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', '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--??
 391                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 392                                 'dl', '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                                                                 $ot = @array_pop( $tagstack );
 461                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 462                                                                         array_push( $optstack, $ot );
 463                                                                         $ot = @array_pop( $tagstack );
 464                                                                 }
 465                                                                 if ( $t != $ot ) {
 466                                                                         # No match. Push the optional elements back again
 467                                                                         $badtag = true;
 468                                                                         while ( $ot = @array_pop( $optstack ) ) {
 469                                                                                 array_push( $tagstack, $ot );
 470                                                                         }
 471                                                                 }
 472                                                         } else {
 473                                                                 @array_push( $tagstack, $ot );
 474                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 475                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 476                                                                         $badtag = true;
 477                                                                 }
 478                                                         }
 479                                                 } else {
 480                                                         if ( $t == 'table' ) {
 481                                                                 $tagstack = array_pop( $tablestack );
 482                                                         }
 483                                                 }
 484                                                 $newparams = '';
 485                                         } else {
 486                                                 # Keep track for later
 487                                                 if ( isset( $tabletags[$t] ) &&
 488                                                 !in_array( 'table', $tagstack ) ) {
 489                                                         $badtag = true;
 490                                                 } elseif ( in_array( $t, $tagstack ) &&
 491                                                 !isset( $htmlnest [$t ] ) ) {
 492                                                         $badtag = true;
 493                                                 # Is it a self closed htmlpair ? (bug 5487)
 494                                                 } elseif ( $brace == '/>' &&
 495                                                 isset( $htmlpairs[$t] ) ) {
 496                                                         $badtag = true;
 497                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 498                                                         # Hack to force empty tag for uncloseable elements
 499                                                         $brace = '/>';
 500                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 501                                                         # Hack to not close $htmlsingle tags
 502                                                         $brace = null;
 503                                                 } elseif ( isset( $tabletags[$t] )
 504                                                 && in_array( $t, $tagstack ) ) {
 505                                                         // New table tag but forgot to close the previous one
 506                                                         $text .= "</$t>";
 507                                                 } else {
 508                                                         if ( $t == 'table' ) {
 509                                                                 array_push( $tablestack, $tagstack );
 510                                                                 $tagstack = array();
 511                                                         }
 512                                                         array_push( $tagstack, $t );
 513                                                 }
 514
 515                                                 # Replace any variables or template parameters with
 516                                                 # plaintext results.
 517                                                 if( is_callable( $processCallback ) ) {
 518                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 519                                                 }
 520
 521                                                 # Strip non-approved attributes from the tag
 522                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 523                                         }
 524                                         if ( !$badtag ) {
 525                                                 $rest = str_replace( '>', '&gt;', $rest );
 526                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 527                                                 $text .= "<$slash$t$newparams$close>$rest";
 528                                                 continue;
 529                                         }
 530                                 }
 531                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 532                         }
 533                         # Close off any remaining tags
 534                         while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
 535                                 $text .= "</$t>\n";
 536                                 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
 537                         }
 538                 } else {
 539                         # this might be possible using tidy itself
 540                         foreach ( $bits as $x ) {
 541                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 542                                 $x, $regs );
 543                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 544                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 545                                         if( is_callable( $processCallback ) ) {
 546                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 547                                         }
 548                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 549                                         $rest = str_replace( '>', '&gt;', $rest );
 550                                         $text .= "<$slash$t$newparams$brace$rest";
 551                                 } else {
 552                                         $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 553                                 }
 554                         }
 555                 }
 556                 wfProfileOut( __METHOD__ );
 557                 return $text;
 558         }
 559
 560         /**
 561          * Remove '<!--', '-->', and everything between.
 562          * To avoid leaving blank lines, when a comment is both preceded
 563          * and followed by a newline (ignoring spaces), trim leading and
 564          * trailing spaces and one of the newlines.
 565          *
 566          * @private
 567          * @param $text String
 568          * @return string
 569          */
 570         static function removeHTMLcomments( $text ) {
 571                 wfProfileIn( __METHOD__ );
 572                 while (($start = strpos($text, '<!--')) !== false) {
 573                         $end = strpos($text, '-->', $start + 4);
 574                         if ($end === false) {
 575                                 # Unterminated comment; bail out
 576                                 break;
 577                         }
 578
 579                         $end += 3;
 580
 581                         # Trim space and newline if the comment is both
 582                         # preceded and followed by a newline
 583                         $spaceStart = max($start - 1, 0);
 584                         $spaceLen = $end - $spaceStart;
 585                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 586                                 $spaceStart--;
 587                                 $spaceLen++;
 588                         }
 589                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 590                                 $spaceLen++;
 591                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 592                                 # Remove the comment, leading and trailing
 593                                 # spaces, and leave only one newline.
 594                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 595                         }
 596                         else {
 597                                 # Remove just the comment.
 598                                 $text = substr_replace($text, '', $start, $end - $start);
 599                         }
 600                 }
 601                 wfProfileOut( __METHOD__ );
 602                 return $text;
 603         }
 604
 605         /**
 606          * Take an array of attribute names and values and fix some deprecated values
 607          * for the given element type.
 608          * This does not validate properties, so you should ensure that you call
 609          * validateTagAttributes AFTER this to ensure that the resulting style rule
 610          * this may add is safe.
 611          *
 612          * - Converts most presentational attributes like align into inline css
 613          *
 614          * @param $attribs Array
 615          * @param $element String
 616          * @return Array
 617          */
 618         static function fixDeprecatedAttributes( $attribs, $element ) {
 619                 global $wgHtml5, $wgCleanupPresentationalAttributes;
 620
 621                 // presentational attributes were removed from html5, we can leave them
 622                 // in when html5 is turned off
 623                 if ( !$wgHtml5 || !$wgCleanupPresentationalAttributes ) {
 624                         return $attribs;
 625                 }
 626
 627                 $table = array( 'table' );
 628                 $cells = array( 'td', 'th' );
 629                 $colls = array( 'col', 'colgroup' );
 630                 $tblocks = array( 'tbody', 'tfoot', 'thead' );
 631                 $h = array( 'h1', 'h2', 'h3', 'h4', 'h5', 'h6' );
 632
 633                 $presentationalAttribs = array(
 634                         'align' => array( 'text-align', array_merge( array( 'caption', 'hr', 'div', 'p', 'tr' ), $table, $cells, $colls, $tblocks, $h ) ),
 635                         'clear' => array( 'clear', array( 'br' ) ),
 636                         'height' => array( 'height', $cells ),
 637                         'nowrap' => array( 'white-space', $cells ),
 638                         'size' => array( 'height', array( 'hr' ) ),
 639                         'type' => array( 'list-style-type', array( 'li', 'ol', 'ul' ) ),
 640                         'valign' => array( 'vertical-align', array_merge( $cells, $colls, $tblocks ) ),
 641                         'width' => array( 'width', array_merge( array( 'hr', 'pre' ), $table, $cells, $colls ) ),
 642                 );
 643
 644                 // Ensure that any upper case or mixed case attributes are converted to lowercase
 645                 foreach ( $attribs as $attribute => $value ) {
 646                         if ( $attribute !== strtolower( $attribute ) && array_key_exists( strtolower( $attribute ), $presentationalAttribs ) ) {
 647                                 $attribs[strtolower( $attribute )] = $value;
 648                                 unset( $attribs[$attribute] );
 649                         }
 650                 }
 651
 652                 $style = "";
 653                 foreach ( $presentationalAttribs as $attribute => $info ) {
 654                         list( $property, $elements ) = $info;
 655
 656                         // Skip if this attribute is not relevant to this element
 657                         if ( !in_array( $element, $elements ) ) {
 658                                 continue;
 659                         }
 660
 661                         // Skip if the attribute is not used
 662                         if ( !array_key_exists( $attribute, $attribs ) ) {
 663                                 continue;
 664                         }
 665
 666                         $value = $attribs[$attribute];
 667
 668                         // For nowrap the value should be nowrap instead of whatever text is in the value
 669                         if ( $attribute === 'nowrap' ) {
 670                                 $value = 'nowrap';
 671                         }
 672
 673                         // clear="all" is clear: both; in css
 674                         if ( $attribute === 'clear' && strtolower( $value ) === 'all' ) {
 675                                 $value = 'both';
 676                         }
 677
 678                         // Size based properties should have px applied to them if they have no unit
 679                         if ( in_array( $attribute, array( 'height', 'width', 'size' ) ) ) {
 680                                 if ( preg_match( '/^[\d.]+$/', $value ) ) {
 681                                         $value = "{$value}px";
 682                                 }
 683                         }
 684
 685                         $style .= " $property: $value;";
 686
 687                         unset( $attribs[$attribute] );
 688                 }
 689
 690                 if ( $style ) {
 691                         // Prepend our style rules so that they can be overridden by user css
 692                         if ( isset($attribs['style']) ) {
 693                                 $style .= " " . $attribs['style'];
 694                         }
 695                         $attribs['style'] = trim($style);
 696                 }
 697
 698                 return $attribs;
 699         }
 700
 701         /**
 702          * Take an array of attribute names and values and normalize or discard
 703          * illegal values for the given element type.
 704          *
 705          * - Discards attributes not on a whitelist for the given element
 706          * - Unsafe style attributes are discarded
 707          * - Invalid id attributes are reencoded
 708          *
 709          * @param $attribs Array
 710          * @param $element String
 711          * @return Array
 712          *
 713          * @todo Check for legal values where the DTD limits things.
 714          * @todo Check for unique id attribute :P
 715          */
 716         static function validateTagAttributes( $attribs, $element ) {
 717                 return Sanitizer::validateAttributes( $attribs,
 718                         Sanitizer::attributeWhitelist( $element ) );
 719         }
 720
 721         /**
 722          * Take an array of attribute names and values and normalize or discard
 723          * illegal values for the given whitelist.
 724          *
 725          * - Discards attributes not the given whitelist
 726          * - Unsafe style attributes are discarded
 727          * - Invalid id attributes are reencoded
 728          *
 729          * @param $attribs Array
 730          * @param $whitelist Array: list of allowed attribute names
 731          * @return Array
 732          *
 733          * @todo Check for legal values where the DTD limits things.
 734          * @todo Check for unique id attribute :P
 735          */
 736         static function validateAttributes( $attribs, $whitelist ) {
 737                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
 738
 739                 $whitelist = array_flip( $whitelist );
 740                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 741
 742                 $out = array();
 743                 foreach( $attribs as $attribute => $value ) {
 744                         #allow XML namespace declaration if RDFa is enabled
 745                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 746                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 747                                         $out[$attribute] = $value;
 748                                 }
 749
 750                                 continue;
 751                         }
 752
 753                         # Allow any attribute beginning with "data-", if in HTML5 mode
 754                         if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
 755                                 continue;
 756                         }
 757
 758                         # Strip javascript "expression" from stylesheets.
 759                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 760                         if( $attribute == 'style' ) {
 761                                 $value = Sanitizer::checkCss( $value );
 762                         }
 763
 764                         if ( $attribute === 'id' ) {
 765                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 766                         }
 767
 768                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 769                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 770                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 771                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 772                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 773                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 774
 775                                 //Paranoia. Allow "simple" values but suppress javascript
 776                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 777                                         continue;
 778                                 }
 779                         }
 780
 781                         # NOTE: even though elements using href/src are not allowed directly, supply
 782                         #       validation code that can be used by tag hook handlers, etc
 783                         if ( $attribute === 'href' || $attribute === 'src' ) {
 784                                 if ( !preg_match( $hrefExp, $value ) ) {
 785                                         continue; //drop any href or src attributes not using an allowed protocol.
 786                                                   //NOTE: this also drops all relative URLs
 787                                 }
 788                         }
 789
 790                         // If this attribute was previously set, override it.
 791                         // Output should only have one attribute of each name.
 792                         $out[$attribute] = $value;
 793                 }
 794
 795                 if ( $wgAllowMicrodataAttributes ) {
 796                         # itemtype, itemid, itemref don't make sense without itemscope
 797                         if ( !array_key_exists( 'itemscope', $out ) ) {
 798                                 unset( $out['itemtype'] );
 799                                 unset( $out['itemid'] );
 800                                 unset( $out['itemref'] );
 801                         }
 802                         # TODO: Strip itemprop if we aren't descendants of an itemscope.
 803                 }
 804                 return $out;
 805         }
 806
 807         /**
 808          * Merge two sets of HTML attributes.  Conflicting items in the second set
 809          * will override those in the first, except for 'class' attributes which
 810          * will be combined (if they're both strings).
 811          *
 812          * @todo implement merging for other attributes such as style
 813          * @param $a Array
 814          * @param $b Array
 815          * @return array
 816          */
 817         static function mergeAttributes( $a, $b ) {
 818                 $out = array_merge( $a, $b );
 819                 if( isset( $a['class'] ) && isset( $b['class'] )
 820                 && is_string( $a['class'] ) && is_string( $b['class'] )
 821                 && $a['class'] !== $b['class'] ) {
 822                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 823                                 -1, PREG_SPLIT_NO_EMPTY );
 824                         $out['class'] = implode( ' ', array_unique( $classes ) );
 825                 }
 826                 return $out;
 827         }
 828
 829         /**
 830          * Pick apart some CSS and check it for forbidden or unsafe structures.
 831          * Returns a sanitized string. This sanitized string will have
 832          * character references and escape sequences decoded, and comments
 833          * stripped. If the input is just too evil, only a comment complaining
 834          * about evilness will be returned.
 835          *
 836          * Currently URL references, 'expression', 'tps' are forbidden.
 837          *
 838          * NOTE: Despite the fact that character references are decoded, the
 839          * returned string may contain character references given certain
 840          * clever input strings. These character references must
 841          * be escaped before the return value is embedded in HTML.
 842          *
 843          * @param $value String
 844          * @return String
 845          */
 846         static function checkCss( $value ) {
 847                 // Decode character references like &#123;
 848                 $value = Sanitizer::decodeCharReferences( $value );
 849
 850                 // Decode escape sequences and line continuation
 851                 // See the grammar in the CSS 2 spec, appendix D.
 852                 // This has to be done AFTER decoding character references.
 853                 // This means it isn't possible for this function to return
 854                 // unsanitized escape sequences. It is possible to manufacture
 855                 // input that contains character references that decode to
 856                 // escape sequences that decode to character references, but
 857                 // it's OK for the return value to contain character references
 858                 // because the caller is supposed to escape those anyway.
 859                 static $decodeRegex;
 860                 if ( !$decodeRegex ) {
 861                         $space = '[\\x20\\t\\r\\n\\f]';
 862                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 863                         $backslash = '\\\\';
 864                         $decodeRegex = "/ $backslash
 865                                 (?:
 866                                         ($nl) |  # 1. Line continuation
 867                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 868                                         (.) | # 3. backslash cancelling special meaning
 869                                         () | # 4. backslash at end of string
 870                                 )/xu";
 871                 }
 872                 $value = preg_replace_callback( $decodeRegex,
 873                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 874
 875                 // Remove any comments; IE gets token splitting wrong
 876                 // This must be done AFTER decoding character references and
 877                 // escape sequences, because those steps can introduce comments
 878                 // This step cannot introduce character references or escape
 879                 // sequences, because it replaces comments with spaces rather
 880                 // than removing them completely.
 881                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 882
 883                 // Remove anything after a comment-start token, to guard against
 884                 // incorrect client implementations.
 885                 $commentPos = strpos( $value, '/*' );
 886                 if ( $commentPos !== false ) {
 887                         $value = substr( $value, 0, $commentPos );
 888                 }
 889
 890                 // Reject problematic keywords and control characters
 891                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 892                         return '/* invalid control char */';
 893                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
 894                         return '/* insecure input */';
 895                 }
 896                 return $value;
 897         }
 898
 899         /**
 900          * @param $matches array
 901          * @return String
 902          */
 903         static function cssDecodeCallback( $matches ) {
 904                 if ( $matches[1] !== '' ) {
 905                         // Line continuation
 906                         return '';
 907                 } elseif ( $matches[2] !== '' ) {
 908                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 909                 } elseif ( $matches[3] !== '' ) {
 910                         $char = $matches[3];
 911                 } else {
 912                         $char = '\\';
 913                 }
 914                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 915                         // These characters need to be escaped in strings
 916                         // Clean up the escape sequence to avoid parsing errors by clients
 917                         return '\\' . dechex( ord( $char ) ) . ' ';
 918                 } else {
 919                         // Decode unnecessary escape
 920                         return $char;
 921                 }
 922         }
 923
 924         /**
 925          * Take a tag soup fragment listing an HTML element's attributes
 926          * and normalize it to well-formed XML, discarding unwanted attributes.
 927          * Output is safe for further wikitext processing, with escaping of
 928          * values that could trigger problems.
 929          *
 930          * - Normalizes attribute names to lowercase
 931          * - Discards attributes not on a whitelist for the given element
 932          * - Turns broken or invalid entities into plaintext
 933          * - Double-quotes all attribute values
 934          * - Attributes without values are given the name as attribute
 935          * - Double attributes are discarded
 936          * - Unsafe style attributes are discarded
 937          * - Prepends space if there are attributes.
 938          *
 939          * @param $text String
 940          * @param $element String
 941          * @return String
 942          */
 943         static function fixTagAttributes( $text, $element ) {
 944                 if( trim( $text ) == '' ) {
 945                         return '';
 946                 }
 947
 948                 $decoded = Sanitizer::decodeTagAttributes( $text );
 949                 $decoded = Sanitizer::fixDeprecatedAttributes( $decoded, $element );
 950                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
 951
 952                 $attribs = array();
 953                 foreach( $stripped as $attribute => $value ) {
 954                         $encAttribute = htmlspecialchars( $attribute );
 955                         $encValue = Sanitizer::safeEncodeAttribute( $value );
 956
 957                         $attribs[] = "$encAttribute=\"$encValue\"";
 958                 }
 959                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
 960         }
 961
 962         /**
 963          * Encode an attribute value for HTML output.
 964          * @param $text String
 965          * @return HTML-encoded text fragment
 966          */
 967         static function encodeAttribute( $text ) {
 968                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
 969
 970                 // Whitespace is normalized during attribute decoding,
 971                 // so if we've been passed non-spaces we must encode them
 972                 // ahead of time or they won't be preserved.
 973                 $encValue = strtr( $encValue, array(
 974                         "\n" => '&#10;',
 975                         "\r" => '&#13;',
 976                         "\t" => '&#9;',
 977                 ) );
 978
 979                 return $encValue;
 980         }
 981
 982         /**
 983          * Encode an attribute value for HTML tags, with extra armoring
 984          * against further wiki processing.
 985          * @param $text String
 986          * @return HTML-encoded text fragment
 987          */
 988         static function safeEncodeAttribute( $text ) {
 989                 $encValue = Sanitizer::encodeAttribute( $text );
 990
 991                 # Templates and links may be expanded in later parsing,
 992                 # creating invalid or dangerous output. Suppress this.
 993                 $encValue = strtr( $encValue, array(
 994                         '<'    => '&lt;',   // This should never happen,
 995                         '>'    => '&gt;',   // we've received invalid input
 996                         '"'    => '&quot;', // which should have been escaped.
 997                         '{'    => '&#123;',
 998                         '['    => '&#91;',
 999                         "''"   => '&#39;&#39;',
1000                         'ISBN' => '&#73;SBN',
1001                         'RFC'  => '&#82;FC',
1002                         'PMID' => '&#80;MID',
1003                         '|'    => '&#124;',
1004                         '__'   => '&#95;_',
1005                 ) );
1006
1007                 # Stupid hack
1008                 $encValue = preg_replace_callback(
1009                         '/(' . wfUrlProtocols() . ')/',
1010                         array( 'Sanitizer', 'armorLinksCallback' ),
1011                         $encValue );
1012                 return $encValue;
1013         }
1014
1015         /**
1016          * Given a value, escape it so that it can be used in an id attribute and
1017          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1018          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1019          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1020          * escaped with lots of dots.
1021          *
1022          * To ensure we don't have to bother escaping anything, we also strip ', ",
1023          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1024          * We also strip # because it upsets IE, and % because it could be
1025          * ambiguous if it's part of something that looks like a percent escape
1026          * (which don't work reliably in fragments cross-browser).
1027          *
1028          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1029          *                                                          in the id and
1030          *                                                          name attributes
1031          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1032          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1033          *   HTML5 definition of id attribute
1034          *
1035          * @param $id String: id to escape
1036          * @param $options Mixed: string or array of strings (default is array()):
1037          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1038          *       so don't pay attention if the first character isn't valid at the
1039          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1040          *       false.
1041          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1042          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1043          *       anchors and links won't break.
1044          * @return String
1045          */
1046         static function escapeId( $id, $options = array() ) {
1047                 global $wgHtml5, $wgExperimentalHtmlIds;
1048                 $options = (array)$options;
1049
1050                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1051                         $id = Sanitizer::decodeCharReferences( $id );
1052                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1053                         $id = trim( $id, '_' );
1054                         if ( $id === '' ) {
1055                                 # Must have been all whitespace to start with.
1056                                 return '_';
1057                         } else {
1058                                 return $id;
1059                         }
1060                 }
1061
1062                 # HTML4-style escaping
1063                 static $replace = array(
1064                         '%3A' => ':',
1065                         '%' => '.'
1066                 );
1067
1068                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1069                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1070
1071                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1072                 && !in_array( 'noninitial', $options ) )  {
1073                         // Initial character must be a letter!
1074                         $id = "x$id";
1075                 }
1076                 return $id;
1077         }
1078
1079         /**
1080          * Given a value, escape it so that it can be used as a CSS class and
1081          * return it.
1082          *
1083          * @todo For extra validity, input should be validated UTF-8.
1084          *
1085          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1086          *
1087          * @param $class String
1088          * @return String
1089          */
1090         static function escapeClass( $class ) {
1091                 // Convert ugly stuff to underscores and kill underscores in ugly places
1092                 return rtrim(preg_replace(
1093                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
1094                         '_',
1095                         $class ), '_');
1096         }
1097
1098         /**
1099          * Given HTML input, escape with htmlspecialchars but un-escape entites.
1100          * This allows (generally harmless) entities like &#160; to survive.
1101          *
1102          * @param $html String to escape
1103          * @return String: escaped input
1104          */
1105         static function escapeHtmlAllowEntities( $html ) {
1106                 $html = Sanitizer::decodeCharReferences( $html );
1107                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1108                 # hurt.
1109                 $html = htmlspecialchars( $html, ENT_QUOTES );
1110                 return $html;
1111         }
1112
1113         /**
1114          * Regex replace callback for armoring links against further processing.
1115          * @param $matches Array
1116          * @return string
1117          */
1118         private static function armorLinksCallback( $matches ) {
1119                 return str_replace( ':', '&#58;', $matches[1] );
1120         }
1121
1122         /**
1123          * Return an associative array of attribute names and values from
1124          * a partial tag string. Attribute names are forces to lowercase,
1125          * character references are decoded to UTF-8 text.
1126          *
1127          * @param $text String
1128          * @return Array
1129          */
1130         public static function decodeTagAttributes( $text ) {
1131                 if( trim( $text ) == '' ) {
1132                         return array();
1133                 }
1134
1135                 $attribs = array();
1136                 $pairs = array();
1137                 if( !preg_match_all(
1138                         self::getAttribsRegex(),
1139                         $text,
1140                         $pairs,
1141                         PREG_SET_ORDER ) ) {
1142                         return $attribs;
1143                 }
1144
1145                 foreach( $pairs as $set ) {
1146                         $attribute = strtolower( $set[1] );
1147                         $value = Sanitizer::getTagAttributeCallback( $set );
1148
1149                         // Normalize whitespace
1150                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1151                         $value = trim( $value );
1152
1153                         // Decode character references
1154                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1155                 }
1156                 return $attribs;
1157         }
1158
1159         /**
1160          * Pick the appropriate attribute value from a match set from the
1161          * attribs regex matches.
1162          *
1163          * @param $set Array
1164          * @return String
1165          */
1166         private static function getTagAttributeCallback( $set ) {
1167                 if( isset( $set[6] ) ) {
1168                         # Illegal #XXXXXX color with no quotes.
1169                         return $set[6];
1170                 } elseif( isset( $set[5] ) ) {
1171                         # No quotes.
1172                         return $set[5];
1173                 } elseif( isset( $set[4] ) ) {
1174                         # Single-quoted
1175                         return $set[4];
1176                 } elseif( isset( $set[3] ) ) {
1177                         # Double-quoted
1178                         return $set[3];
1179                 } elseif( !isset( $set[2] ) ) {
1180                         # In XHTML, attributes must have a value.
1181                         # For 'reduced' form, return explicitly the attribute name here.
1182                         return $set[1];
1183                 } else {
1184                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1185                 }
1186         }
1187
1188         /**
1189          * Normalize whitespace and character references in an XML source-
1190          * encoded text for an attribute value.
1191          *
1192          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1193          * but note that we're not returning the value, but are returning
1194          * XML source fragments that will be slapped into output.
1195          *
1196          * @param $text String
1197          * @return String
1198          */
1199         private static function normalizeAttributeValue( $text ) {
1200                 return str_replace( '"', '&quot;',
1201                         self::normalizeWhitespace(
1202                                 Sanitizer::normalizeCharReferences( $text ) ) );
1203         }
1204
1205         /**
1206          * @param $text string
1207          * @return mixed
1208          */
1209         private static function normalizeWhitespace( $text ) {
1210                 return preg_replace(
1211                         '/\r\n|[\x20\x0d\x0a\x09]/',
1212                         ' ',
1213                         $text );
1214         }
1215
1216         /**
1217          * Normalizes whitespace in a section name, such as might be returned
1218          * by Parser::stripSectionName(), for use in the id's that are used for
1219          * section links.
1220          *
1221          * @param $section String
1222          * @return String
1223          */
1224         static function normalizeSectionNameWhitespace( $section ) {
1225                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1226         }
1227
1228         /**
1229          * Ensure that any entities and character references are legal
1230          * for XML and XHTML specifically. Any stray bits will be
1231          * &amp;-escaped to result in a valid text fragment.
1232          *
1233          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1234          *   numericized (this way we're well-formed even without a DTD)
1235          * b. any numeric char refs must be legal chars, not invalid or forbidden
1236          * c. use &#x, not &#X
1237          * d. fix or reject non-valid attributes
1238          *
1239          * @param $text String
1240          * @return String
1241          * @private
1242          */
1243         static function normalizeCharReferences( $text ) {
1244                 return preg_replace_callback(
1245                         self::CHAR_REFS_REGEX,
1246                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1247                         $text );
1248         }
1249         /**
1250          * @param $matches String
1251          * @return String
1252          */
1253         static function normalizeCharReferencesCallback( $matches ) {
1254                 $ret = null;
1255                 if( $matches[1] != '' ) {
1256                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1257                 } elseif( $matches[2] != '' ) {
1258                         $ret = Sanitizer::decCharReference( $matches[2] );
1259                 } elseif( $matches[3] != ''  ) {
1260                         $ret = Sanitizer::hexCharReference( $matches[3] );
1261                 }
1262                 if( is_null( $ret ) ) {
1263                         return htmlspecialchars( $matches[0] );
1264                 } else {
1265                         return $ret;
1266                 }
1267         }
1268
1269         /**
1270          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1271          * return the equivalent numeric entity reference (except for the core &lt;
1272          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1273          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1274          * pseudo-entity source (eg &amp;foo;)
1275          *
1276          * @param $name String
1277          * @return String
1278          */
1279         static function normalizeEntity( $name ) {
1280                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1281                         return '&' . self::$htmlEntityAliases[$name] . ';';
1282                 } elseif ( in_array( $name,
1283                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1284                         return "&$name;";
1285                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1286                         return '&#' . self::$htmlEntities[$name] . ';';
1287                 } else {
1288                         return "&amp;$name;";
1289                 }
1290         }
1291
1292         /**
1293          * @param $codepoint
1294          * @return null|string
1295          */
1296         static function decCharReference( $codepoint ) {
1297                 $point = intval( $codepoint );
1298                 if( Sanitizer::validateCodepoint( $point ) ) {
1299                         return sprintf( '&#%d;', $point );
1300                 } else {
1301                         return null;
1302                 }
1303         }
1304
1305         /**
1306          * @param $codepoint
1307          * @return null|string
1308          */
1309         static function hexCharReference( $codepoint ) {
1310                 $point = hexdec( $codepoint );
1311                 if( Sanitizer::validateCodepoint( $point ) ) {
1312                         return sprintf( '&#x%x;', $point );
1313                 } else {
1314                         return null;
1315                 }
1316         }
1317
1318         /**
1319          * Returns true if a given Unicode codepoint is a valid character in XML.
1320          * @param $codepoint Integer
1321          * @return Boolean
1322          */
1323         private static function validateCodepoint( $codepoint ) {
1324                 return ($codepoint ==    0x09)
1325                         || ($codepoint ==    0x0a)
1326                         || ($codepoint ==    0x0d)
1327                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1328                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1329                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1330         }
1331
1332         /**
1333          * Decode any character references, numeric or named entities,
1334          * in the text and return a UTF-8 string.
1335          *
1336          * @param $text String
1337          * @return String
1338          */
1339         public static function decodeCharReferences( $text ) {
1340                 return preg_replace_callback(
1341                         self::CHAR_REFS_REGEX,
1342                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1343                         $text );
1344         }
1345
1346         /**
1347          * Decode any character references, numeric or named entities,
1348          * in the next and normalize the resulting string. (bug 14952)
1349          *
1350          * This is useful for page titles, not for text to be displayed,
1351          * MediaWiki allows HTML entities to escape normalization as a feature.
1352          *
1353          * @param $text String (already normalized, containing entities)
1354          * @return String (still normalized, without entities)
1355          */
1356         public static function decodeCharReferencesAndNormalize( $text ) {
1357                 global $wgContLang;
1358                 $text = preg_replace_callback(
1359                         self::CHAR_REFS_REGEX,
1360                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1361                         $text, /* limit */ -1, $count );
1362
1363                 if ( $count ) {
1364                         return $wgContLang->normalize( $text );
1365                 } else {
1366                         return $text;
1367                 }
1368         }
1369
1370         /**
1371          * @param $matches String
1372          * @return String
1373          */
1374         static function decodeCharReferencesCallback( $matches ) {
1375                 if( $matches[1] != '' ) {
1376                         return Sanitizer::decodeEntity( $matches[1] );
1377                 } elseif( $matches[2] != '' ) {
1378                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1379                 } elseif( $matches[3] != ''  ) {
1380                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1381                 }
1382                 # Last case should be an ampersand by itself
1383                 return $matches[0];
1384         }
1385
1386         /**
1387          * Return UTF-8 string for a codepoint if that is a valid
1388          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1389          * @param $codepoint Integer
1390          * @return String
1391          * @private
1392          */
1393         static function decodeChar( $codepoint ) {
1394                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1395                         return codepointToUtf8( $codepoint );
1396                 } else {
1397                         return UTF8_REPLACEMENT;
1398                 }
1399         }
1400
1401         /**
1402          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1403          * return the UTF-8 encoding of that character. Otherwise, returns
1404          * pseudo-entity source (eg &foo;)
1405          *
1406          * @param $name String
1407          * @return String
1408          */
1409         static function decodeEntity( $name ) {
1410                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1411                         $name = self::$htmlEntityAliases[$name];
1412                 }
1413                 if( isset( self::$htmlEntities[$name] ) ) {
1414                         return codepointToUtf8( self::$htmlEntities[$name] );
1415                 } else {
1416                         return "&$name;";
1417                 }
1418         }
1419
1420         /**
1421          * Fetch the whitelist of acceptable attributes for a given element name.
1422          *
1423          * @param $element String
1424          * @return Array
1425          */
1426         static function attributeWhitelist( $element ) {
1427                 static $list;
1428                 if( !isset( $list ) ) {
1429                         $list = Sanitizer::setupAttributeWhitelist();
1430                 }
1431                 return isset( $list[$element] )
1432                         ? $list[$element]
1433                         : array();
1434         }
1435
1436         /**
1437          * Foreach array key (an allowed HTML element), return an array
1438          * of allowed attributes
1439          * @return Array
1440          */
1441         static function setupAttributeWhitelist() {
1442                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1443
1444                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1445
1446                 if ( $wgAllowRdfaAttributes ) {
1447                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1448                         $common = array_merge( $common, array(
1449                             'about', 'property', 'resource', 'datatype', 'typeof',
1450                         ) );
1451                 }
1452
1453                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1454                         # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1455                         $common = array_merge( $common, array(
1456                             'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1457                         ) );
1458                 }
1459
1460                 $block = array_merge( $common, array( 'align' ) );
1461                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1462                 $tablecell = array( 'abbr',
1463                                     'axis',
1464                                     'headers',
1465                                     'scope',
1466                                     'rowspan',
1467                                     'colspan',
1468                                     'nowrap', # deprecated
1469                                     'width',  # deprecated
1470                                     'height', # deprecated
1471                                     'bgcolor' # deprecated
1472                                     );
1473
1474                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1475                 # See: http://www.w3.org/TR/html4/
1476                 $whitelist = array (
1477                         # 7.5.4
1478                         'div'        => $block,
1479                         'center'     => $common, # deprecated
1480                         'span'       => $block, # ??
1481
1482                         # 7.5.5
1483                         'h1'         => $block,
1484                         'h2'         => $block,
1485                         'h3'         => $block,
1486                         'h4'         => $block,
1487                         'h5'         => $block,
1488                         'h6'         => $block,
1489
1490                         # 7.5.6
1491                         # address
1492
1493                         # 8.2.4
1494                         # bdo
1495
1496                         # 9.2.1
1497                         'em'         => $common,
1498                         'strong'     => $common,
1499                         'cite'       => $common,
1500                         'dfn'        => $common,
1501                         'code'       => $common,
1502                         'samp'       => $common,
1503                         'kbd'        => $common,
1504                         'var'        => $common,
1505                         'abbr'       => $common,
1506                         # acronym
1507
1508                         # 9.2.2
1509                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1510                         # q
1511
1512                         # 9.2.3
1513                         'sub'        => $common,
1514                         'sup'        => $common,
1515
1516                         # 9.3.1
1517                         'p'          => $block,
1518
1519                         # 9.3.2
1520                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1521
1522                         # 9.3.4
1523                         'pre'        => array_merge( $common, array( 'width' ) ),
1524
1525                         # 9.4
1526                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1527                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1528
1529                         # 10.2
1530                         'ul'         => array_merge( $common, array( 'type' ) ),
1531                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1532                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1533
1534                         # 10.3
1535                         'dl'         => $common,
1536                         'dd'         => $common,
1537                         'dt'         => $common,
1538
1539                         # 11.2.1
1540                         'table'      => array_merge( $common,
1541                                                                 array( 'summary', 'width', 'border', 'frame',
1542                                                                                 'rules', 'cellspacing', 'cellpadding',
1543                                                                                 'align', 'bgcolor',
1544                                                                 ) ),
1545
1546                         # 11.2.2
1547                         'caption'    => array_merge( $common, array( 'align' ) ),
1548
1549                         # 11.2.3
1550                         'thead'      => array_merge( $common, $tablealign ),
1551                         'tfoot'      => array_merge( $common, $tablealign ),
1552                         'tbody'      => array_merge( $common, $tablealign ),
1553
1554                         # 11.2.4
1555                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1556                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1557
1558                         # 11.2.5
1559                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1560
1561                         # 11.2.6
1562                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1563                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1564
1565                         # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1566                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1567
1568                         # 13.2
1569                         # Not usually allowed, but may be used for extension-style hooks
1570                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1571                         # true
1572                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1573
1574                         # 15.2.1
1575                         'tt'         => $common,
1576                         'b'          => $common,
1577                         'i'          => $common,
1578                         'big'        => $common,
1579                         'small'      => $common,
1580                         'strike'     => $common,
1581                         's'          => $common,
1582                         'u'          => $common,
1583
1584                         # 15.2.2
1585                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1586                         # basefont
1587
1588                         # 15.3
1589                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1590
1591                         # XHTML Ruby annotation text module, simple ruby only.
1592                         # http://www.w3c.org/TR/ruby/
1593                         'ruby'       => $common,
1594                         # rbc
1595                         # rtc
1596                         'rb'         => $common,
1597                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1598                         'rp'         => $common,
1599
1600                         # MathML root element, where used for extensions
1601                         # 'title' may not be 100% valid here; it's XHTML
1602                         # http://www.w3.org/TR/REC-MathML/
1603                         'math'       => array( 'class', 'style', 'id', 'title' ),
1604                         );
1605                 return $whitelist;
1606         }
1607
1608         /**
1609          * Take a fragment of (potentially invalid) HTML and return
1610          * a version with any tags removed, encoded as plain text.
1611          *
1612          * Warning: this return value must be further escaped for literal
1613          * inclusion in HTML output as of 1.10!
1614          *
1615          * @param $text String: HTML fragment
1616          * @return String
1617          */
1618         static function stripAllTags( $text ) {
1619                 # Actual <tags>
1620                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1621
1622                 # Normalize &entities and whitespace
1623                 $text = self::decodeCharReferences( $text );
1624                 $text = self::normalizeWhitespace( $text );
1625
1626                 return $text;
1627         }
1628
1629         /**
1630          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1631          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1632          * PHP 5.1 doesn't.
1633          *
1634          * Use for passing XHTML fragments to PHP's XML parsing functions
1635          *
1636          * @return String
1637          */
1638         static function hackDocType() {
1639                 $out = "<!DOCTYPE html [\n";
1640                 foreach( self::$htmlEntities as $entity => $codepoint ) {
1641                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1642                 }
1643                 $out .= "]>\n";
1644                 return $out;
1645         }
1646
1647         /**
1648          * @param $url string
1649          * @return mixed|string
1650          */
1651         static function cleanUrl( $url ) {
1652                 # Normalize any HTML entities in input. They will be
1653                 # re-escaped by makeExternalLink().
1654                 $url = Sanitizer::decodeCharReferences( $url );
1655
1656                 # Escape any control characters introduced by the above step
1657                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1658                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1659
1660                 # Validate hostname portion
1661                 $matches = array();
1662                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1663                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1664
1665                         // Characters that will be ignored in IDNs.
1666                         // http://tools.ietf.org/html/3454#section-3.1
1667                         // Strip them before further processing so blacklists and such work.
1668                         $strip = "/
1669                                 \\s|          # general whitespace
1670                                 \xc2\xad|     # 00ad SOFT HYPHEN
1671                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1672                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1673                                 \xe2\x81\xa0| # 2060 WORD JOINER
1674                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1675                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1676                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1677                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1678                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1679                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1680                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1681                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe00f VARIATION SELECTOR-1-16
1682                                 /xuD";
1683
1684                         $host = preg_replace( $strip, '', $host );
1685
1686                         // @todo FIXME: Validate hostnames here
1687
1688                         return $protocol . $host . $rest;
1689                 } else {
1690                         return $url;
1691                 }
1692         }
1693
1694         /**
1695          * @param $matches array
1696          * @return string
1697          */
1698         static function cleanUrlCallback( $matches ) {
1699                 return urlencode( $matches[0] );
1700         }
1701
1702         /**
1703          * Does a string look like an e-mail address?
1704          *
1705          * This validates an email address using an HTML5 specification found at:
1706          * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1707          * Which as of 2011-01-24 says:
1708          *
1709          *   A valid e-mail address is a string that matches the ABNF production
1710          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1711          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1712          *   3.5.
1713          *
1714          * This function is an implementation of the specification as requested in
1715          * bug 22449.
1716          *
1717          * Client-side forms will use the same standard validation rules via JS or
1718          * HTML 5 validation; additional restrictions can be enforced server-side
1719          * by extensions via the 'isValidEmailAddr' hook.
1720          *
1721          * Note that this validation doesn't 100% match RFC 2822, but is believed
1722          * to be liberal enough for wide use. Some invalid addresses will still
1723          * pass validation here.
1724          *
1725          * @since 1.18
1726          *
1727          * @param $addr String E-mail address
1728          * @return Bool
1729          */
1730         public static function validateEmail( $addr ) {
1731                 $result = null;
1732                 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1733                         return $result;
1734                 }
1735
1736                 // Please note strings below are enclosed in brackets [], this make the
1737                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1738                 // See bug 26948
1739                 $rfc5322_atext   = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1740                 $rfc1034_ldh_str = "a-z0-9\\-" ;
1741
1742                 $HTML5_email_regexp = "/
1743                 ^                      # start of string
1744                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1745                 @                      # 'apostrophe'
1746                 [$rfc1034_ldh_str]+       # First domain part
1747                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1748                 $                      # End of string
1749                 /ix" ; // case Insensitive, eXtended
1750
1751                 return (bool) preg_match( $HTML5_email_regexp, $addr );
1752         }
1753 }