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         const EVIL_URI_PATTERN = '!(^|\s|\*/\s*)(javascript|vbscript)([^\w]|$)!i';
  43         const XMLNS_ATTRIBUTE_PATTERN = "/^xmlns:[:A-Z_a-z-.0-9]+$/";
  44
  45         /**
  46          * List of all named character entities defined in HTML 4.01
  47          * http://www.w3.org/TR/html4/sgml/entities.html
  48          * As well as &apos; which is only defined starting in XHTML1.
  49          * @private
  50          */
  51         static $htmlEntities = array(
  52                 'Aacute'   => 193,
  53                 'aacute'   => 225,
  54                 'Acirc'    => 194,
  55                 'acirc'    => 226,
  56                 'acute'    => 180,
  57                 'AElig'    => 198,
  58                 'aelig'    => 230,
  59                 'Agrave'   => 192,
  60                 'agrave'   => 224,
  61                 'alefsym'  => 8501,
  62                 'Alpha'    => 913,
  63                 'alpha'    => 945,
  64                 'amp'      => 38,
  65                 'and'      => 8743,
  66                 'ang'      => 8736,
  67                 'apos'     => 39,
  68                 'Aring'    => 197,
  69                 'aring'    => 229,
  70                 'asymp'    => 8776,
  71                 'Atilde'   => 195,
  72                 'atilde'   => 227,
  73                 'Auml'     => 196,
  74                 'auml'     => 228,
  75                 'bdquo'    => 8222,
  76                 'Beta'     => 914,
  77                 'beta'     => 946,
  78                 'brvbar'   => 166,
  79                 'bull'     => 8226,
  80                 'cap'      => 8745,
  81                 'Ccedil'   => 199,
  82                 'ccedil'   => 231,
  83                 'cedil'    => 184,
  84                 'cent'     => 162,
  85                 'Chi'      => 935,
  86                 'chi'      => 967,
  87                 'circ'     => 710,
  88                 'clubs'    => 9827,
  89                 'cong'     => 8773,
  90                 'copy'     => 169,
  91                 'crarr'    => 8629,
  92                 'cup'      => 8746,
  93                 'curren'   => 164,
  94                 'dagger'   => 8224,
  95                 'Dagger'   => 8225,
  96                 'darr'     => 8595,
  97                 'dArr'     => 8659,
  98                 'deg'      => 176,
  99                 'Delta'    => 916,
 100                 'delta'    => 948,
 101                 'diams'    => 9830,
 102                 'divide'   => 247,
 103                 'Eacute'   => 201,
 104                 'eacute'   => 233,
 105                 'Ecirc'    => 202,
 106                 'ecirc'    => 234,
 107                 'Egrave'   => 200,
 108                 'egrave'   => 232,
 109                 'empty'    => 8709,
 110                 'emsp'     => 8195,
 111                 'ensp'     => 8194,
 112                 'Epsilon'  => 917,
 113                 'epsilon'  => 949,
 114                 'equiv'    => 8801,
 115                 'Eta'      => 919,
 116                 'eta'      => 951,
 117                 'ETH'      => 208,
 118                 'eth'      => 240,
 119                 'Euml'     => 203,
 120                 'euml'     => 235,
 121                 'euro'     => 8364,
 122                 'exist'    => 8707,
 123                 'fnof'     => 402,
 124                 'forall'   => 8704,
 125                 'frac12'   => 189,
 126                 'frac14'   => 188,
 127                 'frac34'   => 190,
 128                 'frasl'    => 8260,
 129                 'Gamma'    => 915,
 130                 'gamma'    => 947,
 131                 'ge'       => 8805,
 132                 'gt'       => 62,
 133                 'harr'     => 8596,
 134                 'hArr'     => 8660,
 135                 'hearts'   => 9829,
 136                 'hellip'   => 8230,
 137                 'Iacute'   => 205,
 138                 'iacute'   => 237,
 139                 'Icirc'    => 206,
 140                 'icirc'    => 238,
 141                 'iexcl'    => 161,
 142                 'Igrave'   => 204,
 143                 'igrave'   => 236,
 144                 'image'    => 8465,
 145                 'infin'    => 8734,
 146                 'int'      => 8747,
 147                 'Iota'     => 921,
 148                 'iota'     => 953,
 149                 'iquest'   => 191,
 150                 'isin'     => 8712,
 151                 'Iuml'     => 207,
 152                 'iuml'     => 239,
 153                 'Kappa'    => 922,
 154                 'kappa'    => 954,
 155                 'Lambda'   => 923,
 156                 'lambda'   => 955,
 157                 'lang'     => 9001,
 158                 'laquo'    => 171,
 159                 'larr'     => 8592,
 160                 'lArr'     => 8656,
 161                 'lceil'    => 8968,
 162                 'ldquo'    => 8220,
 163                 'le'       => 8804,
 164                 'lfloor'   => 8970,
 165                 'lowast'   => 8727,
 166                 'loz'      => 9674,
 167                 'lrm'      => 8206,
 168                 'lsaquo'   => 8249,
 169                 'lsquo'    => 8216,
 170                 'lt'       => 60,
 171                 'macr'     => 175,
 172                 'mdash'    => 8212,
 173                 'micro'    => 181,
 174                 'middot'   => 183,
 175                 'minus'    => 8722,
 176                 'Mu'       => 924,
 177                 'mu'       => 956,
 178                 'nabla'    => 8711,
 179                 'nbsp'     => 160,
 180                 'ndash'    => 8211,
 181                 'ne'       => 8800,
 182                 'ni'       => 8715,
 183                 'not'      => 172,
 184                 'notin'    => 8713,
 185                 'nsub'     => 8836,
 186                 'Ntilde'   => 209,
 187                 'ntilde'   => 241,
 188                 'Nu'       => 925,
 189                 'nu'       => 957,
 190                 'Oacute'   => 211,
 191                 'oacute'   => 243,
 192                 'Ocirc'    => 212,
 193                 'ocirc'    => 244,
 194                 'OElig'    => 338,
 195                 'oelig'    => 339,
 196                 'Ograve'   => 210,
 197                 'ograve'   => 242,
 198                 'oline'    => 8254,
 199                 'Omega'    => 937,
 200                 'omega'    => 969,
 201                 'Omicron'  => 927,
 202                 'omicron'  => 959,
 203                 'oplus'    => 8853,
 204                 'or'       => 8744,
 205                 'ordf'     => 170,
 206                 'ordm'     => 186,
 207                 'Oslash'   => 216,
 208                 'oslash'   => 248,
 209                 'Otilde'   => 213,
 210                 'otilde'   => 245,
 211                 'otimes'   => 8855,
 212                 'Ouml'     => 214,
 213                 'ouml'     => 246,
 214                 'para'     => 182,
 215                 'part'     => 8706,
 216                 'permil'   => 8240,
 217                 'perp'     => 8869,
 218                 'Phi'      => 934,
 219                 'phi'      => 966,
 220                 'Pi'       => 928,
 221                 'pi'       => 960,
 222                 'piv'      => 982,
 223                 'plusmn'   => 177,
 224                 'pound'    => 163,
 225                 'prime'    => 8242,
 226                 'Prime'    => 8243,
 227                 'prod'     => 8719,
 228                 'prop'     => 8733,
 229                 'Psi'      => 936,
 230                 'psi'      => 968,
 231                 'quot'     => 34,
 232                 'radic'    => 8730,
 233                 'rang'     => 9002,
 234                 'raquo'    => 187,
 235                 'rarr'     => 8594,
 236                 'rArr'     => 8658,
 237                 'rceil'    => 8969,
 238                 'rdquo'    => 8221,
 239                 'real'     => 8476,
 240                 'reg'      => 174,
 241                 'rfloor'   => 8971,
 242                 'Rho'      => 929,
 243                 'rho'      => 961,
 244                 'rlm'      => 8207,
 245                 'rsaquo'   => 8250,
 246                 'rsquo'    => 8217,
 247                 'sbquo'    => 8218,
 248                 'Scaron'   => 352,
 249                 'scaron'   => 353,
 250                 'sdot'     => 8901,
 251                 'sect'     => 167,
 252                 'shy'      => 173,
 253                 'Sigma'    => 931,
 254                 'sigma'    => 963,
 255                 'sigmaf'   => 962,
 256                 'sim'      => 8764,
 257                 'spades'   => 9824,
 258                 'sub'      => 8834,
 259                 'sube'     => 8838,
 260                 'sum'      => 8721,
 261                 'sup'      => 8835,
 262                 'sup1'     => 185,
 263                 'sup2'     => 178,
 264                 'sup3'     => 179,
 265                 'supe'     => 8839,
 266                 'szlig'    => 223,
 267                 'Tau'      => 932,
 268                 'tau'      => 964,
 269                 'there4'   => 8756,
 270                 'Theta'    => 920,
 271                 'theta'    => 952,
 272                 'thetasym' => 977,
 273                 'thinsp'   => 8201,
 274                 'THORN'    => 222,
 275                 'thorn'    => 254,
 276                 'tilde'    => 732,
 277                 'times'    => 215,
 278                 'trade'    => 8482,
 279                 'Uacute'   => 218,
 280                 'uacute'   => 250,
 281                 'uarr'     => 8593,
 282                 'uArr'     => 8657,
 283                 'Ucirc'    => 219,
 284                 'ucirc'    => 251,
 285                 'Ugrave'   => 217,
 286                 'ugrave'   => 249,
 287                 'uml'      => 168,
 288                 'upsih'    => 978,
 289                 'Upsilon'  => 933,
 290                 'upsilon'  => 965,
 291                 'Uuml'     => 220,
 292                 'uuml'     => 252,
 293                 'weierp'   => 8472,
 294                 'Xi'       => 926,
 295                 'xi'       => 958,
 296                 'Yacute'   => 221,
 297                 'yacute'   => 253,
 298                 'yen'      => 165,
 299                 'Yuml'     => 376,
 300                 'yuml'     => 255,
 301                 'Zeta'     => 918,
 302                 'zeta'     => 950,
 303                 'zwj'      => 8205,
 304                 'zwnj'     => 8204
 305         );
 306
 307         /**
 308          * Character entity aliases accepted by MediaWiki
 309          */
 310         static $htmlEntityAliases = array(
 311                 'רלמ' => 'rlm',
 312                 'رلم' => 'rlm',
 313         );
 314
 315         /**
 316          * Lazy-initialised attributes regex, see getAttribsRegex()
 317          */
 318         static $attribsRegex;
 319
 320         /**
 321          * Regular expression to match HTML/XML attribute pairs within a tag.
 322          * Allows some... latitude.
 323          * Used in Sanitizer::fixTagAttributes and Sanitizer::decodeTagAttributes
 324          */
 325         static function getAttribsRegex() {
 326                 if ( self::$attribsRegex === null ) {
 327                         $attribFirst = '[:A-Z_a-z0-9]';
 328                         $attrib = '[:A-Z_a-z-.0-9]';
 329                         $space = '[\x09\x0a\x0d\x20]';
 330                         self::$attribsRegex =
 331                                 "/(?:^|$space)({$attribFirst}{$attrib}*)
 332                                   ($space*=$space*
 333                                         (?:
 334                                          # The attribute value: quoted or alone
 335                                           \"([^<\"]*)\"
 336                                          | '([^<']*)'
 337                                          |  ([a-zA-Z0-9!#$%&()*,\\-.\\/:;<>?@[\\]^_`{|}~]+)
 338                                          |  (\#[0-9a-fA-F]+) # Technically wrong, but lots of
 339                                                                                  # colors are specified like this.
 340                                                                                  # We'll be normalizing it.
 341                                         )
 342                                 )?(?=$space|\$)/sx";
 343                 }
 344                 return self::$attribsRegex;
 345         }
 346
 347         /**
 348          * Cleans up HTML, removes dangerous tags and attributes, and
 349          * removes HTML comments
 350          * @private
 351          * @param $text String
 352          * @param $processCallback Callback to do any variable or parameter replacements in HTML attribute values
 353          * @param $args Array for the processing callback
 354          * @param $extratags Array for any extra tags to include
 355          * @param $removetags Array for any tags (default or extra) to exclude
 356          * @return string
 357          */
 358         static function removeHTMLtags( $text, $processCallback = null, $args = array(), $extratags = array(), $removetags = array() ) {
 359                 global $wgUseTidy;
 360
 361                 static $htmlpairsStatic, $htmlsingle, $htmlsingleonly, $htmlnest, $tabletags,
 362                         $htmllist, $listtags, $htmlsingleallowed, $htmlelementsStatic, $staticInitialised;
 363
 364                 wfProfileIn( __METHOD__ );
 365
 366                 if ( !$staticInitialised ) {
 367
 368                         $htmlpairsStatic = array( # Tags that must be closed
 369                                 'b', 'del', 'i', 'ins', 'u', 'font', 'big', 'small', 'sub', 'sup', 'h1',
 370                                 'h2', 'h3', 'h4', 'h5', 'h6', 'cite', 'code', 'em', 's',
 371                                 'strike', 'strong', 'tt', 'var', 'div', 'center',
 372                                 'blockquote', 'ol', 'ul', 'dl', 'table', 'caption', 'pre',
 373                                 'ruby', 'rt' , 'rb' , 'rp', 'p', 'span', 'abbr', 'dfn',
 374                                 'kbd', 'samp', 'thead', 'tbody', 'tfoot'
 375                         );
 376                         $htmlsingle = array(
 377                                 'br', 'hr', 'li', 'dt', 'dd'
 378                         );
 379                         $htmlsingleonly = array( # Elements that cannot have close tags
 380                                 'br', 'hr'
 381                         );
 382                         $htmlnest = array( # Tags that can be nested--??
 383                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 384                                 'dl', 'font', 'big', 'small', 'sub', 'sup', 'span'
 385                         );
 386                         $tabletags = array( # Can only appear inside table, we will close them
 387                                 'td', 'th', 'tr',
 388                         );
 389                         $htmllist = array( # Tags used by list
 390                                 'ul','ol',
 391                         );
 392                         $listtags = array( # Tags that can appear in a list
 393                                 'li',
 394                         );
 395
 396                         global $wgAllowImageTag;
 397                         if ( $wgAllowImageTag ) {
 398                                 $htmlsingle[] = 'img';
 399                                 $htmlsingleonly[] = 'img';
 400                         }
 401
 402                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 403                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 404
 405                         # Convert them all to hashtables for faster lookup
 406                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 407                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 408                         foreach ( $vars as $var ) {
 409                                 $$var = array_flip( $$var );
 410                         }
 411                         $staticInitialised = true;
 412                 }
 413                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 414                 $extratags = array_flip( $extratags );
 415                 $removetags = array_flip( $removetags );
 416                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 417                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ) , $removetags );
 418
 419                 # Remove HTML comments
 420                 $text = Sanitizer::removeHTMLcomments( $text );
 421                 $bits = explode( '<', $text );
 422                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 423                 if ( !$wgUseTidy ) {
 424                         $tagstack = $tablestack = array();
 425                         foreach ( $bits as $x ) {
 426                                 $regs = array();
 427                                 # $slash: Does the current element start with a '/'?
 428                                 # $t: Current element name
 429                                 # $params: String between element name and >
 430                                 # $brace: Ending '>' or '/>'
 431                                 # $rest: Everything until the next element of $bits
 432                                 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 433                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 434                                 } else {
 435                                         $slash = $t = $params = $brace = $rest = null;
 436                                 }
 437
 438                                 $badtag = false;
 439                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 440                                         # Check our stack
 441                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 442                                                 $badtag = true;
 443                                         } elseif ( $slash ) {
 444                                                 # Closing a tag... is it the one we just opened?
 445                                                 $ot = @array_pop( $tagstack );
 446                                                 if ( $ot != $t ) {
 447                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 448                                                                 # Pop all elements with an optional close tag
 449                                                                 # and see if we find a match below them
 450                                                                 $optstack = array();
 451                                                                 array_push( $optstack, $ot );
 452                                                                 $ot = @array_pop( $tagstack );
 453                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 454                                                                         array_push( $optstack, $ot );
 455                                                                         $ot = @array_pop( $tagstack );
 456                                                                 }
 457                                                                 if ( $t != $ot ) {
 458                                                                         # No match. Push the optional elements back again
 459                                                                         $badtag = true;
 460                                                                         while ( $ot = @array_pop( $optstack ) ) {
 461                                                                                 array_push( $tagstack, $ot );
 462                                                                         }
 463                                                                 }
 464                                                         } else {
 465                                                                 @array_push( $tagstack, $ot );
 466                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 467                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 468                                                                         $badtag = true;
 469                                                                 }
 470                                                         }
 471                                                 } else {
 472                                                         if ( $t == 'table' ) {
 473                                                                 $tagstack = array_pop( $tablestack );
 474                                                         }
 475                                                 }
 476                                                 $newparams = '';
 477                                         } else {
 478                                                 # Keep track for later
 479                                                 if ( isset( $tabletags[$t] ) &&
 480                                                 !in_array( 'table', $tagstack ) ) {
 481                                                         $badtag = true;
 482                                                 } elseif ( in_array( $t, $tagstack ) &&
 483                                                 !isset( $htmlnest [$t ] ) ) {
 484                                                         $badtag = true;
 485                                                 # Is it a self closed htmlpair ? (bug 5487)
 486                                                 } elseif ( $brace == '/>' &&
 487                                                 isset( $htmlpairs[$t] ) ) {
 488                                                         $badtag = true;
 489                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 490                                                         # Hack to force empty tag for uncloseable elements
 491                                                         $brace = '/>';
 492                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 493                                                         # Hack to not close $htmlsingle tags
 494                                                         $brace = null;
 495                                                 } elseif ( isset( $tabletags[$t] )
 496                                                 && in_array( $t, $tagstack ) ) {
 497                                                         // New table tag but forgot to close the previous one
 498                                                         $text .= "</$t>";
 499                                                 } else {
 500                                                         if ( $t == 'table' ) {
 501                                                                 array_push( $tablestack, $tagstack );
 502                                                                 $tagstack = array();
 503                                                         }
 504                                                         array_push( $tagstack, $t );
 505                                                 }
 506
 507                                                 # Replace any variables or template parameters with
 508                                                 # plaintext results.
 509                                                 if( is_callable( $processCallback ) ) {
 510                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 511                                                 }
 512
 513                                                 # Strip non-approved attributes from the tag
 514                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 515                                         }
 516                                         if ( !$badtag ) {
 517                                                 $rest = str_replace( '>', '&gt;', $rest );
 518                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 519                                                 $text .= "<$slash$t$newparams$close>$rest";
 520                                                 continue;
 521                                         }
 522                                 }
 523                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 524                         }
 525                         # Close off any remaining tags
 526                         while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
 527                                 $text .= "</$t>\n";
 528                                 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
 529                         }
 530                 } else {
 531                         # this might be possible using tidy itself
 532                         foreach ( $bits as $x ) {
 533                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 534                                 $x, $regs );
 535                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 536                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 537                                         if( is_callable( $processCallback ) ) {
 538                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 539                                         }
 540                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 541                                         $rest = str_replace( '>', '&gt;', $rest );
 542                                         $text .= "<$slash$t$newparams$brace$rest";
 543                                 } else {
 544                                         $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 545                                 }
 546                         }
 547                 }
 548                 wfProfileOut( __METHOD__ );
 549                 return $text;
 550         }
 551
 552         /**
 553          * Remove '<!--', '-->', and everything between.
 554          * To avoid leaving blank lines, when a comment is both preceded
 555          * and followed by a newline (ignoring spaces), trim leading and
 556          * trailing spaces and one of the newlines.
 557          *
 558          * @private
 559          * @param $text String
 560          * @return string
 561          */
 562         static function removeHTMLcomments( $text ) {
 563                 wfProfileIn( __METHOD__ );
 564                 while (($start = strpos($text, '<!--')) !== false) {
 565                         $end = strpos($text, '-->', $start + 4);
 566                         if ($end === false) {
 567                                 # Unterminated comment; bail out
 568                                 break;
 569                         }
 570
 571                         $end += 3;
 572
 573                         # Trim space and newline if the comment is both
 574                         # preceded and followed by a newline
 575                         $spaceStart = max($start - 1, 0);
 576                         $spaceLen = $end - $spaceStart;
 577                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 578                                 $spaceStart--;
 579                                 $spaceLen++;
 580                         }
 581                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 582                                 $spaceLen++;
 583                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 584                                 # Remove the comment, leading and trailing
 585                                 # spaces, and leave only one newline.
 586                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 587                         }
 588                         else {
 589                                 # Remove just the comment.
 590                                 $text = substr_replace($text, '', $start, $end - $start);
 591                         }
 592                 }
 593                 wfProfileOut( __METHOD__ );
 594                 return $text;
 595         }
 596
 597         /**
 598          * Take an array of attribute names and values and normalize or discard
 599          * illegal values for the given element type.
 600          *
 601          * - Discards attributes not on a whitelist for the given element
 602          * - Unsafe style attributes are discarded
 603          * - Invalid id attributes are reencoded
 604          *
 605          * @param $attribs Array
 606          * @param $element String
 607          * @return Array
 608          *
 609          * @todo Check for legal values where the DTD limits things.
 610          * @todo Check for unique id attribute :P
 611          */
 612         static function validateTagAttributes( $attribs, $element ) {
 613                 return Sanitizer::validateAttributes( $attribs,
 614                         Sanitizer::attributeWhitelist( $element ) );
 615         }
 616
 617         /**
 618          * Take an array of attribute names and values and normalize or discard
 619          * illegal values for the given whitelist.
 620          *
 621          * - Discards attributes not the given whitelist
 622          * - Unsafe style attributes are discarded
 623          * - Invalid id attributes are reencoded
 624          *
 625          * @param $attribs Array
 626          * @param $whitelist Array: list of allowed attribute names
 627          * @return Array
 628          *
 629          * @todo Check for legal values where the DTD limits things.
 630          * @todo Check for unique id attribute :P
 631          */
 632         static function validateAttributes( $attribs, $whitelist ) {
 633                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
 634
 635                 $whitelist = array_flip( $whitelist );
 636                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 637
 638                 $out = array();
 639                 foreach( $attribs as $attribute => $value ) {
 640                         #allow XML namespace declaration if RDFa is enabled
 641                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 642                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 643                                         $out[$attribute] = $value;
 644                                 }
 645
 646                                 continue;
 647                         }
 648
 649                         # Allow any attribute beginning with "data-", if in HTML5 mode
 650                         if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
 651                                 continue;
 652                         }
 653
 654                         # Strip javascript "expression" from stylesheets.
 655                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 656                         if( $attribute == 'style' ) {
 657                                 $value = Sanitizer::checkCss( $value );
 658                         }
 659
 660                         if ( $attribute === 'id' ) {
 661                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 662                         }
 663
 664                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 665                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 666                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 667                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 668                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 669                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 670
 671                                 //Paranoia. Allow "simple" values but suppress javascript
 672                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 673                                         continue;
 674                                 }
 675                         }
 676
 677                         # NOTE: even though elements using href/src are not allowed directly, supply
 678                         #       validation code that can be used by tag hook handlers, etc
 679                         if ( $attribute === 'href' || $attribute === 'src' ) {
 680                                 if ( !preg_match( $hrefExp, $value ) ) {
 681                                         continue; //drop any href or src attributes not using an allowed protocol.
 682                                                   //NOTE: this also drops all relative URLs
 683                                 }
 684                         }
 685
 686                         // If this attribute was previously set, override it.
 687                         // Output should only have one attribute of each name.
 688                         $out[$attribute] = $value;
 689                 }
 690
 691                 if ( $wgAllowMicrodataAttributes ) {
 692                         # itemtype, itemid, itemref don't make sense without itemscope
 693                         if ( !array_key_exists( 'itemscope', $out ) ) {
 694                                 unset( $out['itemtype'] );
 695                                 unset( $out['itemid'] );
 696                                 unset( $out['itemref'] );
 697                         }
 698                         # TODO: Strip itemprop if we aren't descendants of an itemscope.
 699                 }
 700                 return $out;
 701         }
 702
 703         /**
 704          * Merge two sets of HTML attributes.  Conflicting items in the second set
 705          * will override those in the first, except for 'class' attributes which
 706          * will be combined (if they're both strings).
 707          *
 708          * @todo implement merging for other attributes such as style
 709          * @param $a Array
 710          * @param $b Array
 711          * @return array
 712          */
 713         static function mergeAttributes( $a, $b ) {
 714                 $out = array_merge( $a, $b );
 715                 if( isset( $a['class'] ) && isset( $b['class'] )
 716                 && is_string( $a['class'] ) && is_string( $b['class'] )
 717                 && $a['class'] !== $b['class'] ) {
 718                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 719                                 -1, PREG_SPLIT_NO_EMPTY );
 720                         $out['class'] = implode( ' ', array_unique( $classes ) );
 721                 }
 722                 return $out;
 723         }
 724
 725         /**
 726          * Pick apart some CSS and check it for forbidden or unsafe structures.
 727          * Returns a sanitized string. This sanitized string will have
 728          * character references and escape sequences decoded, and comments
 729          * stripped. If the input is just too evil, only a comment complaining
 730          * about evilness will be returned.
 731          *
 732          * Currently URL references, 'expression', 'tps' are forbidden.
 733          *
 734          * NOTE: Despite the fact that character references are decoded, the
 735          * returned string may contain character references given certain
 736          * clever input strings. These character references must
 737          * be escaped before the return value is embedded in HTML.
 738          *
 739          * @param $value String
 740          * @return String
 741          */
 742         static function checkCss( $value ) {
 743                 // Decode character references like &#123;
 744                 $value = Sanitizer::decodeCharReferences( $value );
 745
 746                 // Decode escape sequences and line continuation
 747                 // See the grammar in the CSS 2 spec, appendix D.
 748                 // This has to be done AFTER decoding character references.
 749                 // This means it isn't possible for this function to return
 750                 // unsanitized escape sequences. It is possible to manufacture
 751                 // input that contains character references that decode to
 752                 // escape sequences that decode to character references, but
 753                 // it's OK for the return value to contain character references
 754                 // because the caller is supposed to escape those anyway.
 755                 static $decodeRegex;
 756                 if ( !$decodeRegex ) {
 757                         $space = '[\\x20\\t\\r\\n\\f]';
 758                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 759                         $backslash = '\\\\';
 760                         $decodeRegex = "/ $backslash
 761                                 (?:
 762                                         ($nl) |  # 1. Line continuation
 763                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 764                                         (.) | # 3. backslash cancelling special meaning
 765                                         () | # 4. backslash at end of string
 766                                 )/xu";
 767                 }
 768                 $value = preg_replace_callback( $decodeRegex,
 769                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 770
 771                 // Remove any comments; IE gets token splitting wrong
 772                 // This must be done AFTER decoding character references and
 773                 // escape sequences, because those steps can introduce comments
 774                 // This step cannot introduce character references or escape
 775                 // sequences, because it replaces comments with spaces rather
 776                 // than removing them completely.
 777                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 778
 779                 // Remove anything after a comment-start token, to guard against
 780                 // incorrect client implementations.
 781                 $commentPos = strpos( $value, '/*' );
 782                 if ( $commentPos !== false ) {
 783                         $value = substr( $value, 0, $commentPos );
 784                 }
 785
 786                 // Reject problematic keywords and control characters
 787                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 788                         return '/* invalid control char */';
 789                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
 790                         return '/* insecure input */';
 791                 }
 792                 return $value;
 793         }
 794
 795         /**
 796          * @param $matches array
 797          * @return String
 798          */
 799         static function cssDecodeCallback( $matches ) {
 800                 if ( $matches[1] !== '' ) {
 801                         // Line continuation
 802                         return '';
 803                 } elseif ( $matches[2] !== '' ) {
 804                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 805                 } elseif ( $matches[3] !== '' ) {
 806                         $char = $matches[3];
 807                 } else {
 808                         $char = '\\';
 809                 }
 810                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 811                         // These characters need to be escaped in strings
 812                         // Clean up the escape sequence to avoid parsing errors by clients
 813                         return '\\' . dechex( ord( $char ) ) . ' ';
 814                 } else {
 815                         // Decode unnecessary escape
 816                         return $char;
 817                 }
 818         }
 819
 820         /**
 821          * Take a tag soup fragment listing an HTML element's attributes
 822          * and normalize it to well-formed XML, discarding unwanted attributes.
 823          * Output is safe for further wikitext processing, with escaping of
 824          * values that could trigger problems.
 825          *
 826          * - Normalizes attribute names to lowercase
 827          * - Discards attributes not on a whitelist for the given element
 828          * - Turns broken or invalid entities into plaintext
 829          * - Double-quotes all attribute values
 830          * - Attributes without values are given the name as attribute
 831          * - Double attributes are discarded
 832          * - Unsafe style attributes are discarded
 833          * - Prepends space if there are attributes.
 834          *
 835          * @param $text String
 836          * @param $element String
 837          * @return String
 838          */
 839         static function fixTagAttributes( $text, $element ) {
 840                 if( trim( $text ) == '' ) {
 841                         return '';
 842                 }
 843
 844                 $stripped = Sanitizer::validateTagAttributes(
 845                         Sanitizer::decodeTagAttributes( $text ), $element );
 846
 847                 $attribs = array();
 848                 foreach( $stripped as $attribute => $value ) {
 849                         $encAttribute = htmlspecialchars( $attribute );
 850                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 851
 852                         $attribs[] = "$encAttribute=\"$encValue\"";
 853                 }
 854                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 855         }
 856
 857         /**
 858          * Encode an attribute value for HTML output.
 859          * @param $text String
 860          * @return HTML-encoded text fragment
 861          */
 862         static function encodeAttribute( $text ) {
 863                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 864
 865                 // Whitespace is normalized during attribute decoding,
 866                 // so if we've been passed non-spaces we must encode them
 867                 // ahead of time or they won't be preserved.
 868                 $encValue = strtr( $encValue, array(
 869                         "\n" => '&#10;',
 870                         "\r" => '&#13;',
 871                         "\t" => '&#9;',
 872                 ) );
 873
 874                 return $encValue;
 875         }
 876
 877         /**
 878          * Encode an attribute value for HTML tags, with extra armoring
 879          * against further wiki processing.
 880          * @param $text String
 881          * @return HTML-encoded text fragment
 882          */
 883         static function safeEncodeAttribute( $text ) {
 884                 $encValue = Sanitizer::encodeAttribute( $text );
 885
 886                 # Templates and links may be expanded in later parsing,
 887                 # creating invalid or dangerous output. Suppress this.
 888                 $encValue = strtr( $encValue, array(
 889                         '<'    => '&lt;',   // This should never happen,
 890                         '>'    => '&gt;',   // we've received invalid input
 891                         '"'    => '&quot;', // which should have been escaped.
 892                         '{'    => '&#123;',
 893                         '['    => '&#91;',
 894                         "''"   => '&#39;&#39;',
 895                         'ISBN' => '&#73;SBN',
 896                         'RFC'  => '&#82;FC',
 897                         'PMID' => '&#80;MID',
 898                         '|'    => '&#124;',
 899                         '__'   => '&#95;_',
 900                 ) );
 901
 902                 # Stupid hack
 903                 $encValue = preg_replace_callback(
 904                         '/(' . wfUrlProtocols() . ')/',
 905                         array( 'Sanitizer', 'armorLinksCallback' ),
 906                         $encValue );
 907                 return $encValue;
 908         }
 909
 910         /**
 911          * Given a value, escape it so that it can be used in an id attribute and
 912          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
 913          * true, allowing anything but ASCII whitespace.  Otherwise it will use
 914          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
 915          * escaped with lots of dots.
 916          *
 917          * To ensure we don't have to bother escaping anything, we also strip ', ",
 918          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
 919          * We also strip # because it upsets IE, and % because it could be
 920          * ambiguous if it's part of something that looks like a percent escape
 921          * (which don't work reliably in fragments cross-browser).
 922          *
 923          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
 924          *                                                          in the id and
 925          *                                                          name attributes
 926          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
 927          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
 928          *   HTML5 definition of id attribute
 929          *
 930          * @param $id String: id to escape
 931          * @param $options Mixed: string or array of strings (default is array()):
 932          *   'noninitial': This is a non-initial fragment of an id, not a full id,
 933          *       so don't pay attention if the first character isn't valid at the
 934          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
 935          *       false.
 936          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
 937          *       if $wgExperimentalHtmlIds is used, so we can generate extra
 938          *       anchors and links won't break.
 939          * @return String
 940          */
 941         static function escapeId( $id, $options = array() ) {
 942                 global $wgHtml5, $wgExperimentalHtmlIds;
 943                 $options = (array)$options;
 944
 945                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
 946                         $id = Sanitizer::decodeCharReferences( $id );
 947                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
 948                         $id = trim( $id, '_' );
 949                         if ( $id === '' ) {
 950                                 # Must have been all whitespace to start with.
 951                                 return '_';
 952                         } else {
 953                                 return $id;
 954                         }
 955                 }
 956
 957                 # HTML4-style escaping
 958                 static $replace = array(
 959                         '%3A' => ':',
 960                         '%' => '.'
 961                 );
 962
 963                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
 964                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
 965
 966                 if ( !preg_match( '/^[a-zA-Z]/', $id )
 967                 && !in_array( 'noninitial', $options ) )  {
 968                         // Initial character must be a letter!
 969                         $id = "x$id";
 970                 }
 971                 return $id;
 972         }
 973
 974         /**
 975          * Given a value, escape it so that it can be used as a CSS class and
 976          * return it.
 977          *
 978          * @todo For extra validity, input should be validated UTF-8.
 979          *
 980          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
 981          *
 982          * @param $class String
 983          * @return String
 984          */
 985         static function escapeClass( $class ) {
 986                 // Convert ugly stuff to underscores and kill underscores in ugly places
 987                 return rtrim(preg_replace(
 988                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
 989                         '_',
 990                         $class ), '_');
 991         }
 992
 993         /**
 994          * Given HTML input, escape with htmlspecialchars but un-escape entites.
 995          * This allows (generally harmless) entities like &#160; to survive.
 996          *
 997          * @param $html String to escape
 998          * @return String: escaped input
 999          */
1000         static function escapeHtmlAllowEntities( $html ) {
1001                 $html = Sanitizer::decodeCharReferences( $html );
1002                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1003                 # hurt.
1004                 $html = htmlspecialchars( $html, ENT_QUOTES );
1005                 return $html;
1006         }
1007
1008         /**
1009          * Regex replace callback for armoring links against further processing.
1010          * @param $matches Array
1011          * @return string
1012          */
1013         private static function armorLinksCallback( $matches ) {
1014                 return str_replace( ':', '&#58;', $matches[1] );
1015         }
1016
1017         /**
1018          * Return an associative array of attribute names and values from
1019          * a partial tag string. Attribute names are forces to lowercase,
1020          * character references are decoded to UTF-8 text.
1021          *
1022          * @param $text String
1023          * @return Array
1024          */
1025         public static function decodeTagAttributes( $text ) {
1026                 if( trim( $text ) == '' ) {
1027                         return array();
1028                 }
1029
1030                 $attribs = array();
1031                 $pairs = array();
1032                 if( !preg_match_all(
1033                         self::getAttribsRegex(),
1034                         $text,
1035                         $pairs,
1036                         PREG_SET_ORDER ) ) {
1037                         return $attribs;
1038                 }
1039
1040                 foreach( $pairs as $set ) {
1041                         $attribute = strtolower( $set[1] );
1042                         $value = Sanitizer::getTagAttributeCallback( $set );
1043
1044                         // Normalize whitespace
1045                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1046                         $value = trim( $value );
1047
1048                         // Decode character references
1049                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1050                 }
1051                 return $attribs;
1052         }
1053
1054         /**
1055          * Pick the appropriate attribute value from a match set from the
1056          * attribs regex matches.
1057          *
1058          * @param $set Array
1059          * @return String
1060          */
1061         private static function getTagAttributeCallback( $set ) {
1062                 if( isset( $set[6] ) ) {
1063                         # Illegal #XXXXXX color with no quotes.
1064                         return $set[6];
1065                 } elseif( isset( $set[5] ) ) {
1066                         # No quotes.
1067                         return $set[5];
1068                 } elseif( isset( $set[4] ) ) {
1069                         # Single-quoted
1070                         return $set[4];
1071                 } elseif( isset( $set[3] ) ) {
1072                         # Double-quoted
1073                         return $set[3];
1074                 } elseif( !isset( $set[2] ) ) {
1075                         # In XHTML, attributes must have a value.
1076                         # For 'reduced' form, return explicitly the attribute name here.
1077                         return $set[1];
1078                 } else {
1079                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1080                 }
1081         }
1082
1083         /**
1084          * Normalize whitespace and character references in an XML source-
1085          * encoded text for an attribute value.
1086          *
1087          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1088          * but note that we're not returning the value, but are returning
1089          * XML source fragments that will be slapped into output.
1090          *
1091          * @param $text String
1092          * @return String
1093          */
1094         private static function normalizeAttributeValue( $text ) {
1095                 return str_replace( '"', '&quot;',
1096                         self::normalizeWhitespace(
1097                                 Sanitizer::normalizeCharReferences( $text ) ) );
1098         }
1099
1100         /**
1101          * @param $text string
1102          * @return mixed
1103          */
1104         private static function normalizeWhitespace( $text ) {
1105                 return preg_replace(
1106                         '/\r\n|[\x20\x0d\x0a\x09]/',
1107                         ' ',
1108                         $text );
1109         }
1110
1111         /**
1112          * Normalizes whitespace in a section name, such as might be returned
1113          * by Parser::stripSectionName(), for use in the id's that are used for
1114          * section links.
1115          *
1116          * @param $section String
1117          * @return String
1118          */
1119         static function normalizeSectionNameWhitespace( $section ) {
1120                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1121         }
1122
1123         /**
1124          * Ensure that any entities and character references are legal
1125          * for XML and XHTML specifically. Any stray bits will be
1126          * &amp;-escaped to result in a valid text fragment.
1127          *
1128          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1129          *   numericized (this way we're well-formed even without a DTD)
1130          * b. any numeric char refs must be legal chars, not invalid or forbidden
1131          * c. use &#x, not &#X
1132          * d. fix or reject non-valid attributes
1133          *
1134          * @param $text String
1135          * @return String
1136          * @private
1137          */
1138         static function normalizeCharReferences( $text ) {
1139                 return preg_replace_callback(
1140                         self::CHAR_REFS_REGEX,
1141                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1142                         $text );
1143         }
1144         /**
1145          * @param $matches String
1146          * @return String
1147          */
1148         static function normalizeCharReferencesCallback( $matches ) {
1149                 $ret = null;
1150                 if( $matches[1] != '' ) {
1151                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1152                 } elseif( $matches[2] != '' ) {
1153                         $ret = Sanitizer::decCharReference( $matches[2] );
1154                 } elseif( $matches[3] != ''  ) {
1155                         $ret = Sanitizer::hexCharReference( $matches[3] );
1156                 }
1157                 if( is_null( $ret ) ) {
1158                         return htmlspecialchars( $matches[0] );
1159                 } else {
1160                         return $ret;
1161                 }
1162         }
1163
1164         /**
1165          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1166          * return the equivalent numeric entity reference (except for the core &lt;
1167          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1168          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1169          * pseudo-entity source (eg &amp;foo;)
1170          *
1171          * @param $name String
1172          * @return String
1173          */
1174         static function normalizeEntity( $name ) {
1175                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1176                         return '&' . self::$htmlEntityAliases[$name] . ';';
1177                 } elseif ( in_array( $name,
1178                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1179                         return "&$name;";
1180                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1181                         return '&#' . self::$htmlEntities[$name] . ';';
1182                 } else {
1183                         return "&amp;$name;";
1184                 }
1185         }
1186
1187         /**
1188          * @param $codepoint
1189          * @return null|string
1190          */
1191         static function decCharReference( $codepoint ) {
1192                 $point = intval( $codepoint );
1193                 if( Sanitizer::validateCodepoint( $point ) ) {
1194                         return sprintf( '&#%d;', $point );
1195                 } else {
1196                         return null;
1197                 }
1198         }
1199
1200         /**
1201          * @param $codepoint
1202          * @return null|string
1203          */
1204         static function hexCharReference( $codepoint ) {
1205                 $point = hexdec( $codepoint );
1206                 if( Sanitizer::validateCodepoint( $point ) ) {
1207                         return sprintf( '&#x%x;', $point );
1208                 } else {
1209                         return null;
1210                 }
1211         }
1212
1213         /**
1214          * Returns true if a given Unicode codepoint is a valid character in XML.
1215          * @param $codepoint Integer
1216          * @return Boolean
1217          */
1218         private static function validateCodepoint( $codepoint ) {
1219                 return ($codepoint ==    0x09)
1220                         || ($codepoint ==    0x0a)
1221                         || ($codepoint ==    0x0d)
1222                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1223                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1224                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1225         }
1226
1227         /**
1228          * Decode any character references, numeric or named entities,
1229          * in the text and return a UTF-8 string.
1230          *
1231          * @param $text String
1232          * @return String
1233          */
1234         public static function decodeCharReferences( $text ) {
1235                 return preg_replace_callback(
1236                         self::CHAR_REFS_REGEX,
1237                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1238                         $text );
1239         }
1240
1241         /**
1242          * Decode any character references, numeric or named entities,
1243          * in the next and normalize the resulting string. (bug 14952)
1244          *
1245          * This is useful for page titles, not for text to be displayed,
1246          * MediaWiki allows HTML entities to escape normalization as a feature.
1247          *
1248          * @param $text String (already normalized, containing entities)
1249          * @return String (still normalized, without entities)
1250          */
1251         public static function decodeCharReferencesAndNormalize( $text ) {
1252                 global $wgContLang;
1253                 $text = preg_replace_callback(
1254                         self::CHAR_REFS_REGEX,
1255                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1256                         $text, /* limit */ -1, $count );
1257
1258                 if ( $count ) {
1259                         return $wgContLang->normalize( $text );
1260                 } else {
1261                         return $text;
1262                 }
1263         }
1264
1265         /**
1266          * @param $matches String
1267          * @return String
1268          */
1269         static function decodeCharReferencesCallback( $matches ) {
1270                 if( $matches[1] != '' ) {
1271                         return Sanitizer::decodeEntity( $matches[1] );
1272                 } elseif( $matches[2] != '' ) {
1273                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1274                 } elseif( $matches[3] != ''  ) {
1275                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1276                 }
1277                 # Last case should be an ampersand by itself
1278                 return $matches[0];
1279         }
1280
1281         /**
1282          * Return UTF-8 string for a codepoint if that is a valid
1283          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1284          * @param $codepoint Integer
1285          * @return String
1286          * @private
1287          */
1288         static function decodeChar( $codepoint ) {
1289                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1290                         return codepointToUtf8( $codepoint );
1291                 } else {
1292                         return UTF8_REPLACEMENT;
1293                 }
1294         }
1295
1296         /**
1297          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1298          * return the UTF-8 encoding of that character. Otherwise, returns
1299          * pseudo-entity source (eg &foo;)
1300          *
1301          * @param $name String
1302          * @return String
1303          */
1304         static function decodeEntity( $name ) {
1305                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1306                         $name = self::$htmlEntityAliases[$name];
1307                 }
1308                 if( isset( self::$htmlEntities[$name] ) ) {
1309                         return codepointToUtf8( self::$htmlEntities[$name] );
1310                 } else {
1311                         return "&$name;";
1312                 }
1313         }
1314
1315         /**
1316          * Fetch the whitelist of acceptable attributes for a given element name.
1317          *
1318          * @param $element String
1319          * @return Array
1320          */
1321         static function attributeWhitelist( $element ) {
1322                 static $list;
1323                 if( !isset( $list ) ) {
1324                         $list = Sanitizer::setupAttributeWhitelist();
1325                 }
1326                 return isset( $list[$element] )
1327                         ? $list[$element]
1328                         : array();
1329         }
1330
1331         /**
1332          * Foreach array key (an allowed HTML element), return an array
1333          * of allowed attributes
1334          * @return Array
1335          */
1336         static function setupAttributeWhitelist() {
1337                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1338
1339                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1340
1341                 if ( $wgAllowRdfaAttributes ) {
1342                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1343                         $common = array_merge( $common, array(
1344                             'about', 'property', 'resource', 'datatype', 'typeof',
1345                         ) );
1346                 }
1347
1348                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1349                         # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1350                         $common = array_merge( $common, array(
1351                             'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1352                         ) );
1353                 }
1354
1355                 $block = array_merge( $common, array( 'align' ) );
1356                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1357                 $tablecell = array( 'abbr',
1358                                     'axis',
1359                                     'headers',
1360                                     'scope',
1361                                     'rowspan',
1362                                     'colspan',
1363                                     'nowrap', # deprecated
1364                                     'width',  # deprecated
1365                                     'height', # deprecated
1366                                     'bgcolor' # deprecated
1367                                     );
1368
1369                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1370                 # See: http://www.w3.org/TR/html4/
1371                 $whitelist = array (
1372                         # 7.5.4
1373                         'div'        => $block,
1374                         'center'     => $common, # deprecated
1375                         'span'       => $block, # ??
1376
1377                         # 7.5.5
1378                         'h1'         => $block,
1379                         'h2'         => $block,
1380                         'h3'         => $block,
1381                         'h4'         => $block,
1382                         'h5'         => $block,
1383                         'h6'         => $block,
1384
1385                         # 7.5.6
1386                         # address
1387
1388                         # 8.2.4
1389                         # bdo
1390
1391                         # 9.2.1
1392                         'em'         => $common,
1393                         'strong'     => $common,
1394                         'cite'       => $common,
1395                         'dfn'        => $common,
1396                         'code'       => $common,
1397                         'samp'       => $common,
1398                         'kbd'        => $common,
1399                         'var'        => $common,
1400                         'abbr'       => $common,
1401                         # acronym
1402
1403                         # 9.2.2
1404                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1405                         # q
1406
1407                         # 9.2.3
1408                         'sub'        => $common,
1409                         'sup'        => $common,
1410
1411                         # 9.3.1
1412                         'p'          => $block,
1413
1414                         # 9.3.2
1415                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1416
1417                         # 9.3.4
1418                         'pre'        => array_merge( $common, array( 'width' ) ),
1419
1420                         # 9.4
1421                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1422                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1423
1424                         # 10.2
1425                         'ul'         => array_merge( $common, array( 'type' ) ),
1426                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1427                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1428
1429                         # 10.3
1430                         'dl'         => $common,
1431                         'dd'         => $common,
1432                         'dt'         => $common,
1433
1434                         # 11.2.1
1435                         'table'      => array_merge( $common,
1436                                                                 array( 'summary', 'width', 'border', 'frame',
1437                                                                                 'rules', 'cellspacing', 'cellpadding',
1438                                                                                 'align', 'bgcolor',
1439                                                                 ) ),
1440
1441                         # 11.2.2
1442                         'caption'    => array_merge( $common, array( 'align' ) ),
1443
1444                         # 11.2.3
1445                         'thead'      => array_merge( $common, $tablealign ),
1446                         'tfoot'      => array_merge( $common, $tablealign ),
1447                         'tbody'      => array_merge( $common, $tablealign ),
1448
1449                         # 11.2.4
1450                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1451                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1452
1453                         # 11.2.5
1454                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1455
1456                         # 11.2.6
1457                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1458                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1459
1460                         # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1461                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1462
1463                         # 13.2
1464                         # Not usually allowed, but may be used for extension-style hooks
1465                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1466                         # true
1467                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1468
1469                         # 15.2.1
1470                         'tt'         => $common,
1471                         'b'          => $common,
1472                         'i'          => $common,
1473                         'big'        => $common,
1474                         'small'      => $common,
1475                         'strike'     => $common,
1476                         's'          => $common,
1477                         'u'          => $common,
1478
1479                         # 15.2.2
1480                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1481                         # basefont
1482
1483                         # 15.3
1484                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1485
1486                         # XHTML Ruby annotation text module, simple ruby only.
1487                         # http://www.w3c.org/TR/ruby/
1488                         'ruby'       => $common,
1489                         # rbc
1490                         # rtc
1491                         'rb'         => $common,
1492                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1493                         'rp'         => $common,
1494
1495                         # MathML root element, where used for extensions
1496                         # 'title' may not be 100% valid here; it's XHTML
1497                         # http://www.w3.org/TR/REC-MathML/
1498                         'math'       => array( 'class', 'style', 'id', 'title' ),
1499                         );
1500                 return $whitelist;
1501         }
1502
1503         /**
1504          * Take a fragment of (potentially invalid) HTML and return
1505          * a version with any tags removed, encoded as plain text.
1506          *
1507          * Warning: this return value must be further escaped for literal
1508          * inclusion in HTML output as of 1.10!
1509          *
1510          * @param $text String: HTML fragment
1511          * @return String
1512          */
1513         static function stripAllTags( $text ) {
1514                 # Actual <tags>
1515                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1516
1517                 # Normalize &entities and whitespace
1518                 $text = self::decodeCharReferences( $text );
1519                 $text = self::normalizeWhitespace( $text );
1520
1521                 return $text;
1522         }
1523
1524         /**
1525          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1526          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1527          * PHP 5.1 doesn't.
1528          *
1529          * Use for passing XHTML fragments to PHP's XML parsing functions
1530          *
1531          * @return String
1532          */
1533         static function hackDocType() {
1534                 $out = "<!DOCTYPE html [\n";
1535                 foreach( self::$htmlEntities as $entity => $codepoint ) {
1536                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1537                 }
1538                 $out .= "]>\n";
1539                 return $out;
1540         }
1541
1542         /**
1543          * @param $url string
1544          * @return mixed|string
1545          */
1546         static function cleanUrl( $url ) {
1547                 # Normalize any HTML entities in input. They will be
1548                 # re-escaped by makeExternalLink().
1549                 $url = Sanitizer::decodeCharReferences( $url );
1550
1551                 # Escape any control characters introduced by the above step
1552                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1553                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1554
1555                 # Validate hostname portion
1556                 $matches = array();
1557                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1558                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1559
1560                         // Characters that will be ignored in IDNs.
1561                         // http://tools.ietf.org/html/3454#section-3.1
1562                         // Strip them before further processing so blacklists and such work.
1563                         $strip = "/
1564                                 \\s|          # general whitespace
1565                                 \xc2\xad|     # 00ad SOFT HYPHEN
1566                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1567                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1568                                 \xe2\x81\xa0| # 2060 WORD JOINER
1569                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1570                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1571                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1572                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1573                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1574                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1575                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1576                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe00f VARIATION SELECTOR-1-16
1577                                 /xuD";
1578
1579                         $host = preg_replace( $strip, '', $host );
1580
1581                         // @todo FIXME: Validate hostnames here
1582
1583                         return $protocol . $host . $rest;
1584                 } else {
1585                         return $url;
1586                 }
1587         }
1588
1589         /**
1590          * @param $matches array
1591          * @return string
1592          */
1593         static function cleanUrlCallback( $matches ) {
1594                 return urlencode( $matches[0] );
1595         }
1596
1597         /**
1598          * Does a string look like an e-mail address?
1599          *
1600          * This validates an email address using an HTML5 specification found at:
1601          * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1602          * Which as of 2011-01-24 says:
1603          *
1604          *   A valid e-mail address is a string that matches the ABNF production
1605          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1606          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1607          *   3.5.
1608          *
1609          * This function is an implementation of the specification as requested in
1610          * bug 22449.
1611          *
1612          * Client-side forms will use the same standard validation rules via JS or
1613          * HTML 5 validation; additional restrictions can be enforced server-side
1614          * by extensions via the 'isValidEmailAddr' hook.
1615          *
1616          * Note that this validation doesn't 100% match RFC 2822, but is believed
1617          * to be liberal enough for wide use. Some invalid addresses will still
1618          * pass validation here.
1619          *
1620          * @param $addr String E-mail address
1621          * @return Bool
1622          */
1623         public static function validateEmail( $addr ) {
1624                 $result = null;
1625                 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1626                         return $result;
1627                 }
1628
1629                 // Please note strings below are enclosed in brackets [], this make the
1630                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1631                 // See bug 26948
1632                 $rfc5322_atext   = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1633                 $rfc1034_ldh_str = "a-z0-9\\-" ;
1634
1635                 $HTML5_email_regexp = "/
1636                 ^                      # start of string
1637                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1638                 @                      # 'apostrophe'
1639                 [$rfc1034_ldh_str]+       # First domain part
1640                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1641                 $                      # End of string
1642                 /ix" ; // case Insensitive, eXtended
1643
1644                 return (bool) preg_match( $HTML5_email_regexp, $addr );
1645         }
1646 }