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