1/*
2 +----------------------------------------------------------------------+
3 | PHP Version 5 |
4 +----------------------------------------------------------------------+
5 | Copyright (c) 1997-2015 The PHP Group |
6 +----------------------------------------------------------------------+
7 | This source file is subject to version 3.01 of the PHP license, |
8 | that is bundled with this package in the file LICENSE, and is |
9 | available through the world-wide-web at the following url: |
10 | http://www.php.net/license/3_01.txt |
11 | If you did not receive a copy of the PHP license and are unable to |
12 | obtain it through the world-wide-web, please send a note to |
13 | license@php.net so we can mail you a copy immediately. |
14 +----------------------------------------------------------------------+
15 | Authors: Rasmus Lerdorf <rasmus@php.net> |
16 | Jaakko Hyvätti <jaakko.hyvatti@iki.fi> |
17 | Wez Furlong <wez@thebrainroom.com> |
18 | Gustavo Lopes <cataphract@php.net> |
19 +----------------------------------------------------------------------+
20*/
21
22/* $Id$ */
23
24/*
25 * HTML entity resources:
26 *
27 * http://www.unicode.org/Public/MAPPINGS/OBSOLETE/UNI2SGML.TXT
28 *
29 * XHTML 1.0 DTD
30 * http://www.w3.org/TR/2002/REC-xhtml1-20020801/dtds.html#h-A2
31 *
32 * From HTML 4.01 strict DTD:
33 * http://www.w3.org/TR/html4/HTMLlat1.ent
34 * http://www.w3.org/TR/html4/HTMLsymbol.ent
35 * http://www.w3.org/TR/html4/HTMLspecial.ent
36 *
37 * HTML 5:
38 * http://dev.w3.org/html5/spec/Overview.html#named-character-references
39 */
40
41#include "php.h"
42#if PHP_WIN32
43#include "config.w32.h"
44#else
45#include <php_config.h>
46#endif
47#include "php_standard.h"
48#include "php_string.h"
49#include "SAPI.h"
50#if HAVE_LOCALE_H
51#include <locale.h>
52#endif
53#if HAVE_LANGINFO_H
54#include <langinfo.h>
55#endif
56
57#include <zend_hash.h>
58#include "html_tables.h"
59
60/* Macro for disabling flag of translation of non-basic entities where this isn't supported.
61 * Not appropriate for html_entity_decode/htmlspecialchars_decode */
62#define LIMIT_ALL(all, doctype, charset) do { \
63 (all) = (all) && !CHARSET_PARTIAL_SUPPORT((charset)) && ((doctype) != ENT_HTML_DOC_XML1); \
64} while (0)
65
66#define MB_FAILURE(pos, advance) do { \
67 *cursor = pos + (advance); \
68 *status = FAILURE; \
69 return 0; \
70} while (0)
71
72#define CHECK_LEN(pos, chars_need) ((str_len - (pos)) >= (chars_need))
73
74/* valid as single byte character or leading byte */
75#define utf8_lead(c) ((c) < 0x80 || ((c) >= 0xC2 && (c) <= 0xF4))
76/* whether it's actually valid depends on other stuff;
77 * this macro cannot check for non-shortest forms, surrogates or
78 * code points above 0x10FFFF */
79#define utf8_trail(c) ((c) >= 0x80 && (c) <= 0xBF)
80
81#define gb2312_lead(c) ((c) != 0x8E && (c) != 0x8F && (c) != 0xA0 && (c) != 0xFF)
82#define gb2312_trail(c) ((c) >= 0xA1 && (c) <= 0xFE)
83
84#define sjis_lead(c) ((c) != 0x80 && (c) != 0xA0 && (c) < 0xFD)
85#define sjis_trail(c) ((c) >= 0x40 && (c) != 0x7F && (c) < 0xFD)
86
87/* {{{ get_default_charset
88 */
89static char *get_default_charset(TSRMLS_D) {
90 if (PG(internal_encoding) && PG(internal_encoding)[0]) {
91 return PG(internal_encoding);
92 } else if (SG(default_charset) && SG(default_charset)[0] ) {
93 return SG(default_charset);
94 }
95 return NULL;
96}
97/* }}} */
98
99/* {{{ get_next_char
100 */
101static inline unsigned int get_next_char(
102 enum entity_charset charset,
103 const unsigned char *str,
104 size_t str_len,
105 size_t *cursor,
106 int *status)
107{
108 size_t pos = *cursor;
109 unsigned int this_char = 0;
110
111 *status = SUCCESS;
112 assert(pos <= str_len);
113
114 if (!CHECK_LEN(pos, 1))
115 MB_FAILURE(pos, 1);
116
117 switch (charset) {
118 case cs_utf_8:
119 {
120 /* We'll follow strategy 2. from section 3.6.1 of UTR #36:
121 * "In a reported illegal byte sequence, do not include any
122 * non-initial byte that encodes a valid character or is a leading
123 * byte for a valid sequence." */
124 unsigned char c;
125 c = str[pos];
126 if (c < 0x80) {
127 this_char = c;
128 pos++;
129 } else if (c < 0xc2) {
130 MB_FAILURE(pos, 1);
131 } else if (c < 0xe0) {
132 if (!CHECK_LEN(pos, 2))
133 MB_FAILURE(pos, 1);
134
135 if (!utf8_trail(str[pos + 1])) {
136 MB_FAILURE(pos, utf8_lead(str[pos + 1]) ? 1 : 2);
137 }
138 this_char = ((c & 0x1f) << 6) | (str[pos + 1] & 0x3f);
139 if (this_char < 0x80) { /* non-shortest form */
140 MB_FAILURE(pos, 2);
141 }
142 pos += 2;
143 } else if (c < 0xf0) {
144 size_t avail = str_len - pos;
145
146 if (avail < 3 ||
147 !utf8_trail(str[pos + 1]) || !utf8_trail(str[pos + 2])) {
148 if (avail < 2 || utf8_lead(str[pos + 1]))
149 MB_FAILURE(pos, 1);
150 else if (avail < 3 || utf8_lead(str[pos + 2]))
151 MB_FAILURE(pos, 2);
152 else
153 MB_FAILURE(pos, 3);
154 }
155
156 this_char = ((c & 0x0f) << 12) | ((str[pos + 1] & 0x3f) << 6) | (str[pos + 2] & 0x3f);
157 if (this_char < 0x800) { /* non-shortest form */
158 MB_FAILURE(pos, 3);
159 } else if (this_char >= 0xd800 && this_char <= 0xdfff) { /* surrogate */
160 MB_FAILURE(pos, 3);
161 }
162 pos += 3;
163 } else if (c < 0xf5) {
164 size_t avail = str_len - pos;
165
166 if (avail < 4 ||
167 !utf8_trail(str[pos + 1]) || !utf8_trail(str[pos + 2]) ||
168 !utf8_trail(str[pos + 3])) {
169 if (avail < 2 || utf8_lead(str[pos + 1]))
170 MB_FAILURE(pos, 1);
171 else if (avail < 3 || utf8_lead(str[pos + 2]))
172 MB_FAILURE(pos, 2);
173 else if (avail < 4 || utf8_lead(str[pos + 3]))
174 MB_FAILURE(pos, 3);
175 else
176 MB_FAILURE(pos, 4);
177 }
178
179 this_char = ((c & 0x07) << 18) | ((str[pos + 1] & 0x3f) << 12) | ((str[pos + 2] & 0x3f) << 6) | (str[pos + 3] & 0x3f);
180 if (this_char < 0x10000 || this_char > 0x10FFFF) { /* non-shortest form or outside range */
181 MB_FAILURE(pos, 4);
182 }
183 pos += 4;
184 } else {
185 MB_FAILURE(pos, 1);
186 }
187 }
188 break;
189
190 case cs_big5:
191 /* reference http://demo.icu-project.org/icu-bin/convexp?conv=big5 */
192 {
193 unsigned char c = str[pos];
194 if (c >= 0x81 && c <= 0xFE) {
195 unsigned char next;
196 if (!CHECK_LEN(pos, 2))
197 MB_FAILURE(pos, 1);
198
199 next = str[pos + 1];
200
201 if ((next >= 0x40 && next <= 0x7E) ||
202 (next >= 0xA1 && next <= 0xFE)) {
203 this_char = (c << 8) | next;
204 } else {
205 MB_FAILURE(pos, 1);
206 }
207 pos += 2;
208 } else {
209 this_char = c;
210 pos += 1;
211 }
212 }
213 break;
214
215 case cs_big5hkscs:
216 {
217 unsigned char c = str[pos];
218 if (c >= 0x81 && c <= 0xFE) {
219 unsigned char next;
220 if (!CHECK_LEN(pos, 2))
221 MB_FAILURE(pos, 1);
222
223 next = str[pos + 1];
224
225 if ((next >= 0x40 && next <= 0x7E) ||
226 (next >= 0xA1 && next <= 0xFE)) {
227 this_char = (c << 8) | next;
228 } else if (next != 0x80 && next != 0xFF) {
229 MB_FAILURE(pos, 1);
230 } else {
231 MB_FAILURE(pos, 2);
232 }
233 pos += 2;
234 } else {
235 this_char = c;
236 pos += 1;
237 }
238 }
239 break;
240
241 case cs_gb2312: /* EUC-CN */
242 {
243 unsigned char c = str[pos];
244 if (c >= 0xA1 && c <= 0xFE) {
245 unsigned char next;
246 if (!CHECK_LEN(pos, 2))
247 MB_FAILURE(pos, 1);
248
249 next = str[pos + 1];
250
251 if (gb2312_trail(next)) {
252 this_char = (c << 8) | next;
253 } else if (gb2312_lead(next)) {
254 MB_FAILURE(pos, 1);
255 } else {
256 MB_FAILURE(pos, 2);
257 }
258 pos += 2;
259 } else if (gb2312_lead(c)) {
260 this_char = c;
261 pos += 1;
262 } else {
263 MB_FAILURE(pos, 1);
264 }
265 }
266 break;
267
268 case cs_sjis:
269 {
270 unsigned char c = str[pos];
271 if ((c >= 0x81 && c <= 0x9F) || (c >= 0xE0 && c <= 0xFC)) {
272 unsigned char next;
273 if (!CHECK_LEN(pos, 2))
274 MB_FAILURE(pos, 1);
275
276 next = str[pos + 1];
277
278 if (sjis_trail(next)) {
279 this_char = (c << 8) | next;
280 } else if (sjis_lead(next)) {
281 MB_FAILURE(pos, 1);
282 } else {
283 MB_FAILURE(pos, 2);
284 }
285 pos += 2;
286 } else if (c < 0x80 || (c >= 0xA1 && c <= 0xDF)) {
287 this_char = c;
288 pos += 1;
289 } else {
290 MB_FAILURE(pos, 1);
291 }
292 }
293 break;
294
295 case cs_eucjp:
296 {
297 unsigned char c = str[pos];
298
299 if (c >= 0xA1 && c <= 0xFE) {
300 unsigned next;
301 if (!CHECK_LEN(pos, 2))
302 MB_FAILURE(pos, 1);
303 next = str[pos + 1];
304
305 if (next >= 0xA1 && next <= 0xFE) {
306 /* this a jis kanji char */
307 this_char = (c << 8) | next;
308 } else {
309 MB_FAILURE(pos, (next != 0xA0 && next != 0xFF) ? 1 : 2);
310 }
311 pos += 2;
312 } else if (c == 0x8E) {
313 unsigned next;
314 if (!CHECK_LEN(pos, 2))
315 MB_FAILURE(pos, 1);
316
317 next = str[pos + 1];
318 if (next >= 0xA1 && next <= 0xDF) {
319 /* JIS X 0201 kana */
320 this_char = (c << 8) | next;
321 } else {
322 MB_FAILURE(pos, (next != 0xA0 && next != 0xFF) ? 1 : 2);
323 }
324 pos += 2;
325 } else if (c == 0x8F) {
326 size_t avail = str_len - pos;
327
328 if (avail < 3 || !(str[pos + 1] >= 0xA1 && str[pos + 1] <= 0xFE) ||
329 !(str[pos + 2] >= 0xA1 && str[pos + 2] <= 0xFE)) {
330 if (avail < 2 || (str[pos + 1] != 0xA0 && str[pos + 1] != 0xFF))
331 MB_FAILURE(pos, 1);
332 else if (avail < 3 || (str[pos + 2] != 0xA0 && str[pos + 2] != 0xFF))
333 MB_FAILURE(pos, 2);
334 else
335 MB_FAILURE(pos, 3);
336 } else {
337 /* JIS X 0212 hojo-kanji */
338 this_char = (c << 16) | (str[pos + 1] << 8) | str[pos + 2];
339 }
340 pos += 3;
341 } else if (c != 0xA0 && c != 0xFF) {
342 /* character encoded in 1 code unit */
343 this_char = c;
344 pos += 1;
345 } else {
346 MB_FAILURE(pos, 1);
347 }
348 }
349 break;
350 default:
351 /* single-byte charsets */
352 this_char = str[pos++];
353 break;
354 }
355
356 *cursor = pos;
357 return this_char;
358}
359/* }}} */
360
361/* {{{ php_next_utf8_char
362 * Public interface for get_next_char used with UTF-8 */
363 PHPAPI unsigned int php_next_utf8_char(
364 const unsigned char *str,
365 size_t str_len,
366 size_t *cursor,
367 int *status)
368{
369 return get_next_char(cs_utf_8, str, str_len, cursor, status);
370}
371/* }}} */
372
373/* {{{ entity_charset determine_charset
374 * returns the charset identifier based on current locale or a hint.
375 * defaults to UTF-8 */
376static enum entity_charset determine_charset(char *charset_hint TSRMLS_DC)
377{
378 int i;
379 enum entity_charset charset = cs_utf_8;
380 int len = 0;
381 const zend_encoding *zenc;
382
383 /* Default is now UTF-8 */
384 if (charset_hint == NULL)
385 return cs_utf_8;
386
387 if ((len = strlen(charset_hint)) != 0) {
388 goto det_charset;
389 }
390
391 zenc = zend_multibyte_get_internal_encoding(TSRMLS_C);
392 if (zenc != NULL) {
393 charset_hint = (char *)zend_multibyte_get_encoding_name(zenc);
394 if (charset_hint != NULL && (len=strlen(charset_hint)) != 0) {
395 if ((len == 4) /* sizeof (none|auto|pass) */ &&
396 (!memcmp("pass", charset_hint, 4) ||
397 !memcmp("auto", charset_hint, 4) ||
398 !memcmp("none", charset_hint, 4))) {
399 charset_hint = NULL;
400 len = 0;
401 } else {
402 goto det_charset;
403 }
404 }
405 }
406
407 charset_hint = SG(default_charset);
408 if (charset_hint != NULL && (len=strlen(charset_hint)) != 0) {
409 goto det_charset;
410 }
411
412 /* try to detect the charset for the locale */
413#if HAVE_NL_LANGINFO && HAVE_LOCALE_H && defined(CODESET)
414 charset_hint = nl_langinfo(CODESET);
415 if (charset_hint != NULL && (len=strlen(charset_hint)) != 0) {
416 goto det_charset;
417 }
418#endif
419
420#if HAVE_LOCALE_H
421 /* try to figure out the charset from the locale */
422 {
423 char *localename;
424 char *dot, *at;
425
426 /* lang[_territory][.codeset][@modifier] */
427 localename = setlocale(LC_CTYPE, NULL);
428
429 dot = strchr(localename, '.');
430 if (dot) {
431 dot++;
432 /* locale specifies a codeset */
433 at = strchr(dot, '@');
434 if (at)
435 len = at - dot;
436 else
437 len = strlen(dot);
438 charset_hint = dot;
439 } else {
440 /* no explicit name; see if the name itself
441 * is the charset */
442 charset_hint = localename;
443 len = strlen(charset_hint);
444 }
445 }
446#endif
447
448det_charset:
449
450 if (charset_hint) {
451 int found = 0;
452
453 /* now walk the charset map and look for the codeset */
454 for (i = 0; charset_map[i].codeset; i++) {
455 if (len == strlen(charset_map[i].codeset) && strncasecmp(charset_hint, charset_map[i].codeset, len) == 0) {
456 charset = charset_map[i].charset;
457 found = 1;
458 break;
459 }
460 }
461 if (!found) {
462 php_error_docref(NULL TSRMLS_CC, E_WARNING, "charset `%s' not supported, assuming utf-8",
463 charset_hint);
464 }
465 }
466 return charset;
467}
468/* }}} */
469
470/* {{{ php_utf32_utf8 */
471static inline size_t php_utf32_utf8(unsigned char *buf, unsigned k)
472{
473 size_t retval = 0;
474
475 /* assert(0x0 <= k <= 0x10FFFF); */
476
477 if (k < 0x80) {
478 buf[0] = k;
479 retval = 1;
480 } else if (k < 0x800) {
481 buf[0] = 0xc0 | (k >> 6);
482 buf[1] = 0x80 | (k & 0x3f);
483 retval = 2;
484 } else if (k < 0x10000) {
485 buf[0] = 0xe0 | (k >> 12);
486 buf[1] = 0x80 | ((k >> 6) & 0x3f);
487 buf[2] = 0x80 | (k & 0x3f);
488 retval = 3;
489 } else {
490 buf[0] = 0xf0 | (k >> 18);
491 buf[1] = 0x80 | ((k >> 12) & 0x3f);
492 buf[2] = 0x80 | ((k >> 6) & 0x3f);
493 buf[3] = 0x80 | (k & 0x3f);
494 retval = 4;
495 }
496 /* UTF-8 has been restricted to max 4 bytes since RFC 3629 */
497
498 return retval;
499}
500/* }}} */
501
502/* {{{ php_mb2_int_to_char
503 * Convert back big endian int representation of sequence of one or two 8-bit code units. */
504static inline size_t php_mb2_int_to_char(unsigned char *buf, unsigned k)
505{
506 assert(k <= 0xFFFFU);
507 /* one or two bytes */
508 if (k <= 0xFFU) { /* 1 */
509 buf[0] = k;
510 return 1U;
511 } else { /* 2 */
512 buf[0] = k >> 8;
513 buf[1] = k & 0xFFU;
514 return 2U;
515 }
516}
517/* }}} */
518
519/* {{{ php_mb3_int_to_char
520 * Convert back big endian int representation of sequence of one to three 8-bit code units.
521 * For EUC-JP. */
522static inline size_t php_mb3_int_to_char(unsigned char *buf, unsigned k)
523{
524 assert(k <= 0xFFFFFFU);
525 /* one to three bytes */
526 if (k <= 0xFFU) { /* 1 */
527 buf[0] = k;
528 return 1U;
529 } else if (k <= 0xFFFFU) { /* 2 */
530 buf[0] = k >> 8;
531 buf[1] = k & 0xFFU;
532 return 2U;
533 } else {
534 buf[0] = k >> 16;
535 buf[1] = (k >> 8) & 0xFFU;
536 buf[2] = k & 0xFFU;
537 return 3U;
538 }
539}
540/* }}} */
541
542
543/* {{{ unimap_bsearc_cmp
544 * Binary search of unicode code points in unicode <--> charset mapping.
545 * Returns the code point in the target charset (whose mapping table was given) or 0 if
546 * the unicode code point is not in the table.
547 */
548static inline unsigned char unimap_bsearch(const uni_to_enc *table, unsigned code_key_a, size_t num)
549{
550 const uni_to_enc *l = table,
551 *h = &table[num-1],
552 *m;
553 unsigned short code_key;
554
555 /* we have no mappings outside the BMP */
556 if (code_key_a > 0xFFFFU)
557 return 0;
558
559 code_key = (unsigned short) code_key_a;
560
561 while (l <= h) {
562 m = l + (h - l) / 2;
563 if (code_key < m->un_code_point)
564 h = m - 1;
565 else if (code_key > m->un_code_point)
566 l = m + 1;
567 else
568 return m->cs_code;
569 }
570 return 0;
571}
572/* }}} */
573
574/* {{{ map_from_unicode */
575static inline int map_from_unicode(unsigned code, enum entity_charset charset, unsigned *res)
576{
577 unsigned char found;
578 const uni_to_enc *table;
579 size_t table_size;
580
581 switch (charset) {
582 case cs_8859_1:
583 /* identity mapping of code points to unicode */
584 if (code > 0xFF) {
585 return FAILURE;
586 }
587 *res = code;
588 break;
589
590 case cs_8859_5:
591 if (code <= 0xA0 || code == 0xAD /* soft hyphen */) {
592 *res = code;
593 } else if (code == 0x2116) {
594 *res = 0xF0; /* numero sign */
595 } else if (code == 0xA7) {
596 *res = 0xFD; /* section sign */
597 } else if (code >= 0x0401 && code <= 0x044F) {
598 if (code == 0x040D || code == 0x0450 || code == 0x045D)
599 return FAILURE;
600 *res = code - 0x360;
601 } else {
602 return FAILURE;
603 }
604 break;
605
606 case cs_8859_15:
607 if (code < 0xA4 || (code > 0xBE && code <= 0xFF)) {
608 *res = code;
609 } else { /* between A4 and 0xBE */
610 found = unimap_bsearch(unimap_iso885915,
611 code, sizeof(unimap_iso885915) / sizeof(*unimap_iso885915));
612 if (found)
613 *res = found;
614 else
615 return FAILURE;
616 }
617 break;
618
619 case cs_cp1252:
620 if (code <= 0x7F || (code >= 0xA0 && code <= 0xFF)) {
621 *res = code;
622 } else {
623 found = unimap_bsearch(unimap_win1252,
624 code, sizeof(unimap_win1252) / sizeof(*unimap_win1252));
625 if (found)
626 *res = found;
627 else
628 return FAILURE;
629 }
630 break;
631
632 case cs_macroman:
633 if (code == 0x7F)
634 return FAILURE;
635 table = unimap_macroman;
636 table_size = sizeof(unimap_macroman) / sizeof(*unimap_macroman);
637 goto table_over_7F;
638 case cs_cp1251:
639 table = unimap_win1251;
640 table_size = sizeof(unimap_win1251) / sizeof(*unimap_win1251);
641 goto table_over_7F;
642 case cs_koi8r:
643 table = unimap_koi8r;
644 table_size = sizeof(unimap_koi8r) / sizeof(*unimap_koi8r);
645 goto table_over_7F;
646 case cs_cp866:
647 table = unimap_cp866;
648 table_size = sizeof(unimap_cp866) / sizeof(*unimap_cp866);
649
650table_over_7F:
651 if (code <= 0x7F) {
652 *res = code;
653 } else {
654 found = unimap_bsearch(table, code, table_size);
655 if (found)
656 *res = found;
657 else
658 return FAILURE;
659 }
660 break;
661
662 /* from here on, only map the possible characters in the ASCII range.
663 * to improve support here, it's a matter of building the unicode mappings.
664 * See <http://www.unicode.org/Public/6.0.0/ucd/Unihan.zip> */
665 case cs_sjis:
666 case cs_eucjp:
667 /* we interpret 0x5C as the Yen symbol. This is not universal.
668 * See <http://www.w3.org/Submission/japanese-xml/#ambiguity_of_yen> */
669 if (code >= 0x20 && code <= 0x7D) {
670 if (code == 0x5C)
671 return FAILURE;
672 *res = code;
673 } else {
674 return FAILURE;
675 }
676 break;
677
678 case cs_big5:
679 case cs_big5hkscs:
680 case cs_gb2312:
681 if (code >= 0x20 && code <= 0x7D) {
682 *res = code;
683 } else {
684 return FAILURE;
685 }
686 break;
687
688 default:
689 return FAILURE;
690 }
691
692 return SUCCESS;
693}
694/* }}} */
695
696/* {{{ */
697static inline void map_to_unicode(unsigned code, const enc_to_uni *table, unsigned *res)
698{
699 /* only single byte encodings are currently supported; assumed code <= 0xFF */
700 *res = table->inner[ENT_ENC_TO_UNI_STAGE1(code)]->uni_cp[ENT_ENC_TO_UNI_STAGE2(code)];
701}
702/* }}} */
703
704/* {{{ unicode_cp_is_allowed */
705static inline int unicode_cp_is_allowed(unsigned uni_cp, int document_type)
706{
707 /* XML 1.0 HTML 4.01 HTML 5
708 * 0x09..0x0A 0x09..0x0A 0x09..0x0A
709 * 0x0D 0x0D 0x0C..0x0D
710 * 0x0020..0xD7FF 0x20..0x7E 0x20..0x7E
711 * 0x00A0..0xD7FF 0x00A0..0xD7FF
712 * 0xE000..0xFFFD 0xE000..0x10FFFF 0xE000..0xFDCF
713 * 0x010000..0x10FFFF 0xFDF0..0x10FFFF (*)
714 *
715 * (*) exclude code points where ((code & 0xFFFF) >= 0xFFFE)
716 *
717 * References:
718 * XML 1.0: <http://www.w3.org/TR/REC-xml/#charsets>
719 * HTML 4.01: <http://www.w3.org/TR/1999/PR-html40-19990824/sgml/sgmldecl.html>
720 * HTML 5: <http://dev.w3.org/html5/spec/Overview.html#preprocessing-the-input-stream>
721 *
722 * Not sure this is the relevant part for HTML 5, though. I opted to
723 * disallow the characters that would result in a parse error when
724 * preprocessing of the input stream. See also section 8.1.3.
725 *
726 * It's unclear if XHTML 1.0 allows C1 characters. I'll opt to apply to
727 * XHTML 1.0 the same rules as for XML 1.0.
728 * See <http://cmsmcq.com/2007/C1.xml>.
729 */
730
731 switch (document_type) {
732 case ENT_HTML_DOC_HTML401:
733 return (uni_cp >= 0x20 && uni_cp <= 0x7E) ||
734 (uni_cp == 0x0A || uni_cp == 0x09 || uni_cp == 0x0D) ||
735 (uni_cp >= 0xA0 && uni_cp <= 0xD7FF) ||
736 (uni_cp >= 0xE000 && uni_cp <= 0x10FFFF);
737 case ENT_HTML_DOC_HTML5:
738 return (uni_cp >= 0x20 && uni_cp <= 0x7E) ||
739 (uni_cp >= 0x09 && uni_cp <= 0x0D && uni_cp != 0x0B) || /* form feed U+0C allowed */
740 (uni_cp >= 0xA0 && uni_cp <= 0xD7FF) ||
741 (uni_cp >= 0xE000 && uni_cp <= 0x10FFFF &&
742 ((uni_cp & 0xFFFF) < 0xFFFE) && /* last two of each plane (nonchars) disallowed */
743 (uni_cp < 0xFDD0 || uni_cp > 0xFDEF)); /* U+FDD0-U+FDEF (nonchars) disallowed */
744 case ENT_HTML_DOC_XHTML:
745 case ENT_HTML_DOC_XML1:
746 return (uni_cp >= 0x20 && uni_cp <= 0xD7FF) ||
747 (uni_cp == 0x0A || uni_cp == 0x09 || uni_cp == 0x0D) ||
748 (uni_cp >= 0xE000 && uni_cp <= 0x10FFFF && uni_cp != 0xFFFE && uni_cp != 0xFFFF);
749 default:
750 return 1;
751 }
752}
753/* }}} */
754
755/* {{{ unicode_cp_is_allowed */
756static inline int numeric_entity_is_allowed(unsigned uni_cp, int document_type)
757{
758 /* less restrictive than unicode_cp_is_allowed */
759 switch (document_type) {
760 case ENT_HTML_DOC_HTML401:
761 /* all non-SGML characters (those marked with UNUSED in DESCSET) should be
762 * representable with numeric entities */
763 return uni_cp <= 0x10FFFF;
764 case ENT_HTML_DOC_HTML5:
765 /* 8.1.4. The numeric character reference forms described above are allowed to
766 * reference any Unicode code point other than U+0000, U+000D, permanently
767 * undefined Unicode characters (noncharacters), and control characters other
768 * than space characters (U+0009, U+000A, U+000C and U+000D) */
769 /* seems to allow surrogate characters, then */
770 return (uni_cp >= 0x20 && uni_cp <= 0x7E) ||
771 (uni_cp >= 0x09 && uni_cp <= 0x0C && uni_cp != 0x0B) || /* form feed U+0C allowed, but not U+0D */
772 (uni_cp >= 0xA0 && uni_cp <= 0x10FFFF &&
773 ((uni_cp & 0xFFFF) < 0xFFFE) && /* last two of each plane (nonchars) disallowed */
774 (uni_cp < 0xFDD0 || uni_cp > 0xFDEF)); /* U+FDD0-U+FDEF (nonchars) disallowed */
775 case ENT_HTML_DOC_XHTML:
776 case ENT_HTML_DOC_XML1:
777 /* OTOH, XML 1.0 requires "character references to match the production for Char
778 * See <http://www.w3.org/TR/REC-xml/#NT-CharRef> */
779 return unicode_cp_is_allowed(uni_cp, document_type);
780 default:
781 return 1;
782 }
783}
784/* }}} */
785
786/* {{{ process_numeric_entity
787 * Auxiliary function to traverse_for_entities.
788 * On input, *buf should point to the first character after # and on output, it's the last
789 * byte read, no matter if there was success or insuccess.
790 */
791static inline int process_numeric_entity(const char **buf, unsigned *code_point)
792{
793 long code_l;
794 int hexadecimal = (**buf == 'x' || **buf == 'X'); /* TODO: XML apparently disallows "X" */
795 char *endptr;
796
797 if (hexadecimal && (**buf != '\0'))
798 (*buf)++;
799
800 /* strtol allows whitespace and other stuff in the beginning
801 * we're not interested */
802 if ((hexadecimal && !isxdigit(**buf)) ||
803 (!hexadecimal && !isdigit(**buf))) {
804 return FAILURE;
805 }
806
807 code_l = strtol(*buf, &endptr, hexadecimal ? 16 : 10);
808 /* we're guaranteed there were valid digits, so *endptr > buf */
809 *buf = endptr;
810
811 if (**buf != ';')
812 return FAILURE;
813
814 /* many more are invalid, but that depends on whether it's HTML
815 * (and which version) or XML. */
816 if (code_l > 0x10FFFFL)
817 return FAILURE;
818
819 if (code_point != NULL)
820 *code_point = (unsigned)code_l;
821
822 return SUCCESS;
823}
824/* }}} */
825
826/* {{{ process_named_entity */
827static inline int process_named_entity_html(const char **buf, const char **start, size_t *length)
828{
829 *start = *buf;
830
831 /* "&" is represented by a 0x26 in all supported encodings. That means
832 * the byte after represents a character or is the leading byte of an
833 * sequence of 8-bit code units. If in the ranges below, it represents
834 * necessarily a alpha character because none of the supported encodings
835 * has an overlap with ASCII in the leading byte (only on the second one) */
836 while ((**buf >= 'a' && **buf <= 'z') ||
837 (**buf >= 'A' && **buf <= 'Z') ||
838 (**buf >= '0' && **buf <= '9')) {
839 (*buf)++;
840 }
841
842 if (**buf != ';')
843 return FAILURE;
844
845 /* cast to size_t OK as the quantity is always non-negative */
846 *length = *buf - *start;
847
848 if (*length == 0)
849 return FAILURE;
850
851 return SUCCESS;
852}
853/* }}} */
854
855/* {{{ resolve_named_entity_html */
856static inline int resolve_named_entity_html(const char *start, size_t length, const entity_ht *ht, unsigned *uni_cp1, unsigned *uni_cp2)
857{
858 const entity_cp_map *s;
859 ulong hash = zend_inline_hash_func(start, length);
860
861 s = ht->buckets[hash % ht->num_elems];
862 while (s->entity) {
863 if (s->entity_len == length) {
864 if (memcmp(start, s->entity, length) == 0) {
865 *uni_cp1 = s->codepoint1;
866 *uni_cp2 = s->codepoint2;
867 return SUCCESS;
868 }
869 }
870 s++;
871 }
872 return FAILURE;
873}
874/* }}} */
875
876static inline size_t write_octet_sequence(unsigned char *buf, enum entity_charset charset, unsigned code) {
877 /* code is not necessarily a unicode code point */
878 switch (charset) {
879 case cs_utf_8:
880 return php_utf32_utf8(buf, code);
881
882 case cs_8859_1:
883 case cs_cp1252:
884 case cs_8859_15:
885 case cs_koi8r:
886 case cs_cp1251:
887 case cs_8859_5:
888 case cs_cp866:
889 case cs_macroman:
890 /* single byte stuff */
891 *buf = code;
892 return 1;
893
894 case cs_big5:
895 case cs_big5hkscs:
896 case cs_sjis:
897 case cs_gb2312:
898 /* we don't have complete unicode mappings for these yet in entity_decode,
899 * and we opt to pass through the octet sequences for these in htmlentities
900 * instead of converting to an int and then converting back. */
901#if 0
902 return php_mb2_int_to_char(buf, code);
903#else
904#if ZEND_DEBUG
905 assert(code <= 0xFFU);
906#endif
907 *buf = code;
908 return 1;
909#endif
910
911 case cs_eucjp:
912#if 0 /* idem */
913 return php_mb2_int_to_char(buf, code);
914#else
915#if ZEND_DEBUG
916 assert(code <= 0xFFU);
917#endif
918 *buf = code;
919 return 1;
920#endif
921
922 default:
923 assert(0);
924 return 0;
925 }
926}
927
928/* {{{ traverse_for_entities
929 * Auxiliary function to php_unescape_html_entities().
930 * - The argument "all" determines if all numeric entities are decode or only those
931 * that correspond to quotes (depending on quote_style).
932 */
933/* maximum expansion (factor 1.2) for HTML 5 with &nGt; and &nLt; */
934/* +2 is 1 because of rest (probably unnecessary), 1 because of terminating 0 */
935#define TRAVERSE_FOR_ENTITIES_EXPAND_SIZE(oldlen) ((oldlen) + (oldlen) / 5 + 2)
936static void traverse_for_entities(
937 const char *old,
938 size_t oldlen,
939 char *ret, /* should have allocated TRAVERSE_FOR_ENTITIES_EXPAND_SIZE(olden) */
940 size_t *retlen,
941 int all,
942 int flags,
943 const entity_ht *inv_map,
944 enum entity_charset charset)
945{
946 const char *p,
947 *lim;
948 char *q;
949 int doctype = flags & ENT_HTML_DOC_TYPE_MASK;
950
951 lim = old + oldlen; /* terminator address */
952 assert(*lim == '\0');
953
954 for (p = old, q = ret; p < lim;) {
955 unsigned code, code2 = 0;
956 const char *next = NULL; /* when set, next > p, otherwise possible inf loop */
957
958 /* Shift JIS, Big5 and HKSCS use multi-byte encodings where an
959 * ASCII range byte can be part of a multi-byte sequence.
960 * However, they start at 0x40, therefore if we find a 0x26 byte,
961 * we're sure it represents the '&' character. */
962
963 /* assumes there are no single-char entities */
964 if (p[0] != '&' || (p + 3 >= lim)) {
965 *(q++) = *(p++);
966 continue;
967 }
968
969 /* now p[3] is surely valid and is no terminator */
970
971 /* numerical entity */
972 if (p[1] == '#') {
973 next = &p[2];
974 if (process_numeric_entity(&next, &code) == FAILURE)
975 goto invalid_code;
976
977 /* If we're in htmlspecialchars_decode, we're only decoding entities
978 * that represent &, <, >, " and '. Is this one of them? */
979 if (!all && (code > 63U ||
980 stage3_table_be_apos_00000[code].data.ent.entity == NULL))
981 goto invalid_code;
982
983 /* are we allowed to decode this entity in this document type?
984 * HTML 5 is the only that has a character that cannot be used in
985 * a numeric entity but is allowed literally (U+000D). The
986 * unoptimized version would be ... || !numeric_entity_is_allowed(code) */
987 if (!unicode_cp_is_allowed(code, doctype) ||
988 (doctype == ENT_HTML_DOC_HTML5 && code == 0x0D))
989 goto invalid_code;
990 } else {
991 const char *start;
992 size_t ent_len;
993
994 next = &p[1];
995 start = next;
996
997 if (process_named_entity_html(&next, &start, &ent_len) == FAILURE)
998 goto invalid_code;
999
1000 if (resolve_named_entity_html(start, ent_len, inv_map, &code, &code2) == FAILURE) {
1001 if (doctype == ENT_HTML_DOC_XHTML && ent_len == 4 && start[0] == 'a'
1002 && start[1] == 'p' && start[2] == 'o' && start[3] == 's') {
1003 /* uses html4 inv_map, which doesn't include apos;. This is a
1004 * hack to support it */
1005 code = (unsigned) '\'';
1006 } else {
1007 goto invalid_code;
1008 }
1009 }
1010 }
1011
1012 assert(*next == ';');
1013
1014 if (((code == '\'' && !(flags & ENT_HTML_QUOTE_SINGLE)) ||
1015 (code == '"' && !(flags & ENT_HTML_QUOTE_DOUBLE)))
1016 /* && code2 == '\0' always true for current maps */)
1017 goto invalid_code;
1018
1019 /* UTF-8 doesn't need mapping (ISO-8859-1 doesn't either, but
1020 * the call is needed to ensure the codepoint <= U+00FF) */
1021 if (charset != cs_utf_8) {
1022 /* replace unicode code point */
1023 if (map_from_unicode(code, charset, &code) == FAILURE || code2 != 0)
1024 goto invalid_code; /* not representable in target charset */
1025 }
1026
1027 q += write_octet_sequence(q, charset, code);
1028 if (code2) {
1029 q += write_octet_sequence(q, charset, code2);
1030 }
1031
1032 /* jump over the valid entity; may go beyond size of buffer; np */
1033 p = next + 1;
1034 continue;
1035
1036invalid_code:
1037 for (; p < next; p++) {
1038 *(q++) = *p;
1039 }
1040 }
1041
1042 *q = '\0';
1043 *retlen = (size_t)(q - ret);
1044}
1045/* }}} */
1046
1047/* {{{ unescape_inverse_map */
1048static const entity_ht *unescape_inverse_map(int all, int flags)
1049{
1050 int document_type = flags & ENT_HTML_DOC_TYPE_MASK;
1051
1052 if (all) {
1053 switch (document_type) {
1054 case ENT_HTML_DOC_HTML401:
1055 case ENT_HTML_DOC_XHTML: /* but watch out for &apos;...*/
1056 return &ent_ht_html4;
1057 case ENT_HTML_DOC_HTML5:
1058 return &ent_ht_html5;
1059 default:
1060 return &ent_ht_be_apos;
1061 }
1062 } else {
1063 switch (document_type) {
1064 case ENT_HTML_DOC_HTML401:
1065 return &ent_ht_be_noapos;
1066 default:
1067 return &ent_ht_be_apos;
1068 }
1069 }
1070}
1071/* }}} */
1072
1073/* {{{ determine_entity_table
1074 * Entity table to use. Note that entity tables are defined in terms of
1075 * unicode code points */
1076static entity_table_opt determine_entity_table(int all, int doctype)
1077{
1078 entity_table_opt retval = {NULL};
1079
1080 assert(!(doctype == ENT_HTML_DOC_XML1 && all));
1081
1082 if (all) {
1083 retval.ms_table = (doctype == ENT_HTML_DOC_HTML5) ?
1084 entity_ms_table_html5 : entity_ms_table_html4;
1085 } else {
1086 retval.table = (doctype == ENT_HTML_DOC_HTML401) ?
1087 stage3_table_be_noapos_00000 : stage3_table_be_apos_00000;
1088 }
1089 return retval;
1090}
1091/* }}} */
1092
1093/* {{{ php_unescape_html_entities
1094 * The parameter "all" should be true to decode all possible entities, false to decode
1095 * only the basic ones, i.e., those in basic_entities_ex + the numeric entities
1096 * that correspond to quotes.
1097 */
1098PHPAPI char *php_unescape_html_entities(unsigned char *old, size_t oldlen, size_t *newlen, int all, int flags, char *hint_charset TSRMLS_DC)
1099{
1100 size_t retlen;
1101 char *ret;
1102 enum entity_charset charset;
1103 const entity_ht *inverse_map = NULL;
1104 size_t new_size = TRAVERSE_FOR_ENTITIES_EXPAND_SIZE(oldlen);
1105
1106 if (all) {
1107 charset = determine_charset(hint_charset TSRMLS_CC);
1108 } else {
1109 charset = cs_8859_1; /* charset shouldn't matter, use ISO-8859-1 for performance */
1110 }
1111
1112 /* don't use LIMIT_ALL! */
1113
1114 if (oldlen > new_size) {
1115 /* overflow, refuse to do anything */
1116 ret = estrndup((char*)old, oldlen);
1117 retlen = oldlen;
1118 goto empty_source;
1119 }
1120 ret = emalloc(new_size);
1121 *ret = '\0';
1122 retlen = oldlen;
1123 if (retlen == 0) {
1124 goto empty_source;
1125 }
1126
1127 inverse_map = unescape_inverse_map(all, flags);
1128
1129 /* replace numeric entities */
1130 traverse_for_entities(old, oldlen, ret, &retlen, all, flags, inverse_map, charset);
1131
1132empty_source:
1133 *newlen = retlen;
1134 return ret;
1135}
1136/* }}} */
1137
1138PHPAPI char *php_escape_html_entities(unsigned char *old, size_t oldlen, size_t *newlen, int all, int flags, char *hint_charset TSRMLS_DC)
1139{
1140 return php_escape_html_entities_ex(old, oldlen, newlen, all, flags, hint_charset, 1 TSRMLS_CC);
1141}
1142
1143/* {{{ find_entity_for_char */
1144static inline void find_entity_for_char(
1145 unsigned int k,
1146 enum entity_charset charset,
1147 const entity_stage1_row *table,
1148 const unsigned char **entity,
1149 size_t *entity_len,
1150 unsigned char *old,
1151 size_t oldlen,
1152 size_t *cursor)
1153{
1154 unsigned stage1_idx = ENT_STAGE1_INDEX(k);
1155 const entity_stage3_row *c;
1156
1157 if (stage1_idx > 0x1D) {
1158 *entity = NULL;
1159 *entity_len = 0;
1160 return;
1161 }
1162
1163 c = &table[stage1_idx][ENT_STAGE2_INDEX(k)][ENT_STAGE3_INDEX(k)];
1164
1165 if (!c->ambiguous) {
1166 *entity = (const unsigned char *)c->data.ent.entity;
1167 *entity_len = c->data.ent.entity_len;
1168 } else {
1169 /* peek at next char */
1170 size_t cursor_before = *cursor;
1171 int status = SUCCESS;
1172 unsigned next_char;
1173
1174 if (!(*cursor < oldlen))
1175 goto no_suitable_2nd;
1176
1177 next_char = get_next_char(charset, old, oldlen, cursor, &status);
1178
1179 if (status == FAILURE)
1180 goto no_suitable_2nd;
1181
1182 {
1183 const entity_multicodepoint_row *s, *e;
1184
1185 s = &c->data.multicodepoint_table[1];
1186 e = s - 1 + c->data.multicodepoint_table[0].leading_entry.size;
1187 /* we could do a binary search but it's not worth it since we have
1188 * at most two entries... */
1189 for ( ; s <= e; s++) {
1190 if (s->normal_entry.second_cp == next_char) {
1191 *entity = s->normal_entry.entity;
1192 *entity_len = s->normal_entry.entity_len;
1193 return;
1194 }
1195 }
1196 }
1197no_suitable_2nd:
1198 *cursor = cursor_before;
1199 *entity = (const unsigned char *)
1200 c->data.multicodepoint_table[0].leading_entry.default_entity;
1201 *entity_len = c->data.multicodepoint_table[0].leading_entry.default_entity_len;
1202 }
1203}
1204/* }}} */
1205
1206/* {{{ find_entity_for_char_basic */
1207static inline void find_entity_for_char_basic(
1208 unsigned int k,
1209 const entity_stage3_row *table,
1210 const unsigned char **entity,
1211 size_t *entity_len)
1212{
1213 if (k >= 64U) {
1214 *entity = NULL;
1215 *entity_len = 0;
1216 return;
1217 }
1218
1219 *entity = table[k].data.ent.entity;
1220 *entity_len = table[k].data.ent.entity_len;
1221}
1222/* }}} */
1223
1224/* {{{ php_escape_html_entities
1225 */
1226PHPAPI char *php_escape_html_entities_ex(unsigned char *old, size_t oldlen, size_t *newlen, int all, int flags, char *hint_charset, zend_bool double_encode TSRMLS_DC)
1227{
1228 size_t cursor, maxlen, len;
1229 char *replaced;
1230 enum entity_charset charset = determine_charset(hint_charset TSRMLS_CC);
1231 int doctype = flags & ENT_HTML_DOC_TYPE_MASK;
1232 entity_table_opt entity_table;
1233 const enc_to_uni *to_uni_table = NULL;
1234 const entity_ht *inv_map = NULL; /* used for !double_encode */
1235 /* only used if flags includes ENT_HTML_IGNORE_ERRORS or ENT_HTML_SUBSTITUTE_DISALLOWED_CHARS */
1236 const unsigned char *replacement = NULL;
1237 size_t replacement_len = 0;
1238
1239 if (all) { /* replace with all named entities */
1240 if (CHARSET_PARTIAL_SUPPORT(charset)) {
1241 php_error_docref0(NULL TSRMLS_CC, E_STRICT, "Only basic entities "
1242 "substitution is supported for multi-byte encodings other than UTF-8; "
1243 "functionality is equivalent to htmlspecialchars");
1244 }
1245 LIMIT_ALL(all, doctype, charset);
1246 }
1247 entity_table = determine_entity_table(all, doctype);
1248 if (all && !CHARSET_UNICODE_COMPAT(charset)) {
1249 to_uni_table = enc_to_uni_index[charset];
1250 }
1251
1252 if (!double_encode) {
1253 /* first arg is 1 because we want to identify valid named entities
1254 * even if we are only encoding the basic ones */
1255 inv_map = unescape_inverse_map(1, flags);
1256 }
1257
1258 if (flags & (ENT_HTML_SUBSTITUTE_ERRORS | ENT_HTML_SUBSTITUTE_DISALLOWED_CHARS)) {
1259 if (charset == cs_utf_8) {
1260 replacement = (const unsigned char*)"\xEF\xBF\xBD";
1261 replacement_len = sizeof("\xEF\xBF\xBD") - 1;
1262 } else {
1263 replacement = (const unsigned char*)"&#xFFFD;";
1264 replacement_len = sizeof("&#xFFFD;") - 1;
1265 }
1266 }
1267
1268 /* initial estimate */
1269 if (oldlen < 64) {
1270 maxlen = 128;
1271 } else {
1272 maxlen = 2 * oldlen;
1273 if (maxlen < oldlen) {
1274 zend_error_noreturn(E_ERROR, "Input string is too long");
1275 return NULL;
1276 }
1277 }
1278
1279 replaced = emalloc(maxlen + 1); /* adding 1 is safe: maxlen is even */
1280 len = 0;
1281 cursor = 0;
1282 while (cursor < oldlen) {
1283 const unsigned char *mbsequence = NULL;
1284 size_t mbseqlen = 0,
1285 cursor_before = cursor;
1286 int status = SUCCESS;
1287 unsigned int this_char = get_next_char(charset, old, oldlen, &cursor, &status);
1288
1289 /* guarantee we have at least 40 bytes to write.
1290 * In HTML5, entities may take up to 33 bytes */
1291 if (len > maxlen - 40) { /* maxlen can never be smaller than 128 */
1292 replaced = safe_erealloc(replaced, maxlen , 1, 128 + 1);
1293 maxlen += 128;
1294 }
1295
1296 if (status == FAILURE) {
1297 /* invalid MB sequence */
1298 if (flags & ENT_HTML_IGNORE_ERRORS) {
1299 continue;
1300 } else if (flags & ENT_HTML_SUBSTITUTE_ERRORS) {
1301 memcpy(&replaced[len], replacement, replacement_len);
1302 len += replacement_len;
1303 continue;
1304 } else {
1305 efree(replaced);
1306 *newlen = 0;
1307 return STR_EMPTY_ALLOC();
1308 }
1309 } else { /* SUCCESS */
1310 mbsequence = &old[cursor_before];
1311 mbseqlen = cursor - cursor_before;
1312 }
1313
1314 if (this_char != '&') { /* no entity on this position */
1315 const unsigned char *rep = NULL;
1316 size_t rep_len = 0;
1317
1318 if (((this_char == '\'' && !(flags & ENT_HTML_QUOTE_SINGLE)) ||
1319 (this_char == '"' && !(flags & ENT_HTML_QUOTE_DOUBLE))))
1320 goto pass_char_through;
1321
1322 if (all) { /* false that CHARSET_PARTIAL_SUPPORT(charset) */
1323 if (to_uni_table != NULL) {
1324 /* !CHARSET_UNICODE_COMPAT therefore not UTF-8; since UTF-8
1325 * is the only multibyte encoding with !CHARSET_PARTIAL_SUPPORT,
1326 * we're using a single byte encoding */
1327 map_to_unicode(this_char, to_uni_table, &this_char);
1328 if (this_char == 0xFFFF) /* no mapping; pass through */
1329 goto pass_char_through;
1330 }
1331 /* the cursor may advance */
1332 find_entity_for_char(this_char, charset, entity_table.ms_table, &rep,
1333 &rep_len, old, oldlen, &cursor);
1334 } else {
1335 find_entity_for_char_basic(this_char, entity_table.table, &rep, &rep_len);
1336 }
1337
1338 if (rep != NULL) {
1339 replaced[len++] = '&';
1340 memcpy(&replaced[len], rep, rep_len);
1341 len += rep_len;
1342 replaced[len++] = ';';
1343 } else {
1344 /* we did not find an entity for this char.
1345 * check for its validity, if its valid pass it unchanged */
1346 if (flags & ENT_HTML_SUBSTITUTE_DISALLOWED_CHARS) {
1347 if (CHARSET_UNICODE_COMPAT(charset)) {
1348 if (!unicode_cp_is_allowed(this_char, doctype)) {
1349 mbsequence = replacement;
1350 mbseqlen = replacement_len;
1351 }
1352 } else if (to_uni_table) {
1353 if (!all) /* otherwise we already did this */
1354 map_to_unicode(this_char, to_uni_table, &this_char);
1355 if (!unicode_cp_is_allowed(this_char, doctype)) {
1356 mbsequence = replacement;
1357 mbseqlen = replacement_len;
1358 }
1359 } else {
1360 /* not a unicode code point, unless, coincidentally, it's in
1361 * the 0x20..0x7D range (except 0x5C in sjis). We know nothing
1362 * about other code points, because we have no tables. Since
1363 * Unicode code points in that range are not disallowed in any
1364 * document type, we could do nothing. However, conversion
1365 * tables frequently map 0x00-0x1F to the respective C0 code
1366 * points. Let's play it safe and admit that's the case */
1367 if (this_char <= 0x7D &&
1368 !unicode_cp_is_allowed(this_char, doctype)) {
1369 mbsequence = replacement;
1370 mbseqlen = replacement_len;
1371 }
1372 }
1373 }
1374pass_char_through:
1375 if (mbseqlen > 1) {
1376 memcpy(replaced + len, mbsequence, mbseqlen);
1377 len += mbseqlen;
1378 } else {
1379 replaced[len++] = mbsequence[0];
1380 }
1381 }
1382 } else { /* this_char == '&' */
1383 if (double_encode) {
1384encode_amp:
1385 memcpy(&replaced[len], "&amp;", sizeof("&amp;") - 1);
1386 len += sizeof("&amp;") - 1;
1387 } else { /* no double encode */
1388 /* check if entity is valid */
1389 size_t ent_len; /* not counting & or ; */
1390 /* peek at next char */
1391 if (old[cursor] == '#') { /* numeric entity */
1392 unsigned code_point;
1393 int valid;
1394 char *pos = (char*)&old[cursor+1];
1395 valid = process_numeric_entity((const char **)&pos, &code_point);
1396 if (valid == FAILURE)
1397 goto encode_amp;
1398 if (flags & ENT_HTML_SUBSTITUTE_DISALLOWED_CHARS) {
1399 if (!numeric_entity_is_allowed(code_point, doctype))
1400 goto encode_amp;
1401 }
1402 ent_len = pos - (char*)&old[cursor];
1403 } else { /* named entity */
1404 /* check for vality of named entity */
1405 const char *start = &old[cursor],
1406 *next = start;
1407 unsigned dummy1, dummy2;
1408
1409 if (process_named_entity_html(&next, &start, &ent_len) == FAILURE)
1410 goto encode_amp;
1411 if (resolve_named_entity_html(start, ent_len, inv_map, &dummy1, &dummy2) == FAILURE) {
1412 if (!(doctype == ENT_HTML_DOC_XHTML && ent_len == 4 && start[0] == 'a'
1413 && start[1] == 'p' && start[2] == 'o' && start[3] == 's')) {
1414 /* uses html4 inv_map, which doesn't include apos;. This is a
1415 * hack to support it */
1416 goto encode_amp;
1417 }
1418 }
1419 }
1420 /* checks passed; copy entity to result */
1421 /* entity size is unbounded, we may need more memory */
1422 /* at this point maxlen - len >= 40 */
1423 if (maxlen - len < ent_len + 2 /* & and ; */) {
1424 /* ent_len < oldlen, which is certainly <= SIZE_MAX/2 */
1425 replaced = safe_erealloc(replaced, maxlen, 1, ent_len + 128 + 1);
1426 maxlen += ent_len + 128;
1427 }
1428 replaced[len++] = '&';
1429 memcpy(&replaced[len], &old[cursor], ent_len);
1430 len += ent_len;
1431 replaced[len++] = ';';
1432 cursor += ent_len + 1;
1433 }
1434 }
1435 }
1436 replaced[len] = '\0';
1437 *newlen = len;
1438
1439 return replaced;
1440}
1441/* }}} */
1442
1443/* {{{ php_html_entities
1444 */
1445static void php_html_entities(INTERNAL_FUNCTION_PARAMETERS, int all)
1446{
1447 char *str, *hint_charset = NULL;
1448 int str_len, hint_charset_len = 0;
1449 size_t new_len;
1450 long flags = ENT_COMPAT;
1451 char *replaced;
1452 zend_bool double_encode = 1;
1453
1454 if (zend_parse_parameters(ZEND_NUM_ARGS() TSRMLS_CC, "s|ls!b", &str, &str_len, &flags, &hint_charset, &hint_charset_len, &double_encode) == FAILURE) {
1455 return;
1456 }
1457
1458 if (!hint_charset) {
1459 hint_charset = get_default_charset(TSRMLS_C);
1460 }
1461 replaced = php_escape_html_entities_ex(str, str_len, &new_len, all, (int) flags, hint_charset, double_encode TSRMLS_CC);
1462
1463 RETVAL_STRINGL(replaced, (int)new_len, 0);
1464}
1465/* }}} */
1466
1467#define HTML_SPECIALCHARS 0
1468#define HTML_ENTITIES 1
1469
1470/* {{{ register_html_constants
1471 */
1472void register_html_constants(INIT_FUNC_ARGS)
1473{
1474 REGISTER_LONG_CONSTANT("HTML_SPECIALCHARS", HTML_SPECIALCHARS, CONST_PERSISTENT|CONST_CS);
1475 REGISTER_LONG_CONSTANT("HTML_ENTITIES", HTML_ENTITIES, CONST_PERSISTENT|CONST_CS);
1476 REGISTER_LONG_CONSTANT("ENT_COMPAT", ENT_COMPAT, CONST_PERSISTENT|CONST_CS);
1477 REGISTER_LONG_CONSTANT("ENT_QUOTES", ENT_QUOTES, CONST_PERSISTENT|CONST_CS);
1478 REGISTER_LONG_CONSTANT("ENT_NOQUOTES", ENT_NOQUOTES, CONST_PERSISTENT|CONST_CS);
1479 REGISTER_LONG_CONSTANT("ENT_IGNORE", ENT_IGNORE, CONST_PERSISTENT|CONST_CS);
1480 REGISTER_LONG_CONSTANT("ENT_SUBSTITUTE", ENT_SUBSTITUTE, CONST_PERSISTENT|CONST_CS);
1481 REGISTER_LONG_CONSTANT("ENT_DISALLOWED", ENT_DISALLOWED, CONST_PERSISTENT|CONST_CS);
1482 REGISTER_LONG_CONSTANT("ENT_HTML401", ENT_HTML401, CONST_PERSISTENT|CONST_CS);
1483 REGISTER_LONG_CONSTANT("ENT_XML1", ENT_XML1, CONST_PERSISTENT|CONST_CS);
1484 REGISTER_LONG_CONSTANT("ENT_XHTML", ENT_XHTML, CONST_PERSISTENT|CONST_CS);
1485 REGISTER_LONG_CONSTANT("ENT_HTML5", ENT_HTML5, CONST_PERSISTENT|CONST_CS);
1486}
1487/* }}} */
1488
1489/* {{{ proto string htmlspecialchars(string string [, int quote_style[, string charset[, bool double_encode]]])
1490 Convert special characters to HTML entities */
1491PHP_FUNCTION(htmlspecialchars)
1492{
1493 php_html_entities(INTERNAL_FUNCTION_PARAM_PASSTHRU, 0);
1494}
1495/* }}} */
1496
1497/* {{{ proto string htmlspecialchars_decode(string string [, int quote_style])
1498 Convert special HTML entities back to characters */
1499PHP_FUNCTION(htmlspecialchars_decode)
1500{
1501 char *str;
1502 int str_len;
1503 size_t new_len = 0;
1504 long quote_style = ENT_COMPAT;
1505 char *replaced;
1506
1507 if (zend_parse_parameters(ZEND_NUM_ARGS() TSRMLS_CC, "s|l", &str, &str_len, &quote_style) == FAILURE) {
1508 return;
1509 }
1510
1511 replaced = php_unescape_html_entities(str, str_len, &new_len, 0 /*!all*/, quote_style, NULL TSRMLS_CC);
1512 if (replaced) {
1513 RETURN_STRINGL(replaced, (int)new_len, 0);
1514 }
1515 RETURN_FALSE;
1516}
1517/* }}} */
1518
1519/* {{{ proto string html_entity_decode(string string [, int quote_style][, string charset])
1520 Convert all HTML entities to their applicable characters */
1521PHP_FUNCTION(html_entity_decode)
1522{
1523 char *str, *hint_charset = NULL;
1524 int str_len, hint_charset_len;
1525 size_t new_len = 0;
1526 long quote_style = ENT_COMPAT;
1527 char *replaced;
1528
1529 if (zend_parse_parameters(ZEND_NUM_ARGS() TSRMLS_CC, "s|ls", &str, &str_len,
1530 &quote_style, &hint_charset, &hint_charset_len) == FAILURE) {
1531 return;
1532 }
1533
1534 if (!hint_charset) {
1535 hint_charset = get_default_charset(TSRMLS_C);
1536 }
1537 replaced = php_unescape_html_entities(str, str_len, &new_len, 1 /*all*/, quote_style, hint_charset TSRMLS_CC);
1538
1539 if (replaced) {
1540 RETURN_STRINGL(replaced, (int)new_len, 0);
1541 }
1542 RETURN_FALSE;
1543}
1544/* }}} */
1545
1546
1547/* {{{ proto string htmlentities(string string [, int quote_style[, string charset[, bool double_encode]]])
1548 Convert all applicable characters to HTML entities */
1549PHP_FUNCTION(htmlentities)
1550{
1551 php_html_entities(INTERNAL_FUNCTION_PARAM_PASSTHRU, 1);
1552}
1553/* }}} */
1554
1555/* {{{ write_s3row_data */
1556static inline void write_s3row_data(
1557 const entity_stage3_row *r,
1558 unsigned orig_cp,
1559 enum entity_charset charset,
1560 zval *arr)
1561{
1562 char key[9] = ""; /* two unicode code points in UTF-8 */
1563 char entity[LONGEST_ENTITY_LENGTH + 2] = {'&'};
1564 size_t written_k1;
1565
1566 written_k1 = write_octet_sequence(key, charset, orig_cp);
1567
1568 if (!r->ambiguous) {
1569 size_t l = r->data.ent.entity_len;
1570 memcpy(&entity[1], r->data.ent.entity, l);
1571 entity[l + 1] = ';';
1572 add_assoc_stringl_ex(arr, key, written_k1 + 1, entity, l + 2, 1);
1573 } else {
1574 unsigned i,
1575 num_entries;
1576 const entity_multicodepoint_row *mcpr = r->data.multicodepoint_table;
1577
1578 if (mcpr[0].leading_entry.default_entity != NULL) {
1579 size_t l = mcpr[0].leading_entry.default_entity_len;
1580 memcpy(&entity[1], mcpr[0].leading_entry.default_entity, l);
1581 entity[l + 1] = ';';
1582 add_assoc_stringl_ex(arr, key, written_k1 + 1, entity, l + 2, 1);
1583 }
1584 num_entries = mcpr[0].leading_entry.size;
1585 for (i = 1; i <= num_entries; i++) {
1586 size_t l,
1587 written_k2;
1588 unsigned uni_cp,
1589 spe_cp;
1590
1591 uni_cp = mcpr[i].normal_entry.second_cp;
1592 l = mcpr[i].normal_entry.entity_len;
1593
1594 if (!CHARSET_UNICODE_COMPAT(charset)) {
1595 if (map_from_unicode(uni_cp, charset, &spe_cp) == FAILURE)
1596 continue; /* non representable in this charset */
1597 } else {
1598 spe_cp = uni_cp;
1599 }
1600
1601 written_k2 = write_octet_sequence(&key[written_k1], charset, spe_cp);
1602 memcpy(&entity[1], mcpr[i].normal_entry.entity, l);
1603 entity[l + 1] = ';';
1604 entity[l + 1] = '\0';
1605 add_assoc_stringl_ex(arr, key, written_k1 + written_k2 + 1, entity, l + 1, 1);
1606 }
1607 }
1608}
1609/* }}} */
1610
1611/* {{{ proto array get_html_translation_table([int table [, int flags [, string charset_hint]]])
1612 Returns the internal translation table used by htmlspecialchars and htmlentities */
1613PHP_FUNCTION(get_html_translation_table)
1614{
1615 long all = HTML_SPECIALCHARS,
1616 flags = ENT_COMPAT;
1617 int doctype;
1618 entity_table_opt entity_table;
1619 const enc_to_uni *to_uni_table = NULL;
1620 char *charset_hint = NULL;
1621 int charset_hint_len;
1622 enum entity_charset charset;
1623
1624 /* in this function we have to jump through some loops because we're
1625 * getting the translated table from data structures that are optimized for
1626 * random access, not traversal */
1627
1628 if (zend_parse_parameters(ZEND_NUM_ARGS() TSRMLS_CC, "|lls",
1629 &all, &flags, &charset_hint, &charset_hint_len) == FAILURE) {
1630 return;
1631 }
1632
1633 charset = determine_charset(charset_hint TSRMLS_CC);
1634 doctype = flags & ENT_HTML_DOC_TYPE_MASK;
1635 LIMIT_ALL(all, doctype, charset);
1636
1637 array_init(return_value);
1638
1639 entity_table = determine_entity_table(all, doctype);
1640 if (all && !CHARSET_UNICODE_COMPAT(charset)) {
1641 to_uni_table = enc_to_uni_index[charset];
1642 }
1643
1644 if (all) { /* HTML_ENTITIES (actually, any non-zero value for 1st param) */
1645 const entity_stage1_row *ms_table = entity_table.ms_table;
1646
1647 if (CHARSET_UNICODE_COMPAT(charset)) {
1648 unsigned i, j, k,
1649 max_i, max_j, max_k;
1650 /* no mapping to unicode required */
1651 if (CHARSET_SINGLE_BYTE(charset)) { /* ISO-8859-1 */
1652 max_i = 1; max_j = 4; max_k = 64;
1653 } else {
1654 max_i = 0x1E; max_j = 64; max_k = 64;
1655 }
1656
1657 for (i = 0; i < max_i; i++) {
1658 if (ms_table[i] == empty_stage2_table)
1659 continue;
1660 for (j = 0; j < max_j; j++) {
1661 if (ms_table[i][j] == empty_stage3_table)
1662 continue;
1663 for (k = 0; k < max_k; k++) {
1664 const entity_stage3_row *r = &ms_table[i][j][k];
1665 unsigned code;
1666
1667 if (r->data.ent.entity == NULL)
1668 continue;
1669
1670 code = ENT_CODE_POINT_FROM_STAGES(i, j, k);
1671 if (((code == '\'' && !(flags & ENT_HTML_QUOTE_SINGLE)) ||
1672 (code == '"' && !(flags & ENT_HTML_QUOTE_DOUBLE))))
1673 continue;
1674 write_s3row_data(r, code, charset, return_value);
1675 }
1676 }
1677 }
1678 } else {
1679 /* we have to iterate through the set of code points for this
1680 * encoding and map them to unicode code points */
1681 unsigned i;
1682 for (i = 0; i <= 0xFF; i++) {
1683 const entity_stage3_row *r;
1684 unsigned uni_cp;
1685
1686 /* can be done before mapping, they're invariant */
1687 if (((i == '\'' && !(flags & ENT_HTML_QUOTE_SINGLE)) ||
1688 (i == '"' && !(flags & ENT_HTML_QUOTE_DOUBLE))))
1689 continue;
1690
1691 map_to_unicode(i, to_uni_table, &uni_cp);
1692 r = &ms_table[ENT_STAGE1_INDEX(uni_cp)][ENT_STAGE2_INDEX(uni_cp)][ENT_STAGE3_INDEX(uni_cp)];
1693 if (r->data.ent.entity == NULL)
1694 continue;
1695
1696 write_s3row_data(r, i, charset, return_value);
1697 }
1698 }
1699 } else {
1700 /* we could use sizeof(stage3_table_be_apos_00000) as well */
1701 unsigned j,
1702 numelems = sizeof(stage3_table_be_noapos_00000) /
1703 sizeof(*stage3_table_be_noapos_00000);
1704
1705 for (j = 0; j < numelems; j++) {
1706 const entity_stage3_row *r = &entity_table.table[j];
1707 if (r->data.ent.entity == NULL)
1708 continue;
1709
1710 if (((j == '\'' && !(flags & ENT_HTML_QUOTE_SINGLE)) ||
1711 (j == '"' && !(flags & ENT_HTML_QUOTE_DOUBLE))))
1712 continue;
1713
1714 /* charset is indifferent, used cs_8859_1 for efficiency */
1715 write_s3row_data(r, j, cs_8859_1, return_value);
1716 }
1717 }
1718}
1719/* }}} */
1720
1721/*
1722 * Local variables:
1723 * tab-width: 4
1724 * c-basic-offset: 4
1725 * End:
1726 * vim600: sw=4 ts=4 fdm=marker
1727 * vim<600: sw=4 ts=4
1728 */
1729