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                 global $wgUseTidy;
 458
 459                 extract( self::getRecognizedTagData( $extratags, $removetags ) );
 460
 461                 # Remove HTML comments
 462                 $text = Sanitizer::removeHTMLcomments( $text );
 463                 $bits = explode( '<', $text );
 464                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 465                 if ( !$wgUseTidy ) {
 466                         $tagstack = $tablestack = array();
 467                         foreach ( $bits as $x ) {
 468                                 $regs = array();
 469                                 # $slash: Does the current element start with a '/'?
 470                                 # $t: Current element name
 471                                 # $params: String between element name and >
 472                                 # $brace: Ending '>' or '/>'
 473                                 # $rest: Everything until the next element of $bits
 474                                 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 475                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 476                                 } else {
 477                                         $slash = $t = $params = $brace = $rest = null;
 478                                 }
 479
 480                                 $badtag = false;
 481                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 482                                         # Check our stack
 483                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 484                                                 $badtag = true;
 485                                         } elseif ( $slash ) {
 486                                                 # Closing a tag... is it the one we just opened?
 487                                                 MediaWiki\suppressWarnings();
 488                                                 $ot = array_pop( $tagstack );
 489                                                 MediaWiki\restoreWarnings();
 490
 491                                                 if ( $ot != $t ) {
 492                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 493                                                                 # Pop all elements with an optional close tag
 494                                                                 # and see if we find a match below them
 495                                                                 $optstack = array();
 496                                                                 array_push( $optstack, $ot );
 497                                                                 MediaWiki\suppressWarnings();
 498                                                                 $ot = array_pop( $tagstack );
 499                                                                 MediaWiki\restoreWarnings();
 500                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 501                                                                         array_push( $optstack, $ot );
 502                                                                         MediaWiki\suppressWarnings();
 503                                                                         $ot = array_pop( $tagstack );
 504                                                                         MediaWiki\restoreWarnings();
 505                                                                 }
 506                                                                 if ( $t != $ot ) {
 507                                                                         # No match. Push the optional elements back again
 508                                                                         $badtag = true;
 509                                                                         MediaWiki\suppressWarnings();
 510                                                                         $ot = array_pop( $optstack );
 511                                                                         MediaWiki\restoreWarnings();
 512                                                                         while ( $ot ) {
 513                                                                                 array_push( $tagstack, $ot );
 514                                                                                 MediaWiki\suppressWarnings();
 515                                                                                 $ot = array_pop( $optstack );
 516                                                                                 MediaWiki\restoreWarnings();
 517                                                                         }
 518                                                                 }
 519                                                         } else {
 520                                                                 MediaWiki\suppressWarnings();
 521                                                                 array_push( $tagstack, $ot );
 522                                                                 MediaWiki\restoreWarnings();
 523
 524                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 525                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 526                                                                         $badtag = true;
 527                                                                 }
 528                                                         }
 529                                                 } else {
 530                                                         if ( $t == 'table' ) {
 531                                                                 $tagstack = array_pop( $tablestack );
 532                                                         }
 533                                                 }
 534                                                 $newparams = '';
 535                                         } else {
 536                                                 # Keep track for later
 537                                                 if ( isset( $tabletags[$t] ) && !in_array( 'table', $tagstack ) ) {
 538                                                         $badtag = true;
 539                                                 } elseif ( in_array( $t, $tagstack ) && !isset( $htmlnest[$t] ) ) {
 540                                                         $badtag = true;
 541                                                 # Is it a self closed htmlpair ? (bug 5487)
 542                                                 } elseif ( $brace == '/>' && isset( $htmlpairs[$t] ) ) {
 543                                                         $badtag = true;
 544                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 545                                                         # Hack to force empty tag for unclosable elements
 546                                                         $brace = '/>';
 547                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 548                                                         # Hack to not close $htmlsingle tags
 549                                                         $brace = null;
 550                                                         # Still need to push this optionally-closed tag to
 551                                                         # the tag stack so that we can match end tags
 552                                                         # instead of marking them as bad.
 553                                                         array_push( $tagstack, $t );
 554                                                 } elseif ( isset( $tabletags[$t] ) && in_array( $t, $tagstack ) ) {
 555                                                         // New table tag but forgot to close the previous one
 556                                                         $text .= "</$t>";
 557                                                 } else {
 558                                                         if ( $t == 'table' ) {
 559                                                                 array_push( $tablestack, $tagstack );
 560                                                                 $tagstack = array();
 561                                                         }
 562                                                         array_push( $tagstack, $t );
 563                                                 }
 564
 565                                                 # Replace any variables or template parameters with
 566                                                 # plaintext results.
 567                                                 if ( is_callable( $processCallback ) ) {
 568                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 569                                                 }
 570
 571                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 572                                                         $badtag = true;
 573                                                 }
 574
 575                                                 # Strip non-approved attributes from the tag
 576                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 577                                         }
 578                                         if ( !$badtag ) {
 579                                                 $rest = str_replace( '>', '&gt;', $rest );
 580                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 581                                                 $text .= "<$slash$t$newparams$close>$rest";
 582                                                 continue;
 583                                         }
 584                                 }
 585                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 586                         }
 587                         # Close off any remaining tags
 588                         while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
 589                                 $text .= "</$t>\n";
 590                                 if ( $t == 'table' ) {
 591                                         $tagstack = array_pop( $tablestack );
 592                                 }
 593                         }
 594                 } else {
 595                         # this might be possible using tidy itself
 596                         foreach ( $bits as $x ) {
 597                                 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 598                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 599
 600                                         $badtag = false;
 601                                         if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 602                                                 if ( is_callable( $processCallback ) ) {
 603                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 604                                                 }
 605
 606                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 607                                                         $badtag = true;
 608                                                 }
 609
 610                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 611                                                 if ( !$badtag ) {
 612                                                         $rest = str_replace( '>', '&gt;', $rest );
 613                                                         $text .= "<$slash$t$newparams$brace$rest";
 614                                                         continue;
 615                                                 }
 616                                         }
 617                                 }
 618                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 619                         }
 620                 }
 621                 return $text;
 622         }
 623
 624         /**
 625          * Remove '<!--', '-->', and everything between.
 626          * To avoid leaving blank lines, when a comment is both preceded
 627          * and followed by a newline (ignoring spaces), trim leading and
 628          * trailing spaces and one of the newlines.
 629          *
 630          * @param string $text
 631          * @return string
 632          */
 633         public static function removeHTMLcomments( $text ) {
 634                 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
 635                         $end = strpos( $text, '-->', $start + 4 );
 636                         if ( $end === false ) {
 637                                 # Unterminated comment; bail out
 638                                 break;
 639                         }
 640
 641                         $end += 3;
 642
 643                         # Trim space and newline if the comment is both
 644                         # preceded and followed by a newline
 645                         $spaceStart = max( $start - 1, 0 );
 646                         $spaceLen = $end - $spaceStart;
 647                         while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
 648                                 $spaceStart--;
 649                                 $spaceLen++;
 650                         }
 651                         while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
 652                                 $spaceLen++;
 653                         }
 654                         if ( substr( $text, $spaceStart, 1 ) === "\n"
 655                                 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
 656                                 # Remove the comment, leading and trailing
 657                                 # spaces, and leave only one newline.
 658                                 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
 659                         } else {
 660                                 # Remove just the comment.
 661                                 $text = substr_replace( $text, '', $start, $end - $start );
 662                         }
 663                 }
 664                 return $text;
 665         }
 666
 667         /**
 668          * Takes attribute names and values for a tag and the tag name and
 669          * validates that the tag is allowed to be present.
 670          * This DOES NOT validate the attributes, nor does it validate the
 671          * tags themselves. This method only handles the special circumstances
 672          * where we may want to allow a tag within content but ONLY when it has
 673          * specific attributes set.
 674          *
 675          * @param string $params
 676          * @param string $element
 677          * @return bool
 678          */
 679         static function validateTag( $params, $element ) {
 680                 $params = Sanitizer::decodeTagAttributes( $params );
 681
 682                 if ( $element == 'meta' || $element == 'link' ) {
 683                         if ( !isset( $params['itemprop'] ) ) {
 684                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 685                                 return false;
 686                         }
 687                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 688                                 // <meta> must have a content="" for the itemprop
 689                                 return false;
 690                         }
 691                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 692                                 // <link> must have an associated href=""
 693                                 return false;
 694                         }
 695                 }
 696
 697                 return true;
 698         }
 699
 700         /**
 701          * Take an array of attribute names and values and normalize or discard
 702          * illegal values for the given element type.
 703          *
 704          * - Discards attributes not on a whitelist for the given element
 705          * - Unsafe style attributes are discarded
 706          * - Invalid id attributes are re-encoded
 707          *
 708          * @param array $attribs
 709          * @param string $element
 710          * @return array
 711          *
 712          * @todo Check for legal values where the DTD limits things.
 713          * @todo Check for unique id attribute :P
 714          */
 715         static function validateTagAttributes( $attribs, $element ) {
 716                 return Sanitizer::validateAttributes( $attribs,
 717                         Sanitizer::attributeWhitelist( $element ) );
 718         }
 719
 720         /**
 721          * Take an array of attribute names and values and normalize or discard
 722          * illegal values for the given whitelist.
 723          *
 724          * - Discards attributes not the given whitelist
 725          * - Unsafe style attributes are discarded
 726          * - Invalid id attributes are re-encoded
 727          *
 728          * @param array $attribs
 729          * @param array $whitelist List of allowed attribute names
 730          * @return array
 731          *
 732          * @todo Check for legal values where the DTD limits things.
 733          * @todo Check for unique id attribute :P
 734          */
 735         static function validateAttributes( $attribs, $whitelist ) {
 736                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
 737
 738                 $whitelist = array_flip( $whitelist );
 739                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 740
 741                 $out = array();
 742                 foreach ( $attribs as $attribute => $value ) {
 743                         #allow XML namespace declaration if RDFa is enabled
 744                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 745                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 746                                         $out[$attribute] = $value;
 747                                 }
 748
 749                                 continue;
 750                         }
 751
 752                         # Allow any attribute beginning with "data-"
 753                         if ( !preg_match( '/^data-(?!ooui)/i', $attribute ) && !isset( $whitelist[$attribute] ) ) {
 754                                 continue;
 755                         }
 756
 757                         # Strip javascript "expression" from stylesheets.
 758                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 759                         if ( $attribute == 'style' ) {
 760                                 $value = Sanitizer::checkCss( $value );
 761                         }
 762
 763                         if ( $attribute === 'id' ) {
 764                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 765                         }
 766
 767                         # WAI-ARIA
 768                         # http://www.w3.org/TR/wai-aria/
 769                         # http://www.whatwg.org/html/elements.html#wai-aria
 770                         # For now we only support role="presentation" until we work out what roles should be
 771                         # usable by content and we ensure that our code explicitly rejects patterns that
 772                         # violate HTML5's ARIA restrictions.
 773                         if ( $attribute === 'role' && $value !== 'presentation' ) {
 774                                 continue;
 775                         }
 776
 777                         // RDFa and microdata properties allow URLs, URIs and/or CURIs.
 778                         // Check them for sanity.
 779                         if ( $attribute === 'rel' || $attribute === 'rev'
 780                                 # RDFa
 781                                 || $attribute === 'about' || $attribute === 'property'
 782                                 || $attribute === 'resource' || $attribute === 'datatype'
 783                                 || $attribute === 'typeof'
 784                                 # HTML5 microdata
 785                                 || $attribute === 'itemid' || $attribute === 'itemprop'
 786                                 || $attribute === 'itemref' || $attribute === 'itemscope'
 787                                 || $attribute === 'itemtype'
 788                         ) {
 789                                 //Paranoia. Allow "simple" values but suppress javascript
 790                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 791                                         continue;
 792                                 }
 793                         }
 794
 795                         # NOTE: even though elements using href/src are not allowed directly, supply
 796                         #       validation code that can be used by tag hook handlers, etc
 797                         if ( $attribute === 'href' || $attribute === 'src' ) {
 798                                 if ( !preg_match( $hrefExp, $value ) ) {
 799                                         continue; //drop any href or src attributes not using an allowed protocol.
 800                                         // NOTE: this also drops all relative URLs
 801                                 }
 802                         }
 803
 804                         // If this attribute was previously set, override it.
 805                         // Output should only have one attribute of each name.
 806                         $out[$attribute] = $value;
 807                 }
 808
 809                 if ( $wgAllowMicrodataAttributes ) {
 810                         # itemtype, itemid, itemref don't make sense without itemscope
 811                         if ( !array_key_exists( 'itemscope', $out ) ) {
 812                                 unset( $out['itemtype'] );
 813                                 unset( $out['itemid'] );
 814                                 unset( $out['itemref'] );
 815                         }
 816                         # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 817                 }
 818                 return $out;
 819         }
 820
 821         /**
 822          * Merge two sets of HTML attributes.  Conflicting items in the second set
 823          * will override those in the first, except for 'class' attributes which
 824          * will be combined (if they're both strings).
 825          *
 826          * @todo implement merging for other attributes such as style
 827          * @param array $a
 828          * @param array $b
 829          * @return array
 830          */
 831         static function mergeAttributes( $a, $b ) {
 832                 $out = array_merge( $a, $b );
 833                 if ( isset( $a['class'] ) && isset( $b['class'] )
 834                         && is_string( $a['class'] ) && is_string( $b['class'] )
 835                         && $a['class'] !== $b['class']
 836                 ) {
 837                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 838                                 -1, PREG_SPLIT_NO_EMPTY );
 839                         $out['class'] = implode( ' ', array_unique( $classes ) );
 840                 }
 841                 return $out;
 842         }
 843
 844         /**
 845          * Normalize CSS into a format we can easily search for hostile input
 846          *  - decode character references
 847          *  - decode escape sequences
 848          *  - convert characters that IE6 interprets into ascii
 849          *  - remove comments, unless the entire value is one single comment
 850          * @param string $value the css string
 851          * @return string normalized css
 852          */
 853         public static function normalizeCss( $value ) {
 854
 855                 // Decode character references like &#123;
 856                 $value = Sanitizer::decodeCharReferences( $value );
 857
 858                 // Decode escape sequences and line continuation
 859                 // See the grammar in the CSS 2 spec, appendix D.
 860                 // This has to be done AFTER decoding character references.
 861                 // This means it isn't possible for this function to return
 862                 // unsanitized escape sequences. It is possible to manufacture
 863                 // input that contains character references that decode to
 864                 // escape sequences that decode to character references, but
 865                 // it's OK for the return value to contain character references
 866                 // because the caller is supposed to escape those anyway.
 867                 static $decodeRegex;
 868                 if ( !$decodeRegex ) {
 869                         $space = '[\\x20\\t\\r\\n\\f]';
 870                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 871                         $backslash = '\\\\';
 872                         $decodeRegex = "/ $backslash
 873                                 (?:
 874                                         ($nl) |  # 1. Line continuation
 875                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 876                                         (.) | # 3. backslash cancelling special meaning
 877                                         () | # 4. backslash at end of string
 878                                 )/xu";
 879                 }
 880                 $value = preg_replace_callback( $decodeRegex,
 881                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 882
 883                 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
 884                 $value = preg_replace_callback(
 885                         '/[！-［］-ｚ]/u', // U+FF01 to U+FF5A, excluding U+FF3C (bug 58088)
 886                         function ( $matches ) {
 887                                 $cp = UtfNormal\Utils::utf8ToCodepoint( $matches[0] );
 888                                 if ( $cp === false ) {
 889                                         return '';
 890                                 }
 891                                 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
 892                         },
 893                         $value
 894                 );
 895
 896                 // Convert more characters IE6 might treat as ascii
 897                 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
 898                 $value = str_replace(
 899                         array( 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ),
 900                         array( 'r', 'n', 'n', 'l', 'i', '(', '(' ),
 901                         $value
 902                 );
 903
 904                 // Let the value through if it's nothing but a single comment, to
 905                 // allow other functions which may reject it to pass some error
 906                 // message through.
 907                 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
 908                         // Remove any comments; IE gets token splitting wrong
 909                         // This must be done AFTER decoding character references and
 910                         // escape sequences, because those steps can introduce comments
 911                         // This step cannot introduce character references or escape
 912                         // sequences, because it replaces comments with spaces rather
 913                         // than removing them completely.
 914                         $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 915
 916                         // Remove anything after a comment-start token, to guard against
 917                         // incorrect client implementations.
 918                         $commentPos = strpos( $value, '/*' );
 919                         if ( $commentPos !== false ) {
 920                                 $value = substr( $value, 0, $commentPos );
 921                         }
 922                 }
 923
 924                 // S followed by repeat, iteration, or prolonged sound marks,
 925                 // which IE will treat as "ss"
 926                 $value = preg_replace(
 927                         '/s(?:
 928                                 \xE3\x80\xB1 | # U+3031
 929                                 \xE3\x82\x9D | # U+309D
 930                                 \xE3\x83\xBC | # U+30FC
 931                                 \xE3\x83\xBD | # U+30FD
 932                                 \xEF\xB9\xBC | # U+FE7C
 933                                 \xEF\xB9\xBD | # U+FE7D
 934                                 \xEF\xBD\xB0   # U+FF70
 935                         )/ix',
 936                         'ss',
 937                         $value
 938                 );
 939
 940                 return $value;
 941         }
 942
 943
 944         /**
 945          * Pick apart some CSS and check it for forbidden or unsafe structures.
 946          * Returns a sanitized string. This sanitized string will have
 947          * character references and escape sequences decoded and comments
 948          * stripped (unless it is itself one valid comment, in which case the value
 949          * will be passed through). If the input is just too evil, only a comment
 950          * complaining about evilness will be returned.
 951          *
 952          * Currently URL references, 'expression', 'tps' are forbidden.
 953          *
 954          * NOTE: Despite the fact that character references are decoded, the
 955          * returned string may contain character references given certain
 956          * clever input strings. These character references must
 957          * be escaped before the return value is embedded in HTML.
 958          *
 959          * @param string $value
 960          * @return string
 961          */
 962         static function checkCss( $value ) {
 963                 $value = self::normalizeCss( $value );
 964
 965                 // Reject problematic keywords and control characters
 966                 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ||
 967                         strpos( $value, UtfNormal\Constants::UTF8_REPLACEMENT ) !== false ) {
 968                         return '/* invalid control char */';
 969                 } elseif ( preg_match(
 970                         '! expression
 971                                 | filter\s*:
 972                                 | accelerator\s*:
 973                                 | -o-link\s*:
 974                                 | -o-link-source\s*:
 975                                 | -o-replace\s*:
 976                                 | url\s*\(
 977                                 | image\s*\(
 978                                 | image-set\s*\(
 979                         !ix', $value ) ) {
 980                         return '/* insecure input */';
 981                 }
 982                 return $value;
 983         }
 984
 985         /**
 986          * @param array $matches
 987          * @return string
 988          */
 989         static function cssDecodeCallback( $matches ) {
 990                 if ( $matches[1] !== '' ) {
 991                         // Line continuation
 992                         return '';
 993                 } elseif ( $matches[2] !== '' ) {
 994                         $char = UtfNormal\Utils::codepointToUtf8( hexdec( $matches[2] ) );
 995                 } elseif ( $matches[3] !== '' ) {
 996                         $char = $matches[3];
 997                 } else {
 998                         $char = '\\';
 999                 }
1000                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
1001                         // These characters need to be escaped in strings
1002                         // Clean up the escape sequence to avoid parsing errors by clients
1003                         return '\\' . dechex( ord( $char ) ) . ' ';
1004                 } else {
1005                         // Decode unnecessary escape
1006                         return $char;
1007                 }
1008         }
1009
1010         /**
1011          * Take a tag soup fragment listing an HTML element's attributes
1012          * and normalize it to well-formed XML, discarding unwanted attributes.
1013          * Output is safe for further wikitext processing, with escaping of
1014          * values that could trigger problems.
1015          *
1016          * - Normalizes attribute names to lowercase
1017          * - Discards attributes not on a whitelist for the given element
1018          * - Turns broken or invalid entities into plaintext
1019          * - Double-quotes all attribute values
1020          * - Attributes without values are given the name as attribute
1021          * - Double attributes are discarded
1022          * - Unsafe style attributes are discarded
1023          * - Prepends space if there are attributes.
1024          *
1025          * @param string $text
1026          * @param string $element
1027          * @return string
1028          */
1029         static function fixTagAttributes( $text, $element ) {
1030                 if ( trim( $text ) == '' ) {
1031                         return '';
1032                 }
1033
1034                 $decoded = Sanitizer::decodeTagAttributes( $text );
1035                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1036
1037                 return Sanitizer::safeEncodeTagAttributes( $stripped );
1038         }
1039
1040         /**
1041          * Encode an attribute value for HTML output.
1042          * @param string $text
1043          * @return string HTML-encoded text fragment
1044          */
1045         static function encodeAttribute( $text ) {
1046                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1047
1048                 // Whitespace is normalized during attribute decoding,
1049                 // so if we've been passed non-spaces we must encode them
1050                 // ahead of time or they won't be preserved.
1051                 $encValue = strtr( $encValue, array(
1052                         "\n" => '&#10;',
1053                         "\r" => '&#13;',
1054                         "\t" => '&#9;',
1055                 ) );
1056
1057                 return $encValue;
1058         }
1059
1060         /**
1061          * Encode an attribute value for HTML tags, with extra armoring
1062          * against further wiki processing.
1063          * @param string $text
1064          * @return string HTML-encoded text fragment
1065          */
1066         static function safeEncodeAttribute( $text ) {
1067                 $encValue = Sanitizer::encodeAttribute( $text );
1068
1069                 # Templates and links may be expanded in later parsing,
1070                 # creating invalid or dangerous output. Suppress this.
1071                 $encValue = strtr( $encValue, array(
1072                         '<'    => '&lt;',   // This should never happen,
1073                         '>'    => '&gt;',   // we've received invalid input
1074                         '"'    => '&quot;', // which should have been escaped.
1075                         '{'    => '&#123;',
1076                         '['    => '&#91;',
1077                         "''"   => '&#39;&#39;',
1078                         'ISBN' => '&#73;SBN',
1079                         'RFC'  => '&#82;FC',
1080                         'PMID' => '&#80;MID',
1081                         '|'    => '&#124;',
1082                         '__'   => '&#95;_',
1083                 ) );
1084
1085                 # Stupid hack
1086                 $encValue = preg_replace_callback(
1087                         '/((?i)' . wfUrlProtocols() . ')/',
1088                         array( 'Sanitizer', 'armorLinksCallback' ),
1089                         $encValue );
1090                 return $encValue;
1091         }
1092
1093         /**
1094          * Given a value, escape it so that it can be used in an id attribute and
1095          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1096          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1097          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1098          * escaped with lots of dots.
1099          *
1100          * To ensure we don't have to bother escaping anything, we also strip ', ",
1101          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1102          * We also strip # because it upsets IE, and % because it could be
1103          * ambiguous if it's part of something that looks like a percent escape
1104          * (which don't work reliably in fragments cross-browser).
1105          *
1106          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1107          *   in the id and name attributes
1108          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1109          *   the id attribute
1110          * @see http://www.whatwg.org/html/elements.html#the-id-attribute
1111          *   HTML5 definition of id attribute
1112          *
1113          * @param string $id Id to escape
1114          * @param string|array $options String or array of strings (default is array()):
1115          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1116          *       so don't pay attention if the first character isn't valid at the
1117          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1118          *       false.
1119          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1120          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1121          *       anchors and links won't break.
1122          * @return string
1123          */
1124         static function escapeId( $id, $options = array() ) {
1125                 global $wgExperimentalHtmlIds;
1126                 $options = (array)$options;
1127
1128                 $id = Sanitizer::decodeCharReferences( $id );
1129
1130                 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1131                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1132                         $id = trim( $id, '_' );
1133                         if ( $id === '' ) {
1134                                 // Must have been all whitespace to start with.
1135                                 return '_';
1136                         } else {
1137                                 return $id;
1138                         }
1139                 }
1140
1141                 // HTML4-style escaping
1142                 static $replace = array(
1143                         '%3A' => ':',
1144                         '%' => '.'
1145                 );
1146
1147                 $id = urlencode( strtr( $id, ' ', '_' ) );
1148                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1149
1150                 if ( !preg_match( '/^[a-zA-Z]/', $id ) && !in_array( 'noninitial', $options ) ) {
1151                         // Initial character must be a letter!
1152                         $id = "x$id";
1153                 }
1154                 return $id;
1155         }
1156
1157         /**
1158          * Given a value, escape it so that it can be used as a CSS class and
1159          * return it.
1160          *
1161          * @todo For extra validity, input should be validated UTF-8.
1162          *
1163          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1164          *
1165          * @param string $class
1166          * @return string
1167          */
1168         static function escapeClass( $class ) {
1169                 // Convert ugly stuff to underscores and kill underscores in ugly places
1170                 return rtrim( preg_replace(
1171                         array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1172                         '_',
1173                         $class ), '_' );
1174         }
1175
1176         /**
1177          * Given HTML input, escape with htmlspecialchars but un-escape entities.
1178          * This allows (generally harmless) entities like &#160; to survive.
1179          *
1180          * @param string $html HTML to escape
1181          * @return string Escaped input
1182          */
1183         static function escapeHtmlAllowEntities( $html ) {
1184                 $html = Sanitizer::decodeCharReferences( $html );
1185                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1186                 # hurt.
1187                 $html = htmlspecialchars( $html, ENT_QUOTES );
1188                 return $html;
1189         }
1190
1191         /**
1192          * Regex replace callback for armoring links against further processing.
1193          * @param array $matches
1194          * @return string
1195          */
1196         private static function armorLinksCallback( $matches ) {
1197                 return str_replace( ':', '&#58;', $matches[1] );
1198         }
1199
1200         /**
1201          * Return an associative array of attribute names and values from
1202          * a partial tag string. Attribute names are forces to lowercase,
1203          * character references are decoded to UTF-8 text.
1204          *
1205          * @param string $text
1206          * @return array
1207          */
1208         public static function decodeTagAttributes( $text ) {
1209                 if ( trim( $text ) == '' ) {
1210                         return array();
1211                 }
1212
1213                 $attribs = array();
1214                 $pairs = array();
1215                 if ( !preg_match_all(
1216                         self::getAttribsRegex(),
1217                         $text,
1218                         $pairs,
1219                         PREG_SET_ORDER ) ) {
1220                         return $attribs;
1221                 }
1222
1223                 foreach ( $pairs as $set ) {
1224                         $attribute = strtolower( $set[1] );
1225                         $value = Sanitizer::getTagAttributeCallback( $set );
1226
1227                         // Normalize whitespace
1228                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1229                         $value = trim( $value );
1230
1231                         // Decode character references
1232                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1233                 }
1234                 return $attribs;
1235         }
1236
1237         /**
1238          * Build a partial tag string from an associative array of attribute
1239          * names and values as returned by decodeTagAttributes.
1240          *
1241          * @param array $assoc_array
1242          * @return string
1243          */
1244         public static function safeEncodeTagAttributes( $assoc_array ) {
1245                 $attribs = array();
1246                 foreach ( $assoc_array as $attribute => $value ) {
1247                         $encAttribute = htmlspecialchars( $attribute );
1248                         $encValue = Sanitizer::safeEncodeAttribute( $value );
1249
1250                         $attribs[] = "$encAttribute=\"$encValue\"";
1251                 }
1252                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1253         }
1254
1255         /**
1256          * Pick the appropriate attribute value from a match set from the
1257          * attribs regex matches.
1258          *
1259          * @param array $set
1260          * @throws MWException When tag conditions are not met.
1261          * @return string
1262          */
1263         private static function getTagAttributeCallback( $set ) {
1264                 if ( isset( $set[5] ) ) {
1265                         # No quotes.
1266                         return $set[5];
1267                 } elseif ( isset( $set[4] ) ) {
1268                         # Single-quoted
1269                         return $set[4];
1270                 } elseif ( isset( $set[3] ) ) {
1271                         # Double-quoted
1272                         return $set[3];
1273                 } elseif ( !isset( $set[2] ) ) {
1274                         # In XHTML, attributes must have a value so return an empty string.
1275                         # See "Empty attribute syntax",
1276                         # http://www.w3.org/TR/html5/syntax.html#syntax-attribute-name
1277                         return "";
1278                 } else {
1279                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1280                 }
1281         }
1282
1283         /**
1284          * @param string $text
1285          * @return string
1286          */
1287         private static function normalizeWhitespace( $text ) {
1288                 return preg_replace(
1289                         '/\r\n|[\x20\x0d\x0a\x09]/',
1290                         ' ',
1291                         $text );
1292         }
1293
1294         /**
1295          * Normalizes whitespace in a section name, such as might be returned
1296          * by Parser::stripSectionName(), for use in the id's that are used for
1297          * section links.
1298          *
1299          * @param string $section
1300          * @return string
1301          */
1302         static function normalizeSectionNameWhitespace( $section ) {
1303                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1304         }
1305
1306         /**
1307          * Ensure that any entities and character references are legal
1308          * for XML and XHTML specifically. Any stray bits will be
1309          * &amp;-escaped to result in a valid text fragment.
1310          *
1311          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1312          *   numericized (this way we're well-formed even without a DTD)
1313          * b. any numeric char refs must be legal chars, not invalid or forbidden
1314          * c. use lower cased "&#x", not "&#X"
1315          * d. fix or reject non-valid attributes
1316          *
1317          * @param string $text
1318          * @return string
1319          * @private
1320          */
1321         static function normalizeCharReferences( $text ) {
1322                 return preg_replace_callback(
1323                         self::CHAR_REFS_REGEX,
1324                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1325                         $text );
1326         }
1327
1328         /**
1329          * @param string $matches
1330          * @return string
1331          */
1332         static function normalizeCharReferencesCallback( $matches ) {
1333                 $ret = null;
1334                 if ( $matches[1] != '' ) {
1335                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1336                 } elseif ( $matches[2] != '' ) {
1337                         $ret = Sanitizer::decCharReference( $matches[2] );
1338                 } elseif ( $matches[3] != '' ) {
1339                         $ret = Sanitizer::hexCharReference( $matches[3] );
1340                 }
1341                 if ( is_null( $ret ) ) {
1342                         return htmlspecialchars( $matches[0] );
1343                 } else {
1344                         return $ret;
1345                 }
1346         }
1347
1348         /**
1349          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1350          * return the equivalent numeric entity reference (except for the core &lt;
1351          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1352          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1353          * pseudo-entity source (eg &amp;foo;)
1354          *
1355          * @param string $name
1356          * @return string
1357          */
1358         static function normalizeEntity( $name ) {
1359                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1360                         return '&' . self::$htmlEntityAliases[$name] . ';';
1361                 } elseif ( in_array( $name, array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1362                         return "&$name;";
1363                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1364                         return '&#' . self::$htmlEntities[$name] . ';';
1365                 } else {
1366                         return "&amp;$name;";
1367                 }
1368         }
1369
1370         /**
1371          * @param int $codepoint
1372          * @return null|string
1373          */
1374         static function decCharReference( $codepoint ) {
1375                 $point = intval( $codepoint );
1376                 if ( Sanitizer::validateCodepoint( $point ) ) {
1377                         return sprintf( '&#%d;', $point );
1378                 } else {
1379                         return null;
1380                 }
1381         }
1382
1383         /**
1384          * @param int $codepoint
1385          * @return null|string
1386          */
1387         static function hexCharReference( $codepoint ) {
1388                 $point = hexdec( $codepoint );
1389                 if ( Sanitizer::validateCodepoint( $point ) ) {
1390                         return sprintf( '&#x%x;', $point );
1391                 } else {
1392                         return null;
1393                 }
1394         }
1395
1396         /**
1397          * Returns true if a given Unicode codepoint is a valid character in
1398          * both HTML5 and XML.
1399          * @param int $codepoint
1400          * @return bool
1401          */
1402         private static function validateCodepoint( $codepoint ) {
1403                 # U+000C is valid in HTML5 but not allowed in XML.
1404                 # U+000D is valid in XML but not allowed in HTML5.
1405                 # U+007F - U+009F are disallowed in HTML5 (control characters).
1406                 return $codepoint == 0x09
1407                         || $codepoint == 0x0a
1408                         || ( $codepoint >= 0x20 && $codepoint <= 0x7e )
1409                         || ( $codepoint >= 0xa0 && $codepoint <= 0xd7ff )
1410                         || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1411                         || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1412         }
1413
1414         /**
1415          * Decode any character references, numeric or named entities,
1416          * in the text and return a UTF-8 string.
1417          *
1418          * @param string $text
1419          * @return string
1420          */
1421         public static function decodeCharReferences( $text ) {
1422                 return preg_replace_callback(
1423                         self::CHAR_REFS_REGEX,
1424                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1425                         $text );
1426         }
1427
1428         /**
1429          * Decode any character references, numeric or named entities,
1430          * in the next and normalize the resulting string. (bug 14952)
1431          *
1432          * This is useful for page titles, not for text to be displayed,
1433          * MediaWiki allows HTML entities to escape normalization as a feature.
1434          *
1435          * @param string $text Already normalized, containing entities
1436          * @return string Still normalized, without entities
1437          */
1438         public static function decodeCharReferencesAndNormalize( $text ) {
1439                 global $wgContLang;
1440                 $text = preg_replace_callback(
1441                         self::CHAR_REFS_REGEX,
1442                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1443                         $text, /* limit */ -1, $count );
1444
1445                 if ( $count ) {
1446                         return $wgContLang->normalize( $text );
1447                 } else {
1448                         return $text;
1449                 }
1450         }
1451
1452         /**
1453          * @param string $matches
1454          * @return string
1455          */
1456         static function decodeCharReferencesCallback( $matches ) {
1457                 if ( $matches[1] != '' ) {
1458                         return Sanitizer::decodeEntity( $matches[1] );
1459                 } elseif ( $matches[2] != '' ) {
1460                         return Sanitizer::decodeChar( intval( $matches[2] ) );
1461                 } elseif ( $matches[3] != '' ) {
1462                         return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1463                 }
1464                 # Last case should be an ampersand by itself
1465                 return $matches[0];
1466         }
1467
1468         /**
1469          * Return UTF-8 string for a codepoint if that is a valid
1470          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1471          * @param int $codepoint
1472          * @return string
1473          * @private
1474          */
1475         static function decodeChar( $codepoint ) {
1476                 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1477                         return UtfNormal\Utils::codepointToUtf8( $codepoint );
1478                 } else {
1479                         return UtfNormal\Constants::UTF8_REPLACEMENT;
1480                 }
1481         }
1482
1483         /**
1484          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1485          * return the UTF-8 encoding of that character. Otherwise, returns
1486          * pseudo-entity source (eg "&foo;")
1487          *
1488          * @param string $name
1489          * @return string
1490          */
1491         static function decodeEntity( $name ) {
1492                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1493                         $name = self::$htmlEntityAliases[$name];
1494                 }
1495                 if ( isset( self::$htmlEntities[$name] ) ) {
1496                         return UtfNormal\Utils::codepointToUtf8( self::$htmlEntities[$name] );
1497                 } else {
1498                         return "&$name;";
1499                 }
1500         }
1501
1502         /**
1503          * Fetch the whitelist of acceptable attributes for a given element name.
1504          *
1505          * @param string $element
1506          * @return array
1507          */
1508         static function attributeWhitelist( $element ) {
1509                 $list = Sanitizer::setupAttributeWhitelist();
1510                 return isset( $list[$element] )
1511                         ? $list[$element]
1512                         : array();
1513         }
1514
1515         /**
1516          * Foreach array key (an allowed HTML element), return an array
1517          * of allowed attributes
1518          * @return array
1519          */
1520         static function setupAttributeWhitelist() {
1521                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
1522                 static $whitelist, $staticInitialised;
1523
1524                 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgAllowMicrodataAttributes' ) );
1525
1526                 if ( $whitelist !== null && $staticInitialised == $globalContext ) {
1527                         return $whitelist;
1528                 }
1529
1530                 $common = array(
1531                         # HTML
1532                         'id',
1533                         'class',
1534                         'style',
1535                         'lang',
1536                         'dir',
1537                         'title',
1538
1539                         # WAI-ARIA
1540                         'role',
1541                 );
1542
1543                 if ( $wgAllowRdfaAttributes ) {
1544                         # RDFa attributes as specified in section 9 of
1545                         # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1546                         $common = array_merge( $common, array(
1547                                 'about', 'property', 'resource', 'datatype', 'typeof',
1548                         ) );
1549                 }
1550
1551                 if ( $wgAllowMicrodataAttributes ) {
1552                         # add HTML5 microdata tags as specified by
1553                         # http://www.whatwg.org/html/microdata.html#the-microdata-model
1554                         $common = array_merge( $common, array(
1555                                 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1556                         ) );
1557                 }
1558
1559                 $block = array_merge( $common, array( 'align' ) );
1560                 $tablealign = array( 'align', 'valign' );
1561                 $tablecell = array(
1562                         'abbr',
1563                         'axis',
1564                         'headers',
1565                         'scope',
1566                         'rowspan',
1567                         'colspan',
1568                         'nowrap', # deprecated
1569                         'width', # deprecated
1570                         'height', # deprecated
1571                         'bgcolor', # deprecated
1572                 );
1573
1574                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1575                 # See: http://www.w3.org/TR/html4/
1576                 $whitelist = array(
1577                         # 7.5.4
1578                         'div'        => $block,
1579                         'center'     => $common, # deprecated
1580                         'span'       => $common,
1581
1582                         # 7.5.5
1583                         'h1'         => $block,
1584                         'h2'         => $block,
1585                         'h3'         => $block,
1586                         'h4'         => $block,
1587                         'h5'         => $block,
1588                         'h6'         => $block,
1589
1590                         # 7.5.6
1591                         # address
1592
1593                         # 8.2.4
1594                         'bdo'        => $common,
1595
1596                         # 9.2.1
1597                         'em'         => $common,
1598                         'strong'     => $common,
1599                         'cite'       => $common,
1600                         'dfn'        => $common,
1601                         'code'       => $common,
1602                         'samp'       => $common,
1603                         'kbd'        => $common,
1604                         'var'        => $common,
1605                         'abbr'       => $common,
1606                         # acronym
1607
1608                         # 9.2.2
1609                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1610                         'q'          => array_merge( $common, array( 'cite' ) ),
1611
1612                         # 9.2.3
1613                         'sub'        => $common,
1614                         'sup'        => $common,
1615
1616                         # 9.3.1
1617                         'p'          => $block,
1618
1619                         # 9.3.2
1620                         'br'         => array_merge( $common, array( 'clear' ) ),
1621
1622                         # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1623                         'wbr'        => $common,
1624
1625                         # 9.3.4
1626                         'pre'        => array_merge( $common, array( 'width' ) ),
1627
1628                         # 9.4
1629                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1630                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1631
1632                         # 10.2
1633                         'ul'         => array_merge( $common, array( 'type' ) ),
1634                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1635                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1636
1637                         # 10.3
1638                         'dl'         => $common,
1639                         'dd'         => $common,
1640                         'dt'         => $common,
1641
1642                         # 11.2.1
1643                         'table'      => array_merge( $common,
1644                                                                 array( 'summary', 'width', 'border', 'frame',
1645                                                                                 'rules', 'cellspacing', 'cellpadding',
1646                                                                                 'align', 'bgcolor',
1647                                                                 ) ),
1648
1649                         # 11.2.2
1650                         'caption'    => $block,
1651
1652                         # 11.2.3
1653                         'thead'      => $common,
1654                         'tfoot'      => $common,
1655                         'tbody'      => $common,
1656
1657                         # 11.2.4
1658                         'colgroup'   => array_merge( $common, array( 'span' ) ),
1659                         'col'        => array_merge( $common, array( 'span' ) ),
1660
1661                         # 11.2.5
1662                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1663
1664                         # 11.2.6
1665                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1666                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1667
1668                         # 12.2
1669                         # NOTE: <a> is not allowed directly, but the attrib
1670                         # whitelist is used from the Parser object
1671                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1672
1673                         # 13.2
1674                         # Not usually allowed, but may be used for extension-style hooks
1675                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1676                         # true
1677                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1678
1679                         # 15.2.1
1680                         'tt'         => $common,
1681                         'b'          => $common,
1682                         'i'          => $common,
1683                         'big'        => $common,
1684                         'small'      => $common,
1685                         'strike'     => $common,
1686                         's'          => $common,
1687                         'u'          => $common,
1688
1689                         # 15.2.2
1690                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1691                         # basefont
1692
1693                         # 15.3
1694                         'hr'         => array_merge( $common, array( 'width' ) ),
1695
1696                         # HTML Ruby annotation text module, simple ruby only.
1697                         # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
1698                         'ruby'       => $common,
1699                         # rbc
1700                         'rb'         => $common,
1701                         'rp'         => $common,
1702                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1703                         'rtc'         => $common,
1704
1705                         # MathML root element, where used for extensions
1706                         # 'title' may not be 100% valid here; it's XHTML
1707                         # http://www.w3.org/TR/REC-MathML/
1708                         'math'       => array( 'class', 'style', 'id', 'title' ),
1709
1710                         # HTML 5 section 4.6
1711                         'bdi' => $common,
1712
1713                         # HTML5 elements, defined by:
1714                         # http://www.whatwg.org/html/
1715                         'data' => array_merge( $common, array( 'value' ) ),
1716                         'time' => array_merge( $common, array( 'datetime' ) ),
1717                         'mark' => $common,
1718
1719                         // meta and link are only permitted by removeHTMLtags when Microdata
1720                         // is enabled so we don't bother adding a conditional to hide these
1721                         // Also meta and link are only valid in WikiText as Microdata elements
1722                         // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1723                         // So we don't bother including $common attributes that have no purpose.
1724                         'meta' => array( 'itemprop', 'content' ),
1725                         'link' => array( 'itemprop', 'href' ),
1726                 );
1727
1728                 $staticInitialised = $globalContext;
1729
1730                 return $whitelist;
1731         }
1732
1733         /**
1734          * Take a fragment of (potentially invalid) HTML and return
1735          * a version with any tags removed, encoded as plain text.
1736          *
1737          * Warning: this return value must be further escaped for literal
1738          * inclusion in HTML output as of 1.10!
1739          *
1740          * @param string $text HTML fragment
1741          * @return string
1742          */
1743         static function stripAllTags( $text ) {
1744                 # Actual <tags>
1745                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1746
1747                 # Normalize &entities and whitespace
1748                 $text = self::decodeCharReferences( $text );
1749                 $text = self::normalizeWhitespace( $text );
1750
1751                 return $text;
1752         }
1753
1754         /**
1755          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1756          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1757          * PHP 5.1 doesn't.
1758          *
1759          * Use for passing XHTML fragments to PHP's XML parsing functions
1760          *
1761          * @return string
1762          */
1763         static function hackDocType() {
1764                 $out = "<!DOCTYPE html [\n";
1765                 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1766                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1767                 }
1768                 $out .= "]>\n";
1769                 return $out;
1770         }
1771
1772         /**
1773          * @param string $url
1774          * @return mixed|string
1775          */
1776         static function cleanUrl( $url ) {
1777                 # Normalize any HTML entities in input. They will be
1778                 # re-escaped by makeExternalLink().
1779                 $url = Sanitizer::decodeCharReferences( $url );
1780
1781                 # Escape any control characters introduced by the above step
1782                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1783                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1784
1785                 # Validate hostname portion
1786                 $matches = array();
1787                 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1788                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1789
1790                         // Characters that will be ignored in IDNs.
1791                         // http://tools.ietf.org/html/3454#section-3.1
1792                         // Strip them before further processing so blacklists and such work.
1793                         $strip = "/
1794                                 \\s|          # general whitespace
1795                                 \xc2\xad|     # 00ad SOFT HYPHEN
1796                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1797                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1798                                 \xe2\x81\xa0| # 2060 WORD JOINER
1799                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1800                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1801                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1802                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1803                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1804                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1805                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1806                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1807                                 /xuD";
1808
1809                         $host = preg_replace( $strip, '', $host );
1810
1811                         // IPv6 host names are bracketed with [].  Url-decode these.
1812                         if ( substr_compare( "//%5B", $host, 0, 5 ) === 0 && preg_match( '!^//%5B([0-9A-Fa-f:.]+)%5D((:\d+)?)$!', $host, $matches ) ) {
1813                                 $host = '//[' . $matches[1] . ']' . $matches[2];
1814                         }
1815
1816                         // @todo FIXME: Validate hostnames here
1817
1818                         return $protocol . $host . $rest;
1819                 } else {
1820                         return $url;
1821                 }
1822         }
1823
1824         /**
1825          * @param array $matches
1826          * @return string
1827          */
1828         static function cleanUrlCallback( $matches ) {
1829                 return urlencode( $matches[0] );
1830         }
1831
1832         /**
1833          * Does a string look like an e-mail address?
1834          *
1835          * This validates an email address using an HTML5 specification found at:
1836          * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1837          * Which as of 2011-01-24 says:
1838          *
1839          *   A valid e-mail address is a string that matches the ABNF production
1840          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1841          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1842          *   3.5.
1843          *
1844          * This function is an implementation of the specification as requested in
1845          * bug 22449.
1846          *
1847          * Client-side forms will use the same standard validation rules via JS or
1848          * HTML 5 validation; additional restrictions can be enforced server-side
1849          * by extensions via the 'isValidEmailAddr' hook.
1850          *
1851          * Note that this validation doesn't 100% match RFC 2822, but is believed
1852          * to be liberal enough for wide use. Some invalid addresses will still
1853          * pass validation here.
1854          *
1855          * @since 1.18
1856          *
1857          * @param string $addr E-mail address
1858          * @return bool
1859          */
1860         public static function validateEmail( $addr ) {
1861                 $result = null;
1862                 if ( !Hooks::run( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1863                         return $result;
1864                 }
1865
1866                 // Please note strings below are enclosed in brackets [], this make the
1867                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1868                 // See bug 26948
1869                 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1870                 $rfc1034_ldh_str = "a-z0-9\\-";
1871
1872                 $html5_email_regexp = "/
1873                 ^                      # start of string
1874                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1875                 @                      # 'apostrophe'
1876                 [$rfc1034_ldh_str]+       # First domain part
1877                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1878                 $                      # End of string
1879                 /ix"; // case Insensitive, eXtended
1880
1881                 return (bool)preg_match( $html5_email_regexp, $addr );
1882         }
1883 }