HTMLPurifier 4.4.0
|
00001 <?php 00002 00007 class HTMLPurifier_Encoder 00008 { 00009 00013 private function __construct() { 00014 trigger_error('Cannot instantiate encoder, call methods statically', E_USER_ERROR); 00015 } 00016 00020 public static function muteErrorHandler() {} 00021 00025 public static function unsafeIconv($in, $out, $text) { 00026 set_error_handler(array('HTMLPurifier_Encoder', 'muteErrorHandler')); 00027 $r = iconv($in, $out, $text); 00028 restore_error_handler(); 00029 return $r; 00030 } 00031 00035 public static function iconv($in, $out, $text, $max_chunk_size = 8000) { 00036 $code = self::testIconvTruncateBug(); 00037 if ($code == self::ICONV_OK) { 00038 return self::unsafeIconv($in, $out, $text); 00039 } elseif ($code == self::ICONV_TRUNCATES) { 00040 // we can only work around this if the input character set 00041 // is utf-8 00042 if ($in == 'utf-8') { 00043 if ($max_chunk_size < 4) { 00044 trigger_error('max_chunk_size is too small', E_USER_WARNING); 00045 return false; 00046 } 00047 // split into 8000 byte chunks, but be careful to handle 00048 // multibyte boundaries properly 00049 if (($c = strlen($text)) <= $max_chunk_size) { 00050 return self::unsafeIconv($in, $out, $text); 00051 } 00052 $r = ''; 00053 $i = 0; 00054 while (true) { 00055 if ($i + $max_chunk_size >= $c) { 00056 $r .= self::unsafeIconv($in, $out, substr($text, $i)); 00057 break; 00058 } 00059 // wibble the boundary 00060 if (0x80 != (0xC0 & ord($text[$i + $max_chunk_size]))) { 00061 $chunk_size = $max_chunk_size; 00062 } elseif (0x80 != (0xC0 & ord($text[$i + $max_chunk_size - 1]))) { 00063 $chunk_size = $max_chunk_size - 1; 00064 } elseif (0x80 != (0xC0 & ord($text[$i + $max_chunk_size - 2]))) { 00065 $chunk_size = $max_chunk_size - 2; 00066 } elseif (0x80 != (0xC0 & ord($text[$i + $max_chunk_size - 3]))) { 00067 $chunk_size = $max_chunk_size - 3; 00068 } else { 00069 return false; // rather confusing UTF-8... 00070 } 00071 $chunk = substr($text, $i, $chunk_size); // substr doesn't mind overlong lengths 00072 $r .= self::unsafeIconv($in, $out, $chunk); 00073 $i += $chunk_size; 00074 } 00075 return $r; 00076 } else { 00077 return false; 00078 } 00079 } else { 00080 return false; 00081 } 00082 } 00083 00109 public static function cleanUTF8($str, $force_php = false) { 00110 00111 // UTF-8 validity is checked since PHP 4.3.5 00112 // This is an optimization: if the string is already valid UTF-8, no 00113 // need to do PHP stuff. 99% of the time, this will be the case. 00114 // The regexp matches the XML char production, as well as well as excluding 00115 // non-SGML codepoints U+007F to U+009F 00116 if (preg_match('/^[\x{9}\x{A}\x{D}\x{20}-\x{7E}\x{A0}-\x{D7FF}\x{E000}-\x{FFFD}\x{10000}-\x{10FFFF}]*$/Du', $str)) { 00117 return $str; 00118 } 00119 00120 $mState = 0; // cached expected number of octets after the current octet 00121 // until the beginning of the next UTF8 character sequence 00122 $mUcs4 = 0; // cached Unicode character 00123 $mBytes = 1; // cached expected number of octets in the current sequence 00124 00125 // original code involved an $out that was an array of Unicode 00126 // codepoints. Instead of having to convert back into UTF-8, we've 00127 // decided to directly append valid UTF-8 characters onto a string 00128 // $out once they're done. $char accumulates raw bytes, while $mUcs4 00129 // turns into the Unicode code point, so there's some redundancy. 00130 00131 $out = ''; 00132 $char = ''; 00133 00134 $len = strlen($str); 00135 for($i = 0; $i < $len; $i++) { 00136 $in = ord($str{$i}); 00137 $char .= $str[$i]; // append byte to char 00138 if (0 == $mState) { 00139 // When mState is zero we expect either a US-ASCII character 00140 // or a multi-octet sequence. 00141 if (0 == (0x80 & ($in))) { 00142 // US-ASCII, pass straight through. 00143 if (($in <= 31 || $in == 127) && 00144 !($in == 9 || $in == 13 || $in == 10) // save \r\t\n 00145 ) { 00146 // control characters, remove 00147 } else { 00148 $out .= $char; 00149 } 00150 // reset 00151 $char = ''; 00152 $mBytes = 1; 00153 } elseif (0xC0 == (0xE0 & ($in))) { 00154 // First octet of 2 octet sequence 00155 $mUcs4 = ($in); 00156 $mUcs4 = ($mUcs4 & 0x1F) << 6; 00157 $mState = 1; 00158 $mBytes = 2; 00159 } elseif (0xE0 == (0xF0 & ($in))) { 00160 // First octet of 3 octet sequence 00161 $mUcs4 = ($in); 00162 $mUcs4 = ($mUcs4 & 0x0F) << 12; 00163 $mState = 2; 00164 $mBytes = 3; 00165 } elseif (0xF0 == (0xF8 & ($in))) { 00166 // First octet of 4 octet sequence 00167 $mUcs4 = ($in); 00168 $mUcs4 = ($mUcs4 & 0x07) << 18; 00169 $mState = 3; 00170 $mBytes = 4; 00171 } elseif (0xF8 == (0xFC & ($in))) { 00172 // First octet of 5 octet sequence. 00173 // 00174 // This is illegal because the encoded codepoint must be 00175 // either: 00176 // (a) not the shortest form or 00177 // (b) outside the Unicode range of 0-0x10FFFF. 00178 // Rather than trying to resynchronize, we will carry on 00179 // until the end of the sequence and let the later error 00180 // handling code catch it. 00181 $mUcs4 = ($in); 00182 $mUcs4 = ($mUcs4 & 0x03) << 24; 00183 $mState = 4; 00184 $mBytes = 5; 00185 } elseif (0xFC == (0xFE & ($in))) { 00186 // First octet of 6 octet sequence, see comments for 5 00187 // octet sequence. 00188 $mUcs4 = ($in); 00189 $mUcs4 = ($mUcs4 & 1) << 30; 00190 $mState = 5; 00191 $mBytes = 6; 00192 } else { 00193 // Current octet is neither in the US-ASCII range nor a 00194 // legal first octet of a multi-octet sequence. 00195 $mState = 0; 00196 $mUcs4 = 0; 00197 $mBytes = 1; 00198 $char = ''; 00199 } 00200 } else { 00201 // When mState is non-zero, we expect a continuation of the 00202 // multi-octet sequence 00203 if (0x80 == (0xC0 & ($in))) { 00204 // Legal continuation. 00205 $shift = ($mState - 1) * 6; 00206 $tmp = $in; 00207 $tmp = ($tmp & 0x0000003F) << $shift; 00208 $mUcs4 |= $tmp; 00209 00210 if (0 == --$mState) { 00211 // End of the multi-octet sequence. mUcs4 now contains 00212 // the final Unicode codepoint to be output 00213 00214 // Check for illegal sequences and codepoints. 00215 00216 // From Unicode 3.1, non-shortest form is illegal 00217 if (((2 == $mBytes) && ($mUcs4 < 0x0080)) || 00218 ((3 == $mBytes) && ($mUcs4 < 0x0800)) || 00219 ((4 == $mBytes) && ($mUcs4 < 0x10000)) || 00220 (4 < $mBytes) || 00221 // From Unicode 3.2, surrogate characters = illegal 00222 (($mUcs4 & 0xFFFFF800) == 0xD800) || 00223 // Codepoints outside the Unicode range are illegal 00224 ($mUcs4 > 0x10FFFF) 00225 ) { 00226 00227 } elseif (0xFEFF != $mUcs4 && // omit BOM 00228 // check for valid Char unicode codepoints 00229 ( 00230 0x9 == $mUcs4 || 00231 0xA == $mUcs4 || 00232 0xD == $mUcs4 || 00233 (0x20 <= $mUcs4 && 0x7E >= $mUcs4) || 00234 // 7F-9F is not strictly prohibited by XML, 00235 // but it is non-SGML, and thus we don't allow it 00236 (0xA0 <= $mUcs4 && 0xD7FF >= $mUcs4) || 00237 (0x10000 <= $mUcs4 && 0x10FFFF >= $mUcs4) 00238 ) 00239 ) { 00240 $out .= $char; 00241 } 00242 // initialize UTF8 cache (reset) 00243 $mState = 0; 00244 $mUcs4 = 0; 00245 $mBytes = 1; 00246 $char = ''; 00247 } 00248 } else { 00249 // ((0xC0 & (*in) != 0x80) && (mState != 0)) 00250 // Incomplete multi-octet sequence. 00251 // used to result in complete fail, but we'll reset 00252 $mState = 0; 00253 $mUcs4 = 0; 00254 $mBytes = 1; 00255 $char =''; 00256 } 00257 } 00258 } 00259 return $out; 00260 } 00261 00275 // +----------+----------+----------+----------+ 00276 // | 33222222 | 22221111 | 111111 | | 00277 // | 10987654 | 32109876 | 54321098 | 76543210 | bit 00278 // +----------+----------+----------+----------+ 00279 // | | | | 0xxxxxxx | 1 byte 0x00000000..0x0000007F 00280 // | | | 110yyyyy | 10xxxxxx | 2 byte 0x00000080..0x000007FF 00281 // | | 1110zzzz | 10yyyyyy | 10xxxxxx | 3 byte 0x00000800..0x0000FFFF 00282 // | 11110www | 10wwzzzz | 10yyyyyy | 10xxxxxx | 4 byte 0x00010000..0x0010FFFF 00283 // +----------+----------+----------+----------+ 00284 // | 00000000 | 00011111 | 11111111 | 11111111 | Theoretical upper limit of legal scalars: 2097151 (0x001FFFFF) 00285 // | 00000000 | 00010000 | 11111111 | 11111111 | Defined upper limit of legal scalar codes 00286 // +----------+----------+----------+----------+ 00287 00288 public static function unichr($code) { 00289 if($code > 1114111 or $code < 0 or 00290 ($code >= 55296 and $code <= 57343) ) { 00291 // bits are set outside the "valid" range as defined 00292 // by UNICODE 4.1.0 00293 return ''; 00294 } 00295 00296 $x = $y = $z = $w = 0; 00297 if ($code < 128) { 00298 // regular ASCII character 00299 $x = $code; 00300 } else { 00301 // set up bits for UTF-8 00302 $x = ($code & 63) | 128; 00303 if ($code < 2048) { 00304 $y = (($code & 2047) >> 6) | 192; 00305 } else { 00306 $y = (($code & 4032) >> 6) | 128; 00307 if($code < 65536) { 00308 $z = (($code >> 12) & 15) | 224; 00309 } else { 00310 $z = (($code >> 12) & 63) | 128; 00311 $w = (($code >> 18) & 7) | 240; 00312 } 00313 } 00314 } 00315 // set up the actual character 00316 $ret = ''; 00317 if($w) $ret .= chr($w); 00318 if($z) $ret .= chr($z); 00319 if($y) $ret .= chr($y); 00320 $ret .= chr($x); 00321 00322 return $ret; 00323 } 00324 00325 public static function iconvAvailable() { 00326 static $iconv = null; 00327 if ($iconv === null) { 00328 $iconv = function_exists('iconv') && self::testIconvTruncateBug() != self::ICONV_UNUSABLE; 00329 } 00330 return $iconv; 00331 } 00332 00336 public static function convertToUTF8($str, $config, $context) { 00337 $encoding = $config->get('Core.Encoding'); 00338 if ($encoding === 'utf-8') return $str; 00339 static $iconv = null; 00340 if ($iconv === null) $iconv = self::iconvAvailable(); 00341 if ($iconv && !$config->get('Test.ForceNoIconv')) { 00342 // unaffected by bugs, since UTF-8 support all characters 00343 $str = self::unsafeIconv($encoding, 'utf-8//IGNORE', $str); 00344 if ($str === false) { 00345 // $encoding is not a valid encoding 00346 trigger_error('Invalid encoding ' . $encoding, E_USER_ERROR); 00347 return ''; 00348 } 00349 // If the string is bjorked by Shift_JIS or a similar encoding 00350 // that doesn't support all of ASCII, convert the naughty 00351 // characters to their true byte-wise ASCII/UTF-8 equivalents. 00352 $str = strtr($str, self::testEncodingSupportsASCII($encoding)); 00353 return $str; 00354 } elseif ($encoding === 'iso-8859-1') { 00355 $str = utf8_encode($str); 00356 return $str; 00357 } 00358 trigger_error('Encoding not supported, please install iconv', E_USER_ERROR); 00359 } 00360 00366 public static function convertFromUTF8($str, $config, $context) { 00367 $encoding = $config->get('Core.Encoding'); 00368 if ($escape = $config->get('Core.EscapeNonASCIICharacters')) { 00369 $str = self::convertToASCIIDumbLossless($str); 00370 } 00371 if ($encoding === 'utf-8') return $str; 00372 static $iconv = null; 00373 if ($iconv === null) $iconv = self::iconvAvailable(); 00374 if ($iconv && !$config->get('Test.ForceNoIconv')) { 00375 // Undo our previous fix in convertToUTF8, otherwise iconv will barf 00376 $ascii_fix = self::testEncodingSupportsASCII($encoding); 00377 if (!$escape && !empty($ascii_fix)) { 00378 $clear_fix = array(); 00379 foreach ($ascii_fix as $utf8 => $native) $clear_fix[$utf8] = ''; 00380 $str = strtr($str, $clear_fix); 00381 } 00382 $str = strtr($str, array_flip($ascii_fix)); 00383 // Normal stuff 00384 $str = self::iconv('utf-8', $encoding . '//IGNORE', $str); 00385 return $str; 00386 } elseif ($encoding === 'iso-8859-1') { 00387 $str = utf8_decode($str); 00388 return $str; 00389 } 00390 trigger_error('Encoding not supported', E_USER_ERROR); 00391 // You might be tempted to assume that the ASCII representation 00392 // might be OK, however, this is *not* universally true over all 00393 // encodings. So we take the conservative route here, rather 00394 // than forcibly turn on %Core.EscapeNonASCIICharacters 00395 } 00396 00413 public static function convertToASCIIDumbLossless($str) { 00414 $bytesleft = 0; 00415 $result = ''; 00416 $working = 0; 00417 $len = strlen($str); 00418 for( $i = 0; $i < $len; $i++ ) { 00419 $bytevalue = ord( $str[$i] ); 00420 if( $bytevalue <= 0x7F ) { //0xxx xxxx 00421 $result .= chr( $bytevalue ); 00422 $bytesleft = 0; 00423 } elseif( $bytevalue <= 0xBF ) { //10xx xxxx 00424 $working = $working << 6; 00425 $working += ($bytevalue & 0x3F); 00426 $bytesleft--; 00427 if( $bytesleft <= 0 ) { 00428 $result .= "&#" . $working . ";"; 00429 } 00430 } elseif( $bytevalue <= 0xDF ) { //110x xxxx 00431 $working = $bytevalue & 0x1F; 00432 $bytesleft = 1; 00433 } elseif( $bytevalue <= 0xEF ) { //1110 xxxx 00434 $working = $bytevalue & 0x0F; 00435 $bytesleft = 2; 00436 } else { //1111 0xxx 00437 $working = $bytevalue & 0x07; 00438 $bytesleft = 3; 00439 } 00440 } 00441 return $result; 00442 } 00443 00445 const ICONV_OK = 0; 00446 00449 const ICONV_TRUNCATES = 1; 00450 00453 const ICONV_UNUSABLE = 2; 00454 00469 public static function testIconvTruncateBug() { 00470 static $code = null; 00471 if ($code === null) { 00472 // better not use iconv, otherwise infinite loop! 00473 $r = self::unsafeIconv('utf-8', 'ascii//IGNORE', "\xCE\xB1" . str_repeat('a', 9000)); 00474 if ($r === false) { 00475 $code = self::ICONV_UNUSABLE; 00476 } elseif (($c = strlen($r)) < 9000) { 00477 $code = self::ICONV_TRUNCATES; 00478 } elseif ($c > 9000) { 00479 trigger_error('Your copy of iconv is extremely buggy. Please notify HTML Purifier maintainers: include your iconv version as per phpversion()', E_USER_ERROR); 00480 } else { 00481 $code = self::ICONV_OK; 00482 } 00483 } 00484 return $code; 00485 } 00486 00498 public static function testEncodingSupportsASCII($encoding, $bypass = false) { 00499 // All calls to iconv here are unsafe, proof by case analysis: 00500 // If ICONV_OK, no difference. 00501 // If ICONV_TRUNCATE, all calls involve one character inputs, 00502 // so bug is not triggered. 00503 // If ICONV_UNUSABLE, this call is irrelevant 00504 static $encodings = array(); 00505 if (!$bypass) { 00506 if (isset($encodings[$encoding])) return $encodings[$encoding]; 00507 $lenc = strtolower($encoding); 00508 switch ($lenc) { 00509 case 'shift_jis': 00510 return array("\xC2\xA5" => '\\', "\xE2\x80\xBE" => '~'); 00511 case 'johab': 00512 return array("\xE2\x82\xA9" => '\\'); 00513 } 00514 if (strpos($lenc, 'iso-8859-') === 0) return array(); 00515 } 00516 $ret = array(); 00517 if (self::unsafeIconv('UTF-8', $encoding, 'a') === false) return false; 00518 for ($i = 0x20; $i <= 0x7E; $i++) { // all printable ASCII chars 00519 $c = chr($i); // UTF-8 char 00520 $r = self::unsafeIconv('UTF-8', "$encoding//IGNORE", $c); // initial conversion 00521 if ( 00522 $r === '' || 00523 // This line is needed for iconv implementations that do not 00524 // omit characters that do not exist in the target character set 00525 ($r === $c && self::unsafeIconv($encoding, 'UTF-8//IGNORE', $r) !== $c) 00526 ) { 00527 // Reverse engineer: what's the UTF-8 equiv of this byte 00528 // sequence? This assumes that there's no variable width 00529 // encoding that doesn't support ASCII. 00530 $ret[self::unsafeIconv($encoding, 'UTF-8//IGNORE', $c)] = $c; 00531 } 00532 } 00533 $encodings[$encoding] = $ret; 00534 return $ret; 00535 } 00536 00537 00538 } 00539 00540 // vim: et sw=4 sts=4