aboutsummaryrefslogtreecommitdiffhomepage
path: root/inc/3rdparty/htmlpurifier/HTMLPurifier/Encoder.php
diff options
context:
space:
mode:
Diffstat (limited to 'inc/3rdparty/htmlpurifier/HTMLPurifier/Encoder.php')
-rw-r--r--inc/3rdparty/htmlpurifier/HTMLPurifier/Encoder.php611
1 files changed, 611 insertions, 0 deletions
diff --git a/inc/3rdparty/htmlpurifier/HTMLPurifier/Encoder.php b/inc/3rdparty/htmlpurifier/HTMLPurifier/Encoder.php
new file mode 100644
index 00000000..3aa00ca5
--- /dev/null
+++ b/inc/3rdparty/htmlpurifier/HTMLPurifier/Encoder.php
@@ -0,0 +1,611 @@
1<?php
2
3/**
4 * A UTF-8 specific character encoder that handles cleaning and transforming.
5 * @note All functions in this class should be static.
6 */
7class HTMLPurifier_Encoder
8{
9
10 /**
11 * Constructor throws fatal error if you attempt to instantiate class
12 */
13 private function __construct()
14 {
15 trigger_error('Cannot instantiate encoder, call methods statically', E_USER_ERROR);
16 }
17
18 /**
19 * Error-handler that mutes errors, alternative to shut-up operator.
20 */
21 public static function muteErrorHandler()
22 {
23 }
24
25 /**
26 * iconv wrapper which mutes errors, but doesn't work around bugs.
27 * @param string $in Input encoding
28 * @param string $out Output encoding
29 * @param string $text The text to convert
30 * @return string
31 */
32 public static function unsafeIconv($in, $out, $text)
33 {
34 set_error_handler(array('HTMLPurifier_Encoder', 'muteErrorHandler'));
35 $r = iconv($in, $out, $text);
36 restore_error_handler();
37 return $r;
38 }
39
40 /**
41 * iconv wrapper which mutes errors and works around bugs.
42 * @param string $in Input encoding
43 * @param string $out Output encoding
44 * @param string $text The text to convert
45 * @param int $max_chunk_size
46 * @return string
47 */
48 public static function iconv($in, $out, $text, $max_chunk_size = 8000)
49 {
50 $code = self::testIconvTruncateBug();
51 if ($code == self::ICONV_OK) {
52 return self::unsafeIconv($in, $out, $text);
53 } elseif ($code == self::ICONV_TRUNCATES) {
54 // we can only work around this if the input character set
55 // is utf-8
56 if ($in == 'utf-8') {
57 if ($max_chunk_size < 4) {
58 trigger_error('max_chunk_size is too small', E_USER_WARNING);
59 return false;
60 }
61 // split into 8000 byte chunks, but be careful to handle
62 // multibyte boundaries properly
63 if (($c = strlen($text)) <= $max_chunk_size) {
64 return self::unsafeIconv($in, $out, $text);
65 }
66 $r = '';
67 $i = 0;
68 while (true) {
69 if ($i + $max_chunk_size >= $c) {
70 $r .= self::unsafeIconv($in, $out, substr($text, $i));
71 break;
72 }
73 // wibble the boundary
74 if (0x80 != (0xC0 & ord($text[$i + $max_chunk_size]))) {
75 $chunk_size = $max_chunk_size;
76 } elseif (0x80 != (0xC0 & ord($text[$i + $max_chunk_size - 1]))) {
77 $chunk_size = $max_chunk_size - 1;
78 } elseif (0x80 != (0xC0 & ord($text[$i + $max_chunk_size - 2]))) {
79 $chunk_size = $max_chunk_size - 2;
80 } elseif (0x80 != (0xC0 & ord($text[$i + $max_chunk_size - 3]))) {
81 $chunk_size = $max_chunk_size - 3;
82 } else {
83 return false; // rather confusing UTF-8...
84 }
85 $chunk = substr($text, $i, $chunk_size); // substr doesn't mind overlong lengths
86 $r .= self::unsafeIconv($in, $out, $chunk);
87 $i += $chunk_size;
88 }
89 return $r;
90 } else {
91 return false;
92 }
93 } else {
94 return false;
95 }
96 }
97
98 /**
99 * Cleans a UTF-8 string for well-formedness and SGML validity
100 *
101 * It will parse according to UTF-8 and return a valid UTF8 string, with
102 * non-SGML codepoints excluded.
103 *
104 * @param string $str The string to clean
105 * @param bool $force_php
106 * @return string
107 *
108 * @note Just for reference, the non-SGML code points are 0 to 31 and
109 * 127 to 159, inclusive. However, we allow code points 9, 10
110 * and 13, which are the tab, line feed and carriage return
111 * respectively. 128 and above the code points map to multibyte
112 * UTF-8 representations.
113 *
114 * @note Fallback code adapted from utf8ToUnicode by Henri Sivonen and
115 * hsivonen@iki.fi at <http://iki.fi/hsivonen/php-utf8/> under the
116 * LGPL license. Notes on what changed are inside, but in general,
117 * the original code transformed UTF-8 text into an array of integer
118 * Unicode codepoints. Understandably, transforming that back to
119 * a string would be somewhat expensive, so the function was modded to
120 * directly operate on the string. However, this discourages code
121 * reuse, and the logic enumerated here would be useful for any
122 * function that needs to be able to understand UTF-8 characters.
123 * As of right now, only smart lossless character encoding converters
124 * would need that, and I'm probably not going to implement them.
125 * Once again, PHP 6 should solve all our problems.
126 */
127 public static function cleanUTF8($str, $force_php = false)
128 {
129 // UTF-8 validity is checked since PHP 4.3.5
130 // This is an optimization: if the string is already valid UTF-8, no
131 // need to do PHP stuff. 99% of the time, this will be the case.
132 // The regexp matches the XML char production, as well as well as excluding
133 // non-SGML codepoints U+007F to U+009F
134 if (preg_match(
135 '/^[\x{9}\x{A}\x{D}\x{20}-\x{7E}\x{A0}-\x{D7FF}\x{E000}-\x{FFFD}\x{10000}-\x{10FFFF}]*$/Du',
136 $str
137 )) {
138 return $str;
139 }
140
141 $mState = 0; // cached expected number of octets after the current octet
142 // until the beginning of the next UTF8 character sequence
143 $mUcs4 = 0; // cached Unicode character
144 $mBytes = 1; // cached expected number of octets in the current sequence
145
146 // original code involved an $out that was an array of Unicode
147 // codepoints. Instead of having to convert back into UTF-8, we've
148 // decided to directly append valid UTF-8 characters onto a string
149 // $out once they're done. $char accumulates raw bytes, while $mUcs4
150 // turns into the Unicode code point, so there's some redundancy.
151
152 $out = '';
153 $char = '';
154
155 $len = strlen($str);
156 for ($i = 0; $i < $len; $i++) {
157 $in = ord($str{$i});
158 $char .= $str[$i]; // append byte to char
159 if (0 == $mState) {
160 // When mState is zero we expect either a US-ASCII character
161 // or a multi-octet sequence.
162 if (0 == (0x80 & ($in))) {
163 // US-ASCII, pass straight through.
164 if (($in <= 31 || $in == 127) &&
165 !($in == 9 || $in == 13 || $in == 10) // save \r\t\n
166 ) {
167 // control characters, remove
168 } else {
169 $out .= $char;
170 }
171 // reset
172 $char = '';
173 $mBytes = 1;
174 } elseif (0xC0 == (0xE0 & ($in))) {
175 // First octet of 2 octet sequence
176 $mUcs4 = ($in);
177 $mUcs4 = ($mUcs4 & 0x1F) << 6;
178 $mState = 1;
179 $mBytes = 2;
180 } elseif (0xE0 == (0xF0 & ($in))) {
181 // First octet of 3 octet sequence
182 $mUcs4 = ($in);
183 $mUcs4 = ($mUcs4 & 0x0F) << 12;
184 $mState = 2;
185 $mBytes = 3;
186 } elseif (0xF0 == (0xF8 & ($in))) {
187 // First octet of 4 octet sequence
188 $mUcs4 = ($in);
189 $mUcs4 = ($mUcs4 & 0x07) << 18;
190 $mState = 3;
191 $mBytes = 4;
192 } elseif (0xF8 == (0xFC & ($in))) {
193 // First octet of 5 octet sequence.
194 //
195 // This is illegal because the encoded codepoint must be
196 // either:
197 // (a) not the shortest form or
198 // (b) outside the Unicode range of 0-0x10FFFF.
199 // Rather than trying to resynchronize, we will carry on
200 // until the end of the sequence and let the later error
201 // handling code catch it.
202 $mUcs4 = ($in);
203 $mUcs4 = ($mUcs4 & 0x03) << 24;
204 $mState = 4;
205 $mBytes = 5;
206 } elseif (0xFC == (0xFE & ($in))) {
207 // First octet of 6 octet sequence, see comments for 5
208 // octet sequence.
209 $mUcs4 = ($in);
210 $mUcs4 = ($mUcs4 & 1) << 30;
211 $mState = 5;
212 $mBytes = 6;
213 } else {
214 // Current octet is neither in the US-ASCII range nor a
215 // legal first octet of a multi-octet sequence.
216 $mState = 0;
217 $mUcs4 = 0;
218 $mBytes = 1;
219 $char = '';
220 }
221 } else {
222 // When mState is non-zero, we expect a continuation of the
223 // multi-octet sequence
224 if (0x80 == (0xC0 & ($in))) {
225 // Legal continuation.
226 $shift = ($mState - 1) * 6;
227 $tmp = $in;
228 $tmp = ($tmp & 0x0000003F) << $shift;
229 $mUcs4 |= $tmp;
230
231 if (0 == --$mState) {
232 // End of the multi-octet sequence. mUcs4 now contains
233 // the final Unicode codepoint to be output
234
235 // Check for illegal sequences and codepoints.
236
237 // From Unicode 3.1, non-shortest form is illegal
238 if (((2 == $mBytes) && ($mUcs4 < 0x0080)) ||
239 ((3 == $mBytes) && ($mUcs4 < 0x0800)) ||
240 ((4 == $mBytes) && ($mUcs4 < 0x10000)) ||
241 (4 < $mBytes) ||
242 // From Unicode 3.2, surrogate characters = illegal
243 (($mUcs4 & 0xFFFFF800) == 0xD800) ||
244 // Codepoints outside the Unicode range are illegal
245 ($mUcs4 > 0x10FFFF)
246 ) {
247
248 } elseif (0xFEFF != $mUcs4 && // omit BOM
249 // check for valid Char unicode codepoints
250 (
251 0x9 == $mUcs4 ||
252 0xA == $mUcs4 ||
253 0xD == $mUcs4 ||
254 (0x20 <= $mUcs4 && 0x7E >= $mUcs4) ||
255 // 7F-9F is not strictly prohibited by XML,
256 // but it is non-SGML, and thus we don't allow it
257 (0xA0 <= $mUcs4 && 0xD7FF >= $mUcs4) ||
258 (0x10000 <= $mUcs4 && 0x10FFFF >= $mUcs4)
259 )
260 ) {
261 $out .= $char;
262 }
263 // initialize UTF8 cache (reset)
264 $mState = 0;
265 $mUcs4 = 0;
266 $mBytes = 1;
267 $char = '';
268 }
269 } else {
270 // ((0xC0 & (*in) != 0x80) && (mState != 0))
271 // Incomplete multi-octet sequence.
272 // used to result in complete fail, but we'll reset
273 $mState = 0;
274 $mUcs4 = 0;
275 $mBytes = 1;
276 $char ='';
277 }
278 }
279 }
280 return $out;
281 }
282
283 /**
284 * Translates a Unicode codepoint into its corresponding UTF-8 character.
285 * @note Based on Feyd's function at
286 * <http://forums.devnetwork.net/viewtopic.php?p=191404#191404>,
287 * which is in public domain.
288 * @note While we're going to do code point parsing anyway, a good
289 * optimization would be to refuse to translate code points that
290 * are non-SGML characters. However, this could lead to duplication.
291 * @note This is very similar to the unichr function in
292 * maintenance/generate-entity-file.php (although this is superior,
293 * due to its sanity checks).
294 */
295
296 // +----------+----------+----------+----------+
297 // | 33222222 | 22221111 | 111111 | |
298 // | 10987654 | 32109876 | 54321098 | 76543210 | bit
299 // +----------+----------+----------+----------+
300 // | | | | 0xxxxxxx | 1 byte 0x00000000..0x0000007F
301 // | | | 110yyyyy | 10xxxxxx | 2 byte 0x00000080..0x000007FF
302 // | | 1110zzzz | 10yyyyyy | 10xxxxxx | 3 byte 0x00000800..0x0000FFFF
303 // | 11110www | 10wwzzzz | 10yyyyyy | 10xxxxxx | 4 byte 0x00010000..0x0010FFFF
304 // +----------+----------+----------+----------+
305 // | 00000000 | 00011111 | 11111111 | 11111111 | Theoretical upper limit of legal scalars: 2097151 (0x001FFFFF)
306 // | 00000000 | 00010000 | 11111111 | 11111111 | Defined upper limit of legal scalar codes
307 // +----------+----------+----------+----------+
308
309 public static function unichr($code)
310 {
311 if ($code > 1114111 or $code < 0 or
312 ($code >= 55296 and $code <= 57343) ) {
313 // bits are set outside the "valid" range as defined
314 // by UNICODE 4.1.0
315 return '';
316 }
317
318 $x = $y = $z = $w = 0;
319 if ($code < 128) {
320 // regular ASCII character
321 $x = $code;
322 } else {
323 // set up bits for UTF-8
324 $x = ($code & 63) | 128;
325 if ($code < 2048) {
326 $y = (($code & 2047) >> 6) | 192;
327 } else {
328 $y = (($code & 4032) >> 6) | 128;
329 if ($code < 65536) {
330 $z = (($code >> 12) & 15) | 224;
331 } else {
332 $z = (($code >> 12) & 63) | 128;
333 $w = (($code >> 18) & 7) | 240;
334 }
335 }
336 }
337 // set up the actual character
338 $ret = '';
339 if ($w) {
340 $ret .= chr($w);
341 }
342 if ($z) {
343 $ret .= chr($z);
344 }
345 if ($y) {
346 $ret .= chr($y);
347 }
348 $ret .= chr($x);
349
350 return $ret;
351 }
352
353 /**
354 * @return bool
355 */
356 public static function iconvAvailable()
357 {
358 static $iconv = null;
359 if ($iconv === null) {
360 $iconv = function_exists('iconv') && self::testIconvTruncateBug() != self::ICONV_UNUSABLE;
361 }
362 return $iconv;
363 }
364
365 /**
366 * Convert a string to UTF-8 based on configuration.
367 * @param string $str The string to convert
368 * @param HTMLPurifier_Config $config
369 * @param HTMLPurifier_Context $context
370 * @return string
371 */
372 public static function convertToUTF8($str, $config, $context)
373 {
374 $encoding = $config->get('Core.Encoding');
375 if ($encoding === 'utf-8') {
376 return $str;
377 }
378 static $iconv = null;
379 if ($iconv === null) {
380 $iconv = self::iconvAvailable();
381 }
382 if ($iconv && !$config->get('Test.ForceNoIconv')) {
383 // unaffected by bugs, since UTF-8 support all characters
384 $str = self::unsafeIconv($encoding, 'utf-8//IGNORE', $str);
385 if ($str === false) {
386 // $encoding is not a valid encoding
387 trigger_error('Invalid encoding ' . $encoding, E_USER_ERROR);
388 return '';
389 }
390 // If the string is bjorked by Shift_JIS or a similar encoding
391 // that doesn't support all of ASCII, convert the naughty
392 // characters to their true byte-wise ASCII/UTF-8 equivalents.
393 $str = strtr($str, self::testEncodingSupportsASCII($encoding));
394 return $str;
395 } elseif ($encoding === 'iso-8859-1') {
396 $str = utf8_encode($str);
397 return $str;
398 }
399 $bug = HTMLPurifier_Encoder::testIconvTruncateBug();
400 if ($bug == self::ICONV_OK) {
401 trigger_error('Encoding not supported, please install iconv', E_USER_ERROR);
402 } else {
403 trigger_error(
404 'You have a buggy version of iconv, see https://bugs.php.net/bug.php?id=48147 ' .
405 'and http://sourceware.org/bugzilla/show_bug.cgi?id=13541',
406 E_USER_ERROR
407 );
408 }
409 }
410
411 /**
412 * Converts a string from UTF-8 based on configuration.
413 * @param string $str The string to convert
414 * @param HTMLPurifier_Config $config
415 * @param HTMLPurifier_Context $context
416 * @return string
417 * @note Currently, this is a lossy conversion, with unexpressable
418 * characters being omitted.
419 */
420 public static function convertFromUTF8($str, $config, $context)
421 {
422 $encoding = $config->get('Core.Encoding');
423 if ($escape = $config->get('Core.EscapeNonASCIICharacters')) {
424 $str = self::convertToASCIIDumbLossless($str);
425 }
426 if ($encoding === 'utf-8') {
427 return $str;
428 }
429 static $iconv = null;
430 if ($iconv === null) {
431 $iconv = self::iconvAvailable();
432 }
433 if ($iconv && !$config->get('Test.ForceNoIconv')) {
434 // Undo our previous fix in convertToUTF8, otherwise iconv will barf
435 $ascii_fix = self::testEncodingSupportsASCII($encoding);
436 if (!$escape && !empty($ascii_fix)) {
437 $clear_fix = array();
438 foreach ($ascii_fix as $utf8 => $native) {
439 $clear_fix[$utf8] = '';
440 }
441 $str = strtr($str, $clear_fix);
442 }
443 $str = strtr($str, array_flip($ascii_fix));
444 // Normal stuff
445 $str = self::iconv('utf-8', $encoding . '//IGNORE', $str);
446 return $str;
447 } elseif ($encoding === 'iso-8859-1') {
448 $str = utf8_decode($str);
449 return $str;
450 }
451 trigger_error('Encoding not supported', E_USER_ERROR);
452 // You might be tempted to assume that the ASCII representation
453 // might be OK, however, this is *not* universally true over all
454 // encodings. So we take the conservative route here, rather
455 // than forcibly turn on %Core.EscapeNonASCIICharacters
456 }
457
458 /**
459 * Lossless (character-wise) conversion of HTML to ASCII
460 * @param string $str UTF-8 string to be converted to ASCII
461 * @return string ASCII encoded string with non-ASCII character entity-ized
462 * @warning Adapted from MediaWiki, claiming fair use: this is a common
463 * algorithm. If you disagree with this license fudgery,
464 * implement it yourself.
465 * @note Uses decimal numeric entities since they are best supported.
466 * @note This is a DUMB function: it has no concept of keeping
467 * character entities that the projected character encoding
468 * can allow. We could possibly implement a smart version
469 * but that would require it to also know which Unicode
470 * codepoints the charset supported (not an easy task).
471 * @note Sort of with cleanUTF8() but it assumes that $str is
472 * well-formed UTF-8
473 */
474 public static function convertToASCIIDumbLossless($str)
475 {
476 $bytesleft = 0;
477 $result = '';
478 $working = 0;
479 $len = strlen($str);
480 for ($i = 0; $i < $len; $i++) {
481 $bytevalue = ord($str[$i]);
482 if ($bytevalue <= 0x7F) { //0xxx xxxx
483 $result .= chr($bytevalue);
484 $bytesleft = 0;
485 } elseif ($bytevalue <= 0xBF) { //10xx xxxx
486 $working = $working << 6;
487 $working += ($bytevalue & 0x3F);
488 $bytesleft--;
489 if ($bytesleft <= 0) {
490 $result .= "&#" . $working . ";";
491 }
492 } elseif ($bytevalue <= 0xDF) { //110x xxxx
493 $working = $bytevalue & 0x1F;
494 $bytesleft = 1;
495 } elseif ($bytevalue <= 0xEF) { //1110 xxxx
496 $working = $bytevalue & 0x0F;
497 $bytesleft = 2;
498 } else { //1111 0xxx
499 $working = $bytevalue & 0x07;
500 $bytesleft = 3;
501 }
502 }
503 return $result;
504 }
505
506 /** No bugs detected in iconv. */
507 const ICONV_OK = 0;
508
509 /** Iconv truncates output if converting from UTF-8 to another
510 * character set with //IGNORE, and a non-encodable character is found */
511 const ICONV_TRUNCATES = 1;
512
513 /** Iconv does not support //IGNORE, making it unusable for
514 * transcoding purposes */
515 const ICONV_UNUSABLE = 2;
516
517 /**
518 * glibc iconv has a known bug where it doesn't handle the magic
519 * //IGNORE stanza correctly. In particular, rather than ignore
520 * characters, it will return an EILSEQ after consuming some number
521 * of characters, and expect you to restart iconv as if it were
522 * an E2BIG. Old versions of PHP did not respect the errno, and
523 * returned the fragment, so as a result you would see iconv
524 * mysteriously truncating output. We can work around this by
525 * manually chopping our input into segments of about 8000
526 * characters, as long as PHP ignores the error code. If PHP starts
527 * paying attention to the error code, iconv becomes unusable.
528 *
529 * @return int Error code indicating severity of bug.
530 */
531 public static function testIconvTruncateBug()
532 {
533 static $code = null;
534 if ($code === null) {
535 // better not use iconv, otherwise infinite loop!
536 $r = self::unsafeIconv('utf-8', 'ascii//IGNORE', "\xCE\xB1" . str_repeat('a', 9000));
537 if ($r === false) {
538 $code = self::ICONV_UNUSABLE;
539 } elseif (($c = strlen($r)) < 9000) {
540 $code = self::ICONV_TRUNCATES;
541 } elseif ($c > 9000) {
542 trigger_error(
543 'Your copy of iconv is extremely buggy. Please notify HTML Purifier maintainers: ' .
544 'include your iconv version as per phpversion()',
545 E_USER_ERROR
546 );
547 } else {
548 $code = self::ICONV_OK;
549 }
550 }
551 return $code;
552 }
553
554 /**
555 * This expensive function tests whether or not a given character
556 * encoding supports ASCII. 7/8-bit encodings like Shift_JIS will
557 * fail this test, and require special processing. Variable width
558 * encodings shouldn't ever fail.
559 *
560 * @param string $encoding Encoding name to test, as per iconv format
561 * @param bool $bypass Whether or not to bypass the precompiled arrays.
562 * @return Array of UTF-8 characters to their corresponding ASCII,
563 * which can be used to "undo" any overzealous iconv action.
564 */
565 public static function testEncodingSupportsASCII($encoding, $bypass = false)
566 {
567 // All calls to iconv here are unsafe, proof by case analysis:
568 // If ICONV_OK, no difference.
569 // If ICONV_TRUNCATE, all calls involve one character inputs,
570 // so bug is not triggered.
571 // If ICONV_UNUSABLE, this call is irrelevant
572 static $encodings = array();
573 if (!$bypass) {
574 if (isset($encodings[$encoding])) {
575 return $encodings[$encoding];
576 }
577 $lenc = strtolower($encoding);
578 switch ($lenc) {
579 case 'shift_jis':
580 return array("\xC2\xA5" => '\\', "\xE2\x80\xBE" => '~');
581 case 'johab':
582 return array("\xE2\x82\xA9" => '\\');
583 }
584 if (strpos($lenc, 'iso-8859-') === 0) {
585 return array();
586 }
587 }
588 $ret = array();
589 if (self::unsafeIconv('UTF-8', $encoding, 'a') === false) {
590 return false;
591 }
592 for ($i = 0x20; $i <= 0x7E; $i++) { // all printable ASCII chars
593 $c = chr($i); // UTF-8 char
594 $r = self::unsafeIconv('UTF-8', "$encoding//IGNORE", $c); // initial conversion
595 if ($r === '' ||
596 // This line is needed for iconv implementations that do not
597 // omit characters that do not exist in the target character set
598 ($r === $c && self::unsafeIconv($encoding, 'UTF-8//IGNORE', $r) !== $c)
599 ) {
600 // Reverse engineer: what's the UTF-8 equiv of this byte
601 // sequence? This assumes that there's no variable width
602 // encoding that doesn't support ASCII.
603 $ret[self::unsafeIconv($encoding, 'UTF-8//IGNORE', $c)] = $c;
604 }
605 }
606 $encodings[$encoding] = $ret;
607 return $ret;
608 }
609}
610
611// vim: et sw=4 sts=4