2 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
3 * Copyright 2012 Garrett D'Amore <garrett@damore.org> All rights reserved.
4 * Copyright 2015 John Marino <draco@marino.st>
6 * This source code is derived from the illumos localedef command, and
7 * provided under BSD-style license terms by Nexenta Systems, Inc.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
19 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
20 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
23 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29 * POSSIBILITY OF SUCH DAMAGE.
33 * The functions in this file convert from the standard multibyte forms
34 * to the wide character forms used internally by libc. Unfortunately,
35 * this approach means that we need a method for each and every encoding.
41 #include <sys/types.h>
42 #include "localedef.h"
44 static int towide_none(wchar_t *, const char *, unsigned);
45 static int towide_utf8(wchar_t *, const char *, unsigned);
46 static int towide_big5(wchar_t *, const char *, unsigned);
47 static int towide_gbk(wchar_t *, const char *, unsigned);
48 static int towide_gb2312(wchar_t *, const char *, unsigned);
49 static int towide_gb18030(wchar_t *, const char *, unsigned);
50 static int towide_mskanji(wchar_t *, const char *, unsigned);
51 static int towide_euccn(wchar_t *, const char *, unsigned);
52 static int towide_eucjp(wchar_t *, const char *, unsigned);
53 static int towide_euckr(wchar_t *, const char *, unsigned);
54 static int towide_euctw(wchar_t *, const char *, unsigned);
56 static int tomb_none(char *, wchar_t);
57 static int tomb_utf8(char *, wchar_t);
58 static int tomb_mbs(char *, wchar_t);
60 static int (*_towide)(wchar_t *, const char *, unsigned) = towide_none;
61 static int (*_tomb)(char *, wchar_t) = tomb_none;
62 static const char *_encoding = "NONE";
63 static int _nbits = 7;
66 * Table of supported encodings. We only bother to list the multibyte
67 * encodings here, because single byte locales are handed by "NONE".
71 /* the name that the underlying libc implemenation uses */
73 /* the maximum number of bits required for priorities */
75 int (*towide)(wchar_t *, const char *, unsigned);
76 int (*tomb)(char *, wchar_t);
79 * UTF8 values max out at 0x1fffff (although in theory there could
80 * be later extensions, but it won't happen.) This means we only need
81 * 21 bits to be able to encode the entire range of priorities.
83 { "UTF-8", "UTF-8", 21, towide_utf8, tomb_utf8 },
84 { "UTF8", "UTF-8", 21, towide_utf8, tomb_utf8 },
85 { "utf8", "UTF-8", 21, towide_utf8, tomb_utf8 },
86 { "utf-8", "UTF-8", 21, towide_utf8, tomb_utf8 },
88 { "EUC-CN", "EUC-CN", 16, towide_euccn, tomb_mbs },
89 { "eucCN", "EUC-CN", 16, towide_euccn, tomb_mbs },
91 * Becuase the 3-byte form of EUC-JP use the same leading byte,
92 * only 17 bits required to provide unique priorities. (The low
93 * bit of that first byte is set.) By setting this value low,
94 * we can get by with only 3 bytes in the strxfrm expansion.
96 { "EUC-JP", "EUC-JP", 17, towide_eucjp, tomb_mbs },
97 { "eucJP", "EUC-JP", 17, towide_eucjp, tomb_mbs },
99 { "EUC-KR", "EUC-KR", 16, towide_euckr, tomb_mbs },
100 { "eucKR", "EUC-KR", 16, towide_euckr, tomb_mbs },
102 * EUC-TW uses 2 bytes most of the time, but 4 bytes if the
103 * high order byte is 0x8E. However, with 4 byte encodings,
104 * the third byte will be A0-B0. So we only need to consider
105 * the lower order 24 bits for collation.
107 { "EUC-TW", "EUC-TW", 24, towide_euctw, tomb_mbs },
108 { "eucTW", "EUC-TW", 24, towide_euctw, tomb_mbs },
110 { "MS_Kanji", "MSKanji", 16, towide_mskanji, tomb_mbs },
111 { "MSKanji", "MSKanji", 16, towide_mskanji, tomb_mbs },
112 { "PCK", "MSKanji", 16, towide_mskanji, tomb_mbs },
113 { "SJIS", "MSKanji", 16, towide_mskanji, tomb_mbs },
114 { "Shift_JIS", "MSKanji", 16, towide_mskanji, tomb_mbs },
116 { "BIG5", "BIG5", 16, towide_big5, tomb_mbs },
117 { "big5", "BIG5", 16, towide_big5, tomb_mbs },
118 { "Big5", "BIG5", 16, towide_big5, tomb_mbs },
120 { "GBK", "GBK", 16, towide_gbk, tomb_mbs },
123 * GB18030 can get away with just 31 bits. This is because the
124 * high order bit is always set for 4 byte values, and the
125 * at least one of the other bits in that 4 byte value will
128 { "GB18030", "GB18030", 31, towide_gb18030, tomb_mbs },
131 * This should probably be an aliase for euc-cn, or vice versa.
133 { "GB2312", "GB2312", 16, towide_gb2312, tomb_mbs },
135 { NULL, NULL, 0, 0, 0 },
139 show_mb(const char *mb)
143 /* ASCII stuff we just print */
144 if (isascii(*mb) && isgraph(*mb)) {
152 (void) snprintf(scr, sizeof (scr), "\\x%02x", *mb);
153 (void) strlcat(buf, scr, sizeof (buf));
159 static char *widemsg;
162 werr(const char *fmt, ...)
168 (void) vasprintf(&msg, fmt, va);
176 * This is used for 8-bit encodings.
179 towide_none(wchar_t *c, const char *mb, unsigned n __unused)
181 if (mb_cur_max != 1) {
182 werr("invalid or unsupported multibyte locale");
190 tomb_none(char *mb, wchar_t wc)
192 if (mb_cur_max != 1) {
193 werr("invalid or unsupported multibyte locale");
196 *(uint8_t *)mb = (wc & 0xff);
202 * UTF-8 stores wide characters in UTF-32 form.
205 towide_utf8(wchar_t *wc, const char *mb, unsigned n)
209 int lv; /* lowest legal value */
211 const uint8_t *s = (const uint8_t *)mb;
215 if ((c & 0x80) == 0) {
219 } else if ((c & 0xe0) == 0xc0) {
220 /* u80-u7ff - two bytes encoded */
224 } else if ((c & 0xf0) == 0xe0) {
225 /* u800-uffff - three bytes encoded */
229 } else if ((c & 0xf8) == 0xf0) {
230 /* u1000-u1fffff - four bytes encoded */
235 /* 5 and 6 byte encodings are not legal unicode */
236 werr("utf8 encoding too large (%s)", show_mb(mb));
240 werr("incomplete utf8 sequence (%s)", show_mb(mb));
244 for (i = 1; i < nb; i++) {
245 if (((s[i]) & 0xc0) != 0x80) {
246 werr("illegal utf8 byte (%x)", s[i]);
254 werr("illegal redundant utf8 encoding (%s)", show_mb(mb));
262 tomb_utf8(char *mb, wchar_t wc)
264 uint8_t *s = (uint8_t *)mb;
277 } else if (wc <= 0xffff) {
280 } else if (wc <= 0x1fffff) {
284 werr("illegal uf8 char (%x)", wc);
287 for (i = cnt - 1; i; i--) {
288 s[i] = (wc & 0x3f) | 0x80;
297 * Several encodings share a simplistic dual byte encoding. In these
298 * forms, they all indicate that a two byte sequence is to be used if
299 * the first byte has its high bit set. They all store this simple
300 * encoding as a 16-bit value, although a great many of the possible
301 * code points are not used in most character sets. This gives a possible
302 * set of just over 32,000 valid code points.
304 * 0x00 - 0x7f - 1 byte encoding
305 * 0x80 - 0x7fff - illegal
306 * 0x8000 - 0xffff - 2 byte encoding
309 #pragma GCC diagnostic push
310 #pragma GCC diagnostic ignored "-Wcast-qual"
313 towide_dbcs(wchar_t *wc, const char *mb, unsigned n)
319 if ((c & 0x80) == 0) {
325 werr("incomplete character sequence (%s)", show_mb(mb));
329 /* Store both bytes as a single 16-bit wide. */
331 c |= (uint8_t)(mb[1]);
337 * Most multibyte locales just convert the wide character to the multibyte
338 * form by stripping leading null bytes, and writing the 32-bit quantity
339 * in big-endian order.
342 tomb_mbs(char *mb, wchar_t wc)
344 uint8_t *s = (uint8_t *)mb;
347 if ((wc & 0xff000000U) != 0) {
349 } else if ((wc & 0x00ff0000U) != 0) {
351 } else if ((wc & 0x0000ff00U) != 0) {
362 /* ensure null termination */
369 * big5 is a simple dual byte character set.
372 towide_big5(wchar_t *wc, const char *mb, unsigned n)
374 return (towide_dbcs(wc, mb, n));
378 * GBK encodes wides in the same way that big5 does, the high order
379 * bit of the first byte indicates a double byte character.
382 towide_gbk(wchar_t *wc, const char *mb, unsigned n)
384 return (towide_dbcs(wc, mb, n));
388 * GB2312 is another DBCS. Its cleaner than others in that the second
389 * byte does not encode ASCII, but it supports characters.
392 towide_gb2312(wchar_t *wc, const char *mb, unsigned n)
394 return (towide_dbcs(wc, mb, n));
398 * GB18030. This encodes as 8, 16, or 32-bits.
399 * 7-bit values are in 1 byte, 4 byte sequences are used when
400 * the second byte encodes 0x30-39 and all other sequences are 2 bytes.
403 towide_gb18030(wchar_t *wc, const char *mb, unsigned n)
409 if ((c & 0x80) == 0) {
415 werr("incomplete character sequence (%s)", show_mb(mb));
419 /* pull in the second byte */
421 c |= (uint8_t)(mb[1]);
423 if (((c & 0xff) >= 0x30) && ((c & 0xff) <= 0x39)) {
425 werr("incomplete 4-byte character sequence (%s)",
430 c |= (uint8_t)(mb[2]);
432 c |= (uint8_t)(mb[3]);
442 * MS-Kanji (aka SJIS) is almost a clean DBCS like the others, but it
443 * also has a range of single byte characters above 0x80. (0xa1-0xdf).
446 towide_mskanji(wchar_t *wc, const char *mb, unsigned n)
452 if ((c < 0x80) || ((c > 0xa0) && (c < 0xe0))) {
459 werr("incomplete character sequence (%s)", show_mb(mb));
463 /* Store both bytes as a single 16-bit wide. */
465 c |= (uint8_t)(mb[1]);
471 * EUC forms. EUC encodings are "variable". FreeBSD carries some additional
472 * variable data to encode these, but we're going to treat each as independent
473 * instead. Its the only way we can sensibly move forward.
475 * Note that the way in which the different EUC forms vary is how wide
476 * CS2 and CS3 are and what the first byte of them is.
479 towide_euc_impl(wchar_t *wc, const char *mb, unsigned n,
480 uint8_t cs2, uint8_t cs2width, uint8_t cs3, uint8_t cs3width)
489 * All variations of EUC encode 7-bit ASCII as one byte, and use
490 * additional bytes for more than that.
492 if ((c & 0x80) == 0) {
499 * All EUC variants reserve 0xa1-0xff to identify CS1, which
500 * is always two bytes wide. Note that unused CS will be zero,
501 * and that cannot be true because we know that the high order
506 } else if (c == cs2) {
508 } else if (c == cs3) {
512 if ((int)n < width) {
513 werr("incomplete character sequence (%s)", show_mb(mb));
517 for (i = 1; i < width; i++) {
518 /* pull in the next byte */
520 c |= (uint8_t)(mb[i]);
527 #pragma GCC diagnostic pop
530 * EUC-CN encodes as follows:
532 * Code set 0 (ASCII): 0x21-0x7E
533 * Code set 1 (CNS 11643-1992 Plane 1): 0xA1A1-0xFEFE
538 towide_euccn(wchar_t *wc, const char *mb, unsigned n)
540 return (towide_euc_impl(wc, mb, n, 0x8e, 4, 0, 0));
544 * EUC-JP encodes as follows:
546 * Code set 0 (ASCII or JIS X 0201-1976 Roman): 0x21-0x7E
547 * Code set 1 (JIS X 0208): 0xA1A1-0xFEFE
548 * Code set 2 (half-width katakana): 0x8EA1-0x8EDF
549 * Code set 3 (JIS X 0212-1990): 0x8FA1A1-0x8FFEFE
552 towide_eucjp(wchar_t *wc, const char *mb, unsigned n)
554 return (towide_euc_impl(wc, mb, n, 0x8e, 2, 0x8f, 3));
558 * EUC-KR encodes as follows:
560 * Code set 0 (ASCII or KS C 5636-1993): 0x21-0x7E
561 * Code set 1 (KS C 5601-1992): 0xA1A1-0xFEFE
566 towide_euckr(wchar_t *wc, const char *mb, unsigned n)
568 return (towide_euc_impl(wc, mb, n, 0, 0, 0, 0));
572 * EUC-TW encodes as follows:
574 * Code set 0 (ASCII): 0x21-0x7E
575 * Code set 1 (CNS 11643-1992 Plane 1): 0xA1A1-0xFEFE
576 * Code set 2 (CNS 11643-1992 Planes 1-16): 0x8EA1A1A1-0x8EB0FEFE
580 towide_euctw(wchar_t *wc, const char *mb, unsigned n)
582 return (towide_euc_impl(wc, mb, n, 0x8e, 4, 0, 0));
586 * Public entry points.
590 to_wide(wchar_t *wc, const char *mb)
592 /* this won't fail hard */
593 return (_towide(wc, mb, strlen(mb)));
597 to_mb(char *mb, wchar_t wc)
601 if ((rv = _tomb(mb, wc)) < 0) {
610 to_mb_string(const wchar_t *wcs)
616 mbs = malloc((wcslen(wcs) * mb_cur_max) + 1);
618 errf("out of memory");
623 if ((len = to_mb(ptr, *wcs)) < 0) {
636 set_wide_encoding(const char *encoding)
640 _towide = towide_none;
645 for (i = 0; mb_encodings[i].name; i++) {
646 if (strcasecmp(encoding, mb_encodings[i].name) == 0) {
647 _towide = mb_encodings[i].towide;
648 _tomb = mb_encodings[i].tomb;
649 _encoding = mb_encodings[i].cname;
650 _nbits = mb_encodings[i].nbits;
657 get_wide_encoding(void)
665 return ((int)((1U << _nbits) - 1));