includes/Sanitizer.php

   1 <?php
   2 /**
   3  * XHTML sanitizer for %MediaWiki.
   4  *
   5  * Copyright © 2002-2005 Brion Vibber <brion@pobox.com> et al
   6  * http://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  * XHTML 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          * Blacklist for evil uris like javascript:
  44          * WARNING: DO NOT use this in any place that actually requires blacklisting
  45          * for security reasons. There are NUMEROUS[1] ways to bypass blacklisting, the
  46          * only way to be secure from javascript: uri based xss vectors is to whitelist
  47          * things that you know are safe and deny everything else.
  48          * [1]: http://ha.ckers.org/xss.html
  49          */
  50         const EVIL_URI_PATTERN = '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
  51         const XMLNS_ATTRIBUTE_PATTERN = "/^xmlns:[:A-Z_a-z-.0-9]+$/";
  52
  53         /**
  54          * List of all named character entities defined in HTML 4.01
  55          * http://www.w3.org/TR/html4/sgml/entities.html
  56          * As well as &apos; which is only defined starting in XHTML1.
  57          * @private
  58          */
  59         static $htmlEntities = array(
  60                 'Aacute'   => 193,
  61                 'aacute'   => 225,
  62                 'Acirc'    => 194,
  63                 'acirc'    => 226,
  64                 'acute'    => 180,
  65                 'AElig'    => 198,
  66                 'aelig'    => 230,
  67                 'Agrave'   => 192,
  68                 'agrave'   => 224,
  69                 'alefsym'  => 8501,
  70                 'Alpha'    => 913,
  71                 'alpha'    => 945,
  72                 'amp'      => 38,
  73                 'and'      => 8743,
  74                 'ang'      => 8736,
  75                 'apos'     => 39, // New in XHTML & HTML 5; avoid in output for compatibility with IE.
  76                 'Aring'    => 197,
  77                 'aring'    => 229,
  78                 'asymp'    => 8776,
  79                 'Atilde'   => 195,
  80                 'atilde'   => 227,
  81                 'Auml'     => 196,
  82                 'auml'     => 228,
  83                 'bdquo'    => 8222,
  84                 'Beta'     => 914,
  85                 'beta'     => 946,
  86                 'brvbar'   => 166,
  87                 'bull'     => 8226,
  88                 'cap'      => 8745,
  89                 'Ccedil'   => 199,
  90                 'ccedil'   => 231,
  91                 'cedil'    => 184,
  92                 'cent'     => 162,
  93                 'Chi'      => 935,
  94                 'chi'      => 967,
  95                 'circ'     => 710,
  96                 'clubs'    => 9827,
  97                 'cong'     => 8773,
  98                 'copy'     => 169,
  99                 'crarr'    => 8629,
 100                 'cup'      => 8746,
 101                 'curren'   => 164,
 102                 'dagger'   => 8224,
 103                 'Dagger'   => 8225,
 104                 'darr'     => 8595,
 105                 'dArr'     => 8659,
 106                 'deg'      => 176,
 107                 'Delta'    => 916,
 108                 'delta'    => 948,
 109                 'diams'    => 9830,
 110                 'divide'   => 247,
 111                 'Eacute'   => 201,
 112                 'eacute'   => 233,
 113                 'Ecirc'    => 202,
 114                 'ecirc'    => 234,
 115                 'Egrave'   => 200,
 116                 'egrave'   => 232,
 117                 'empty'    => 8709,
 118                 'emsp'     => 8195,
 119                 'ensp'     => 8194,
 120                 'Epsilon'  => 917,
 121                 'epsilon'  => 949,
 122                 'equiv'    => 8801,
 123                 'Eta'      => 919,
 124                 'eta'      => 951,
 125                 'ETH'      => 208,
 126                 'eth'      => 240,
 127                 'Euml'     => 203,
 128                 'euml'     => 235,
 129                 'euro'     => 8364,
 130                 'exist'    => 8707,
 131                 'fnof'     => 402,
 132                 'forall'   => 8704,
 133                 'frac12'   => 189,
 134                 'frac14'   => 188,
 135                 'frac34'   => 190,
 136                 'frasl'    => 8260,
 137                 'Gamma'    => 915,
 138                 'gamma'    => 947,
 139                 'ge'       => 8805,
 140                 'gt'       => 62,
 141                 'harr'     => 8596,
 142                 'hArr'     => 8660,
 143                 'hearts'   => 9829,
 144                 'hellip'   => 8230,
 145                 'Iacute'   => 205,
 146                 'iacute'   => 237,
 147                 'Icirc'    => 206,
 148                 'icirc'    => 238,
 149                 'iexcl'    => 161,
 150                 'Igrave'   => 204,
 151                 'igrave'   => 236,
 152                 'image'    => 8465,
 153                 'infin'    => 8734,
 154                 'int'      => 8747,
 155                 'Iota'     => 921,
 156                 'iota'     => 953,
 157                 'iquest'   => 191,
 158                 'isin'     => 8712,
 159                 'Iuml'     => 207,
 160                 'iuml'     => 239,
 161                 'Kappa'    => 922,
 162                 'kappa'    => 954,
 163                 'Lambda'   => 923,
 164                 'lambda'   => 955,
 165                 'lang'     => 9001,
 166                 'laquo'    => 171,
 167                 'larr'     => 8592,
 168                 'lArr'     => 8656,
 169                 'lceil'    => 8968,
 170                 'ldquo'    => 8220,
 171                 'le'       => 8804,
 172                 'lfloor'   => 8970,
 173                 'lowast'   => 8727,
 174                 'loz'      => 9674,
 175                 'lrm'      => 8206,
 176                 'lsaquo'   => 8249,
 177                 'lsquo'    => 8216,
 178                 'lt'       => 60,
 179                 'macr'     => 175,
 180                 'mdash'    => 8212,
 181                 'micro'    => 181,
 182                 'middot'   => 183,
 183                 'minus'    => 8722,
 184                 'Mu'       => 924,
 185                 'mu'       => 956,
 186                 'nabla'    => 8711,
 187                 'nbsp'     => 160,
 188                 'ndash'    => 8211,
 189                 'ne'       => 8800,
 190                 'ni'       => 8715,
 191                 'not'      => 172,
 192                 'notin'    => 8713,
 193                 'nsub'     => 8836,
 194                 'Ntilde'   => 209,
 195                 'ntilde'   => 241,
 196                 'Nu'       => 925,
 197                 'nu'       => 957,
 198                 'Oacute'   => 211,
 199                 'oacute'   => 243,
 200                 'Ocirc'    => 212,
 201                 'ocirc'    => 244,
 202                 'OElig'    => 338,
 203                 'oelig'    => 339,
 204                 'Ograve'   => 210,
 205                 'ograve'   => 242,
 206                 'oline'    => 8254,
 207                 'Omega'    => 937,
 208                 'omega'    => 969,
 209                 'Omicron'  => 927,
 210                 'omicron'  => 959,
 211                 'oplus'    => 8853,
 212                 'or'       => 8744,
 213                 'ordf'     => 170,
 214                 'ordm'     => 186,
 215                 'Oslash'   => 216,
 216                 'oslash'   => 248,
 217                 'Otilde'   => 213,
 218                 'otilde'   => 245,
 219                 'otimes'   => 8855,
 220                 'Ouml'     => 214,
 221                 'ouml'     => 246,
 222                 'para'     => 182,
 223                 'part'     => 8706,
 224                 'permil'   => 8240,
 225                 'perp'     => 8869,
 226                 'Phi'      => 934,
 227                 'phi'      => 966,
 228                 'Pi'       => 928,
 229                 'pi'       => 960,
 230                 'piv'      => 982,
 231                 'plusmn'   => 177,
 232                 'pound'    => 163,
 233                 'prime'    => 8242,
 234                 'Prime'    => 8243,
 235                 'prod'     => 8719,
 236                 'prop'     => 8733,
 237                 'Psi'      => 936,
 238                 'psi'      => 968,
 239                 'quot'     => 34,
 240                 'radic'    => 8730,
 241                 'rang'     => 9002,
 242                 'raquo'    => 187,
 243                 'rarr'     => 8594,
 244                 'rArr'     => 8658,
 245                 'rceil'    => 8969,
 246                 'rdquo'    => 8221,
 247                 'real'     => 8476,
 248                 'reg'      => 174,
 249                 'rfloor'   => 8971,
 250                 'Rho'      => 929,
 251                 'rho'      => 961,
 252                 'rlm'      => 8207,
 253                 'rsaquo'   => 8250,
 254                 'rsquo'    => 8217,
 255                 'sbquo'    => 8218,
 256                 'Scaron'   => 352,
 257                 'scaron'   => 353,
 258                 'sdot'     => 8901,
 259                 'sect'     => 167,
 260                 'shy'      => 173,
 261                 'Sigma'    => 931,
 262                 'sigma'    => 963,
 263                 'sigmaf'   => 962,
 264                 'sim'      => 8764,
 265                 'spades'   => 9824,
 266                 'sub'      => 8834,
 267                 'sube'     => 8838,
 268                 'sum'      => 8721,
 269                 'sup'      => 8835,
 270                 'sup1'     => 185,
 271                 'sup2'     => 178,
 272                 'sup3'     => 179,
 273                 'supe'     => 8839,
 274                 'szlig'    => 223,
 275                 'Tau'      => 932,
 276                 'tau'      => 964,
 277                 'there4'   => 8756,
 278                 'Theta'    => 920,
 279                 'theta'    => 952,
 280                 'thetasym' => 977,
 281                 'thinsp'   => 8201,
 282                 'THORN'    => 222,
 283                 'thorn'    => 254,
 284                 'tilde'    => 732,
 285                 'times'    => 215,
 286                 'trade'    => 8482,
 287                 'Uacute'   => 218,
 288                 'uacute'   => 250,
 289                 'uarr'     => 8593,
 290                 'uArr'     => 8657,
 291                 'Ucirc'    => 219,
 292                 'ucirc'    => 251,
 293                 'Ugrave'   => 217,
 294                 'ugrave'   => 249,
 295                 'uml'      => 168,
 296                 'upsih'    => 978,
 297                 'Upsilon'  => 933,
 298                 'upsilon'  => 965,
 299                 'Uuml'     => 220,
 300                 'uuml'     => 252,
 301                 'weierp'   => 8472,
 302                 'Xi'       => 926,
 303                 'xi'       => 958,
 304                 'Yacute'   => 221,
 305                 'yacute'   => 253,
 306                 'yen'      => 165,
 307                 'Yuml'     => 376,
 308                 'yuml'     => 255,
 309                 'Zeta'     => 918,
 310                 'zeta'     => 950,
 311                 'zwj'      => 8205,
 312                 'zwnj'     => 8204
 313         );
 314
 315         /**
 316          * Character entity aliases accepted by MediaWiki
 317          */
 318         static $htmlEntityAliases = array(
 319                 'רלמ' => 'rlm',
 320                 'رلم' => 'rlm',
 321         );
 322
 323         /**
 324          * Lazy-initialised attributes regex, see getAttribsRegex()
 325          */
 326         static $attribsRegex;
 327
 328         /**
 329          * Regular expression to match HTML/XML attribute pairs within a tag.
 330          * Allows some... latitude.
 331          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 332          */
 333         static function getAttribsRegex() {
 334                 if ( self::$attribsRegex === null ) {
 335                         $attribFirst = '[:A-Z_a-z0-9]';
 336                         $attrib = '[:A-Z_a-z-.0-9]';
 337                         $space = '[\x09\x0a\x0d\x20]';
 338                         self::$attribsRegex =
 339                                 "/(?:^|$space)({$attribFirst}{$attrib}*)
 340                                   ($space*=$space*
 341                                         (?:
 342                                          # The attribute value: quoted or alone
 343                                           \"([^<\"]*)\"
 344                                          | '([^<']*)'
 345                                          |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
 346                                          |  (\#[0-9a-fA-F]+) # Technically wrong, but lots of
 347                                                                                  # colors are specified like this.
 348                                                                                  # We'll be normalizing it.
 349                                         )
 350                                 )?(?=$space|\$)/sx";
 351                 }
 352                 return self::$attribsRegex;
 353         }
 354
 355         /**
 356          * Cleans up HTML, removes dangerous tags and attributes, and
 357          * removes HTML comments
 358          * @private
 359          * @param $text String
 360          * @param $processCallback Callback to do any variable or parameter replacements in HTML attribute values
 361          * @param $args Array for the processing callback
 362          * @param $extratags Array for any extra tags to include
 363          * @param $removetags Array for any tags (default or extra) to exclude
 364          * @return string
 365          */
 366         static function removeHTMLtags( $text, $processCallback = null, $args = array(), $extratags = array(), $removetags = array() ) {
 367                 global $wgUseTidy;
 368
 369                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 370                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 371
 372                 wfProfileIn( __METHOD__ );
 373
 374                 if ( !$staticInitialised ) {
 375
 376                         $htmlpairsStatic = array( # Tags that must be closed
 377                                 'b', 'bdi', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 378                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 379                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 380                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 381                                 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'abbr', 'dfn',
 382                                 'kbd', 'samp'
 383                         );
 384                         $htmlsingle = array(
 385                                 'br', 'hr', 'li', 'dt', 'dd'
 386                         );
 387                         $htmlsingleonly = array( # Elements that cannot have close tags
 388                                 'br', 'hr'
 389                         );
 390                         $htmlnest = array( # Tags that can be nested directly or indirectly
 391                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 392                                 'li', 'dl', 'dt', 'dd', 'font', 'big', 'small', 'sub', 'sup', 'span'
 393                         );
 394                         $tabletags = array( # Can only appear inside table, we will close them
 395                                 'td', 'th', 'tr',
 396                         );
 397                         $htmllist = array( # Tags used by list
 398                                 'ul','ol',
 399                         );
 400                         $listtags = array( # Tags that can appear in a list
 401                                 'li',
 402                         );
 403
 404                         global $wgAllowImageTag;
 405                         if ( $wgAllowImageTag ) {
 406                                 $htmlsingle[] = 'img';
 407                                 $htmlsingleonly[] = 'img';
 408                         }
 409
 410                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 411                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 412
 413                         # Convert them all to hashtables for faster lookup
 414                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 415                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 416                         foreach ( $vars as $var ) {
 417                                 $$var = array_flip( $$var );
 418                         }
 419                         $staticInitialised = true;
 420                 }
 421                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 422                 $extratags = array_flip( $extratags );
 423                 $removetags = array_flip( $removetags );
 424                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 425                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ) , $removetags );
 426
 427                 # Remove HTML comments
 428                 $text = Sanitizer::removeHTMLcomments( $text );
 429                 $bits = explode( '<', $text );
 430                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 431                 if ( !$wgUseTidy ) {
 432                         $tagstack = $tablestack = array();
 433                         foreach ( $bits as $x ) {
 434                                 $regs = array();
 435                                 # $slash: Does the current element start with a '/'?
 436                                 # $t: Current element name
 437                                 # $params: String between element name and >
 438                                 # $brace: Ending '>' or '/>'
 439                                 # $rest: Everything until the next element of $bits
 440                                 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 441                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 442                                 } else {
 443                                         $slash = $t = $params = $brace = $rest = null;
 444                                 }
 445
 446                                 $badtag = false;
 447                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 448                                         # Check our stack
 449                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 450                                                 $badtag = true;
 451                                         } elseif ( $slash ) {
 452                                                 # Closing a tag... is it the one we just opened?
 453                                                 $ot = @array_pop( $tagstack );
 454                                                 if ( $ot != $t ) {
 455                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 456                                                                 # Pop all elements with an optional close tag
 457                                                                 # and see if we find a match below them
 458                                                                 $optstack = array();
 459                                                                 array_push( $optstack, $ot );
 460                                                                 wfSuppressWarnings();
 461                                                                 $ot = array_pop( $tagstack );
 462                                                                 wfRestoreWarnings();
 463                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 464                                                                         array_push( $optstack, $ot );
 465                                                                         wfSuppressWarnings();
 466                                                                         $ot = array_pop( $tagstack );
 467                                                                         wfRestoreWarnings();
 468                                                                 }
 469                                                                 if ( $t != $ot ) {
 470                                                                         # No match. Push the optional elements back again
 471                                                                         $badtag = true;
 472                                                                         wfSuppressWarnings();
 473                                                                         $ot = array_pop( $optstack );
 474                                                                         wfRestoreWarnings();
 475                                                                         while ( $ot ) {
 476                                                                                 array_push( $tagstack, $ot );
 477                                                                                 wfSuppressWarnings();
 478                                                                                 $ot = array_pop( $optstack );
 479                                                                                 wfRestoreWarnings();
 480                                                                         }
 481                                                                 }
 482                                                         } else {
 483                                                                 @array_push( $tagstack, $ot );
 484                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 485                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 486                                                                         $badtag = true;
 487                                                                 }
 488                                                         }
 489                                                 } else {
 490                                                         if ( $t == 'table' ) {
 491                                                                 $tagstack = array_pop( $tablestack );
 492                                                         }
 493                                                 }
 494                                                 $newparams = '';
 495                                         } else {
 496                                                 # Keep track for later
 497                                                 if ( isset( $tabletags[$t] ) &&
 498                                                 !in_array( 'table', $tagstack ) ) {
 499                                                         $badtag = true;
 500                                                 } elseif ( in_array( $t, $tagstack ) &&
 501                                                 !isset( $htmlnest [$t ] ) ) {
 502                                                         $badtag = true;
 503                                                 # Is it a self closed htmlpair ? (bug 5487)
 504                                                 } elseif ( $brace == '/>' &&
 505                                                 isset( $htmlpairs[$t] ) ) {
 506                                                         $badtag = true;
 507                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 508                                                         # Hack to force empty tag for uncloseable elements
 509                                                         $brace = '/>';
 510                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 511                                                         # Hack to not close $htmlsingle tags
 512                                                         $brace = null;
 513                                                         # Still need to push this optionally-closed tag to
 514                                                         # the tag stack so that we can match end tags
 515                                                         # instead of marking them as bad.
 516                                                         array_push( $tagstack, $t );
 517                                                 } elseif ( isset( $tabletags[$t] )
 518                                                 && in_array( $t, $tagstack ) ) {
 519                                                         // New table tag but forgot to close the previous one
 520                                                         $text .= "</$t>";
 521                                                 } else {
 522                                                         if ( $t == 'table' ) {
 523                                                                 array_push( $tablestack, $tagstack );
 524                                                                 $tagstack = array();
 525                                                         }
 526                                                         array_push( $tagstack, $t );
 527                                                 }
 528
 529                                                 # Replace any variables or template parameters with
 530                                                 # plaintext results.
 531                                                 if( is_callable( $processCallback ) ) {
 532                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 533                                                 }
 534
 535                                                 # Strip non-approved attributes from the tag
 536                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 537                                         }
 538                                         if ( !$badtag ) {
 539                                                 $rest = str_replace( '>', '&gt;', $rest );
 540                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 541                                                 $text .= "<$slash$t$newparams$close>$rest";
 542                                                 continue;
 543                                         }
 544                                 }
 545                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 546                         }
 547                         # Close off any remaining tags
 548                         while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
 549                                 $text .= "</$t>\n";
 550                                 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
 551                         }
 552                 } else {
 553                         # this might be possible using tidy itself
 554                         foreach ( $bits as $x ) {
 555                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 556                                 $x, $regs );
 557                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 558                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 559                                         if( is_callable( $processCallback ) ) {
 560                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 561                                         }
 562                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 563                                         $rest = str_replace( '>', '&gt;', $rest );
 564                                         $text .= "<$slash$t$newparams$brace$rest";
 565                                 } else {
 566                                         $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 567                                 }
 568                         }
 569                 }
 570                 wfProfileOut( __METHOD__ );
 571                 return $text;
 572         }
 573
 574         /**
 575          * Remove '<!--', '-->', and everything between.
 576          * To avoid leaving blank lines, when a comment is both preceded
 577          * and followed by a newline (ignoring spaces), trim leading and
 578          * trailing spaces and one of the newlines.
 579          *
 580          * @private
 581          * @param $text String
 582          * @return string
 583          */
 584         static function removeHTMLcomments( $text ) {
 585                 wfProfileIn( __METHOD__ );
 586                 while (($start = strpos($text, '<!--')) !== false) {
 587                         $end = strpos($text, '-->', $start + 4);
 588                         if ($end === false) {
 589                                 # Unterminated comment; bail out
 590                                 break;
 591                         }
 592
 593                         $end += 3;
 594
 595                         # Trim space and newline if the comment is both
 596                         # preceded and followed by a newline
 597                         $spaceStart = max($start - 1, 0);
 598                         $spaceLen = $end - $spaceStart;
 599                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 600                                 $spaceStart--;
 601                                 $spaceLen++;
 602                         }
 603                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 604                                 $spaceLen++;
 605                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 606                                 # Remove the comment, leading and trailing
 607                                 # spaces, and leave only one newline.
 608                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 609                         }
 610                         else {
 611                                 # Remove just the comment.
 612                                 $text = substr_replace($text, '', $start, $end - $start);
 613                         }
 614                 }
 615                 wfProfileOut( __METHOD__ );
 616                 return $text;
 617         }
 618
 619         /**
 620          * Take an array of attribute names and values and fix some deprecated values
 621          * for the given element type.
 622          * This does not validate properties, so you should ensure that you call
 623          * validateTagAttributes AFTER this to ensure that the resulting style rule
 624          * this may add is safe.
 625          *
 626          * - Converts most presentational attributes like align into inline css
 627          *
 628          * @param $attribs Array
 629          * @param $element String
 630          * @return Array
 631          */
 632         static function fixDeprecatedAttributes( $attribs, $element ) {
 633                 global $wgHtml5, $wgCleanupPresentationalAttributes;
 634
 635                 // presentational attributes were removed from html5, we can leave them
 636                 // in when html5 is turned off
 637                 if ( !$wgHtml5 || !$wgCleanupPresentationalAttributes ) {
 638                         return $attribs;
 639                 }
 640
 641                 $table = array( 'table' );
 642                 $cells = array( 'td', 'th' );
 643                 $colls = array( 'col', 'colgroup' );
 644                 $tblocks = array( 'tbody', 'tfoot', 'thead' );
 645                 $h = array( 'h1', 'h2', 'h3', 'h4', 'h5', 'h6' );
 646
 647                 $presentationalAttribs = array(
 648                         'align' => array( 'text-align', array_merge( array( 'caption', 'hr', 'div', 'p', 'tr' ), $table, $cells, $colls, $tblocks, $h ) ),
 649                         'clear' => array( 'clear', array( 'br' ) ),
 650                         'height' => array( 'height', $cells ),
 651                         'nowrap' => array( 'white-space', $cells ),
 652                         'size' => array( 'height', array( 'hr' ) ),
 653                         'type' => array( 'list-style-type', array( 'li', 'ol', 'ul' ) ),
 654                         'valign' => array( 'vertical-align', array_merge( $cells, $colls, $tblocks ) ),
 655                         'width' => array( 'width', array_merge( array( 'hr', 'pre' ), $table, $cells, $colls ) ),
 656                 );
 657
 658                 // Ensure that any upper case or mixed case attributes are converted to lowercase
 659                 foreach ( $attribs as $attribute => $value ) {
 660                         if ( $attribute !== strtolower( $attribute ) && array_key_exists( strtolower( $attribute ), $presentationalAttribs ) ) {
 661                                 $attribs[strtolower( $attribute )] = $value;
 662                                 unset( $attribs[$attribute] );
 663                         }
 664                 }
 665
 666                 $style = "";
 667                 foreach ( $presentationalAttribs as $attribute => $info ) {
 668                         list( $property, $elements ) = $info;
 669
 670                         // Skip if this attribute is not relevant to this element
 671                         if ( !in_array( $element, $elements ) ) {
 672                                 continue;
 673                         }
 674
 675                         // Skip if the attribute is not used
 676                         if ( !array_key_exists( $attribute, $attribs ) ) {
 677                                 continue;
 678                         }
 679
 680                         $value = $attribs[$attribute];
 681
 682                         // For nowrap the value should be nowrap instead of whatever text is in the value
 683                         if ( $attribute === 'nowrap' ) {
 684                                 $value = 'nowrap';
 685                         }
 686
 687                         // clear="all" is clear: both; in css
 688                         if ( $attribute === 'clear' && strtolower( $value ) === 'all' ) {
 689                                 $value = 'both';
 690                         }
 691
 692                         // Size based properties should have px applied to them if they have no unit
 693                         if ( in_array( $attribute, array( 'height', 'width', 'size' ) ) ) {
 694                                 if ( preg_match( '/^[\d.]+$/', $value ) ) {
 695                                         $value = "{$value}px";
 696                                 }
 697                         }
 698
 699                         // Table align is special, it's about block alignment instead of
 700                         // content align (see also bug 40306)
 701                         if ( $attribute === 'align' && in_array( $element, $table ) ) {
 702                                 if ( $value === 'center' ) {
 703                                         $style .= ' margin-left: auto;';
 704                                         $property = 'margin-right';
 705                                         $value = 'auto';
 706                                 } else {
 707                                         $property = 'float';
 708                                 }
 709                         }
 710
 711                         $style .= " $property: $value;";
 712
 713                         unset( $attribs[$attribute] );
 714                 }
 715
 716                 if ( $style ) {
 717                         // Prepend our style rules so that they can be overridden by user css
 718                         if ( isset($attribs['style']) ) {
 719                                 $style .= " " . $attribs['style'];
 720                         }
 721                         $attribs['style'] = trim($style);
 722                 }
 723
 724                 return $attribs;
 725         }
 726
 727         /**
 728          * Take an array of attribute names and values and normalize or discard
 729          * illegal values for the given element type.
 730          *
 731          * - Discards attributes not on a whitelist for the given element
 732          * - Unsafe style attributes are discarded
 733          * - Invalid id attributes are reencoded
 734          *
 735          * @param $attribs Array
 736          * @param $element String
 737          * @return Array
 738          *
 739          * @todo Check for legal values where the DTD limits things.
 740          * @todo Check for unique id attribute :P
 741          */
 742         static function validateTagAttributes( $attribs, $element ) {
 743                 return Sanitizer::validateAttributes( $attribs,
 744                         Sanitizer::attributeWhitelist( $element ) );
 745         }
 746
 747         /**
 748          * Take an array of attribute names and values and normalize or discard
 749          * illegal values for the given whitelist.
 750          *
 751          * - Discards attributes not the given whitelist
 752          * - Unsafe style attributes are discarded
 753          * - Invalid id attributes are reencoded
 754          *
 755          * @param $attribs Array
 756          * @param $whitelist Array: list of allowed attribute names
 757          * @return Array
 758          *
 759          * @todo Check for legal values where the DTD limits things.
 760          * @todo Check for unique id attribute :P
 761          */
 762         static function validateAttributes( $attribs, $whitelist ) {
 763                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
 764
 765                 $whitelist = array_flip( $whitelist );
 766                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 767
 768                 $out = array();
 769                 foreach( $attribs as $attribute => $value ) {
 770                         #allow XML namespace declaration if RDFa is enabled
 771                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 772                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 773                                         $out[$attribute] = $value;
 774                                 }
 775
 776                                 continue;
 777                         }
 778
 779                         # Allow any attribute beginning with "data-", if in HTML5 mode
 780                         if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
 781                                 continue;
 782                         }
 783
 784                         # Strip javascript "expression" from stylesheets.
 785                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 786                         if( $attribute == 'style' ) {
 787                                 $value = Sanitizer::checkCss( $value );
 788                         }
 789
 790                         if ( $attribute === 'id' ) {
 791                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 792                         }
 793
 794                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 795                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 796                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 797                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 798                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 799                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 800
 801                                 //Paranoia. Allow "simple" values but suppress javascript
 802                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 803                                         continue;
 804                                 }
 805                         }
 806
 807                         # NOTE: even though elements using href/src are not allowed directly, supply
 808                         #       validation code that can be used by tag hook handlers, etc
 809                         if ( $attribute === 'href' || $attribute === 'src' ) {
 810                                 if ( !preg_match( $hrefExp, $value ) ) {
 811                                         continue; //drop any href or src attributes not using an allowed protocol.
 812                                                   //NOTE: this also drops all relative URLs
 813                                 }
 814                         }
 815
 816                         // If this attribute was previously set, override it.
 817                         // Output should only have one attribute of each name.
 818                         $out[$attribute] = $value;
 819                 }
 820
 821                 if ( $wgAllowMicrodataAttributes ) {
 822                         # itemtype, itemid, itemref don't make sense without itemscope
 823                         if ( !array_key_exists( 'itemscope', $out ) ) {
 824                                 unset( $out['itemtype'] );
 825                                 unset( $out['itemid'] );
 826                                 unset( $out['itemref'] );
 827                         }
 828                         # TODO: Strip itemprop if we aren't descendants of an itemscope.
 829                 }
 830                 return $out;
 831         }
 832
 833         /**
 834          * Merge two sets of HTML attributes.  Conflicting items in the second set
 835          * will override those in the first, except for 'class' attributes which
 836          * will be combined (if they're both strings).
 837          *
 838          * @todo implement merging for other attributes such as style
 839          * @param $a Array
 840          * @param $b Array
 841          * @return array
 842          */
 843         static function mergeAttributes( $a, $b ) {
 844                 $out = array_merge( $a, $b );
 845                 if( isset( $a['class'] ) && isset( $b['class'] )
 846                 && is_string( $a['class'] ) && is_string( $b['class'] )
 847                 && $a['class'] !== $b['class'] ) {
 848                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 849                                 -1, PREG_SPLIT_NO_EMPTY );
 850                         $out['class'] = implode( ' ', array_unique( $classes ) );
 851                 }
 852                 return $out;
 853         }
 854
 855         /**
 856          * Pick apart some CSS and check it for forbidden or unsafe structures.
 857          * Returns a sanitized string. This sanitized string will have
 858          * character references and escape sequences decoded, and comments
 859          * stripped. If the input is just too evil, only a comment complaining
 860          * about evilness will be returned.
 861          *
 862          * Currently URL references, 'expression', 'tps' are forbidden.
 863          *
 864          * NOTE: Despite the fact that character references are decoded, the
 865          * returned string may contain character references given certain
 866          * clever input strings. These character references must
 867          * be escaped before the return value is embedded in HTML.
 868          *
 869          * @param $value String
 870          * @return String
 871          */
 872         static function checkCss( $value ) {
 873                 // Decode character references like &#123;
 874                 $value = Sanitizer::decodeCharReferences( $value );
 875
 876                 // Decode escape sequences and line continuation
 877                 // See the grammar in the CSS 2 spec, appendix D.
 878                 // This has to be done AFTER decoding character references.
 879                 // This means it isn't possible for this function to return
 880                 // unsanitized escape sequences. It is possible to manufacture
 881                 // input that contains character references that decode to
 882                 // escape sequences that decode to character references, but
 883                 // it's OK for the return value to contain character references
 884                 // because the caller is supposed to escape those anyway.
 885                 static $decodeRegex;
 886                 if ( !$decodeRegex ) {
 887                         $space = '[\\x20\\t\\r\\n\\f]';
 888                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 889                         $backslash = '\\\\';
 890                         $decodeRegex = "/ $backslash
 891                                 (?:
 892                                         ($nl) |  # 1. Line continuation
 893                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 894                                         (.) | # 3. backslash cancelling special meaning
 895                                         () | # 4. backslash at end of string
 896                                 )/xu";
 897                 }
 898                 $value = preg_replace_callback( $decodeRegex,
 899                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 900
 901                 // Remove any comments; IE gets token splitting wrong
 902                 // This must be done AFTER decoding character references and
 903                 // escape sequences, because those steps can introduce comments
 904                 // This step cannot introduce character references or escape
 905                 // sequences, because it replaces comments with spaces rather
 906                 // than removing them completely.
 907                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 908
 909                 // Remove anything after a comment-start token, to guard against
 910                 // incorrect client implementations.
 911                 $commentPos = strpos( $value, '/*' );
 912                 if ( $commentPos !== false ) {
 913                         $value = substr( $value, 0, $commentPos );
 914                 }
 915
 916                 // Reject problematic keywords and control characters
 917                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 918                         return '/* invalid control char */';
 919                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( | image\s*\( !ix', $value ) ) {
 920                         return '/* insecure input */';
 921                 }
 922                 return $value;
 923         }
 924
 925         /**
 926          * @param $matches array
 927          * @return String
 928          */
 929         static function cssDecodeCallback( $matches ) {
 930                 if ( $matches[1] !== '' ) {
 931                         // Line continuation
 932                         return '';
 933                 } elseif ( $matches[2] !== '' ) {
 934                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 935                 } elseif ( $matches[3] !== '' ) {
 936                         $char = $matches[3];
 937                 } else {
 938                         $char = '\\';
 939                 }
 940                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 941                         // These characters need to be escaped in strings
 942                         // Clean up the escape sequence to avoid parsing errors by clients
 943                         return '\\' . dechex( ord( $char ) ) . ' ';
 944                 } else {
 945                         // Decode unnecessary escape
 946                         return $char;
 947                 }
 948         }
 949
 950         /**
 951          * Take a tag soup fragment listing an HTML element's attributes
 952          * and normalize it to well-formed XML, discarding unwanted attributes.
 953          * Output is safe for further wikitext processing, with escaping of
 954          * values that could trigger problems.
 955          *
 956          * - Normalizes attribute names to lowercase
 957          * - Discards attributes not on a whitelist for the given element
 958          * - Turns broken or invalid entities into plaintext
 959          * - Double-quotes all attribute values
 960          * - Attributes without values are given the name as attribute
 961          * - Double attributes are discarded
 962          * - Unsafe style attributes are discarded
 963          * - Prepends space if there are attributes.
 964          *
 965          * @param $text String
 966          * @param $element String
 967          * @return String
 968          */
 969         static function fixTagAttributes( $text, $element ) {
 970                 if( trim( $text ) == '' ) {
 971                         return '';
 972                 }
 973
 974                 $decoded = Sanitizer::decodeTagAttributes( $text );
 975                 $decoded = Sanitizer::fixDeprecatedAttributes( $decoded, $element );
 976                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
 977
 978                 $attribs = array();
 979                 foreach( $stripped as $attribute => $value ) {
 980                         $encAttribute = htmlspecialchars( $attribute );
 981                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 982
 983                         $attribs[] = "$encAttribute=\"$encValue\"";
 984                 }
 985                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 986         }
 987
 988         /**
 989          * Encode an attribute value for HTML output.
 990          * @param $text String
 991          * @return HTML-encoded text fragment
 992          */
 993         static function encodeAttribute( $text ) {
 994                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 995
 996                 // Whitespace is normalized during attribute decoding,
 997                 // so if we've been passed non-spaces we must encode them
 998                 // ahead of time or they won't be preserved.
 999                 $encValue = strtr( $encValue, array(
1000                         "\n" => '&#10;',
1001                         "\r" => '&#13;',
1002                         "\t" => '&#9;',
1003                 ) );
1004
1005                 return $encValue;
1006         }
1007
1008         /**
1009          * Encode an attribute value for HTML tags, with extra armoring
1010          * against further wiki processing.
1011          * @param $text String
1012          * @return HTML-encoded text fragment
1013          */
1014         static function safeEncodeAttribute( $text ) {
1015                 $encValue = Sanitizer::encodeAttribute( $text );
1016
1017                 # Templates and links may be expanded in later parsing,
1018                 # creating invalid or dangerous output. Suppress this.
1019                 $encValue = strtr( $encValue, array(
1020                         '<'    => '&lt;',   // This should never happen,
1021                         '>'    => '&gt;',   // we've received invalid input
1022                         '"'    => '&quot;', // which should have been escaped.
1023                         '{'    => '&#123;',
1024                         '['    => '&#91;',
1025                         "''"   => '&#39;&#39;',
1026                         'ISBN' => '&#73;SBN',
1027                         'RFC'  => '&#82;FC',
1028                         'PMID' => '&#80;MID',
1029                         '|'    => '&#124;',
1030                         '__'   => '&#95;_',
1031                 ) );
1032
1033                 # Stupid hack
1034                 $encValue = preg_replace_callback(
1035                         '/((?i)' . wfUrlProtocols() . ')/',
1036                         array( 'Sanitizer', 'armorLinksCallback' ),
1037                         $encValue );
1038                 return $encValue;
1039         }
1040
1041         /**
1042          * Given a value, escape it so that it can be used in an id attribute and
1043          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1044          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1045          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1046          * escaped with lots of dots.
1047          *
1048          * To ensure we don't have to bother escaping anything, we also strip ', ",
1049          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1050          * We also strip # because it upsets IE, and % because it could be
1051          * ambiguous if it's part of something that looks like a percent escape
1052          * (which don't work reliably in fragments cross-browser).
1053          *
1054          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1055          *                                                          in the id and
1056          *                                                          name attributes
1057          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1058          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1059          *   HTML5 definition of id attribute
1060          *
1061          * @param $id String: id to escape
1062          * @param $options Mixed: string or array of strings (default is array()):
1063          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1064          *       so don't pay attention if the first character isn't valid at the
1065          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1066          *       false.
1067          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1068          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1069          *       anchors and links won't break.
1070          * @return String
1071          */
1072         static function escapeId( $id, $options = array() ) {
1073                 global $wgHtml5, $wgExperimentalHtmlIds;
1074                 $options = (array)$options;
1075
1076                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1077                         $id = Sanitizer::decodeCharReferences( $id );
1078                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1079                         $id = trim( $id, '_' );
1080                         if ( $id === '' ) {
1081                                 # Must have been all whitespace to start with.
1082                                 return '_';
1083                         } else {
1084                                 return $id;
1085                         }
1086                 }
1087
1088                 # HTML4-style escaping
1089                 static $replace = array(
1090                         '%3A' => ':',
1091                         '%' => '.'
1092                 );
1093
1094                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1095                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1096
1097                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1098                 && !in_array( 'noninitial', $options ) )  {
1099                         // Initial character must be a letter!
1100                         $id = "x$id";
1101                 }
1102                 return $id;
1103         }
1104
1105         /**
1106          * Given a value, escape it so that it can be used as a CSS class and
1107          * return it.
1108          *
1109          * @todo For extra validity, input should be validated UTF-8.
1110          *
1111          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1112          *
1113          * @param $class String
1114          * @return String
1115          */
1116         static function escapeClass( $class ) {
1117                 // Convert ugly stuff to underscores and kill underscores in ugly places
1118                 return rtrim(preg_replace(
1119                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
1120                         '_',
1121                         $class ), '_');
1122         }
1123
1124         /**
1125          * Given HTML input, escape with htmlspecialchars but un-escape entites.
1126          * This allows (generally harmless) entities like &#160; to survive.
1127          *
1128          * @param $html String to escape
1129          * @return String: escaped input
1130          */
1131         static function escapeHtmlAllowEntities( $html ) {
1132                 $html = Sanitizer::decodeCharReferences( $html );
1133                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1134                 # hurt.
1135                 $html = htmlspecialchars( $html, ENT_QUOTES );
1136                 return $html;
1137         }
1138
1139         /**
1140          * Regex replace callback for armoring links against further processing.
1141          * @param $matches Array
1142          * @return string
1143          */
1144         private static function armorLinksCallback( $matches ) {
1145                 return str_replace( ':', '&#58;', $matches[1] );
1146         }
1147
1148         /**
1149          * Return an associative array of attribute names and values from
1150          * a partial tag string. Attribute names are forces to lowercase,
1151          * character references are decoded to UTF-8 text.
1152          *
1153          * @param $text String
1154          * @return Array
1155          */
1156         public static function decodeTagAttributes( $text ) {
1157                 if( trim( $text ) == '' ) {
1158                         return array();
1159                 }
1160
1161                 $attribs = array();
1162                 $pairs = array();
1163                 if( !preg_match_all(
1164                         self::getAttribsRegex(),
1165                         $text,
1166                         $pairs,
1167                         PREG_SET_ORDER ) ) {
1168                         return $attribs;
1169                 }
1170
1171                 foreach( $pairs as $set ) {
1172                         $attribute = strtolower( $set[1] );
1173                         $value = Sanitizer::getTagAttributeCallback( $set );
1174
1175                         // Normalize whitespace
1176                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1177                         $value = trim( $value );
1178
1179                         // Decode character references
1180                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1181                 }
1182                 return $attribs;
1183         }
1184
1185         /**
1186          * Pick the appropriate attribute value from a match set from the
1187          * attribs regex matches.
1188          *
1189          * @param $set Array
1190          * @throws MWException
1191          * @return String
1192          */
1193         private static function getTagAttributeCallback( $set ) {
1194                 if( isset( $set[6] ) ) {
1195                         # Illegal #XXXXXX color with no quotes.
1196                         return $set[6];
1197                 } elseif( isset( $set[5] ) ) {
1198                         # No quotes.
1199                         return $set[5];
1200                 } elseif( isset( $set[4] ) ) {
1201                         # Single-quoted
1202                         return $set[4];
1203                 } elseif( isset( $set[3] ) ) {
1204                         # Double-quoted
1205                         return $set[3];
1206                 } elseif( !isset( $set[2] ) ) {
1207                         # In XHTML, attributes must have a value.
1208                         # For 'reduced' form, return explicitly the attribute name here.
1209                         return $set[1];
1210                 } else {
1211                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1212                 }
1213         }
1214
1215         /**
1216          * Normalize whitespace and character references in an XML source-
1217          * encoded text for an attribute value.
1218          *
1219          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1220          * but note that we're not returning the value, but are returning
1221          * XML source fragments that will be slapped into output.
1222          *
1223          * @param $text String
1224          * @return String
1225          */
1226         private static function normalizeAttributeValue( $text ) {
1227                 return str_replace( '"', '&quot;',
1228                         self::normalizeWhitespace(
1229                                 Sanitizer::normalizeCharReferences( $text ) ) );
1230         }
1231
1232         /**
1233          * @param $text string
1234          * @return mixed
1235          */
1236         private static function normalizeWhitespace( $text ) {
1237                 return preg_replace(
1238                         '/\r\n|[\x20\x0d\x0a\x09]/',
1239                         ' ',
1240                         $text );
1241         }
1242
1243         /**
1244          * Normalizes whitespace in a section name, such as might be returned
1245          * by Parser::stripSectionName(), for use in the id's that are used for
1246          * section links.
1247          *
1248          * @param $section String
1249          * @return String
1250          */
1251         static function normalizeSectionNameWhitespace( $section ) {
1252                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1253         }
1254
1255         /**
1256          * Ensure that any entities and character references are legal
1257          * for XML and XHTML specifically. Any stray bits will be
1258          * &amp;-escaped to result in a valid text fragment.
1259          *
1260          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1261          *   numericized (this way we're well-formed even without a DTD)
1262          * b. any numeric char refs must be legal chars, not invalid or forbidden
1263          * c. use lower cased "&#x", not "&#X"
1264          * d. fix or reject non-valid attributes
1265          *
1266          * @param $text String
1267          * @return String
1268          * @private
1269          */
1270         static function normalizeCharReferences( $text ) {
1271                 return preg_replace_callback(
1272                         self::CHAR_REFS_REGEX,
1273                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1274                         $text );
1275         }
1276         /**
1277          * @param $matches String
1278          * @return String
1279          */
1280         static function normalizeCharReferencesCallback( $matches ) {
1281                 $ret = null;
1282                 if( $matches[1] != '' ) {
1283                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1284                 } elseif( $matches[2] != '' ) {
1285                         $ret = Sanitizer::decCharReference( $matches[2] );
1286                 } elseif( $matches[3] != ''  ) {
1287                         $ret = Sanitizer::hexCharReference( $matches[3] );
1288                 }
1289                 if( is_null( $ret ) ) {
1290                         return htmlspecialchars( $matches[0] );
1291                 } else {
1292                         return $ret;
1293                 }
1294         }
1295
1296         /**
1297          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1298          * return the equivalent numeric entity reference (except for the core &lt;
1299          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1300          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1301          * pseudo-entity source (eg &amp;foo;)
1302          *
1303          * @param $name String
1304          * @return String
1305          */
1306         static function normalizeEntity( $name ) {
1307                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1308                         return '&' . self::$htmlEntityAliases[$name] . ';';
1309                 } elseif ( in_array( $name,
1310                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1311                         return "&$name;";
1312                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1313                         return '&#' . self::$htmlEntities[$name] . ';';
1314                 } else {
1315                         return "&amp;$name;";
1316                 }
1317         }
1318
1319         /**
1320          * @param $codepoint
1321          * @return null|string
1322          */
1323         static function decCharReference( $codepoint ) {
1324                 $point = intval( $codepoint );
1325                 if( Sanitizer::validateCodepoint( $point ) ) {
1326                         return sprintf( '&#%d;', $point );
1327                 } else {
1328                         return null;
1329                 }
1330         }
1331
1332         /**
1333          * @param $codepoint
1334          * @return null|string
1335          */
1336         static function hexCharReference( $codepoint ) {
1337                 $point = hexdec( $codepoint );
1338                 if( Sanitizer::validateCodepoint( $point ) ) {
1339                         return sprintf( '&#x%x;', $point );
1340                 } else {
1341                         return null;
1342                 }
1343         }
1344
1345         /**
1346          * Returns true if a given Unicode codepoint is a valid character in XML.
1347          * @param $codepoint Integer
1348          * @return Boolean
1349          */
1350         private static function validateCodepoint( $codepoint ) {
1351                 return ($codepoint ==    0x09)
1352                         || ($codepoint ==    0x0a)
1353                         || ($codepoint ==    0x0d)
1354                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1355                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1356                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1357         }
1358
1359         /**
1360          * Decode any character references, numeric or named entities,
1361          * in the text and return a UTF-8 string.
1362          *
1363          * @param $text String
1364          * @return String
1365          */
1366         public static function decodeCharReferences( $text ) {
1367                 return preg_replace_callback(
1368                         self::CHAR_REFS_REGEX,
1369                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1370                         $text );
1371         }
1372
1373         /**
1374          * Decode any character references, numeric or named entities,
1375          * in the next and normalize the resulting string. (bug 14952)
1376          *
1377          * This is useful for page titles, not for text to be displayed,
1378          * MediaWiki allows HTML entities to escape normalization as a feature.
1379          *
1380          * @param $text String (already normalized, containing entities)
1381          * @return String (still normalized, without entities)
1382          */
1383         public static function decodeCharReferencesAndNormalize( $text ) {
1384                 global $wgContLang;
1385                 $text = preg_replace_callback(
1386                         self::CHAR_REFS_REGEX,
1387                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1388                         $text, /* limit */ -1, $count );
1389
1390                 if ( $count ) {
1391                         return $wgContLang->normalize( $text );
1392                 } else {
1393                         return $text;
1394                 }
1395         }
1396
1397         /**
1398          * @param $matches String
1399          * @return String
1400          */
1401         static function decodeCharReferencesCallback( $matches ) {
1402                 if( $matches[1] != '' ) {
1403                         return Sanitizer::decodeEntity( $matches[1] );
1404                 } elseif( $matches[2] != '' ) {
1405                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1406                 } elseif( $matches[3] != ''  ) {
1407                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1408                 }
1409                 # Last case should be an ampersand by itself
1410                 return $matches[0];
1411         }
1412
1413         /**
1414          * Return UTF-8 string for a codepoint if that is a valid
1415          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1416          * @param $codepoint Integer
1417          * @return String
1418          * @private
1419          */
1420         static function decodeChar( $codepoint ) {
1421                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1422                         return codepointToUtf8( $codepoint );
1423                 } else {
1424                         return UTF8_REPLACEMENT;
1425                 }
1426         }
1427
1428         /**
1429          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1430          * return the UTF-8 encoding of that character. Otherwise, returns
1431          * pseudo-entity source (eg "&foo;")
1432          *
1433          * @param $name String
1434          * @return String
1435          */
1436         static function decodeEntity( $name ) {
1437                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1438                         $name = self::$htmlEntityAliases[$name];
1439                 }
1440                 if( isset( self::$htmlEntities[$name] ) ) {
1441                         return codepointToUtf8( self::$htmlEntities[$name] );
1442                 } else {
1443                         return "&$name;";
1444                 }
1445         }
1446
1447         /**
1448          * Fetch the whitelist of acceptable attributes for a given element name.
1449          *
1450          * @param $element String
1451          * @return Array
1452          */
1453         static function attributeWhitelist( $element ) {
1454                 static $list;
1455                 if( !isset( $list ) ) {
1456                         $list = Sanitizer::setupAttributeWhitelist();
1457                 }
1458                 return isset( $list[$element] )
1459                         ? $list[$element]
1460                         : array();
1461         }
1462
1463         /**
1464          * Foreach array key (an allowed HTML element), return an array
1465          * of allowed attributes
1466          * @return Array
1467          */
1468         static function setupAttributeWhitelist() {
1469                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1470
1471                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1472
1473                 if ( $wgAllowRdfaAttributes ) {
1474                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1475                         $common = array_merge( $common, array(
1476                             'about', 'property', 'resource', 'datatype', 'typeof',
1477                         ) );
1478                 }
1479
1480                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1481                         # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1482                         $common = array_merge( $common, array(
1483                             'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1484                         ) );
1485                 }
1486
1487                 $block = array_merge( $common, array( 'align' ) );
1488                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1489                 $tablecell = array( 'abbr',
1490                                     'axis',
1491                                     'headers',
1492                                     'scope',
1493                                     'rowspan',
1494                                     'colspan',
1495                                     'nowrap', # deprecated
1496                                     'width',  # deprecated
1497                                     'height', # deprecated
1498                                     'bgcolor' # deprecated
1499                                     );
1500
1501                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1502                 # See: http://www.w3.org/TR/html4/
1503                 $whitelist = array (
1504                         # 7.5.4
1505                         'div'        => $block,
1506                         'center'     => $common, # deprecated
1507                         'span'       => $block, # ??
1508
1509                         # 7.5.5
1510                         'h1'         => $block,
1511                         'h2'         => $block,
1512                         'h3'         => $block,
1513                         'h4'         => $block,
1514                         'h5'         => $block,
1515                         'h6'         => $block,
1516
1517                         # 7.5.6
1518                         # address
1519
1520                         # 8.2.4
1521                         # bdo
1522
1523                         # 9.2.1
1524                         'em'         => $common,
1525                         'strong'     => $common,
1526                         'cite'       => $common,
1527                         'dfn'        => $common,
1528                         'code'       => $common,
1529                         'samp'       => $common,
1530                         'kbd'        => $common,
1531                         'var'        => $common,
1532                         'abbr'       => $common,
1533                         # acronym
1534
1535                         # 9.2.2
1536                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1537                         # q
1538
1539                         # 9.2.3
1540                         'sub'        => $common,
1541                         'sup'        => $common,
1542
1543                         # 9.3.1
1544                         'p'          => $block,
1545
1546                         # 9.3.2
1547                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1548
1549                         # 9.3.4
1550                         'pre'        => array_merge( $common, array( 'width' ) ),
1551
1552                         # 9.4
1553                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1554                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1555
1556                         # 10.2
1557                         'ul'         => array_merge( $common, array( 'type' ) ),
1558                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1559                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1560
1561                         # 10.3
1562                         'dl'         => $common,
1563                         'dd'         => $common,
1564                         'dt'         => $common,
1565
1566                         # 11.2.1
1567                         'table'      => array_merge( $common,
1568                                                                 array( 'summary', 'width', 'border', 'frame',
1569                                                                                 'rules', 'cellspacing', 'cellpadding',
1570                                                                                 'align', 'bgcolor',
1571                                                                 ) ),
1572
1573                         # 11.2.2
1574                         'caption'    => array_merge( $common, array( 'align' ) ),
1575
1576                         # 11.2.3
1577                         'thead'      => array_merge( $common, $tablealign ),
1578                         'tfoot'      => array_merge( $common, $tablealign ),
1579                         'tbody'      => array_merge( $common, $tablealign ),
1580
1581                         # 11.2.4
1582                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1583                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1584
1585                         # 11.2.5
1586                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1587
1588                         # 11.2.6
1589                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1590                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1591
1592                         # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1593                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1594
1595                         # 13.2
1596                         # Not usually allowed, but may be used for extension-style hooks
1597                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1598                         # true
1599                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1600
1601                         # 15.2.1
1602                         'tt'         => $common,
1603                         'b'          => $common,
1604                         'i'          => $common,
1605                         'big'        => $common,
1606                         'small'      => $common,
1607                         'strike'     => $common,
1608                         's'          => $common,
1609                         'u'          => $common,
1610
1611                         # 15.2.2
1612                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1613                         # basefont
1614
1615                         # 15.3
1616                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1617
1618                         # XHTML Ruby annotation text module, simple ruby only.
1619                         # http://www.w3c.org/TR/ruby/
1620                         'ruby'       => $common,
1621                         # rbc
1622                         # rtc
1623                         'rb'         => $common,
1624                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1625                         'rp'         => $common,
1626
1627                         # MathML root element, where used for extensions
1628                         # 'title' may not be 100% valid here; it's XHTML
1629                         # http://www.w3.org/TR/REC-MathML/
1630                         'math'       => array( 'class', 'style', 'id', 'title' ),
1631
1632                         # HTML 5 section 4.6
1633                         'bdi' => $common,
1634
1635                         );
1636                 return $whitelist;
1637         }
1638
1639         /**
1640          * Take a fragment of (potentially invalid) HTML and return
1641          * a version with any tags removed, encoded as plain text.
1642          *
1643          * Warning: this return value must be further escaped for literal
1644          * inclusion in HTML output as of 1.10!
1645          *
1646          * @param $text String: HTML fragment
1647          * @return String
1648          */
1649         static function stripAllTags( $text ) {
1650                 # Actual <tags>
1651                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1652
1653                 # Normalize &entities and whitespace
1654                 $text = self::decodeCharReferences( $text );
1655                 $text = self::normalizeWhitespace( $text );
1656
1657                 return $text;
1658         }
1659
1660         /**
1661          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1662          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1663          * PHP 5.1 doesn't.
1664          *
1665          * Use for passing XHTML fragments to PHP's XML parsing functions
1666          *
1667          * @return String
1668          */
1669         static function hackDocType() {
1670                 $out = "<!DOCTYPE html [\n";
1671                 foreach( self::$htmlEntities as $entity => $codepoint ) {
1672                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1673                 }
1674                 $out .= "]>\n";
1675                 return $out;
1676         }
1677
1678         /**
1679          * @param $url string
1680          * @return mixed|string
1681          */
1682         static function cleanUrl( $url ) {
1683                 # Normalize any HTML entities in input. They will be
1684                 # re-escaped by makeExternalLink().
1685                 $url = Sanitizer::decodeCharReferences( $url );
1686
1687                 # Escape any control characters introduced by the above step
1688                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1689                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1690
1691                 # Validate hostname portion
1692                 $matches = array();
1693                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1694                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1695
1696                         // Characters that will be ignored in IDNs.
1697                         // http://tools.ietf.org/html/3454#section-3.1
1698                         // Strip them before further processing so blacklists and such work.
1699                         $strip = "/
1700                                 \\s|          # general whitespace
1701                                 \xc2\xad|     # 00ad SOFT HYPHEN
1702                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1703                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1704                                 \xe2\x81\xa0| # 2060 WORD JOINER
1705                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1706                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1707                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1708                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1709                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1710                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1711                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1712                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1713                                 /xuD";
1714
1715                         $host = preg_replace( $strip, '', $host );
1716
1717                         // @todo FIXME: Validate hostnames here
1718
1719                         return $protocol . $host . $rest;
1720                 } else {
1721                         return $url;
1722                 }
1723         }
1724
1725         /**
1726          * @param $matches array
1727          * @return string
1728          */
1729         static function cleanUrlCallback( $matches ) {
1730                 return urlencode( $matches[0] );
1731         }
1732
1733         /**
1734          * Does a string look like an e-mail address?
1735          *
1736          * This validates an email address using an HTML5 specification found at:
1737          * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1738          * Which as of 2011-01-24 says:
1739          *
1740          *   A valid e-mail address is a string that matches the ABNF production
1741          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1742          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1743          *   3.5.
1744          *
1745          * This function is an implementation of the specification as requested in
1746          * bug 22449.
1747          *
1748          * Client-side forms will use the same standard validation rules via JS or
1749          * HTML 5 validation; additional restrictions can be enforced server-side
1750          * by extensions via the 'isValidEmailAddr' hook.
1751          *
1752          * Note that this validation doesn't 100% match RFC 2822, but is believed
1753          * to be liberal enough for wide use. Some invalid addresses will still
1754          * pass validation here.
1755          *
1756          * @since 1.18
1757          *
1758          * @param $addr String E-mail address
1759          * @return Bool
1760          */
1761         public static function validateEmail( $addr ) {
1762                 $result = null;
1763                 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1764                         return $result;
1765                 }
1766
1767                 // Please note strings below are enclosed in brackets [], this make the
1768                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1769                 // See bug 26948
1770                 $rfc5322_atext   = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1771                 $rfc1034_ldh_str = "a-z0-9\\-" ;
1772
1773                 $HTML5_email_regexp = "/
1774                 ^                      # start of string
1775                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1776                 @                      # 'apostrophe'
1777                 [$rfc1034_ldh_str]+       # First domain part
1778                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1779                 $                      # End of string
1780                 /ix" ; // case Insensitive, eXtended
1781
1782                 return (bool) preg_match( $HTML5_email_regexp, $addr );
1783         }
1784 }