2 * Copyright 2010 Nexenta Systems, Inc. All rights reserved.
3 * Copyright (c) 1995 Alex Tatmanjants <alex@elvisti.kiev.ua>
4 * at Electronni Visti IA, Kiev, Ukraine.
7 * Copyright (c) 2011 The FreeBSD Foundation
9 * Portions of this software were developed by David Chisnall
10 * under sponsorship from the FreeBSD Foundation.
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
21 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * Adapted to xlocale by John Marino <draco@marino.st>
36 #include "namespace.h"
37 #include <sys/types.h>
48 #include "un-namespace.h"
51 #include "setlocale.h"
53 #include "libc_private.h"
55 struct xlocale_collate __xlocale_global_collate = {
56 {{0}, "C"}, 1, 0, 0, 0
59 struct xlocale_collate __xlocale_C_collate = {
60 {{0}, "C"}, 1, 0, 0, 0
64 __collate_load_tables_l(const char *encoding, struct xlocale_collate *table);
67 destruct_collate(void *t)
69 struct xlocale_collate *table = t;
70 if (table->map && (table->maplen > 0)) {
71 (void) munmap(table->map, table->maplen);
77 __collate_load(const char *encoding, __unused locale_t unused)
79 if (strcmp(encoding, "C") == 0 || strcmp(encoding, "POSIX") == 0) {
80 return &__xlocale_C_collate;
82 struct xlocale_collate *table = calloc(sizeof(struct xlocale_collate), 1);
83 table->header.header.destructor = destruct_collate;
84 // FIXME: Make sure that _LDP_CACHE is never returned. We should be doing
85 // the caching outside of this section
86 if (__collate_load_tables_l(encoding, table) != _LDP_LOADED) {
87 xlocale_release(table);
94 * Load the collation tables for the specified encoding into the global table.
97 __collate_load_tables(const char *encoding)
99 return (__collate_load_tables_l(encoding, &__xlocale_global_collate));
103 __collate_load_tables_l(const char *encoding, struct xlocale_collate *table)
109 collate_info_t *info;
113 table->__collate_load_error = 1;
115 /* 'encoding' must be already checked. */
116 if (strcmp(encoding, "C") == 0 || strcmp(encoding, "POSIX") == 0) {
120 (void) snprintf(buf, sizeof (buf), "%s/%s/LC_COLLATE",
121 _PathLocale, encoding);
123 if ((fd = _open(buf, O_RDONLY)) < 0)
125 if (_fstat(fd, &sbuf) < 0) {
129 if (sbuf.st_size < (COLLATE_STR_LEN + sizeof (info))) {
134 map = mmap(NULL, sbuf.st_size, PROT_READ, MAP_PRIVATE, fd, 0);
136 if ((TMP = map) == NULL) {
140 if (strncmp(TMP, COLLATE_VERSION, COLLATE_STR_LEN) != 0) {
141 (void) munmap(map, sbuf.st_size);
145 TMP += COLLATE_STR_LEN;
148 TMP += sizeof (*info);
150 if ((info->directive_count < 1) ||
151 (info->directive_count >= COLL_WEIGHTS_MAX) ||
152 ((chains = info->chain_count) < 0)) {
153 (void) munmap(map, sbuf.st_size);
158 i = (sizeof (collate_char_t) * (UCHAR_MAX + 1)) +
159 (sizeof (collate_chain_t) * chains) +
160 (sizeof (collate_large_t) * info->large_count);
161 for (z = 0; z < info->directive_count; z++) {
162 i += sizeof (collate_subst_t) * info->subst_count[z];
164 if (i != (sbuf.st_size - (TMP - map))) {
165 (void) munmap(map, sbuf.st_size);
171 table->char_pri_table = (void *)TMP;
172 TMP += sizeof (collate_char_t) * (UCHAR_MAX + 1);
174 for (z = 0; z < info->directive_count; z++) {
175 if (info->subst_count[z] > 0) {
176 table->subst_table[z] = (void *)TMP;
177 TMP += info->subst_count[z] * sizeof (collate_subst_t);
179 table->subst_table[z] = NULL;
184 table->chain_pri_table = (void *)TMP;
185 TMP += chains * sizeof (collate_chain_t);
187 table->chain_pri_table = NULL;
188 if (info->large_count > 0)
189 table->large_pri_table = (void *)TMP;
191 table->large_pri_table = NULL;
193 table->__collate_load_error = 0;
194 return (_LDP_LOADED);
197 static const int32_t *
198 substsearch(struct xlocale_collate *table, const wchar_t key, int pass)
200 const collate_subst_t *p;
201 int n = table->info->subst_count[pass];
206 if (pass >= table->info->directive_count)
209 if (!(key & COLLATE_SUBST_PRIORITY))
212 p = table->subst_table[pass] + (key & ~COLLATE_SUBST_PRIORITY);
213 assert(p->key == key);
217 static collate_chain_t *
218 chainsearch(struct xlocale_collate *table, const wchar_t *key, int *len)
221 int high = table->info->chain_count - 1;;
224 collate_chain_t *tab = table->chain_pri_table;
229 while (low <= high) {
230 next = (low + high) / 2;
232 compar = *key - *p->str;
234 l = wcsnlen(p->str, COLLATE_STR_LEN);
235 compar = wcsncmp(key, p->str, l);
249 static collate_large_t *
250 largesearch(struct xlocale_collate *table, const wchar_t key)
253 int high = table->info->large_count - 1;
256 collate_large_t *tab = table->large_pri_table;
261 while (low <= high) {
262 next = (low + high) / 2;
264 compar = key - p->val;
276 _collate_lookup(struct xlocale_collate *table, const wchar_t *t, int *len,
277 int *pri, int which, const int **state)
280 collate_large_t *match;
285 * If this is the "last" pass for the UNDEFINED, then
286 * we just return the priority itself.
288 if (which >= table->info->directive_count) {
296 * If we have remaining substitution data from a previous
297 * call, consume it first.
299 if ((sptr = *state) != NULL) {
302 if ((sptr == *state) || (sptr == NULL))
310 /* No active substitutions */
314 * Check for composites such as diphthongs that collate as a
315 * single element (aka chains or collating-elements).
317 if (((p2 = chainsearch(table, t, &l)) != NULL) &&
318 ((p = p2->pri[which]) >= 0)) {
323 } else if (*t <= UCHAR_MAX) {
326 * Character is a small (8-bit) character.
327 * We just look these up directly for speed.
329 *pri = table->char_pri_table[*t].pri[which];
331 } else if ((table->info->large_count > 0) &&
332 ((match = largesearch(table, *t)) != NULL)) {
335 * Character was found in the extended table.
337 *pri = match->pri.pri[which];
341 * Character lacks a specific definition.
343 if (table->info->directive[which] & DIRECTIVE_UNDEFINED) {
344 /* Mask off sign bit to prevent ordering confusion. */
345 *pri = (*t & COLLATE_MAX_PRIORITY);
347 *pri = table->info->undef_pri[which];
349 /* No substitutions for undefined characters! */
354 * Try substituting (expanding) the character. We are
355 * currently doing this *after* the chain compression. I
356 * think it should not matter, but this way might be slightly
359 * We do this after the priority search, as this will help us
360 * to identify a single key value. In order for this to work,
361 * its important that the priority assigned to a given element
362 * to be substituted be unique for that level. The localedef
363 * code ensures this for us.
365 if ((sptr = substsearch(table, *pri, which)) != NULL) {
366 if ((*pri = *sptr) > 0) {
368 *state = *sptr ? sptr : NULL;
375 * This is the meaty part of wcsxfrm & strxfrm. Note that it does
376 * NOT NULL terminate. That is left to the caller.
379 _collate_wxfrm(struct xlocale_collate *table, const wchar_t *src, wchar_t *xf,
388 const int32_t *state;
391 int ndir = table->info->directive_count;
395 for (pass = 0; pass <= ndir; pass++) {
400 /* insert level separator from the previous pass */
408 /* special pass for undefined */
410 direc = DIRECTIVE_FORWARD | DIRECTIVE_UNDEFINED;
412 direc = table->info->directive[pass];
417 if (direc & DIRECTIVE_BACKWARD) {
421 if ((tr = wcsdup(t)) == NULL) {
426 fp = tr + wcslen(tr) - 1;
432 t = (const wchar_t *)tr;
435 if (direc & DIRECTIVE_POSITION) {
436 while (*t || state) {
437 _collate_lookup(table, t, &len, &pri, pass, &state);
445 pri = COLLATE_MAX_PRIORITY;
455 while (*t || state) {
456 _collate_lookup(table, t, &len, &pri, pass, &state);
480 return ((size_t)(-1));
484 * In the non-POSIX case, we transform each character into a string of
485 * characters representing the character's priority. Since char is usually
486 * signed, we are limited by 7 bits per byte. To avoid zero, we need to add
487 * XFRM_OFFSET, so we can't use a full 7 bits. For simplicity, we choose 6
490 * It turns out that we sometimes have real priorities that are
491 * 31-bits wide. (But: be careful using priorities where the high
492 * order bit is set -- i.e. the priority is negative. The sort order
493 * may be surprising!)
495 * TODO: This would be a good area to optimize somewhat. It turns out
496 * that real prioririties *except for the last UNDEFINED pass* are generally
497 * very small. We need the localedef code to precalculate the max
498 * priority for us, and ideally also give us a mask, and then we could
499 * severely limit what we expand to.
502 #define XFRM_OFFSET ('0') /* make all printable characters */
504 #define XFRM_MASK ((1 << XFRM_SHIFT) - 1)
505 #define XFRM_SEP ('.') /* chosen to be less than XFRM_OFFSET */
508 xfrm(struct xlocale_collate *table, unsigned char *p, int pri, int pass)
510 /* we use unsigned to ensure zero fill on right shift */
511 uint32_t val = (uint32_t)table->info->pri_count[pass];
515 *p = (pri & XFRM_MASK) + XFRM_OFFSET;
525 _collate_sxfrm(struct xlocale_collate *table, const wchar_t *src, char *xf,
534 const int32_t *state;
538 uint8_t buf[XFRM_BYTES];
539 int ndir = table->info->directive_count;
543 for (pass = 0; pass <= ndir; pass++) {
548 /* insert level separator from the previous pass */
556 /* special pass for undefined */
558 direc = DIRECTIVE_FORWARD | DIRECTIVE_UNDEFINED;
560 direc = table->info->directive[pass];
565 if (direc & DIRECTIVE_BACKWARD) {
569 if ((tr = wcsdup(t)) == NULL) {
574 fp = tr + wcslen(tr) - 1;
580 t = (const wchar_t *)tr;
583 if (direc & DIRECTIVE_POSITION) {
584 while (*t || state) {
586 _collate_lookup(table, t, &len, &pri, pass, &state);
594 pri = COLLATE_MAX_PRIORITY;
597 b = xfrm(table, buf, pri, pass);
611 while (*t || state) {
612 _collate_lookup(table, t, &len, &pri, pass, &state);
623 b = xfrm(table, buf, pri, pass);
644 return ((size_t)(-1));
648 * __collate_equiv_value returns the primary collation value for the given
649 * collating symbol specified by str and len. Zero or negative is returned
650 * if the collating symbol was not found. This function is used by bracket
651 * code in the TRE regex library.
654 __collate_equiv_value(locale_t locale, const wchar_t *str, size_t len)
658 if (len < 1 || len >= COLLATE_STR_LEN)
662 struct xlocale_collate *table =
663 (struct xlocale_collate*)locale->components[XLC_COLLATE];
665 if (table->__collate_load_error)
666 return ((len == 1 && *str <= UCHAR_MAX) ? *str : -1);
670 if (*str <= UCHAR_MAX)
671 e = table->char_pri_table[*str].pri[0];
672 else if (table->info->large_count > 0) {
673 collate_large_t *match_large;
674 match_large = largesearch(table, *str);
676 e = match_large->pri.pri[0];
680 return (e > 0 ? e : 0);
682 if (table->info->chain_count > 0) {
683 wchar_t name[COLLATE_STR_LEN];
684 collate_chain_t *match_chain;
687 wcsncpy (name, str, len);
689 match_chain = chainsearch(table, name, &clen);
691 e = match_chain->pri[0];
694 return (e < 0 ? -e : e);