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                         !ix', $value ) ) {
1058                         return '/* insecure input */';
1059                 }
1060                 return $value;
1061         }
1062
1063         /**
1064          * @param array $matches
1065          * @return string
1066          */
1067         static function cssDecodeCallback( $matches ) {
1068                 if ( $matches[1] !== '' ) {
1069                         // Line continuation
1070                         return '';
1071                 } elseif ( $matches[2] !== '' ) {
1072                         $char = UtfNormal\Utils::codepointToUtf8( hexdec( $matches[2] ) );
1073                 } elseif ( $matches[3] !== '' ) {
1074                         $char = $matches[3];
1075                 } else {
1076                         $char = '\\';
1077                 }
1078                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
1079                         // These characters need to be escaped in strings
1080                         // Clean up the escape sequence to avoid parsing errors by clients
1081                         return '\\' . dechex( ord( $char ) ) . ' ';
1082                 } else {
1083                         // Decode unnecessary escape
1084                         return $char;
1085                 }
1086         }
1087
1088         /**
1089          * Take a tag soup fragment listing an HTML element's attributes
1090          * and normalize it to well-formed XML, discarding unwanted attributes.
1091          * Output is safe for further wikitext processing, with escaping of
1092          * values that could trigger problems.
1093          *
1094          * - Normalizes attribute names to lowercase
1095          * - Discards attributes not on a whitelist for the given element
1096          * - Turns broken or invalid entities into plaintext
1097          * - Double-quotes all attribute values
1098          * - Attributes without values are given the name as attribute
1099          * - Double attributes are discarded
1100          * - Unsafe style attributes are discarded
1101          * - Prepends space if there are attributes.
1102          * - (Optionally) Sorts attributes by name.
1103          *
1104          * @param string $text
1105          * @param string $element
1106          * @param bool $sorted Whether to sort the attributes (default: false)
1107          * @return string
1108          */
1109         static function fixTagAttributes( $text, $element, $sorted = false ) {
1110                 if ( trim( $text ) == '' ) {
1111                         return '';
1112                 }
1113
1114                 $decoded = self::decodeTagAttributes( $text );
1115                 $stripped = self::validateTagAttributes( $decoded, $element );
1116
1117                 if ( $sorted ) {
1118                         ksort( $stripped );
1119                 }
1120
1121                 return self::safeEncodeTagAttributes( $stripped );
1122         }
1123
1124         /**
1125          * Encode an attribute value for HTML output.
1126          * @param string $text
1127          * @return string HTML-encoded text fragment
1128          */
1129         static function encodeAttribute( $text ) {
1130                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1131
1132                 // Whitespace is normalized during attribute decoding,
1133                 // so if we've been passed non-spaces we must encode them
1134                 // ahead of time or they won't be preserved.
1135                 $encValue = strtr( $encValue, [
1136                         "\n" => '&#10;',
1137                         "\r" => '&#13;',
1138                         "\t" => '&#9;',
1139                 ] );
1140
1141                 return $encValue;
1142         }
1143
1144         /**
1145          * Armor French spaces with a replacement character
1146          *
1147          * @since 1.32
1148          * @param string $text Text to armor
1149          * @param string $space Space character for the French spaces, defaults to '&#160;'
1150          * @return string Armored text
1151          */
1152         public static function armorFrenchSpaces( $text, $space = '&#160;' ) {
1153                 // Replace $ with \$ and \ with \\
1154                 $space = preg_replace( '#(?<!\\\\)(\\$|\\\\)#', '\\\\$1', $space );
1155                 $fixtags = [
1156                         # French spaces, last one Guillemet-left
1157                         # only if there is something before the space
1158                         '/(.) (?=[?:;!%»])/u' => "\\1$space",
1159                         # French spaces, Guillemet-right
1160                         '/(«) /u' => "\\1$space",
1161                 ];
1162                 return preg_replace( array_keys( $fixtags ), array_values( $fixtags ), $text );
1163         }
1164
1165         /**
1166          * Encode an attribute value for HTML tags, with extra armoring
1167          * against further wiki processing.
1168          * @param string $text
1169          * @return string HTML-encoded text fragment
1170          */
1171         static function safeEncodeAttribute( $text ) {
1172                 $encValue = self::encodeAttribute( $text );
1173
1174                 # Templates and links may be expanded in later parsing,
1175                 # creating invalid or dangerous output. Suppress this.
1176                 $encValue = strtr( $encValue, [
1177                         '<'    => '&lt;',   // This should never happen,
1178                         '>'    => '&gt;',   // we've received invalid input
1179                         '"'    => '&quot;', // which should have been escaped.
1180                         '{'    => '&#123;',
1181                         '}'    => '&#125;', // prevent unpaired language conversion syntax
1182                         '['    => '&#91;',
1183                         ']'    => '&#93;',
1184                         "''"   => '&#39;&#39;',
1185                         'ISBN' => '&#73;SBN',
1186                         'RFC'  => '&#82;FC',
1187                         'PMID' => '&#80;MID',
1188                         '|'    => '&#124;',
1189                         '__'   => '&#95;_',
1190                 ] );
1191
1192                 # Armor against French spaces detection (T5158)
1193                 $encValue = self::armorFrenchSpaces( $encValue, '&#32;' );
1194
1195                 # Stupid hack
1196                 $encValue = preg_replace_callback(
1197                         '/((?i)' . wfUrlProtocols() . ')/',
1198                         function ( $matches ) {
1199                                 return str_replace( ':', '&#58;', $matches[1] );
1200                         },
1201                         $encValue );
1202                 return $encValue;
1203         }
1204
1205         /**
1206          * Given a value, escape it so that it can be used in an id attribute and
1207          * return it.  This will use HTML5 validation, allowing anything but ASCII
1208          * whitespace.
1209          *
1210          * To ensure we don't have to bother escaping anything, we also strip ', ".
1211          * TODO: Is this the best tactic?
1212          *
1213          * We also strip # because it upsets IE, and % because it could be
1214          * ambiguous if it's part of something that looks like a percent escape
1215          * (which don't work reliably in fragments cross-browser).
1216          *
1217          * @deprecated since 1.30, use one of this class' escapeIdFor*() functions
1218          *
1219          * @see https://www.w3.org/TR/html401/types.html#type-name Valid characters
1220          *   in the id and name attributes
1221          * @see https://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1222          *   the id attribute
1223          * @see https://www.w3.org/TR/html5/dom.html#the-id-attribute
1224          *   HTML5 definition of id attribute
1225          *
1226          * @param string $id Id to escape
1227          * @param string|array $options String or array of strings (default is array()):
1228          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1229          *       so don't pay attention if the first character isn't valid at the
1230          *       beginning of an id.
1231          * @return string
1232          */
1233         static function escapeId( $id, $options = [] ) {
1234                 $options = (array)$options;
1235
1236                 // HTML4-style escaping
1237                 static $replace = [
1238                         '%3A' => ':',
1239                         '%' => '.'
1240                 ];
1241
1242                 $id = urlencode( strtr( $id, ' ', '_' ) );
1243                 $id = strtr( $id, $replace );
1244
1245                 if ( !preg_match( '/^[a-zA-Z]/', $id ) && !in_array( 'noninitial', $options ) ) {
1246                         // Initial character must be a letter!
1247                         $id = "x$id";
1248                 }
1249                 return $id;
1250         }
1251
1252         /**
1253          * Given a section name or other user-generated or otherwise unsafe string, escapes it to be
1254          * a valid HTML id attribute.
1255          *
1256          * WARNING: unlike escapeId(), the output of this function is not guaranteed to be HTML safe,
1257          * be sure to use proper escaping.
1258          *
1259          * @param string $id String to escape
1260          * @param int $mode One of ID_* constants, specifying whether the primary or fallback encoding
1261          *     should be used.
1262          * @return string|bool Escaped ID or false if fallback encoding is requested but it's not
1263          *     configured.
1264          *
1265          * @since 1.30
1266          */
1267         public static function escapeIdForAttribute( $id, $mode = self::ID_PRIMARY ) {
1268                 global $wgFragmentMode;
1269
1270                 if ( !isset( $wgFragmentMode[$mode] ) ) {
1271                         if ( $mode === self::ID_PRIMARY ) {
1272                                 throw new UnexpectedValueException( '$wgFragmentMode is configured with no primary mode' );
1273                         }
1274                         return false;
1275                 }
1276
1277                 $internalMode = $wgFragmentMode[$mode];
1278
1279                 return self::escapeIdInternal( $id, $internalMode );
1280         }
1281
1282         /**
1283          * Given a section name or other user-generated or otherwise unsafe string, escapes it to be
1284          * a valid URL fragment.
1285          *
1286          * WARNING: unlike escapeId(), the output of this function is not guaranteed to be HTML safe,
1287          * be sure to use proper escaping.
1288          *
1289          * @param string $id String to escape
1290          * @return string Escaped ID
1291          *
1292          * @since 1.30
1293          */
1294         public static function escapeIdForLink( $id ) {
1295                 global $wgFragmentMode;
1296
1297                 if ( !isset( $wgFragmentMode[self::ID_PRIMARY] ) ) {
1298                         throw new UnexpectedValueException( '$wgFragmentMode is configured with no primary mode' );
1299                 }
1300
1301                 $mode = $wgFragmentMode[self::ID_PRIMARY];
1302
1303                 $id = self::escapeIdInternal( $id, $mode );
1304
1305                 return $id;
1306         }
1307
1308         /**
1309          * Given a section name or other user-generated or otherwise unsafe string, escapes it to be
1310          * a valid URL fragment for external interwikis.
1311          *
1312          * @param string $id String to escape
1313          * @return string Escaped ID
1314          *
1315          * @since 1.30
1316          */
1317         public static function escapeIdForExternalInterwiki( $id ) {
1318                 global $wgExternalInterwikiFragmentMode;
1319
1320                 $id = self::escapeIdInternal( $id, $wgExternalInterwikiFragmentMode );
1321
1322                 return $id;
1323         }
1324
1325         /**
1326          * Helper for escapeIdFor*() functions. Performs most of the actual escaping.
1327          *
1328          * @param string $id String to escape
1329          * @param string $mode One of modes from $wgFragmentMode
1330          * @return string
1331          */
1332         private static function escapeIdInternal( $id, $mode ) {
1333                 switch ( $mode ) {
1334                         case 'html5':
1335                                 $id = str_replace( ' ', '_', $id );
1336                                 break;
1337                         case 'legacy':
1338                                 // This corresponds to 'noninitial' mode of the old escapeId()
1339                                 static $replace = [
1340                                         '%3A' => ':',
1341                                         '%' => '.'
1342                                 ];
1343
1344                                 $id = urlencode( str_replace( ' ', '_', $id ) );
1345                                 $id = strtr( $id, $replace );
1346                                 break;
1347                         default:
1348                                 throw new InvalidArgumentException( "Invalid mode '$mode' passed to '" . __METHOD__ );
1349                 }
1350
1351                 return $id;
1352         }
1353
1354         /**
1355          * Given a string containing a space delimited list of ids, escape each id
1356          * to match ids escaped by the escapeId() function.
1357          *
1358          * @todo remove $options completely in 1.32
1359          *
1360          * @since 1.27
1361          *
1362          * @param string $referenceString Space delimited list of ids
1363          * @param string|array $options Deprecated and does nothing.
1364          * @return string
1365          */
1366         static function escapeIdReferenceList( $referenceString, $options = [] ) {
1367                 if ( $options ) {
1368                         wfDeprecated( __METHOD__ . ' with $options', '1.31' );
1369                 }
1370                 # Explode the space delimited list string into an array of tokens
1371                 $references = preg_split( '/\s+/', "{$referenceString}", -1, PREG_SPLIT_NO_EMPTY );
1372
1373                 # Escape each token as an id
1374                 foreach ( $references as &$ref ) {
1375                         $ref = self::escapeIdForAttribute( $ref );
1376                 }
1377
1378                 # Merge the array back to a space delimited list string
1379                 # If the array is empty, the result will be an empty string ('')
1380                 $referenceString = implode( ' ', $references );
1381
1382                 return $referenceString;
1383         }
1384
1385         /**
1386          * Given a value, escape it so that it can be used as a CSS class and
1387          * return it.
1388          *
1389          * @todo For extra validity, input should be validated UTF-8.
1390          *
1391          * @see https://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1392          *
1393          * @param string $class
1394          * @return string
1395          */
1396         static function escapeClass( $class ) {
1397                 // Convert ugly stuff to underscores and kill underscores in ugly places
1398                 return rtrim( preg_replace(
1399                         [ '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ],
1400                         '_',
1401                         $class ), '_' );
1402         }
1403
1404         /**
1405          * Given HTML input, escape with htmlspecialchars but un-escape entities.
1406          * This allows (generally harmless) entities like &#160; to survive.
1407          *
1408          * @param string $html HTML to escape
1409          * @return string Escaped input
1410          */
1411         static function escapeHtmlAllowEntities( $html ) {
1412                 $html = self::decodeCharReferences( $html );
1413                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1414                 # hurt. Use ENT_SUBSTITUTE so that incorrectly truncated multibyte characters
1415                 # don't cause the entire string to disappear.
1416                 $html = htmlspecialchars( $html, ENT_QUOTES | ENT_SUBSTITUTE );
1417                 return $html;
1418         }
1419
1420         /**
1421          * Return an associative array of attribute names and values from
1422          * a partial tag string. Attribute names are forced to lowercase,
1423          * character references are decoded to UTF-8 text.
1424          *
1425          * @param string $text
1426          * @return array
1427          */
1428         public static function decodeTagAttributes( $text ) {
1429                 if ( trim( $text ) == '' ) {
1430                         return [];
1431                 }
1432
1433                 $attribs = [];
1434                 $pairs = [];
1435                 if ( !preg_match_all(
1436                         self::getAttribsRegex(),
1437                         $text,
1438                         $pairs,
1439                         PREG_SET_ORDER ) ) {
1440                         return $attribs;
1441                 }
1442
1443                 foreach ( $pairs as $set ) {
1444                         $attribute = strtolower( $set[1] );
1445                         $value = self::getTagAttributeCallback( $set );
1446
1447                         // Normalize whitespace
1448                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1449                         $value = trim( $value );
1450
1451                         // Decode character references
1452                         $attribs[$attribute] = self::decodeCharReferences( $value );
1453                 }
1454                 return $attribs;
1455         }
1456
1457         /**
1458          * Build a partial tag string from an associative array of attribute
1459          * names and values as returned by decodeTagAttributes.
1460          *
1461          * @param array $assoc_array
1462          * @return string
1463          */
1464         public static function safeEncodeTagAttributes( $assoc_array ) {
1465                 $attribs = [];
1466                 foreach ( $assoc_array as $attribute => $value ) {
1467                         $encAttribute = htmlspecialchars( $attribute );
1468                         $encValue = self::safeEncodeAttribute( $value );
1469
1470                         $attribs[] = "$encAttribute=\"$encValue\"";
1471                 }
1472                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1473         }
1474
1475         /**
1476          * Pick the appropriate attribute value from a match set from the
1477          * attribs regex matches.
1478          *
1479          * @param array $set
1480          * @throws MWException When tag conditions are not met.
1481          * @return string
1482          */
1483         private static function getTagAttributeCallback( $set ) {
1484                 if ( isset( $set[5] ) ) {
1485                         # No quotes.
1486                         return $set[5];
1487                 } elseif ( isset( $set[4] ) ) {
1488                         # Single-quoted
1489                         return $set[4];
1490                 } elseif ( isset( $set[3] ) ) {
1491                         # Double-quoted
1492                         return $set[3];
1493                 } elseif ( !isset( $set[2] ) ) {
1494                         # In XHTML, attributes must have a value so return an empty string.
1495                         # See "Empty attribute syntax",
1496                         # https://www.w3.org/TR/html5/syntax.html#syntax-attribute-name
1497                         return "";
1498                 } else {
1499                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1500                 }
1501         }
1502
1503         /**
1504          * @param string $text
1505          * @return string
1506          */
1507         private static function normalizeWhitespace( $text ) {
1508                 return preg_replace(
1509                         '/\r\n|[\x20\x0d\x0a\x09]/',
1510                         ' ',
1511                         $text );
1512         }
1513
1514         /**
1515          * Normalizes whitespace in a section name, such as might be returned
1516          * by Parser::stripSectionName(), for use in the id's that are used for
1517          * section links.
1518          *
1519          * @param string $section
1520          * @return string
1521          */
1522         static function normalizeSectionNameWhitespace( $section ) {
1523                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1524         }
1525
1526         /**
1527          * Ensure that any entities and character references are legal
1528          * for XML and XHTML specifically. Any stray bits will be
1529          * &amp;-escaped to result in a valid text fragment.
1530          *
1531          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1532          *   numericized (this way we're well-formed even without a DTD)
1533          * b. any numeric char refs must be legal chars, not invalid or forbidden
1534          * c. use lower cased "&#x", not "&#X"
1535          * d. fix or reject non-valid attributes
1536          *
1537          * @param string $text
1538          * @return string
1539          * @private
1540          */
1541         static function normalizeCharReferences( $text ) {
1542                 return preg_replace_callback(
1543                         self::CHAR_REFS_REGEX,
1544                         [ self::class, 'normalizeCharReferencesCallback' ],
1545                         $text );
1546         }
1547
1548         /**
1549          * @param string $matches
1550          * @return string
1551          */
1552         static function normalizeCharReferencesCallback( $matches ) {
1553                 $ret = null;
1554                 if ( $matches[1] != '' ) {
1555                         $ret = self::normalizeEntity( $matches[1] );
1556                 } elseif ( $matches[2] != '' ) {
1557                         $ret = self::decCharReference( $matches[2] );
1558                 } elseif ( $matches[3] != '' ) {
1559                         $ret = self::hexCharReference( $matches[3] );
1560                 }
1561                 if ( is_null( $ret ) ) {
1562                         return htmlspecialchars( $matches[0] );
1563                 } else {
1564                         return $ret;
1565                 }
1566         }
1567
1568         /**
1569          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1570          * return the equivalent numeric entity reference (except for the core &lt;
1571          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1572          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1573          * pseudo-entity source (eg &amp;foo;)
1574          *
1575          * @param string $name
1576          * @return string
1577          */
1578         static function normalizeEntity( $name ) {
1579                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1580                         return '&' . self::$htmlEntityAliases[$name] . ';';
1581                 } elseif ( in_array( $name, [ 'lt', 'gt', 'amp', 'quot' ] ) ) {
1582                         return "&$name;";
1583                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1584                         return '&#' . self::$htmlEntities[$name] . ';';
1585                 } else {
1586                         return "&amp;$name;";
1587                 }
1588         }
1589
1590         /**
1591          * @param int $codepoint
1592          * @return null|string
1593          */
1594         static function decCharReference( $codepoint ) {
1595                 $point = intval( $codepoint );
1596                 if ( self::validateCodepoint( $point ) ) {
1597                         return sprintf( '&#%d;', $point );
1598                 } else {
1599                         return null;
1600                 }
1601         }
1602
1603         /**
1604          * @param int $codepoint
1605          * @return null|string
1606          */
1607         static function hexCharReference( $codepoint ) {
1608                 $point = hexdec( $codepoint );
1609                 if ( self::validateCodepoint( $point ) ) {
1610                         return sprintf( '&#x%x;', $point );
1611                 } else {
1612                         return null;
1613                 }
1614         }
1615
1616         /**
1617          * Returns true if a given Unicode codepoint is a valid character in
1618          * both HTML5 and XML.
1619          * @param int $codepoint
1620          * @return bool
1621          */
1622         private static function validateCodepoint( $codepoint ) {
1623                 # U+000C is valid in HTML5 but not allowed in XML.
1624                 # U+000D is valid in XML but not allowed in HTML5.
1625                 # U+007F - U+009F are disallowed in HTML5 (control characters).
1626                 return $codepoint == 0x09
1627                         || $codepoint == 0x0a
1628                         || ( $codepoint >= 0x20 && $codepoint <= 0x7e )
1629                         || ( $codepoint >= 0xa0 && $codepoint <= 0xd7ff )
1630                         || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1631                         || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1632         }
1633
1634         /**
1635          * Decode any character references, numeric or named entities,
1636          * in the text and return a UTF-8 string.
1637          *
1638          * @param string $text
1639          * @return string
1640          */
1641         public static function decodeCharReferences( $text ) {
1642                 return preg_replace_callback(
1643                         self::CHAR_REFS_REGEX,
1644                         [ self::class, 'decodeCharReferencesCallback' ],
1645                         $text );
1646         }
1647
1648         /**
1649          * Decode any character references, numeric or named entities,
1650          * in the next and normalize the resulting string. (T16952)
1651          *
1652          * This is useful for page titles, not for text to be displayed,
1653          * MediaWiki allows HTML entities to escape normalization as a feature.
1654          *
1655          * @param string $text Already normalized, containing entities
1656          * @return string Still normalized, without entities
1657          */
1658         public static function decodeCharReferencesAndNormalize( $text ) {
1659                 global $wgContLang;
1660                 $text = preg_replace_callback(
1661                         self::CHAR_REFS_REGEX,
1662                         [ self::class, 'decodeCharReferencesCallback' ],
1663                         $text,
1664                         -1, //limit
1665                         $count
1666                 );
1667
1668                 if ( $count ) {
1669                         return $wgContLang->normalize( $text );
1670                 } else {
1671                         return $text;
1672                 }
1673         }
1674
1675         /**
1676          * @param string $matches
1677          * @return string
1678          */
1679         static function decodeCharReferencesCallback( $matches ) {
1680                 if ( $matches[1] != '' ) {
1681                         return self::decodeEntity( $matches[1] );
1682                 } elseif ( $matches[2] != '' ) {
1683                         return self::decodeChar( intval( $matches[2] ) );
1684                 } elseif ( $matches[3] != '' ) {
1685                         return self::decodeChar( hexdec( $matches[3] ) );
1686                 }
1687                 # Last case should be an ampersand by itself
1688                 return $matches[0];
1689         }
1690
1691         /**
1692          * Return UTF-8 string for a codepoint if that is a valid
1693          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1694          * @param int $codepoint
1695          * @return string
1696          * @private
1697          */
1698         static function decodeChar( $codepoint ) {
1699                 if ( self::validateCodepoint( $codepoint ) ) {
1700                         return UtfNormal\Utils::codepointToUtf8( $codepoint );
1701                 } else {
1702                         return UtfNormal\Constants::UTF8_REPLACEMENT;
1703                 }
1704         }
1705
1706         /**
1707          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1708          * return the UTF-8 encoding of that character. Otherwise, returns
1709          * pseudo-entity source (eg "&foo;")
1710          *
1711          * @param string $name
1712          * @return string
1713          */
1714         static function decodeEntity( $name ) {
1715                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1716                         $name = self::$htmlEntityAliases[$name];
1717                 }
1718                 if ( isset( self::$htmlEntities[$name] ) ) {
1719                         return UtfNormal\Utils::codepointToUtf8( self::$htmlEntities[$name] );
1720                 } else {
1721                         return "&$name;";
1722                 }
1723         }
1724
1725         /**
1726          * Fetch the whitelist of acceptable attributes for a given element name.
1727          *
1728          * @param string $element
1729          * @return array
1730          */
1731         static function attributeWhitelist( $element ) {
1732                 $list = self::setupAttributeWhitelist();
1733                 return $list[$element] ?? [];
1734         }
1735
1736         /**
1737          * Foreach array key (an allowed HTML element), return an array
1738          * of allowed attributes
1739          * @return array
1740          */
1741         static function setupAttributeWhitelist() {
1742                 static $whitelist;
1743
1744                 if ( $whitelist !== null ) {
1745                         return $whitelist;
1746                 }
1747
1748                 $common = [
1749                         # HTML
1750                         'id',
1751                         'class',
1752                         'style',
1753                         'lang',
1754                         'dir',
1755                         'title',
1756
1757                         # WAI-ARIA
1758                         'aria-describedby',
1759                         'aria-flowto',
1760                         'aria-label',
1761                         'aria-labelledby',
1762                         'aria-owns',
1763                         'role',
1764
1765                         # RDFa
1766                         # These attributes are specified in section 9 of
1767                         # https://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1768                         'about',
1769                         'property',
1770                         'resource',
1771                         'datatype',
1772                         'typeof',
1773
1774                         # Microdata. These are specified by
1775                         # https://html.spec.whatwg.org/multipage/microdata.html#the-microdata-model
1776                         'itemid',
1777                         'itemprop',
1778                         'itemref',
1779                         'itemscope',
1780                         'itemtype',
1781                 ];
1782
1783                 $block = array_merge( $common, [ 'align' ] );
1784                 $tablealign = [ 'align', 'valign' ];
1785                 $tablecell = [
1786                         'abbr',
1787                         'axis',
1788                         'headers',
1789                         'scope',
1790                         'rowspan',
1791                         'colspan',
1792                         'nowrap', # deprecated
1793                         'width', # deprecated
1794                         'height', # deprecated
1795                         'bgcolor', # deprecated
1796                 ];
1797
1798                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1799                 # See: https://www.w3.org/TR/html4/
1800                 $whitelist = [
1801                         # 7.5.4
1802                         'div'        => $block,
1803                         'center'     => $common, # deprecated
1804                         'span'       => $common,
1805
1806                         # 7.5.5
1807                         'h1'         => $block,
1808                         'h2'         => $block,
1809                         'h3'         => $block,
1810                         'h4'         => $block,
1811                         'h5'         => $block,
1812                         'h6'         => $block,
1813
1814                         # 7.5.6
1815                         # address
1816
1817                         # 8.2.4
1818                         'bdo'        => $common,
1819
1820                         # 9.2.1
1821                         'em'         => $common,
1822                         'strong'     => $common,
1823                         'cite'       => $common,
1824                         'dfn'        => $common,
1825                         'code'       => $common,
1826                         'samp'       => $common,
1827                         'kbd'        => $common,
1828                         'var'        => $common,
1829                         'abbr'       => $common,
1830                         # acronym
1831
1832                         # 9.2.2
1833                         'blockquote' => array_merge( $common, [ 'cite' ] ),
1834                         'q'          => array_merge( $common, [ 'cite' ] ),
1835
1836                         # 9.2.3
1837                         'sub'        => $common,
1838                         'sup'        => $common,
1839
1840                         # 9.3.1
1841                         'p'          => $block,
1842
1843                         # 9.3.2
1844                         'br'         => array_merge( $common, [ 'clear' ] ),
1845
1846                         # https://www.w3.org/TR/html5/text-level-semantics.html#the-wbr-element
1847                         'wbr'        => $common,
1848
1849                         # 9.3.4
1850                         'pre'        => array_merge( $common, [ 'width' ] ),
1851
1852                         # 9.4
1853                         'ins'        => array_merge( $common, [ 'cite', 'datetime' ] ),
1854                         'del'        => array_merge( $common, [ 'cite', 'datetime' ] ),
1855
1856                         # 10.2
1857                         'ul'         => array_merge( $common, [ 'type' ] ),
1858                         'ol'         => array_merge( $common, [ 'type', 'start', 'reversed' ] ),
1859                         'li'         => array_merge( $common, [ 'type', 'value' ] ),
1860
1861                         # 10.3
1862                         'dl'         => $common,
1863                         'dd'         => $common,
1864                         'dt'         => $common,
1865
1866                         # 11.2.1
1867                         'table'      => array_merge( $common,
1868                                                                 [ 'summary', 'width', 'border', 'frame',
1869                                                                                 'rules', 'cellspacing', 'cellpadding',
1870                                                                                 'align', 'bgcolor',
1871                                                                 ] ),
1872
1873                         # 11.2.2
1874                         'caption'    => $block,
1875
1876                         # 11.2.3
1877                         'thead'      => $common,
1878                         'tfoot'      => $common,
1879                         'tbody'      => $common,
1880
1881                         # 11.2.4
1882                         'colgroup'   => array_merge( $common, [ 'span' ] ),
1883                         'col'        => array_merge( $common, [ 'span' ] ),
1884
1885                         # 11.2.5
1886                         'tr'         => array_merge( $common, [ 'bgcolor' ], $tablealign ),
1887
1888                         # 11.2.6
1889                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1890                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1891
1892                         # 12.2
1893                         # NOTE: <a> is not allowed directly, but the attrib
1894                         # whitelist is used from the Parser object
1895                         'a'          => array_merge( $common, [ 'href', 'rel', 'rev' ] ), # rel/rev esp. for RDFa
1896
1897                         # 13.2
1898                         # Not usually allowed, but may be used for extension-style hooks
1899                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1900                         # true
1901                         'img'        => array_merge( $common, [ 'alt', 'src', 'width', 'height', 'srcset' ] ),
1902
1903                         'video'      => array_merge( $common, [ 'poster', 'controls', 'preload', 'width', 'height' ] ),
1904                         'source'     => array_merge( $common, [ 'type', 'src' ] ),
1905                         'track'      => array_merge( $common, [ 'type', 'src', 'srclang', 'kind', 'label' ] ),
1906
1907                         # 15.2.1
1908                         'tt'         => $common,
1909                         'b'          => $common,
1910                         'i'          => $common,
1911                         'big'        => $common,
1912                         'small'      => $common,
1913                         'strike'     => $common,
1914                         's'          => $common,
1915                         'u'          => $common,
1916
1917                         # 15.2.2
1918                         'font'       => array_merge( $common, [ 'size', 'color', 'face' ] ),
1919                         # basefont
1920
1921                         # 15.3
1922                         'hr'         => array_merge( $common, [ 'width' ] ),
1923
1924                         # HTML Ruby annotation text module, simple ruby only.
1925                         # https://www.w3.org/TR/html5/text-level-semantics.html#the-ruby-element
1926                         'ruby'       => $common,
1927                         # rbc
1928                         'rb'         => $common,
1929                         'rp'         => $common,
1930                         'rt'         => $common, # array_merge( $common, array( 'rbspan' ) ),
1931                         'rtc'        => $common,
1932
1933                         # MathML root element, where used for extensions
1934                         # 'title' may not be 100% valid here; it's XHTML
1935                         # https://www.w3.org/TR/REC-MathML/
1936                         'math'       => [ 'class', 'style', 'id', 'title' ],
1937
1938                         // HTML 5 section 4.5
1939                         'figure'     => $common,
1940                         'figcaption' => $common,
1941
1942                         # HTML 5 section 4.6
1943                         'bdi' => $common,
1944
1945                         # HTML5 elements, defined by:
1946                         # https://html.spec.whatwg.org/multipage/semantics.html#the-data-element
1947                         'data' => array_merge( $common, [ 'value' ] ),
1948                         'time' => array_merge( $common, [ 'datetime' ] ),
1949                         'mark' => $common,
1950
1951                         // meta and link are only permitted by removeHTMLtags when Microdata
1952                         // is enabled so we don't bother adding a conditional to hide these
1953                         // Also meta and link are only valid in WikiText as Microdata elements
1954                         // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1955                         // So we don't bother including $common attributes that have no purpose.
1956                         'meta' => [ 'itemprop', 'content' ],
1957                         'link' => [ 'itemprop', 'href', 'title' ],
1958                 ];
1959
1960                 return $whitelist;
1961         }
1962
1963         /**
1964          * Take a fragment of (potentially invalid) HTML and return
1965          * a version with any tags removed, encoded as plain text.
1966          *
1967          * Warning: this return value must be further escaped for literal
1968          * inclusion in HTML output as of 1.10!
1969          *
1970          * @param string $html HTML fragment
1971          * @return string
1972          */
1973         static function stripAllTags( $html ) {
1974                 // Use RemexHtml to tokenize $html and extract the text
1975                 $handler = new RemexStripTagHandler;
1976                 $tokenizer = new RemexHtml\Tokenizer\Tokenizer( $handler, $html, [
1977                         'ignoreErrors' => true,
1978                         // don't ignore char refs, we want them to be decoded
1979                         'ignoreNulls' => true,
1980                         'skipPreprocess' => true,
1981                 ] );
1982                 $tokenizer->execute();
1983                 $text = $handler->getResult();
1984
1985                 $text = self::normalizeWhitespace( $text );
1986                 return $text;
1987         }
1988
1989         /**
1990          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1991          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1992          * PHP 5.1 doesn't.
1993          *
1994          * Use for passing XHTML fragments to PHP's XML parsing functions
1995          *
1996          * @return string
1997          */
1998         static function hackDocType() {
1999                 $out = "<!DOCTYPE html [\n";
2000                 foreach ( self::$htmlEntities as $entity => $codepoint ) {
2001                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
2002                 }
2003                 $out .= "]>\n";
2004                 return $out;
2005         }
2006
2007         /**
2008          * @param string $url
2009          * @return mixed|string
2010          */
2011         static function cleanUrl( $url ) {
2012                 # Normalize any HTML entities in input. They will be
2013                 # re-escaped by makeExternalLink().
2014                 $url = self::decodeCharReferences( $url );
2015
2016                 # Escape any control characters introduced by the above step
2017                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
2018                         [ __CLASS__, 'cleanUrlCallback' ], $url );
2019
2020                 # Validate hostname portion
2021                 $matches = [];
2022                 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
2023                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
2024
2025                         // Characters that will be ignored in IDNs.
2026                         // https://tools.ietf.org/html/rfc3454#section-3.1
2027                         // Strip them before further processing so blacklists and such work.
2028                         $strip = "/
2029                                 \\s|          # general whitespace
2030                                 \xc2\xad|     # 00ad SOFT HYPHEN
2031                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
2032                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
2033                                 \xe2\x81\xa0| # 2060 WORD JOINER
2034                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
2035                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
2036                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
2037                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
2038                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
2039                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
2040                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
2041                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
2042                                 /xuD";
2043
2044                         $host = preg_replace( $strip, '', $host );
2045
2046                         // IPv6 host names are bracketed with [].  Url-decode these.
2047                         if ( substr_compare( "//%5B", $host, 0, 5 ) === 0 &&
2048                                 preg_match( '!^//%5B([0-9A-Fa-f:.]+)%5D((:\d+)?)$!', $host, $matches )
2049                         ) {
2050                                 $host = '//[' . $matches[1] . ']' . $matches[2];
2051                         }
2052
2053                         // @todo FIXME: Validate hostnames here
2054
2055                         return $protocol . $host . $rest;
2056                 } else {
2057                         return $url;
2058                 }
2059         }
2060
2061         /**
2062          * @param array $matches
2063          * @return string
2064          */
2065         static function cleanUrlCallback( $matches ) {
2066                 return urlencode( $matches[0] );
2067         }
2068
2069         /**
2070          * Does a string look like an e-mail address?
2071          *
2072          * This validates an email address using an HTML5 specification found at:
2073          * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
2074          * Which as of 2011-01-24 says:
2075          *
2076          *   A valid e-mail address is a string that matches the ABNF production
2077          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
2078          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
2079          *   3.5.
2080          *
2081          * This function is an implementation of the specification as requested in
2082          * T24449.
2083          *
2084          * Client-side forms will use the same standard validation rules via JS or
2085          * HTML 5 validation; additional restrictions can be enforced server-side
2086          * by extensions via the 'isValidEmailAddr' hook.
2087          *
2088          * Note that this validation doesn't 100% match RFC 2822, but is believed
2089          * to be liberal enough for wide use. Some invalid addresses will still
2090          * pass validation here.
2091          *
2092          * @since 1.18
2093          *
2094          * @param string $addr E-mail address
2095          * @return bool
2096          */
2097         public static function validateEmail( $addr ) {
2098                 $result = null;
2099                 if ( !Hooks::run( 'isValidEmailAddr', [ $addr, &$result ] ) ) {
2100                         return $result;
2101                 }
2102
2103                 // Please note strings below are enclosed in brackets [], this make the
2104                 // hyphen "-" a range indicator. Hence it is double backslashed below.
2105                 // See T28948
2106                 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
2107                 $rfc1034_ldh_str = "a-z0-9\\-";
2108
2109                 $html5_email_regexp = "/
2110                 ^                      # start of string
2111                 [$rfc5322_atext\\.]+    # user part which is liberal :p
2112                 @                      # 'apostrophe'
2113                 [$rfc1034_ldh_str]+       # First domain part
2114                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
2115                 $                      # End of string
2116                 /ix"; // case Insensitive, eXtended
2117
2118                 return (bool)preg_match( $html5_email_regexp, $addr );
2119         }
2120 }