting wrong
 963                         // This must be done AFTER decoding character references and
 964                         // escape sequences, because those steps can introduce comments
 965                         // This step cannot introduce character references or escape
 966                         // sequences, because it replaces comments with spaces rather
 967                         // than removing them completely.
 968                         $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 969
 970                         // Remove anything after a comment-start token, to guard against
 971                         // incorrect client implementations.
 972                         $commentPos = strpos( $value, '/*' );
 973                         if ( $commentPos !== false ) {
 974                                 $value = substr( $value, 0, $commentPos );
 975                         }
 976                 }
 977
 978                 // S followed by repeat, iteration, or prolonged sound marks,
 979                 // which IE will treat as "ss"
 980                 $value = preg_replace(
 981                         '/s(?:
 982                                 \xE3\x80\xB1 | # U+3031
 983                                 \xE3\x82\x9D | # U+309D
 984                                 \xE3\x83\xBC | # U+30FC
 985                                 \xE3\x83\xBD | # U+30FD
 986                                 \xEF\xB9\xBC | # U+FE7C
 987                                 \xEF\xB9\xBD | # U+FE7D
 988                                 \xEF\xBD\xB0   # U+FF70
 989                         )/ix',
 990                         'ss',
 991                         $value
 992                 );
 993
 994                 return $value;
 995         }
 996
 997         /**
 998          * Pick apart some CSS and check it for forbidden or unsafe structures.
 999          * Returns a sanitized string. This sanitized string will have
1000          * character references and escape sequences decoded and comments
1001          * stripped (unless it is itself one valid comment, in which case the value
1002          * will be passed through). If the input is just too evil, only a comment
1003          * complaining about evilness will be returned.
1004          *
1005          * Currently URL references, 'expression', 'tps' are forbidden.
1006          *
1007          * NOTE: Despite the fact that character references are decoded, the
1008          * returned string may contain character references given certain
1009          * clever input strings. These character references must
1010          * be escaped before the return value is embedded in HTML.
1011          *
1012          * @param string $value
1013          * @return string
1014          */
1015         static function checkCss( $value ) {
1016                 $value = self::normalizeCss( $value );
1017
1018                 // Reject problematic keywords and control characters
1019                 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ||
1020                         strpos( $value, UtfNormal\Constants::UTF8_REPLACEMENT ) !== false ) {
1021                         return '/* invalid control char */';
1022                 } elseif ( preg_match(
1023                         '! expression
1024                                 | filter\s*:
1025                                 | accelerator\s*:
1026                                 | -o-link\s*:
1027                                 | -o-link-source\s*:
1028                                 | -o-replace\s*:
1029                                 | url\s*\(
1030                                 | image\s*\(
1031                                 | image-set\s*\(
1032                                 | attr\s*\([^)]+[\s,]+url
1033                         !ix', $value ) ) {
1034                         return '/* insecure input */';
1035                 }
1036                 return $value;
1037         }
1038
1039         /**
1040          * @param array $matches
1041          * @return string
1042          */
1043         static function cssDecodeCallback( $matches ) {
1044                 if ( $matches[1] !== '' ) {
1045                         // Line continuation
1046                         return '';
1047                 } elseif ( $matches[2] !== '' ) {
1048                         $char = UtfNormal\Utils::codepointToUtf8( hexdec( $matches[2] ) );
1049                 } elseif ( $matches[3] !== '' ) {
1050                         $char = $matches[3];
1051                 } else {
1052                         $char = '\\';
1053                 }
1054                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
1055                         // These characters need to be escaped in strings
1056                         // Clean up the escape sequence to avoid parsing errors by clients
1057                         return '\\' . dechex( ord( $char ) ) . ' ';
1058                 } else {
1059                         // Decode unnecessary escape
1060                         return $char;
1061                 }
1062         }
1063
1064         /**
1065          * Take a tag soup fragment listing an HTML element's attributes
1066          * and normalize it to well-formed XML, discarding unwanted attributes.
1067          * Output is safe for further wikitext processing, with escaping of
1068          * values that could trigger problems.
1069          *
1070          * - Normalizes attribute names to lowercase
1071          * - Discards attributes not on a whitelist for the given element
1072          * - Turns broken or invalid entities into plaintext
1073          * - Double-quotes all attribute values
1074          * - Attributes without values are given the name as attribute
1075          * - Double attributes are discarded
1076          * - Unsafe style attributes are discarded
1077          * - Prepends space if there are attributes.
1078          * - (Optionally) Sorts attributes by name.
1079          *
1080          * @param string $text
1081          * @param string $element
1082          * @param bool $sorted Whether to sort the attributes (default: false)
1083          * @return string
1084          */
1085         static function fixTagAttributes( $text, $element, $sorted = false ) {
1086                 if ( trim( $text ) == '' ) {
1087                         return '';
1088                 }
1089
1090                 $decoded = Sanitizer::decodeTagAttributes( $text );
1091                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1092
1093                 if ( $sorted ) {
1094                         ksort( $stripped );
1095                 }
1096
1097                 return Sanitizer::safeEncodeTagAttributes( $stripped );
1098         }
1099
1100         /**
1101          * Encode an attribute value for HTML output.
1102          * @param string $text
1103          * @return string HTML-encoded text fragment
1104          */
1105         static function encodeAttribute( $text ) {
1106                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1107
1108                 // Whitespace is normalized during attribute decoding,
1109                 // so if we've been passed non-spaces we must encode them
1110                 // ahead of time or they won't be preserved.
1111                 $encValue = strtr( $encValue, [
1112                         "\n" => '&#10;',
1113                         "\r" => '&#13;',
1114                         "\t" => '&#9;',
1115                 ] );
1116
1117                 return $encValue;
1118         }
1119
1120         /**
1121          * Encode an attribute value for HTML tags, with extra armoring
1122          * against further wiki processing.
1123          * @param string $text
1124          * @return string HTML-encoded text fragment
1125          */
1126         static function safeEncodeAttribute( $text ) {
1127                 $encValue = Sanitizer::encodeAttribute( $text );
1128
1129                 # Templates and links may be expanded in later parsing,
1130                 # creating invalid or dangerous output. Suppress this.
1131                 $encValue = strtr( $encValue, [
1132                         '<'    => '&lt;',   // This should never happen,
1133                         '>'    => '&gt;',   // we've received invalid input
1134                         '"'    => '&quot;', // which should have been escaped.
1135                         '{'    => '&#123;',
1136                         '}'    => '&#125;', // prevent unpaired language conversion syntax
1137                         '['    => '&#91;',
1138                         "''"   => '&#39;&#39;',
1139                         'ISBN' => '&#73;SBN',
1140                         'RFC'  => '&#82;FC',
1141                         'PMID' => '&#80;MID',
1142                         '|'    => '&#124;',
1143                         '__'   => '&#95;_',
1144                 ] );
1145
1146                 # Stupid hack
1147                 $encValue = preg_replace_callback(
1148                         '/((?i)' . wfUrlProtocols() . ')/',
1149                         [ 'Sanitizer', 'armorLinksCallback' ],
1150                         $encValue );
1151                 return $encValue;
1152         }
1153
1154         /**
1155          * Given a value, escape it so that it can be used in an id attribute and
1156          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1157          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1158          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1159          * escaped with lots of dots.
1160          *
1161          * To ensure we don't have to bother escaping anything, we also strip ', ",
1162          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1163          * We also strip # because it upsets IE, and % because it could be
1164          * ambiguous if it's part of something that looks like a percent escape
1165          * (which don't work reliably in fragments cross-browser).
1166          *
1167          * @see https://www.w3.org/TR/html401/types.html#type-name Valid characters
1168          *   in the id and name attributes
1169          * @see https://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1170          *   the id attribute
1171          * @see https://www.w3.org/TR/html5/dom.html#the-id-attribute
1172          *   HTML5 definition of id attribute
1173          *
1174          * @param string $id Id to escape
1175          * @param string|array $options String or array of strings (default is array()):
1176          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1177          *       so don't pay attention if the first character isn't valid at the
1178          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1179          *       false.
1180          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1181          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1182          *       anchors and links won't break.
1183          * @return string
1184          */
1185         static function escapeId( $id, $options = [] ) {
1186                 global $wgExperimentalHtmlIds;
1187                 $options = (array)$options;
1188
1189                 $id = Sanitizer::decodeCharReferences( $id );
1190
1191                 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1192                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1193                         $id = trim( $id, '_' );
1194                         if ( $id === '' ) {
1195                                 // Must have been all whitespace to start with.
1196                                 return '_';
1197                         } else {
1198                                 return $id;
1199                         }
1200                 }
1201
1202                 // HTML4-style escaping
1203                 static $replace = [
1204                         '%3A' => ':',
1205                         '%' => '.'
1206                 ];
1207
1208                 $id = urlencode( strtr( $id, ' ', '_' ) );
1209                 $id = strtr( $id, $replace );
1210
1211                 if ( !preg_match( '/^[a-zA-Z]/', $id ) && !in_array( 'noninitial', $options ) ) {
1212                         // Initial character must be a letter!
1213                         $id = "x$id";
1214                 }
1215                 return $id;
1216         }
1217
1218         /**
1219          * Given a string containing a space delimited list of ids, escape each id
1220          * to match ids escaped by the escapeId() function.
1221          *
1222          * @since 1.27
1223          *
1224          * @param string $referenceString Space delimited list of ids
1225          * @param string|array $options String or array of strings (default is array()):
1226          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1227          *       so don't pay attention if the first character isn't valid at the
1228          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1229          *       false.
1230          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1231          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1232          *       anchors and links won't break.
1233          * @return string
1234          */
1235         static function escapeIdReferenceList( $referenceString, $options = [] ) {
1236                 # Explode the space delimited list string into an array of tokens
1237                 $references = preg_split( '/\s+/', "{$referenceString}", -1, PREG_SPLIT_NO_EMPTY );
1238
1239                 # Escape each token as an id
1240                 foreach ( $references as &$ref ) {
1241                         $ref = Sanitizer::escapeId( $ref, $options );
1242                 }
1243
1244                 # Merge the array back to a space delimited list string
1245                 # If the array is empty, the result will be an empty string ('')
1246                 $referenceString = implode( ' ', $references );
1247
1248                 return $referenceString;
1249         }
1250
1251         /**
1252          * Given a value, escape it so that it can be used as a CSS class and
1253          * return it.
1254          *
1255          * @todo For extra validity, input should be validated UTF-8.
1256          *
1257          * @see https://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1258          *
1259          * @param string $class
1260          * @return string
1261          */
1262         static function escapeClass( $class ) {
1263                 // Convert ugly stuff to underscores and kill underscores in ugly places
1264                 return rtrim( preg_replace(
1265                         [ '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ],
1266                         '_',
1267                         $class ), '_' );
1268         }
1269
1270         /**
1271          * Given HTML input, escape with htmlspecialchars but un-escape entities.
1272          * This allows (generally harmless) entities like &#160; to survive.
1273          *
1274          * @param string $html HTML to escape
1275          * @return string Escaped input
1276          */
1277         static function escapeHtmlAllowEntities( $html ) {
1278                 $html = Sanitizer::decodeCharReferences( $html );
1279                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1280                 # hurt. Use ENT_SUBSTITUTE so that incorrectly truncated multibyte characters
1281                 # don't cause the entire string to disappear.
1282                 $html = htmlspecialchars( $html, ENT_QUOTES | ENT_SUBSTITUTE );
1283                 return $html;
1284         }
1285
1286         /**
1287          * Regex replace callback for armoring links against further processing.
1288          * @param array $matches
1289          * @return string
1290          */
1291         private static function armorLinksCallback( $matches ) {
1292                 return str_replace( ':', '&#58;', $matches[1] );
1293         }
1294
1295         /**
1296          * Return an associative array of attribute names and values from
1297          * a partial tag string. Attribute names are forced to lowercase,
1298          * character references are decoded to UTF-8 text.
1299          *
1300          * @param string $text
1301          * @return array
1302          */
1303         public static function decodeTagAttributes( $text ) {
1304                 if ( trim( $text ) == '' ) {
1305                         return [];
1306                 }
1307
1308                 $attribs = [];
1309                 $pairs = [];
1310                 if ( !preg_match_all(
1311                         self::getAttribsRegex(),
1312                         $text,
1313                         $pairs,
1314                         PREG_SET_ORDER ) ) {
1315                         return $attribs;
1316                 }
1317
1318                 foreach ( $pairs as $set ) {
1319                         $attribute = strtolower( $set[1] );
1320                         $value = Sanitizer::getTagAttributeCallback( $set );
1321
1322                         // Normalize whitespace
1323                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1324                         $value = trim( $value );
1325
1326                         // Decode character references
1327                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1328                 }
1329                 return $attribs;
1330         }
1331
1332         /**
1333          * Build a partial tag string from an associative array of attribute
1334          * names and values as returned by decodeTagAttributes.
1335          *
1336          * @param array $assoc_array
1337          * @return string
1338          */
1339         public static function safeEncodeTagAttributes( $assoc_array ) {
1340                 $attribs = [];
1341                 foreach ( $assoc_array as $attribute => $value ) {
1342                         $encAttribute = htmlspecialchars( $attribute );
1343                         $encValue = Sanitizer::safeEncodeAttribute( $value );
1344
1345                         $attribs[] = "$encAttribute=\"$encValue\"";
1346                 }
1347                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1348         }
1349
1350         /**
1351          * Pick the appropriate attribute value from a match set from the
1352          * attribs regex matches.
1353          *
1354          * @param array $set
1355          * @throws MWException When tag conditions are not met.
1356          * @return string
1357          */
1358         private static function getTagAttributeCallback( $set ) {
1359                 if ( isset( $set[5] ) ) {
1360                         # No quotes.
1361                         return $set[5];
1362                 } elseif ( isset( $set[4] ) ) {
1363                         # Single-quoted
1364                         return $set[4];
1365                 } elseif ( isset( $set[3] ) ) {
1366                         # Double-quoted
1367                         return $set[3];
1368                 } elseif ( !isset( $set[2] ) ) {
1369                         # In XHTML, attributes must have a value so return an empty string.
1370                         # See "Empty attribute syntax",
1371                         # https://www.w3.org/TR/html5/syntax.html#syntax-attribute-name
1372                         return "";
1373                 } else {
1374                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1375                 }
1376         }
1377
1378         /**
1379          * @param string $text
1380          * @return string
1381          */
1382         private static function normalizeWhitespace( $text ) {
1383                 return preg_replace(
1384                         '/\r\n|[\x20\x0d\x0a\x09]/',
1385                         ' ',
1386                         $text );
1387         }
1388
1389         /**
1390          * Normalizes whitespace in a section name, such as might be returned
1391          * by Parser::stripSectionName(), for use in the id's that are used for
1392          * section links.
1393          *
1394          * @param string $section
1395          * @return string
1396          */
1397         static function normalizeSectionNameWhitespace( $section ) {
1398                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1399         }
1400
1401         /**
1402          * Ensure that any entities and character references are legal
1403          * for XML and XHTML specifically. Any stray bits will be
1404          * &amp;-escaped to result in a valid text fragment.
1405          *
1406          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1407          *   numericized (this way we're well-formed even without a DTD)
1408          * b. any numeric char refs must be legal chars, not invalid or forbidden
1409          * c. use lower cased "&#x", not "&#X"
1410          * d. fix or reject non-valid attributes
1411          *
1412          * @param string $text
1413          * @return string
1414          * @private
1415          */
1416         static function normalizeCharReferences( $text ) {
1417                 return preg_replace_callback(
1418                         self::CHAR_REFS_REGEX,
1419                         [ 'Sanitizer', 'normalizeCharReferencesCallback' ],
1420                         $text );
1421         }
1422
1423         /**
1424          * @param string $matches
1425          * @return string
1426          */
1427         static function normalizeCharReferencesCallback( $matches ) {
1428                 $ret = null;
1429                 if ( $matches[1] != '' ) {
1430                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1431                 } elseif ( $matches[2] != '' ) {
1432                         $ret = Sanitizer::decCharReference( $matches[2] );
1433                 } elseif ( $matches[3] != '' ) {
1434                         $ret = Sanitizer::hexCharReference( $matches[3] );
1435                 }
1436                 if ( is_null( $ret ) ) {
1437                         return htmlspecialchars( $matches[0] );
1438                 } else {
1439                         return $ret;
1440                 }
1441         }
1442
1443         /**
1444          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1445          * return the equivalent numeric entity reference (except for the core &lt;
1446          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1447          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1448          * pseudo-entity source (eg &amp;foo;)
1449          *
1450          * @param string $name
1451          * @return string
1452          */
1453         static function normalizeEntity( $name ) {
1454                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1455                         return '&' . self::$htmlEntityAliases[$name] . ';';
1456                 } elseif ( in_array( $name, [ 'lt', 'gt', 'amp', 'quot' ] ) ) {
1457                         return "&$name;";
1458                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1459                         return '&#' . self::$htmlEntities[$name] . ';';
1460                 } else {
1461                         return "&amp;$name;";
1462                 }
1463         }
1464
1465         /**
1466          * @param int $codepoint
1467          * @return null|string
1468          */
1469         static function decCharReference( $codepoint ) {
1470                 $point = intval( $codepoint );
1471                 if ( Sanitizer::validateCodepoint( $point ) ) {
1472                         return sprintf( '&#%d;', $point );
1473                 } else {
1474                         return null;
1475                 }
1476         }
1477
1478         /**
1479          * @param int $codepoint
1480          * @return null|string
1481          */
1482         static function hexCharReference( $codepoint ) {
1483                 $point = hexdec( $codepoint );
1484                 if ( Sanitizer::validateCodepoint( $point ) ) {
1485                         return sprintf( '&#x%x;', $point );
1486                 } else {
1487                         return null;
1488                 }
1489         }
1490
1491         /**
1492          * Returns true if a given Unicode codepoint is a valid character in
1493          * both HTML5 and XML.
1494          * @param int $codepoint
1495          * @return bool
1496          */
1497         private static function validateCodepoint( $codepoint ) {
1498                 # U+000C is valid in HTML5 but not allowed in XML.
1499                 # U+000D is valid in XML but not allowed in HTML5.
1500                 # U+007F - U+009F are disallowed in HTML5 (control characters).
1501                 return $codepoint == 0x09
1502                         || $codepoint == 0x0a
1503                         || ( $codepoint >= 0x20 && $codepoint <= 0x7e )
1504                         || ( $codepoint >= 0xa0 && $codepoint <= 0xd7ff )
1505                         || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1506                         || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1507         }
1508
1509         /**
1510          * Decode any character references, numeric or named entities,
1511          * in the text and return a UTF-8 string.
1512          *
1513          * @param string $text
1514          * @return string
1515          */
1516         public static function decodeCharReferences( $text ) {
1517                 return preg_replace_callback(
1518                         self::CHAR_REFS_REGEX,
1519                         [ 'Sanitizer', 'decodeCharReferencesCallback' ],
1520                         $text );
1521         }
1522
1523         /**
1524          * Decode any character references, numeric or named entities,
1525          * in the next and normalize the resulting string. (T16952)
1526          *
1527          * This is useful for page titles, not for text to be displayed,
1528          * MediaWiki allows HTML entities to escape normalization as a feature.
1529          *
1530          * @param string $text Already normalized, containing entities
1531          * @return string Still normalized, without entities
1532          */
1533         public static function decodeCharReferencesAndNormalize( $text ) {
1534                 global $wgContLang;
1535                 $text = preg_replace_callback(
1536                         self::CHAR_REFS_REGEX,
1537                         [ 'Sanitizer', 'decodeCharReferencesCallback' ],
1538                         $text, /* limit */ -1, $count );
1539
1540                 if ( $count ) {
1541                         return $wgContLang->normalize( $text );
1542                 } else {
1543                         return $text;
1544                 }
1545         }
1546
1547         /**
1548          * @param string $matches
1549          * @return string
1550          */
1551         static function decodeCharReferencesCallback( $matches ) {
1552                 if ( $matches[1] != '' ) {
1553                         return Sanitizer::decodeEntity( $matches[1] );
1554                 } elseif ( $matches[2] != '' ) {
1555                         return Sanitizer::decodeChar( intval( $matches[2] ) );
1556                 } elseif ( $matches[3] != '' ) {
1557                         return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1558                 }
1559                 # Last case should be an ampersand by itself
1560                 return $matches[0];
1561         }
1562
1563         /**
1564          * Return UTF-8 string for a codepoint if that is a valid
1565          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1566          * @param int $codepoint
1567          * @return string
1568          * @private
1569          */
1570         static function decodeChar( $codepoint ) {
1571                 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1572                         return UtfNormal\Utils::codepointToUtf8( $codepoint );
1573                 } else {
1574                         return UtfNormal\Constants::UTF8_REPLACEMENT;
1575                 }
1576         }
1577
1578         /**
1579          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1580          * return the UTF-8 encoding of that character. Otherwise, returns
1581          * pseudo-entity source (eg "&foo;")
1582          *
1583          * @param string $name
1584          * @return string
1585          */
1586         static function decodeEntity( $name ) {
1587                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1588                         $name = self::$htmlEntityAliases[$name];
1589                 }
1590                 if ( isset( self::$htmlEntities[$name] ) ) {
1591                         return UtfNormal\Utils::codepointToUtf8( self::$htmlEntities[$name] );
1592                 } else {
1593                         return "&$name;";
1594                 }
1595         }
1596
1597         /**
1598          * Fetch the whitelist of acceptable attributes for a given element name.
1599          *
1600          * @param string $element
1601          * @return array
1602          */
1603         static function attributeWhitelist( $element ) {
1604                 $list = Sanitizer::setupAttributeWhitelist();
1605                 return isset( $list[$element] )
1606                         ? $list[$element]
1607                         : [];
1608         }
1609
1610         /**
1611          * Foreach array key (an allowed HTML element), return an array
1612          * of allowed attributes
1613          * @return array
1614          */
1615         static function setupAttributeWhitelist() {
1616                 static $whitelist;
1617
1618                 if ( $whitelist !== null ) {
1619                         return $whitelist;
1620                 }
1621
1622                 $common = [
1623                         # HTML
1624                         'id',
1625                         'class',
1626                         'style',
1627                         'lang',
1628                         'dir',
1629                         'title',
1630
1631                         # WAI-ARIA
1632                         'aria-describedby',
1633                         'aria-flowto',
1634                         'aria-label',
1635                         'aria-labelledby',
1636                         'aria-owns',
1637                         'role',
1638
1639                         # RDFa
1640                         # These attributes are specified in section 9 of
1641                         # https://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1642                         'about',
1643                         'property',
1644                         'resource',
1645                         'datatype',
1646                         'typeof',
1647
1648                         # Microdata. These are specified by
1649                         # https://html.spec.whatwg.org/multipage/microdata.html#the-microdata-model
1650                         'itemid',
1651                         'itemprop',
1652                         'itemref',
1653                         'itemscope',
1654                         'itemtype',
1655                 ];
1656
1657                 $block = array_merge( $common, [ 'align' ] );
1658                 $tablealign = [ 'align', 'valign' ];
1659                 $tablecell = [
1660                         'abbr',
1661                         'axis',
1662                         'headers',
1663                         'scope',
1664                         'rowspan',
1665                         'colspan',
1666                         'nowrap', # deprecated
1667                         'width', # deprecated
1668                         'height', # deprecated
1669                         'bgcolor', # deprecated
1670                 ];
1671
1672                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1673                 # See: https://www.w3.org/TR/html4/
1674                 $whitelist = [
1675                         # 7.5.4
1676                         'div'        => $block,
1677                         'center'     => $common, # deprecated
1678                         'span'       => $common,
1679
1680                         # 7.5.5
1681                         'h1'         => $block,
1682                         'h2'         => $block,
1683                         'h3'         => $block,
1684                         'h4'         => $block,
1685                         'h5'         => $block,
1686                         'h6'         => $block,
1687
1688                         # 7.5.6
1689                         # address
1690
1691                         # 8.2.4
1692                         'bdo'        => $common,
1693
1694                         # 9.2.1
1695                         'em'         => $common,
1696                         'strong'     => $common,
1697                         'cite'       => $common,
1698                         'dfn'        => $common,
1699                         'code'       => $common,
1700                         'samp'       => $common,
1701                         'kbd'        => $common,
1702                         'var'        => $common,
1703                         'abbr'       => $common,
1704                         # acronym
1705
1706                         # 9.2.2
1707                         'blockquote' => array_merge( $common, [ 'cite' ] ),
1708                         'q'          => array_merge( $common, [ 'cite' ] ),
1709
1710                         # 9.2.3
1711                         'sub'        => $common,
1712                         'sup'        => $common,
1713
1714                         # 9.3.1
1715                         'p'          => $block,
1716
1717                         # 9.3.2
1718                         'br'         => array_merge( $common, [ 'clear' ] ),
1719
1720                         # https://www.w3.org/TR/html5/text-level-semantics.html#the-wbr-element
1721                         'wbr'        => $common,
1722
1723                         # 9.3.4
1724                         'pre'        => array_merge( $common, [ 'width' ] ),
1725
1726                         # 9.4
1727                         'ins'        => array_merge( $common, [ 'cite', 'datetime' ] ),
1728                         'del'        => array_merge( $common, [ 'cite', 'datetime' ] ),
1729
1730                         # 10.2
1731                         'ul'         => array_merge( $common, [ 'type' ] ),
1732                         'ol'         => array_merge( $common, [ 'type', 'start', 'reversed' ] ),
1733                         'li'         => array_merge( $common, [ 'type', 'value' ] ),
1734
1735                         # 10.3
1736                         'dl'         => $common,
1737                         'dd'         => $common,
1738                         'dt'         => $common,
1739
1740                         # 11.2.1
1741                         'table'      => array_merge( $common,
1742                                                                 [ 'summary', 'width', 'border', 'frame',
1743                                                                                 'rules', 'cellspacing', 'cellpadding',
1744                                                                                 'align', 'bgcolor',
1745                                                                 ] ),
1746
1747                         # 11.2.2
1748                         'caption'    => $block,
1749
1750                         # 11.2.3
1751                         'thead'      => $common,
1752                         'tfoot'      => $common,
1753                         'tbody'      => $common,
1754
1755                         # 11.2.4
1756                         'colgroup'   => array_merge( $common, [ 'span' ] ),
1757                         'col'        => array_merge( $common, [ 'span' ] ),
1758
1759                         # 11.2.5
1760                         'tr'         => array_merge( $common, [ 'bgcolor' ], $tablealign ),
1761
1762                         # 11.2.6
1763                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1764                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1765
1766                         # 12.2
1767                         # NOTE: <a> is not allowed directly, but the attrib
1768                         # whitelist is used from the Parser object
1769                         'a'          => array_merge( $common, [ 'href', 'rel', 'rev' ] ), # rel/rev esp. for RDFa
1770
1771                         # 13.2
1772                         # Not usually allowed, but may be used for extension-style hooks
1773                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1774                         # true
1775                         'img'        => array_merge( $common, [ 'alt', 'src', 'width', 'height' ] ),
1776
1777                         'video'      => array_merge( $common, [ 'poster', 'controls', 'preload', 'width', 'height' ] ),
1778                         'source'     => array_merge( $common, [ 'type', 'src' ] ),
1779                         'track'      => array_merge( $common, [ 'type', 'src', 'srclang', 'kind', 'label' ] ),
1780
1781                         # 15.2.1
1782                         'tt'         => $common,
1783                         'b'          => $common,
1784                         'i'          => $common,
1785                         'big'        => $common,
1786                         'small'      => $common,
1787                         'strike'     => $common,
1788                         's'          => $common,
1789                         'u'          => $common,
1790
1791                         # 15.2.2
1792                         'font'       => array_merge( $common, [ 'size', 'color', 'face' ] ),
1793                         # basefont
1794
1795                         # 15.3
1796                         'hr'         => array_merge( $common, [ 'width' ] ),
1797
1798                         # HTML Ruby annotation text module, simple ruby only.
1799                         # https://www.w3.org/TR/html5/text-level-semantics.html#the-ruby-element
1800                         'ruby'       => $common,
1801                         # rbc
1802                         'rb'         => $common,
1803                         'rp'         => $common,
1804                         'rt'         => $common, # array_merge( $common, array( 'rbspan' ) ),
1805                         'rtc'        => $common,
1806
1807                         # MathML root element, where used for extensions
1808                         # 'title' may not be 100% valid here; it's XHTML
1809                         # https://www.w3.org/TR/REC-MathML/
1810                         'math'       => [ 'class', 'style', 'id', 'title' ],
1811
1812                         # HTML 5 section 4.6
1813                         'bdi' => $common,
1814
1815                         # HTML5 elements, defined by:
1816                         # https://html.spec.whatwg.org/multipage/semantics.html#the-data-element
1817                         'data' => array_merge( $common, [ 'value' ] ),
1818                         'time' => array_merge( $common, [ 'datetime' ] ),
1819                         'mark' => $common,
1820
1821                         // meta and link are only permitted by removeHTMLtags when Microdata
1822                         // is enabled so we don't bother adding a conditional to hide these
1823                         // Also meta and link are only valid in WikiText as Microdata elements
1824                         // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1825                         // So we don't bother including $common attributes that have no purpose.
1826                         'meta' => [ 'itemprop', 'content' ],
1827                         'link' => [ 'itemprop', 'href' ],
1828                 ];
1829
1830                 return $whitelist;
1831         }
1832
1833         /**
1834          * Take a fragment of (potentially invalid) HTML and return
1835          * a version with any tags removed, encoded as plain text.
1836          *
1837          * Warning: this return value must be further escaped for literal
1838          * inclusion in HTML output as of 1.10!
1839          *
1840          * @param string $text HTML fragment
1841          * @return string
1842          */
1843         static function stripAllTags( $text ) {
1844                 # Actual <tags>
1845                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1846
1847                 # Normalize &entities and whitespace
1848                 $text = self::decodeCharReferences( $text );
1849                 $text = self::normalizeWhitespace( $text );
1850
1851                 return $text;
1852         }
1853
1854         /**
1855          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1856          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1857          * PHP 5.1 doesn't.
1858          *
1859          * Use for passing XHTML fragments to PHP's XML parsing functions
1860          *
1861          * @return string
1862          */
1863         static function hackDocType() {
1864                 $out = "<!DOCTYPE html [\n";
1865                 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1866                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1867                 }
1868                 $out .= "]>\n";
1869                 return $out;
1870         }
1871
1872         /**
1873          * @param string $url
1874          * @return mixed|string
1875          */
1876         static function cleanUrl( $url ) {
1877                 # Normalize any HTML entities in input. They will be
1878                 # re-escaped by makeExternalLink().
1879                 $url = Sanitizer::decodeCharReferences( $url );
1880
1881                 # Escape any control characters introduced by the above step
1882                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1883                         [ __CLASS__, 'cleanUrlCallback' ], $url );
1884
1885                 # Validate hostname portion
1886                 $matches = [];
1887                 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1888                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1889
1890                         // Characters that will be ignored in IDNs.
1891                         // https://tools.ietf.org/html/rfc3454#section-3.1
1892                         // Strip them before further processing so blacklists and such work.
1893                         $strip = "/
1894                                 \\s|          # general whitespace
1895                                 \xc2\xad|     # 00ad SOFT HYPHEN
1896                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1897                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1898                                 \xe2\x81\xa0| # 2060 WORD JOINER
1899                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1900                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1901                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1902                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1903                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1904                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1905                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1906                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1907                                 /xuD";
1908
1909                         $host = preg_replace( $strip, '', $host );
1910
1911                         // IPv6 host names are bracketed with [].  Url-decode these.
1912                         if ( substr_compare( "//%5B", $host, 0, 5 ) === 0 &&
1913                                 preg_match( '!^//%5B([0-9A-Fa-f:.]+)%5D((:\d+)?)$!', $host, $matches )
1914                         ) {
1915                                 $host = '//[' . $matches[1] . ']' . $matches[2];
1916                         }
1917
1918                         // @todo FIXME: Validate hostnames here
1919
1920                         return $protocol . $host . $rest;
1921                 } else {
1922                         return $url;
1923                 }
1924         }
1925
1926         /**
1927          * @param array $matches
1928          * @return string
1929          */
1930         static function cleanUrlCallback( $matches ) {
1931                 return urlencode( $matches[0] );
1932         }
1933
1934         /**
1935          * Does a string look like an e-mail address?
1936          *
1937          * This validates an email address using an HTML5 specification found at:
1938          * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1939          * Which as of 2011-01-24 says:
1940          *
1941          *   A valid e-mail address is a string that matches the ABNF production
1942          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1943          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1944          *   3.5.
1945          *
1946          * This function is an implementation of the specification as requested in
1947          * T24449.
1948          *
1949          * Client-side forms will use the same standard validation rules via JS or
1950          * HTML 5 validation; additional restrictions can be enforced server-side
1951          * by extensions via the 'isValidEmailAddr' hook.
1952          *
1953          * Note that this validation doesn't 100% match RFC 2822, but is believed
1954          * to be liberal enough for wide use. Some invalid addresses will still
1955          * pass validation here.
1956          *
1957          * @since 1.18
1958          *
1959          * @param string $addr E-mail address
1960          * @return bool
1961          */
1962         public static function validateEmail( $addr ) {
1963                 $result = null;
1964                 if ( !Hooks::run( 'isValidEmailAddr', [ $addr, &$result ] ) ) {
1965                         return $result;
1966                 }
1967
1968                 // Please note strings below are enclosed in brackets [], this make the
1969                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1970                 // See T28948
1971                 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1972                 $rfc1034_ldh_str = "a-z0-9\\-";
1973
1974                 $html5_email_regexp = "/
1975                 ^                      # start of string
1976                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1977                 @                      # 'apostrophe'
1978                 [$rfc1034_ldh_str]+       # First domain part
1979                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1980                 $                      # End of string
1981                 /ix"; // case Insensitive, eXtended
1982
1983                 return (bool)preg_match( $html5_email_regexp, $addr );
1984         }
1985 }