includes/Sanitizer.php

   1 <?php
   2 /**
   3  * HTML sanitizer for %MediaWiki.
   4  *
   5  * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
   6  * https://www.mediawiki.org/
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License as published by
  10  * the Free Software Foundation; either version 2 of the License, or
  11  * (at your option) any later version.
  12  *
  13  * This program is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  16  * GNU General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU General Public License along
  19  * with this program; if not, write to the Free Software Foundation, Inc.,
  20  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  21  * http://www.gnu.org/copyleft/gpl.html
  22  *
  23  * @file
  24  * @ingroup Parser
  25  */
  26
  27 /**
  28  * HTML sanitizer for MediaWiki
  29  * @ingroup Parser
  30  */
  31 class Sanitizer {
  32         /**
  33          * Regular expression to match various types of character references in
  34          * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
  35          */
  36         const CHAR_REFS_REGEX =
  37                 '/&([A-Za-z0-9\x80-\xff]+);
  38                  |&\#([0-9]+);
  39                  |&\#[xX]([0-9A-Fa-f]+);
  40                  |(&)/x';
  41
  42         /**
  43          * Acceptable tag name charset from HTML5 parsing spec
  44          * http://www.w3.org/TR/html5/syntax.html#tag-open-state
  45          */
  46         const ELEMENT_BITS_REGEX = '!^(/?)([A-Za-z][^\t\n\v />\0]*+)([^>]*?)(/?>)([^<]*)$!';
  47
  48         /**
  49          * Blacklist for evil uris like javascript:
  50          * WARNING: DO NOT use this in any place that actually requires blacklisting
  51          * for security reasons. There are NUMEROUS[1] ways to bypass blacklisting, the
  52          * only way to be secure from javascript: uri based xss vectors is to whitelist
  53          * things that you know are safe and deny everything else.
  54          * [1]: http://ha.ckers.org/xss.html
  55          */
  56         const EVIL_URI_PATTERN = '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
  57         const XMLNS_ATTRIBUTE_PATTERN = "/^xmlns:[:A-Z_a-z-.0-9]+$/";
  58
  59         /**
  60          * List of all named character entities defined in HTML 4.01
  61          * http://www.w3.org/TR/html4/sgml/entities.html
  62          * As well as &apos; which is only defined starting in XHTML1.
  63          */
  64         private static $htmlEntities = array(
  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 = array(
 324                 'רלמ' => 'rlm',
 325                 'رلم' => 'rlm',
 326         );
 327
 328         /**
 329          * Lazy-initialised attributes regex, see getAttribsRegex()
 330          */
 331         private static $attribsRegex;
 332
 333         /**
 334          * Regular expression to match HTML/XML attribute pairs within a tag.
 335          * Allows some... latitude.
 336          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 337          * @return string
 338          */
 339         static function getAttribsRegex() {
 340                 if ( self::$attribsRegex === null ) {
 341                         $attribFirst = '[:A-Z_a-z0-9]';
 342                         $attrib = '[:A-Z_a-z-.0-9]';
 343                         $space = '[\x09\x0a\x0d\x20]';
 344                         self::$attribsRegex =
 345                                 "/(?:^|$space)({$attribFirst}{$attrib}*)
 346                                   ($space*=$space*
 347                                         (?:
 348                                          # The attribute value: quoted or alone
 349                                           \"([^<\"]*)(?:\"|\$)
 350                                          | '([^<']*)(?:'|\$)
 351                                          |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
 352                                         )
 353                                 )?(?=$space|\$)/sx";
 354                 }
 355                 return self::$attribsRegex;
 356         }
 357
 358         /**
 359          * Return the various lists of recognized tags
 360          * @param array $extratags For any extra tags to include
 361          * @param array $removetags For any tags (default or extra) to exclude
 362          * @return array
 363          */
 364         public static function getRecognizedTagData( $extratags = array(), $removetags = array() ) {
 365                 global $wgAllowMicrodataAttributes, $wgAllowImageTag;
 366
 367                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 368                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 369
 370                 // Base our staticInitialised variable off of the global config state so that if the globals
 371                 // are changed (like in the screwed up test system) we will re-initialise the settings.
 372                 $globalContext = implode( '-', compact( 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
 373                 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
 374                         $htmlpairsStatic = array( # Tags that must be closed
 375                                 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 376                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 377                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 378                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 379                                 'ruby', 'rb', 'rp', 'rt', 'rtc', 'p', 'span', 'abbr', 'dfn',
 380                                 'kbd', 'samp', 'data', 'time', 'mark'
 381                         );
 382                         $htmlsingle = array(
 383                                 'br', 'wbr', 'hr', 'li', 'dt', 'dd'
 384                         );
 385                         $htmlsingleonly = array( # Elements that cannot have close tags
 386                                 'br', 'wbr', 'hr'
 387                         );
 388                         if ( $wgAllowMicrodataAttributes ) {
 389                                 $htmlsingle[] = $htmlsingleonly[] = 'meta';
 390                                 $htmlsingle[] = $htmlsingleonly[] = 'link';
 391                         }
 392                         $htmlnest = array( # Tags that can be nested--??
 393                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 394                                 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
 395                                 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
 396                         );
 397                         $tabletags = array( # Can only appear inside table, we will close them
 398                                 'td', 'th', 'tr',
 399                         );
 400                         $htmllist = array( # Tags used by list
 401                                 'ul', 'ol',
 402                         );
 403                         $listtags = array( # Tags that can appear in a list
 404                                 'li',
 405                         );
 406
 407                         if ( $wgAllowImageTag ) {
 408                                 $htmlsingle[] = 'img';
 409                                 $htmlsingleonly[] = 'img';
 410                         }
 411
 412                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 413                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 414
 415                         # Convert them all to hashtables for faster lookup
 416                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 417                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 418                         foreach ( $vars as $var ) {
 419                                 $$var = array_flip( $$var );
 420                         }
 421                         $staticInitialised = $globalContext;
 422                 }
 423
 424                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 425                 $extratags = array_flip( $extratags );
 426                 $removetags = array_flip( $removetags );
 427                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 428                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
 429
 430                 return array(
 431                         'htmlpairs' => $htmlpairs,
 432                         'htmlsingle' => $htmlsingle,
 433                         'htmlsingleonly' => $htmlsingleonly,
 434                         'htmlnest' => $htmlnest,
 435                         'tabletags' => $tabletags,
 436                         'htmllist' => $htmllist,
 437                         'listtags' => $listtags,
 438                         'htmlsingleallowed' => $htmlsingleallowed,
 439                         'htmlelements' => $htmlelements,
 440                 );
 441         }
 442
 443         /**
 444          * Cleans up HTML, removes dangerous tags and attributes, and
 445          * removes HTML comments
 446          * @param string $text
 447          * @param callable $processCallback Callback to do any variable or parameter
 448          *   replacements in HTML attribute values
 449          * @param array|bool $args Arguments for the processing callback
 450          * @param array $extratags For any extra tags to include
 451          * @param array $removetags For any tags (default or extra) to exclude
 452          * @return string
 453          */
 454         public static function removeHTMLtags( $text, $processCallback = null,
 455                 $args = array(), $extratags = array(), $removetags = array()
 456         ) {
 457                 extract( self::getRecognizedTagData( $extratags, $removetags ) );
 458
 459                 # Remove HTML comments
 460                 $text = Sanitizer::removeHTMLcomments( $text );
 461                 $bits = explode( '<', $text );
 462                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 463                 if ( !MWTidy::isEnabled() ) {
 464                         $tagstack = $tablestack = array();
 465                         foreach ( $bits as $x ) {
 466                                 $regs = array();
 467                                 # $slash: Does the current element start with a '/'?
 468                                 # $t: Current element name
 469                                 # $params: String between element name and >
 470                                 # $brace: Ending '>' or '/>'
 471                                 # $rest: Everything until the next element of $bits
 472                                 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 473                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 474                                 } else {
 475                                         $slash = $t = $params = $brace = $rest = null;
 476                                 }
 477
 478                                 $badtag = false;
 479                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 480                                         # Check our stack
 481                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 482                                                 $badtag = true;
 483                                         } elseif ( $slash ) {
 484                                                 # Closing a tag... is it the one we just opened?
 485                                                 MediaWiki\suppressWarnings();
 486                                                 $ot = array_pop( $tagstack );
 487                                                 MediaWiki\restoreWarnings();
 488
 489                                                 if ( $ot != $t ) {
 490                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 491                                                                 # Pop all elements with an optional close tag
 492                                                                 # and see if we find a match below them
 493                                                                 $optstack = array();
 494                                                                 array_push( $optstack, $ot );
 495                                                                 MediaWiki\suppressWarnings();
 496                                                                 $ot = array_pop( $tagstack );
 497                                                                 MediaWiki\restoreWarnings();
 498                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 499                                                                         array_push( $optstack, $ot );
 500                                                                         MediaWiki\suppressWarnings();
 501                                                                         $ot = array_pop( $tagstack );
 502                                                                         MediaWiki\restoreWarnings();
 503                                                                 }
 504                                                                 if ( $t != $ot ) {
 505                                                                         # No match. Push the optional elements back again
 506                                                                         $badtag = true;
 507                                                                         MediaWiki\suppressWarnings();
 508                                                                         $ot = array_pop( $optstack );
 509                                                                         MediaWiki\restoreWarnings();
 510                                                                         while ( $ot ) {
 511                                                                                 array_push( $tagstack, $ot );
 512                                                                                 MediaWiki\suppressWarnings();
 513                                                                                 $ot = array_pop( $optstack );
 514                                                                                 MediaWiki\restoreWarnings();
 515                                                                         }
 516                                                                 }
 517                                                         } else {
 518                                                                 MediaWiki\suppressWarnings();
 519                                                                 array_push( $tagstack, $ot );
 520                                                                 MediaWiki\restoreWarnings();
 521
 522                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 523                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 524                                                                         $badtag = true;
 525                                                                 }
 526                                                         }
 527                                                 } else {
 528                                                         if ( $t == 'table' ) {
 529                                                                 $tagstack = array_pop( $tablestack );
 530                                                         }
 531                                                 }
 532                                                 $newparams = '';
 533                                         } else {
 534                                                 # Keep track for later
 535                                                 if ( isset( $tabletags[$t] ) && !in_array( 'table', $tagstack ) ) {
 536                                                         $badtag = true;
 537                                                 } elseif ( in_array( $t, $tagstack ) && !isset( $htmlnest[$t] ) ) {
 538                                                         $badtag = true;
 539                                                 # Is it a self closed htmlpair ? (bug 5487)
 540                                                 } elseif ( $brace == '/>' && isset( $htmlpairs[$t] ) ) {
 541                                                         $badtag = true;
 542                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 543                                                         # Hack to force empty tag for unclosable elements
 544                                                         $brace = '/>';
 545                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 546                                                         # Hack to not close $htmlsingle tags
 547                                                         $brace = null;
 548                                                         # Still need to push this optionally-closed tag to
 549                                                         # the tag stack so that we can match end tags
 550                                                         # instead of marking them as bad.
 551                                                         array_push( $tagstack, $t );
 552                                                 } elseif ( isset( $tabletags[$t] ) && in_array( $t, $tagstack ) ) {
 553                                                         // New table tag but forgot to close the previous one
 554                                                         $text .= "</$t>";
 555                                                 } else {
 556                                                         if ( $t == 'table' ) {
 557                                                                 array_push( $tablestack, $tagstack );
 558                                                                 $tagstack = array();
 559                                                         }
 560                                                         array_push( $tagstack, $t );
 561                                                 }
 562
 563                                                 # Replace any variables or template parameters with
 564                                                 # plaintext results.
 565                                                 if ( is_callable( $processCallback ) ) {
 566                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 567                                                 }
 568
 569                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 570                                                         $badtag = true;
 571                                                 }
 572
 573                                                 # Strip non-approved attributes from the tag
 574                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 575                                         }
 576                                         if ( !$badtag ) {
 577                                                 $rest = str_replace( '>', '&gt;', $rest );
 578                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 579                                                 $text .= "<$slash$t$newparams$close>$rest";
 580                                                 continue;
 581                                         }
 582                                 }
 583                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 584                         }
 585                         # Close off any remaining tags
 586                         while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
 587                                 $text .= "</$t>\n";
 588                                 if ( $t == 'table' ) {
 589                                         $tagstack = array_pop( $tablestack );
 590                                 }
 591                         }
 592                 } else {
 593                         # this might be possible using tidy itself
 594                         foreach ( $bits as $x ) {
 595                                 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 596                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 597
 598                                         $badtag = false;
 599                                         if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 600                                                 if ( is_callable( $processCallback ) ) {
 601                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 602                                                 }
 603
 604                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 605                                                         $badtag = true;
 606                                                 }
 607
 608                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 609                                                 if ( !$badtag ) {
 610                                                         $rest = str_replace( '>', '&gt;', $rest );
 611                                                         $text .= "<$slash$t$newparams$brace$rest";
 612                                                         continue;
 613                                                 }
 614                                         }
 615                                 }
 616                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 617                         }
 618                 }
 619                 return $text;
 620         }
 621
 622         /**
 623          * Remove '<!--', '-->', and everything between.
 624          * To avoid leaving blank lines, when a comment is both preceded
 625          * and followed by a newline (ignoring spaces), trim leading and
 626          * trailing spaces and one of the newlines.
 627          *
 628          * @param string $text
 629          * @return string
 630          */
 631         public static function removeHTMLcomments( $text ) {
 632                 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
 633                         $end = strpos( $text, '-->', $start + 4 );
 634                         if ( $end === false ) {
 635                                 # Unterminated comment; bail out
 636                                 break;
 637                         }
 638
 639                         $end += 3;
 640
 641                         # Trim space and newline if the comment is both
 642                         # preceded and followed by a newline
 643                         $spaceStart = max( $start - 1, 0 );
 644                         $spaceLen = $end - $spaceStart;
 645                         while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
 646                                 $spaceStart--;
 647                                 $spaceLen++;
 648                         }
 649                         while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
 650                                 $spaceLen++;
 651                         }
 652                         if ( substr( $text, $spaceStart, 1 ) === "\n"
 653                                 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
 654                                 # Remove the comment, leading and trailing
 655                                 # spaces, and leave only one newline.
 656                                 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
 657                         } else {
 658                                 # Remove just the comment.
 659                                 $text = substr_replace( $text, '', $start, $end - $start );
 660                         }
 661                 }
 662                 return $text;
 663         }
 664
 665         /**
 666          * Takes attribute names and values for a tag and the tag name and
 667          * validates that the tag is allowed to be present.
 668          * This DOES NOT validate the attributes, nor does it validate the
 669          * tags themselves. This method only handles the special circumstances
 670          * where we may want to allow a tag within content but ONLY when it has
 671          * specific attributes set.
 672          *
 673          * @param string $params
 674          * @param string $element
 675          * @return bool
 676          */
 677         static function validateTag( $params, $element ) {
 678                 $params = Sanitizer::decodeTagAttributes( $params );
 679
 680                 if ( $element == 'meta' || $element == 'link' ) {
 681                         if ( !isset( $params['itemprop'] ) ) {
 682                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 683                                 return false;
 684                         }
 685                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 686                                 // <meta> must have a content="" for the itemprop
 687                                 return false;
 688                         }
 689                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 690                                 // <link> must have an associated href=""
 691                                 return false;
 692                         }
 693                 }
 694
 695                 return true;
 696         }
 697
 698         /**
 699          * Take an array of attribute names and values and normalize or discard
 700          * illegal values for the given element type.
 701          *
 702          * - Discards attributes not on a whitelist for the given element
 703          * - Unsafe style attributes are discarded
 704          * - Invalid id attributes are re-encoded
 705          *
 706          * @param array $attribs
 707          * @param string $element
 708          * @return array
 709          *
 710          * @todo Check for legal values where the DTD limits things.
 711          * @todo Check for unique id attribute :P
 712          */
 713         static function validateTagAttributes( $attribs, $element ) {
 714                 return Sanitizer::validateAttributes( $attribs,
 715                         Sanitizer::attributeWhitelist( $element ) );
 716         }
 717
 718         /**
 719          * Take an array of attribute names and values and normalize or discard
 720          * illegal values for the given whitelist.
 721          *
 722          * - Discards attributes not the given whitelist
 723          * - Unsafe style attributes are discarded
 724          * - Invalid id attributes are re-encoded
 725          *
 726          * @param array $attribs
 727          * @param array $whitelist List of allowed attribute names
 728          * @return array
 729          *
 730          * @todo Check for legal values where the DTD limits things.
 731          * @todo Check for unique id attribute :P
 732          */
 733         static function validateAttributes( $attribs, $whitelist ) {
 734                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
 735
 736                 $whitelist = array_flip( $whitelist );
 737                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 738
 739                 $out = array();
 740                 foreach ( $attribs as $attribute => $value ) {
 741                         #allow XML namespace declaration if RDFa is enabled
 742                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 743                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 744                                         $out[$attribute] = $value;
 745                                 }
 746
 747                                 continue;
 748                         }
 749
 750                         # Allow any attribute beginning with "data-"
 751                         if ( !preg_match( '/^data-(?!ooui)/i', $attribute ) && !isset( $whitelist[$attribute] ) ) {
 752                                 continue;
 753                         }
 754
 755                         # Strip javascript "expression" from stylesheets.
 756                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 757                         if ( $attribute == 'style' ) {
 758                                 $value = Sanitizer::checkCss( $value );
 759                         }
 760
 761                         if ( $attribute === 'id' ) {
 762                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 763                         }
 764
 765                         # WAI-ARIA
 766                         # http://www.w3.org/TR/wai-aria/
 767                         # http://www.whatwg.org/html/elements.html#wai-aria
 768                         # For now we only support role="presentation" until we work out what roles should be
 769                         # usable by content and we ensure that our code explicitly rejects patterns that
 770                         # violate HTML5's ARIA restrictions.
 771                         if ( $attribute === 'role' && $value !== 'presentation' ) {
 772                                 continue;
 773                         }
 774
 775                         // RDFa and microdata properties allow URLs, URIs and/or CURIs.
 776                         // Check them for sanity.
 777                         if ( $attribute === 'rel' || $attribute === 'rev'
 778                                 # RDFa
 779                                 || $attribute === 'about' || $attribute === 'property'
 780                                 || $attribute === 'resource' || $attribute === 'datatype'
 781                                 || $attribute === 'typeof'
 782                                 # HTML5 microdata
 783                                 || $attribute === 'itemid' || $attribute === 'itemprop'
 784                                 || $attribute === 'itemref' || $attribute === 'itemscope'
 785                                 || $attribute === 'itemtype'
 786                         ) {
 787                                 //Paranoia. Allow "simple" values but suppress javascript
 788                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 789                                         continue;
 790                                 }
 791                         }
 792
 793                         # NOTE: even though elements using href/src are not allowed directly, supply
 794                         #       validation code that can be used by tag hook handlers, etc
 795                         if ( $attribute === 'href' || $attribute === 'src' ) {
 796                                 if ( !preg_match( $hrefExp, $value ) ) {
 797                                         continue; //drop any href or src attributes not using an allowed protocol.
 798                                         // NOTE: this also drops all relative URLs
 799                                 }
 800                         }
 801
 802                         // If this attribute was previously set, override it.
 803                         // Output should only have one attribute of each name.
 804                         $out[$attribute] = $value;
 805                 }
 806
 807                 if ( $wgAllowMicrodataAttributes ) {
 808                         # itemtype, itemid, itemref don't make sense without itemscope
 809                         if ( !array_key_exists( 'itemscope', $out ) ) {
 810                                 unset( $out['itemtype'] );
 811                                 unset( $out['itemid'] );
 812                                 unset( $out['itemref'] );
 813                         }
 814                         # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 815                 }
 816                 return $out;
 817         }
 818
 819         /**
 820          * Merge two sets of HTML attributes.  Conflicting items in the second set
 821          * will override those in the first, except for 'class' attributes which
 822          * will be combined (if they're both strings).
 823          *
 824          * @todo implement merging for other attributes such as style
 825          * @param array $a
 826          * @param array $b
 827          * @return array
 828          */
 829         static function mergeAttributes( $a, $b ) {
 830                 $out = array_merge( $a, $b );
 831                 if ( isset( $a['class'] ) && isset( $b['class'] )
 832                         && is_string( $a['class'] ) && is_string( $b['class'] )
 833                         && $a['class'] !== $b['class']
 834                 ) {
 835                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 836                                 -1, PREG_SPLIT_NO_EMPTY );
 837                         $out['class'] = implode( ' ', array_unique( $classes ) );
 838                 }
 839                 return $out;
 840         }
 841
 842         /**
 843          * Normalize CSS into a format we can easily search for hostile input
 844          *  - decode character references
 845          *  - decode escape sequences
 846          *  - convert characters that IE6 interprets into ascii
 847          *  - remove comments, unless the entire value is one single comment
 848          * @param string $value the css string
 849          * @return string normalized css
 850          */
 851         public static function normalizeCss( $value ) {
 852
 853                 // Decode character references like &#123;
 854                 $value = Sanitizer::decodeCharReferences( $value );
 855
 856                 // Decode escape sequences and line continuation
 857                 // See the grammar in the CSS 2 spec, appendix D.
 858                 // This has to be done AFTER decoding character references.
 859                 // This means it isn't possible for this function to return
 860                 // unsanitized escape sequences. It is possible to manufacture
 861                 // input that contains character references that decode to
 862                 // escape sequences that decode to character references, but
 863                 // it's OK for the return value to contain character references
 864                 // because the caller is supposed to escape those anyway.
 865                 static $decodeRegex;
 866                 if ( !$decodeRegex ) {
 867                         $space = '[\\x20\\t\\r\\n\\f]';
 868                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 869                         $backslash = '\\\\';
 870                         $decodeRegex = "/ $backslash
 871                                 (?:
 872                                         ($nl) |  # 1. Line continuation
 873                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 874                                         (.) | # 3. backslash cancelling special meaning
 875                                         () | # 4. backslash at end of string
 876                                 )/xu";
 877                 }
 878                 $value = preg_replace_callback( $decodeRegex,
 879                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 880
 881                 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
 882                 $value = preg_replace_callback(
 883                         '/[！-［］-ｚ]/u', // U+FF01 to U+FF5A, excluding U+FF3C (bug 58088)
 884                         function ( $matches ) {
 885                                 $cp = UtfNormal\Utils::utf8ToCodepoint( $matches[0] );
 886                                 if ( $cp === false ) {
 887                                         return '';
 888                                 }
 889                                 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
 890                         },
 891                         $value
 892                 );
 893
 894                 // Convert more characters IE6 might treat as ascii
 895                 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
 896                 $value = str_replace(
 897                         array( 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ),
 898                         array( 'r', 'n', 'n', 'l', 'i', '(', '(' ),
 899                         $value
 900                 );
 901
 902                 // Let the value through if it's nothing but a single comment, to
 903                 // allow other functions which may reject it to pass some error
 904                 // message through.
 905                 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
 906                         // Remove any comments; IE gets token splitting wrong
 907                         // This must be done AFTER decoding character references and
 908                         // escape sequences, because those steps can introduce comments
 909                         // This step cannot introduce character references or escape
 910                         // sequences, because it replaces comments with spaces rather
 911                         // than removing them completely.
 912                         $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 913
 914                         // Remove anything after a comment-start token, to guard against
 915                         // incorrect client implementations.
 916                         $commentPos = strpos( $value, '/*' );
 917                         if ( $commentPos !== false ) {
 918                                 $value = substr( $value, 0, $commentPos );
 919                         }
 920                 }
 921
 922                 // S followed by repeat, iteration, or prolonged sound marks,
 923                 // which IE will treat as "ss"
 924                 $value = preg_replace(
 925                         '/s(?:
 926                                 \xE3\x80\xB1 | # U+3031
 927                                 \xE3\x82\x9D | # U+309D
 928                                 \xE3\x83\xBC | # U+30FC
 929                                 \xE3\x83\xBD | # U+30FD
 930                                 \xEF\xB9\xBC | # U+FE7C
 931                                 \xEF\xB9\xBD | # U+FE7D
 932                                 \xEF\xBD\xB0   # U+FF70
 933                         )/ix',
 934                         'ss',
 935                         $value
 936                 );
 937
 938                 return $value;
 939         }
 940
 941
 942         /**
 943          * Pick apart some CSS and check it for forbidden or unsafe structures.
 944          * Returns a sanitized string. This sanitized string will have
 945          * character references and escape sequences decoded and comments
 946          * stripped (unless it is itself one valid comment, in which case the value
 947          * will be passed through). If the input is just too evil, only a comment
 948          * complaining about evilness will be returned.
 949          *
 950          * Currently URL references, 'expression', 'tps' are forbidden.
 951          *
 952          * NOTE: Despite the fact that character references are decoded, the
 953          * returned string may contain character references given certain
 954          * clever input strings. These character references must
 955          * be escaped before the return value is embedded in HTML.
 956          *
 957          * @param string $value
 958          * @return string
 959          */
 960         static function checkCss( $value ) {
 961                 $value = self::normalizeCss( $value );
 962
 963                 // Reject problematic keywords and control characters
 964                 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ||
 965                         strpos( $value, UtfNormal\Constants::UTF8_REPLACEMENT ) !== false ) {
 966                         return '/* invalid control char */';
 967                 } elseif ( preg_match(
 968                         '! expression
 969                                 | filter\s*:
 970                                 | accelerator\s*:
 971                                 | -o-link\s*:
 972                                 | -o-link-source\s*:
 973                                 | -o-replace\s*:
 974                                 | url\s*\(
 975                                 | image\s*\(
 976                                 | image-set\s*\(
 977                         !ix', $value ) ) {
 978                         return '/* insecure input */';
 979                 }
 980                 return $value;
 981         }
 982
 983         /**
 984          * @param array $matches
 985          * @return string
 986          */
 987         static function cssDecodeCallback( $matches ) {
 988                 if ( $matches[1] !== '' ) {
 989                         // Line continuation
 990                         return '';
 991                 } elseif ( $matches[2] !== '' ) {
 992                         $char = UtfNormal\Utils::codepointToUtf8( hexdec( $matches[2] ) );
 993                 } elseif ( $matches[3] !== '' ) {
 994                         $char = $matches[3];
 995                 } else {
 996                         $char = '\\';
 997                 }
 998                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 999                         // These characters need to be escaped in strings
1000                         // Clean up the escape sequence to avoid parsing errors by clients
1001                         return '\\' . dechex( ord( $char ) ) . ' ';
1002                 } else {
1003                         // Decode unnecessary escape
1004                         return $char;
1005                 }
1006         }
1007
1008         /**
1009          * Take a tag soup fragment listing an HTML element's attributes
1010          * and normalize it to well-formed XML, discarding unwanted attributes.
1011          * Output is safe for further wikitext processing, with escaping of
1012          * values that could trigger problems.
1013          *
1014          * - Normalizes attribute names to lowercase
1015          * - Discards attributes not on a whitelist for the given element
1016          * - Turns broken or invalid entities into plaintext
1017          * - Double-quotes all attribute values
1018          * - Attributes without values are given the name as attribute
1019          * - Double attributes are discarded
1020          * - Unsafe style attributes are discarded
1021          * - Prepends space if there are attributes.
1022          *
1023          * @param string $text
1024          * @param string $element
1025          * @return string
1026          */
1027         static function fixTagAttributes( $text, $element ) {
1028                 if ( trim( $text ) == '' ) {
1029                         return '';
1030                 }
1031
1032                 $decoded = Sanitizer::decodeTagAttributes( $text );
1033                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1034
1035                 return Sanitizer::safeEncodeTagAttributes( $stripped );
1036         }
1037
1038         /**
1039          * Encode an attribute value for HTML output.
1040          * @param string $text
1041          * @return string HTML-encoded text fragment
1042          */
1043         static function encodeAttribute( $text ) {
1044                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1045
1046                 // Whitespace is normalized during attribute decoding,
1047                 // so if we've been passed non-spaces we must encode them
1048                 // ahead of time or they won't be preserved.
1049                 $encValue = strtr( $encValue, array(
1050                         "\n" => '&#10;',
1051                         "\r" => '&#13;',
1052                         "\t" => '&#9;',
1053                 ) );
1054
1055                 return $encValue;
1056         }
1057
1058         /**
1059          * Encode an attribute value for HTML tags, with extra armoring
1060          * against further wiki processing.
1061          * @param string $text
1062          * @return string HTML-encoded text fragment
1063          */
1064         static function safeEncodeAttribute( $text ) {
1065                 $encValue = Sanitizer::encodeAttribute( $text );
1066
1067                 # Templates and links may be expanded in later parsing,
1068                 # creating invalid or dangerous output. Suppress this.
1069                 $encValue = strtr( $encValue, array(
1070                         '<'    => '&lt;',   // This should never happen,
1071                         '>'    => '&gt;',   // we've received invalid input
1072                         '"'    => '&quot;', // which should have been escaped.
1073                         '{'    => '&#123;',
1074                         '['    => '&#91;',
1075                         "''"   => '&#39;&#39;',
1076                         'ISBN' => '&#73;SBN',
1077                         'RFC'  => '&#82;FC',
1078                         'PMID' => '&#80;MID',
1079                         '|'    => '&#124;',
1080                         '__'   => '&#95;_',
1081                 ) );
1082
1083                 # Stupid hack
1084                 $encValue = preg_replace_callback(
1085                         '/((?i)' . wfUrlProtocols() . ')/',
1086                         array( 'Sanitizer', 'armorLinksCallback' ),
1087                         $encValue );
1088                 return $encValue;
1089         }
1090
1091         /**
1092          * Given a value, escape it so that it can be used in an id attribute and
1093          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1094          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1095          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1096          * escaped with lots of dots.
1097          *
1098          * To ensure we don't have to bother escaping anything, we also strip ', ",
1099          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1100          * We also strip # because it upsets IE, and % because it could be
1101          * ambiguous if it's part of something that looks like a percent escape
1102          * (which don't work reliably in fragments cross-browser).
1103          *
1104          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1105          *   in the id and name attributes
1106          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1107          *   the id attribute
1108          * @see http://www.whatwg.org/html/elements.html#the-id-attribute
1109          *   HTML5 definition of id attribute
1110          *
1111          * @param string $id Id to escape
1112          * @param string|array $options String or array of strings (default is array()):
1113          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1114          *       so don't pay attention if the first character isn't valid at the
1115          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1116          *       false.
1117          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1118          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1119          *       anchors and links won't break.
1120          * @return string
1121          */
1122         static function escapeId( $id, $options = array() ) {
1123                 global $wgExperimentalHtmlIds;
1124                 $options = (array)$options;
1125
1126                 $id = Sanitizer::decodeCharReferences( $id );
1127
1128                 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1129                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1130                         $id = trim( $id, '_' );
1131                         if ( $id === '' ) {
1132                                 // Must have been all whitespace to start with.
1133                                 return '_';
1134                         } else {
1135                                 return $id;
1136                         }
1137                 }
1138
1139                 // HTML4-style escaping
1140                 static $replace = array(
1141                         '%3A' => ':',
1142                         '%' => '.'
1143                 );
1144
1145                 $id = urlencode( strtr( $id, ' ', '_' ) );
1146                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1147
1148                 if ( !preg_match( '/^[a-zA-Z]/', $id ) && !in_array( 'noninitial', $options ) ) {
1149                         // Initial character must be a letter!
1150                         $id = "x$id";
1151                 }
1152                 return $id;
1153         }
1154
1155         /**
1156          * Given a value, escape it so that it can be used as a CSS class and
1157          * return it.
1158          *
1159          * @todo For extra validity, input should be validated UTF-8.
1160          *
1161          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1162          *
1163          * @param string $class
1164          * @return string
1165          */
1166         static function escapeClass( $class ) {
1167                 // Convert ugly stuff to underscores and kill underscores in ugly places
1168                 return rtrim( preg_replace(
1169                         array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1170                         '_',
1171                         $class ), '_' );
1172         }
1173
1174         /**
1175          * Given HTML input, escape with htmlspecialchars but un-escape entities.
1176          * This allows (generally harmless) entities like &#160; to survive.
1177          *
1178          * @param string $html HTML to escape
1179          * @return string Escaped input
1180          */
1181         static function escapeHtmlAllowEntities( $html ) {
1182                 $html = Sanitizer::decodeCharReferences( $html );
1183                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1184                 # hurt.
1185                 $html = htmlspecialchars( $html, ENT_QUOTES );
1186                 return $html;
1187         }
1188
1189         /**
1190          * Regex replace callback for armoring links against further processing.
1191          * @param array $matches
1192          * @return string
1193          */
1194         private static function armorLinksCallback( $matches ) {
1195                 return str_replace( ':', '&#58;', $matches[1] );
1196         }
1197
1198         /**
1199          * Return an associative array of attribute names and values from
1200          * a partial tag string. Attribute names are forces to lowercase,
1201          * character references are decoded to UTF-8 text.
1202          *
1203          * @param string $text
1204          * @return array
1205          */
1206         public static function decodeTagAttributes( $text ) {
1207                 if ( trim( $text ) == '' ) {
1208                         return array();
1209                 }
1210
1211                 $attribs = array();
1212                 $pairs = array();
1213                 if ( !preg_match_all(
1214                         self::getAttribsRegex(),
1215                         $text,
1216                         $pairs,
1217                         PREG_SET_ORDER ) ) {
1218                         return $attribs;
1219                 }
1220
1221                 foreach ( $pairs as $set ) {
1222                         $attribute = strtolower( $set[1] );
1223                         $value = Sanitizer::getTagAttributeCallback( $set );
1224
1225                         // Normalize whitespace
1226                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1227                         $value = trim( $value );
1228
1229                         // Decode character references
1230                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1231                 }
1232                 return $attribs;
1233         }
1234
1235         /**
1236          * Build a partial tag string from an associative array of attribute
1237          * names and values as returned by decodeTagAttributes.
1238          *
1239          * @param array $assoc_array
1240          * @return string
1241          */
1242         public static function safeEncodeTagAttributes( $assoc_array ) {
1243                 $attribs = array();
1244                 foreach ( $assoc_array as $attribute => $value ) {
1245                         $encAttribute = htmlspecialchars( $attribute );
1246                         $encValue = Sanitizer::safeEncodeAttribute( $value );
1247
1248                         $attribs[] = "$encAttribute=\"$encValue\"";
1249                 }
1250                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1251         }
1252
1253         /**
1254          * Pick the appropriate attribute value from a match set from the
1255          * attribs regex matches.
1256          *
1257          * @param array $set
1258          * @throws MWException When tag conditions are not met.
1259          * @return string
1260          */
1261         private static function getTagAttributeCallback( $set ) {
1262                 if ( isset( $set[5] ) ) {
1263                         # No quotes.
1264                         return $set[5];
1265                 } elseif ( isset( $set[4] ) ) {
1266                         # Single-quoted
1267                         return $set[4];
1268                 } elseif ( isset( $set[3] ) ) {
1269                         # Double-quoted
1270                         return $set[3];
1271                 } elseif ( !isset( $set[2] ) ) {
1272                         # In XHTML, attributes must have a value so return an empty string.
1273                         # See "Empty attribute syntax",
1274                         # http://www.w3.org/TR/html5/syntax.html#syntax-attribute-name
1275                         return "";
1276                 } else {
1277                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1278                 }
1279         }
1280
1281         /**
1282          * @param string $text
1283          * @return string
1284          */
1285         private static function normalizeWhitespace( $text ) {
1286                 return preg_replace(
1287                         '/\r\n|[\x20\x0d\x0a\x09]/',
1288                         ' ',
1289                         $text );
1290         }
1291
1292         /**
1293          * Normalizes whitespace in a section name, such as might be returned
1294          * by Parser::stripSectionName(), for use in the id's that are used for
1295          * section links.
1296          *
1297          * @param string $section
1298          * @return string
1299          */
1300         static function normalizeSectionNameWhitespace( $section ) {
1301                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1302         }
1303
1304         /**
1305          * Ensure that any entities and character references are legal
1306          * for XML and XHTML specifically. Any stray bits will be
1307          * &amp;-escaped to result in a valid text fragment.
1308          *
1309          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1310          *   numericized (this way we're well-formed even without a DTD)
1311          * b. any numeric char refs must be legal chars, not invalid or forbidden
1312          * c. use lower cased "&#x", not "&#X"
1313          * d. fix or reject non-valid attributes
1314          *
1315          * @param string $text
1316          * @return string
1317          * @private
1318          */
1319         static function normalizeCharReferences( $text ) {
1320                 return preg_replace_callback(
1321                         self::CHAR_REFS_REGEX,
1322                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1323                         $text );
1324         }
1325
1326         /**
1327          * @param string $matches
1328          * @return string
1329          */
1330         static function normalizeCharReferencesCallback( $matches ) {
1331                 $ret = null;
1332                 if ( $matches[1] != '' ) {
1333                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1334                 } elseif ( $matches[2] != '' ) {
1335                         $ret = Sanitizer::decCharReference( $matches[2] );
1336                 } elseif ( $matches[3] != '' ) {
1337                         $ret = Sanitizer::hexCharReference( $matches[3] );
1338                 }
1339                 if ( is_null( $ret ) ) {
1340                         return htmlspecialchars( $matches[0] );
1341                 } else {
1342                         return $ret;
1343                 }
1344         }
1345
1346         /**
1347          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1348          * return the equivalent numeric entity reference (except for the core &lt;
1349          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1350          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1351          * pseudo-entity source (eg &amp;foo;)
1352          *
1353          * @param string $name
1354          * @return string
1355          */
1356         static function normalizeEntity( $name ) {
1357                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1358                         return '&' . self::$htmlEntityAliases[$name] . ';';
1359                 } elseif ( in_array( $name, array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1360                         return "&$name;";
1361                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1362                         return '&#' . self::$htmlEntities[$name] . ';';
1363                 } else {
1364                         return "&amp;$name;";
1365                 }
1366         }
1367
1368         /**
1369          * @param int $codepoint
1370          * @return null|string
1371          */
1372         static function decCharReference( $codepoint ) {
1373                 $point = intval( $codepoint );
1374                 if ( Sanitizer::validateCodepoint( $point ) ) {
1375                         return sprintf( '&#%d;', $point );
1376                 } else {
1377                         return null;
1378                 }
1379         }
1380
1381         /**
1382          * @param int $codepoint
1383          * @return null|string
1384          */
1385         static function hexCharReference( $codepoint ) {
1386                 $point = hexdec( $codepoint );
1387                 if ( Sanitizer::validateCodepoint( $point ) ) {
1388                         return sprintf( '&#x%x;', $point );
1389                 } else {
1390                         return null;
1391                 }
1392         }
1393
1394         /**
1395          * Returns true if a given Unicode codepoint is a valid character in
1396          * both HTML5 and XML.
1397          * @param int $codepoint
1398          * @return bool
1399          */
1400         private static function validateCodepoint( $codepoint ) {
1401                 # U+000C is valid in HTML5 but not allowed in XML.
1402                 # U+000D is valid in XML but not allowed in HTML5.
1403                 # U+007F - U+009F are disallowed in HTML5 (control characters).
1404                 return $codepoint == 0x09
1405                         || $codepoint == 0x0a
1406                         || ( $codepoint >= 0x20 && $codepoint <= 0x7e )
1407                         || ( $codepoint >= 0xa0 && $codepoint <= 0xd7ff )
1408                         || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1409                         || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1410         }
1411
1412         /**
1413          * Decode any character references, numeric or named entities,
1414          * in the text and return a UTF-8 string.
1415          *
1416          * @param string $text
1417          * @return string
1418          */
1419         public static function decodeCharReferences( $text ) {
1420                 return preg_replace_callback(
1421                         self::CHAR_REFS_REGEX,
1422                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1423                         $text );
1424         }
1425
1426         /**
1427          * Decode any character references, numeric or named entities,
1428          * in the next and normalize the resulting string. (bug 14952)
1429          *
1430          * This is useful for page titles, not for text to be displayed,
1431          * MediaWiki allows HTML entities to escape normalization as a feature.
1432          *
1433          * @param string $text Already normalized, containing entities
1434          * @return string Still normalized, without entities
1435          */
1436         public static function decodeCharReferencesAndNormalize( $text ) {
1437                 global $wgContLang;
1438                 $text = preg_replace_callback(
1439                         self::CHAR_REFS_REGEX,
1440                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1441                         $text, /* limit */ -1, $count );
1442
1443                 if ( $count ) {
1444                         return $wgContLang->normalize( $text );
1445                 } else {
1446                         return $text;
1447                 }
1448         }
1449
1450         /**
1451          * @param string $matches
1452          * @return string
1453          */
1454         static function decodeCharReferencesCallback( $matches ) {
1455                 if ( $matches[1] != '' ) {
1456                         return Sanitizer::decodeEntity( $matches[1] );
1457                 } elseif ( $matches[2] != '' ) {
1458                         return Sanitizer::decodeChar( intval( $matches[2] ) );
1459                 } elseif ( $matches[3] != '' ) {
1460                         return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1461                 }
1462                 # Last case should be an ampersand by itself
1463                 return $matches[0];
1464         }
1465
1466         /**
1467          * Return UTF-8 string for a codepoint if that is a valid
1468          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1469          * @param int $codepoint
1470          * @return string
1471          * @private
1472          */
1473         static function decodeChar( $codepoint ) {
1474                 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1475                         return UtfNormal\Utils::codepointToUtf8( $codepoint );
1476                 } else {
1477                         return UtfNormal\Constants::UTF8_REPLACEMENT;
1478                 }
1479         }
1480
1481         /**
1482          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1483          * return the UTF-8 encoding of that character. Otherwise, returns
1484          * pseudo-entity source (eg "&foo;")
1485          *
1486          * @param string $name
1487          * @return string
1488          */
1489         static function decodeEntity( $name ) {
1490                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1491                         $name = self::$htmlEntityAliases[$name];
1492                 }
1493                 if ( isset( self::$htmlEntities[$name] ) ) {
1494                         return UtfNormal\Utils::codepointToUtf8( self::$htmlEntities[$name] );
1495                 } else {
1496                         return "&$name;";
1497                 }
1498         }
1499
1500         /**
1501          * Fetch the whitelist of acceptable attributes for a given element name.
1502          *
1503          * @param string $element
1504          * @return array
1505          */
1506         static function attributeWhitelist( $element ) {
1507                 $list = Sanitizer::setupAttributeWhitelist();
1508                 return isset( $list[$element] )
1509                         ? $list[$element]
1510                         : array();
1511         }
1512
1513         /**
1514          * Foreach array key (an allowed HTML element), return an array
1515          * of allowed attributes
1516          * @return array
1517          */
1518         static function setupAttributeWhitelist() {
1519                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
1520                 static $whitelist, $staticInitialised;
1521
1522                 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgAllowMicrodataAttributes' ) );
1523
1524                 if ( $whitelist !== null && $staticInitialised == $globalContext ) {
1525                         return $whitelist;
1526                 }
1527
1528                 $common = array(
1529                         # HTML
1530                         'id',
1531                         'class',
1532                         'style',
1533                         'lang',
1534                         'dir',
1535                         'title',
1536
1537                         # WAI-ARIA
1538                         'role',
1539                 );
1540
1541                 if ( $wgAllowRdfaAttributes ) {
1542                         # RDFa attributes as specified in section 9 of
1543                         # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1544                         $common = array_merge( $common, array(
1545                                 'about', 'property', 'resource', 'datatype', 'typeof',
1546                         ) );
1547                 }
1548
1549                 if ( $wgAllowMicrodataAttributes ) {
1550                         # add HTML5 microdata tags as specified by
1551                         # http://www.whatwg.org/html/microdata.html#the-microdata-model
1552                         $common = array_merge( $common, array(
1553                                 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1554                         ) );
1555                 }
1556
1557                 $block = array_merge( $common, array( 'align' ) );
1558                 $tablealign = array( 'align', 'valign' );
1559                 $tablecell = array(
1560                         'abbr',
1561                         'axis',
1562                         'headers',
1563                         'scope',
1564                         'rowspan',
1565                         'colspan',
1566                         'nowrap', # deprecated
1567                         'width', # deprecated
1568                         'height', # deprecated
1569                         'bgcolor', # deprecated
1570                 );
1571
1572                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1573                 # See: http://www.w3.org/TR/html4/
1574                 $whitelist = array(
1575                         # 7.5.4
1576                         'div'        => $block,
1577                         'center'     => $common, # deprecated
1578                         'span'       => $common,
1579
1580                         # 7.5.5
1581                         'h1'         => $block,
1582                         'h2'         => $block,
1583                         'h3'         => $block,
1584                         'h4'         => $block,
1585                         'h5'         => $block,
1586                         'h6'         => $block,
1587
1588                         # 7.5.6
1589                         # address
1590
1591                         # 8.2.4
1592                         'bdo'        => $common,
1593
1594                         # 9.2.1
1595                         'em'         => $common,
1596                         'strong'     => $common,
1597                         'cite'       => $common,
1598                         'dfn'        => $common,
1599                         'code'       => $common,
1600                         'samp'       => $common,
1601                         'kbd'        => $common,
1602                         'var'        => $common,
1603                         'abbr'       => $common,
1604                         # acronym
1605
1606                         # 9.2.2
1607                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1608                         'q'          => array_merge( $common, array( 'cite' ) ),
1609
1610                         # 9.2.3
1611                         'sub'        => $common,
1612                         'sup'        => $common,
1613
1614                         # 9.3.1
1615                         'p'          => $block,
1616
1617                         # 9.3.2
1618                         'br'         => array_merge( $common, array( 'clear' ) ),
1619
1620                         # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1621                         'wbr'        => $common,
1622
1623                         # 9.3.4
1624                         'pre'        => array_merge( $common, array( 'width' ) ),
1625
1626                         # 9.4
1627                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1628                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1629
1630                         # 10.2
1631                         'ul'         => array_merge( $common, array( 'type' ) ),
1632                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1633                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1634
1635                         # 10.3
1636                         'dl'         => $common,
1637                         'dd'         => $common,
1638                         'dt'         => $common,
1639
1640                         # 11.2.1
1641                         'table'      => array_merge( $common,
1642                                                                 array( 'summary', 'width', 'border', 'frame',
1643                                                                                 'rules', 'cellspacing', 'cellpadding',
1644                                                                                 'align', 'bgcolor',
1645                                                                 ) ),
1646
1647                         # 11.2.2
1648                         'caption'    => $block,
1649
1650                         # 11.2.3
1651                         'thead'      => $common,
1652                         'tfoot'      => $common,
1653                         'tbody'      => $common,
1654
1655                         # 11.2.4
1656                         'colgroup'   => array_merge( $common, array( 'span' ) ),
1657                         'col'        => array_merge( $common, array( 'span' ) ),
1658
1659                         # 11.2.5
1660                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1661
1662                         # 11.2.6
1663                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1664                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1665
1666                         # 12.2
1667                         # NOTE: <a> is not allowed directly, but the attrib
1668                         # whitelist is used from the Parser object
1669                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1670
1671                         # 13.2
1672                         # Not usually allowed, but may be used for extension-style hooks
1673                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1674                         # true
1675                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1676
1677                         # 15.2.1
1678                         'tt'         => $common,
1679                         'b'          => $common,
1680                         'i'          => $common,
1681                         'big'        => $common,
1682                         'small'      => $common,
1683                         'strike'     => $common,
1684                         's'          => $common,
1685                         'u'          => $common,
1686
1687                         # 15.2.2
1688                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1689                         # basefont
1690
1691                         # 15.3
1692                         'hr'         => array_merge( $common, array( 'width' ) ),
1693
1694                         # HTML Ruby annotation text module, simple ruby only.
1695                         # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
1696                         'ruby'       => $common,
1697                         # rbc
1698                         'rb'         => $common,
1699                         'rp'         => $common,
1700                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1701                         'rtc'         => $common,
1702
1703                         # MathML root element, where used for extensions
1704                         # 'title' may not be 100% valid here; it's XHTML
1705                         # http://www.w3.org/TR/REC-MathML/
1706                         'math'       => array( 'class', 'style', 'id', 'title' ),
1707
1708                         # HTML 5 section 4.6
1709                         'bdi' => $common,
1710
1711                         # HTML5 elements, defined by:
1712                         # http://www.whatwg.org/html/
1713                         'data' => array_merge( $common, array( 'value' ) ),
1714                         'time' => array_merge( $common, array( 'datetime' ) ),
1715                         'mark' => $common,
1716
1717                         // meta and link are only permitted by removeHTMLtags when Microdata
1718                         // is enabled so we don't bother adding a conditional to hide these
1719                         // Also meta and link are only valid in WikiText as Microdata elements
1720                         // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1721                         // So we don't bother including $common attributes that have no purpose.
1722                         'meta' => array( 'itemprop', 'content' ),
1723                         'link' => array( 'itemprop', 'href' ),
1724                 );
1725
1726                 $staticInitialised = $globalContext;
1727
1728                 return $whitelist;
1729         }
1730
1731         /**
1732          * Take a fragment of (potentially invalid) HTML and return
1733          * a version with any tags removed, encoded as plain text.
1734          *
1735          * Warning: this return value must be further escaped for literal
1736          * inclusion in HTML output as of 1.10!
1737          *
1738          * @param string $text HTML fragment
1739          * @return string
1740          */
1741         static function stripAllTags( $text ) {
1742                 # Actual <tags>
1743                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1744
1745                 # Normalize &entities and whitespace
1746                 $text = self::decodeCharReferences( $text );
1747                 $text = self::normalizeWhitespace( $text );
1748
1749                 return $text;
1750         }
1751
1752         /**
1753          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1754          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1755          * PHP 5.1 doesn't.
1756          *
1757          * Use for passing XHTML fragments to PHP's XML parsing functions
1758          *
1759          * @return string
1760          */
1761         static function hackDocType() {
1762                 $out = "<!DOCTYPE html [\n";
1763                 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1764                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1765                 }
1766                 $out .= "]>\n";
1767                 return $out;
1768         }
1769
1770         /**
1771          * @param string $url
1772          * @return mixed|string
1773          */
1774         static function cleanUrl( $url ) {
1775                 # Normalize any HTML entities in input. They will be
1776                 # re-escaped by makeExternalLink().
1777                 $url = Sanitizer::decodeCharReferences( $url );
1778
1779                 # Escape any control characters introduced by the above step
1780                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1781                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1782
1783                 # Validate hostname portion
1784                 $matches = array();
1785                 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1786                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1787
1788                         // Characters that will be ignored in IDNs.
1789                         // http://tools.ietf.org/html/3454#section-3.1
1790                         // Strip them before further processing so blacklists and such work.
1791                         $strip = "/
1792                                 \\s|          # general whitespace
1793                                 \xc2\xad|     # 00ad SOFT HYPHEN
1794                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1795                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1796                                 \xe2\x81\xa0| # 2060 WORD JOINER
1797                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1798                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1799                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1800                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1801                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1802                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1803                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1804                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1805                                 /xuD";
1806
1807                         $host = preg_replace( $strip, '', $host );
1808
1809                         // IPv6 host names are bracketed with [].  Url-decode these.
1810                         if ( substr_compare( "//%5B", $host, 0, 5 ) === 0 && preg_match( '!^//%5B([0-9A-Fa-f:.]+)%5D((:\d+)?)$!', $host, $matches ) ) {
1811                                 $host = '//[' . $matches[1] . ']' . $matches[2];
1812                         }
1813
1814                         // @todo FIXME: Validate hostnames here
1815
1816                         return $protocol . $host . $rest;
1817                 } else {
1818                         return $url;
1819                 }
1820         }
1821
1822         /**
1823          * @param array $matches
1824          * @return string
1825          */
1826         static function cleanUrlCallback( $matches ) {
1827                 return urlencode( $matches[0] );
1828         }
1829
1830         /**
1831          * Does a string look like an e-mail address?
1832          *
1833          * This validates an email address using an HTML5 specification found at:
1834          * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1835          * Which as of 2011-01-24 says:
1836          *
1837          *   A valid e-mail address is a string that matches the ABNF production
1838          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1839          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1840          *   3.5.
1841          *
1842          * This function is an implementation of the specification as requested in
1843          * bug 22449.
1844          *
1845          * Client-side forms will use the same standard validation rules via JS or
1846          * HTML 5 validation; additional restrictions can be enforced server-side
1847          * by extensions via the 'isValidEmailAddr' hook.
1848          *
1849          * Note that this validation doesn't 100% match RFC 2822, but is believed
1850          * to be liberal enough for wide use. Some invalid addresses will still
1851          * pass validation here.
1852          *
1853          * @since 1.18
1854          *
1855          * @param string $addr E-mail address
1856          * @return bool
1857          */
1858         public static function validateEmail( $addr ) {
1859                 $result = null;
1860                 if ( !Hooks::run( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1861                         return $result;
1862                 }
1863
1864                 // Please note strings below are enclosed in brackets [], this make the
1865                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1866                 // See bug 26948
1867                 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1868                 $rfc1034_ldh_str = "a-z0-9\\-";
1869
1870                 $html5_email_regexp = "/
1871                 ^                      # start of string
1872                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1873                 @                      # 'apostrophe'
1874                 [$rfc1034_ldh_str]+       # First domain part
1875                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1876                 $                      # End of string
1877                 /ix"; // case Insensitive, eXtended
1878
1879                 return (bool)preg_match( $html5_email_regexp, $addr );
1880         }
1881 }