6 * Copyright (C) 2012, 2013 Eric Biggers
8 * This file is part of wimlib, a library for working with WIM files.
10 * wimlib is free software; you can redistribute it and/or modify it under the
11 * terms of the GNU General Public License as published by the Free
12 * Software Foundation; either version 3 of the License, or (at your option)
15 * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
16 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
17 * A PARTICULAR PURPOSE. See the GNU General Public License for more
20 * You should have received a copy of the GNU General Public License
21 * along with wimlib; if not, see http://www.gnu.org/licenses/.
29 #include "wimlib/assert.h"
30 #include "wimlib/encoding.h"
31 #include "wimlib/endianness.h"
32 #include "wimlib/error.h"
33 #include "wimlib/list.h"
34 #include "wimlib/util.h"
46 bool wimlib_mbs_is_utf8 = !TCHAR_IS_UTF16LE;
48 /* List of iconv_t conversion descriptors for a specific character conversion.
49 * The idea is that it is not thread-safe to have just one conversion
50 * descriptor, but it also is inefficient to open a new conversion descriptor to
51 * convert every string. Both these problems can be solved by maintaining a
52 * list of conversion descriptors; then, a thread can use an existing conversion
53 * descriptor if available. */
54 struct iconv_list_head {
55 const char *from_encoding;
56 const char *to_encoding;
57 struct list_head list;
58 pthread_mutex_t mutex;
63 struct list_head list;
64 struct iconv_list_head *head;
67 #define ICONV_LIST(name, from, to) \
68 struct iconv_list_head name = { \
69 .from_encoding = from, \
71 .list = LIST_HEAD_INIT(name.list), \
72 .mutex = PTHREAD_MUTEX_INITIALIZER, \
76 get_iconv(struct iconv_list_head *head)
82 pthread_mutex_lock(&head->mutex);
83 if (list_empty(&head->list)) {
84 cd = iconv_open(head->to_encoding, head->from_encoding);
85 if (cd == (iconv_t)-1) {
86 ERROR_WITH_ERRNO("Failed to open iconv from %s to %s",
87 head->from_encoding, head->to_encoding);
90 i = MALLOC(sizeof(struct iconv_node));
101 i = container_of(head->list.next, struct iconv_node, list);
102 list_del(head->list.next);
105 pthread_mutex_unlock(&head->mutex);
110 put_iconv(iconv_t *cd)
112 int errno_save = errno;
113 struct iconv_node *i = container_of(cd, struct iconv_node, cd);
114 struct iconv_list_head *head = i->head;
116 pthread_mutex_lock(&head->mutex);
117 list_add(&i->list, &head->list);
118 pthread_mutex_unlock(&head->mutex);
122 #define DEFINE_CHAR_CONVERSION_FUNCTIONS(varname1, longname1, chartype1,\
123 varname2, longname2, chartype2,\
124 earlyreturn_on_utf8_locale, \
126 worst_case_len_expr, \
130 static ICONV_LIST(iconv_##varname1##_to_##varname2, \
131 longname1, longname2); \
134 varname1##_to_##varname2##_nbytes(const chartype1 *in, size_t in_nbytes,\
135 size_t *out_nbytes_ret) \
137 iconv_t *cd = get_iconv(&iconv_##varname1##_to_##varname2); \
139 return WIMLIB_ERR_ICONV_NOT_AVAILABLE; \
144 bufsize = (worst_case_len_expr) * sizeof(chartype2); \
145 /* Worst case length */ \
146 if (bufsize <= STACK_MAX) { \
147 buf = alloca(bufsize); \
148 buf_onheap = false; \
150 buf = MALLOC(bufsize); \
152 return WIMLIB_ERR_NOMEM; \
156 char *inbuf = (char*)in; \
157 size_t inbytesleft = in_nbytes; \
158 char *outbuf = (char*)buf; \
159 size_t outbytesleft = bufsize; \
163 len = iconv(*cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft); \
164 if (len == (size_t)-1) { \
168 *out_nbytes_ret = bufsize - outbytesleft; \
178 varname1##_to_##varname2##_buf(const chartype1 *in, size_t in_nbytes, \
181 iconv_t *cd = get_iconv(&iconv_##varname1##_to_##varname2); \
183 return WIMLIB_ERR_ICONV_NOT_AVAILABLE; \
185 char *inbuf = (char*)in; \
186 size_t inbytesleft = in_nbytes; \
187 char *outbuf = (char*)out; \
188 const size_t LARGE_NUMBER = 1000000000; \
189 size_t outbytesleft = LARGE_NUMBER; \
193 len = iconv(*cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft); \
194 if (len == (size_t)-1) { \
198 out[(LARGE_NUMBER-outbytesleft)/sizeof(chartype2)] = 0; \
206 varname1##_to_##varname2(const chartype1 *in, size_t in_nbytes, \
207 chartype2 **out_ret, \
208 size_t *out_nbytes_ret) \
214 if (earlyreturn_on_utf8_locale && wimlib_mbs_is_utf8) { \
216 /* Out same as in */ \
217 out = MALLOC(in_nbytes + sizeof(chartype2)); \
219 return WIMLIB_ERR_NOMEM; \
220 memcpy(out, in, in_nbytes); \
221 out[in_nbytes / sizeof(chartype2)] = 0; \
223 *out_nbytes_ret = in_nbytes; \
227 ret = varname1##_to_##varname2##_nbytes(in, in_nbytes, \
232 out = MALLOC(out_nbytes + sizeof(chartype2)); \
234 return WIMLIB_ERR_NOMEM; \
236 ret = varname1##_to_##varname2##_buf(in, in_nbytes, out); \
241 *out_nbytes_ret = out_nbytes; \
246 #if !TCHAR_IS_UTF16LE
250 DEFINE_CHAR_CONVERSION_FUNCTIONS(utf8, "UTF-8", tchar,
251 utf16le, "UTF-16LE", utf16lechar,
255 WIMLIB_ERR_INVALID_UTF8_STRING,
256 ERROR_WITH_ERRNO("Failed to convert UTF-8 string "
257 "to UTF-16LE string!"),
260 DEFINE_CHAR_CONVERSION_FUNCTIONS(utf16le, "UTF-16LE", utf16lechar,
261 utf8, "UTF-8", tchar,
265 WIMLIB_ERR_INVALID_UTF16_STRING,
266 ERROR_WITH_ERRNO("Failed to convert UTF-16LE string "
270 DEFINE_CHAR_CONVERSION_FUNCTIONS(tstr, "", tchar,
271 utf16le, "UTF-16LE", utf16lechar,
273 return utf8_to_utf16le(in, in_nbytes, out_ret, out_nbytes_ret),
275 WIMLIB_ERR_INVALID_MULTIBYTE_STRING,
276 ERROR_WITH_ERRNO("Failed to convert multibyte "
277 "string \"%"TS"\" to UTF-16LE string!", in);
278 ERROR("If the data you provided was UTF-8, please make sure "
279 "the character encoding\n"
280 " of your current locale is UTF-8."),
283 DEFINE_CHAR_CONVERSION_FUNCTIONS(utf16le, "UTF-16LE", utf16lechar,
286 return utf16le_to_utf8(in, in_nbytes, out_ret, out_nbytes_ret),
288 WIMLIB_ERR_UNICODE_STRING_NOT_REPRESENTABLE,
289 ERROR("Failed to convert UTF-16LE string to "
290 "multibyte string!");
291 ERROR("This may be because the UTF-16LE string "
292 "could not be represented\n"
293 " in your locale's character encoding."),
297 /* tchar to UTF-8 and back */
301 DEFINE_CHAR_CONVERSION_FUNCTIONS(tstr, "UTF-16LE", tchar,
306 WIMLIB_ERR_INVALID_UTF16_STRING,
307 ERROR_WITH_ERRNO("Failed to convert UTF-16LE "
308 "string \"%"TS"\" to UTF-8 string!", in),
311 DEFINE_CHAR_CONVERSION_FUNCTIONS(utf8, "UTF-8", char,
312 tstr, "UTF-16LE", tchar,
316 WIMLIB_ERR_INVALID_UTF8_STRING,
317 ERROR_WITH_ERRNO("Failed to convert UTF-8 string "
318 "to UTF-16LE string!"),
324 DEFINE_CHAR_CONVERSION_FUNCTIONS(tstr, "", tchar,
329 WIMLIB_ERR_INVALID_MULTIBYTE_STRING,
330 ERROR_WITH_ERRNO("Failed to convert multibyte "
331 "string \"%"TS"\" to UTF-8 string!", in);
332 ERROR("If the data you provided was UTF-8, please make sure "
334 " encoding of your current locale is UTF-8."),
337 DEFINE_CHAR_CONVERSION_FUNCTIONS(utf8, "UTF-8", char,
342 WIMLIB_ERR_UNICODE_STRING_NOT_REPRESENTABLE,
343 ERROR("Failed to convert UTF-8 string to "
344 "multibyte string!");
345 ERROR("This may be because the UTF-8 data "
346 "could not be represented\n"
347 " in your locale's character encoding."),
352 tstr_to_utf8_simple(const tchar *tstr, char **out)
355 return tstr_to_utf8(tstr, tstrlen(tstr) * sizeof(tchar),
360 utf8_to_tstr_simple(const char *utf8str, tchar **out)
363 return utf8_to_tstr(utf8str, strlen(utf8str), out, &out_nbytes);
367 iconv_cleanup(struct iconv_list_head *head)
369 pthread_mutex_destroy(&head->mutex);
370 while (!list_empty(&head->list)) {
371 struct iconv_node *i;
373 i = container_of(head->list.next, struct iconv_node, list);
381 iconv_global_cleanup(void)
383 iconv_cleanup(&iconv_utf8_to_tstr);
384 iconv_cleanup(&iconv_tstr_to_utf8);
385 #if !TCHAR_IS_UTF16LE
386 iconv_cleanup(&iconv_utf16le_to_tstr);
387 iconv_cleanup(&iconv_tstr_to_utf16le);
388 iconv_cleanup(&iconv_utf16le_to_utf8);
389 iconv_cleanup(&iconv_utf8_to_utf16le);
393 /* A table that maps from UCS-2 characters to their upper case equivalents.
394 * Index and array values are both CPU endian.
395 * Note: this is only an *approximation* of real UTF-16 case folding.
397 static u16 upcase[65536];
400 * Initialize the 'upcase' table.
402 * Code modified from NTFS-3g, original copyright notices below:
404 * Copyright (c) 2000-2004 Anton Altaparmakov
405 * Copyright (c) 2002-2009 Szabolcs Szakacsits
406 * Copyright (c) 2008-2011 Jean-Pierre Andre
407 * Copyright (c) 2008 Bernhard Kaindl
409 * License GPLv2 or later.
411 * The expanded table itself is the same as that used by Windows Vista for
412 * NTFS case insensitivity.
418 * This is the table as defined by Vista
421 * "Start" is inclusive and "End" is exclusive, every value has the
422 * value of "Add" added to it.
424 static const u16 uc_run_table[][3] = { /* Start, End, Add */
425 {0x0061, 0x007b, -32}, {0x00e0, 0x00f7, -32}, {0x00f8, 0x00ff, -32},
426 {0x0256, 0x0258, -205}, {0x028a, 0x028c, -217}, {0x037b, 0x037e, 130},
427 {0x03ac, 0x03ad, -38}, {0x03ad, 0x03b0, -37}, {0x03b1, 0x03c2, -32},
428 {0x03c2, 0x03c3, -31}, {0x03c3, 0x03cc, -32}, {0x03cc, 0x03cd, -64},
429 {0x03cd, 0x03cf, -63}, {0x0430, 0x0450, -32}, {0x0450, 0x0460, -80},
430 {0x0561, 0x0587, -48}, {0x1f00, 0x1f08, 8}, {0x1f10, 0x1f16, 8},
431 {0x1f20, 0x1f28, 8}, {0x1f30, 0x1f38, 8}, {0x1f40, 0x1f46, 8},
432 {0x1f51, 0x1f52, 8}, {0x1f53, 0x1f54, 8}, {0x1f55, 0x1f56, 8},
433 {0x1f57, 0x1f58, 8}, {0x1f60, 0x1f68, 8}, {0x1f70, 0x1f72, 74},
434 {0x1f72, 0x1f76, 86}, {0x1f76, 0x1f78, 100}, {0x1f78, 0x1f7a, 128},
435 {0x1f7a, 0x1f7c, 112}, {0x1f7c, 0x1f7e, 126}, {0x1f80, 0x1f88, 8},
436 {0x1f90, 0x1f98, 8}, {0x1fa0, 0x1fa8, 8}, {0x1fb0, 0x1fb2, 8},
437 {0x1fb3, 0x1fb4, 9}, {0x1fcc, 0x1fcd, -9}, {0x1fd0, 0x1fd2, 8},
438 {0x1fe0, 0x1fe2, 8}, {0x1fe5, 0x1fe6, 7}, {0x1ffc, 0x1ffd, -9},
439 {0x2170, 0x2180, -16}, {0x24d0, 0x24ea, -26}, {0x2c30, 0x2c5f, -48},
440 {0x2d00, 0x2d26, -7264}, {0xff41, 0xff5b, -32},
443 * "Start" is exclusive and "End" is inclusive, every second value is
444 * decremented by one.
446 static const u16 uc_dup_table[][2] = { /* Start, End */
447 {0x0100, 0x012f}, {0x0132, 0x0137}, {0x0139, 0x0149}, {0x014a, 0x0178},
448 {0x0179, 0x017e}, {0x01a0, 0x01a6}, {0x01b3, 0x01b7}, {0x01cd, 0x01dd},
449 {0x01de, 0x01ef}, {0x01f4, 0x01f5}, {0x01f8, 0x01f9}, {0x01fa, 0x0220},
450 {0x0222, 0x0234}, {0x023b, 0x023c}, {0x0241, 0x0242}, {0x0246, 0x024f},
451 {0x03d8, 0x03ef}, {0x03f7, 0x03f8}, {0x03fa, 0x03fb}, {0x0460, 0x0481},
452 {0x048a, 0x04bf}, {0x04c1, 0x04c4}, {0x04c5, 0x04c8}, {0x04c9, 0x04ce},
453 {0x04ec, 0x04ed}, {0x04d0, 0x04eb}, {0x04ee, 0x04f5}, {0x04f6, 0x0513},
454 {0x1e00, 0x1e95}, {0x1ea0, 0x1ef9}, {0x2183, 0x2184}, {0x2c60, 0x2c61},
455 {0x2c67, 0x2c6c}, {0x2c75, 0x2c76}, {0x2c80, 0x2ce3},
458 * Set the Unicode character at offset "Offset" to "Value". Note,
459 * "Value" is host endian.
461 static const u16 uc_byte_table[][2] = { /* Offset, Value */
462 {0x00ff, 0x0178}, {0x0180, 0x0243}, {0x0183, 0x0182}, {0x0185, 0x0184},
463 {0x0188, 0x0187}, {0x018c, 0x018b}, {0x0192, 0x0191}, {0x0195, 0x01f6},
464 {0x0199, 0x0198}, {0x019a, 0x023d}, {0x019e, 0x0220}, {0x01a8, 0x01a7},
465 {0x01ad, 0x01ac}, {0x01b0, 0x01af}, {0x01b9, 0x01b8}, {0x01bd, 0x01bc},
466 {0x01bf, 0x01f7}, {0x01c6, 0x01c4}, {0x01c9, 0x01c7}, {0x01cc, 0x01ca},
467 {0x01dd, 0x018e}, {0x01f3, 0x01f1}, {0x023a, 0x2c65}, {0x023e, 0x2c66},
468 {0x0253, 0x0181}, {0x0254, 0x0186}, {0x0259, 0x018f}, {0x025b, 0x0190},
469 {0x0260, 0x0193}, {0x0263, 0x0194}, {0x0268, 0x0197}, {0x0269, 0x0196},
470 {0x026b, 0x2c62}, {0x026f, 0x019c}, {0x0272, 0x019d}, {0x0275, 0x019f},
471 {0x027d, 0x2c64}, {0x0280, 0x01a6}, {0x0283, 0x01a9}, {0x0288, 0x01ae},
472 {0x0289, 0x0244}, {0x028c, 0x0245}, {0x0292, 0x01b7}, {0x03f2, 0x03f9},
473 {0x04cf, 0x04c0}, {0x1d7d, 0x2c63}, {0x214e, 0x2132},
476 for (u32 i = 0; i < 65536; i++)
479 for (u32 r = 0; r < ARRAY_LEN(uc_run_table); r++)
480 for (u32 i = uc_run_table[r][0]; i < uc_run_table[r][1]; i++)
481 upcase[i] = i + uc_run_table[r][2];
483 for (u32 r = 0; r < ARRAY_LEN(uc_dup_table); r++)
484 for (u32 i = uc_dup_table[r][0]; i < uc_dup_table[r][1]; i += 2)
487 for (u32 r = 0; r < ARRAY_LEN(uc_byte_table); r++)
488 upcase[uc_byte_table[r][0]] = uc_byte_table[r][1];
492 wimlib_assert(upcase['a'] == 'A');
493 wimlib_assert(upcase['A'] == 'A');
494 wimlib_assert(upcase['z'] == 'Z');
495 wimlib_assert(upcase['Z'] == 'Z');
496 wimlib_assert(upcase['1'] == '1');
497 wimlib_assert(upcase[0x00e9] == 0x00c9); /* Latin letter e, with acute accent */
498 wimlib_assert(upcase[0x00c9] == 0x00c9);
499 wimlib_assert(upcase[0x03c1] == 0x03a1); /* Greek letter rho */
500 wimlib_assert(upcase[0x03a1] == 0x03a1);
501 wimlib_assert(upcase[0x0436] == 0x0416); /* Cyrillic letter zhe */
502 wimlib_assert(upcase[0x0416] == 0x0416);
503 wimlib_assert(upcase[0x0567] == 0x0537); /* Armenian letter eh */
504 wimlib_assert(upcase[0x0537] == 0x0537);
505 wimlib_assert(upcase[0x24d0] == 0x24b6); /* Circled Latin letter A
506 (is that a real character???) */
507 wimlib_assert(upcase[0x24b6] == 0x24b6);
508 wimlib_assert(upcase[0x2603] == 0x2603); /* Note to self: Upper case
509 snowman symbol does not
514 /* Compare UTF-16LE strings case-sensitively (%ignore_case == false) or
515 * case-insensitively (%ignore_case == true).
517 * This is implemented using the default upper-case table used by NTFS. It does
518 * not handle all possible cases allowed by UTF-16LE. For example, different
519 * normalizations of the same sequence of "characters" are not considered equal.
520 * It hopefully does the right thing most of the time though. */
522 cmp_utf16le_strings(const utf16lechar *s1, size_t n1,
523 const utf16lechar *s2, size_t n2,
526 size_t n = min(n1, n2);
529 for (size_t i = 0; i < n; i++) {
530 u16 c1 = upcase[le16_to_cpu(s1[i])];
531 u16 c2 = upcase[le16_to_cpu(s2[i])];
533 return (c1 < c2) ? -1 : 1;
536 for (size_t i = 0; i < n; i++) {
537 u16 c1 = le16_to_cpu(s1[i]);
538 u16 c2 = le16_to_cpu(s2[i]);
540 return (c1 < c2) ? -1 : 1;
545 return (n1 < n2) ? -1 : 1;
548 /* Duplicate a UTF16-LE string which may not be null-terminated. */
550 utf16le_dupz(const utf16lechar *ustr, size_t usize)
552 utf16lechar *dup = MALLOC(usize + sizeof(utf16lechar));
554 memcpy(dup, ustr, usize);
555 dup[usize / sizeof(utf16lechar)] = 0;