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, $wgHtml5, $wgAllowMicrodataAttributes;
 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                         if ( $wgHtml5 ) {
 385                                 $htmlpairsStatic = array_merge( $htmlpairsStatic, array( 'data', 'time' ) );
 386                         }
 387                         $htmlsingle = array(
 388                                 'br', 'hr', 'li', 'dt', 'dd'
 389                         );
 390                         $htmlsingleonly = array( # Elements that cannot have close tags
 391                                 'br', 'hr'
 392                         );
 393                         if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
 394                                 $htmlsingle[] = $htmlsingleonly[] = 'meta';
 395                                 $htmlsingle[] = $htmlsingleonly[] = 'link';
 396                         }
 397                         $htmlnest = array( # Tags that can be nested--??
 398                                 'table', 'tr', 'td', 'th', 'div', 'blockquote', 'ol', 'ul',
 399                                 'dl', 'font', 'big', 'small', 'sub', 'sup', 'span'
 400                         );
 401                         $tabletags = array( # Can only appear inside table, we will close them
 402                                 'td', 'th', 'tr',
 403                         );
 404                         $htmllist = array( # Tags used by list
 405                                 'ul','ol',
 406                         );
 407                         $listtags = array( # Tags that can appear in a list
 408                                 'li',
 409                         );
 410
 411                         global $wgAllowImageTag;
 412                         if ( $wgAllowImageTag ) {
 413                                 $htmlsingle[] = 'img';
 414                                 $htmlsingleonly[] = 'img';
 415                         }
 416
 417                         $htmlsingleallowed = array_unique( array_merge( $htmlsingle, $tabletags ) );
 418                         $htmlelementsStatic = array_unique( array_merge( $htmlsingle, $htmlpairsStatic, $htmlnest ) );
 419
 420                         # Convert them all to hashtables for faster lookup
 421                         $vars = array( 'htmlpairsStatic', 'htmlsingle', 'htmlsingleonly', 'htmlnest', 'tabletags',
 422                                 'htmllist', 'listtags', 'htmlsingleallowed', 'htmlelementsStatic' );
 423                         foreach ( $vars as $var ) {
 424                                 $$var = array_flip( $$var );
 425                         }
 426                         $staticInitialised = true;
 427                 }
 428                 # Populate $htmlpairs and $htmlelements with the $extratags and $removetags arrays
 429                 $extratags = array_flip( $extratags );
 430                 $removetags = array_flip( $removetags );
 431                 $htmlpairs = array_merge( $extratags, $htmlpairsStatic );
 432                 $htmlelements = array_diff_key( array_merge( $extratags, $htmlelementsStatic ) , $removetags );
 433
 434                 # Remove HTML comments
 435                 $text = Sanitizer::removeHTMLcomments( $text );
 436                 $bits = explode( '<', $text );
 437                 $text = str_replace( '>', '&gt;', array_shift( $bits ) );
 438                 if ( !$wgUseTidy ) {
 439                         $tagstack = $tablestack = array();
 440                         foreach ( $bits as $x ) {
 441                                 $regs = array();
 442                                 # $slash: Does the current element start with a '/'?
 443                                 # $t: Current element name
 444                                 # $params: String between element name and >
 445                                 # $brace: Ending '>' or '/>'
 446                                 # $rest: Everything until the next element of $bits
 447                                 if( preg_match( '!^(/?)(\\w+)([^>]*?)(/{0,1}>)([^<]*)$!', $x, $regs ) ) {
 448                                         list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 449                                 } else {
 450                                         $slash = $t = $params = $brace = $rest = null;
 451                                 }
 452
 453                                 $badtag = false;
 454                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 455                                         # Check our stack
 456                                         if ( $slash && isset( $htmlsingleonly[$t] ) ) {
 457                                                 $badtag = true;
 458                                         } elseif ( $slash ) {
 459                                                 # Closing a tag... is it the one we just opened?
 460                                                 $ot = @array_pop( $tagstack );
 461                                                 if ( $ot != $t ) {
 462                                                         if ( isset( $htmlsingleallowed[$ot] ) ) {
 463                                                                 # Pop all elements with an optional close tag
 464                                                                 # and see if we find a match below them
 465                                                                 $optstack = array();
 466                                                                 array_push( $optstack, $ot );
 467                                                                 wfSuppressWarnings();
 468                                                                 $ot = array_pop( $tagstack );
 469                                                                 wfRestoreWarnings();
 470                                                                 while ( $ot != $t && isset( $htmlsingleallowed[$ot] ) ) {
 471                                                                         array_push( $optstack, $ot );
 472                                                                         wfSuppressWarnings();
 473                                                                         $ot = array_pop( $tagstack );
 474                                                                         wfRestoreWarnings();
 475                                                                 }
 476                                                                 if ( $t != $ot ) {
 477                                                                         # No match. Push the optional elements back again
 478                                                                         $badtag = true;
 479                                                                         wfSuppressWarnings();
 480                                                                         $ot = array_pop( $optstack );
 481                                                                         wfRestoreWarnings();
 482                                                                         while ( $ot ) {
 483                                                                                 array_push( $tagstack, $ot );
 484                                                                                 wfSuppressWarnings();
 485                                                                                 $ot = array_pop( $optstack );
 486                                                                                 wfRestoreWarnings();
 487                                                                         }
 488                                                                 }
 489                                                         } else {
 490                                                                 @array_push( $tagstack, $ot );
 491                                                                 # <li> can be nested in <ul> or <ol>, skip those cases:
 492                                                                 if ( !isset( $htmllist[$ot] ) || !isset( $listtags[$t] ) ) {
 493                                                                         $badtag = true;
 494                                                                 }
 495                                                         }
 496                                                 } else {
 497                                                         if ( $t == 'table' ) {
 498                                                                 $tagstack = array_pop( $tablestack );
 499                                                         }
 500                                                 }
 501                                                 $newparams = '';
 502                                         } else {
 503                                                 # Keep track for later
 504                                                 if ( isset( $tabletags[$t] ) &&
 505                                                 !in_array( 'table', $tagstack ) ) {
 506                                                         $badtag = true;
 507                                                 } elseif ( in_array( $t, $tagstack ) &&
 508                                                 !isset( $htmlnest [$t ] ) ) {
 509                                                         $badtag = true;
 510                                                 # Is it a self closed htmlpair ? (bug 5487)
 511                                                 } elseif ( $brace == '/>' &&
 512                                                 isset( $htmlpairs[$t] ) ) {
 513                                                         $badtag = true;
 514                                                 } elseif ( isset( $htmlsingleonly[$t] ) ) {
 515                                                         # Hack to force empty tag for uncloseable elements
 516                                                         $brace = '/>';
 517                                                 } elseif ( isset( $htmlsingle[$t] ) ) {
 518                                                         # Hack to not close $htmlsingle tags
 519                                                         $brace = null;
 520                                                 } elseif ( isset( $tabletags[$t] )
 521                                                 && in_array( $t, $tagstack ) ) {
 522                                                         // New table tag but forgot to close the previous one
 523                                                         $text .= "</$t>";
 524                                                 } else {
 525                                                         if ( $t == 'table' ) {
 526                                                                 array_push( $tablestack, $tagstack );
 527                                                                 $tagstack = array();
 528                                                         }
 529                                                         array_push( $tagstack, $t );
 530                                                 }
 531
 532                                                 # Replace any variables or template parameters with
 533                                                 # plaintext results.
 534                                                 if( is_callable( $processCallback ) ) {
 535                                                         call_user_func_array( $processCallback, array( &$params, $args ) );
 536                                                 }
 537
 538                                                 if ( !Sanitizer::validateTag( $params, $t ) ) {
 539                                                         $badtag = true;
 540                                                 }
 541
 542                                                 # Strip non-approved attributes from the tag
 543                                                 $newparams = Sanitizer::fixTagAttributes( $params, $t );
 544                                         }
 545                                         if ( !$badtag ) {
 546                                                 $rest = str_replace( '>', '&gt;', $rest );
 547                                                 $close = ( $brace == '/>' && !$slash ) ? ' /' : '';
 548                                                 $text .= "<$slash$t$newparams$close>$rest";
 549                                                 continue;
 550                                         }
 551                                 }
 552                                 $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 553                         }
 554                         # Close off any remaining tags
 555                         while ( is_array( $tagstack ) && ($t = array_pop( $tagstack )) ) {
 556                                 $text .= "</$t>\n";
 557                                 if ( $t == 'table' ) { $tagstack = array_pop( $tablestack ); }
 558                         }
 559                 } else {
 560                         # this might be possible using tidy itself
 561                         foreach ( $bits as $x ) {
 562                                 preg_match( '/^(\\/?)(\\w+)([^>]*?)(\\/{0,1}>)([^<]*)$/',
 563                                 $x, $regs );
 564                                 @list( /* $qbar */, $slash, $t, $params, $brace, $rest ) = $regs;
 565                                 if ( isset( $htmlelements[$t = strtolower( $t )] ) ) {
 566                                         if( is_callable( $processCallback ) ) {
 567                                                 call_user_func_array( $processCallback, array( &$params, $args ) );
 568                                         }
 569                                         $newparams = Sanitizer::fixTagAttributes( $params, $t );
 570                                         $rest = str_replace( '>', '&gt;', $rest );
 571                                         $text .= "<$slash$t$newparams$brace$rest";
 572                                 } else {
 573                                         $text .= '&lt;' . str_replace( '>', '&gt;', $x);
 574                                 }
 575                         }
 576                 }
 577                 wfProfileOut( __METHOD__ );
 578                 return $text;
 579         }
 580
 581         /**
 582          * Remove '<!--', '-->', and everything between.
 583          * To avoid leaving blank lines, when a comment is both preceded
 584          * and followed by a newline (ignoring spaces), trim leading and
 585          * trailing spaces and one of the newlines.
 586          *
 587          * @private
 588          * @param $text String
 589          * @return string
 590          */
 591         static function removeHTMLcomments( $text ) {
 592                 wfProfileIn( __METHOD__ );
 593                 while (($start = strpos($text, '<!--')) !== false) {
 594                         $end = strpos($text, '-->', $start + 4);
 595                         if ($end === false) {
 596                                 # Unterminated comment; bail out
 597                                 break;
 598                         }
 599
 600                         $end += 3;
 601
 602                         # Trim space and newline if the comment is both
 603                         # preceded and followed by a newline
 604                         $spaceStart = max($start - 1, 0);
 605                         $spaceLen = $end - $spaceStart;
 606                         while (substr($text, $spaceStart, 1) === ' ' && $spaceStart > 0) {
 607                                 $spaceStart--;
 608                                 $spaceLen++;
 609                         }
 610                         while (substr($text, $spaceStart + $spaceLen, 1) === ' ')
 611                                 $spaceLen++;
 612                         if (substr($text, $spaceStart, 1) === "\n" and substr($text, $spaceStart + $spaceLen, 1) === "\n") {
 613                                 # Remove the comment, leading and trailing
 614                                 # spaces, and leave only one newline.
 615                                 $text = substr_replace($text, "\n", $spaceStart, $spaceLen + 1);
 616                         }
 617                         else {
 618                                 # Remove just the comment.
 619                                 $text = substr_replace($text, '', $start, $end - $start);
 620                         }
 621                 }
 622                 wfProfileOut( __METHOD__ );
 623                 return $text;
 624         }
 625
 626         /**
 627          * Take an array of attribute names and values and fix some deprecated values
 628          * for the given element type.
 629          * This does not validate properties, so you should ensure that you call
 630          * validateTagAttributes AFTER this to ensure that the resulting style rule
 631          * this may add is safe.
 632          *
 633          * - Converts most presentational attributes like align into inline css
 634          *
 635          * @param $attribs Array
 636          * @param $element String
 637          * @return Array
 638          */
 639         static function fixDeprecatedAttributes( $attribs, $element ) {
 640                 global $wgHtml5, $wgCleanupPresentationalAttributes;
 641
 642                 // presentational attributes were removed from html5, we can leave them
 643                 // in when html5 is turned off
 644                 if ( !$wgHtml5 || !$wgCleanupPresentationalAttributes ) {
 645                         return $attribs;
 646                 }
 647
 648                 $table = array( 'table' );
 649                 $cells = array( 'td', 'th' );
 650                 $colls = array( 'col', 'colgroup' );
 651                 $tblocks = array( 'tbody', 'tfoot', 'thead' );
 652                 $h = array( 'h1', 'h2', 'h3', 'h4', 'h5', 'h6' );
 653
 654                 $presentationalAttribs = array(
 655                         'align' => array( 'text-align', array_merge( array( 'caption', 'hr', 'div', 'p', 'tr' ), $table, $cells, $colls, $tblocks, $h ) ),
 656                         'clear' => array( 'clear', array( 'br' ) ),
 657                         'height' => array( 'height', $cells ),
 658                         'nowrap' => array( 'white-space', $cells ),
 659                         'size' => array( 'height', array( 'hr' ) ),
 660                         'type' => array( 'list-style-type', array( 'li', 'ol', 'ul' ) ),
 661                         'valign' => array( 'vertical-align', array_merge( $cells, $colls, $tblocks ) ),
 662                         'width' => array( 'width', array_merge( array( 'hr', 'pre' ), $table, $cells, $colls ) ),
 663                 );
 664
 665                 // Ensure that any upper case or mixed case attributes are converted to lowercase
 666                 foreach ( $attribs as $attribute => $value ) {
 667                         if ( $attribute !== strtolower( $attribute ) && array_key_exists( strtolower( $attribute ), $presentationalAttribs ) ) {
 668                                 $attribs[strtolower( $attribute )] = $value;
 669                                 unset( $attribs[$attribute] );
 670                         }
 671                 }
 672
 673                 $style = "";
 674                 foreach ( $presentationalAttribs as $attribute => $info ) {
 675                         list( $property, $elements ) = $info;
 676
 677                         // Skip if this attribute is not relevant to this element
 678                         if ( !in_array( $element, $elements ) ) {
 679                                 continue;
 680                         }
 681
 682                         // Skip if the attribute is not used
 683                         if ( !array_key_exists( $attribute, $attribs ) ) {
 684                                 continue;
 685                         }
 686
 687                         $value = $attribs[$attribute];
 688
 689                         // For nowrap the value should be nowrap instead of whatever text is in the value
 690                         if ( $attribute === 'nowrap' ) {
 691                                 $value = 'nowrap';
 692                         }
 693
 694                         // clear="all" is clear: both; in css
 695                         if ( $attribute === 'clear' && strtolower( $value ) === 'all' ) {
 696                                 $value = 'both';
 697                         }
 698
 699                         // Size based properties should have px applied to them if they have no unit
 700                         if ( in_array( $attribute, array( 'height', 'width', 'size' ) ) ) {
 701                                 if ( preg_match( '/^[\d.]+$/', $value ) ) {
 702                                         $value = "{$value}px";
 703                                 }
 704                         }
 705
 706                         $style .= " $property: $value;";
 707
 708                         unset( $attribs[$attribute] );
 709                 }
 710
 711                 if ( $style ) {
 712                         // Prepend our style rules so that they can be overridden by user css
 713                         if ( isset($attribs['style']) ) {
 714                                 $style .= " " . $attribs['style'];
 715                         }
 716                         $attribs['style'] = trim($style);
 717                 }
 718
 719                 return $attribs;
 720         }
 721
 722         /**
 723          * Takes attribute names and values for a tag and the tah name and
 724          * validates that the tag is allowed to be present.
 725          * This DOES NOT validate the attributes, nor does it validate the
 726          * tags themselves. This method only handles the special circumstances
 727          * where we may want to allow a tag within content but ONLY when it has
 728          * specific attributes set.
 729          *
 730          * @param $
 731          */
 732         static function validateTag( $params, $element ) {
 733                 $params = Sanitizer::decodeTagAttributes( $params );
 734
 735                 if ( $element == 'meta' || $element == 'link' ) {
 736                         if ( !isset( $params['itemprop'] ) ) {
 737                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
 738                                 return false;
 739                         }
 740                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 741                                 // <meta> must have a content="" for the itemprop
 742                                 return false;
 743                         }
 744                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 745                                 // <link> must have an associated href=""
 746                                 return false;
 747                         }
 748                 }
 749
 750                 return true;
 751         }
 752
 753         /**
 754          * Take an array of attribute names and values and normalize or discard
 755          * illegal values for the given element type.
 756          *
 757          * - Discards attributes not on a whitelist for the given element
 758          * - Unsafe style attributes are discarded
 759          * - Invalid id attributes are reencoded
 760          *
 761          * @param $attribs Array
 762          * @param $element String
 763          * @return Array
 764          *
 765          * @todo Check for legal values where the DTD limits things.
 766          * @todo Check for unique id attribute :P
 767          */
 768         static function validateTagAttributes( $attribs, $element ) {
 769                 return Sanitizer::validateAttributes( $attribs,
 770                         Sanitizer::attributeWhitelist( $element ) );
 771         }
 772
 773         /**
 774          * Take an array of attribute names and values and normalize or discard
 775          * illegal values for the given whitelist.
 776          *
 777          * - Discards attributes not the given whitelist
 778          * - Unsafe style attributes are discarded
 779          * - Invalid id attributes are reencoded
 780          *
 781          * @param $attribs Array
 782          * @param $whitelist Array: list of allowed attribute names
 783          * @return Array
 784          *
 785          * @todo Check for legal values where the DTD limits things.
 786          * @todo Check for unique id attribute :P
 787          */
 788         static function validateAttributes( $attribs, $whitelist ) {
 789                 global $wgAllowRdfaAttributes, $wgAllowMicrodataAttributes, $wgHtml5;
 790
 791                 $whitelist = array_flip( $whitelist );
 792                 $hrefExp = '/^(' . wfUrlProtocols() . ')[^\s]+$/';
 793
 794                 $out = array();
 795                 foreach( $attribs as $attribute => $value ) {
 796                         #allow XML namespace declaration if RDFa is enabled
 797                         if ( $wgAllowRdfaAttributes && preg_match( self::XMLNS_ATTRIBUTE_PATTERN, $attribute ) ) {
 798                                 if ( !preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 799                                         $out[$attribute] = $value;
 800                                 }
 801
 802                                 continue;
 803                         }
 804
 805                         # Allow any attribute beginning with "data-", if in HTML5 mode
 806                         if ( !($wgHtml5 && preg_match( '/^data-/i', $attribute )) && !isset( $whitelist[$attribute] ) ) {
 807                                 continue;
 808                         }
 809
 810                         # Strip javascript "expression" from stylesheets.
 811                         # http://msdn.microsoft.com/workshop/author/dhtml/overview/recalc.asp
 812                         if( $attribute == 'style' ) {
 813                                 $value = Sanitizer::checkCss( $value );
 814                         }
 815
 816                         if ( $attribute === 'id' ) {
 817                                 $value = Sanitizer::escapeId( $value, 'noninitial' );
 818                         }
 819
 820                         //RDFa and microdata properties allow URLs, URIs and/or CURIs. check them for sanity
 821                         if ( $attribute === 'rel' || $attribute === 'rev' ||
 822                                 $attribute === 'about' || $attribute === 'property' || $attribute === 'resource' || #RDFa
 823                                 $attribute === 'datatype' || $attribute === 'typeof' ||                             #RDFa
 824                                 $attribute === 'itemid' || $attribute === 'itemprop' || $attribute === 'itemref' || #HTML5 microdata
 825                                 $attribute === 'itemscope' || $attribute === 'itemtype' ) {                         #HTML5 microdata
 826
 827                                 //Paranoia. Allow "simple" values but suppress javascript
 828                                 if ( preg_match( self::EVIL_URI_PATTERN, $value ) ) {
 829                                         continue;
 830                                 }
 831                         }
 832
 833                         # NOTE: even though elements using href/src are not allowed directly, supply
 834                         #       validation code that can be used by tag hook handlers, etc
 835                         if ( $attribute === 'href' || $attribute === 'src' ) {
 836                                 if ( !preg_match( $hrefExp, $value ) ) {
 837                                         continue; //drop any href or src attributes not using an allowed protocol.
 838                                                   //NOTE: this also drops all relative URLs
 839                                 }
 840                         }
 841
 842                         // If this attribute was previously set, override it.
 843                         // Output should only have one attribute of each name.
 844                         $out[$attribute] = $value;
 845                 }
 846
 847                 if ( $wgAllowMicrodataAttributes ) {
 848                         # itemtype, itemid, itemref don't make sense without itemscope
 849                         if ( !array_key_exists( 'itemscope', $out ) ) {
 850                                 unset( $out['itemtype'] );
 851                                 unset( $out['itemid'] );
 852                                 unset( $out['itemref'] );
 853                         }
 854                         # TODO: Strip itemprop if we aren't descendants of an itemscope or pointed to by an itemref.
 855                 }
 856                 return $out;
 857         }
 858
 859         /**
 860          * Merge two sets of HTML attributes.  Conflicting items in the second set
 861          * will override those in the first, except for 'class' attributes which
 862          * will be combined (if they're both strings).
 863          *
 864          * @todo implement merging for other attributes such as style
 865          * @param $a Array
 866          * @param $b Array
 867          * @return array
 868          */
 869         static function mergeAttributes( $a, $b ) {
 870                 $out = array_merge( $a, $b );
 871                 if( isset( $a['class'] ) && isset( $b['class'] )
 872                 && is_string( $a['class'] ) && is_string( $b['class'] )
 873                 && $a['class'] !== $b['class'] ) {
 874                         $classes = preg_split( '/\s+/', "{$a['class']} {$b['class']}",
 875                                 -1, PREG_SPLIT_NO_EMPTY );
 876                         $out['class'] = implode( ' ', array_unique( $classes ) );
 877                 }
 878                 return $out;
 879         }
 880
 881         /**
 882          * Pick apart some CSS and check it for forbidden or unsafe structures.
 883          * Returns a sanitized string. This sanitized string will have
 884          * character references and escape sequences decoded, and comments
 885          * stripped. If the input is just too evil, only a comment complaining
 886          * about evilness will be returned.
 887          *
 888          * Currently URL references, 'expression', 'tps' are forbidden.
 889          *
 890          * NOTE: Despite the fact that character references are decoded, the
 891          * returned string may contain character references given certain
 892          * clever input strings. These character references must
 893          * be escaped before the return value is embedded in HTML.
 894          *
 895          * @param $value String
 896          * @return String
 897          */
 898         static function checkCss( $value ) {
 899                 // Decode character references like &#123;
 900                 $value = Sanitizer::decodeCharReferences( $value );
 901
 902                 // Decode escape sequences and line continuation
 903                 // See the grammar in the CSS 2 spec, appendix D.
 904                 // This has to be done AFTER decoding character references.
 905                 // This means it isn't possible for this function to return
 906                 // unsanitized escape sequences. It is possible to manufacture
 907                 // input that contains character references that decode to
 908                 // escape sequences that decode to character references, but
 909                 // it's OK for the return value to contain character references
 910                 // because the caller is supposed to escape those anyway.
 911                 static $decodeRegex;
 912                 if ( !$decodeRegex ) {
 913                         $space = '[\\x20\\t\\r\\n\\f]';
 914                         $nl = '(?:\\n|\\r\\n|\\r|\\f)';
 915                         $backslash = '\\\\';
 916                         $decodeRegex = "/ $backslash
 917                                 (?:
 918                                         ($nl) |  # 1. Line continuation
 919                                         ([0-9A-Fa-f]{1,6})$space? |  # 2. character number
 920                                         (.) | # 3. backslash cancelling special meaning
 921                                         () | # 4. backslash at end of string
 922                                 )/xu";
 923                 }
 924                 $value = preg_replace_callback( $decodeRegex,
 925                         array( __CLASS__, 'cssDecodeCallback' ), $value );
 926
 927                 // Remove any comments; IE gets token splitting wrong
 928                 // This must be done AFTER decoding character references and
 929                 // escape sequences, because those steps can introduce comments
 930                 // This step cannot introduce character references or escape
 931                 // sequences, because it replaces comments with spaces rather
 932                 // than removing them completely.
 933                 $value = StringUtils::delimiterReplace( '/*', '*/', ' ', $value );
 934
 935                 // Remove anything after a comment-start token, to guard against
 936                 // incorrect client implementations.
 937                 $commentPos = strpos( $value, '/*' );
 938                 if ( $commentPos !== false ) {
 939                         $value = substr( $value, 0, $commentPos );
 940                 }
 941
 942                 // Reject problematic keywords and control characters
 943                 if ( preg_match( '/[\000-\010\016-\037\177]/', $value ) ) {
 944                         return '/* invalid control char */';
 945                 } elseif ( preg_match( '! expression | filter\s*: | accelerator\s*: | url\s*\( !ix', $value ) ) {
 946                         return '/* insecure input */';
 947                 }
 948                 return $value;
 949         }
 950
 951         /**
 952          * @param $matches array
 953          * @return String
 954          */
 955         static function cssDecodeCallback( $matches ) {
 956                 if ( $matches[1] !== '' ) {
 957                         // Line continuation
 958                         return '';
 959                 } elseif ( $matches[2] !== '' ) {
 960                         $char = codepointToUtf8( hexdec( $matches[2] ) );
 961                 } elseif ( $matches[3] !== '' ) {
 962                         $char = $matches[3];
 963                 } else {
 964                         $char = '\\';
 965                 }
 966                 if ( $char == "\n" || $char == '"' || $char == "'" || $char == '\\' ) {
 967                         // These characters need to be escaped in strings
 968                         // Clean up the escape sequence to avoid parsing errors by clients
 969                         return '\\' . dechex( ord( $char ) ) . ' ';
 970                 } else {
 971                         // Decode unnecessary escape
 972                         return $char;
 973                 }
 974         }
 975
 976         /**
 977          * Take a tag soup fragment listing an HTML element's attributes
 978          * and normalize it to well-formed XML, discarding unwanted attributes.
 979          * Output is safe for further wikitext processing, with escaping of
 980          * values that could trigger problems.
 981          *
 982          * - Normalizes attribute names to lowercase
 983          * - Discards attributes not on a whitelist for the given element
 984          * - Turns broken or invalid entities into plaintext
 985          * - Double-quotes all attribute values
 986          * - Attributes without values are given the name as attribute
 987          * - Double attributes are discarded
 988          * - Unsafe style attributes are discarded
 989          * - Prepends space if there are attributes.
 990          *
 991          * @param $text String
 992          * @param $element String
 993          * @return String
 994          */
 995         static function fixTagAttributes( $text, $element ) {
 996                 if( trim( $text ) == '' ) {
 997                         return '';
 998                 }
 999
1000                 $decoded = Sanitizer::decodeTagAttributes( $text );
1001                 $decoded = Sanitizer::fixDeprecatedAttributes( $decoded, $element );
1002                 $stripped = Sanitizer::validateTagAttributes( $decoded, $element );
1003
1004                 $attribs = array();
1005                 foreach( $stripped as $attribute => $value ) {
1006                         $encAttribute = htmlspecialchars( $attribute );
1007                         $encValue = Sanitizer::safeEncodeAttribute( $value );
1008
1009                         $attribs[] = "$encAttribute=\"$encValue\"";
1010                 }
1011                 return count( $attribs ) ? ' ' . implode( ' ', $attribs ) : '';
1012         }
1013
1014         /**
1015          * Encode an attribute value for HTML output.
1016          * @param $text String
1017          * @return HTML-encoded text fragment
1018          */
1019         static function encodeAttribute( $text ) {
1020                 $encValue = htmlspecialchars( $text, ENT_QUOTES );
1021
1022                 // Whitespace is normalized during attribute decoding,
1023                 // so if we've been passed non-spaces we must encode them
1024                 // ahead of time or they won't be preserved.
1025                 $encValue = strtr( $encValue, array(
1026                         "\n" => '&#10;',
1027                         "\r" => '&#13;',
1028                         "\t" => '&#9;',
1029                 ) );
1030
1031                 return $encValue;
1032         }
1033
1034         /**
1035          * Encode an attribute value for HTML tags, with extra armoring
1036          * against further wiki processing.
1037          * @param $text String
1038          * @return HTML-encoded text fragment
1039          */
1040         static function safeEncodeAttribute( $text ) {
1041                 $encValue = Sanitizer::encodeAttribute( $text );
1042
1043                 # Templates and links may be expanded in later parsing,
1044                 # creating invalid or dangerous output. Suppress this.
1045                 $encValue = strtr( $encValue, array(
1046                         '<'    => '&lt;',   // This should never happen,
1047                         '>'    => '&gt;',   // we've received invalid input
1048                         '"'    => '&quot;', // which should have been escaped.
1049                         '{'    => '&#123;',
1050                         '['    => '&#91;',
1051                         "''"   => '&#39;&#39;',
1052                         'ISBN' => '&#73;SBN',
1053                         'RFC'  => '&#82;FC',
1054                         'PMID' => '&#80;MID',
1055                         '|'    => '&#124;',
1056                         '__'   => '&#95;_',
1057                 ) );
1058
1059                 # Stupid hack
1060                 $encValue = preg_replace_callback(
1061                         '/(' . wfUrlProtocols() . ')/',
1062                         array( 'Sanitizer', 'armorLinksCallback' ),
1063                         $encValue );
1064                 return $encValue;
1065         }
1066
1067         /**
1068          * Given a value, escape it so that it can be used in an id attribute and
1069          * return it.  This will use HTML5 validation if $wgExperimentalHtmlIds is
1070          * true, allowing anything but ASCII whitespace.  Otherwise it will use
1071          * HTML 4 rules, which means a narrow subset of ASCII, with bad characters
1072          * escaped with lots of dots.
1073          *
1074          * To ensure we don't have to bother escaping anything, we also strip ', ",
1075          * & even if $wgExperimentalIds is true.  TODO: Is this the best tactic?
1076          * We also strip # because it upsets IE, and % because it could be
1077          * ambiguous if it's part of something that looks like a percent escape
1078          * (which don't work reliably in fragments cross-browser).
1079          *
1080          * @see http://www.w3.org/TR/html401/types.html#type-name Valid characters
1081          *                                                          in the id and
1082          *                                                          name attributes
1083          * @see http://www.w3.org/TR/html401/struct/links.html#h-12.2.3 Anchors with the id attribute
1084          * @see http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-id-attribute
1085          *   HTML5 definition of id attribute
1086          *
1087          * @param $id String: id to escape
1088          * @param $options Mixed: string or array of strings (default is array()):
1089          *   'noninitial': This is a non-initial fragment of an id, not a full id,
1090          *       so don't pay attention if the first character isn't valid at the
1091          *       beginning of an id.  Only matters if $wgExperimentalHtmlIds is
1092          *       false.
1093          *   'legacy': Behave the way the old HTML 4-based ID escaping worked even
1094          *       if $wgExperimentalHtmlIds is used, so we can generate extra
1095          *       anchors and links won't break.
1096          * @return String
1097          */
1098         static function escapeId( $id, $options = array() ) {
1099                 global $wgHtml5, $wgExperimentalHtmlIds;
1100                 $options = (array)$options;
1101
1102                 if ( $wgHtml5 && $wgExperimentalHtmlIds && !in_array( 'legacy', $options ) ) {
1103                         $id = Sanitizer::decodeCharReferences( $id );
1104                         $id = preg_replace( '/[ \t\n\r\f_\'"&#%]+/', '_', $id );
1105                         $id = trim( $id, '_' );
1106                         if ( $id === '' ) {
1107                                 # Must have been all whitespace to start with.
1108                                 return '_';
1109                         } else {
1110                                 return $id;
1111                         }
1112                 }
1113
1114                 # HTML4-style escaping
1115                 static $replace = array(
1116                         '%3A' => ':',
1117                         '%' => '.'
1118                 );
1119
1120                 $id = urlencode( Sanitizer::decodeCharReferences( strtr( $id, ' ', '_' ) ) );
1121                 $id = str_replace( array_keys( $replace ), array_values( $replace ), $id );
1122
1123                 if ( !preg_match( '/^[a-zA-Z]/', $id )
1124                 && !in_array( 'noninitial', $options ) )  {
1125                         // Initial character must be a letter!
1126                         $id = "x$id";
1127                 }
1128                 return $id;
1129         }
1130
1131         /**
1132          * Given a value, escape it so that it can be used as a CSS class and
1133          * return it.
1134          *
1135          * @todo For extra validity, input should be validated UTF-8.
1136          *
1137          * @see http://www.w3.org/TR/CSS21/syndata.html Valid characters/format
1138          *
1139          * @param $class String
1140          * @return String
1141          */
1142         static function escapeClass( $class ) {
1143                 // Convert ugly stuff to underscores and kill underscores in ugly places
1144                 return rtrim(preg_replace(
1145                         array('/(^[0-9\\-])|[\\x00-\\x20!"#$%&\'()*+,.\\/:;<=>?@[\\]^`{|}~]|\\xC2\\xA0/','/_+/'),
1146                         '_',
1147                         $class ), '_');
1148         }
1149
1150         /**
1151          * Given HTML input, escape with htmlspecialchars but un-escape entites.
1152          * This allows (generally harmless) entities like &#160; to survive.
1153          *
1154          * @param $html String to escape
1155          * @return String: escaped input
1156          */
1157         static function escapeHtmlAllowEntities( $html ) {
1158                 $html = Sanitizer::decodeCharReferences( $html );
1159                 # It seems wise to escape ' as well as ", as a matter of course.  Can't
1160                 # hurt.
1161                 $html = htmlspecialchars( $html, ENT_QUOTES );
1162                 return $html;
1163         }
1164
1165         /**
1166          * Regex replace callback for armoring links against further processing.
1167          * @param $matches Array
1168          * @return string
1169          */
1170         private static function armorLinksCallback( $matches ) {
1171                 return str_replace( ':', '&#58;', $matches[1] );
1172         }
1173
1174         /**
1175          * Return an associative array of attribute names and values from
1176          * a partial tag string. Attribute names are forces to lowercase,
1177          * character references are decoded to UTF-8 text.
1178          *
1179          * @param $text String
1180          * @return Array
1181          */
1182         public static function decodeTagAttributes( $text ) {
1183                 if( trim( $text ) == '' ) {
1184                         return array();
1185                 }
1186
1187                 $attribs = array();
1188                 $pairs = array();
1189                 if( !preg_match_all(
1190                         self::getAttribsRegex(),
1191                         $text,
1192                         $pairs,
1193                         PREG_SET_ORDER ) ) {
1194                         return $attribs;
1195                 }
1196
1197                 foreach( $pairs as $set ) {
1198                         $attribute = strtolower( $set[1] );
1199                         $value = Sanitizer::getTagAttributeCallback( $set );
1200
1201                         // Normalize whitespace
1202                         $value = preg_replace( '/[\t\r\n ]+/', ' ', $value );
1203                         $value = trim( $value );
1204
1205                         // Decode character references
1206                         $attribs[$attribute] = Sanitizer::decodeCharReferences( $value );
1207                 }
1208                 return $attribs;
1209         }
1210
1211         /**
1212          * Pick the appropriate attribute value from a match set from the
1213          * attribs regex matches.
1214          *
1215          * @param $set Array
1216          * @return String
1217          */
1218         private static function getTagAttributeCallback( $set ) {
1219                 if( isset( $set[6] ) ) {
1220                         # Illegal #XXXXXX color with no quotes.
1221                         return $set[6];
1222                 } elseif( isset( $set[5] ) ) {
1223                         # No quotes.
1224                         return $set[5];
1225                 } elseif( isset( $set[4] ) ) {
1226                         # Single-quoted
1227                         return $set[4];
1228                 } elseif( isset( $set[3] ) ) {
1229                         # Double-quoted
1230                         return $set[3];
1231                 } elseif( !isset( $set[2] ) ) {
1232                         # In XHTML, attributes must have a value.
1233                         # For 'reduced' form, return explicitly the attribute name here.
1234                         return $set[1];
1235                 } else {
1236                         throw new MWException( "Tag conditions not met. This should never happen and is a bug." );
1237                 }
1238         }
1239
1240         /**
1241          * Normalize whitespace and character references in an XML source-
1242          * encoded text for an attribute value.
1243          *
1244          * See http://www.w3.org/TR/REC-xml/#AVNormalize for background,
1245          * but note that we're not returning the value, but are returning
1246          * XML source fragments that will be slapped into output.
1247          *
1248          * @param $text String
1249          * @return String
1250          */
1251         private static function normalizeAttributeValue( $text ) {
1252                 return str_replace( '"', '&quot;',
1253                         self::normalizeWhitespace(
1254                                 Sanitizer::normalizeCharReferences( $text ) ) );
1255         }
1256
1257         /**
1258          * @param $text string
1259          * @return mixed
1260          */
1261         private static function normalizeWhitespace( $text ) {
1262                 return preg_replace(
1263                         '/\r\n|[\x20\x0d\x0a\x09]/',
1264                         ' ',
1265                         $text );
1266         }
1267
1268         /**
1269          * Normalizes whitespace in a section name, such as might be returned
1270          * by Parser::stripSectionName(), for use in the id's that are used for
1271          * section links.
1272          *
1273          * @param $section String
1274          * @return String
1275          */
1276         static function normalizeSectionNameWhitespace( $section ) {
1277                 return trim( preg_replace( '/[ _]+/', ' ', $section ) );
1278         }
1279
1280         /**
1281          * Ensure that any entities and character references are legal
1282          * for XML and XHTML specifically. Any stray bits will be
1283          * &amp;-escaped to result in a valid text fragment.
1284          *
1285          * a. named char refs can only be &lt; &gt; &amp; &quot;, others are
1286          *   numericized (this way we're well-formed even without a DTD)
1287          * b. any numeric char refs must be legal chars, not invalid or forbidden
1288          * c. use &#x, not &#X
1289          * d. fix or reject non-valid attributes
1290          *
1291          * @param $text String
1292          * @return String
1293          * @private
1294          */
1295         static function normalizeCharReferences( $text ) {
1296                 return preg_replace_callback(
1297                         self::CHAR_REFS_REGEX,
1298                         array( 'Sanitizer', 'normalizeCharReferencesCallback' ),
1299                         $text );
1300         }
1301         /**
1302          * @param $matches String
1303          * @return String
1304          */
1305         static function normalizeCharReferencesCallback( $matches ) {
1306                 $ret = null;
1307                 if( $matches[1] != '' ) {
1308                         $ret = Sanitizer::normalizeEntity( $matches[1] );
1309                 } elseif( $matches[2] != '' ) {
1310                         $ret = Sanitizer::decCharReference( $matches[2] );
1311                 } elseif( $matches[3] != ''  ) {
1312                         $ret = Sanitizer::hexCharReference( $matches[3] );
1313                 }
1314                 if( is_null( $ret ) ) {
1315                         return htmlspecialchars( $matches[0] );
1316                 } else {
1317                         return $ret;
1318                 }
1319         }
1320
1321         /**
1322          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1323          * return the equivalent numeric entity reference (except for the core &lt;
1324          * &gt; &amp; &quot;). If the entity is a MediaWiki-specific alias, returns
1325          * the HTML equivalent. Otherwise, returns HTML-escaped text of
1326          * pseudo-entity source (eg &amp;foo;)
1327          *
1328          * @param $name String
1329          * @return String
1330          */
1331         static function normalizeEntity( $name ) {
1332                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1333                         return '&' . self::$htmlEntityAliases[$name] . ';';
1334                 } elseif ( in_array( $name,
1335                 array( 'lt', 'gt', 'amp', 'quot' ) ) ) {
1336                         return "&$name;";
1337                 } elseif ( isset( self::$htmlEntities[$name] ) ) {
1338                         return '&#' . self::$htmlEntities[$name] . ';';
1339                 } else {
1340                         return "&amp;$name;";
1341                 }
1342         }
1343
1344         /**
1345          * @param $codepoint
1346          * @return null|string
1347          */
1348         static function decCharReference( $codepoint ) {
1349                 $point = intval( $codepoint );
1350                 if( Sanitizer::validateCodepoint( $point ) ) {
1351                         return sprintf( '&#%d;', $point );
1352                 } else {
1353                         return null;
1354                 }
1355         }
1356
1357         /**
1358          * @param $codepoint
1359          * @return null|string
1360          */
1361         static function hexCharReference( $codepoint ) {
1362                 $point = hexdec( $codepoint );
1363                 if( Sanitizer::validateCodepoint( $point ) ) {
1364                         return sprintf( '&#x%x;', $point );
1365                 } else {
1366                         return null;
1367                 }
1368         }
1369
1370         /**
1371          * Returns true if a given Unicode codepoint is a valid character in XML.
1372          * @param $codepoint Integer
1373          * @return Boolean
1374          */
1375         private static function validateCodepoint( $codepoint ) {
1376                 return ($codepoint ==    0x09)
1377                         || ($codepoint ==    0x0a)
1378                         || ($codepoint ==    0x0d)
1379                         || ($codepoint >=    0x20 && $codepoint <=   0xd7ff)
1380                         || ($codepoint >=  0xe000 && $codepoint <=   0xfffd)
1381                         || ($codepoint >= 0x10000 && $codepoint <= 0x10ffff);
1382         }
1383
1384         /**
1385          * Decode any character references, numeric or named entities,
1386          * in the text and return a UTF-8 string.
1387          *
1388          * @param $text String
1389          * @return String
1390          */
1391         public static function decodeCharReferences( $text ) {
1392                 return preg_replace_callback(
1393                         self::CHAR_REFS_REGEX,
1394                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1395                         $text );
1396         }
1397
1398         /**
1399          * Decode any character references, numeric or named entities,
1400          * in the next and normalize the resulting string. (bug 14952)
1401          *
1402          * This is useful for page titles, not for text to be displayed,
1403          * MediaWiki allows HTML entities to escape normalization as a feature.
1404          *
1405          * @param $text String (already normalized, containing entities)
1406          * @return String (still normalized, without entities)
1407          */
1408         public static function decodeCharReferencesAndNormalize( $text ) {
1409                 global $wgContLang;
1410                 $text = preg_replace_callback(
1411                         self::CHAR_REFS_REGEX,
1412                         array( 'Sanitizer', 'decodeCharReferencesCallback' ),
1413                         $text, /* limit */ -1, $count );
1414
1415                 if ( $count ) {
1416                         return $wgContLang->normalize( $text );
1417                 } else {
1418                         return $text;
1419                 }
1420         }
1421
1422         /**
1423          * @param $matches String
1424          * @return String
1425          */
1426         static function decodeCharReferencesCallback( $matches ) {
1427                 if( $matches[1] != '' ) {
1428                         return Sanitizer::decodeEntity( $matches[1] );
1429                 } elseif( $matches[2] != '' ) {
1430                         return  Sanitizer::decodeChar( intval( $matches[2] ) );
1431                 } elseif( $matches[3] != ''  ) {
1432                         return  Sanitizer::decodeChar( hexdec( $matches[3] ) );
1433                 }
1434                 # Last case should be an ampersand by itself
1435                 return $matches[0];
1436         }
1437
1438         /**
1439          * Return UTF-8 string for a codepoint if that is a valid
1440          * character reference, otherwise U+FFFD REPLACEMENT CHARACTER.
1441          * @param $codepoint Integer
1442          * @return String
1443          * @private
1444          */
1445         static function decodeChar( $codepoint ) {
1446                 if( Sanitizer::validateCodepoint( $codepoint ) ) {
1447                         return codepointToUtf8( $codepoint );
1448                 } else {
1449                         return UTF8_REPLACEMENT;
1450                 }
1451         }
1452
1453         /**
1454          * If the named entity is defined in the HTML 4.0/XHTML 1.0 DTD,
1455          * return the UTF-8 encoding of that character. Otherwise, returns
1456          * pseudo-entity source (eg &foo;)
1457          *
1458          * @param $name String
1459          * @return String
1460          */
1461         static function decodeEntity( $name ) {
1462                 if ( isset( self::$htmlEntityAliases[$name] ) ) {
1463                         $name = self::$htmlEntityAliases[$name];
1464                 }
1465                 if( isset( self::$htmlEntities[$name] ) ) {
1466                         return codepointToUtf8( self::$htmlEntities[$name] );
1467                 } else {
1468                         return "&$name;";
1469                 }
1470         }
1471
1472         /**
1473          * Fetch the whitelist of acceptable attributes for a given element name.
1474          *
1475          * @param $element String
1476          * @return Array
1477          */
1478         static function attributeWhitelist( $element ) {
1479                 static $list;
1480                 if( !isset( $list ) ) {
1481                         $list = Sanitizer::setupAttributeWhitelist();
1482                 }
1483                 return isset( $list[$element] )
1484                         ? $list[$element]
1485                         : array();
1486         }
1487
1488         /**
1489          * Foreach array key (an allowed HTML element), return an array
1490          * of allowed attributes
1491          * @return Array
1492          */
1493         static function setupAttributeWhitelist() {
1494                 global $wgAllowRdfaAttributes, $wgHtml5, $wgAllowMicrodataAttributes;
1495
1496                 $common = array( 'id', 'class', 'lang', 'dir', 'title', 'style' );
1497
1498                 if ( $wgAllowRdfaAttributes ) {
1499                         #RDFa attributes as specified in section 9 of http://www.w3.org/TR/2008/REC-rdfa-syntax-20081014
1500                         $common = array_merge( $common, array(
1501                             'about', 'property', 'resource', 'datatype', 'typeof',
1502                         ) );
1503                 }
1504
1505                 if ( $wgHtml5 && $wgAllowMicrodataAttributes ) {
1506                         # add HTML5 microdata tages as pecified by http://www.whatwg.org/specs/web-apps/current-work/multipage/microdata.html#the-microdata-model
1507                         $common = array_merge( $common, array(
1508                             'itemid', 'itemprop', 'itemref', 'itemscope', 'itemtype'
1509                         ) );
1510                 }
1511
1512                 $block = array_merge( $common, array( 'align' ) );
1513                 $tablealign = array( 'align', 'char', 'charoff', 'valign' );
1514                 $tablecell = array( 'abbr',
1515                                     'axis',
1516                                     'headers',
1517                                     'scope',
1518                                     'rowspan',
1519                                     'colspan',
1520                                     'nowrap', # deprecated
1521                                     'width',  # deprecated
1522                                     'height', # deprecated
1523                                     'bgcolor' # deprecated
1524                                     );
1525
1526                 # Numbers refer to sections in HTML 4.01 standard describing the element.
1527                 # See: http://www.w3.org/TR/html4/
1528                 $whitelist = array(
1529                         # 7.5.4
1530                         'div'        => $block,
1531                         'center'     => $common, # deprecated
1532                         'span'       => $block, # ??
1533
1534                         # 7.5.5
1535                         'h1'         => $block,
1536                         'h2'         => $block,
1537                         'h3'         => $block,
1538                         'h4'         => $block,
1539                         'h5'         => $block,
1540                         'h6'         => $block,
1541
1542                         # 7.5.6
1543                         # address
1544
1545                         # 8.2.4
1546                         # bdo
1547
1548                         # 9.2.1
1549                         'em'         => $common,
1550                         'strong'     => $common,
1551                         'cite'       => $common,
1552                         'dfn'        => $common,
1553                         'code'       => $common,
1554                         'samp'       => $common,
1555                         'kbd'        => $common,
1556                         'var'        => $common,
1557                         'abbr'       => $common,
1558                         # acronym
1559
1560                         # 9.2.2
1561                         'blockquote' => array_merge( $common, array( 'cite' ) ),
1562                         # q
1563
1564                         # 9.2.3
1565                         'sub'        => $common,
1566                         'sup'        => $common,
1567
1568                         # 9.3.1
1569                         'p'          => $block,
1570
1571                         # 9.3.2
1572                         'br'         => array( 'id', 'class', 'title', 'style', 'clear' ),
1573
1574                         # 9.3.4
1575                         'pre'        => array_merge( $common, array( 'width' ) ),
1576
1577                         # 9.4
1578                         'ins'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1579                         'del'        => array_merge( $common, array( 'cite', 'datetime' ) ),
1580
1581                         # 10.2
1582                         'ul'         => array_merge( $common, array( 'type' ) ),
1583                         'ol'         => array_merge( $common, array( 'type', 'start' ) ),
1584                         'li'         => array_merge( $common, array( 'type', 'value' ) ),
1585
1586                         # 10.3
1587                         'dl'         => $common,
1588                         'dd'         => $common,
1589                         'dt'         => $common,
1590
1591                         # 11.2.1
1592                         'table'      => array_merge( $common,
1593                                                                 array( 'summary', 'width', 'border', 'frame',
1594                                                                                 'rules', 'cellspacing', 'cellpadding',
1595                                                                                 'align', 'bgcolor',
1596                                                                 ) ),
1597
1598                         # 11.2.2
1599                         'caption'    => array_merge( $common, array( 'align' ) ),
1600
1601                         # 11.2.3
1602                         'thead'      => array_merge( $common, $tablealign ),
1603                         'tfoot'      => array_merge( $common, $tablealign ),
1604                         'tbody'      => array_merge( $common, $tablealign ),
1605
1606                         # 11.2.4
1607                         'colgroup'   => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1608                         'col'        => array_merge( $common, array( 'span', 'width' ), $tablealign ),
1609
1610                         # 11.2.5
1611                         'tr'         => array_merge( $common, array( 'bgcolor' ), $tablealign ),
1612
1613                         # 11.2.6
1614                         'td'         => array_merge( $common, $tablecell, $tablealign ),
1615                         'th'         => array_merge( $common, $tablecell, $tablealign ),
1616
1617                         # 12.2 # NOTE: <a> is not allowed directly, but the attrib whitelist is used from the Parser object
1618                         'a'          => array_merge( $common, array( 'href', 'rel', 'rev' ) ), # rel/rev esp. for RDFa
1619
1620                         # 13.2
1621                         # Not usually allowed, but may be used for extension-style hooks
1622                         # such as <math> when it is rasterized, or if $wgAllowImageTag is
1623                         # true
1624                         'img'        => array_merge( $common, array( 'alt', 'src', 'width', 'height' ) ),
1625
1626                         # 15.2.1
1627                         'tt'         => $common,
1628                         'b'          => $common,
1629                         'i'          => $common,
1630                         'big'        => $common,
1631                         'small'      => $common,
1632                         'strike'     => $common,
1633                         's'          => $common,
1634                         'u'          => $common,
1635
1636                         # 15.2.2
1637                         'font'       => array_merge( $common, array( 'size', 'color', 'face' ) ),
1638                         # basefont
1639
1640                         # 15.3
1641                         'hr'         => array_merge( $common, array( 'noshade', 'size', 'width' ) ),
1642
1643                         # XHTML Ruby annotation text module, simple ruby only.
1644                         # http://www.w3c.org/TR/ruby/
1645                         'ruby'       => $common,
1646                         # rbc
1647                         # rtc
1648                         'rb'         => $common,
1649                         'rt'         => $common, #array_merge( $common, array( 'rbspan' ) ),
1650                         'rp'         => $common,
1651
1652                         # MathML root element, where used for extensions
1653                         # 'title' may not be 100% valid here; it's XHTML
1654                         # http://www.w3.org/TR/REC-MathML/
1655                         'math'       => array( 'class', 'style', 'id', 'title' ),
1656                 );
1657
1658                 if ( $wgHtml5 ) {
1659                         # HTML5 elements, defined by:
1660                         # http://www.whatwg.org/specs/web-apps/current-work/multipage/
1661                         $whitelist += array(
1662                                 'data' => array_merge( $common, array( 'value' ) ),
1663                                 'time' => array_merge( $common, array( 'datetime' ) ),
1664
1665                                 // meta and link are only present when Microdata is allowed anyways
1666                                 // so we don't bother adding another condition here
1667                                 // meta and link are only valid for use as Microdata so we do not
1668                                 // allow the common attributes here.
1669                                 'meta' => array( 'itemprop', 'content' ),
1670                                 'link' => array( 'itemprop', 'href' ),
1671                         );
1672                 }
1673
1674                 return $whitelist;
1675         }
1676
1677         /**
1678          * Take a fragment of (potentially invalid) HTML and return
1679          * a version with any tags removed, encoded as plain text.
1680          *
1681          * Warning: this return value must be further escaped for literal
1682          * inclusion in HTML output as of 1.10!
1683          *
1684          * @param $text String: HTML fragment
1685          * @return String
1686          */
1687         static function stripAllTags( $text ) {
1688                 # Actual <tags>
1689                 $text = StringUtils::delimiterReplace( '<', '>', '', $text );
1690
1691                 # Normalize &entities and whitespace
1692                 $text = self::decodeCharReferences( $text );
1693                 $text = self::normalizeWhitespace( $text );
1694
1695                 return $text;
1696         }
1697
1698         /**
1699          * Hack up a private DOCTYPE with HTML's standard entity declarations.
1700          * PHP 4 seemed to know these if you gave it an HTML doctype, but
1701          * PHP 5.1 doesn't.
1702          *
1703          * Use for passing XHTML fragments to PHP's XML parsing functions
1704          *
1705          * @return String
1706          */
1707         static function hackDocType() {
1708                 $out = "<!DOCTYPE html [\n";
1709                 foreach( self::$htmlEntities as $entity => $codepoint ) {
1710                         $out .= "<!ENTITY $entity \"&#$codepoint;\">";
1711                 }
1712                 $out .= "]>\n";
1713                 return $out;
1714         }
1715
1716         /**
1717          * @param $url string
1718          * @return mixed|string
1719          */
1720         static function cleanUrl( $url ) {
1721                 # Normalize any HTML entities in input. They will be
1722                 # re-escaped by makeExternalLink().
1723                 $url = Sanitizer::decodeCharReferences( $url );
1724
1725                 # Escape any control characters introduced by the above step
1726                 $url = preg_replace_callback( '/[\][<>"\\x00-\\x20\\x7F\|]/',
1727                         array( __CLASS__, 'cleanUrlCallback' ), $url );
1728
1729                 # Validate hostname portion
1730                 $matches = array();
1731                 if( preg_match( '!^([^:]+:)(//[^/]+)?(.*)$!iD', $url, $matches ) ) {
1732                         list( /* $whole */, $protocol, $host, $rest ) = $matches;
1733
1734                         // Characters that will be ignored in IDNs.
1735                         // http://tools.ietf.org/html/3454#section-3.1
1736                         // Strip them before further processing so blacklists and such work.
1737                         $strip = "/
1738                                 \\s|          # general whitespace
1739                                 \xc2\xad|     # 00ad SOFT HYPHEN
1740                                 \xe1\xa0\x86| # 1806 MONGOLIAN TODO SOFT HYPHEN
1741                                 \xe2\x80\x8b| # 200b ZERO WIDTH SPACE
1742                                 \xe2\x81\xa0| # 2060 WORD JOINER
1743                                 \xef\xbb\xbf| # feff ZERO WIDTH NO-BREAK SPACE
1744                                 \xcd\x8f|     # 034f COMBINING GRAPHEME JOINER
1745                                 \xe1\xa0\x8b| # 180b MONGOLIAN FREE VARIATION SELECTOR ONE
1746                                 \xe1\xa0\x8c| # 180c MONGOLIAN FREE VARIATION SELECTOR TWO
1747                                 \xe1\xa0\x8d| # 180d MONGOLIAN FREE VARIATION SELECTOR THREE
1748                                 \xe2\x80\x8c| # 200c ZERO WIDTH NON-JOINER
1749                                 \xe2\x80\x8d| # 200d ZERO WIDTH JOINER
1750                                 [\xef\xb8\x80-\xef\xb8\x8f] # fe00-fe0f VARIATION SELECTOR-1-16
1751                                 /xuD";
1752
1753                         $host = preg_replace( $strip, '', $host );
1754
1755                         // @todo FIXME: Validate hostnames here
1756
1757                         return $protocol . $host . $rest;
1758                 } else {
1759                         return $url;
1760                 }
1761         }
1762
1763         /**
1764          * @param $matches array
1765          * @return string
1766          */
1767         static function cleanUrlCallback( $matches ) {
1768                 return urlencode( $matches[0] );
1769         }
1770
1771         /**
1772          * Does a string look like an e-mail address?
1773          *
1774          * This validates an email address using an HTML5 specification found at:
1775          * http://www.whatwg.org/specs/web-apps/current-work/multipage/states-of-the-type-attribute.html#valid-e-mail-address
1776          * Which as of 2011-01-24 says:
1777          *
1778          *   A valid e-mail address is a string that matches the ABNF production
1779          *   1*( atext / "." ) "@" ldh-str *( "." ldh-str ) where atext is defined
1780          *   in RFC 5322 section 3.2.3, and ldh-str is defined in RFC 1034 section
1781          *   3.5.
1782          *
1783          * This function is an implementation of the specification as requested in
1784          * bug 22449.
1785          *
1786          * Client-side forms will use the same standard validation rules via JS or
1787          * HTML 5 validation; additional restrictions can be enforced server-side
1788          * by extensions via the 'isValidEmailAddr' hook.
1789          *
1790          * Note that this validation doesn't 100% match RFC 2822, but is believed
1791          * to be liberal enough for wide use. Some invalid addresses will still
1792          * pass validation here.
1793          *
1794          * @since 1.18
1795          *
1796          * @param $addr String E-mail address
1797          * @return Bool
1798          */
1799         public static function validateEmail( $addr ) {
1800                 $result = null;
1801                 if( !wfRunHooks( 'isValidEmailAddr', array( $addr, &$result ) ) ) {
1802                         return $result;
1803                 }
1804
1805                 // Please note strings below are enclosed in brackets [], this make the
1806                 // hyphen "-" a range indicator. Hence it is double backslashed below.
1807                 // See bug 26948
1808                 $rfc5322_atext   = "a-z0-9!#$%&'*+\\-\/=?^_`{|}~" ;
1809                 $rfc1034_ldh_str = "a-z0-9\\-" ;
1810
1811                 $HTML5_email_regexp = "/
1812                 ^                      # start of string
1813                 [$rfc5322_atext\\.]+    # user part which is liberal :p
1814                 @                      # 'apostrophe'
1815                 [$rfc1034_ldh_str]+       # First domain part
1816                 (\\.[$rfc1034_ldh_str]+)*  # Following part prefixed with a dot
1817                 $                      # End of string
1818                 /ix" ; // case Insensitive, eXtended
1819
1820                 return (bool) preg_match( $HTML5_email_regexp, $addr );
1821         }
1822 }