includes/Sanitizer.php

   1 <?php
   2 /**
   3  * HTML sanitizer for %MediaWiki.
   4  *
   5  * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
   6  * https://www.mediawiki.org/
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License as published by
  10  * the Free Software Foundation; either version 2 of the License, or
  11  * (at your option) any later version.
  12  *
  13  * This program is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  16  * GNU General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU General Public License along
  19  * with this program; if not, write to the Free Software Foundation, Inc.,
  20  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  21  * http://www.gnu.org/copyleft/gpl.html
  22  *
  23  * @file
  24  * @ingroup Parser
  25  */
  26
  27 /**
  28  * HTML sanitizer for MediaWiki
  29  * @ingroup Parser
  30  */
  31 class Sanitizer {
  32         /**
  33          * Regular expression to match various types of character references in
  34          * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
  35          */
  36         const CHAR_REFS_REGEX =
  37                 '/&([A-Za-z0-9\x80-\xff]+);
  38                  |&\#([0-9]+);
  39                  |&\#[xX]([0-9A-Fa-f]+);
  40                  |(&)/x';
  41
  42         /**
  43          * Acceptable tag name charset from HTML5 parsing spec
  44          * https://www.w3.org/TR/html5/syntax.html#tag-open-state
  45          */
  46         const ELEMENT_BITS_REGEX = '!^(/?)([A-Za-z][^\t\n\v />\0]*+)([^>]*?)(/?>)([^<]*)$!';
  47
  48         /**
  49          * Blacklist for evil uris like javascript:
  50          * WARNING: DO NOT use this in any place that actually requires blacklisting
  51          * for security reasons. There are NUMEROUS[1] ways to bypass blacklisting, the
  52          * only way to be secure from javascript: uri based xss vectors is to whitelist
  53          * things that you know are safe and deny everything else.
  54          * [1]: http://ha.ckers.org/xss.html
  55          */
  56         const EVIL_URI_PATTERN = '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
  57         const XMLNS_ATTRIBUTE_PATTERN = "/^xmlns:[:A-Z_a-z-.0-9]+$/";
  58
  59         /**
  60          * List of all named character entities defined in HTML 4.01
  61          * https://www.w3.org/TR/html4/sgml/entities.html
  62          * As well as &apos; which is only defined starting in XHTML1.
  63          */
  64         private static $htmlEntities = [
  65                 'Aacute'   => 193,
  66                 'aacute'   => 225,
  67                 'Acirc'    => 194,
  68                 'acirc'    => 226,
  69                 'acute'    => 180,
  70                 'AElig'    => 198,
  71                 'aelig'    => 230,
  72                 'Agrave'   => 192,
  73                 'agrave'   => 224,
  74                 'alefsym'  => 8501,
  75                 'Alpha'    => 913,
  76                 'alpha'    => 945,
  77                 'amp'      => 38,
  78                 'and'      => 8743,
  79                 'ang'      => 8736,
  80                 'apos'     => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
  81                 'Aring'    => 197,
  82                 'aring'    => 229,
  83                 'asymp'    => 8776,
  84                 'Atilde'   => 195,
  85                 'atilde'   => 227,
  86                 'Auml'     => 196,
  87                 'auml'     => 228,
  88                 'bdquo'    => 8222,
  89                 'Beta'     => 914,
  90                 'beta'     => 946,
  91                 'brvbar'   => 166,
  92                 'bull'     => 8226,
  93                 'cap'      => 8745,
  94                 'Ccedil'   => 199,
  95                 'ccedil'   => 231,
  96                 'cedil'    => 184,
  97                 'cent'     => 162,
  98                 'Chi'      => 935,
  99                 'chi'      => 967,
 100                 'circ'     => 710,
 101                 'clubs'    => 9827,
 102                 'cong'     => 8773,
 103                 'copy'     => 169,
 104                 'crarr'    => 8629,
 105                 'cup'      => 8746,
 106                 'curren'   => 164,
 107                 'dagger'   => 8224,
 108                 'Dagger'   => 8225,
 109                 'darr'     => 8595,
 110                 'dArr'     => 8659,
 111                 'deg'      => 176,
 112                 'Delta'    => 916,
 113                 'delta'    => 948,
 114                 'diams'    => 9830,
 115                 'divide'   => 247,
 116                 'Eacute'   => 201,
 117                 'eacute'   => 233,
 118                 'Ecirc'    => 202,
 119                 'ecirc'    => 234,
 120                 'Egrave'   => 200,
 121                 'egrave'   => 232,
 122                 'empty'    => 8709,
 123                 'emsp'     => 8195,
 124                 'ensp'     => 8194,
 125                 'Epsilon'  => 917,
 126                 'epsilon'  => 949,
 127                 'equiv'    => 8801,
 128                 'Eta'      => 919,
 129                 'eta'      => 951,
 130                 'ETH'      => 208,
 131                 'eth'      => 240,
 132                 'Euml'     => 203,
 133                 'euml'     => 235,
 134                 'euro'     => 8364,
 135                 'exist'    => 8707,
 136                 'fnof'     => 402,
 137                 'forall'   => 8704,
 138                 'frac12'   => 189,
 139                 'frac14'   => 188,
 140                 'frac34'   => 190,
 141                 'frasl'    => 8260,
 142                 'Gamma'    => 915,
 143                 'gamma'    => 947,
 144                 'ge'       => 8805,
 145                 'gt'       => 62,
 146                 'harr'     => 8596,
 147                 'hArr'     => 8660,
 148                 'hearts'   => 9829,
 149                 'hellip'   => 8230,
 150                 'Iacute'   => 205,
 151                 'iacute'   => 237,
 152                 'Icirc'    => 206,
 153                 'icirc'    => 238,
 154                 'iexcl'    => 161,
 155                 'Igrave'   => 204,
 156                 'igrave'   => 236,
 157                 'image'    => 8465,
 158                 'infin'    => 8734,
 159                 'int'      => 8747,
 160                 'Iota'     => 921,
 161                 'iota'     => 953,
 162                 'iquest'   => 191,
 163                 'isin'     => 8712,
 164                 'Iuml'     => 207,
 165                 'iuml'     => 239,
 166                 'Kappa'    => 922,
 167                 'kappa'    => 954,
 168                 'Lambda'   => 923,
 169                 'lambda'   => 955,
 170                 'lang'     => 9001,
 171                 'laquo'    => 171,
 172                 'larr'     => 8592,
 173                 'lArr'     => 8656,
 174                 'lceil'    => 8968,
 175                 'ldquo'    => 8220,
 176                 'le'       => 8804,
 177                 'lfloor'   => 8970,
 178                 'lowast'   => 8727,
 179                 'loz'      => 9674,
 180                 'lrm'      => 8206,
 181                 'lsaquo'   => 8249,
 182                 'lsquo'    => 8216,
 183                 'lt'       => 60,
 184                 'macr'     => 175,
 185                 'mdash'    => 8212,
 186                 'micro'    => 181,
 187                 'middot'   => 183,
 188                 'minus'    => 8722,
 189                 'Mu'       => 924,
 190                 'mu'       => 956,
 191                 'nabla'    => 8711,
 192                 'nbsp'     => 160,
 193                 'ndash'    => 8211,
 194                 'ne'       => 8800,
 195                 'ni'       => 8715,
 196                 'not'      => 172,
 197                 'notin'    => 8713,
 198                 'nsub'     => 8836,
 199                 'Ntilde'   => 209,
 200                 'ntilde'   => 241,
 201                 'Nu'       => 925,
 202                 'nu'       => 957,
 203                 'Oacute'   => 211,
 204                 'oacute'   => 243,
 205                 'Ocirc'    => 212,
 206                 'ocirc'    => 244,
 207                 'OElig'    => 338,
 208                 'oelig'    => 339,
 209                 'Ograve'   => 210,
 210                 'ograve'   => 242,
 211                 'oline'    => 8254,
 212                 'Omega'    => 937,
 213                 'omega'    => 969,
 214                 'Omicron'  => 927,
 215                 'omicron'  => 959,
 216                 'oplus'    => 8853,
 217                 'or'       => 8744,
 218                 'ordf'     => 170,
 219                 'ordm'     => 186,
 220                 'Oslash'   => 216,
 221                 'oslash'   => 248,
 222                 'Otilde'   => 213,
 223                 'otilde'   => 245,
 224                 'otimes'   => 8855,
 225                 'Ouml'     => 214,
 226                 'ouml'     => 246,
 227                 'para'     => 182,
 228                 'part'     => 8706,
 229                 'permil'   => 8240,
 230                 'perp'     => 8869,
 231                 'Phi'      => 934,
 232                 'phi'      => 966,
 233                 'Pi'       => 928,
 234                 'pi'       => 960,
 235                 'piv'      => 982,
 236                 'plusmn'   => 177,
 237                 'pound'    => 163,
 238                 'prime'    => 8242,
 239                 'Prime'    => 8243,
 240                 'prod'     => 8719,
 241                 'prop'     => 8733,
 242                 'Psi'      => 936,
 243                 'psi'      => 968,
 244                 'quot'     => 34,
 245                 'radic'    => 8730,
 246                 'rang'     => 9002,
 247                 'raquo'    => 187,
 248                 'rarr'     => 8594,
 249                 'rArr'     => 8658,
 250                 'rceil'    => 8969,
 251                 'rdquo'    => 8221,
 252                 'real'     => 8476,
 253                 'reg'      => 174,
 254                 'rfloor'   => 8971,
 255                 'Rho'      => 929,
 256                 'rho'      => 961,
 257                 'rlm'      => 8207,
 258                 'rsaquo'   => 8250,
 259                 'rsquo'    => 8217,
 260                 'sbquo'    => 8218,
 261                 'Scaron'   => 352,
 262                 'scaron'   => 353,
 263                 'sdot'     => 8901,
 264                 'sect'     => 167,
 265                 'shy'      => 173,
 266                 'Sigma'    => 931,
 267                 'sigma'    => 963,
 268                 'sigmaf'   => 962,
 269                 'sim'      => 8764,
 270                 'spades'   => 9824,
 271                 'sub'      => 8834,
 272                 'sube'     => 8838,
 273                 'sum'      => 8721,
 274                 'sup'      => 8835,
 275                 'sup1'     => 185,
 276                 'sup2'     => 178,
 277                 'sup3'     => 179,
 278                 'supe'     => 8839,
 279                 'szlig'    => 223,
 280                 'Tau'      => 932,
 281                 'tau'      => 964,
 282                 'there4'   => 8756,
 283                 'Theta'    => 920,
 284                 'theta'    => 952,
 285                 'thetasym' => 977,
 286                 'thinsp'   => 8201,
 287                 'THORN'    => 222,
 288                 'thorn'    => 254,
 289                 'tilde'    => 732,
 290                 'times'    => 215,
 291                 'trade'    => 8482,
 292                 'Uacute'   => 218,
 293                 'uacute'   => 250,
 294                 'uarr'     => 8593,
 295                 'uArr'     => 8657,
 296                 'Ucirc'    => 219,
 297                 'ucirc'    => 251,
 298                 'Ugrave'   => 217,
 299                 'ugrave'   => 249,
 300                 'uml'      => 168,
 301                 'upsih'    => 978,
 302                 'Upsilon'  => 933,
 303                 'upsilon'  => 965,
 304                 'Uuml'     => 220,
 305                 'uuml'     => 252,
 306                 'weierp'   => 8472,
 307                 'Xi'       => 926,
 308                 'xi'       => 958,
 309                 'Yacute'   => 221,
 310                 'yacute'   => 253,
 311                 'yen'      => 165,
 312                 'Yuml'     => 376,
 313                 'yuml'     => 255,
 314                 'Zeta'     => 918,
 315                 'zeta'     => 950,
 316                 'zwj'      => 8205,
 317                 'zwnj'     => 8204
 318         ];
 319
 320         /**
 321          * Character entity aliases accepted by MediaWiki
 322          */
 323         private static $htmlEntityAliases = [
 324                 'רלמ' => 'rlm',
 325                 'رلم' => 'rlm',
 326         ];
 327
 328         /**
 329          * Lazy-initialised attributes regex, see getAttribsRegex()
 330          */
 331         private static $attribsRegex;
 332
 333         /**
 334          * Regular expression to match HTML/XML attribute pairs within a tag.
 335          * Allows some... latitude. Based on,
 336          * https://www.w3.org/TR/html5/syntax.html#before-attribute-value-state
 337          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 338          * @return string
 339          */
 340         static function getAttribsRegex() {
 341                 if ( self::$attribsRegex === null ) {
 342                         $attribFirst = '[:A-Z_a-z0-9]';
 343                         $attrib = '[:A-Z_a-z-.0-9]';
 344                         $space = '[\x09\x0a\x0c\x0d\x20]';
 345                         self::$attribsRegex =
 346                                 "/(?:^|$space)({$attribFirst}{$attrib}*)
 347                                         ($space*=$space*
 348                                         (?:
 349                                                 # The attribute value: quoted or alone
 350                                                 \"([^\"]*)(?:\"|\$)
 351                                                 | '([^']*)(?:'|\$)
 352                                                 | (((?!$space|>).)*)
 353                                         )
 354                                 )?(?=$space|\$)/sx";
 355                 }
 356                 return self::$attribsRegex;
 357         }
 358
 359         /**
 360          * Return the various lists of recognized tags
 361          * @param array $extratags For any extra tags to include
 362          * @param array $removetags For any tags (default or extra) to exclude
 363          * @return array
 364          */
 365         public static function getRecognizedTagData( $extratags = [], $removetags = [] ) {
 366                 global $wgAllowImageTag;
 367
 368                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 369                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 370
 371                 // Base our staticInitialised variable off of the global config state so that if the globals
 372                 // are changed (like in the screwed up test system) we will re-initialise the settings.
 373                 $globalContext = $wgAllowImageTag;
 374                 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
 375                         $htmlpairsStatic = [ # Tags that must be closed
 376                                 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 377                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 378                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 379                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 380                                 'ruby', 'rb', 'rp', 'rt', 'rtc', 'p', 'span', 'abbr', 'dfn',
 381                                 'kbd', 'samp', 'data', 'time', 'mark'
 382                         ];
 383                         $htmlsingle = [
 384                                 'br', 'wbr', 'hr', 'li', 'dt', 'dd', 'meta', 'link'
 385                         ];
 386
 387                         # Elements that cannot have close tags. This is (not coincidentally)
 388                         # also the list of tags for which the HTML 5 parsing algorithm
 389                         # requires you to "acknowledge the token's self-closing flag", i.e.
 390                         # a self-closing tag like <br/> is not an HTML 5 parse error only
 391                         # for this list.
 392                         $htmlsingleonly = [
 393                                 'br', 'wbr', 'hr', 'meta', 'link'
 394                         ];
 395
 396                         $htmlnest = [ # Tags that can be nested--??
 397                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 398                                 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
 399                                 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
 400                         ];
 401                         $tabletags = [ # Can only appear inside table, we will close them
 402                                 'td', 'th', 'tr',
 403                         ];
 404                         $htmllist = [ # Tags used by list
 405                                 'ul', 'ol',
 406                         ];
 407                         $listtags = [ # Tags that can appear in a list
 408                                 'li',
 409                         ];
 410
 411                         if ( $wgAllowImageTag ) {
 412                                 $htmlsingle[] = 'img';
 413                                 $htmlsingleonly[] = 'img';
 414                         }
 415
 416                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 417                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 418
 419                         # Convert them all to hashtables for faster lookup
 420                         $vars = [ 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 421                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' ];
 422                         foreach ( $vars as $var ) {
 423                                 $$var = array_flip( $$var );
 424                         }
 425                         $staticInitialised = $globalContext;
 426                 }
 427
 428                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 429                 $extratags = array_flip( $extratags );
 430                 $removetags = array_flip( $removetags );
 431                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 432                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
 433
 434                 return [
 435                         'htmlpairs' => $htmlpairs,
 436                         'htmlsingle' => $htmlsingle,
 437                         'htmlsingleonly' => $htmlsingleonly,
 438                         'htmlnest' => $htmlnest,
 439                         'tabletags' => $tabletags,
 440                         'htmllist' => $htmllist,
 441                         'listtags' => $listtags,
 442                         'htmlsingleallowed' => $htmlsingleallowed,
 443                         'htmlelements' => $htmlelements,
 444                 ];
 445         }
 446
 447         /**
 448          * Cleans up HTML, removes dangerous tags and attributes, and
 449          * removes HTML comments
 450          * @param string $text
 451          * @param callable $processCallback Callback to do any variable or parameter
 452          *   replacements in HTML attribute values
 453          * @param array|bool $args Arguments for the processing callback
 454          * @param array $extratags For any extra tags to include
 455          * @param array $removetags For any tags (default or extra) to exclude
 456          * @param callable $warnCallback (Deprecated) Callback allowing the
 457          *   addition of a tracking category when bad input is encountered.
 458          *   DO NOT ADD NEW PARAMETERS AFTER $warnCallback, since it will be
 459          *   removed shortly.
 460          * @return string
 461          */
 462         public static function removeHTMLtags( $text, $processCallback = null,
 463                 $args = [], $extratags = [], $removetags = [], $warnCallback = null
 464         ) {
 465                 extract( self::getRecognizedTagData( $extratags, $removetags ) );
 466
 467                 # Remove HTML comments
 468                 $text = Sanitizer::removeHTMLcomments( $text );
 469                 $bits = explode( '<', $text );
 470                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 471                 if ( !MWTidy::isEnabled() ) {
 472                         $tagstack = $tablestack = [];
 473                         foreach ( $bits as $x ) {
 474                                 $regs = [];
 475                                 # $slash: Does the current element start with a '/'?
 476                                 # $t: Current element name
 477                                 # $params: String between element name and >
 478                                 # $brace: Ending '>' or '/>'
 479                                 # $rest: Everything until the next element of $bits
 480                                 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 481                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 482                                 } else {
 483                                         $slash = $t = $params = $brace = $rest = null;
 484                                 }
 485
 486                                 $badtag = false;
 487                                 $t = strtolower( $t );
 488                                 if ( isset( $htmlelements[$t] ) ) {
 489                                         # Check our stack
 490                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 491                                                 $badtag = true;
 492                                         } elseif ( $slash ) {
 493                                                 # Closing a tag... is it the one we just opened?
 494                                                 MediaWiki\suppressWarnings();
 495                                                 $ot = array_pop( $tagstack );
 496                                                 MediaWiki\restoreWarnings();
 497
 498                                                 if ( $ot != $t ) {
 499                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 500                                                                 # Pop all elements with an optional close tag
 501                                                                 # and see if we find a match below them
 502                                                                 $optstack = [];
 503                                                                 array_push( $optstack, $ot );
 504                                                                 MediaWiki\suppressWarnings();
 505                                                                 $ot = array_pop( $tagstack );
 506                                                                 MediaWiki\restoreWarnings();
 507                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 508                                                                         array_push( $optstack, $ot );
 509                                                                         MediaWiki\suppressWarnings();
 510                                                                         $ot = array_pop( $tagstack );
 511                                                                         MediaWiki\restoreWarnings();
 512                                                                 }
 513                                                                 if ( $t != $ot ) {
 514                                                                         # No match. Push the optional elements back again
 515                                                                         $badtag = true;
 516                                                                         MediaWiki\suppressWarnings();
 517                                                                         $ot = array_pop( $optstack );
 518                                                                         MediaWiki\restoreWarnings();
 519                                                                         while ( $ot ) {
 520                                                                                 array_push( $tagstack, $ot );
 521                                                                                 MediaWiki\suppressWarnings();
 522                                                                                 $ot = array_pop( $optstack );
 523                                                                                 MediaWiki\restoreWarnings();
 524                                                                         }
 525                                                                 }
 526                                                         } else {
 527                                                                 MediaWiki\suppressWarnings();
 528                                                                 array_push( $tagstack, $ot );
 529                                                                 MediaWiki\restoreWarnings();
 530
 531                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 532                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 533                                                                         $badtag = true;
 534                                                                 }
 535                                                         }
 536                                                 } else {
 537                                                         if ( $t == 'table' ) {
 538                                                                 $tagstack = array_pop( $tablestack );
 539                                                         }
 540                                                 }
 541                                                 $newparams = '';
 542                                         } else {
 543                                                 # Keep track for later
 544                                                 if ( isset( $tabletags[$t] ) && !in_array( 'table', $tagstack ) ) {
 545                                                         $badtag = true;
 546                                                 } elseif ( in_array( $t, $tagstack ) && !isset( $htmlnest[$t] ) ) {
 547                                                         $badtag = true;
 548                                                 #  Is it a self closed htmlpair ? (T7487)
 549                                                 } elseif ( $brace == '/>' && isset( $htmlpairs[$t] ) ) {
 550                                                         // Eventually we'll just remove the self-closing
 551                                                         // slash, in order to be consistent with HTML5
 552                                                         // semantics.
 553                                                         // $brace = '>';
 554                                                         // For now, let's just warn authors to clean up.
 555                                                         if ( is_callable( $warnCallback ) ) {
 556                                                                 call_user_func_array( $warnCallback, [ 'deprecated-self-close-category' ] );
 557                                                         }
 558                                                         $badtag = true;
 559                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 560                                                         # Hack to force empty tag for unclosable elements
 561                                                         $brace = '/>';
 562                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 563                                                         # Hack to not close $htmlsingle tags
 564                                                         $brace = null;
 565                                                         # Still need to push this optionally-closed tag to
 566                                                         # the tag stack so that we can match end tags
 567                                                         # instead of marking them as bad.
 568                                                         array_push( $tagstack, $t );
 569                                                 } elseif ( isset( $tabletags[$t] ) && in_array( $t, $tagstack ) ) {
 570                                                         // New table tag but forgot to close the previous one
 571                                                         $text .= "</$t>";
 572                                                 } else {
 573                                                         if ( $t == 'table' ) {
 574                                                                 array_push( $tablestack, $tagstack );
 575                                                                 $tagstack = [];
 576                                                         }
 577                                                         array_push( $tagstack, $t );
 578                                                 }
 579
 580                                                 # Replace any variables or template parameters with
 581                                                 # plaintext results.
 582                                                 if ( is_callable( $processCallback ) ) {
 583                                                         call_user_func_array( $processCallback, [ &$params, $args ] );
 584                                                 }
 585
 586                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 587                                                         $badtag = true;
 588                                                 }
 589
 590                                                 # Strip non-approved attributes from the tag
 591                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 592                                         }
 593                                         if ( !$badtag ) {
 594                                                 $rest = str_replace( '>', '&gt;', $rest );
 595                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 596                                                 $text .= "<$slash$t$newparams$close>$rest";
 597                                                 continue;
 598                                         }
 599                                 }
 600                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 601                         }
 602                         # Close off any remaining tags
 603                         while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
 604                                 $text .= "</$t>\n";
 605                                 if ( $t == 'table' ) {
 606                                         $tagstack = array_pop( $tablestack );
 607                                 }
 608                         }
 609                 } else {
 610                         # this might be possible using tidy itself
 611                         foreach ( $bits as $x ) {
 612                                 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 613                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 614
 615                                         $badtag = false;
 616                                         $t = strtolower( $t );
 617                                         if ( isset( $htmlelements[$t] ) ) {
 618                                                 if ( is_callable( $processCallback ) ) {
 619                                                         call_user_func_array( $processCallback, [ &$params, $args ] );
 620                                                 }
 621
 622                                                 if ( $brace == '/>' && !( isset( $htmlsingle[$t] ) || isset( $htmlsingleonly[$t] ) ) ) {
 623                                                         // Eventually we'll just remove the self-closing
 624                                                         // slash, in order to be consistent with HTML5
 625                                                         // semantics.
 626                                                         // $brace = '>';
 627                                                         // For now, let's just warn authors to clean up.
 628                                                         if ( is_callable( $warnCallback ) ) {
 629                                                                 call_user_func_array( $warnCallback, [ 'deprecated-self-close-category' ] );
 630                                                         }
 631                                                 }
 632                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 633                                                         $badtag = true;
 634                                                 }
 635
 636                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 637                                                 if ( !$badtag ) {
 638                                                         if ( $brace === '/>' && !isset( $htmlsingleonly[$t] ) ) {
 639                                                                 # Interpret self-closing tags as empty tags even when
 640                                                                 # HTML 5 would interpret them as start tags. Such input
 641                                                                 # is commonly seen on Wikimedia wikis with this intention.
 642                                                                 $brace = "></$t>";
 643                                                         }
 644
 645                                                         $rest = str_replace( '>', '&gt;', $rest );
 646                                                         $text .= "<$slash$t$newparams$brace$rest";
 647                                                         continue;
 648                                                 }
 649                                         }
 650                                 }
 651                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 652                         }
 653                 }
 654                 return $text;
 655         }
 656
 657         /**
 658          * Remove '<!--', '-->', and everything between.
 659          * To avoid leaving blank lines, when a comment is both preceded
 660          * and followed by a newline (ignoring spaces), trim leading and
 661          * trailing spaces and one of the newlines.
 662          *
 663          * @param string $text
 664          * @return string
 665          */
 666         public static function removeHTMLcomments( $text ) {
 667                 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
 668                         $end = strpos( $text, '-->', $start + 4 );
 669                         if ( $end === false ) {
 670                                 # Unterminated comment; bail out
 671                                 break;
 672                         }
 673
 674                         $end += 3;
 675
 676                         # Trim space and newline if the comment is both
 677                         # preceded and followed by a newline
 678                         $spaceStart = max( $start - 1, 0 );
 679                         $spaceLen = $end - $spaceStart;
 680                         while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
 681                                 $spaceStart--;
 682                                 $spaceLen++;
 683                         }
 684                         while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
 685                                 $spaceLen++;
 686                         }
 687                         if ( substr( $text, $spaceStart, 1 ) === "\n"
 688                                 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
 689                                 # Remove the comment, leading and trailing
 690                                 # spaces, and leave only one newline.
 691                                 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
 692                         } else {
 693                                 # Remove just the comment.
 694                                 $text = substr_replace( $text, '', $start, $end - $start );
 695                         }
 696                 }
 697                 return $text;
 698         }
 699
 700         /**
 701          * Takes attribute names and values for a tag and the tag name and
 702          * validates that the tag is allowed to be present.
 703          * This DOES NOT validate the attributes, nor does it validate the
 704          * tags themselves. This method only handles the special circumstances
 705          * where we may want to allow a tag within content but ONLY when it has
 706          * specific attributes set.
 707          *
 708          * @param string $params
 709          * @param string $element
 710          * @return bool
 711          */
 712         static function validateTag( $params, $element ) {
 713                 $params = Sanitizer::decodeTagAttributes( $params );
 714
 715                 if ( $element == 'meta' || $element == 'link' ) {
 716                         if ( !isset( $params['itemprop'] ) ) {
 717                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 718                                 return false;
 719                         }
 720                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 721                                 // <meta> must have a content="" for the itemprop
 722                                 return false;
 723                         }
 724                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 725                                 // <link> must have an associated href=""
 726                                 return false;
 727                         }
 728                 }
 729
 730                 return true;
 731         }
 732
 733         /**
 734          * Take an array of attribute names and values and normalize or discard
 735          * illegal values for the given element type.
 736          *
 737          * - Discards attributes not on a whitelist for the given element
 738          * - Unsafe style attributes are discarded
 739          * - Invalid id attributes are re-encoded
 740          *
 741          * @param array $attribs
 742          * @param string $element
 743          * @return array
 744          *
 745          * @todo Check for legal values where the DTD limits things.
 746          * @todo Check for unique id attribute :P
 747          */
 748         static function validateTagAttributes( $attribs, $element ) {
 749                 return Sanitizer::validateAttributes( $attribs,
 750                         Sanitizer::attributeWhitelist( $element ) );
 751         }
 752
 753         /**
 754          * Take an array of attribute names and values and normalize or discard
 755          * illegal values for the given whitelist.
 756          *
 757          * - Discards attributes not on the given whitelist
 758          * - Unsafe style attributes are discarded
 759          * - Invalid id attributes are re-encoded
 760          *
 761          * @param array $attribs
 762          * @param array $whitelist List of allowed attribute names
 763          * @return array
 764          *
 765          * @todo Check for legal values where the DTD limits things.
 766          * @todo Check for unique id attribute :P
 767          */
 768         static function validateAttributes( $attribs, $whitelist ) {
 769                 $whitelist = array_flip( $whitelist );
 770                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 771
 772                 $out = [];
 773                 foreach ( $attribs as $attribute => $value ) {
 774                         # Allow XML namespace declaration to allow RDFa
 775                         if ( preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 776                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 777                                         $out[$attribute] = $value;
 778                                 }
 779
 780                                 continue;
 781                         }
 782
 783                         # Allow any attribute beginning with "data-"
 784                         # However:
 785                         # * Disallow data attributes used by MediaWiki code
 786                         # * Ensure that the attribute is not namespaced by banning
 787                         #   colons.
 788                         if ( !preg_match( '/^data-[^:]*$/i', $attribute )
 789                                 && !isset( $whitelist[$attribute] )
 790                                 || self::isReservedDataAttribute( $attribute )
 791                         ) {
 792                                 continue;
 793                         }
 794
 795                         # Strip javascript "expression" from stylesheets.
 796                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 797                         if ( $attribute == 'style' ) {
 798                                 $value = Sanitizer::checkCss( $value );
 799                         }
 800
 801                         # Escape HTML id attributes
 802                         if ( $attribute === 'id' ) {
 803                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 804                         }
 805
 806                         # Escape HTML id reference lists
 807                         if ( $attribute === 'aria-describedby'
 808                                 || $attribute === 'aria-flowto'
 809                                 || $attribute === 'aria-labelledby'
 810                                 || $attribute === 'aria-owns'
 811                         ) {
 812                                 $value = Sanitizer::escapeIdReferenceList( $value, 'noninitial' );
 813                         }
 814
 815                         // RDFa and microdata properties allow URLs, URIs and/or CURIs.
 816                         // Check them for sanity.
 817                         if ( $attribute === 'rel' || $attribute === 'rev'
 818                                 # RDFa
 819                                 || $attribute === 'about' || $attribute === 'property'
 820                                 || $attribute === 'resource' || $attribute === 'datatype'
 821                                 || $attribute === 'typeof'
 822                                 # HTML5 microdata
 823                                 || $attribute === 'itemid' || $attribute === 'itemprop'
 824                                 || $attribute === 'itemref' || $attribute === 'itemscope'
 825                                 || $attribute === 'itemtype'
 826                         ) {
 827                                 // Paranoia. Allow "simple" values but suppress javascript
 828                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 829                                         continue;
 830                                 }
 831                         }
 832
 833                         # NOTE: even though elements using href/src are not allowed directly, supply
 834                         #       validation code that can be used by tag hook handlers, etc
 835                         if ( $attribute === 'href' || $attribute === 'src' || $attribute === 'poster' ) {
 836                                 if ( !preg_match( $hrefExp, $value ) ) {
 837                                         continue; // drop any href or src attributes not using an allowed protocol.
 838                                         // NOTE: this also drops all relative URLs
 839                                 }
 840                         }
 841
 842                         // If this attribute was previously set, override it.
 843                         // Output should only have one attribute of each name.
 844                         $out[$attribute] = $value;
 845                 }
 846
 847                 # itemtype, itemid, itemref don't make sense without itemscope
 848                 if ( !array_key_exists( 'itemscope', $out ) ) {
 849                         unset( $out['itemtype'] );
 850                         unset( $out['itemid'] );
 851                         unset( $out['itemref'] );
 852                 }
 853                 # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 854
 855                 return $out;
 856         }
 857
 858         /**
 859          * Given an attribute name, checks whether it is a reserved data attribute
 860          * (such as data-mw-foo) which is unavailable to user-generated HTML so MediaWiki
 861          * core and extension code can safely use it to communicate with frontend code.
 862          * @param string $attr Attribute name.
 863          * @return bool
 864          */
 865         public static function isReservedDataAttribute( $attr ) {
 866                 // data-ooui is reserved for ooui.
 867                 // data-mw and data-parsoid are reserved for parsoid.
 868                 // data-mw-<name here> is reserved for extensions (or core) if
 869                 // they need to communicate some data to the client and want to be
 870                 // sure that it isn't coming from an untrusted user.
 871                 // We ignore the possibility of namespaces since user-generated HTML
 872                 // can't use them anymore.
 873                 return (bool)preg_match( '/^data-(ooui|mw|parsoid)/i', $attr );
 874         }
 875
 876         /**
 877          * Merge two sets of HTML attributes.  Conflicting items in the second set
 878          * will override those in the first, except for 'class' attributes which
 879          * will be combined (if they're both strings).
 880          *
 881          * @todo implement merging for other attributes such as style
 882          * @param array $a
 883          * @param array $b
 884          * @return array
 885          */
 886         static function mergeAttributes( $a, $b ) {
 887                 $out = array_merge( $a, $b );
 888                 if ( isset( $a['class'] ) && isset( $b['class'] )
 889                         && is_string( $a['class'] ) && is_string( $b['class'] )
 890                         && $a['class'] !== $b['class']
 891                 ) {
 892                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 893                                 -1, PREG_SPLIT_NO_EMPTY );
 894                         $out['class'] = implode( ' ', array_unique( $classes ) );
 895                 }
 896                 return $out;
 897         }
 898
 899         /**
 900          * Normalize CSS into a format we can easily search for hostile input
 901          *  - decode character references
 902          *  - decode escape sequences
 903          *  - convert characters that IE6 interprets into ascii
 904          *  - remove comments, unless the entire value is one single comment
 905          * @param string $value the css string
 906          * @return string normalized css
 907          */
 908         public static function normalizeCss( $value ) {
 909
 910                 // Decode character references like &#123;
 911                 $value = Sanitizer::decodeCharReferences( $value );
 912
 913                 // Decode escape sequences and line continuation
 914                 // See the grammar in the CSS 2 spec, appendix D.
 915                 // This has to be done AFTER decoding character references.
 916                 // This means it isn't possible for this function to return
 917                 // unsanitized escape sequences. It is possible to manufacture
 918                 // input that contains character references that decode to
 919                 // escape sequences that decode to character references, but
 920                 // it's OK for the return value to contain character references
 921                 // because the caller is supposed to escape those anyway.
 922                 static $decodeRegex;
 923                 if ( !$decodeRegex ) {
 924                         $space = '[\\x20\\t\\r\\n\\f]';
 925                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 926                         $backslash = '\\\\';
 927                         $decodeRegex = "/ $backslash
 928                                 (?:
 929                                         ($nl) |  # 1. Line continuation
 930                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 931                                         (.) | # 3. backslash cancelling special meaning
 932                                         () | # 4. backslash at end of string
 933                                 )/xu";
 934                 }
 935                 $value = preg_replace_callback( $decodeRegex,
 936                         [ __CLASS__, 'cssDecodeCallback' ], $value );
 937
 938                 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
 939                 $value = preg_replace_callback(
 940                         '/[！-［］-ｚ]/u', // U+FF01 to U+FF5A, excluding U+FF3C (T60088)
 941                         function ( $matches ) {
 942                                 $cp = UtfNormal\Utils::utf8ToCodepoint( $matches[0] );
 943                                 if ( $cp === false ) {
 944                                         return '';
 945                                 }
 946                                 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
 947                         },
 948                         $value
 949                 );
 950
 951                 // Convert more characters IE6 might treat as ascii
 952                 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
 953                 $value = str_replace(
 954                         [ 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ],
 955                         [ 'r', 'n', 'n', 'l', 'i', '(', '(' ],
 956                         $value
 957                 );
 958
 959                 // Let the value through if it's nothing but a single comment, to
 960                 // allow other functions which may reject it to pass some error
 961                 // message through.
 962                 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
 963                         // Remove any comments; IE gets token splitting wrong
 964                         // This must be done AFTER decoding character references and
 965                         // escape sequences, because those steps can introduce comments
 966                         // This step cannot introduce character references or escape
 967                         // sequences, because it replaces comments with spaces rather
 968                         // than removing them completely.
 969                         $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 970
 971                         // Remove anything after a comment-start token, to guard against
 972                         // incorrect client implementations.
 973                         $commentPos = strpos( $value, '/*' );
 974                         if ( $commentPos !== false ) {
 975                                 $value = substr( $value, 0, $commentPos );
 976                         }
 977                 }
 978
 979                 // S followed by repeat, iteration, or prolonged sound marks,
 980                 // which IE will treat as "ss"
 981                 $value = preg_replace(
 982                         '/s(?:
 983                                 \xE3\x80\xB1 | # U+3031
 984                                 \xE3\x82\x9D | # U+309D
 985                                 \xE3\x83\xBC | # U+30FC
 986                                 \xE3\x83\xBD | # U+30FD
 987                                 \xEF\xB9\xBC | # U+FE7C
 988                                 \xEF\xB9\xBD | # U+FE7D
 989                                 \xEF\xBD\xB0   # U+FF70
 990                         )/ix',
 991                         'ss',
 992                         $value
 993                 );
 994
 995                 return $value;
 996         }
 997
 998         /**
 999          * Pick apart some CSS and check it for forbidden or unsafe structures.
1000          * Returns a sanitized string. This sanitized string will have
1001          * character references and escape sequences decoded and comments
1002          * stripped (unless it is itself one valid comment, in which case the value
1003          * will be passed through). If the input is just too evil, only a comment
1004          * complaining about evilness will be returned.
1005          *
1006          * Currently URL references, 'expression', 'tps' are forbidden.
1007          *
1008          * NOTE: Despite the fact that character references are decoded, the
1009          * returned string may contain character references given certain
1010          * clever input strings. These character references must
1011          * be escaped before the return value is embedded in HTML.
1012          *
1013          * @param string $value
1014          * @return string
1015          */
1016         static function checkCss( $value ) {
1017                 $value = self::normalizeCss( $value );
1018
1019                 // Reject problematic keywords and control characters
1020                 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ||
1021                         strpos( $value, UtfNormal\Constants::UTF8_REPLACEMENT ) !== false ) {
1022                         return '/* invalid control char */';
1023                 } elseif ( preg_match(
1024                         '! expression
1025                                 | filter\s*:
1026                                 | accelerator\s*:
1027                                 | -o-link\s*:
1028                                 | -o-link-source\s*:
1029                                 | -o-replace\s*:
1030                                 | url\s*\(
1031                                 | image\s*\(
1032                                 | image-set\s*\(
1033                                 | attr\s*\([^)]+[\s,]+url
1034                         !ix', $value ) ) {
1035                         return '/* insecure input */';
1036                 }
1037                 return $value;
1038         }
1039
1040         /**
1041          * @param array $matches
1042          * @return string
1043          */
1044         static function cssDecodeCallback( $matches ) {
1045                 if ( $matches[1] !== '' ) {
1046                         // Line continuation
1047                         return '';
1048                 } elseif ( $matches[2] !== '' ) {
1049                         $char = UtfNormal\Utils::codepointToUtf8( hexdec( $matches[2] ) );
1050                 } elseif ( $matches[3] !== '' ) {
1051                         $char = $matches[3];
1052                 } else {
1053                         $char = '\\';
1054                 }
1055                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
1056                         // These characters need to be escaped in strings
1057                         // Clean up the escape sequence to avoid parsing errors by clients
1058                         return '\\' . dechex( ord( $char ) ) . ' ';
1059                 } else {
1060                         // Decode unnecessary escape
1061                         return $char;
1062                 }
1063         }
1064
1065         /**
1066          * Take a tag soup fragment listing an HTML element's attributes
1067          * and normalize it to well-formed XML, discarding unwanted attributes.
1068          * Output is safe for further wikitext processing, with escaping of
1069          * values that could trigger problems.
1070          *
1071          * - Normalizes attribute names to lowercase
1072          * - Discards attributes not on a whitelist for the given element
1073          * - Turns broken or invalid entities into plaintext
1074          * - Double-quotes all attribute values
1075          * - Attributes without values are given the name as attribute
1076          * - Double attributes are discarded
1077          * - Unsafe style attributes are discarded
1078          * - Prepends space if there are attributes.
1079          * - (Optionally) Sorts attributes by name.
1080          *
1081          * @param string $text
1082          * @param string $element
1083          * @param bool $sorted Whether to sort the attributes (default: false)
1084          * @return string
1085          */
1086         static function fixTagAttributes( $text, $element, $sorted = false ) {
1087                 if ( trim( $text ) == '' ) {
1088                         return '';
1089                 }
1090
1091                 $decoded = Sanitizer::decodeTagAttributes( $text );
1092                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1093
1094                 if ( $sorted ) {
1095                         ksort( $stripped );
1096                 }
1097
1098                 return Sanitizer::safeEncodeTagAttributes( $stripped );
1099         }
1100
1101         /**
1102          * Encode an attribute value for HTML output.
1103          * @param string $text
1104          * @return string HTML-encoded text fragment
1105          */
1106         static function encodeAttribute( $text ) {
1107                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1108
1109                 // Whitespace is normalized during attribute decoding,
1110                 // so if we've been passed non-spaces we must encode them
1111                 // ahead of time or they won't be preserved.
1112                 $encValue = strtr( $encValue, [
1113                         "\n" => '&#10;',
1114                         "\r" => '&#13;',
1115                         "\t" => '&#9;',
1116                 ] );
1117
1118                 return $encValue;
1119         }
1120
1121         /**
1122          * Encode an attribute value for HTML tags, with extra armoring
1123          * against further wiki processing.
1124          * @param string $text
1125          * @return string HTML-encoded text fragment
1126          */
1127         static function safeEncodeAttribute( $text ) {
1128                 $encValue = Sanitizer::encodeAttribute( $text );
1129
1130                 # Templates and links may be expanded in later parsing,
1131                 # creating invalid or dangerous output. Suppress this.
1132                 $encValue = strtr( $encValue, [
1133                         '<'    => '&lt;',   // This should never happen,
1134                         '>'    => '&gt;',   // we've received invalid input
1135                         '"'    => '&quot;', // which should have been escaped.
1136                         '{'    => '&#123;',
1137                         '}'    => '&#125;', // prevent unpaired language conversion syntax
1138                         '['    => '&#91;',
1139                         "''"   => '&#39;&#39;',
1140                         'ISBN' => '&#73;SBN',
1141                         'RFC'  => '&#82;FC',
1142                         'PMID' => '&#80;MID',
1143                         '|'    => '&#124;',
1144                         '__'   => '&#95;_',
1145                 ] );
1146
1147                 # Stupid hack
1148                 $encValue = preg_replace_callback(
1149                         '/((?i)' . wfUrlProtocols() . ')/',
1150                         [ 'Sanitizer', 'armorLinksCallback' ],
1151                         $encValue );
1152                 return $encValue;
1153         }
1154
1155         /**
1156          * Given a value, escape it so that it can be used in an id attribute and
1157          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1158          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1159          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1160          * escaped with lots of dots.
1161          *
1162          * To ensure we don't have to bother escaping anything, we also strip ', ",
1163          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1164          * We also strip # because it upsets IE, and % because it could be
1165          * ambiguous if it's part of something that looks like a percent escape
1166          * (which don't work reliably in fragments cross-browser).
1167          *
1168          * @see https://www.w3.org/TR/html401/types.html#type-name Valid characters
1169          *   in the id and name attributes
1170          * @see https://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1171          *   the id attribute
1172          * @see https://www.w3.org/TR/html5/dom.html#the-id-attribute
1173          *   HTML5 definition of id attribute
1174          *
1175          * @param string $id Id to escape
1176          * @param string|array $options String or array of strings (default is array()):
1177          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1178          *       so don't pay attention if the first character isn't valid at the
1179          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1180          *       false.
1181          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1182          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1183          *       anchors and links won't break.
1184          * @return string
1185          */
1186         static function escapeId( $id, $options = [] ) {
1187                 global $wgExperimentalHtmlIds;
1188                 $options = (array)$options;
1189
1190                 $id = Sanitizer::decodeCharReferences( $id );
1191
1192                 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1193                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1194                         $id = trim( $id, '_' );
1195                         if ( $id === '' ) {
1196                                 // Must have been all whitespace to start with.
1197                                 return '_';
1198                         } else {
1199                                 return $id;
1200                         }
1201                 }
1202
1203                 // HTML4-style escaping
1204                 static $replace = [
1205                         '%3A' => ':',
1206                         '%' => '.'
1207                 ];
1208
1209                 $id = urlencode( strtr( $id, ' ', '_' ) );
1210                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1211
1212                 if ( !preg_match( '/^[a-zA-Z]/', $id ) && !in_array( 'noninitial', $options ) ) {
1213                         // Initial character must be a letter!
1214                         $id = "x$id";
1215                 }
1216                 return $id;
1217         }
1218
1219         /**
1220          * Given a string containing a space delimited list of ids, escape each id
1221          * to match ids escaped by the escapeId() function.
1222          *
1223          * @since 1.27
1224          *
1225          * @param string $referenceString Space delimited list of ids
1226          * @param string|array $options String or array of strings (default is array()):
1227          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1228          *       so don't pay attention if the first character isn't valid at the
1229          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1230          *       false.
1231          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1232          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1233          *       anchors and links won't break.
1234          * @return string
1235          */
1236         static function escapeIdReferenceList( $referenceString, $options = [] ) {
1237                 # Explode the space delimited list string into an array of tokens
1238                 $references = preg_split( '/\s+/', "{$referenceString}", -1, PREG_SPLIT_NO_EMPTY );
1239
1240                 # Escape each token as an id
1241                 foreach ( $references as &$ref ) {
1242                         $ref = Sanitizer::escapeId( $ref, $options );
1243                 }
1244
1245                 # Merge the array back to a space delimited list string
1246                 # If the array is empty, the result will be an empty string ('')
1247                 $referenceString = implode( ' ', $references );
1248
1249                 return $referenceString;
1250         }
1251
1252         /**
1253          * Given a value, escape it so that it can be used as a CSS class and
1254          * return it.
1255          *
1256          * @todo For extra validity, input should be validated UTF-8.
1257          *
1258          * @see https://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1259          *
1260          * @param string $class
1261          * @return string
1262          */
1263         static function escapeClass( $class ) {
1264                 // Convert ugly stuff to underscores and kill underscores in ugly places
1265                 return rtrim( preg_replace(
1266                         [ '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ],
1267                         '_',
1268                         $class ), '_' );
1269         }
1270
1271         /**
1272          * Given HTML input, escape with htmlspecialchars but un-escape entities.
1273          * This allows (generally harmless) entities like &#160; to survive.
1274          *
1275          * @param string $html HTML to escape
1276          * @return string Escaped input
1277          */
1278         static function escapeHtmlAllowEntities( $html ) {
1279                 $html = Sanitizer::decodeCharReferences( $html );
1280                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1281                 # hurt. Use ENT_SUBSTITUTE so that incorrectly truncated multibyte characters
1282                 # don't cause the entire string to disappear.
1283                 $html = htmlspecialchars( $html, ENT_QUOTES | ENT_SUBSTITUTE );
1284                 return $html;
1285         }
1286
1287         /**
1288          * Regex replace callback for armoring links against further processing.
1289          * @param array $matches
1290          * @return string
1291          */
1292         private static function armorLinksCallback( $matches ) {
1293                 return str_replace( ':', '&#58;', $matches[1] );
1294         }
1295
1296         /**
1297          * Return an associative array of attribute names and values from
1298          * a partial tag string. Attribute names are forced to lowercase,
1299          * character references are decoded to UTF-8 text.
1300          *
1301          * @param string $text
1302          * @return array
1303          */
1304         public static function decodeTagAttributes( $text ) {
1305                 if ( trim( $text ) == '' ) {
1306                         return [];
1307                 }
1308
1309                 $attribs = [];
1310                 $pairs = [];
1311                 if ( !preg_match_all(
1312                         self::getAttribsRegex(),
1313                         $text,
1314                         $pairs,
1315                         PREG_SET_ORDER ) ) {
1316                         return $attribs;
1317                 }
1318
1319                 foreach ( $pairs as $set ) {
1320                         $attribute = strtolower( $set[1] );
1321                         $value = Sanitizer::getTagAttributeCallback( $set );
1322
1323                         // Normalize whitespace
1324                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1325                         $value = trim( $value );
1326
1327                         // Decode character references
1328                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1329                 }
1330                 return $attribs;
1331         }
1332
1333         /**
1334          * Build a partial tag string from an associative array of attribute
1335          * names and values as returned by decodeTagAttributes.
1336          *
1337          * @param array $assoc_array
1338          * @return string
1339          */
1340         public static function safeEncodeTagAttributes( $assoc_array ) {
1341                 $attribs = [];
1342                 foreach ( $assoc_array as $attribute => $value ) {
1343                         $encAttribute = htmlspecialchars( $attribute );
1344                         $encValue = Sanitizer::safeEncodeAttribute( $value );
1345
1346                         $attribs[] = "$encAttribute=\"$encValue\"";
1347                 }
1348                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1349         }
1350
1351         /**
1352          * Pick the appropriate attribute value from a match set from the
1353          * attribs regex matches.
1354          *
1355          * @param array $set
1356          * @throws MWException When tag conditions are not met.
1357          * @return string
1358          */
1359         private static function getTagAttributeCallback( $set ) {
1360                 if ( isset( $set[5] ) ) {
1361                         # No quotes.
1362                         return $set[5];
1363                 } elseif ( isset( $set[4] ) ) {
1364                         # Single-quoted
1365                         return $set[4];
1366                 } elseif ( isset( $set[3] ) ) {
1367                         # Double-quoted
1368                         return $set[3];
1369                 } elseif ( !isset( $set[2] ) ) {
1370                         # In XHTML, attributes must have a value so return an empty string.
1371                         # See "Empty attribute syntax",
1372                         # https://www.w3.org/TR/html5/syntax.html#syntax-attribute-name
1373                         return "";
1374                 } else {
1375                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1376                 }
1377         }
1378
1379         /**
1380          * @param string $text
1381          * @return string
1382          */
1383         private static function normalizeWhitespace( $text ) {
1384                 return preg_replace(
1385                         '/\r\n|[\x20\x0d\x0a\x09]/',
1386                         ' ',
1387                         $text );
1388         }
1389
1390         /**
1391          * Normalizes whitespace in a section name, such as might be returned
1392          * by Parser::stripSectionName(), for use in the id's that are used for
1393          * section links.
1394          *
1395          * @param string $section
1396          * @return string
1397          */
1398         static function normalizeSectionNameWhitespace( $section ) {
1399                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1400         }
1401
1402         /**
1403          * Ensure that any entities and character references are legal
1404          * for XML and XHTML specifically. Any stray bits will be
1405          * &amp;-escaped to result in a valid text fragment.
1406          *
1407          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1408          *   numericized (this way we're well-formed even without a DTD)
1409          * b. any numeric char refs must be legal chars, not invalid or forbidden
1410          * c. use lower cased "&#x", not "&#X"
1411          * d. fix or reject non-valid attributes
1412          *
1413          * @param string $text
1414          * @return string
1415          * @private
1416          */
1417         static function normalizeCharReferences( $text ) {
1418                 return preg_replace_callback(
1419                         self::CHAR_REFS_REGEX,
1420                         [ 'Sanitizer', 'normalizeCharReferencesCallback' ],
1421                         $text );
1422         }
1423
1424         /**
1425          * @param string $matches
1426          * @return string
1427          */
1428         static function normalizeCharReferencesCallback( $matches ) {
1429                 $ret = null;
1430                 if ( $matches[1] != '' ) {
1431                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1432                 } elseif ( $matches[2] != '' ) {
1433                         $ret = Sanitizer::decCharReference( $matches[2] );
1434                 } elseif ( $matches[3] != '' ) {
1435                         $ret = Sanitizer::hexCharReference( $matches[3] );
1436                 }
1437                 if ( is_null( $ret ) ) {
1438                         return htmlspecialchars( $matches[0] );
1439                 } else {
1440                         return $ret;
1441                 }
1442         }
1443
1444         /**
1445          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1446          * return the equivalent numeric entity reference (except for the core &lt;
1447          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1448          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1449          * pseudo-entity source (eg &amp;foo;)
1450          *
1451          * @param string $name
1452          * @return string
1453          */
1454         static function normalizeEntity( $name ) {
1455                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1456                         return '&' . self::$htmlEntityAliases[$name] . ';';
1457                 } elseif ( in_array( $name, [ 'lt', 'gt', 'amp', 'quot' ] ) ) {
1458                         return "&$name;";
1459                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1460                         return '&#' . self::$htmlEntities[$name] . ';';
1461                 } else {
1462                         return "&amp;$name;";
1463                 }
1464         }
1465
1466         /**
1467          * @param int $codepoint
1468          * @return null|string
1469          */
1470         static function decCharReference( $codepoint ) {
1471                 $point = intval( $codepoint );
1472                 if ( Sanitizer::validateCodepoint( $point ) ) {
1473                         return sprintf( '&#%d;', $point );
1474                 } else {
1475                         return null;
1476                 }
1477         }
1478
1479         /**
1480          * @param int $codepoint
1481          * @return null|string
1482          */
1483         static function hexCharReference( $codepoint ) {
1484                 $point = hexdec( $codepoint );
1485                 if ( Sanitizer::validateCodepoint( $point ) ) {
1486                         return sprintf( '&#x%x;', $point );
1487                 } else {
1488                         return null;
1489                 }
1490         }
1491
1492         /**
1493          * Returns true if a given Unicode codepoint is a valid character in
1494          * both HTML5 and XML.
1495          * @param int $codepoint
1496          * @return bool
1497          */
1498         private static function validateCodepoint( $codepoint ) {
1499                 # U+000C is valid in HTML5 but not allowed in XML.
1500                 # U+000D is valid in XML but not allowed in HTML5.
1501                 # U+007F - U+009F are disallowed in HTML5 (control characters).
1502                 return $codepoint == 0x09
1503                         || $codepoint == 0x0a
1504                         || ( $codepoint >= 0x20 && $codepoint <= 0x7e )
1505                         || ( $codepoint >= 0xa0 && $codepoint <= 0xd7ff )
1506                         || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1507                         || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1508         }
1509
1510         /**
1511          * Decode any character references, numeric or named entities,
1512          * in the text and return a UTF-8 string.
1513          *
1514          * @param string $text
1515          * @return string
1516          */
1517         public static function decodeCharReferences( $text ) {
1518                 return preg_replace_callback(
1519                         self::CHAR_REFS_REGEX,
1520                         [ 'Sanitizer', 'decodeCharReferencesCallback' ],
1521                         $text );
1522         }
1523
1524         /**
1525          * Decode any character references, numeric or named entities,
1526          * in the next and normalize the resulting string. (T16952)
1527          *
1528          * This is useful for page titles, not for text to be displayed,
1529          * MediaWiki allows HTML entities to escape normalization as a feature.
1530          *
1531          * @param string $text Already normalized, containing entities
1532          * @return string Still normalized, without entities
1533          */
1534         public static function decodeCharReferencesAndNormalize( $text ) {
1535                 global $wgContLang;
1536                 $text = preg_replace_callback(
1537                         self::CHAR_REFS_REGEX,
1538                         [ 'Sanitizer', 'decodeCharReferencesCallback' ],
1539                         $text, /* limit */ -1, $count );
1540
1541                 if ( $count ) {
1542                         return $wgContLang->normalize( $text );
1543                 } else {
1544                         return $text;
1545                 }
1546         }
1547
1548         /**
1549          * @param string $matches
1550          * @return string
1551          */
1552         static function decodeCharReferencesCallback( $matches ) {
1553                 if ( $matches[1] != '' ) {
1554                         return Sanitizer::decodeEntity( $matches[1] );
1555                 } elseif ( $matches[2] != '' ) {
1556                         return Sanitizer::decodeChar( intval( $matches[2] ) );
1557                 } elseif ( $matches[3] != '' ) {
1558                         return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1559                 }
1560                 # Last case should be an ampersand by itself
1561                 return $matches[0];
1562         }
1563
1564         /**
1565          * Return UTF-8 string for a codepoint if that is a valid
1566          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1567          * @param int $codepoint
1568          * @return string
1569          * @private
1570          */
1571         static function decodeChar( $codepoint ) {
1572                 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1573                         return UtfNormal\Utils::codepointToUtf8( $codepoint );
1574                 } else {
1575                         return UtfNormal\Constants::UTF8_REPLACEMENT;
1576                 }
1577         }
1578
1579         /**
1580          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1581          * return the UTF-8 encoding of that character. Otherwise, returns
1582          * pseudo-entity source (eg "&foo;")
1583          *
1584          * @param string $name
1585          * @return string
1586          */
1587         static function decodeEntity( $name ) {
1588                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1589                         $name = self::$htmlEntityAliases[$name];
1590                 }
1591                 if ( isset( self::$htmlEntities[$name] ) ) {
1592                         return UtfNormal\Utils::codepointToUtf8( self::$htmlEntities[$name] );
1593                 } else {
1594                         return "&$name;";
1595                 }
1596         }
1597
1598         /**
1599          * Fetch the whitelist of acceptable attributes for a given element name.
1600          *
1601          * @param string $element
1602          * @return array
1603          */
1604         static function attributeWhitelist( $element ) {
1605                 $list = Sanitizer::setupAttributeWhitelist();
1606                 return isset( $list[$element] )
1607                         ? $list[$element]
1608                         : [];
1609         }
1610
1611         /**
1612          * Foreach array key (an allowed HTML element), return an array
1613          * of allowed attributes
1614          * @return array
1615          */
1616         static function setupAttributeWhitelist() {
1617                 static $whitelist;
1618
1619                 if ( $whitelist !== null ) {
1620                         return $whitelist;
1621                 }
1622
1623                 $common = [
1624                         # HTML
1625                         'id',
1626                         'class',
1627                         'style',
1628                         'lang',
1629                         'dir',
1630                         'title',
1631
1632                         # WAI-ARIA
1633                         'aria-describedby',
1634                         'aria-flowto',
1635                         'aria-label',
1636                         'aria-labelledby',
1637                         'aria-owns',
1638                         'role',
1639
1640                         # RDFa
1641                         # These attributes are specified in section 9 of
1642                         # https://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1643                         'about',
1644                         'property',
1645                         'resource',
1646                         'datatype',
1647                         'typeof',
1648
1649                         # Microdata. These are specified by
1650                         # https://html.spec.whatwg.org/multipage/microdata.html#the-microdata-model
1651                         'itemid',
1652                         'itemprop',
1653                         'itemref',
1654                         'itemscope',
1655                         'itemtype',
1656                 ];
1657
1658                 $block = array_merge( $common, [ 'align' ] );
1659                 $tablealign = [ 'align', 'valign' ];
1660                 $tablecell = [
1661                         'abbr',
1662                         'axis',
1663                         'headers',
1664                         'scope',
1665                         'rowspan',
1666                         'colspan',
1667                         'nowrap', # deprecated
1668                         'width', # deprecated
1669                         'height', # deprecated
1670                         'bgcolor', # deprecated
1671                 ];
1672
1673                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1674                 # See: https://www.w3.org/TR/html4/
1675                 $whitelist = [
1676                         # 7.5.4
1677                         'div'        => $block,
1678                         'center'     => $common, # deprecated
1679                         'span'       => $common,
1680
1681                         # 7.5.5
1682                         'h1'         => $block,
1683                         'h2'         => $block,
1684                         'h3'         => $block,
1685                         'h4'         => $block,
1686                         'h5'         => $block,
1687                         'h6'         => $block,
1688
1689                         # 7.5.6
1690                         # address
1691
1692                         # 8.2.4
1693                         'bdo'        => $common,
1694
1695                         # 9.2.1
1696                         'em'         => $common,
1697                         'strong'     => $common,
1698                         'cite'       => $common,
1699                         'dfn'        => $common,
1700                         'code'       => $common,
1701                         'samp'       => $common,
1702                         'kbd'        => $common,
1703                         'var'        => $common,
1704                         'abbr'       => $common,
1705                         # acronym
1706
1707                         # 9.2.2
1708                         'blockquote' => array_merge( $common, [ 'cite' ] ),
1709                         'q'          => array_merge( $common, [ 'cite' ] ),
1710
1711                         # 9.2.3
1712                         'sub'        => $common,
1713                         'sup'        => $common,
1714
1715                         # 9.3.1
1716                         'p'          => $block,
1717
1718                         # 9.3.2
1719                         'br'         => array_merge( $common, [ 'clear' ] ),
1720
1721                         # https://www.w3.org/TR/html5/text-level-semantics.html#the-wbr-element
1722                         'wbr'        => $common,
1723
1724                         # 9.3.4
1725                         'pre'        => array_merge( $common, [ 'width' ] ),
1726
1727                         # 9.4
1728                         'ins'        => array_merge( $common, [ 'cite', 'datetime' ] ),
1729                         'del'        => array_merge( $common, [ 'cite', 'datetime' ] ),
1730
1731                         # 10.2
1732                         'ul'         => array_merge( $common, [ 'type' ] ),
1733                         'ol'         => array_merge( $common, [ 'type', 'start', 'reversed' ] ),
1734                         'li'         => array_merge( $common, [ 'type', 'value' ] ),
1735
1736                         # 10.3
1737                         'dl'         => $common,
1738                         'dd'         => $common,
1739                         'dt'         => $common,
1740
1741                         # 11.2.1
1742                         'table'      => array_merge( $common,
1743                                                                 [ 'summary', 'width', 'border', 'frame',
1744                                                                                 'rules', 'cellspacing', 'cellpadding',
1745                                                                                 'align', 'bgcolor',
1746                                                                 ] ),
1747
1748                         # 11.2.2
1749                         'caption'    => $block,
1750
1751                         # 11.2.3
1752                         'thead'      => $common,
1753                         'tfoot'      => $common,
1754                         'tbody'      => $common,
1755
1756                         # 11.2.4
1757                         'colgroup'   => array_merge( $common, [ 'span' ] ),
1758                         'col'        => array_merge( $common, [ 'span' ] ),
1759
1760                         # 11.2.5
1761                         'tr'         => array_merge( $common, [ 'bgcolor' ], $tablealign ),
1762
1763                         # 11.2.6
1764                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1765                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1766
1767                         # 12.2
1768                         # NOTE: <a> is not allowed directly, but the attrib
1769                         # whitelist is used from the Parser object
1770                         'a'          => array_merge( $common, [ 'href', 'rel', 'rev' ] ), # rel/rev esp. for RDFa
1771
1772                         # 13.2
1773                         # Not usually allowed, but may be used for extension-style hooks
1774                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1775                         # true
1776                         'img'        => array_merge( $common, [ 'alt', 'src', 'width', 'height' ] ),
1777
1778                         'video'      => array_merge( $common, [ 'poster', 'controls', 'preload', 'width', 'height' ] ),
1779                         'source'     => array_merge( $common, [ 'type', 'src' ] ),
1780                         'track'      => array_merge( $common, [ 'type', 'src', 'srclang', 'kind', 'label' ] ),
1781
1782                         # 15.2.1
1783                         'tt'         => $common,
1784                         'b'          => $common,
1785                         'i'          => $common,
1786                         'big'        => $common,
1787                         'small'      => $common,
1788                         'strike'     => $common,
1789                         's'          => $common,
1790                         'u'          => $common,
1791
1792                         # 15.2.2
1793                         'font'       => array_merge( $common, [ 'size', 'color', 'face' ] ),
1794                         # basefont
1795
1796                         # 15.3
1797                         'hr'         => array_merge( $common, [ 'width' ] ),
1798
1799                         # HTML Ruby annotation text module, simple ruby only.
1800                         # https://www.w3.org/TR/html5/text-level-semantics.html#the-ruby-element
1801                         'ruby'       => $common,
1802                         # rbc
1803                         'rb'         => $common,
1804                         'rp'         => $common,
1805                         'rt'         => $common, # array_merge( $common, array( 'rbspan' ) ),
1806                         'rtc'        => $common,
1807
1808                         # MathML root element, where used for extensions
1809                         # 'title' may not be 100% valid here; it's XHTML
1810                         # https://www.w3.org/TR/REC-MathML/
1811                         'math'       => [ 'class', 'style', 'id', 'title' ],
1812
1813                         # HTML 5 section 4.6
1814                         'bdi' => $common,
1815
1816                         # HTML5 elements, defined by:
1817                         # https://html.spec.whatwg.org/multipage/semantics.html#the-data-element
1818                         'data' => array_merge( $common, [ 'value' ] ),
1819                         'time' => array_merge( $common, [ 'datetime' ] ),
1820                         'mark' => $common,
1821
1822                         // meta and link are only permitted by removeHTMLtags when Microdata
1823                         // is enabled so we don't bother adding a conditional to hide these
1824                         // Also meta and link are only valid in WikiText as Microdata elements
1825                         // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1826                         // So we don't bother including $common attributes that have no purpose.
1827                         'meta' => [ 'itemprop', 'content' ],
1828                         'link' => [ 'itemprop', 'href' ],
1829                 ];
1830
1831                 return $whitelist;
1832         }
1833
1834         /**
1835          * Take a fragment of (potentially invalid) HTML and return
1836          * a version with any tags removed, encoded as plain text.
1837          *
1838          * Warning: this return value must be further escaped for literal
1839          * inclusion in HTML output as of 1.10!
1840          *
1841          * @param string $text HTML fragment
1842          * @return string
1843          */
1844         static function stripAllTags( $text ) {
1845                 # Actual <tags>
1846                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1847
1848                 # Normalize &entities and whitespace
1849                 $text = self::decodeCharReferences( $text );
1850                 $text = self::normalizeWhitespace( $text );
1851
1852                 return $text;
1853         }
1854
1855         /**
1856          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1857          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1858          * PHP 5.1 doesn't.
1859          *
1860          * Use for passing XHTML fragments to PHP's XML parsing functions
1861          *
1862          * @return string
1863          */
1864         static function hackDocType() {
1865                 $out = "<!DOCTYPE html [\n";
1866                 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1867                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1868                 }
1869                 $out .= "]>\n";
1870                 return $out;
1871         }
1872
1873         /**
1874          * @param string $url
1875          * @return mixed|string
1876          */
1877         static function cleanUrl( $url ) {
1878                 # Normalize any HTML entities in input. They will be
1879                 # re-escaped by makeExternalLink().
1880                 $url = Sanitizer::decodeCharReferences( $url );
1881
1882                 # Escape any control characters introduced by the above step
1883                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1884                         [ __CLASS__, 'cleanUrlCallback' ], $url );
1885
1886                 # Validate hostname portion
1887                 $matches = [];
1888                 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1889                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1890
1891                         // Characters that will be ignored in IDNs.
1892                         // https://tools.ietf.org/html/rfc3454#section-3.1
1893                         // Strip them before further processing so blacklists and such work.
1894                         $strip = "/
1895                                 \\s|          # general whitespace
1896                                 \xc2\xad|     # 00ad SOFT HYPHEN
1897                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1898                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1899                                 \xe2\x81\xa0| # 2060 WORD JOINER
1900                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1901                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1902                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1903                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1904                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1905                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1906                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1907                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1908                                 /xuD";
1909
1910                         $host = preg_replace( $strip, '', $host );
1911
1912                         // IPv6 host names are bracketed with [].  Url-decode these.
1913                         if ( substr_compare( "//%5B", $host, 0, 5 ) === 0 &&
1914                                 preg_match( '!^//%5B([0-9A-Fa-f:.]+)%5D((:\d+)?)$!', $host, $matches )
1915                         ) {
1916                                 $host = '//[' . $matches[1] . ']' . $matches[2];
1917                         }
1918
1919                         // @todo FIXME: Validate hostnames here
1920
1921                         return $protocol . $host . $rest;
1922                 } else {
1923                         return $url;
1924                 }
1925         }
1926
1927         /**
1928          * @param array $matches
1929          * @return string
1930          */
1931         static function cleanUrlCallback( $matches ) {
1932                 return urlencode( $matches[0] );
1933         }
1934
1935         /**
1936          * Does a string look like an e-mail address?
1937          *
1938          * This validates an email address using an HTML5 specification found at:
1939          * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1940          * Which as of 2011-01-24 says:
1941          *
1942          *   A valid e-mail address is a string that matches the ABNF production
1943          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1944          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1945          *   3.5.
1946          *
1947          * This function is an implementation of the specification as requested in
1948          * T24449.
1949          *
1950          * Client-side forms will use the same standard validation rules via JS or
1951          * HTML 5 validation; additional restrictions can be enforced server-side
1952          * by extensions via the 'isValidEmailAddr' hook.
1953          *
1954          * Note that this validation doesn't 100% match RFC 2822, but is believed
1955          * to be liberal enough for wide use. Some invalid addresses will still
1956          * pass validation here.
1957          *
1958          * @since 1.18
1959          *
1960          * @param string $addr E-mail address
1961          * @return bool
1962          */
1963         public static function validateEmail( $addr ) {
1964                 $result = null;
1965                 if ( !Hooks::run( 'isValidEmailAddr', [ $addr, &$result ] ) ) {
1966                         return $result;
1967                 }
1968
1969                 // Please note strings below are enclosed in brackets [], this make the
1970                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1971                 // See T28948
1972                 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1973                 $rfc1034_ldh_str = "a-z0-9\\-";
1974
1975                 $html5_email_regexp = "/
1976                 ^                      # start of string
1977                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1978                 @                      # 'apostrophe'
1979                 [$rfc1034_ldh_str]+       # First domain part
1980                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1981                 $                      # End of string
1982                 /ix"; // case Insensitive, eXtended
1983
1984                 return (bool)preg_match( $html5_email_regexp, $addr );
1985         }
1986 }