includes/parser/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          * Tells escapeUrlForHtml() to encode the ID using the wiki's primary encoding.
  61          *
  62          * @since 1.30
  63          */
  64         const ID_PRIMARY = 0;
  65
  66         /**
  67          * Tells escapeUrlForHtml() to encode the ID using the fallback encoding, or return false
  68          * if no fallback is configured.
  69          *
  70          * @since 1.30
  71          */
  72         const ID_FALLBACK = 1;
  73
  74         /**
  75          * List of all named character entities defined in HTML 4.01
  76          * https://www.w3.org/TR/html4/sgml/entities.html
  77          * As well as &apos; which is only defined starting in XHTML1.
  78          */
  79         private static $htmlEntities = [
  80                 'Aacute'   => 193,
  81                 'aacute'   => 225,
  82                 'Acirc'    => 194,
  83                 'acirc'    => 226,
  84                 'acute'    => 180,
  85                 'AElig'    => 198,
  86                 'aelig'    => 230,
  87                 'Agrave'   => 192,
  88                 'agrave'   => 224,
  89                 'alefsym'  => 8501,
  90                 'Alpha'    => 913,
  91                 'alpha'    => 945,
  92                 'amp'      => 38,
  93                 'and'      => 8743,
  94                 'ang'      => 8736,
  95                 'apos'     => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
  96                 'Aring'    => 197,
  97                 'aring'    => 229,
  98                 'asymp'    => 8776,
  99                 'Atilde'   => 195,
 100                 'atilde'   => 227,
 101                 'Auml'     => 196,
 102                 'auml'     => 228,
 103                 'bdquo'    => 8222,
 104                 'Beta'     => 914,
 105                 'beta'     => 946,
 106                 'brvbar'   => 166,
 107                 'bull'     => 8226,
 108                 'cap'      => 8745,
 109                 'Ccedil'   => 199,
 110                 'ccedil'   => 231,
 111                 'cedil'    => 184,
 112                 'cent'     => 162,
 113                 'Chi'      => 935,
 114                 'chi'      => 967,
 115                 'circ'     => 710,
 116                 'clubs'    => 9827,
 117                 'cong'     => 8773,
 118                 'copy'     => 169,
 119                 'crarr'    => 8629,
 120                 'cup'      => 8746,
 121                 'curren'   => 164,
 122                 'dagger'   => 8224,
 123                 'Dagger'   => 8225,
 124                 'darr'     => 8595,
 125                 'dArr'     => 8659,
 126                 'deg'      => 176,
 127                 'Delta'    => 916,
 128                 'delta'    => 948,
 129                 'diams'    => 9830,
 130                 'divide'   => 247,
 131                 'Eacute'   => 201,
 132                 'eacute'   => 233,
 133                 'Ecirc'    => 202,
 134                 'ecirc'    => 234,
 135                 'Egrave'   => 200,
 136                 'egrave'   => 232,
 137                 'empty'    => 8709,
 138                 'emsp'     => 8195,
 139                 'ensp'     => 8194,
 140                 'Epsilon'  => 917,
 141                 'epsilon'  => 949,
 142                 'equiv'    => 8801,
 143                 'Eta'      => 919,
 144                 'eta'      => 951,
 145                 'ETH'      => 208,
 146                 'eth'      => 240,
 147                 'Euml'     => 203,
 148                 'euml'     => 235,
 149                 'euro'     => 8364,
 150                 'exist'    => 8707,
 151                 'fnof'     => 402,
 152                 'forall'   => 8704,
 153                 'frac12'   => 189,
 154                 'frac14'   => 188,
 155                 'frac34'   => 190,
 156                 'frasl'    => 8260,
 157                 'Gamma'    => 915,
 158                 'gamma'    => 947,
 159                 'ge'       => 8805,
 160                 'gt'       => 62,
 161                 'harr'     => 8596,
 162                 'hArr'     => 8660,
 163                 'hearts'   => 9829,
 164                 'hellip'   => 8230,
 165                 'Iacute'   => 205,
 166                 'iacute'   => 237,
 167                 'Icirc'    => 206,
 168                 'icirc'    => 238,
 169                 'iexcl'    => 161,
 170                 'Igrave'   => 204,
 171                 'igrave'   => 236,
 172                 'image'    => 8465,
 173                 'infin'    => 8734,
 174                 'int'      => 8747,
 175                 'Iota'     => 921,
 176                 'iota'     => 953,
 177                 'iquest'   => 191,
 178                 'isin'     => 8712,
 179                 'Iuml'     => 207,
 180                 'iuml'     => 239,
 181                 'Kappa'    => 922,
 182                 'kappa'    => 954,
 183                 'Lambda'   => 923,
 184                 'lambda'   => 955,
 185                 'lang'     => 9001,
 186                 'laquo'    => 171,
 187                 'larr'     => 8592,
 188                 'lArr'     => 8656,
 189                 'lceil'    => 8968,
 190                 'ldquo'    => 8220,
 191                 'le'       => 8804,
 192                 'lfloor'   => 8970,
 193                 'lowast'   => 8727,
 194                 'loz'      => 9674,
 195                 'lrm'      => 8206,
 196                 'lsaquo'   => 8249,
 197                 'lsquo'    => 8216,
 198                 'lt'       => 60,
 199                 'macr'     => 175,
 200                 'mdash'    => 8212,
 201                 'micro'    => 181,
 202                 'middot'   => 183,
 203                 'minus'    => 8722,
 204                 'Mu'       => 924,
 205                 'mu'       => 956,
 206                 'nabla'    => 8711,
 207                 'nbsp'     => 160,
 208                 'ndash'    => 8211,
 209                 'ne'       => 8800,
 210                 'ni'       => 8715,
 211                 'not'      => 172,
 212                 'notin'    => 8713,
 213                 'nsub'     => 8836,
 214                 'Ntilde'   => 209,
 215                 'ntilde'   => 241,
 216                 'Nu'       => 925,
 217                 'nu'       => 957,
 218                 'Oacute'   => 211,
 219                 'oacute'   => 243,
 220                 'Ocirc'    => 212,
 221                 'ocirc'    => 244,
 222                 'OElig'    => 338,
 223                 'oelig'    => 339,
 224                 'Ograve'   => 210,
 225                 'ograve'   => 242,
 226                 'oline'    => 8254,
 227                 'Omega'    => 937,
 228                 'omega'    => 969,
 229                 'Omicron'  => 927,
 230                 'omicron'  => 959,
 231                 'oplus'    => 8853,
 232                 'or'       => 8744,
 233                 'ordf'     => 170,
 234                 'ordm'     => 186,
 235                 'Oslash'   => 216,
 236                 'oslash'   => 248,
 237                 'Otilde'   => 213,
 238                 'otilde'   => 245,
 239                 'otimes'   => 8855,
 240                 'Ouml'     => 214,
 241                 'ouml'     => 246,
 242                 'para'     => 182,
 243                 'part'     => 8706,
 244                 'permil'   => 8240,
 245                 'perp'     => 8869,
 246                 'Phi'      => 934,
 247                 'phi'      => 966,
 248                 'Pi'       => 928,
 249                 'pi'       => 960,
 250                 'piv'      => 982,
 251                 'plusmn'   => 177,
 252                 'pound'    => 163,
 253                 'prime'    => 8242,
 254                 'Prime'    => 8243,
 255                 'prod'     => 8719,
 256                 'prop'     => 8733,
 257                 'Psi'      => 936,
 258                 'psi'      => 968,
 259                 'quot'     => 34,
 260                 'radic'    => 8730,
 261                 'rang'     => 9002,
 262                 'raquo'    => 187,
 263                 'rarr'     => 8594,
 264                 'rArr'     => 8658,
 265                 'rceil'    => 8969,
 266                 'rdquo'    => 8221,
 267                 'real'     => 8476,
 268                 'reg'      => 174,
 269                 'rfloor'   => 8971,
 270                 'Rho'      => 929,
 271                 'rho'      => 961,
 272                 'rlm'      => 8207,
 273                 'rsaquo'   => 8250,
 274                 'rsquo'    => 8217,
 275                 'sbquo'    => 8218,
 276                 'Scaron'   => 352,
 277                 'scaron'   => 353,
 278                 'sdot'     => 8901,
 279                 'sect'     => 167,
 280                 'shy'      => 173,
 281                 'Sigma'    => 931,
 282                 'sigma'    => 963,
 283                 'sigmaf'   => 962,
 284                 'sim'      => 8764,
 285                 'spades'   => 9824,
 286                 'sub'      => 8834,
 287                 'sube'     => 8838,
 288                 'sum'      => 8721,
 289                 'sup'      => 8835,
 290                 'sup1'     => 185,
 291                 'sup2'     => 178,
 292                 'sup3'     => 179,
 293                 'supe'     => 8839,
 294                 'szlig'    => 223,
 295                 'Tau'      => 932,
 296                 'tau'      => 964,
 297                 'there4'   => 8756,
 298                 'Theta'    => 920,
 299                 'theta'    => 952,
 300                 'thetasym' => 977,
 301                 'thinsp'   => 8201,
 302                 'THORN'    => 222,
 303                 'thorn'    => 254,
 304                 'tilde'    => 732,
 305                 'times'    => 215,
 306                 'trade'    => 8482,
 307                 'Uacute'   => 218,
 308                 'uacute'   => 250,
 309                 'uarr'     => 8593,
 310                 'uArr'     => 8657,
 311                 'Ucirc'    => 219,
 312                 'ucirc'    => 251,
 313                 'Ugrave'   => 217,
 314                 'ugrave'   => 249,
 315                 'uml'      => 168,
 316                 'upsih'    => 978,
 317                 'Upsilon'  => 933,
 318                 'upsilon'  => 965,
 319                 'Uuml'     => 220,
 320                 'uuml'     => 252,
 321                 'weierp'   => 8472,
 322                 'Xi'       => 926,
 323                 'xi'       => 958,
 324                 'Yacute'   => 221,
 325                 'yacute'   => 253,
 326                 'yen'      => 165,
 327                 'Yuml'     => 376,
 328                 'yuml'     => 255,
 329                 'Zeta'     => 918,
 330                 'zeta'     => 950,
 331                 'zwj'      => 8205,
 332                 'zwnj'     => 8204
 333         ];
 334
 335         /**
 336          * Character entity aliases accepted by MediaWiki
 337          */
 338         private static $htmlEntityAliases = [
 339                 'רלמ' => 'rlm',
 340                 'رلم' => 'rlm',
 341         ];
 342
 343         /**
 344          * Lazy-initialised attributes regex, see getAttribsRegex()
 345          */
 346         private static $attribsRegex;
 347
 348         /**
 349          * Regular expression to match HTML/XML attribute pairs within a tag.
 350          * Allows some... latitude. Based on,
 351          * https://www.w3.org/TR/html5/syntax.html#before-attribute-value-state
 352          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 353          * @return string
 354          */
 355         static function getAttribsRegex() {
 356                 if ( self::$attribsRegex === null ) {
 357                         $attribFirst = "[:_\p{L}\p{N}]";
 358                         $attrib = "[:_\.\-\p{L}\p{N}]";
 359                         $space = '[\x09\x0a\x0c\x0d\x20]';
 360                         self::$attribsRegex =
 361                                 "/(?:^|$space)({$attribFirst}{$attrib}*)
 362                                         ($space*=$space*
 363                                         (?:
 364                                                 # The attribute value: quoted or alone
 365                                                 \"([^\"]*)(?:\"|\$)
 366                                                 | '([^']*)(?:'|\$)
 367                                                 | (((?!$space|>).)*)
 368                                         )
 369                                 )?(?=$space|\$)/sxu";
 370                 }
 371                 return self::$attribsRegex;
 372         }
 373
 374         /**
 375          * Return the various lists of recognized tags
 376          * @param array $extratags For any extra tags to include
 377          * @param array $removetags For any tags (default or extra) to exclude
 378          * @return array
 379          */
 380         public static function getRecognizedTagData( $extratags = [], $removetags = [] ) {
 381                 global $wgAllowImageTag;
 382
 383                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 384                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 385
 386                 // Base our staticInitialised variable off of the global config state so that if the globals
 387                 // are changed (like in the screwed up test system) we will re-initialise the settings.
 388                 $globalContext = $wgAllowImageTag;
 389                 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
 390                         $htmlpairsStatic = [ # Tags that must be closed
 391                                 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 392                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 393                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 394                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 395                                 'ruby', 'rb', 'rp', 'rt', 'rtc', 'p', 'span', 'abbr', 'dfn',
 396                                 'kbd', 'samp', 'data', 'time', 'mark'
 397                         ];
 398                         $htmlsingle = [
 399                                 'br', 'wbr', 'hr', 'li', 'dt', 'dd', 'meta', 'link'
 400                         ];
 401
 402                         # Elements that cannot have close tags. This is (not coincidentally)
 403                         # also the list of tags for which the HTML 5 parsing algorithm
 404                         # requires you to "acknowledge the token's self-closing flag", i.e.
 405                         # a self-closing tag like <br/> is not an HTML 5 parse error only
 406                         # for this list.
 407                         $htmlsingleonly = [
 408                                 'br', 'wbr', 'hr', 'meta', 'link'
 409                         ];
 410
 411                         $htmlnest = [ # Tags that can be nested--??
 412                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 413                                 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
 414                                 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
 415                         ];
 416                         $tabletags = [ # Can only appear inside table, we will close them
 417                                 'td', 'th', 'tr',
 418                         ];
 419                         $htmllist = [ # Tags used by list
 420                                 'ul', 'ol',
 421                         ];
 422                         $listtags = [ # Tags that can appear in a list
 423                                 'li',
 424                         ];
 425
 426                         if ( $wgAllowImageTag ) {
 427                                 $htmlsingle[] = 'img';
 428                                 $htmlsingleonly[] = 'img';
 429                         }
 430
 431                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 432                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 433
 434                         # Convert them all to hashtables for faster lookup
 435                         $vars = [ 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 436                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' ];
 437                         foreach ( $vars as $var ) {
 438                                 $$var = array_flip( $$var );
 439                         }
 440                         $staticInitialised = $globalContext;
 441                 }
 442
 443                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 444                 $extratags = array_flip( $extratags );
 445                 $removetags = array_flip( $removetags );
 446                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 447                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
 448
 449                 return [
 450                         'htmlpairs' => $htmlpairs,
 451                         'htmlsingle' => $htmlsingle,
 452                         'htmlsingleonly' => $htmlsingleonly,
 453                         'htmlnest' => $htmlnest,
 454                         'tabletags' => $tabletags,
 455                         'htmllist' => $htmllist,
 456                         'listtags' => $listtags,
 457                         'htmlsingleallowed' => $htmlsingleallowed,
 458                         'htmlelements' => $htmlelements,
 459                 ];
 460         }
 461
 462         /**
 463          * Cleans up HTML, removes dangerous tags and attributes, and
 464          * removes HTML comments
 465          * @param string $text
 466          * @param callable $processCallback Callback to do any variable or parameter
 467          *   replacements in HTML attribute values
 468          * @param array|bool $args Arguments for the processing callback
 469          * @param array $extratags For any extra tags to include
 470          * @param array $removetags For any tags (default or extra) to exclude
 471          * @param callable $warnCallback (Deprecated) Callback allowing the
 472          *   addition of a tracking category when bad input is encountered.
 473          *   DO NOT ADD NEW PARAMETERS AFTER $warnCallback, since it will be
 474          *   removed shortly.
 475          * @return string
 476          */
 477         public static function removeHTMLtags( $text, $processCallback = null,
 478                 $args = [], $extratags = [], $removetags = [], $warnCallback = null
 479         ) {
 480                 $tagData = self::getRecognizedTagData( $extratags, $removetags );
 481                 $htmlpairs = $tagData['htmlpairs'];
 482                 $htmlsingle = $tagData['htmlsingle'];
 483                 $htmlsingleonly = $tagData['htmlsingleonly'];
 484                 $htmlnest = $tagData['htmlnest'];
 485                 $tabletags = $tagData['tabletags'];
 486                 $htmllist = $tagData['htmllist'];
 487                 $listtags = $tagData['listtags'];
 488                 $htmlsingleallowed = $tagData['htmlsingleallowed'];
 489                 $htmlelements = $tagData['htmlelements'];
 490
 491                 # Remove HTML comments
 492                 $text = self::removeHTMLcomments( $text );
 493                 $bits = explode( '<', $text );
 494                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 495                 if ( !MWTidy::isEnabled() ) {
 496                         $tagstack = $tablestack = [];
 497                         foreach ( $bits as $x ) {
 498                                 $regs = [];
 499                                 # $slash: Does the current element start with a '/'?
 500                                 # $t: Current element name
 501                                 # $params: String between element name and >
 502                                 # $brace: Ending '>' or '/>'
 503                                 # $rest: Everything until the next element of $bits
 504                                 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 505                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 506                                 } else {
 507                                         $slash = $t = $params = $brace = $rest = null;
 508                                 }
 509
 510                                 $badtag = false;
 511                                 $t = strtolower( $t );
 512                                 if ( isset( $htmlelements[$t] ) ) {
 513                                         # Check our stack
 514                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 515                                                 $badtag = true;
 516                                         } elseif ( $slash ) {
 517                                                 # Closing a tag... is it the one we just opened?
 518                                                 Wikimedia\suppressWarnings();
 519                                                 $ot = array_pop( $tagstack );
 520                                                 Wikimedia\restoreWarnings();
 521
 522                                                 if ( $ot != $t ) {
 523                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 524                                                                 # Pop all elements with an optional close tag
 525                                                                 # and see if we find a match below them
 526                                                                 $optstack = [];
 527                                                                 array_push( $optstack, $ot );
 528                                                                 Wikimedia\suppressWarnings();
 529                                                                 $ot = array_pop( $tagstack );
 530                                                                 Wikimedia\restoreWarnings();
 531                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 532                                                                         array_push( $optstack, $ot );
 533                                                                         Wikimedia\suppressWarnings();
 534                                                                         $ot = array_pop( $tagstack );
 535                                                                         Wikimedia\restoreWarnings();
 536                                                                 }
 537                                                                 if ( $t != $ot ) {
 538                                                                         # No match. Push the optional elements back again
 539                                                                         $badtag = true;
 540                                                                         Wikimedia\suppressWarnings();
 541                                                                         $ot = array_pop( $optstack );
 542                                                                         Wikimedia\restoreWarnings();
 543                                                                         while ( $ot ) {
 544                                                                                 array_push( $tagstack, $ot );
 545                                                                                 Wikimedia\suppressWarnings();
 546                                                                                 $ot = array_pop( $optstack );
 547                                                                                 Wikimedia\restoreWarnings();
 548                                                                         }
 549                                                                 }
 550                                                         } else {
 551                                                                 Wikimedia\suppressWarnings();
 552                                                                 array_push( $tagstack, $ot );
 553                                                                 Wikimedia\restoreWarnings();
 554
 555                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 556                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 557                                                                         $badtag = true;
 558                                                                 }
 559                                                         }
 560                                                 } else {
 561                                                         if ( $t == 'table' ) {
 562                                                                 $tagstack = array_pop( $tablestack );
 563                                                         }
 564                                                 }
 565                                                 $newparams = '';
 566                                         } else {
 567                                                 # Keep track for later
 568                                                 if ( isset( $tabletags[$t] ) && !in_array( 'table', $tagstack ) ) {
 569                                                         $badtag = true;
 570                                                 } elseif ( in_array( $t, $tagstack ) && !isset( $htmlnest[$t] ) ) {
 571                                                         $badtag = true;
 572                                                 #  Is it a self closed htmlpair ? (T7487)
 573                                                 } elseif ( $brace == '/>' && isset( $htmlpairs[$t] ) ) {
 574                                                         // Eventually we'll just remove the self-closing
 575                                                         // slash, in order to be consistent with HTML5
 576                                                         // semantics.
 577                                                         // $brace = '>';
 578                                                         // For now, let's just warn authors to clean up.
 579                                                         if ( is_callable( $warnCallback ) ) {
 580                                                                 call_user_func_array( $warnCallback, [ 'deprecated-self-close-category' ] );
 581                                                         }
 582                                                         $badtag = true;
 583                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 584                                                         # Hack to force empty tag for unclosable elements
 585                                                         $brace = '/>';
 586                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 587                                                         # Hack to not close $htmlsingle tags
 588                                                         $brace = null;
 589                                                         # Still need to push this optionally-closed tag to
 590                                                         # the tag stack so that we can match end tags
 591                                                         # instead of marking them as bad.
 592                                                         array_push( $tagstack, $t );
 593                                                 } elseif ( isset( $tabletags[$t] ) && in_array( $t, $tagstack ) ) {
 594                                                         // New table tag but forgot to close the previous one
 595                                                         $text .= "</$t>";
 596                                                 } else {
 597                                                         if ( $t == 'table' ) {
 598                                                                 array_push( $tablestack, $tagstack );
 599                                                                 $tagstack = [];
 600                                                         }
 601                                                         array_push( $tagstack, $t );
 602                                                 }
 603
 604                                                 # Replace any variables or template parameters with
 605                                                 # plaintext results.
 606                                                 if ( is_callable( $processCallback ) ) {
 607                                                         call_user_func_array( $processCallback, [ &$params, $args ] );
 608                                                 }
 609
 610                                                 if ( !self::validateTag( $params, $t ) ) {
 611                                                         $badtag = true;
 612                                                 }
 613
 614                                                 # Strip non-approved attributes from the tag
 615                                                 $newparams = self::fixTagAttributes( $params, $t );
 616                                         }
 617                                         if ( !$badtag ) {
 618                                                 $rest = str_replace( '>', '&gt;', $rest );
 619                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 620                                                 $text .= "<$slash$t$newparams$close>$rest";
 621                                                 continue;
 622                                         }
 623                                 }
 624                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 625                         }
 626                         # Close off any remaining tags
 627                         while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
 628                                 $text .= "</$t>\n";
 629                                 if ( $t == 'table' ) {
 630                                         $tagstack = array_pop( $tablestack );
 631                                 }
 632                         }
 633                 } else {
 634                         # this might be possible using tidy itself
 635                         foreach ( $bits as $x ) {
 636                                 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 637                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 638
 639                                         $badtag = false;
 640                                         $t = strtolower( $t );
 641                                         if ( isset( $htmlelements[$t] ) ) {
 642                                                 if ( is_callable( $processCallback ) ) {
 643                                                         call_user_func_array( $processCallback, [ &$params, $args ] );
 644                                                 }
 645
 646                                                 if ( $brace == '/>' && !( isset( $htmlsingle[$t] ) || isset( $htmlsingleonly[$t] ) ) ) {
 647                                                         // Eventually we'll just remove the self-closing
 648                                                         // slash, in order to be consistent with HTML5
 649                                                         // semantics.
 650                                                         // $brace = '>';
 651                                                         // For now, let's just warn authors to clean up.
 652                                                         if ( is_callable( $warnCallback ) ) {
 653                                                                 call_user_func_array( $warnCallback, [ 'deprecated-self-close-category' ] );
 654                                                         }
 655                                                 }
 656                                                 if ( !self::validateTag( $params, $t ) ) {
 657                                                         $badtag = true;
 658                                                 }
 659
 660                                                 $newparams = self::fixTagAttributes( $params, $t );
 661                                                 if ( !$badtag ) {
 662                                                         if ( $brace === '/>' && !isset( $htmlsingleonly[$t] ) ) {
 663                                                                 # Interpret self-closing tags as empty tags even when
 664                                                                 # HTML 5 would interpret them as start tags. Such input
 665                                                                 # is commonly seen on Wikimedia wikis with this intention.
 666                                                                 $brace = "></$t>";
 667                                                         }
 668
 669                                                         $rest = str_replace( '>', '&gt;', $rest );
 670                                                         $text .= "<$slash$t$newparams$brace$rest";
 671                                                         continue;
 672                                                 }
 673                                         }
 674                                 }
 675                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 676                         }
 677                 }
 678                 return $text;
 679         }
 680
 681         /**
 682          * Remove '<!--', '-->', and everything between.
 683          * To avoid leaving blank lines, when a comment is both preceded
 684          * and followed by a newline (ignoring spaces), trim leading and
 685          * trailing spaces and one of the newlines.
 686          *
 687          * @param string $text
 688          * @return string
 689          */
 690         public static function removeHTMLcomments( $text ) {
 691                 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
 692                         $end = strpos( $text, '-->', $start + 4 );
 693                         if ( $end === false ) {
 694                                 # Unterminated comment; bail out
 695                                 break;
 696                         }
 697
 698                         $end += 3;
 699
 700                         # Trim space and newline if the comment is both
 701                         # preceded and followed by a newline
 702                         $spaceStart = max( $start - 1, 0 );
 703                         $spaceLen = $end - $spaceStart;
 704                         while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
 705                                 $spaceStart--;
 706                                 $spaceLen++;
 707                         }
 708                         while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
 709                                 $spaceLen++;
 710                         }
 711                         if ( substr( $text, $spaceStart, 1 ) === "\n"
 712                                 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
 713                                 # Remove the comment, leading and trailing
 714                                 # spaces, and leave only one newline.
 715                                 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
 716                         } else {
 717                                 # Remove just the comment.
 718                                 $text = substr_replace( $text, '', $start, $end - $start );
 719                         }
 720                 }
 721                 return $text;
 722         }
 723
 724         /**
 725          * Takes attribute names and values for a tag and the tag name and
 726          * validates that the tag is allowed to be present.
 727          * This DOES NOT validate the attributes, nor does it validate the
 728          * tags themselves. This method only handles the special circumstances
 729          * where we may want to allow a tag within content but ONLY when it has
 730          * specific attributes set.
 731          *
 732          * @param string $params
 733          * @param string $element
 734          * @return bool
 735          */
 736         static function validateTag( $params, $element ) {
 737                 $params = self::decodeTagAttributes( $params );
 738
 739                 if ( $element == 'meta' || $element == 'link' ) {
 740                         if ( !isset( $params['itemprop'] ) ) {
 741                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 742                                 return false;
 743                         }
 744                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 745                                 // <meta> must have a content="" for the itemprop
 746                                 return false;
 747                         }
 748                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 749                                 // <link> must have an associated href=""
 750                                 return false;
 751                         }
 752                 }
 753
 754                 return true;
 755         }
 756
 757         /**
 758          * Take an array of attribute names and values and normalize or discard
 759          * illegal values for the given element type.
 760          *
 761          * - Discards attributes not on a whitelist for the given element
 762          * - Unsafe style attributes are discarded
 763          * - Invalid id attributes are re-encoded
 764          *
 765          * @param array $attribs
 766          * @param string $element
 767          * @return array
 768          *
 769          * @todo Check for legal values where the DTD limits things.
 770          * @todo Check for unique id attribute :P
 771          */
 772         static function validateTagAttributes( $attribs, $element ) {
 773                 return self::validateAttributes( $attribs,
 774                         self::attributeWhitelist( $element ) );
 775         }
 776
 777         /**
 778          * Take an array of attribute names and values and normalize or discard
 779          * illegal values for the given whitelist.
 780          *
 781          * - Discards attributes not on the given whitelist
 782          * - Unsafe style attributes are discarded
 783          * - Invalid id attributes are re-encoded
 784          *
 785          * @param array $attribs
 786          * @param array $whitelist List of allowed attribute names
 787          * @return array
 788          *
 789          * @todo Check for legal values where the DTD limits things.
 790          * @todo Check for unique id attribute :P
 791          */
 792         static function validateAttributes( $attribs, $whitelist ) {
 793                 $whitelist = array_flip( $whitelist );
 794                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 795
 796                 $out = [];
 797                 foreach ( $attribs as $attribute => $value ) {
 798                         # Allow XML namespace declaration to allow RDFa
 799                         if ( preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 800                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 801                                         $out[$attribute] = $value;
 802                                 }
 803
 804                                 continue;
 805                         }
 806
 807                         # Allow any attribute beginning with "data-"
 808                         # However:
 809                         # * Disallow data attributes used by MediaWiki code
 810                         # * Ensure that the attribute is not namespaced by banning
 811                         #   colons.
 812                         if ( !preg_match( '/^data-[^:]*$/i', $attribute )
 813                                 && !isset( $whitelist[$attribute] )
 814                                 || self::isReservedDataAttribute( $attribute )
 815                         ) {
 816                                 continue;
 817                         }
 818
 819                         # Strip javascript "expression" from stylesheets.
 820                         # https://msdn.microsoft.com/en-us/library/ms537634.aspx
 821                         if ( $attribute == 'style' ) {
 822                                 $value = self::checkCss( $value );
 823                         }
 824
 825                         # Escape HTML id attributes
 826                         if ( $attribute === 'id' ) {
 827                                 $value = self::escapeIdForAttribute( $value, self::ID_PRIMARY );
 828                         }
 829
 830                         # Escape HTML id reference lists
 831                         if ( $attribute === 'aria-describedby'
 832                                 || $attribute === 'aria-flowto'
 833                                 || $attribute === 'aria-labelledby'
 834                                 || $attribute === 'aria-owns'
 835                         ) {
 836                                 $value = self::escapeIdReferenceList( $value );
 837                         }
 838
 839                         // RDFa and microdata properties allow URLs, URIs and/or CURIs.
 840                         // Check them for sanity.
 841                         if ( $attribute === 'rel' || $attribute === 'rev'
 842                                 # RDFa
 843                                 || $attribute === 'about' || $attribute === 'property'
 844                                 || $attribute === 'resource' || $attribute === 'datatype'
 845                                 || $attribute === 'typeof'
 846                                 # HTML5 microdata
 847                                 || $attribute === 'itemid' || $attribute === 'itemprop'
 848                                 || $attribute === 'itemref' || $attribute === 'itemscope'
 849                                 || $attribute === 'itemtype'
 850                         ) {
 851                                 // Paranoia. Allow "simple" values but suppress javascript
 852                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 853                                         continue;
 854                                 }
 855                         }
 856
 857                         # NOTE: even though elements using href/src are not allowed directly, supply
 858                         #       validation code that can be used by tag hook handlers, etc
 859                         if ( $attribute === 'href' || $attribute === 'src' || $attribute === 'poster' ) {
 860                                 if ( !preg_match( $hrefExp, $value ) ) {
 861                                         continue; // drop any href or src attributes not using an allowed protocol.
 862                                         // NOTE: this also drops all relative URLs
 863                                 }
 864                         }
 865
 866                         // If this attribute was previously set, override it.
 867                         // Output should only have one attribute of each name.
 868                         $out[$attribute] = $value;
 869                 }
 870
 871                 # itemtype, itemid, itemref don't make sense without itemscope
 872                 if ( !array_key_exists( 'itemscope', $out ) ) {
 873                         unset( $out['itemtype'] );
 874                         unset( $out['itemid'] );
 875                         unset( $out['itemref'] );
 876                 }
 877                 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 878
 879                 return $out;
 880         }
 881
 882         /**
 883          * Given an attribute name, checks whether it is a reserved data attribute
 884          * (such as data-mw-foo) which is unavailable to user-generated HTML so MediaWiki
 885          * core and extension code can safely use it to communicate with frontend code.
 886          * @param string $attr Attribute name.
 887          * @return bool
 888          */
 889         public static function isReservedDataAttribute( $attr ) {
 890                 // data-ooui is reserved for ooui.
 891                 // data-mw and data-parsoid are reserved for parsoid.
 892                 // data-mw-<name here> is reserved for extensions (or core) if
 893                 // they need to communicate some data to the client and want to be
 894                 // sure that it isn't coming from an untrusted user.
 895                 // We ignore the possibility of namespaces since user-generated HTML
 896                 // can't use them anymore.
 897                 return (bool)preg_match( '/^data-(ooui|mw|parsoid)/i', $attr );
 898         }
 899
 900         /**
 901          * Merge two sets of HTML attributes.  Conflicting items in the second set
 902          * will override those in the first, except for 'class' attributes which
 903          * will be combined (if they're both strings).
 904          *
 905          * @todo implement merging for other attributes such as style
 906          * @param array $a
 907          * @param array $b
 908          * @return array
 909          */
 910         static function mergeAttributes( $a, $b ) {
 911                 $out = array_merge( $a, $b );
 912                 if ( isset( $a['class'] ) && isset( $b['class'] )
 913                         && is_string( $a['class'] ) && is_string( $b['class'] )
 914                         && $a['class'] !== $b['class']
 915                 ) {
 916                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 917                                 -1, PREG_SPLIT_NO_EMPTY );
 918                         $out['class'] = implode( ' ', array_unique( $classes ) );
 919                 }
 920                 return $out;
 921         }
 922
 923         /**
 924          * Normalize CSS into a format we can easily search for hostile input
 925          *  - decode character references
 926          *  - decode escape sequences
 927          *  - convert characters that IE6 interprets into ascii
 928          *  - remove comments, unless the entire value is one single comment
 929          * @param string $value the css string
 930          * @return string normalized css
 931          */
 932         public static function normalizeCss( $value ) {
 933                 // Decode character references like &#123;
 934                 $value = self::decodeCharReferences( $value );
 935
 936                 // Decode escape sequences and line continuation
 937                 // See the grammar in the CSS 2 spec, appendix D.
 938                 // This has to be done AFTER decoding character references.
 939                 // This means it isn't possible for this function to return
 940                 // unsanitized escape sequences. It is possible to manufacture
 941                 // input that contains character references that decode to
 942                 // escape sequences that decode to character references, but
 943                 // it's OK for the return value to contain character references
 944                 // because the caller is supposed to escape those anyway.
 945                 static $decodeRegex;
 946                 if ( !$decodeRegex ) {
 947                         $space = '[\\x20\\t\\r\\n\\f]';
 948                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 949                         $backslash = '\\\\';
 950                         $decodeRegex = "/ $backslash
 951                                 (?:
 952                                         ($nl) |  # 1. Line continuation
 953                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 954                                         (.) | # 3. backslash cancelling special meaning
 955                                         () | # 4. backslash at end of string
 956                                 )/xu";
 957                 }
 958                 $value = preg_replace_callback( $decodeRegex,
 959                         [ __CLASS__, 'cssDecodeCallback' ], $value );
 960
 961                 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
 962                 $value = preg_replace_callback(
 963                         '/[！-［］-ｚ]/u', // U+FF01 to U+FF5A, excluding U+FF3C (T60088)
 964                         function ( $matches ) {
 965                                 $cp = UtfNormal\Utils::utf8ToCodepoint( $matches[0] );
 966                                 if ( $cp === false ) {
 967                                         return '';
 968                                 }
 969                                 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
 970                         },
 971                         $value
 972                 );
 973
 974                 // Convert more characters IE6 might treat as ascii
 975                 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
 976                 $value = str_replace(
 977                         [ 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ],
 978                         [ 'r', 'n', 'n', 'l', 'i', '(', '(' ],
 979                         $value
 980                 );
 981
 982                 // Let the value through if it's nothing but a single comment, to
 983                 // allow other functions which may reject it to pass some error
 984                 // message through.
 985                 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
 986                         // Remove any comments; IE gets token splitting wrong
 987                         // This must be done AFTER decoding character references and
 988                         // escape sequences, because those steps can introduce comments
 989                         // This step cannot introduce character references or escape
 990                         // sequences, because it replaces comments with spaces rather
 991                         // than removing them completely.
 992                         $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 993
 994                         // Remove anything after a comment-start token, to guard against
 995                         // incorrect client implementations.
 996                         $commentPos = strpos( $value, '/*' );
 997                         if ( $commentPos !== false ) {
 998                                 $value = substr( $value, 0, $commentPos );
 999                         }
1000                 }
1001
1002                 // S followed by repeat, iteration, or prolonged sound marks,
1003                 // which IE will treat as "ss"
1004                 $value = preg_replace(
1005                         '/s(?:
1006                                 \xE3\x80\xB1 | # U+3031
1007                                 \xE3\x82\x9D | # U+309D
1008                                 \xE3\x83\xBC | # U+30FC
1009                                 \xE3\x83\xBD | # U+30FD
1010                                 \xEF\xB9\xBC | # U+FE7C
1011                                 \xEF\xB9\xBD | # U+FE7D
1012                                 \xEF\xBD\xB0   # U+FF70
1013                         )/ix',
1014                         'ss',
1015                         $value
1016                 );
1017
1018                 return $value;
1019         }
1020
1021         /**
1022          * Pick apart some CSS and check it for forbidden or unsafe structures.
1023          * Returns a sanitized string. This sanitized string will have
1024          * character references and escape sequences decoded and comments
1025          * stripped (unless it is itself one valid comment, in which case the value
1026          * will be passed through). If the input is just too evil, only a comment
1027          * complaining about evilness will be returned.
1028          *
1029          * Currently URL references, 'expression', 'tps' are forbidden.
1030          *
1031          * NOTE: Despite the fact that character references are decoded, the
1032          * returned string may contain character references given certain
1033          * clever input strings. These character references must
1034          * be escaped before the return value is embedded in HTML.
1035          *
1036          * @param string $value
1037          * @return string
1038          */
1039         static function checkCss( $value ) {
1040                 $value = self::normalizeCss( $value );
1041
1042                 // Reject problematic keywords and control characters
1043                 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ||
1044                         strpos( $value, UtfNormal\Constants::UTF8_REPLACEMENT ) !== false ) {
1045                         return '/* invalid control char */';
1046                 } elseif ( preg_match(
1047                         '! expression
1048                                 | filter\s*:
1049                                 | accelerator\s*:
1050                                 | -o-link\s*:
1051                                 | -o-link-source\s*:
1052                                 | -o-replace\s*:
1053                                 | url\s*\(
1054                                 | image\s*\(
1055                                 | image-set\s*\(
1056                                 | attr\s*\([^)]+[\s,]+url
1057                                 | var\s*\(
1058                         !ix', $value ) ) {
1059                         return '/* insecure input */';
1060                 }
1061                 return $value;
1062         }
1063
1064         /**
1065          * @param array $matches
1066          * @return string
1067          */
1068         static function cssDecodeCallback( $matches ) {
1069                 if ( $matches[1] !== '' ) {
1070                         // Line continuation
1071                         return '';
1072                 } elseif ( $matches[2] !== '' ) {
1073                         $char = UtfNormal\Utils::codepointToUtf8( hexdec( $matches[2] ) );
1074                 } elseif ( $matches[3] !== '' ) {
1075                         $char = $matches[3];
1076                 } else {
1077                         $char = '\\';
1078                 }
1079                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
1080                         // These characters need to be escaped in strings
1081                         // Clean up the escape sequence to avoid parsing errors by clients
1082                         return '\\' . dechex( ord( $char ) ) . ' ';
1083                 } else {
1084                         // Decode unnecessary escape
1085                         return $char;
1086                 }
1087         }
1088
1089         /**
1090          * Take a tag soup fragment listing an HTML element's attributes
1091          * and normalize it to well-formed XML, discarding unwanted attributes.
1092          * Output is safe for further wikitext processing, with escaping of
1093          * values that could trigger problems.
1094          *
1095          * - Normalizes attribute names to lowercase
1096          * - Discards attributes not on a whitelist for the given element
1097          * - Turns broken or invalid entities into plaintext
1098          * - Double-quotes all attribute values
1099          * - Attributes without values are given the name as attribute
1100          * - Double attributes are discarded
1101          * - Unsafe style attributes are discarded
1102          * - Prepends space if there are attributes.
1103          * - (Optionally) Sorts attributes by name.
1104          *
1105          * @param string $text
1106          * @param string $element
1107          * @param bool $sorted Whether to sort the attributes (default: false)
1108          * @return string
1109          */
1110         static function fixTagAttributes( $text, $element, $sorted = false ) {
1111                 if ( trim( $text ) == '' ) {
1112                         return '';
1113                 }
1114
1115                 $decoded = self::decodeTagAttributes( $text );
1116                 $stripped = self::validateTagAttributes( $decoded, $element );
1117
1118                 if ( $sorted ) {
1119                         ksort( $stripped );
1120                 }
1121
1122                 return self::safeEncodeTagAttributes( $stripped );
1123         }
1124
1125         /**
1126          * Encode an attribute value for HTML output.
1127          * @param string $text
1128          * @return string HTML-encoded text fragment
1129          */
1130         static function encodeAttribute( $text ) {
1131                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1132
1133                 // Whitespace is normalized during attribute decoding,
1134                 // so if we've been passed non-spaces we must encode them
1135                 // ahead of time or they won't be preserved.
1136                 $encValue = strtr( $encValue, [
1137                         "\n" => '&#10;',
1138                         "\r" => '&#13;',
1139                         "\t" => '&#9;',
1140                 ] );
1141
1142                 return $encValue;
1143         }
1144
1145         /**
1146          * Encode an attribute value for HTML tags, with extra armoring
1147          * against further wiki processing.
1148          * @param string $text
1149          * @return string HTML-encoded text fragment
1150          */
1151         static function safeEncodeAttribute( $text ) {
1152                 $encValue = self::encodeAttribute( $text );
1153
1154                 # Templates and links may be expanded in later parsing,
1155                 # creating invalid or dangerous output. Suppress this.
1156                 $encValue = strtr( $encValue, [
1157                         '<'    => '&lt;',   // This should never happen,
1158                         '>'    => '&gt;',   // we've received invalid input
1159                         '"'    => '&quot;', // which should have been escaped.
1160                         '{'    => '&#123;',
1161                         '}'    => '&#125;', // prevent unpaired language conversion syntax
1162                         '['    => '&#91;',
1163                         ']'    => '&#93;',
1164                         "''"   => '&#39;&#39;',
1165                         'ISBN' => '&#73;SBN',
1166                         'RFC'  => '&#82;FC',
1167                         'PMID' => '&#80;MID',
1168                         '|'    => '&#124;',
1169                         '__'   => '&#95;_',
1170                 ] );
1171
1172                 # Stupid hack
1173                 $encValue = preg_replace_callback(
1174                         '/((?i)' . wfUrlProtocols() . ')/',
1175                         function ( $matches ) {
1176                                 return str_replace( ':', '&#58;', $matches[1] );
1177                         },
1178                         $encValue );
1179                 return $encValue;
1180         }
1181
1182         /**
1183          * Given a value, escape it so that it can be used in an id attribute and
1184          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1185          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1186          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1187          * escaped with lots of dots.
1188          *
1189          * To ensure we don't have to bother escaping anything, we also strip ', ",
1190          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1191          * We also strip # because it upsets IE, and % because it could be
1192          * ambiguous if it's part of something that looks like a percent escape
1193          * (which don't work reliably in fragments cross-browser).
1194          *
1195          * @deprecated since 1.30, use one of this class' escapeIdFor*() functions
1196          *
1197          * @see https://www.w3.org/TR/html401/types.html#type-name Valid characters
1198          *   in the id and name attributes
1199          * @see https://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1200          *   the id attribute
1201          * @see https://www.w3.org/TR/html5/dom.html#the-id-attribute
1202          *   HTML5 definition of id attribute
1203          *
1204          * @param string $id Id to escape
1205          * @param string|array $options String or array of strings (default is array()):
1206          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1207          *       so don't pay attention if the first character isn't valid at the
1208          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1209          *       false.
1210          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1211          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1212          *       anchors and links won't break.
1213          * @return string
1214          */
1215         static function escapeId( $id, $options = [] ) {
1216                 global $wgExperimentalHtmlIds;
1217                 $options = (array)$options;
1218
1219                 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1220                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1221                         $id = trim( $id, '_' );
1222                         if ( $id === '' ) {
1223                                 // Must have been all whitespace to start with.
1224                                 return '_';
1225                         } else {
1226                                 return $id;
1227                         }
1228                 }
1229
1230                 // HTML4-style escaping
1231                 static $replace = [
1232                         '%3A' => ':',
1233                         '%' => '.'
1234                 ];
1235
1236                 $id = urlencode( strtr( $id, ' ', '_' ) );
1237                 $id = strtr( $id, $replace );
1238
1239                 if ( !preg_match( '/^[a-zA-Z]/', $id ) && !in_array( 'noninitial', $options ) ) {
1240                         // Initial character must be a letter!
1241                         $id = "x$id";
1242                 }
1243                 return $id;
1244         }
1245
1246         /**
1247          * Given a section name or other user-generated or otherwise unsafe string, escapes it to be
1248          * a valid HTML id attribute.
1249          *
1250          * WARNING: unlike escapeId(), the output of this function is not guaranteed to be HTML safe,
1251          * be sure to use proper escaping.
1252          *
1253          * @param string $id String to escape
1254          * @param int $mode One of ID_* constants, specifying whether the primary or fallback encoding
1255          *     should be used.
1256          * @return string|bool Escaped ID or false if fallback encoding is requested but it's not
1257          *     configured.
1258          *
1259          * @since 1.30
1260          */
1261         public static function escapeIdForAttribute( $id, $mode = self::ID_PRIMARY ) {
1262                 global $wgFragmentMode;
1263
1264                 if ( !isset( $wgFragmentMode[$mode] ) ) {
1265                         if ( $mode === self::ID_PRIMARY ) {
1266                                 throw new UnexpectedValueException( '$wgFragmentMode is configured with no primary mode' );
1267                         }
1268                         return false;
1269                 }
1270
1271                 $internalMode = $wgFragmentMode[$mode];
1272
1273                 return self::escapeIdInternal( $id, $internalMode );
1274         }
1275
1276         /**
1277          * Given a section name or other user-generated or otherwise unsafe string, escapes it to be
1278          * a valid URL fragment.
1279          *
1280          * WARNING: unlike escapeId(), the output of this function is not guaranteed to be HTML safe,
1281          * be sure to use proper escaping.
1282          *
1283          * @param string $id String to escape
1284          * @return string Escaped ID
1285          *
1286          * @since 1.30
1287          */
1288         public static function escapeIdForLink( $id ) {
1289                 global $wgFragmentMode;
1290
1291                 if ( !isset( $wgFragmentMode[self::ID_PRIMARY] ) ) {
1292                         throw new UnexpectedValueException( '$wgFragmentMode is configured with no primary mode' );
1293                 }
1294
1295                 $mode = $wgFragmentMode[self::ID_PRIMARY];
1296
1297                 $id = self::escapeIdInternal( $id, $mode );
1298
1299                 return $id;
1300         }
1301
1302         /**
1303          * Given a section name or other user-generated or otherwise unsafe string, escapes it to be
1304          * a valid URL fragment for external interwikis.
1305          *
1306          * @param string $id String to escape
1307          * @return string Escaped ID
1308          *
1309          * @since 1.30
1310          */
1311         public static function escapeIdForExternalInterwiki( $id ) {
1312                 global $wgExternalInterwikiFragmentMode;
1313
1314                 $id = self::escapeIdInternal( $id, $wgExternalInterwikiFragmentMode );
1315
1316                 return $id;
1317         }
1318
1319         /**
1320          * Helper for escapeIdFor*() functions. Performs most of the actual escaping.
1321          *
1322          * @param string $id String to escape
1323          * @param string $mode One of modes from $wgFragmentMode
1324          * @return string
1325          */
1326         private static function escapeIdInternal( $id, $mode ) {
1327                 switch ( $mode ) {
1328                         case 'html5':
1329                                 $id = str_replace( ' ', '_', $id );
1330                                 break;
1331                         case 'legacy':
1332                                 // This corresponds to 'noninitial' mode of the old escapeId()
1333                                 static $replace = [
1334                                         '%3A' => ':',
1335                                         '%' => '.'
1336                                 ];
1337
1338                                 $id = urlencode( str_replace( ' ', '_', $id ) );
1339                                 $id = strtr( $id, $replace );
1340                                 break;
1341                         case 'html5-legacy':
1342                                 $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1343                                 $id = trim( $id, '_' );
1344                                 if ( $id === '' ) {
1345                                         // Must have been all whitespace to start with.
1346                                         $id = '_';
1347                                 }
1348                                 break;
1349                         default:
1350                                 throw new InvalidArgumentException( "Invalid mode '$mode' passed to '" . __METHOD__ );
1351                 }
1352
1353                 return $id;
1354         }
1355
1356         /**
1357          * Given a string containing a space delimited list of ids, escape each id
1358          * to match ids escaped by the escapeId() function.
1359          *
1360          * @todo remove $options completely in 1.32
1361          *
1362          * @since 1.27
1363          *
1364          * @param string $referenceString Space delimited list of ids
1365          * @param string|array $options Deprecated and does nothing.
1366          * @return string
1367          */
1368         static function escapeIdReferenceList( $referenceString, $options = [] ) {
1369                 if ( $options ) {
1370                         wfDeprecated( __METHOD__ . ' with $options', '1.31' );
1371                 }
1372                 # Explode the space delimited list string into an array of tokens
1373                 $references = preg_split( '/\s+/', "{$referenceString}", -1, PREG_SPLIT_NO_EMPTY );
1374
1375                 # Escape each token as an id
1376                 foreach ( $references as &$ref ) {
1377                         $ref = self::escapeIdForAttribute( $ref );
1378                 }
1379
1380                 # Merge the array back to a space delimited list string
1381                 # If the array is empty, the result will be an empty string ('')
1382                 $referenceString = implode( ' ', $references );
1383
1384                 return $referenceString;
1385         }
1386
1387         /**
1388          * Given a value, escape it so that it can be used as a CSS class and
1389          * return it.
1390          *
1391          * @todo For extra validity, input should be validated UTF-8.
1392          *
1393          * @see https://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1394          *
1395          * @param string $class
1396          * @return string
1397          */
1398         static function escapeClass( $class ) {
1399                 // Convert ugly stuff to underscores and kill underscores in ugly places
1400                 return rtrim( preg_replace(
1401                         [ '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ],
1402                         '_',
1403                         $class ), '_' );
1404         }
1405
1406         /**
1407          * Given HTML input, escape with htmlspecialchars but un-escape entities.
1408          * This allows (generally harmless) entities like &#160; to survive.
1409          *
1410          * @param string $html HTML to escape
1411          * @return string Escaped input
1412          */
1413         static function escapeHtmlAllowEntities( $html ) {
1414                 $html = self::decodeCharReferences( $html );
1415                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1416                 # hurt. Use ENT_SUBSTITUTE so that incorrectly truncated multibyte characters
1417                 # don't cause the entire string to disappear.
1418                 $html = htmlspecialchars( $html, ENT_QUOTES | ENT_SUBSTITUTE );
1419                 return $html;
1420         }
1421
1422         /**
1423          * Return an associative array of attribute names and values from
1424          * a partial tag string. Attribute names are forced to lowercase,
1425          * character references are decoded to UTF-8 text.
1426          *
1427          * @param string $text
1428          * @return array
1429          */
1430         public static function decodeTagAttributes( $text ) {
1431                 if ( trim( $text ) == '' ) {
1432                         return [];
1433                 }
1434
1435                 $attribs = [];
1436                 $pairs = [];
1437                 if ( !preg_match_all(
1438                         self::getAttribsRegex(),
1439                         $text,
1440                         $pairs,
1441                         PREG_SET_ORDER ) ) {
1442                         return $attribs;
1443                 }
1444
1445                 foreach ( $pairs as $set ) {
1446                         $attribute = strtolower( $set[1] );
1447                         $value = self::getTagAttributeCallback( $set );
1448
1449                         // Normalize whitespace
1450                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1451                         $value = trim( $value );
1452
1453                         // Decode character references
1454                         $attribs[$attribute] = self::decodeCharReferences( $value );
1455                 }
1456                 return $attribs;
1457         }
1458
1459         /**
1460          * Build a partial tag string from an associative array of attribute
1461          * names and values as returned by decodeTagAttributes.
1462          *
1463          * @param array $assoc_array
1464          * @return string
1465          */
1466         public static function safeEncodeTagAttributes( $assoc_array ) {
1467                 $attribs = [];
1468                 foreach ( $assoc_array as $attribute => $value ) {
1469                         $encAttribute = htmlspecialchars( $attribute );
1470                         $encValue = self::safeEncodeAttribute( $value );
1471
1472                         $attribs[] = "$encAttribute=\"$encValue\"";
1473                 }
1474                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1475         }
1476
1477         /**
1478          * Pick the appropriate attribute value from a match set from the
1479          * attribs regex matches.
1480          *
1481          * @param array $set
1482          * @throws MWException When tag conditions are not met.
1483          * @return string
1484          */
1485         private static function getTagAttributeCallback( $set ) {
1486                 if ( isset( $set[5] ) ) {
1487                         # No quotes.
1488                         return $set[5];
1489                 } elseif ( isset( $set[4] ) ) {
1490                         # Single-quoted
1491                         return $set[4];
1492                 } elseif ( isset( $set[3] ) ) {
1493                         # Double-quoted
1494                         return $set[3];
1495                 } elseif ( !isset( $set[2] ) ) {
1496                         # In XHTML, attributes must have a value so return an empty string.
1497                         # See "Empty attribute syntax",
1498                         # https://www.w3.org/TR/html5/syntax.html#syntax-attribute-name
1499                         return "";
1500                 } else {
1501                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1502                 }
1503         }
1504
1505         /**
1506          * @param string $text
1507          * @return string
1508          */
1509         private static function normalizeWhitespace( $text ) {
1510                 return preg_replace(
1511                         '/\r\n|[\x20\x0d\x0a\x09]/',
1512                         ' ',
1513                         $text );
1514         }
1515
1516         /**
1517          * Normalizes whitespace in a section name, such as might be returned
1518          * by Parser::stripSectionName(), for use in the id's that are used for
1519          * section links.
1520          *
1521          * @param string $section
1522          * @return string
1523          */
1524         static function normalizeSectionNameWhitespace( $section ) {
1525                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1526         }
1527
1528         /**
1529          * Ensure that any entities and character references are legal
1530          * for XML and XHTML specifically. Any stray bits will be
1531          * &amp;-escaped to result in a valid text fragment.
1532          *
1533          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1534          *   numericized (this way we're well-formed even without a DTD)
1535          * b. any numeric char refs must be legal chars, not invalid or forbidden
1536          * c. use lower cased "&#x", not "&#X"
1537          * d. fix or reject non-valid attributes
1538          *
1539          * @param string $text
1540          * @return string
1541          * @private
1542          */
1543         static function normalizeCharReferences( $text ) {
1544                 return preg_replace_callback(
1545                         self::CHAR_REFS_REGEX,
1546                         [ self::class, 'normalizeCharReferencesCallback' ],
1547                         $text );
1548         }
1549
1550         /**
1551          * @param string $matches
1552          * @return string
1553          */
1554         static function normalizeCharReferencesCallback( $matches ) {
1555                 $ret = null;
1556                 if ( $matches[1] != '' ) {
1557                         $ret = self::normalizeEntity( $matches[1] );
1558                 } elseif ( $matches[2] != '' ) {
1559                         $ret = self::decCharReference( $matches[2] );
1560                 } elseif ( $matches[3] != '' ) {
1561                         $ret = self::hexCharReference( $matches[3] );
1562                 }
1563                 if ( is_null( $ret ) ) {
1564                         return htmlspecialchars( $matches[0] );
1565                 } else {
1566                         return $ret;
1567                 }
1568         }
1569
1570         /**
1571          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1572          * return the equivalent numeric entity reference (except for the core &lt;
1573          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1574          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1575          * pseudo-entity source (eg &amp;foo;)
1576          *
1577          * @param string $name
1578          * @return string
1579          */
1580         static function normalizeEntity( $name ) {
1581                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1582                         return '&' . self::$htmlEntityAliases[$name] . ';';
1583                 } elseif ( in_array( $name, [ 'lt', 'gt', 'amp', 'quot' ] ) ) {
1584                         return "&$name;";
1585                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1586                         return '&#' . self::$htmlEntities[$name] . ';';
1587                 } else {
1588                         return "&amp;$name;";
1589                 }
1590         }
1591
1592         /**
1593          * @param int $codepoint
1594          * @return null|string
1595          */
1596         static function decCharReference( $codepoint ) {
1597                 $point = intval( $codepoint );
1598                 if ( self::validateCodepoint( $point ) ) {
1599                         return sprintf( '&#%d;', $point );
1600                 } else {
1601                         return null;
1602                 }
1603         }
1604
1605         /**
1606          * @param int $codepoint
1607          * @return null|string
1608          */
1609         static function hexCharReference( $codepoint ) {
1610                 $point = hexdec( $codepoint );
1611                 if ( self::validateCodepoint( $point ) ) {
1612                         return sprintf( '&#x%x;', $point );
1613                 } else {
1614                         return null;
1615                 }
1616         }
1617
1618         /**
1619          * Returns true if a given Unicode codepoint is a valid character in
1620          * both HTML5 and XML.
1621          * @param int $codepoint
1622          * @return bool
1623          */
1624         private static function validateCodepoint( $codepoint ) {
1625                 # U+000C is valid in HTML5 but not allowed in XML.
1626                 # U+000D is valid in XML but not allowed in HTML5.
1627                 # U+007F - U+009F are disallowed in HTML5 (control characters).
1628                 return $codepoint == 0x09
1629                         || $codepoint == 0x0a
1630                         || ( $codepoint >= 0x20 && $codepoint <= 0x7e )
1631                         || ( $codepoint >= 0xa0 && $codepoint <= 0xd7ff )
1632                         || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1633                         || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1634         }
1635
1636         /**
1637          * Decode any character references, numeric or named entities,
1638          * in the text and return a UTF-8 string.
1639          *
1640          * @param string $text
1641          * @return string
1642          */
1643         public static function decodeCharReferences( $text ) {
1644                 return preg_replace_callback(
1645                         self::CHAR_REFS_REGEX,
1646                         [ self::class, 'decodeCharReferencesCallback' ],
1647                         $text );
1648         }
1649
1650         /**
1651          * Decode any character references, numeric or named entities,
1652          * in the next and normalize the resulting string. (T16952)
1653          *
1654          * This is useful for page titles, not for text to be displayed,
1655          * MediaWiki allows HTML entities to escape normalization as a feature.
1656          *
1657          * @param string $text Already normalized, containing entities
1658          * @return string Still normalized, without entities
1659          */
1660         public static function decodeCharReferencesAndNormalize( $text ) {
1661                 global $wgContLang;
1662                 $text = preg_replace_callback(
1663                         self::CHAR_REFS_REGEX,
1664                         [ self::class, 'decodeCharReferencesCallback' ],
1665                         $text,
1666                         -1, //limit
1667                         $count
1668                 );
1669
1670                 if ( $count ) {
1671                         return $wgContLang->normalize( $text );
1672                 } else {
1673                         return $text;
1674                 }
1675         }
1676
1677         /**
1678          * @param string $matches
1679          * @return string
1680          */
1681         static function decodeCharReferencesCallback( $matches ) {
1682                 if ( $matches[1] != '' ) {
1683                         return self::decodeEntity( $matches[1] );
1684                 } elseif ( $matches[2] != '' ) {
1685                         return self::decodeChar( intval( $matches[2] ) );
1686                 } elseif ( $matches[3] != '' ) {
1687                         return self::decodeChar( hexdec( $matches[3] ) );
1688                 }
1689                 # Last case should be an ampersand by itself
1690                 return $matches[0];
1691         }
1692
1693         /**
1694          * Return UTF-8 string for a codepoint if that is a valid
1695          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1696          * @param int $codepoint
1697          * @return string
1698          * @private
1699          */
1700         static function decodeChar( $codepoint ) {
1701                 if ( self::validateCodepoint( $codepoint ) ) {
1702                         return UtfNormal\Utils::codepointToUtf8( $codepoint );
1703                 } else {
1704                         return UtfNormal\Constants::UTF8_REPLACEMENT;
1705                 }
1706         }
1707
1708         /**
1709          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1710          * return the UTF-8 encoding of that character. Otherwise, returns
1711          * pseudo-entity source (eg "&foo;")
1712          *
1713          * @param string $name
1714          * @return string
1715          */
1716         static function decodeEntity( $name ) {
1717                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1718                         $name = self::$htmlEntityAliases[$name];
1719                 }
1720                 if ( isset( self::$htmlEntities[$name] ) ) {
1721                         return UtfNormal\Utils::codepointToUtf8( self::$htmlEntities[$name] );
1722                 } else {
1723                         return "&$name;";
1724                 }
1725         }
1726
1727         /**
1728          * Fetch the whitelist of acceptable attributes for a given element name.
1729          *
1730          * @param string $element
1731          * @return array
1732          */
1733         static function attributeWhitelist( $element ) {
1734                 $list = self::setupAttributeWhitelist();
1735                 return isset( $list[$element] )
1736                         ? $list[$element]
1737                         : [];
1738         }
1739
1740         /**
1741          * Foreach array key (an allowed HTML element), return an array
1742          * of allowed attributes
1743          * @return array
1744          */
1745         static function setupAttributeWhitelist() {
1746                 static $whitelist;
1747
1748                 if ( $whitelist !== null ) {
1749                         return $whitelist;
1750                 }
1751
1752                 $common = [
1753                         # HTML
1754                         'id',
1755                         'class',
1756                         'style',
1757                         'lang',
1758                         'dir',
1759                         'title',
1760
1761                         # WAI-ARIA
1762                         'aria-describedby',
1763                         'aria-flowto',
1764                         'aria-label',
1765                         'aria-labelledby',
1766                         'aria-owns',
1767                         'role',
1768
1769                         # RDFa
1770                         # These attributes are specified in section 9 of
1771                         # https://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1772                         'about',
1773                         'property',
1774                         'resource',
1775                         'datatype',
1776                         'typeof',
1777
1778                         # Microdata. These are specified by
1779                         # https://html.spec.whatwg.org/multipage/microdata.html#the-microdata-model
1780                         'itemid',
1781                         'itemprop',
1782                         'itemref',
1783                         'itemscope',
1784                         'itemtype',
1785                 ];
1786
1787                 $block = array_merge( $common, [ 'align' ] );
1788                 $tablealign = [ 'align', 'valign' ];
1789                 $tablecell = [
1790                         'abbr',
1791                         'axis',
1792                         'headers',
1793                         'scope',
1794                         'rowspan',
1795                         'colspan',
1796                         'nowrap', # deprecated
1797                         'width', # deprecated
1798                         'height', # deprecated
1799                         'bgcolor', # deprecated
1800                 ];
1801
1802                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1803                 # See: https://www.w3.org/TR/html4/
1804                 $whitelist = [
1805                         # 7.5.4
1806                         'div'        => $block,
1807                         'center'     => $common, # deprecated
1808                         'span'       => $common,
1809
1810                         # 7.5.5
1811                         'h1'         => $block,
1812                         'h2'         => $block,
1813                         'h3'         => $block,
1814                         'h4'         => $block,
1815                         'h5'         => $block,
1816                         'h6'         => $block,
1817
1818                         # 7.5.6
1819                         # address
1820
1821                         # 8.2.4
1822                         'bdo'        => $common,
1823
1824                         # 9.2.1
1825                         'em'         => $common,
1826                         'strong'     => $common,
1827                         'cite'       => $common,
1828                         'dfn'        => $common,
1829                         'code'       => $common,
1830                         'samp'       => $common,
1831                         'kbd'        => $common,
1832                         'var'        => $common,
1833                         'abbr'       => $common,
1834                         # acronym
1835
1836                         # 9.2.2
1837                         'blockquote' => array_merge( $common, [ 'cite' ] ),
1838                         'q'          => array_merge( $common, [ 'cite' ] ),
1839
1840                         # 9.2.3
1841                         'sub'        => $common,
1842                         'sup'        => $common,
1843
1844                         # 9.3.1
1845                         'p'          => $block,
1846
1847                         # 9.3.2
1848                         'br'         => array_merge( $common, [ 'clear' ] ),
1849
1850                         # https://www.w3.org/TR/html5/text-level-semantics.html#the-wbr-element
1851                         'wbr'        => $common,
1852
1853                         # 9.3.4
1854                         'pre'        => array_merge( $common, [ 'width' ] ),
1855
1856                         # 9.4
1857                         'ins'        => array_merge( $common, [ 'cite', 'datetime' ] ),
1858                         'del'        => array_merge( $common, [ 'cite', 'datetime' ] ),
1859
1860                         # 10.2
1861                         'ul'         => array_merge( $common, [ 'type' ] ),
1862                         'ol'         => array_merge( $common, [ 'type', 'start', 'reversed' ] ),
1863                         'li'         => array_merge( $common, [ 'type', 'value' ] ),
1864
1865                         # 10.3
1866                         'dl'         => $common,
1867                         'dd'         => $common,
1868                         'dt'         => $common,
1869
1870                         # 11.2.1
1871                         'table'      => array_merge( $common,
1872                                                                 [ 'summary', 'width', 'border', 'frame',
1873                                                                                 'rules', 'cellspacing', 'cellpadding',
1874                                                                                 'align', 'bgcolor',
1875                                                                 ] ),
1876
1877                         # 11.2.2
1878                         'caption'    => $block,
1879
1880                         # 11.2.3
1881                         'thead'      => $common,
1882                         'tfoot'      => $common,
1883                         'tbody'      => $common,
1884
1885                         # 11.2.4
1886                         'colgroup'   => array_merge( $common, [ 'span' ] ),
1887                         'col'        => array_merge( $common, [ 'span' ] ),
1888
1889                         # 11.2.5
1890                         'tr'         => array_merge( $common, [ 'bgcolor' ], $tablealign ),
1891
1892                         # 11.2.6
1893                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1894                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1895
1896                         # 12.2
1897                         # NOTE: <a> is not allowed directly, but the attrib
1898                         # whitelist is used from the Parser object
1899                         'a'          => array_merge( $common, [ 'href', 'rel', 'rev' ] ), # rel/rev esp. for RDFa
1900
1901                         # 13.2
1902                         # Not usually allowed, but may be used for extension-style hooks
1903                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1904                         # true
1905                         'img'        => array_merge( $common, [ 'alt', 'src', 'width', 'height', 'srcset' ] ),
1906
1907                         'video'      => array_merge( $common, [ 'poster', 'controls', 'preload', 'width', 'height' ] ),
1908                         'source'     => array_merge( $common, [ 'type', 'src' ] ),
1909                         'track'      => array_merge( $common, [ 'type', 'src', 'srclang', 'kind', 'label' ] ),
1910
1911                         # 15.2.1
1912                         'tt'         => $common,
1913                         'b'          => $common,
1914                         'i'          => $common,
1915                         'big'        => $common,
1916                         'small'      => $common,
1917                         'strike'     => $common,
1918                         's'          => $common,
1919                         'u'          => $common,
1920
1921                         # 15.2.2
1922                         'font'       => array_merge( $common, [ 'size', 'color', 'face' ] ),
1923                         # basefont
1924
1925                         # 15.3
1926                         'hr'         => array_merge( $common, [ 'width' ] ),
1927
1928                         # HTML Ruby annotation text module, simple ruby only.
1929                         # https://www.w3.org/TR/html5/text-level-semantics.html#the-ruby-element
1930                         'ruby'       => $common,
1931                         # rbc
1932                         'rb'         => $common,
1933                         'rp'         => $common,
1934                         'rt'         => $common, # array_merge( $common, array( 'rbspan' ) ),
1935                         'rtc'        => $common,
1936
1937                         # MathML root element, where used for extensions
1938                         # 'title' may not be 100% valid here; it's XHTML
1939                         # https://www.w3.org/TR/REC-MathML/
1940                         'math'       => [ 'class', 'style', 'id', 'title' ],
1941
1942                         // HTML 5 section 4.5
1943                         'figure'     => $common,
1944                         'figcaption' => $common,
1945
1946                         # HTML 5 section 4.6
1947                         'bdi' => $common,
1948
1949                         # HTML5 elements, defined by:
1950                         # https://html.spec.whatwg.org/multipage/semantics.html#the-data-element
1951                         'data' => array_merge( $common, [ 'value' ] ),
1952                         'time' => array_merge( $common, [ 'datetime' ] ),
1953                         'mark' => $common,
1954
1955                         // meta and link are only permitted by removeHTMLtags when Microdata
1956                         // is enabled so we don't bother adding a conditional to hide these
1957                         // Also meta and link are only valid in WikiText as Microdata elements
1958                         // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1959                         // So we don't bother including $common attributes that have no purpose.
1960                         'meta' => [ 'itemprop', 'content' ],
1961                         'link' => [ 'itemprop', 'href', 'title' ],
1962                 ];
1963
1964                 return $whitelist;
1965         }
1966
1967         /**
1968          * Take a fragment of (potentially invalid) HTML and return
1969          * a version with any tags removed, encoded as plain text.
1970          *
1971          * Warning: this return value must be further escaped for literal
1972          * inclusion in HTML output as of 1.10!
1973          *
1974          * @param string $html HTML fragment
1975          * @return string
1976          */
1977         static function stripAllTags( $html ) {
1978                 // Use RemexHtml to tokenize $html and extract the text
1979                 $handler = new RemexStripTagHandler;
1980                 $tokenizer = new RemexHtml\Tokenizer\Tokenizer( $handler, $html, [
1981                         'ignoreErrors' => true,
1982                         // don't ignore char refs, we want them to be decoded
1983                         'ignoreNulls' => true,
1984                         'skipPreprocess' => true,
1985                 ] );
1986                 $tokenizer->execute();
1987                 $text = $handler->getResult();
1988
1989                 $text = self::normalizeWhitespace( $text );
1990                 return $text;
1991         }
1992
1993         /**
1994          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1995          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1996          * PHP 5.1 doesn't.
1997          *
1998          * Use for passing XHTML fragments to PHP's XML parsing functions
1999          *
2000          * @return string
2001          */
2002         static function hackDocType() {
2003                 $out = "<!DOCTYPE html [\n";
2004                 foreach ( self::$htmlEntities as $entity => $codepoint ) {
2005                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
2006                 }
2007                 $out .= "]>\n";
2008                 return $out;
2009         }
2010
2011         /**
2012          * @param string $url
2013          * @return mixed|string
2014          */
2015         static function cleanUrl( $url ) {
2016                 # Normalize any HTML entities in input. They will be
2017                 # re-escaped by makeExternalLink().
2018                 $url = self::decodeCharReferences( $url );
2019
2020                 # Escape any control characters introduced by the above step
2021                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
2022                         [ __CLASS__, 'cleanUrlCallback' ], $url );
2023
2024                 # Validate hostname portion
2025                 $matches = [];
2026                 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
2027                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
2028
2029                         // Characters that will be ignored in IDNs.
2030                         // https://tools.ietf.org/html/rfc3454#section-3.1
2031                         // Strip them before further processing so blacklists and such work.
2032                         $strip = "/
2033                                 \\s|          # general whitespace
2034                                 \xc2\xad|     # 00ad SOFT HYPHEN
2035                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
2036                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
2037                                 \xe2\x81\xa0| # 2060 WORD JOINER
2038                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
2039                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
2040                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
2041                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
2042                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
2043                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
2044                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
2045                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
2046                                 /xuD";
2047
2048                         $host = preg_replace( $strip, '', $host );
2049
2050                         // IPv6 host names are bracketed with [].  Url-decode these.
2051                         if ( substr_compare( "//%5B", $host, 0, 5 ) === 0 &&
2052                                 preg_match( '!^//%5B([0-9A-Fa-f:.]+)%5D((:\d+)?)$!', $host, $matches )
2053                         ) {
2054                                 $host = '//[' . $matches[1] . ']' . $matches[2];
2055                         }
2056
2057                         // @todo FIXME: Validate hostnames here
2058
2059                         return $protocol . $host . $rest;
2060                 } else {
2061                         return $url;
2062                 }
2063         }
2064
2065         /**
2066          * @param array $matches
2067          * @return string
2068          */
2069         static function cleanUrlCallback( $matches ) {
2070                 return urlencode( $matches[0] );
2071         }
2072
2073         /**
2074          * Does a string look like an e-mail address?
2075          *
2076          * This validates an email address using an HTML5 specification found at:
2077          * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
2078          * Which as of 2011-01-24 says:
2079          *
2080          *   A valid e-mail address is a string that matches the ABNF production
2081          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
2082          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
2083          *   3.5.
2084          *
2085          * This function is an implementation of the specification as requested in
2086          * T24449.
2087          *
2088          * Client-side forms will use the same standard validation rules via JS or
2089          * HTML 5 validation; additional restrictions can be enforced server-side
2090          * by extensions via the 'isValidEmailAddr' hook.
2091          *
2092          * Note that this validation doesn't 100% match RFC 2822, but is believed
2093          * to be liberal enough for wide use. Some invalid addresses will still
2094          * pass validation here.
2095          *
2096          * @since 1.18
2097          *
2098          * @param string $addr E-mail address
2099          * @return bool
2100          */
2101         public static function validateEmail( $addr ) {
2102                 $result = null;
2103                 if ( !Hooks::run( 'isValidEmailAddr', [ $addr, &$result ] ) ) {
2104                         return $result;
2105                 }
2106
2107                 // Please note strings below are enclosed in brackets [], this make the
2108                 // hyphen "-" a range indicator. Hence it is double backslashed below.
2109                 // See T28948
2110                 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
2111                 $rfc1034_ldh_str = "a-z0-9\\-";
2112
2113                 $html5_email_regexp = "/
2114                 ^                      # start of string
2115                 [$rfc5322_atext\\.]+    # user part which is liberal :p
2116                 @                      # 'apostrophe'
2117                 [$rfc1034_ldh_str]+       # First domain part
2118                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
2119                 $                      # End of string
2120                 /ix"; // case Insensitive, eXtended
2121
2122                 return (bool)preg_match( $html5_email_regexp, $addr );
2123         }
2124 }