includes/Sanitizer.php

   1 <?php
   2 /**
   3  * HTML sanitizer for %MediaWiki.
   4  *
   5  * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
   6  * https://www.mediawiki.org/
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License as published by
  10  * the Free Software Foundation; either version 2 of the License, or
  11  * (at your option) any later version.
  12  *
  13  * This program is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  16  * GNU General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU General Public License along
  19  * with this program; if not, write to the Free Software Foundation, Inc.,
  20  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  21  * http://www.gnu.org/copyleft/gpl.html
  22  *
  23  * @file
  24  * @ingroup Parser
  25  */
  26
  27 /**
  28  * HTML sanitizer for MediaWiki
  29  * @ingroup Parser
  30  */
  31 class Sanitizer {
  32         /**
  33          * Regular expression to match various types of character references in
  34          * Sanitizer::normalizeCharReferences and Sanitizer::decodeCharReferences
  35          */
  36         const CHAR_REFS_REGEX =
  37                 '/&([A-Za-z0-9\x80-\xff]+);
  38                  |&\#([0-9]+);
  39                  |&\#[xX]([0-9A-Fa-f]+);
  40                  |(&)/x';
  41
  42         /**
  43          * Acceptable tag name charset from HTML5 parsing spec
  44          * http://www.w3.org/TR/html5/syntax.html#tag-open-state
  45          */
  46         const ELEMENT_BITS_REGEX = '!^(/?)([A-Za-z][^\t\n\v />\0]*+)([^>]*?)(/?>)([^<]*)$!';
  47
  48         /**
  49          * Blacklist for evil uris like javascript:
  50          * WARNING: DO NOT use this in any place that actually requires blacklisting
  51          * for security reasons. There are NUMEROUS[1] ways to bypass blacklisting, the
  52          * only way to be secure from javascript: uri based xss vectors is to whitelist
  53          * things that you know are safe and deny everything else.
  54          * [1]: http://ha.ckers.org/xss.html
  55          */
  56         const EVIL_URI_PATTERN = '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
  57         const XMLNS_ATTRIBUTE_PATTERN = "/^xmlns:[:A-Z_a-z-.0-9]+$/";
  58
  59         /**
  60          * List of all named character entities defined in HTML 4.01
  61          * http://www.w3.org/TR/html4/sgml/entities.html
  62          * As well as &apos; which is only defined starting in XHTML1.
  63          */
  64         private static $htmlEntities = array(
  65                 'Aacute'   => 193,
  66                 'aacute'   => 225,
  67                 'Acirc'    => 194,
  68                 'acirc'    => 226,
  69                 'acute'    => 180,
  70                 'AElig'    => 198,
  71                 'aelig'    => 230,
  72                 'Agrave'   => 192,
  73                 'agrave'   => 224,
  74                 'alefsym'  => 8501,
  75                 'Alpha'    => 913,
  76                 'alpha'    => 945,
  77                 'amp'      => 38,
  78                 'and'      => 8743,
  79                 'ang'      => 8736,
  80                 'apos'     => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
  81                 'Aring'    => 197,
  82                 'aring'    => 229,
  83                 'asymp'    => 8776,
  84                 'Atilde'   => 195,
  85                 'atilde'   => 227,
  86                 'Auml'     => 196,
  87                 'auml'     => 228,
  88                 'bdquo'    => 8222,
  89                 'Beta'     => 914,
  90                 'beta'     => 946,
  91                 'brvbar'   => 166,
  92                 'bull'     => 8226,
  93                 'cap'      => 8745,
  94                 'Ccedil'   => 199,
  95                 'ccedil'   => 231,
  96                 'cedil'    => 184,
  97                 'cent'     => 162,
  98                 'Chi'      => 935,
  99                 'chi'      => 967,
 100                 'circ'     => 710,
 101                 'clubs'    => 9827,
 102                 'cong'     => 8773,
 103                 'copy'     => 169,
 104                 'crarr'    => 8629,
 105                 'cup'      => 8746,
 106                 'curren'   => 164,
 107                 'dagger'   => 8224,
 108                 'Dagger'   => 8225,
 109                 'darr'     => 8595,
 110                 'dArr'     => 8659,
 111                 'deg'      => 176,
 112                 'Delta'    => 916,
 113                 'delta'    => 948,
 114                 'diams'    => 9830,
 115                 'divide'   => 247,
 116                 'Eacute'   => 201,
 117                 'eacute'   => 233,
 118                 'Ecirc'    => 202,
 119                 'ecirc'    => 234,
 120                 'Egrave'   => 200,
 121                 'egrave'   => 232,
 122                 'empty'    => 8709,
 123                 'emsp'     => 8195,
 124                 'ensp'     => 8194,
 125                 'Epsilon'  => 917,
 126                 'epsilon'  => 949,
 127                 'equiv'    => 8801,
 128                 'Eta'      => 919,
 129                 'eta'      => 951,
 130                 'ETH'      => 208,
 131                 'eth'      => 240,
 132                 'Euml'     => 203,
 133                 'euml'     => 235,
 134                 'euro'     => 8364,
 135                 'exist'    => 8707,
 136                 'fnof'     => 402,
 137                 'forall'   => 8704,
 138                 'frac12'   => 189,
 139                 'frac14'   => 188,
 140                 'frac34'   => 190,
 141                 'frasl'    => 8260,
 142                 'Gamma'    => 915,
 143                 'gamma'    => 947,
 144                 'ge'       => 8805,
 145                 'gt'       => 62,
 146                 'harr'     => 8596,
 147                 'hArr'     => 8660,
 148                 'hearts'   => 9829,
 149                 'hellip'   => 8230,
 150                 'Iacute'   => 205,
 151                 'iacute'   => 237,
 152                 'Icirc'    => 206,
 153                 'icirc'    => 238,
 154                 'iexcl'    => 161,
 155                 'Igrave'   => 204,
 156                 'igrave'   => 236,
 157                 'image'    => 8465,
 158                 'infin'    => 8734,
 159                 'int'      => 8747,
 160                 'Iota'     => 921,
 161                 'iota'     => 953,
 162                 'iquest'   => 191,
 163                 'isin'     => 8712,
 164                 'Iuml'     => 207,
 165                 'iuml'     => 239,
 166                 'Kappa'    => 922,
 167                 'kappa'    => 954,
 168                 'Lambda'   => 923,
 169                 'lambda'   => 955,
 170                 'lang'     => 9001,
 171                 'laquo'    => 171,
 172                 'larr'     => 8592,
 173                 'lArr'     => 8656,
 174                 'lceil'    => 8968,
 175                 'ldquo'    => 8220,
 176                 'le'       => 8804,
 177                 'lfloor'   => 8970,
 178                 'lowast'   => 8727,
 179                 'loz'      => 9674,
 180                 'lrm'      => 8206,
 181                 'lsaquo'   => 8249,
 182                 'lsquo'    => 8216,
 183                 'lt'       => 60,
 184                 'macr'     => 175,
 185                 'mdash'    => 8212,
 186                 'micro'    => 181,
 187                 'middot'   => 183,
 188                 'minus'    => 8722,
 189                 'Mu'       => 924,
 190                 'mu'       => 956,
 191                 'nabla'    => 8711,
 192                 'nbsp'     => 160,
 193                 'ndash'    => 8211,
 194                 'ne'       => 8800,
 195                 'ni'       => 8715,
 196                 'not'      => 172,
 197                 'notin'    => 8713,
 198                 'nsub'     => 8836,
 199                 'Ntilde'   => 209,
 200                 'ntilde'   => 241,
 201                 'Nu'       => 925,
 202                 'nu'       => 957,
 203                 'Oacute'   => 211,
 204                 'oacute'   => 243,
 205                 'Ocirc'    => 212,
 206                 'ocirc'    => 244,
 207                 'OElig'    => 338,
 208                 'oelig'    => 339,
 209                 'Ograve'   => 210,
 210                 'ograve'   => 242,
 211                 'oline'    => 8254,
 212                 'Omega'    => 937,
 213                 'omega'    => 969,
 214                 'Omicron'  => 927,
 215                 'omicron'  => 959,
 216                 'oplus'    => 8853,
 217                 'or'       => 8744,
 218                 'ordf'     => 170,
 219                 'ordm'     => 186,
 220                 'Oslash'   => 216,
 221                 'oslash'   => 248,
 222                 'Otilde'   => 213,
 223                 'otilde'   => 245,
 224                 'otimes'   => 8855,
 225                 'Ouml'     => 214,
 226                 'ouml'     => 246,
 227                 'para'     => 182,
 228                 'part'     => 8706,
 229                 'permil'   => 8240,
 230                 'perp'     => 8869,
 231                 'Phi'      => 934,
 232                 'phi'      => 966,
 233                 'Pi'       => 928,
 234                 'pi'       => 960,
 235                 'piv'      => 982,
 236                 'plusmn'   => 177,
 237                 'pound'    => 163,
 238                 'prime'    => 8242,
 239                 'Prime'    => 8243,
 240                 'prod'     => 8719,
 241                 'prop'     => 8733,
 242                 'Psi'      => 936,
 243                 'psi'      => 968,
 244                 'quot'     => 34,
 245                 'radic'    => 8730,
 246                 'rang'     => 9002,
 247                 'raquo'    => 187,
 248                 'rarr'     => 8594,
 249                 'rArr'     => 8658,
 250                 'rceil'    => 8969,
 251                 'rdquo'    => 8221,
 252                 'real'     => 8476,
 253                 'reg'      => 174,
 254                 'rfloor'   => 8971,
 255                 'Rho'      => 929,
 256                 'rho'      => 961,
 257                 'rlm'      => 8207,
 258                 'rsaquo'   => 8250,
 259                 'rsquo'    => 8217,
 260                 'sbquo'    => 8218,
 261                 'Scaron'   => 352,
 262                 'scaron'   => 353,
 263                 'sdot'     => 8901,
 264                 'sect'     => 167,
 265                 'shy'      => 173,
 266                 'Sigma'    => 931,
 267                 'sigma'    => 963,
 268                 'sigmaf'   => 962,
 269                 'sim'      => 8764,
 270                 'spades'   => 9824,
 271                 'sub'      => 8834,
 272                 'sube'     => 8838,
 273                 'sum'      => 8721,
 274                 'sup'      => 8835,
 275                 'sup1'     => 185,
 276                 'sup2'     => 178,
 277                 'sup3'     => 179,
 278                 'supe'     => 8839,
 279                 'szlig'    => 223,
 280                 'Tau'      => 932,
 281                 'tau'      => 964,
 282                 'there4'   => 8756,
 283                 'Theta'    => 920,
 284                 'theta'    => 952,
 285                 'thetasym' => 977,
 286                 'thinsp'   => 8201,
 287                 'THORN'    => 222,
 288                 'thorn'    => 254,
 289                 'tilde'    => 732,
 290                 'times'    => 215,
 291                 'trade'    => 8482,
 292                 'Uacute'   => 218,
 293                 'uacute'   => 250,
 294                 'uarr'     => 8593,
 295                 'uArr'     => 8657,
 296                 'Ucirc'    => 219,
 297                 'ucirc'    => 251,
 298                 'Ugrave'   => 217,
 299                 'ugrave'   => 249,
 300                 'uml'      => 168,
 301                 'upsih'    => 978,
 302                 'Upsilon'  => 933,
 303                 'upsilon'  => 965,
 304                 'Uuml'     => 220,
 305                 'uuml'     => 252,
 306                 'weierp'   => 8472,
 307                 'Xi'       => 926,
 308                 'xi'       => 958,
 309                 'Yacute'   => 221,
 310                 'yacute'   => 253,
 311                 'yen'      => 165,
 312                 'Yuml'     => 376,
 313                 'yuml'     => 255,
 314                 'Zeta'     => 918,
 315                 'zeta'     => 950,
 316                 'zwj'      => 8205,
 317                 'zwnj'     => 8204
 318         );
 319
 320         /**
 321          * Character entity aliases accepted by MediaWiki
 322          */
 323         private static $htmlEntityAliases = array(
 324                 'רלמ' => 'rlm',
 325                 'رلم' => 'rlm',
 326         );
 327
 328         /**
 329          * Lazy-initialised attributes regex, see getAttribsRegex()
 330          */
 331         private static $attribsRegex;
 332
 333         /**
 334          * Regular expression to match HTML/XML attribute pairs within a tag.
 335          * Allows some... latitude.
 336          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 337          * @return string
 338          */
 339         static function getAttribsRegex() {
 340                 if ( self::$attribsRegex === null ) {
 341                         $attribFirst = '[:A-Z_a-z0-9]';
 342                         $attrib = '[:A-Z_a-z-.0-9]';
 343                         $space = '[\x09\x0a\x0d\x20]';
 344                         self::$attribsRegex =
 345                                 "/(?:^|$space)({$attribFirst}{$attrib}*)
 346                                   ($space*=$space*
 347                                         (?:
 348                                          # The attribute value: quoted or alone
 349                                           \"([^<\"]*)(?:\"|\$)
 350                                          | '([^<']*)(?:'|\$)
 351                                          |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
 352                                         )
 353                                 )?(?=$space|\$)/sx";
 354                 }
 355                 return self::$attribsRegex;
 356         }
 357
 358         /**
 359          * Return the various lists of recognized tags
 360          * @param array $extratags For any extra tags to include
 361          * @param array $removetags For any tags (default or extra) to exclude
 362          * @return array
 363          */
 364         public static function getRecognizedTagData( $extratags = array(), $removetags = array() ) {
 365                 global $wgAllowMicrodataAttributes, $wgAllowImageTag;
 366
 367                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 368                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 369
 370                 // Base our staticInitialised variable off of the global config state so that if the globals
 371                 // are changed (like in the screwed up test system) we will re-initialise the settings.
 372                 $globalContext = implode( '-', compact( 'wgAllowMicrodataAttributes', 'wgAllowImageTag' ) );
 373                 if ( !$staticInitialised || $staticInitialised != $globalContext ) {
 374                         $htmlpairsStatic = array( # Tags that must be closed
 375                                 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 376                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 377                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 378                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 379                                 'ruby', 'rb', 'rp', 'rt', 'rtc', 'p', 'span', 'abbr', 'dfn',
 380                                 'kbd', 'samp', 'data', 'time', 'mark'
 381                         );
 382                         $htmlsingle = array(
 383                                 'br', 'wbr', 'hr', 'li', 'dt', 'dd'
 384                         );
 385                         $htmlsingleonly = array( # Elements that cannot have close tags
 386                                 'br', 'wbr', 'hr'
 387                         );
 388                         if ( $wgAllowMicrodataAttributes ) {
 389                                 $htmlsingle[] = $htmlsingleonly[] = 'meta';
 390                                 $htmlsingle[] = $htmlsingleonly[] = 'link';
 391                         }
 392                         $htmlnest = array( # Tags that can be nested--??
 393                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 394                                 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span',
 395                                 'var', 'kbd', 'samp', 'em', 'strong', 'q', 'ruby', 'bdo'
 396                         );
 397                         $tabletags = array( # Can only appear inside table, we will close them
 398                                 'td', 'th', 'tr',
 399                         );
 400                         $htmllist = array( # Tags used by list
 401                                 'ul', 'ol',
 402                         );
 403                         $listtags = array( # Tags that can appear in a list
 404                                 'li',
 405                         );
 406
 407                         if ( $wgAllowImageTag ) {
 408                                 $htmlsingle[] = 'img';
 409                                 $htmlsingleonly[] = 'img';
 410                         }
 411
 412                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 413                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 414
 415                         # Convert them all to hashtables for faster lookup
 416                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 417                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 418                         foreach ( $vars as $var ) {
 419                                 $$var = array_flip( $$var );
 420                         }
 421                         $staticInitialised = $globalContext;
 422                 }
 423
 424                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 425                 $extratags = array_flip( $extratags );
 426                 $removetags = array_flip( $removetags );
 427                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 428                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ), $removetags );
 429
 430                 return array(
 431                         'htmlpairs' => $htmlpairs,
 432                         'htmlsingle' => $htmlsingle,
 433                         'htmlsingleonly' => $htmlsingleonly,
 434                         'htmlnest' => $htmlnest,
 435                         'tabletags' => $tabletags,
 436                         'htmllist' => $htmllist,
 437                         'listtags' => $listtags,
 438                         'htmlsingleallowed' => $htmlsingleallowed,
 439                         'htmlelements' => $htmlelements,
 440                 );
 441         }
 442
 443         /**
 444          * Cleans up HTML, removes dangerous tags and attributes, and
 445          * removes HTML comments
 446          * @param string $text
 447          * @param callable $processCallback Callback to do any variable or parameter
 448          *   replacements in HTML attribute values
 449          * @param array|bool $args Arguments for the processing callback
 450          * @param array $extratags For any extra tags to include
 451          * @param array $removetags For any tags (default or extra) to exclude
 452          * @return string
 453          */
 454         public static function removeHTMLtags( $text, $processCallback = null,
 455                 $args = array(), $extratags = array(), $removetags = array()
 456         ) {
 457                 extract( self::getRecognizedTagData( $extratags, $removetags ) );
 458
 459                 # Remove HTML comments
 460                 $text = Sanitizer::removeHTMLcomments( $text );
 461                 $bits = explode( '<', $text );
 462                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 463                 if ( !MWTidy::isEnabled() ) {
 464                         $tagstack = $tablestack = array();
 465                         foreach ( $bits as $x ) {
 466                                 $regs = array();
 467                                 # $slash: Does the current element start with a '/'?
 468                                 # $t: Current element name
 469                                 # $params: String between element name and >
 470                                 # $brace: Ending '>' or '/>'
 471                                 # $rest: Everything until the next element of $bits
 472                                 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 473                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 474                                 } else {
 475                                         $slash = $t = $params = $brace = $rest = null;
 476                                 }
 477
 478                                 $badtag = false;
 479                                 $t = strtolower( $t );
 480                                 if ( isset( $htmlelements[$t] ) ) {
 481                                         # Check our stack
 482                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 483                                                 $badtag = true;
 484                                         } elseif ( $slash ) {
 485                                                 # Closing a tag... is it the one we just opened?
 486                                                 MediaWiki\suppressWarnings();
 487                                                 $ot = array_pop( $tagstack );
 488                                                 MediaWiki\restoreWarnings();
 489
 490                                                 if ( $ot != $t ) {
 491                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 492                                                                 # Pop all elements with an optional close tag
 493                                                                 # and see if we find a match below them
 494                                                                 $optstack = array();
 495                                                                 array_push( $optstack, $ot );
 496                                                                 MediaWiki\suppressWarnings();
 497                                                                 $ot = array_pop( $tagstack );
 498                                                                 MediaWiki\restoreWarnings();
 499                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 500                                                                         array_push( $optstack, $ot );
 501                                                                         MediaWiki\suppressWarnings();
 502                                                                         $ot = array_pop( $tagstack );
 503                                                                         MediaWiki\restoreWarnings();
 504                                                                 }
 505                                                                 if ( $t != $ot ) {
 506                                                                         # No match. Push the optional elements back again
 507                                                                         $badtag = true;
 508                                                                         MediaWiki\suppressWarnings();
 509                                                                         $ot = array_pop( $optstack );
 510                                                                         MediaWiki\restoreWarnings();
 511                                                                         while ( $ot ) {
 512                                                                                 array_push( $tagstack, $ot );
 513                                                                                 MediaWiki\suppressWarnings();
 514                                                                                 $ot = array_pop( $optstack );
 515                                                                                 MediaWiki\restoreWarnings();
 516                                                                         }
 517                                                                 }
 518                                                         } else {
 519                                                                 MediaWiki\suppressWarnings();
 520                                                                 array_push( $tagstack, $ot );
 521                                                                 MediaWiki\restoreWarnings();
 522
 523                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 524                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 525                                                                         $badtag = true;
 526                                                                 }
 527                                                         }
 528                                                 } else {
 529                                                         if ( $t == 'table' ) {
 530                                                                 $tagstack = array_pop( $tablestack );
 531                                                         }
 532                                                 }
 533                                                 $newparams = '';
 534                                         } else {
 535                                                 # Keep track for later
 536                                                 if ( isset( $tabletags[$t] ) && !in_array( 'table', $tagstack ) ) {
 537                                                         $badtag = true;
 538                                                 } elseif ( in_array( $t, $tagstack ) && !isset( $htmlnest[$t] ) ) {
 539                                                         $badtag = true;
 540                                                 #  Is it a self closed htmlpair ? (bug 5487)
 541                                                 } elseif ( $brace == '/>' && isset( $htmlpairs[$t] ) ) {
 542                                                         $badtag = true;
 543                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 544                                                         # Hack to force empty tag for unclosable elements
 545                                                         $brace = '/>';
 546                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 547                                                         # Hack to not close $htmlsingle tags
 548                                                         $brace = null;
 549                                                         # Still need to push this optionally-closed tag to
 550                                                         # the tag stack so that we can match end tags
 551                                                         # instead of marking them as bad.
 552                                                         array_push( $tagstack, $t );
 553                                                 } elseif ( isset( $tabletags[$t] ) && in_array( $t, $tagstack ) ) {
 554                                                         // New table tag but forgot to close the previous one
 555                                                         $text .= "</$t>";
 556                                                 } else {
 557                                                         if ( $t == 'table' ) {
 558                                                                 array_push( $tablestack, $tagstack );
 559                                                                 $tagstack = array();
 560                                                         }
 561                                                         array_push( $tagstack, $t );
 562                                                 }
 563
 564                                                 # Replace any variables or template parameters with
 565                                                 # plaintext results.
 566                                                 if ( is_callable( $processCallback ) ) {
 567                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 568                                                 }
 569
 570                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 571                                                         $badtag = true;
 572                                                 }
 573
 574                                                 # Strip non-approved attributes from the tag
 575                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 576                                         }
 577                                         if ( !$badtag ) {
 578                                                 $rest = str_replace( '>', '&gt;', $rest );
 579                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 580                                                 $text .= "<$slash$t$newparams$close>$rest";
 581                                                 continue;
 582                                         }
 583                                 }
 584                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 585                         }
 586                         # Close off any remaining tags
 587                         while ( is_array( $tagstack ) && ( $t = array_pop( $tagstack ) ) ) {
 588                                 $text .= "</$t>\n";
 589                                 if ( $t == 'table' ) {
 590                                         $tagstack = array_pop( $tablestack );
 591                                 }
 592                         }
 593                 } else {
 594                         # this might be possible using tidy itself
 595                         foreach ( $bits as $x ) {
 596                                 if ( preg_match( self::ELEMENT_BITS_REGEX, $x, $regs ) ) {
 597                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 598
 599                                         $badtag = false;
 600                                         $t = strtolower( $t );
 601                                         if ( isset( $htmlelements[$t] ) ) {
 602                                                 if ( is_callable( $processCallback ) ) {
 603                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 604                                                 }
 605
 606                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 607                                                         $badtag = true;
 608                                                 }
 609
 610                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 611                                                 if ( !$badtag ) {
 612                                                         $rest = str_replace( '>', '&gt;', $rest );
 613                                                         $text .= "<$slash$t$newparams$brace$rest";
 614                                                         continue;
 615                                                 }
 616                                         }
 617                                 }
 618                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x );
 619                         }
 620                 }
 621                 return $text;
 622         }
 623
 624         /**
 625          * Remove '<!--', '-->', and everything between.
 626          * To avoid leaving blank lines, when a comment is both preceded
 627          * and followed by a newline (ignoring spaces), trim leading and
 628          * trailing spaces and one of the newlines.
 629          *
 630          * @param string $text
 631          * @return string
 632          */
 633         public static function removeHTMLcomments( $text ) {
 634                 while ( ( $start = strpos( $text, '<!--' ) ) !== false ) {
 635                         $end = strpos( $text, '-->', $start + 4 );
 636                         if ( $end === false ) {
 637                                 # Unterminated comment; bail out
 638                                 break;
 639                         }
 640
 641                         $end += 3;
 642
 643                         # Trim space and newline if the comment is both
 644                         # preceded and followed by a newline
 645                         $spaceStart = max( $start - 1, 0 );
 646                         $spaceLen = $end - $spaceStart;
 647                         while ( substr( $text, $spaceStart, 1 ) === ' ' && $spaceStart > 0 ) {
 648                                 $spaceStart--;
 649                                 $spaceLen++;
 650                         }
 651                         while ( substr( $text, $spaceStart + $spaceLen, 1 ) === ' ' ) {
 652                                 $spaceLen++;
 653                         }
 654                         if ( substr( $text, $spaceStart, 1 ) === "\n"
 655                                 && substr( $text, $spaceStart + $spaceLen, 1 ) === "\n" ) {
 656                                 # Remove the comment, leading and trailing
 657                                 # spaces, and leave only one newline.
 658                                 $text = substr_replace( $text, "\n", $spaceStart, $spaceLen + 1 );
 659                         } else {
 660                                 # Remove just the comment.
 661                                 $text = substr_replace( $text, '', $start, $end - $start );
 662                         }
 663                 }
 664                 return $text;
 665         }
 666
 667         /**
 668          * Takes attribute names and values for a tag and the tag name and
 669          * validates that the tag is allowed to be present.
 670          * This DOES NOT validate the attributes, nor does it validate the
 671          * tags themselves. This method only handles the special circumstances
 672          * where we may want to allow a tag within content but ONLY when it has
 673          * specific attributes set.
 674          *
 675          * @param string $params
 676          * @param string $element
 677          * @return bool
 678          */
 679         static function validateTag( $params, $element ) {
 680                 $params = Sanitizer::decodeTagAttributes( $params );
 681
 682                 if ( $element == 'meta' || $element == 'link' ) {
 683                         if ( !isset( $params['itemprop'] ) ) {
 684                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 685                                 return false;
 686                         }
 687                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 688                                 // <meta> must have a content="" for the itemprop
 689                                 return false;
 690                         }
 691                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 692                                 // <link> must have an associated href=""
 693                                 return false;
 694                         }
 695                 }
 696
 697                 return true;
 698         }
 699
 700         /**
 701          * Take an array of attribute names and values and normalize or discard
 702          * illegal values for the given element type.
 703          *
 704          * - Discards attributes not on a whitelist for the given element
 705          * - Unsafe style attributes are discarded
 706          * - Invalid id attributes are re-encoded
 707          *
 708          * @param array $attribs
 709          * @param string $element
 710          * @return array
 711          *
 712          * @todo Check for legal values where the DTD limits things.
 713          * @todo Check for unique id attribute :P
 714          */
 715         static function validateTagAttributes( $attribs, $element ) {
 716                 return Sanitizer::validateAttributes( $attribs,
 717                         Sanitizer::attributeWhitelist( $element ) );
 718         }
 719
 720         /**
 721          * Take an array of attribute names and values and normalize or discard
 722          * illegal values for the given whitelist.
 723          *
 724          * - Discards attributes not the given whitelist
 725          * - Unsafe style attributes are discarded
 726          * - Invalid id attributes are re-encoded
 727          *
 728          * @param array $attribs
 729          * @param array $whitelist List of allowed attribute names
 730          * @return array
 731          *
 732          * @todo Check for legal values where the DTD limits things.
 733          * @todo Check for unique id attribute :P
 734          */
 735         static function validateAttributes( $attribs, $whitelist ) {
 736                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
 737
 738                 $whitelist = array_flip( $whitelist );
 739                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 740
 741                 $out = array();
 742                 foreach ( $attribs as $attribute => $value ) {
 743                         # allow XML namespace declaration if RDFa is enabled
 744                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 745                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 746                                         $out[$attribute] = $value;
 747                                 }
 748
 749                                 continue;
 750                         }
 751
 752                         # Allow any attribute beginning with "data-"
 753                         # However:
 754                         # * data-ooui is reserved for ooui
 755                         # * data-mw and data-parsoid are reserved for parsoid
 756                         # * data-mw-<name here> is reserved for extensions (or core) if
 757                         #   they need to communicate some data to the client and want to be
 758                         #   sure that it isn't coming from an untrusted user.
 759                         # * Ensure that the attribute is not namespaced by banning
 760                         #   colons.
 761                         if ( !preg_match( '/^data-(?!ooui|mw|parsoid)[^:]*$/i', $attribute )
 762                                 && !isset( $whitelist[$attribute] )
 763                         ) {
 764                                 continue;
 765                         }
 766
 767                         # Strip javascript "expression" from stylesheets.
 768                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 769                         if ( $attribute == 'style' ) {
 770                                 $value = Sanitizer::checkCss( $value );
 771                         }
 772
 773                         if ( $attribute === 'id' ) {
 774                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 775                         }
 776
 777                         # WAI-ARIA
 778                         # http://www.w3.org/TR/wai-aria/
 779                         # http://www.whatwg.org/html/elements.html#wai-aria
 780                         # For now we only support role="presentation" until we work out what roles should be
 781                         # usable by content and we ensure that our code explicitly rejects patterns that
 782                         # violate HTML5's ARIA restrictions.
 783                         if ( $attribute === 'role' && $value !== 'presentation' ) {
 784                                 continue;
 785                         }
 786
 787                         // RDFa and microdata properties allow URLs, URIs and/or CURIs.
 788                         // Check them for sanity.
 789                         if ( $attribute === 'rel' || $attribute === 'rev'
 790                                 # RDFa
 791                                 || $attribute === 'about' || $attribute === 'property'
 792                                 || $attribute === 'resource' || $attribute === 'datatype'
 793                                 || $attribute === 'typeof'
 794                                 # HTML5 microdata
 795                                 || $attribute === 'itemid' || $attribute === 'itemprop'
 796                                 || $attribute === 'itemref' || $attribute === 'itemscope'
 797                                 || $attribute === 'itemtype'
 798                         ) {
 799                                 // Paranoia. Allow "simple" values but suppress javascript
 800                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 801                                         continue;
 802                                 }
 803                         }
 804
 805                         # NOTE: even though elements using href/src are not allowed directly, supply
 806                         #       validation code that can be used by tag hook handlers, etc
 807                         if ( $attribute === 'href' || $attribute === 'src' ) {
 808                                 if ( !preg_match( $hrefExp, $value ) ) {
 809                                         continue; // drop any href or src attributes not using an allowed protocol.
 810                                         // NOTE: this also drops all relative URLs
 811                                 }
 812                         }
 813
 814                         // If this attribute was previously set, override it.
 815                         // Output should only have one attribute of each name.
 816                         $out[$attribute] = $value;
 817                 }
 818
 819                 if ( $wgAllowMicrodataAttributes ) {
 820                         # itemtype, itemid, itemref don't make sense without itemscope
 821                         if ( !array_key_exists( 'itemscope', $out ) ) {
 822                                 unset( $out['itemtype'] );
 823                                 unset( $out['itemid'] );
 824                                 unset( $out['itemref'] );
 825                         }
 826                         # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 827                 }
 828                 return $out;
 829         }
 830
 831         /**
 832          * Merge two sets of HTML attributes.  Conflicting items in the second set
 833          * will override those in the first, except for 'class' attributes which
 834          * will be combined (if they're both strings).
 835          *
 836          * @todo implement merging for other attributes such as style
 837          * @param array $a
 838          * @param array $b
 839          * @return array
 840          */
 841         static function mergeAttributes( $a, $b ) {
 842                 $out = array_merge( $a, $b );
 843                 if ( isset( $a['class'] ) && isset( $b['class'] )
 844                         && is_string( $a['class'] ) && is_string( $b['class'] )
 845                         && $a['class'] !== $b['class']
 846                 ) {
 847                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 848                                 -1, PREG_SPLIT_NO_EMPTY );
 849                         $out['class'] = implode( ' ', array_unique( $classes ) );
 850                 }
 851                 return $out;
 852         }
 853
 854         /**
 855          * Normalize CSS into a format we can easily search for hostile input
 856          *  - decode character references
 857          *  - decode escape sequences
 858          *  - convert characters that IE6 interprets into ascii
 859          *  - remove comments, unless the entire value is one single comment
 860          * @param string $value the css string
 861          * @return string normalized css
 862          */
 863         public static function normalizeCss( $value ) {
 864
 865                 // Decode character references like &#123;
 866                 $value = Sanitizer::decodeCharReferences( $value );
 867
 868                 // Decode escape sequences and line continuation
 869                 // See the grammar in the CSS 2 spec, appendix D.
 870                 // This has to be done AFTER decoding character references.
 871                 // This means it isn't possible for this function to return
 872                 // unsanitized escape sequences. It is possible to manufacture
 873                 // input that contains character references that decode to
 874                 // escape sequences that decode to character references, but
 875                 // it's OK for the return value to contain character references
 876                 // because the caller is supposed to escape those anyway.
 877                 static $decodeRegex;
 878                 if ( !$decodeRegex ) {
 879                         $space = '[\\x20\\t\\r\\n\\f]';
 880                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 881                         $backslash = '\\\\';
 882                         $decodeRegex = "/ $backslash
 883                                 (?:
 884                                         ($nl) |  # 1. Line continuation
 885                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 886                                         (.) | # 3. backslash cancelling special meaning
 887                                         () | # 4. backslash at end of string
 888                                 )/xu";
 889                 }
 890                 $value = preg_replace_callback( $decodeRegex,
 891                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 892
 893                 // Normalize Halfwidth and Fullwidth Unicode block that IE6 might treat as ascii
 894                 $value = preg_replace_callback(
 895                         '/[！-［］-ｚ]/u', // U+FF01 to U+FF5A, excluding U+FF3C (bug 58088)
 896                         function ( $matches ) {
 897                                 $cp = UtfNormal\Utils::utf8ToCodepoint( $matches[0] );
 898                                 if ( $cp === false ) {
 899                                         return '';
 900                                 }
 901                                 return chr( $cp - 65248 ); // ASCII range \x21-\x7A
 902                         },
 903                         $value
 904                 );
 905
 906                 // Convert more characters IE6 might treat as ascii
 907                 // U+0280, U+0274, U+207F, U+029F, U+026A, U+207D, U+208D
 908                 $value = str_replace(
 909                         array( 'ʀ', 'ɴ', 'ⁿ', 'ʟ', 'ɪ', '⁽', '₍' ),
 910                         array( 'r', 'n', 'n', 'l', 'i', '(', '(' ),
 911                         $value
 912                 );
 913
 914                 // Let the value through if it's nothing but a single comment, to
 915                 // allow other functions which may reject it to pass some error
 916                 // message through.
 917                 if ( !preg_match( '! ^ \s* /\* [^*\\/]* \*/ \s* $ !x', $value ) ) {
 918                         // Remove any comments; IE gets token splitting wrong
 919                         // This must be done AFTER decoding character references and
 920                         // escape sequences, because those steps can introduce comments
 921                         // This step cannot introduce character references or escape
 922                         // sequences, because it replaces comments with spaces rather
 923                         // than removing them completely.
 924                         $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 925
 926                         // Remove anything after a comment-start token, to guard against
 927                         // incorrect client implementations.
 928                         $commentPos = strpos( $value, '/*' );
 929                         if ( $commentPos !== false ) {
 930                                 $value = substr( $value, 0, $commentPos );
 931                         }
 932                 }
 933
 934                 // S followed by repeat, iteration, or prolonged sound marks,
 935                 // which IE will treat as "ss"
 936                 $value = preg_replace(
 937                         '/s(?:
 938                                 \xE3\x80\xB1 | # U+3031
 939                                 \xE3\x82\x9D | # U+309D
 940                                 \xE3\x83\xBC | # U+30FC
 941                                 \xE3\x83\xBD | # U+30FD
 942                                 \xEF\xB9\xBC | # U+FE7C
 943                                 \xEF\xB9\xBD | # U+FE7D
 944                                 \xEF\xBD\xB0   # U+FF70
 945                         )/ix',
 946                         'ss',
 947                         $value
 948                 );
 949
 950                 return $value;
 951         }
 952
 953
 954         /**
 955          * Pick apart some CSS and check it for forbidden or unsafe structures.
 956          * Returns a sanitized string. This sanitized string will have
 957          * character references and escape sequences decoded and comments
 958          * stripped (unless it is itself one valid comment, in which case the value
 959          * will be passed through). If the input is just too evil, only a comment
 960          * complaining about evilness will be returned.
 961          *
 962          * Currently URL references, 'expression', 'tps' are forbidden.
 963          *
 964          * NOTE: Despite the fact that character references are decoded, the
 965          * returned string may contain character references given certain
 966          * clever input strings. These character references must
 967          * be escaped before the return value is embedded in HTML.
 968          *
 969          * @param string $value
 970          * @return string
 971          */
 972         static function checkCss( $value ) {
 973                 $value = self::normalizeCss( $value );
 974
 975                 // Reject problematic keywords and control characters
 976                 if ( preg_match( '/[\000-\010\013\016-\037\177]/', $value ) ||
 977                         strpos( $value, UtfNormal\Constants::UTF8_REPLACEMENT ) !== false ) {
 978                         return '/* invalid control char */';
 979                 } elseif ( preg_match(
 980                         '! expression
 981                                 | filter\s*:
 982                                 | accelerator\s*:
 983                                 | -o-link\s*:
 984                                 | -o-link-source\s*:
 985                                 | -o-replace\s*:
 986                                 | url\s*\(
 987                                 | image\s*\(
 988                                 | image-set\s*\(
 989                         !ix', $value ) ) {
 990                         return '/* insecure input */';
 991                 }
 992                 return $value;
 993         }
 994
 995         /**
 996          * @param array $matches
 997          * @return string
 998          */
 999         static function cssDecodeCallback( $matches ) {
1000                 if ( $matches[1] !== '' ) {
1001                         // Line continuation
1002                         return '';
1003                 } elseif ( $matches[2] !== '' ) {
1004                         $char = UtfNormal\Utils::codepointToUtf8( hexdec( $matches[2] ) );
1005                 } elseif ( $matches[3] !== '' ) {
1006                         $char = $matches[3];
1007                 } else {
1008                         $char = '\\';
1009                 }
1010                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
1011                         // These characters need to be escaped in strings
1012                         // Clean up the escape sequence to avoid parsing errors by clients
1013                         return '\\' . dechex( ord( $char ) ) . ' ';
1014                 } else {
1015                         // Decode unnecessary escape
1016                         return $char;
1017                 }
1018         }
1019
1020         /**
1021          * Take a tag soup fragment listing an HTML element's attributes
1022          * and normalize it to well-formed XML, discarding unwanted attributes.
1023          * Output is safe for further wikitext processing, with escaping of
1024          * values that could trigger problems.
1025          *
1026          * - Normalizes attribute names to lowercase
1027          * - Discards attributes not on a whitelist for the given element
1028          * - Turns broken or invalid entities into plaintext
1029          * - Double-quotes all attribute values
1030          * - Attributes without values are given the name as attribute
1031          * - Double attributes are discarded
1032          * - Unsafe style attributes are discarded
1033          * - Prepends space if there are attributes.
1034          *
1035          * @param string $text
1036          * @param string $element
1037          * @return string
1038          */
1039         static function fixTagAttributes( $text, $element ) {
1040                 if ( trim( $text ) == '' ) {
1041                         return '';
1042                 }
1043
1044                 $decoded = Sanitizer::decodeTagAttributes( $text );
1045                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1046
1047                 return Sanitizer::safeEncodeTagAttributes( $stripped );
1048         }
1049
1050         /**
1051          * Encode an attribute value for HTML output.
1052          * @param string $text
1053          * @return string HTML-encoded text fragment
1054          */
1055         static function encodeAttribute( $text ) {
1056                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1057
1058                 // Whitespace is normalized during attribute decoding,
1059                 // so if we've been passed non-spaces we must encode them
1060                 // ahead of time or they won't be preserved.
1061                 $encValue = strtr( $encValue, array(
1062                         "\n" => '&#10;',
1063                         "\r" => '&#13;',
1064                         "\t" => '&#9;',
1065                 ) );
1066
1067                 return $encValue;
1068         }
1069
1070         /**
1071          * Encode an attribute value for HTML tags, with extra armoring
1072          * against further wiki processing.
1073          * @param string $text
1074          * @return string HTML-encoded text fragment
1075          */
1076         static function safeEncodeAttribute( $text ) {
1077                 $encValue = Sanitizer::encodeAttribute( $text );
1078
1079                 # Templates and links may be expanded in later parsing,
1080                 # creating invalid or dangerous output. Suppress this.
1081                 $encValue = strtr( $encValue, array(
1082                         '<'    => '&lt;',   // This should never happen,
1083                         '>'    => '&gt;',   // we've received invalid input
1084                         '"'    => '&quot;', // which should have been escaped.
1085                         '{'    => '&#123;',
1086                         '['    => '&#91;',
1087                         "''"   => '&#39;&#39;',
1088                         'ISBN' => '&#73;SBN',
1089                         'RFC'  => '&#82;FC',
1090                         'PMID' => '&#80;MID',
1091                         '|'    => '&#124;',
1092                         '__'   => '&#95;_',
1093                 ) );
1094
1095                 # Stupid hack
1096                 $encValue = preg_replace_callback(
1097                         '/((?i)' . wfUrlProtocols() . ')/',
1098                         array( 'Sanitizer', 'armorLinksCallback' ),
1099                         $encValue );
1100                 return $encValue;
1101         }
1102
1103         /**
1104          * Given a value, escape it so that it can be used in an id attribute and
1105          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1106          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1107          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1108          * escaped with lots of dots.
1109          *
1110          * To ensure we don't have to bother escaping anything, we also strip ', ",
1111          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1112          * We also strip # because it upsets IE, and % because it could be
1113          * ambiguous if it's part of something that looks like a percent escape
1114          * (which don't work reliably in fragments cross-browser).
1115          *
1116          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1117          *   in the id and name attributes
1118          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with
1119          *   the id attribute
1120          * @see http://www.whatwg.org/html/elements.html#the-id-attribute
1121          *   HTML5 definition of id attribute
1122          *
1123          * @param string $id Id to escape
1124          * @param string|array $options String or array of strings (default is array()):
1125          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1126          *       so don't pay attention if the first character isn't valid at the
1127          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1128          *       false.
1129          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1130          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1131          *       anchors and links won't break.
1132          * @return string
1133          */
1134         static function escapeId( $id, $options = array() ) {
1135                 global $wgExperimentalHtmlIds;
1136                 $options = (array)$options;
1137
1138                 $id = Sanitizer::decodeCharReferences( $id );
1139
1140                 if ( $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1141                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1142                         $id = trim( $id, '_' );
1143                         if ( $id === '' ) {
1144                                 // Must have been all whitespace to start with.
1145                                 return '_';
1146                         } else {
1147                                 return $id;
1148                         }
1149                 }
1150
1151                 // HTML4-style escaping
1152                 static $replace = array(
1153                         '%3A' => ':',
1154                         '%' => '.'
1155                 );
1156
1157                 $id = urlencode( strtr( $id, ' ', '_' ) );
1158                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1159
1160                 if ( !preg_match( '/^[a-zA-Z]/', $id ) && !in_array( 'noninitial', $options ) ) {
1161                         // Initial character must be a letter!
1162                         $id = "x$id";
1163                 }
1164                 return $id;
1165         }
1166
1167         /**
1168          * Given a value, escape it so that it can be used as a CSS class and
1169          * return it.
1170          *
1171          * @todo For extra validity, input should be validated UTF-8.
1172          *
1173          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1174          *
1175          * @param string $class
1176          * @return string
1177          */
1178         static function escapeClass( $class ) {
1179                 // Convert ugly stuff to underscores and kill underscores in ugly places
1180                 return rtrim( preg_replace(
1181                         array( '/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/', '/_+/' ),
1182                         '_',
1183                         $class ), '_' );
1184         }
1185
1186         /**
1187          * Given HTML input, escape with htmlspecialchars but un-escape entities.
1188          * This allows (generally harmless) entities like &#160; to survive.
1189          *
1190          * @param string $html HTML to escape
1191          * @return string Escaped input
1192          */
1193         static function escapeHtmlAllowEntities( $html ) {
1194                 $html = Sanitizer::decodeCharReferences( $html );
1195                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1196                 # hurt.
1197                 $html = htmlspecialchars( $html, ENT_QUOTES );
1198                 return $html;
1199         }
1200
1201         /**
1202          * Regex replace callback for armoring links against further processing.
1203          * @param array $matches
1204          * @return string
1205          */
1206         private static function armorLinksCallback( $matches ) {
1207                 return str_replace( ':', '&#58;', $matches[1] );
1208         }
1209
1210         /**
1211          * Return an associative array of attribute names and values from
1212          * a partial tag string. Attribute names are forces to lowercase,
1213          * character references are decoded to UTF-8 text.
1214          *
1215          * @param string $text
1216          * @return array
1217          */
1218         public static function decodeTagAttributes( $text ) {
1219                 if ( trim( $text ) == '' ) {
1220                         return array();
1221                 }
1222
1223                 $attribs = array();
1224                 $pairs = array();
1225                 if ( !preg_match_all(
1226                         self::getAttribsRegex(),
1227                         $text,
1228                         $pairs,
1229                         PREG_SET_ORDER ) ) {
1230                         return $attribs;
1231                 }
1232
1233                 foreach ( $pairs as $set ) {
1234                         $attribute = strtolower( $set[1] );
1235                         $value = Sanitizer::getTagAttributeCallback( $set );
1236
1237                         // Normalize whitespace
1238                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1239                         $value = trim( $value );
1240
1241                         // Decode character references
1242                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1243                 }
1244                 return $attribs;
1245         }
1246
1247         /**
1248          * Build a partial tag string from an associative array of attribute
1249          * names and values as returned by decodeTagAttributes.
1250          *
1251          * @param array $assoc_array
1252          * @return string
1253          */
1254         public static function safeEncodeTagAttributes( $assoc_array ) {
1255                 $attribs = array();
1256                 foreach ( $assoc_array as $attribute => $value ) {
1257                         $encAttribute = htmlspecialchars( $attribute );
1258                         $encValue = Sanitizer::safeEncodeAttribute( $value );
1259
1260                         $attribs[] = "$encAttribute=\"$encValue\"";
1261                 }
1262                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1263         }
1264
1265         /**
1266          * Pick the appropriate attribute value from a match set from the
1267          * attribs regex matches.
1268          *
1269          * @param array $set
1270          * @throws MWException When tag conditions are not met.
1271          * @return string
1272          */
1273         private static function getTagAttributeCallback( $set ) {
1274                 if ( isset( $set[5] ) ) {
1275                         # No quotes.
1276                         return $set[5];
1277                 } elseif ( isset( $set[4] ) ) {
1278                         # Single-quoted
1279                         return $set[4];
1280                 } elseif ( isset( $set[3] ) ) {
1281                         # Double-quoted
1282                         return $set[3];
1283                 } elseif ( !isset( $set[2] ) ) {
1284                         # In XHTML, attributes must have a value so return an empty string.
1285                         # See "Empty attribute syntax",
1286                         # http://www.w3.org/TR/html5/syntax.html#syntax-attribute-name
1287                         return "";
1288                 } else {
1289                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1290                 }
1291         }
1292
1293         /**
1294          * @param string $text
1295          * @return string
1296          */
1297         private static function normalizeWhitespace( $text ) {
1298                 return preg_replace(
1299                         '/\r\n|[\x20\x0d\x0a\x09]/',
1300                         ' ',
1301                         $text );
1302         }
1303
1304         /**
1305          * Normalizes whitespace in a section name, such as might be returned
1306          * by Parser::stripSectionName(), for use in the id's that are used for
1307          * section links.
1308          *
1309          * @param string $section
1310          * @return string
1311          */
1312         static function normalizeSectionNameWhitespace( $section ) {
1313                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1314         }
1315
1316         /**
1317          * Ensure that any entities and character references are legal
1318          * for XML and XHTML specifically. Any stray bits will be
1319          * &amp;-escaped to result in a valid text fragment.
1320          *
1321          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1322          *   numericized (this way we're well-formed even without a DTD)
1323          * b. any numeric char refs must be legal chars, not invalid or forbidden
1324          * c. use lower cased "&#x", not "&#X"
1325          * d. fix or reject non-valid attributes
1326          *
1327          * @param string $text
1328          * @return string
1329          * @private
1330          */
1331         static function normalizeCharReferences( $text ) {
1332                 return preg_replace_callback(
1333                         self::CHAR_REFS_REGEX,
1334                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1335                         $text );
1336         }
1337
1338         /**
1339          * @param string $matches
1340          * @return string
1341          */
1342         static function normalizeCharReferencesCallback( $matches ) {
1343                 $ret = null;
1344                 if ( $matches[1] != '' ) {
1345                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1346                 } elseif ( $matches[2] != '' ) {
1347                         $ret = Sanitizer::decCharReference( $matches[2] );
1348                 } elseif ( $matches[3] != '' ) {
1349                         $ret = Sanitizer::hexCharReference( $matches[3] );
1350                 }
1351                 if ( is_null( $ret ) ) {
1352                         return htmlspecialchars( $matches[0] );
1353                 } else {
1354                         return $ret;
1355                 }
1356         }
1357
1358         /**
1359          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1360          * return the equivalent numeric entity reference (except for the core &lt;
1361          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1362          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1363          * pseudo-entity source (eg &amp;foo;)
1364          *
1365          * @param string $name
1366          * @return string
1367          */
1368         static function normalizeEntity( $name ) {
1369                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1370                         return '&' . self::$htmlEntityAliases[$name] . ';';
1371                 } elseif ( in_array( $name, array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1372                         return "&$name;";
1373                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1374                         return '&#' . self::$htmlEntities[$name] . ';';
1375                 } else {
1376                         return "&amp;$name;";
1377                 }
1378         }
1379
1380         /**
1381          * @param int $codepoint
1382          * @return null|string
1383          */
1384         static function decCharReference( $codepoint ) {
1385                 $point = intval( $codepoint );
1386                 if ( Sanitizer::validateCodepoint( $point ) ) {
1387                         return sprintf( '&#%d;', $point );
1388                 } else {
1389                         return null;
1390                 }
1391         }
1392
1393         /**
1394          * @param int $codepoint
1395          * @return null|string
1396          */
1397         static function hexCharReference( $codepoint ) {
1398                 $point = hexdec( $codepoint );
1399                 if ( Sanitizer::validateCodepoint( $point ) ) {
1400                         return sprintf( '&#x%x;', $point );
1401                 } else {
1402                         return null;
1403                 }
1404         }
1405
1406         /**
1407          * Returns true if a given Unicode codepoint is a valid character in
1408          * both HTML5 and XML.
1409          * @param int $codepoint
1410          * @return bool
1411          */
1412         private static function validateCodepoint( $codepoint ) {
1413                 # U+000C is valid in HTML5 but not allowed in XML.
1414                 # U+000D is valid in XML but not allowed in HTML5.
1415                 # U+007F - U+009F are disallowed in HTML5 (control characters).
1416                 return $codepoint == 0x09
1417                         || $codepoint == 0x0a
1418                         || ( $codepoint >= 0x20 && $codepoint <= 0x7e )
1419                         || ( $codepoint >= 0xa0 && $codepoint <= 0xd7ff )
1420                         || ( $codepoint >= 0xe000 && $codepoint <= 0xfffd )
1421                         || ( $codepoint >= 0x10000 && $codepoint <= 0x10ffff );
1422         }
1423
1424         /**
1425          * Decode any character references, numeric or named entities,
1426          * in the text and return a UTF-8 string.
1427          *
1428          * @param string $text
1429          * @return string
1430          */
1431         public static function decodeCharReferences( $text ) {
1432                 return preg_replace_callback(
1433                         self::CHAR_REFS_REGEX,
1434                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1435                         $text );
1436         }
1437
1438         /**
1439          * Decode any character references, numeric or named entities,
1440          * in the next and normalize the resulting string. (bug 14952)
1441          *
1442          * This is useful for page titles, not for text to be displayed,
1443          * MediaWiki allows HTML entities to escape normalization as a feature.
1444          *
1445          * @param string $text Already normalized, containing entities
1446          * @return string Still normalized, without entities
1447          */
1448         public static function decodeCharReferencesAndNormalize( $text ) {
1449                 global $wgContLang;
1450                 $text = preg_replace_callback(
1451                         self::CHAR_REFS_REGEX,
1452                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1453                         $text, /* limit */ -1, $count );
1454
1455                 if ( $count ) {
1456                         return $wgContLang->normalize( $text );
1457                 } else {
1458                         return $text;
1459                 }
1460         }
1461
1462         /**
1463          * @param string $matches
1464          * @return string
1465          */
1466         static function decodeCharReferencesCallback( $matches ) {
1467                 if ( $matches[1] != '' ) {
1468                         return Sanitizer::decodeEntity( $matches[1] );
1469                 } elseif ( $matches[2] != '' ) {
1470                         return Sanitizer::decodeChar( intval( $matches[2] ) );
1471                 } elseif ( $matches[3] != '' ) {
1472                         return Sanitizer::decodeChar( hexdec( $matches[3] ) );
1473                 }
1474                 # Last case should be an ampersand by itself
1475                 return $matches[0];
1476         }
1477
1478         /**
1479          * Return UTF-8 string for a codepoint if that is a valid
1480          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1481          * @param int $codepoint
1482          * @return string
1483          * @private
1484          */
1485         static function decodeChar( $codepoint ) {
1486                 if ( Sanitizer::validateCodepoint( $codepoint ) ) {
1487                         return UtfNormal\Utils::codepointToUtf8( $codepoint );
1488                 } else {
1489                         return UtfNormal\Constants::UTF8_REPLACEMENT;
1490                 }
1491         }
1492
1493         /**
1494          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1495          * return the UTF-8 encoding of that character. Otherwise, returns
1496          * pseudo-entity source (eg "&foo;")
1497          *
1498          * @param string $name
1499          * @return string
1500          */
1501         static function decodeEntity( $name ) {
1502                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1503                         $name = self::$htmlEntityAliases[$name];
1504                 }
1505                 if ( isset( self::$htmlEntities[$name] ) ) {
1506                         return UtfNormal\Utils::codepointToUtf8( self::$htmlEntities[$name] );
1507                 } else {
1508                         return "&$name;";
1509                 }
1510         }
1511
1512         /**
1513          * Fetch the whitelist of acceptable attributes for a given element name.
1514          *
1515          * @param string $element
1516          * @return array
1517          */
1518         static function attributeWhitelist( $element ) {
1519                 $list = Sanitizer::setupAttributeWhitelist();
1520                 return isset( $list[$element] )
1521                         ? $list[$element]
1522                         : array();
1523         }
1524
1525         /**
1526          * Foreach array key (an allowed HTML element), return an array
1527          * of allowed attributes
1528          * @return array
1529          */
1530         static function setupAttributeWhitelist() {
1531                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes;
1532                 static $whitelist, $staticInitialised;
1533
1534                 $globalContext = implode( '-', compact( 'wgAllowRdfaAttributes', 'wgAllowMicrodataAttributes' ) );
1535
1536                 if ( $whitelist !== null && $staticInitialised == $globalContext ) {
1537                         return $whitelist;
1538                 }
1539
1540                 $common = array(
1541                         # HTML
1542                         'id',
1543                         'class',
1544                         'style',
1545                         'lang',
1546                         'dir',
1547                         'title',
1548
1549                         # WAI-ARIA
1550                         'role',
1551                 );
1552
1553                 if ( $wgAllowRdfaAttributes ) {
1554                         # RDFa attributes as specified in section 9 of
1555                         # http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1556                         $common = array_merge( $common, array(
1557                                 'about', 'property', 'resource', 'datatype', 'typeof',
1558                         ) );
1559                 }
1560
1561                 if ( $wgAllowMicrodataAttributes ) {
1562                         # add HTML5 microdata tags as specified by
1563                         # http://www.whatwg.org/html/microdata.html#the-microdata-model
1564                         $common = array_merge( $common, array(
1565                                 'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1566                         ) );
1567                 }
1568
1569                 $block = array_merge( $common, array( 'align' ) );
1570                 $tablealign = array( 'align', 'valign' );
1571                 $tablecell = array(
1572                         'abbr',
1573                         'axis',
1574                         'headers',
1575                         'scope',
1576                         'rowspan',
1577                         'colspan',
1578                         'nowrap', # deprecated
1579                         'width', # deprecated
1580                         'height', # deprecated
1581                         'bgcolor', # deprecated
1582                 );
1583
1584                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1585                 # See: http://www.w3.org/TR/html4/
1586                 $whitelist = array(
1587                         # 7.5.4
1588                         'div'        => $block,
1589                         'center'     => $common, # deprecated
1590                         'span'       => $common,
1591
1592                         # 7.5.5
1593                         'h1'         => $block,
1594                         'h2'         => $block,
1595                         'h3'         => $block,
1596                         'h4'         => $block,
1597                         'h5'         => $block,
1598                         'h6'         => $block,
1599
1600                         # 7.5.6
1601                         # address
1602
1603                         # 8.2.4
1604                         'bdo'        => $common,
1605
1606                         # 9.2.1
1607                         'em'         => $common,
1608                         'strong'     => $common,
1609                         'cite'       => $common,
1610                         'dfn'        => $common,
1611                         'code'       => $common,
1612                         'samp'       => $common,
1613                         'kbd'        => $common,
1614                         'var'        => $common,
1615                         'abbr'       => $common,
1616                         # acronym
1617
1618                         # 9.2.2
1619                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1620                         'q'          => array_merge( $common, array( 'cite' ) ),
1621
1622                         # 9.2.3
1623                         'sub'        => $common,
1624                         'sup'        => $common,
1625
1626                         # 9.3.1
1627                         'p'          => $block,
1628
1629                         # 9.3.2
1630                         'br'         => array_merge( $common, array( 'clear' ) ),
1631
1632                         # http://www.whatwg.org/html/text-level-semantics.html#the-wbr-element
1633                         'wbr'        => $common,
1634
1635                         # 9.3.4
1636                         'pre'        => array_merge( $common, array( 'width' ) ),
1637
1638                         # 9.4
1639                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1640                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1641
1642                         # 10.2
1643                         'ul'         => array_merge( $common, array( 'type' ) ),
1644                         'ol'         => array_merge( $common, array( 'type', 'start', 'reversed' ) ),
1645                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1646
1647                         # 10.3
1648                         'dl'         => $common,
1649                         'dd'         => $common,
1650                         'dt'         => $common,
1651
1652                         # 11.2.1
1653                         'table'      => array_merge( $common,
1654                                                                 array( 'summary', 'width', 'border', 'frame',
1655                                                                                 'rules', 'cellspacing', 'cellpadding',
1656                                                                                 'align', 'bgcolor',
1657                                                                 ) ),
1658
1659                         # 11.2.2
1660                         'caption'    => $block,
1661
1662                         # 11.2.3
1663                         'thead'      => $common,
1664                         'tfoot'      => $common,
1665                         'tbody'      => $common,
1666
1667                         # 11.2.4
1668                         'colgroup'   => array_merge( $common, array( 'span' ) ),
1669                         'col'        => array_merge( $common, array( 'span' ) ),
1670
1671                         # 11.2.5
1672                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1673
1674                         # 11.2.6
1675                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1676                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1677
1678                         # 12.2
1679                         # NOTE: <a> is not allowed directly, but the attrib
1680                         # whitelist is used from the Parser object
1681                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1682
1683                         # 13.2
1684                         # Not usually allowed, but may be used for extension-style hooks
1685                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1686                         # true
1687                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1688
1689                         # 15.2.1
1690                         'tt'         => $common,
1691                         'b'          => $common,
1692                         'i'          => $common,
1693                         'big'        => $common,
1694                         'small'      => $common,
1695                         'strike'     => $common,
1696                         's'          => $common,
1697                         'u'          => $common,
1698
1699                         # 15.2.2
1700                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1701                         # basefont
1702
1703                         # 15.3
1704                         'hr'         => array_merge( $common, array( 'width' ) ),
1705
1706                         # HTML Ruby annotation text module, simple ruby only.
1707                         # http://www.whatwg.org/html/text-level-semantics.html#the-ruby-element
1708                         'ruby'       => $common,
1709                         # rbc
1710                         'rb'         => $common,
1711                         'rp'         => $common,
1712                         'rt'         => $common, # array_merge( $common, array( 'rbspan' ) ),
1713                         'rtc'         => $common,
1714
1715                         # MathML root element, where used for extensions
1716                         # 'title' may not be 100% valid here; it's XHTML
1717                         # http://www.w3.org/TR/REC-MathML/
1718                         'math'       => array( 'class', 'style', 'id', 'title' ),
1719
1720                         # HTML 5 section 4.6
1721                         'bdi' => $common,
1722
1723                         # HTML5 elements, defined by:
1724                         # http://www.whatwg.org/html/
1725                         'data' => array_merge( $common, array( 'value' ) ),
1726                         'time' => array_merge( $common, array( 'datetime' ) ),
1727                         'mark' => $common,
1728
1729                         // meta and link are only permitted by removeHTMLtags when Microdata
1730                         // is enabled so we don't bother adding a conditional to hide these
1731                         // Also meta and link are only valid in WikiText as Microdata elements
1732                         // (ie: validateTag rejects tags missing the attributes needed for Microdata)
1733                         // So we don't bother including $common attributes that have no purpose.
1734                         'meta' => array( 'itemprop', 'content' ),
1735                         'link' => array( 'itemprop', 'href' ),
1736                 );
1737
1738                 $staticInitialised = $globalContext;
1739
1740                 return $whitelist;
1741         }
1742
1743         /**
1744          * Take a fragment of (potentially invalid) HTML and return
1745          * a version with any tags removed, encoded as plain text.
1746          *
1747          * Warning: this return value must be further escaped for literal
1748          * inclusion in HTML output as of 1.10!
1749          *
1750          * @param string $text HTML fragment
1751          * @return string
1752          */
1753         static function stripAllTags( $text ) {
1754                 # Actual <tags>
1755                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1756
1757                 # Normalize &entities and whitespace
1758                 $text = self::decodeCharReferences( $text );
1759                 $text = self::normalizeWhitespace( $text );
1760
1761                 return $text;
1762         }
1763
1764         /**
1765          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1766          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1767          * PHP 5.1 doesn't.
1768          *
1769          * Use for passing XHTML fragments to PHP's XML parsing functions
1770          *
1771          * @return string
1772          */
1773         static function hackDocType() {
1774                 $out = "<!DOCTYPE html [\n";
1775                 foreach ( self::$htmlEntities as $entity => $codepoint ) {
1776                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1777                 }
1778                 $out .= "]>\n";
1779                 return $out;
1780         }
1781
1782         /**
1783          * @param string $url
1784          * @return mixed|string
1785          */
1786         static function cleanUrl( $url ) {
1787                 # Normalize any HTML entities in input. They will be
1788                 # re-escaped by makeExternalLink().
1789                 $url = Sanitizer::decodeCharReferences( $url );
1790
1791                 # Escape any control characters introduced by the above step
1792                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1793                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1794
1795                 # Validate hostname portion
1796                 $matches = array();
1797                 if ( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1798                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1799
1800                         // Characters that will be ignored in IDNs.
1801                         // http://tools.ietf.org/html/3454#section-3.1
1802                         // Strip them before further processing so blacklists and such work.
1803                         $strip = "/
1804                                 \\s|          # general whitespace
1805                                 \xc2\xad|     # 00ad SOFT HYPHEN
1806                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1807                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1808                                 \xe2\x81\xa0| # 2060 WORD JOINER
1809                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1810                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1811                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1812                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1813                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1814                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1815                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1816                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1817                                 /xuD";
1818
1819                         $host = preg_replace( $strip, '', $host );
1820
1821                         // IPv6 host names are bracketed with [].  Url-decode these.
1822                         if ( substr_compare( "//%5B", $host, 0, 5 ) === 0 &&
1823                                 preg_match( '!^//%5B([0-9A-Fa-f:.]+)%5D((:\d+)?)$!', $host, $matches )
1824                         ) {
1825                                 $host = '//[' . $matches[1] . ']' . $matches[2];
1826                         }
1827
1828                         // @todo FIXME: Validate hostnames here
1829
1830                         return $protocol . $host . $rest;
1831                 } else {
1832                         return $url;
1833                 }
1834         }
1835
1836         /**
1837          * @param array $matches
1838          * @return string
1839          */
1840         static function cleanUrlCallback( $matches ) {
1841                 return urlencode( $matches[0] );
1842         }
1843
1844         /**
1845          * Does a string look like an e-mail address?
1846          *
1847          * This validates an email address using an HTML5 specification found at:
1848          * http://www.whatwg.org/html/states-of-the-type-attribute.html#valid-e-mail-address
1849          * Which as of 2011-01-24 says:
1850          *
1851          *   A valid e-mail address is a string that matches the ABNF production
1852          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1853          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1854          *   3.5.
1855          *
1856          * This function is an implementation of the specification as requested in
1857          * bug 22449.
1858          *
1859          * Client-side forms will use the same standard validation rules via JS or
1860          * HTML 5 validation; additional restrictions can be enforced server-side
1861          * by extensions via the 'isValidEmailAddr' hook.
1862          *
1863          * Note that this validation doesn't 100% match RFC 2822, but is believed
1864          * to be liberal enough for wide use. Some invalid addresses will still
1865          * pass validation here.
1866          *
1867          * @since 1.18
1868          *
1869          * @param string $addr E-mail address
1870          * @return bool
1871          */
1872         public static function validateEmail( $addr ) {
1873                 $result = null;
1874                 if ( !Hooks::run( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1875                         return $result;
1876                 }
1877
1878                 // Please note strings below are enclosed in brackets [], this make the
1879                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1880                 // See bug 26948
1881                 $rfc5322_atext = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~";
1882                 $rfc1034_ldh_str = "a-z0-9\\-";
1883
1884                 $html5_email_regexp = "/
1885                 ^                      # start of string
1886                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1887                 @                      # 'apostrophe'
1888                 [$rfc1034_ldh_str]+       # First domain part
1889                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1890                 $                      # End of string
1891                 /ix"; // case Insensitive, eXtended
1892
1893                 return (bool)preg_match( $html5_email_regexp, $addr );
1894         }
1895 }