4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
23 * Use is subject to license terms.
30 * UTF-8 text preparation functions (PSARC/2007/149, PSARC/2007/458).
32 * Man pages: u8_textprep_open(9F), u8_textprep_buf(9F), u8_textprep_close(9F),
33 * u8_textprep_str(9F), u8_strcmp(9F), and u8_validate(9F). See also
34 * the section 3C man pages.
35 * Interface stability: Committed.
38 #include <sys/types.h>
40 #include <sys/param.h>
41 #include <sys/sysmacros.h>
42 #include <sys/systm.h>
43 #include <sys/debug.h>
46 #include <sys/sunddi.h>
48 #include <sys/u8_textprep.h>
51 #include <sys/byteorder.h>
52 #include <sys/errno.h>
53 #include <sys/u8_textprep_data.h>
56 /* The maximum possible number of bytes in a UTF-8 character. */
57 #define U8_MB_CUR_MAX (4)
60 * The maximum number of bytes needed for a UTF-8 character to cover
61 * U+0000 - U+FFFF, i.e., the coding space of now deprecated UCS-2.
63 #define U8_MAX_BYTES_UCS2 (3)
65 /* The maximum possible number of bytes in a Stream-Safe Text. */
66 #define U8_STREAM_SAFE_TEXT_MAX (128)
69 * The maximum number of characters in a combining/conjoining sequence and
70 * the actual upperbound limit of a combining/conjoining sequence.
72 #define U8_MAX_CHARS_A_SEQ (32)
73 #define U8_UPPER_LIMIT_IN_A_SEQ (31)
75 /* The combining class value for Starter. */
76 #define U8_COMBINING_CLASS_STARTER (0)
79 * Some Hangul related macros at below.
81 * The first and the last of Hangul syllables, Hangul Jamo Leading consonants,
82 * Vowels, and optional Trailing consonants in Unicode scalar values.
84 * Please be noted that the U8_HANGUL_JAMO_T_FIRST is 0x11A7 at below not
85 * the actual U+11A8. This is due to that the trailing consonant is optional
86 * and thus we are doing a pre-calculation of subtracting one.
88 * Each of 19 modern leading consonants has total 588 possible syllables since
89 * Hangul has 21 modern vowels and 27 modern trailing consonants plus 1 for
90 * no trailing consonant case, i.e., 21 x 28 = 588.
92 * We also have bunch of Hangul related macros at below. Please bear in mind
93 * that the U8_HANGUL_JAMO_1ST_BYTE can be used to check whether it is
94 * a Hangul Jamo or not but the value does not guarantee that it is a Hangul
95 * Jamo; it just guarantee that it will be most likely.
97 #define U8_HANGUL_SYL_FIRST (0xAC00U)
98 #define U8_HANGUL_SYL_LAST (0xD7A3U)
100 #define U8_HANGUL_JAMO_L_FIRST (0x1100U)
101 #define U8_HANGUL_JAMO_L_LAST (0x1112U)
102 #define U8_HANGUL_JAMO_V_FIRST (0x1161U)
103 #define U8_HANGUL_JAMO_V_LAST (0x1175U)
104 #define U8_HANGUL_JAMO_T_FIRST (0x11A7U)
105 #define U8_HANGUL_JAMO_T_LAST (0x11C2U)
107 #define U8_HANGUL_V_COUNT (21)
108 #define U8_HANGUL_VT_COUNT (588)
109 #define U8_HANGUL_T_COUNT (28)
111 #define U8_HANGUL_JAMO_1ST_BYTE (0xE1U)
113 #define U8_SAVE_HANGUL_AS_UTF8(s, i, j, k, b) \
114 (s)[(i)] = (uchar_t)(0xE0U | ((uint32_t)(b) & 0xF000U) >> 12); \
115 (s)[(j)] = (uchar_t)(0x80U | ((uint32_t)(b) & 0x0FC0U) >> 6); \
116 (s)[(k)] = (uchar_t)(0x80U | ((uint32_t)(b) & 0x003FU));
118 #define U8_HANGUL_JAMO_L(u) \
119 ((u) >= U8_HANGUL_JAMO_L_FIRST && (u) <= U8_HANGUL_JAMO_L_LAST)
121 #define U8_HANGUL_JAMO_V(u) \
122 ((u) >= U8_HANGUL_JAMO_V_FIRST && (u) <= U8_HANGUL_JAMO_V_LAST)
124 #define U8_HANGUL_JAMO_T(u) \
125 ((u) > U8_HANGUL_JAMO_T_FIRST && (u) <= U8_HANGUL_JAMO_T_LAST)
127 #define U8_HANGUL_JAMO(u) \
128 ((u) >= U8_HANGUL_JAMO_L_FIRST && (u) <= U8_HANGUL_JAMO_T_LAST)
130 #define U8_HANGUL_SYLLABLE(u) \
131 ((u) >= U8_HANGUL_SYL_FIRST && (u) <= U8_HANGUL_SYL_LAST)
133 #define U8_HANGUL_COMPOSABLE_L_V(s, u) \
134 ((s) == U8_STATE_HANGUL_L && U8_HANGUL_JAMO_V((u)))
136 #define U8_HANGUL_COMPOSABLE_LV_T(s, u) \
137 ((s) == U8_STATE_HANGUL_LV && U8_HANGUL_JAMO_T((u)))
139 /* The types of decomposition mappings. */
140 #define U8_DECOMP_BOTH (0xF5U)
141 #define U8_DECOMP_CANONICAL (0xF6U)
143 /* The indicator for 16-bit table. */
144 #define U8_16BIT_TABLE_INDICATOR (0x8000U)
146 /* The following are some convenience macros. */
147 #define U8_PUT_3BYTES_INTO_UTF32(u, b1, b2, b3) \
148 (u) = ((((uint32_t)(b1) & 0x0F) << 12) | \
149 (((uint32_t)(b2) & 0x3F) << 6) | \
150 ((uint32_t)(b3) & 0x3F));
152 #define U8_SIMPLE_SWAP(a, b, t) \
157 #define U8_ASCII_TOUPPER(c) \
158 (((c) >= 'a' && (c) <= 'z') ? (c) - 'a' + 'A' : (c))
160 #define U8_ASCII_TOLOWER(c) \
161 (((c) >= 'A' && (c) <= 'Z') ? (c) - 'A' + 'a' : (c))
163 #define U8_ISASCII(c) (((uchar_t)(c)) < 0x80U)
165 * The following macro assumes that the two characters that are to be
166 * swapped are adjacent to each other and 'a' comes before 'b'.
168 * If the assumptions are not met, then, the macro will fail.
170 #define U8_SWAP_COMB_MARKS(a, b) \
171 for (k = 0; k < disp[(a)]; k++) \
172 u8t[k] = u8s[start[(a)] + k]; \
173 for (k = 0; k < disp[(b)]; k++) \
174 u8s[start[(a)] + k] = u8s[start[(b)] + k]; \
175 start[(b)] = start[(a)] + disp[(b)]; \
176 for (k = 0; k < disp[(a)]; k++) \
177 u8s[start[(b)] + k] = u8t[k]; \
178 U8_SIMPLE_SWAP(comb_class[(a)], comb_class[(b)], tc); \
179 U8_SIMPLE_SWAP(disp[(a)], disp[(b)], tc);
181 /* The possible states during normalization. */
184 U8_STATE_HANGUL_L
= 1,
185 U8_STATE_HANGUL_LV
= 2,
186 U8_STATE_HANGUL_LVT
= 3,
187 U8_STATE_HANGUL_V
= 4,
188 U8_STATE_HANGUL_T
= 5,
189 U8_STATE_COMBINING_MARK
= 6
190 } u8_normalization_states_t
;
193 * The three vectors at below are used to check bytes of a given UTF-8
194 * character are valid and not containing any malformed byte values.
196 * We used to have a quite relaxed UTF-8 binary representation but then there
197 * was some security related issues and so the Unicode Consortium defined
198 * and announced the UTF-8 Corrigendum at Unicode 3.1 and then refined it
199 * one more time at the Unicode 3.2. The following three tables are based on
203 #define U8_ILLEGAL_NEXT_BYTE_COMMON(c) ((c) < 0x80 || (c) > 0xBF)
205 #define I_ U8_ILLEGAL_CHAR
206 #define O_ U8_OUT_OF_RANGE_CHAR
208 const int8_t u8_number_of_bytes
[0x100] = {
209 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
210 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
211 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
212 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
213 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
214 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
215 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
216 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
218 /* 80 81 82 83 84 85 86 87 88 89 8A 8B 8C 8D 8E 8F */
219 I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
,
221 /* 90 91 92 93 94 95 96 97 98 99 9A 9B 9C 9D 9E 9F */
222 I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
,
224 /* A0 A1 A2 A3 A4 A5 A6 A7 A8 A9 AA AB AC AD AE AF */
225 I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
,
227 /* B0 B1 B2 B3 B4 B5 B6 B7 B8 B9 BA BB BC BD BE BF */
228 I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
, I_
,
230 /* C0 C1 C2 C3 C4 C5 C6 C7 C8 C9 CA CB CC CD CE CF */
231 I_
, I_
, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
233 /* D0 D1 D2 D3 D4 D5 D6 D7 D8 D9 DA DB DC DD DE DF */
234 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
236 /* E0 E1 E2 E3 E4 E5 E6 E7 E8 E9 EA EB EC ED EE EF */
237 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
239 /* F0 F1 F2 F3 F4 F5 F6 F7 F8 F9 FA FB FC FD FE FF */
240 4, 4, 4, 4, 4, O_
, O_
, O_
, O_
, O_
, O_
, O_
, O_
, O_
, O_
, O_
,
246 const uint8_t u8_valid_min_2nd_byte
[0x100] = {
247 0, 0, 0, 0, 0, 0, 0, 0,
248 0, 0, 0, 0, 0, 0, 0, 0,
249 0, 0, 0, 0, 0, 0, 0, 0,
250 0, 0, 0, 0, 0, 0, 0, 0,
251 0, 0, 0, 0, 0, 0, 0, 0,
252 0, 0, 0, 0, 0, 0, 0, 0,
253 0, 0, 0, 0, 0, 0, 0, 0,
254 0, 0, 0, 0, 0, 0, 0, 0,
255 0, 0, 0, 0, 0, 0, 0, 0,
256 0, 0, 0, 0, 0, 0, 0, 0,
257 0, 0, 0, 0, 0, 0, 0, 0,
258 0, 0, 0, 0, 0, 0, 0, 0,
259 0, 0, 0, 0, 0, 0, 0, 0,
260 0, 0, 0, 0, 0, 0, 0, 0,
261 0, 0, 0, 0, 0, 0, 0, 0,
262 0, 0, 0, 0, 0, 0, 0, 0,
263 0, 0, 0, 0, 0, 0, 0, 0,
264 0, 0, 0, 0, 0, 0, 0, 0,
265 0, 0, 0, 0, 0, 0, 0, 0,
266 0, 0, 0, 0, 0, 0, 0, 0,
267 0, 0, 0, 0, 0, 0, 0, 0,
268 0, 0, 0, 0, 0, 0, 0, 0,
269 0, 0, 0, 0, 0, 0, 0, 0,
270 0, 0, 0, 0, 0, 0, 0, 0,
271 /* C0 C1 C2 C3 C4 C5 C6 C7 */
272 0, 0, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
273 /* C8 C9 CA CB CC CD CE CF */
274 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
275 /* D0 D1 D2 D3 D4 D5 D6 D7 */
276 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
277 /* D8 D9 DA DB DC DD DE DF */
278 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
279 /* E0 E1 E2 E3 E4 E5 E6 E7 */
280 0xa0, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
281 /* E8 E9 EA EB EC ED EE EF */
282 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
283 /* F0 F1 F2 F3 F4 F5 F6 F7 */
284 0x90, 0x80, 0x80, 0x80, 0x80, 0, 0, 0,
285 0, 0, 0, 0, 0, 0, 0, 0,
288 const uint8_t u8_valid_max_2nd_byte
[0x100] = {
289 0, 0, 0, 0, 0, 0, 0, 0,
290 0, 0, 0, 0, 0, 0, 0, 0,
291 0, 0, 0, 0, 0, 0, 0, 0,
292 0, 0, 0, 0, 0, 0, 0, 0,
293 0, 0, 0, 0, 0, 0, 0, 0,
294 0, 0, 0, 0, 0, 0, 0, 0,
295 0, 0, 0, 0, 0, 0, 0, 0,
296 0, 0, 0, 0, 0, 0, 0, 0,
297 0, 0, 0, 0, 0, 0, 0, 0,
298 0, 0, 0, 0, 0, 0, 0, 0,
299 0, 0, 0, 0, 0, 0, 0, 0,
300 0, 0, 0, 0, 0, 0, 0, 0,
301 0, 0, 0, 0, 0, 0, 0, 0,
302 0, 0, 0, 0, 0, 0, 0, 0,
303 0, 0, 0, 0, 0, 0, 0, 0,
304 0, 0, 0, 0, 0, 0, 0, 0,
305 0, 0, 0, 0, 0, 0, 0, 0,
306 0, 0, 0, 0, 0, 0, 0, 0,
307 0, 0, 0, 0, 0, 0, 0, 0,
308 0, 0, 0, 0, 0, 0, 0, 0,
309 0, 0, 0, 0, 0, 0, 0, 0,
310 0, 0, 0, 0, 0, 0, 0, 0,
311 0, 0, 0, 0, 0, 0, 0, 0,
312 0, 0, 0, 0, 0, 0, 0, 0,
313 /* C0 C1 C2 C3 C4 C5 C6 C7 */
314 0, 0, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf,
315 /* C8 C9 CA CB CC CD CE CF */
316 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf,
317 /* D0 D1 D2 D3 D4 D5 D6 D7 */
318 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf,
319 /* D8 D9 DA DB DC DD DE DF */
320 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf,
321 /* E0 E1 E2 E3 E4 E5 E6 E7 */
322 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0xbf,
323 /* E8 E9 EA EB EC ED EE EF */
324 0xbf, 0xbf, 0xbf, 0xbf, 0xbf, 0x9f, 0xbf, 0xbf,
325 /* F0 F1 F2 F3 F4 F5 F6 F7 */
326 0xbf, 0xbf, 0xbf, 0xbf, 0x8f, 0, 0, 0,
327 0, 0, 0, 0, 0, 0, 0, 0,
332 * The u8_validate() validates on the given UTF-8 character string and
333 * calculate the byte length. It is quite similar to mblen(3C) except that
334 * this will validate against the list of characters if required and
335 * specific to UTF-8 and Unicode.
338 u8_validate(char *u8str
, size_t n
, char **list
, int flag
, int *errnum
)
350 boolean_t no_need_to_validate_entire
;
351 boolean_t check_additional
;
352 boolean_t validate_ucs2_range_only
;
357 ib
= (uchar_t
*)u8str
;
362 no_need_to_validate_entire
= ! (flag
& U8_VALIDATE_ENTIRE
);
363 check_additional
= flag
& U8_VALIDATE_CHECK_ADDITIONAL
;
364 validate_ucs2_range_only
= flag
& U8_VALIDATE_UCS2_RANGE
;
366 while (ib
< ibtail
) {
368 * The first byte of a UTF-8 character tells how many
369 * bytes will follow for the character. If the first byte
370 * is an illegal byte value or out of range value, we just
371 * return -1 with an appropriate error number.
373 sz
= u8_number_of_bytes
[*ib
];
374 if (sz
== U8_ILLEGAL_CHAR
) {
379 if (sz
== U8_OUT_OF_RANGE_CHAR
||
380 (validate_ucs2_range_only
&& sz
> U8_MAX_BYTES_UCS2
)) {
386 * If we don't have enough bytes to check on, that's also
387 * an error. As you can see, we give illegal byte sequence
388 * checking higher priority then EINVAL cases.
390 if ((ibtail
- ib
) < sz
) {
400 * Check on the multi-byte UTF-8 character. For more
401 * details on this, see comment added for the used
402 * data structures at the beginning of the file.
407 for (i
= 1; i
< sz
; i
++) {
409 if (*ib
< u8_valid_min_2nd_byte
[f
] ||
410 *ib
> u8_valid_max_2nd_byte
[f
]) {
415 } else if (U8_ILLEGAL_NEXT_BYTE_COMMON(*ib
)) {
424 if (check_additional
) {
425 for (p
= (uchar_t
**)list
, i
= 0; p
[i
]; i
++) {
429 if (*s1
!= *s2
|| *s2
== '\0')
435 if (s1
>= ib
&& *s2
== '\0') {
442 if (no_need_to_validate_entire
)
450 * The do_case_conv() looks at the mapping tables and returns found
451 * bytes if any. If not found, the input bytes are returned. The function
452 * always terminate the return bytes with a null character assuming that
453 * there are plenty of room to do so.
455 * The case conversions are simple case conversions mapping a character to
456 * another character as specified in the Unicode data. The byte size of
457 * the mapped character could be different from that of the input character.
459 * The return value is the byte length of the returned character excluding
460 * the terminating null byte.
463 do_case_conv(int uv
, uchar_t
*u8s
, uchar_t
*s
, int sz
, boolean_t is_it_toupper
)
476 * At this point, the only possible values for sz are 2, 3, and 4.
477 * The u8s should point to a vector that is well beyond the size of
483 } else if (sz
== 3) {
487 } else if (sz
== 4) {
493 /* This is not possible but just in case as a fallback. */
495 *u8s
= U8_ASCII_TOUPPER(*s
);
497 *u8s
= U8_ASCII_TOLOWER(*s
);
505 * Let's find out if we have a corresponding character.
507 b1
= u8_common_b1_tbl
[uv
][b1
];
508 if (b1
== U8_TBL_ELEMENT_NOT_DEF
)
511 b2
= u8_case_common_b2_tbl
[uv
][b1
][b2
];
512 if (b2
== U8_TBL_ELEMENT_NOT_DEF
)
516 b3_tbl
= u8_toupper_b3_tbl
[uv
][b2
][b3
].tbl_id
;
517 if (b3_tbl
== U8_TBL_ELEMENT_NOT_DEF
)
520 start_id
= u8_toupper_b4_tbl
[uv
][b3_tbl
][b4
];
521 end_id
= u8_toupper_b4_tbl
[uv
][b3_tbl
][b4
+ 1];
523 /* Either there is no match or an error at the table. */
524 if (start_id
>= end_id
|| (end_id
- start_id
) > U8_MB_CUR_MAX
)
527 b3_base
= u8_toupper_b3_tbl
[uv
][b2
][b3
].base
;
529 for (i
= 0; start_id
< end_id
; start_id
++)
530 u8s
[i
++] = u8_toupper_final_tbl
[uv
][b3_base
+ start_id
];
532 b3_tbl
= u8_tolower_b3_tbl
[uv
][b2
][b3
].tbl_id
;
533 if (b3_tbl
== U8_TBL_ELEMENT_NOT_DEF
)
536 start_id
= u8_tolower_b4_tbl
[uv
][b3_tbl
][b4
];
537 end_id
= u8_tolower_b4_tbl
[uv
][b3_tbl
][b4
+ 1];
539 if (start_id
>= end_id
|| (end_id
- start_id
) > U8_MB_CUR_MAX
)
542 b3_base
= u8_tolower_b3_tbl
[uv
][b2
][b3
].base
;
544 for (i
= 0; start_id
< end_id
; start_id
++)
545 u8s
[i
++] = u8_tolower_final_tbl
[uv
][b3_base
+ start_id
];
549 * If i is still zero, that means there is no corresponding character.
560 * The do_case_compare() function compares the two input strings, s1 and s2,
561 * one character at a time doing case conversions if applicable and return
562 * the comparison result as like strcmp().
564 * Since, in empirical sense, most of text data are 7-bit ASCII characters,
565 * we treat the 7-bit ASCII characters as a special case trying to yield
566 * faster processing time.
569 do_case_compare(size_t uv
, uchar_t
*s1
, uchar_t
*s2
, size_t n1
,
570 size_t n2
, boolean_t is_it_toupper
, int *errnum
)
578 uchar_t u8s1
[U8_MB_CUR_MAX
+ 1];
579 uchar_t u8s2
[U8_MB_CUR_MAX
+ 1];
582 while (i1
< n1
&& i2
< n2
) {
584 * Find out what would be the byte length for this UTF-8
585 * character at string s1 and also find out if this is
586 * an illegal start byte or not and if so, issue a proper
587 * error number and yet treat this byte as a character.
589 sz1
= u8_number_of_bytes
[*s1
];
596 * For 7-bit ASCII characters mainly, we do a quick case
597 * conversion right at here.
599 * If we don't have enough bytes for this character, issue
600 * an EINVAL error and use what are available.
602 * If we have enough bytes, find out if there is
603 * a corresponding uppercase character and if so, copy over
604 * the bytes for a comparison later. If there is no
605 * corresponding uppercase character, then, use what we have
606 * for the comparison.
610 u8s1
[0] = U8_ASCII_TOUPPER(*s1
);
612 u8s1
[0] = U8_ASCII_TOLOWER(*s1
);
615 } else if ((i1
+ sz1
) > n1
) {
617 for (j
= 0; (i1
+ j
) < n1
; )
621 (void) do_case_conv(uv
, u8s1
, s1
, sz1
, is_it_toupper
);
625 /* Do the same for the string s2. */
626 sz2
= u8_number_of_bytes
[*s2
];
634 u8s2
[0] = U8_ASCII_TOUPPER(*s2
);
636 u8s2
[0] = U8_ASCII_TOLOWER(*s2
);
639 } else if ((i2
+ sz2
) > n2
) {
641 for (j
= 0; (i2
+ j
) < n2
; )
645 (void) do_case_conv(uv
, u8s2
, s2
, sz2
, is_it_toupper
);
649 /* Now compare the two characters. */
650 if (sz1
== 1 && sz2
== 1) {
656 f
= strcmp((const char *)u8s1
, (const char *)u8s2
);
662 * They were the same. Let's move on to the next
670 * We compared until the end of either or both strings.
672 * If we reached to or went over the ends for the both, that means
675 * If we reached only one of the two ends, that means the other string
676 * has something which then the fact can be used to determine
688 * The combining_class() function checks on the given bytes and find out
689 * the corresponding Unicode combining class value. The return value 0 means
690 * it is a Starter. Any illegal UTF-8 character will also be treated as
694 combining_class(size_t uv
, uchar_t
*s
, size_t sz
)
701 if (sz
== 1 || sz
> 4)
707 } else if (sz
== 3) {
711 } else if (sz
== 4) {
718 b1
= u8_common_b1_tbl
[uv
][b1
];
719 if (b1
== U8_TBL_ELEMENT_NOT_DEF
)
722 b2
= u8_combining_class_b2_tbl
[uv
][b1
][b2
];
723 if (b2
== U8_TBL_ELEMENT_NOT_DEF
)
726 b3
= u8_combining_class_b3_tbl
[uv
][b2
][b3
];
727 if (b3
== U8_TBL_ELEMENT_NOT_DEF
)
730 return (u8_combining_class_b4_tbl
[uv
][b3
][b4
]);
734 * The do_decomp() function finds out a matching decomposition if any
735 * and return. If there is no match, the input bytes are copied and returned.
736 * The function also checks if there is a Hangul, decomposes it if necessary
739 * To save time, a single byte 7-bit ASCII character should be handled by
742 * The function returns the number of bytes returned sans always terminating
743 * the null byte. It will also return a state that will tell if there was
744 * a Hangul character decomposed which then will be used by the caller.
747 do_decomp(size_t uv
, uchar_t
*u8s
, uchar_t
*s
, int sz
,
748 boolean_t canonical_decomposition
, u8_normalization_states_t
*state
)
765 } else if (sz
== 3) {
766 /* Convert it to a Unicode scalar value. */
767 U8_PUT_3BYTES_INTO_UTF32(u1
, s
[0], s
[1], s
[2]);
770 * If this is a Hangul syllable, we decompose it into
771 * a leading consonant, a vowel, and an optional trailing
772 * consonant and then return.
774 if (U8_HANGUL_SYLLABLE(u1
)) {
775 u1
-= U8_HANGUL_SYL_FIRST
;
777 b1
= U8_HANGUL_JAMO_L_FIRST
+ u1
/ U8_HANGUL_VT_COUNT
;
778 b2
= U8_HANGUL_JAMO_V_FIRST
+ (u1
% U8_HANGUL_VT_COUNT
)
780 b3
= u1
% U8_HANGUL_T_COUNT
;
782 U8_SAVE_HANGUL_AS_UTF8(u8s
, 0, 1, 2, b1
);
783 U8_SAVE_HANGUL_AS_UTF8(u8s
, 3, 4, 5, b2
);
785 b3
+= U8_HANGUL_JAMO_T_FIRST
;
786 U8_SAVE_HANGUL_AS_UTF8(u8s
, 6, 7, 8, b3
);
789 *state
= U8_STATE_HANGUL_LVT
;
794 *state
= U8_STATE_HANGUL_LV
;
804 * If this is a Hangul Jamo, we know there is nothing
805 * further that we can decompose.
807 if (U8_HANGUL_JAMO_L(u1
)) {
808 *state
= U8_STATE_HANGUL_L
;
812 if (U8_HANGUL_JAMO_V(u1
)) {
813 if (*state
== U8_STATE_HANGUL_L
)
814 *state
= U8_STATE_HANGUL_LV
;
816 *state
= U8_STATE_HANGUL_V
;
820 if (U8_HANGUL_JAMO_T(u1
)) {
821 if (*state
== U8_STATE_HANGUL_LV
)
822 *state
= U8_STATE_HANGUL_LVT
;
824 *state
= U8_STATE_HANGUL_T
;
827 } else if (sz
== 4) {
835 * This is a fallback and should not happen if the function
836 * was called properly.
840 *state
= U8_STATE_START
;
845 * At this point, this routine does not know what it would get.
846 * The caller should sort it out if the state isn't a Hangul one.
848 *state
= U8_STATE_START
;
850 /* Try to find matching decomposition mapping byte sequence. */
851 b1
= u8_common_b1_tbl
[uv
][b1
];
852 if (b1
== U8_TBL_ELEMENT_NOT_DEF
)
855 b2
= u8_decomp_b2_tbl
[uv
][b1
][b2
];
856 if (b2
== U8_TBL_ELEMENT_NOT_DEF
)
859 b3_tbl
= u8_decomp_b3_tbl
[uv
][b2
][b3
].tbl_id
;
860 if (b3_tbl
== U8_TBL_ELEMENT_NOT_DEF
)
864 * If b3_tbl is bigger than or equal to U8_16BIT_TABLE_INDICATOR
865 * which is 0x8000, this means we couldn't fit the mappings into
866 * the cardinality of a unsigned byte.
868 if (b3_tbl
>= U8_16BIT_TABLE_INDICATOR
) {
869 b3_tbl
-= U8_16BIT_TABLE_INDICATOR
;
870 start_id
= u8_decomp_b4_16bit_tbl
[uv
][b3_tbl
][b4
];
871 end_id
= u8_decomp_b4_16bit_tbl
[uv
][b3_tbl
][b4
+ 1];
873 start_id
= u8_decomp_b4_tbl
[uv
][b3_tbl
][b4
];
874 end_id
= u8_decomp_b4_tbl
[uv
][b3_tbl
][b4
+ 1];
877 /* This also means there wasn't any matching decomposition. */
878 if (start_id
>= end_id
)
882 * The final table for decomposition mappings has three types of
883 * byte sequences depending on whether a mapping is for compatibility
884 * decomposition, canonical decomposition, or both like the following:
886 * (1) Compatibility decomposition mappings:
888 * +---+---+-...-+---+
889 * | B0| B1| ... | Bm|
890 * +---+---+-...-+---+
892 * The first byte, B0, is always less then 0xF5 (U8_DECOMP_BOTH).
894 * (2) Canonical decomposition mappings:
896 * +---+---+---+-...-+---+
897 * | T | b0| b1| ... | bn|
898 * +---+---+---+-...-+---+
900 * where the first byte, T, is 0xF6 (U8_DECOMP_CANONICAL).
904 * +---+---+---+---+-...-+---+---+---+-...-+---+
905 * | T | D | b0| b1| ... | bn| B0| B1| ... | Bm|
906 * +---+---+---+---+-...-+---+---+---+-...-+---+
908 * where T is 0xF5 (U8_DECOMP_BOTH) and D is a displacement
909 * byte, b0 to bn are canonical mapping bytes and B0 to Bm are
910 * compatibility mapping bytes.
912 * Note that compatibility decomposition means doing recursive
913 * decompositions using both compatibility decomposition mappings and
914 * canonical decomposition mappings. On the other hand, canonical
915 * decomposition means doing recursive decompositions using only
916 * canonical decomposition mappings. Since the table we have has gone
917 * through the recursions already, we do not need to do so during
918 * runtime, i.e., the table has been completely flattened out
922 b3_base
= u8_decomp_b3_tbl
[uv
][b2
][b3
].base
;
924 /* Get the type, T, of the byte sequence. */
925 b1
= u8_decomp_final_tbl
[uv
][b3_base
+ start_id
];
928 * If necessary, adjust start_id, end_id, or both. Note that if
929 * this is compatibility decomposition mapping, there is no
932 if (canonical_decomposition
) {
933 /* Is the mapping only for compatibility decomposition? */
934 if (b1
< U8_DECOMP_BOTH
)
939 if (b1
== U8_DECOMP_BOTH
) {
941 u8_decomp_final_tbl
[uv
][b3_base
+ start_id
];
946 * Unless this is a compatibility decomposition mapping,
947 * we adjust the start_id.
949 if (b1
== U8_DECOMP_BOTH
) {
951 start_id
+= u8_decomp_final_tbl
[uv
][b3_base
+ start_id
];
952 } else if (b1
== U8_DECOMP_CANONICAL
) {
957 for (i
= 0; start_id
< end_id
; start_id
++)
958 u8s
[i
++] = u8_decomp_final_tbl
[uv
][b3_base
+ start_id
];
965 * The find_composition_start() function uses the character bytes given and
966 * find out the matching composition mappings if any and return the address
967 * to the composition mappings as explained in the do_composition().
970 find_composition_start(size_t uv
, uchar_t
*s
, size_t sz
)
983 } else if (sz
== 2) {
986 } else if (sz
== 3) {
990 } else if (sz
== 4) {
997 * This is a fallback and should not happen if the function
998 * was called properly.
1003 b1
= u8_composition_b1_tbl
[uv
][b1
];
1004 if (b1
== U8_TBL_ELEMENT_NOT_DEF
)
1007 b2
= u8_composition_b2_tbl
[uv
][b1
][b2
];
1008 if (b2
== U8_TBL_ELEMENT_NOT_DEF
)
1011 b3_tbl
= u8_composition_b3_tbl
[uv
][b2
][b3
].tbl_id
;
1012 if (b3_tbl
== U8_TBL_ELEMENT_NOT_DEF
)
1015 if (b3_tbl
>= U8_16BIT_TABLE_INDICATOR
) {
1016 b3_tbl
-= U8_16BIT_TABLE_INDICATOR
;
1017 start_id
= u8_composition_b4_16bit_tbl
[uv
][b3_tbl
][b4
];
1018 end_id
= u8_composition_b4_16bit_tbl
[uv
][b3_tbl
][b4
+ 1];
1020 start_id
= u8_composition_b4_tbl
[uv
][b3_tbl
][b4
];
1021 end_id
= u8_composition_b4_tbl
[uv
][b3_tbl
][b4
+ 1];
1024 if (start_id
>= end_id
)
1027 b3_base
= u8_composition_b3_tbl
[uv
][b2
][b3
].base
;
1029 return ((uchar_t
*)&(u8_composition_final_tbl
[uv
][b3_base
+ start_id
]));
1033 * The blocked() function checks on the combining class values of previous
1034 * characters in this sequence and return whether it is blocked or not.
1037 blocked(uchar_t
*comb_class
, size_t last
)
1039 uchar_t my_comb_class
;
1042 my_comb_class
= comb_class
[last
];
1043 for (i
= 1; i
< last
; i
++)
1044 if (comb_class
[i
] >= my_comb_class
||
1045 comb_class
[i
] == U8_COMBINING_CLASS_STARTER
)
1052 * The do_composition() reads the character string pointed by 's' and
1053 * do necessary canonical composition and then copy over the result back to
1056 * The input argument 's' cannot contain more than 32 characters.
1059 do_composition(size_t uv
, uchar_t
*s
, uchar_t
*comb_class
, uchar_t
*start
,
1060 uchar_t
*disp
, size_t last
, uchar_t
**os
, uchar_t
*oslast
)
1062 uchar_t t
[U8_STREAM_SAFE_TEXT_MAX
+ 1];
1063 uchar_t tc
[U8_MB_CUR_MAX
] = { '\0' };
1064 uint8_t saved_marks
[U8_MAX_CHARS_A_SEQ
];
1065 size_t saved_marks_count
;
1079 boolean_t match_not_found
= B_TRUE
;
1082 * This should never happen unless the callers are doing some strange
1083 * and unexpected things.
1085 * The "last" is the index pointing to the last character not last + 1.
1087 if (last
>= U8_MAX_CHARS_A_SEQ
)
1088 last
= U8_UPPER_LIMIT_IN_A_SEQ
;
1090 for (i
= l
= 0; i
<= last
; i
++) {
1092 * The last or any non-Starters at the beginning, we don't
1093 * have any chance to do composition and so we just copy them
1094 * to the temporary buffer.
1096 if (i
>= last
|| comb_class
[i
] != U8_COMBINING_CLASS_STARTER
) {
1100 for (k
= 0; k
< size
; k
++)
1106 * If this could be a start of Hangul Jamos, then, we try to
1109 if (s
[start
[i
]] == U8_HANGUL_JAMO_1ST_BYTE
) {
1110 U8_PUT_3BYTES_INTO_UTF32(u1
, s
[start
[i
]],
1111 s
[start
[i
] + 1], s
[start
[i
] + 2]);
1112 U8_PUT_3BYTES_INTO_UTF32(u2
, s
[start
[i
] + 3],
1113 s
[start
[i
] + 4], s
[start
[i
] + 5]);
1115 if (U8_HANGUL_JAMO_L(u1
) && U8_HANGUL_JAMO_V(u2
)) {
1116 u1
-= U8_HANGUL_JAMO_L_FIRST
;
1117 u2
-= U8_HANGUL_JAMO_V_FIRST
;
1118 u1
= U8_HANGUL_SYL_FIRST
+
1119 (u1
* U8_HANGUL_V_COUNT
+ u2
) *
1124 U8_PUT_3BYTES_INTO_UTF32(u2
,
1125 s
[start
[i
]], s
[start
[i
] + 1],
1128 if (U8_HANGUL_JAMO_T(u2
)) {
1130 U8_HANGUL_JAMO_T_FIRST
;
1135 U8_SAVE_HANGUL_AS_UTF8(t
+ l
, 0, 1, 2, u1
);
1143 * Let's then find out if this Starter has composition
1146 p
= find_composition_start(uv
, s
+ start
[i
], disp
[i
]);
1151 * We have a Starter with composition mapping and the next
1152 * character is a non-Starter. Let's try to find out if
1153 * we can do composition.
1159 saved_marks_count
= 0;
1166 * The next for() loop compares the non-Starter pointed by
1167 * 'q' with the possible (joinable) characters pointed by 'p'.
1169 * The composition final table entry pointed by the 'p'
1170 * looks like the following:
1172 * +---+---+---+-...-+---+---+---+---+-...-+---+---+
1173 * | C | b0| b2| ... | bn| F | B0| B1| ... | Bm| F |
1174 * +---+---+---+-...-+---+---+---+---+-...-+---+---+
1176 * where C is the count byte indicating the number of
1177 * mapping pairs where each pair would be look like
1178 * (b0-bn F, B0-Bm F). The b0-bn are the bytes of the second
1179 * character of a canonical decomposition and the B0-Bm are
1180 * the bytes of a matching composite character. The F is
1181 * a filler byte after each character as the separator.
1184 match_not_found
= B_TRUE
;
1186 for (C
= *p
++; C
> 0; C
--) {
1187 for (k
= 0; k
< size
; p
++, k
++)
1191 /* Have we found it? */
1192 if (k
>= size
&& *p
== U8_TBL_ELEMENT_FILLER
) {
1193 match_not_found
= B_FALSE
;
1197 while (*++p
!= U8_TBL_ELEMENT_FILLER
)
1203 /* We didn't find; skip to the next pair. */
1204 if (*p
!= U8_TBL_ELEMENT_FILLER
)
1205 while (*++p
!= U8_TBL_ELEMENT_FILLER
)
1207 while (*++p
!= U8_TBL_ELEMENT_FILLER
)
1213 * If there was no match, we will need to save the combining
1214 * mark for later appending. After that, if the next one
1215 * is a non-Starter and not blocked, then, we try once
1216 * again to do composition with the next non-Starter.
1218 * If there was no match and this was a Starter, then,
1219 * this is a new start.
1221 * If there was a match and a composition done and we have
1222 * more to check on, then, we retrieve a new composition final
1223 * table entry for the composite and then try to do the
1224 * composition again.
1227 if (match_not_found
) {
1228 if (comb_class
[i
] == U8_COMBINING_CLASS_STARTER
) {
1233 saved_marks
[saved_marks_count
++] = i
;
1238 if (blocked(comb_class
, i
+ 1))
1239 saved_marks
[saved_marks_count
++] = ++i
;
1245 goto TRY_THE_NEXT_MARK
;
1247 } else if (i
< last
) {
1248 p
= find_composition_start(uv
, t
+ saved_l
,
1252 goto TRY_THE_NEXT_MARK
;
1257 * There is no more composition possible.
1259 * If there was no composition what so ever then we copy
1260 * over the original Starter and then append any non-Starters
1261 * remaining at the target string sequentially after that.
1265 p
= s
+ start
[saved_i
];
1266 size
= disp
[saved_i
];
1267 for (j
= 0; j
< size
; j
++)
1271 for (k
= 0; k
< saved_marks_count
; k
++) {
1272 p
= s
+ start
[saved_marks
[k
]];
1273 size
= disp
[saved_marks
[k
]];
1274 for (j
= 0; j
< size
; j
++)
1280 * If the last character is a Starter and if we have a character
1281 * (possibly another Starter) that can be turned into a composite,
1282 * we do so and we do so until there is no more of composition
1285 if (comb_class
[last
] == U8_COMBINING_CLASS_STARTER
) {
1287 saved_l
= l
- disp
[last
];
1289 while (p
< oslast
) {
1290 size
= u8_number_of_bytes
[*p
];
1291 if (size
<= 1 || (p
+ size
) > oslast
)
1296 for (i
= 0; i
< size
; i
++)
1299 q
= find_composition_start(uv
, t
+ saved_l
,
1306 match_not_found
= B_TRUE
;
1308 for (C
= *q
++; C
> 0; C
--) {
1309 for (k
= 0; k
< size
; q
++, k
++)
1313 if (k
>= size
&& *q
== U8_TBL_ELEMENT_FILLER
) {
1314 match_not_found
= B_FALSE
;
1318 while (*++q
!= U8_TBL_ELEMENT_FILLER
) {
1320 * This is practically
1321 * impossible but we don't
1322 * want to take any chances.
1325 U8_STREAM_SAFE_TEXT_MAX
) {
1335 if (*q
!= U8_TBL_ELEMENT_FILLER
)
1336 while (*++q
!= U8_TBL_ELEMENT_FILLER
)
1338 while (*++q
!= U8_TBL_ELEMENT_FILLER
)
1343 if (match_not_found
) {
1353 * Now we copy over the temporary string to the target string.
1354 * Since composition always reduces the number of characters or
1355 * the number of characters stay, we don't need to worry about
1356 * the buffer overflow here.
1358 for (i
= 0; i
< l
; i
++)
1366 * The collect_a_seq() function checks on the given string s, collect
1367 * a sequence of characters at u8s, and return the sequence. While it collects
1368 * a sequence, it also applies case conversion, canonical or compatibility
1369 * decomposition, canonical decomposition, or some or all of them and
1372 * The collected sequence cannot be bigger than 32 characters since if
1373 * it is having more than 31 characters, the sequence will be terminated
1374 * with a U+034F COMBINING GRAPHEME JOINER (CGJ) character and turned into
1375 * a Stream-Safe Text. The collected sequence is always terminated with
1376 * a null byte and the return value is the byte length of the sequence
1377 * including 0. The return value does not include the terminating
1381 collect_a_seq(size_t uv
, uchar_t
*u8s
, uchar_t
**source
, uchar_t
*slast
,
1382 boolean_t is_it_toupper
, boolean_t is_it_tolower
,
1383 boolean_t canonical_decomposition
, boolean_t compatibility_decomposition
,
1384 boolean_t canonical_composition
,
1385 int *errnum
, u8_normalization_states_t
*state
)
1394 uchar_t comb_class
[U8_MAX_CHARS_A_SEQ
];
1395 uchar_t disp
[U8_MAX_CHARS_A_SEQ
];
1396 uchar_t start
[U8_MAX_CHARS_A_SEQ
];
1397 uchar_t u8t
[U8_MB_CUR_MAX
] = { '\0' };
1398 uchar_t uts
[U8_STREAM_SAFE_TEXT_MAX
+ 1];
1405 * Save the source string pointer which we will return a changed
1406 * pointer if we do processing.
1411 * The following is a fallback for just in case callers are not
1412 * checking the string boundaries before the calling.
1421 * As the first thing, let's collect a character and do case
1422 * conversion if necessary.
1425 sz
= u8_number_of_bytes
[*s
];
1440 u8s
[0] = U8_ASCII_TOUPPER(*s
);
1441 else if (is_it_tolower
)
1442 u8s
[0] = U8_ASCII_TOLOWER(*s
);
1447 } else if ((s
+ sz
) > slast
) {
1450 for (i
= 0; s
< slast
; )
1458 if (is_it_toupper
|| is_it_tolower
) {
1459 i
= do_case_conv(uv
, u8s
, s
, sz
, is_it_toupper
);
1463 for (i
= 0; i
< sz
; )
1470 * And then canonical/compatibility decomposition followed by
1471 * an optional canonical composition. Please be noted that
1472 * canonical composition is done only when a decomposition is
1475 if (canonical_decomposition
|| compatibility_decomposition
) {
1477 *state
= U8_STATE_START
;
1487 saved_sz
= do_decomp(uv
, u8s
, u8s
, sz
,
1488 canonical_decomposition
, state
);
1492 for (i
= 0; i
< saved_sz
; ) {
1493 sz
= u8_number_of_bytes
[u8s
[i
]];
1495 comb_class
[last
] = combining_class(uv
,
1505 * Decomposition yields various Hangul related
1506 * states but not on combining marks. We need to
1507 * find out at here by checking on the last
1510 if (*state
== U8_STATE_START
) {
1511 if (comb_class
[last
- 1])
1512 *state
= U8_STATE_COMBINING_MARK
;
1519 sz
= u8_number_of_bytes
[*s
];
1522 * If this is an illegal character, an incomplete
1523 * character, or an 7-bit ASCII Starter character,
1524 * then we have collected a sequence; break and let
1525 * the next call deal with the two cases.
1527 * Note that this is okay only if you are using this
1528 * function with a fixed length string, not on
1529 * a buffer with multiple calls of one chunk at a time.
1533 } else if ((s
+ sz
) > slast
) {
1537 * If the previous character was a Hangul Jamo
1538 * and this character is a Hangul Jamo that
1539 * can be conjoined, we collect the Jamo.
1541 if (*s
== U8_HANGUL_JAMO_1ST_BYTE
) {
1542 U8_PUT_3BYTES_INTO_UTF32(u1
,
1543 *s
, *(s
+ 1), *(s
+ 2));
1545 if (U8_HANGUL_COMPOSABLE_L_V(*state
,
1548 *state
= U8_STATE_HANGUL_LV
;
1549 goto COLLECT_A_HANGUL
;
1552 if (U8_HANGUL_COMPOSABLE_LV_T(*state
,
1555 *state
= U8_STATE_HANGUL_LVT
;
1556 goto COLLECT_A_HANGUL
;
1561 * Regardless of whatever it was, if this is
1562 * a Starter, we don't collect the character
1563 * since that's a new start and we will deal
1564 * with it at the next time.
1566 i
= combining_class(uv
, s
, sz
);
1567 if (i
== U8_COMBINING_CLASS_STARTER
)
1571 * We know the current character is a combining
1572 * mark. If the previous character wasn't
1573 * a Starter (not Hangul) or a combining mark,
1574 * then, we don't collect this combining mark.
1576 if (*state
!= U8_STATE_START
&&
1577 *state
!= U8_STATE_COMBINING_MARK
)
1580 *state
= U8_STATE_COMBINING_MARK
;
1583 * If we collected a Starter and combining
1584 * marks up to 30, i.e., total 31 characters,
1585 * then, we terminate this degenerately long
1586 * combining sequence with a U+034F COMBINING
1587 * GRAPHEME JOINER (CGJ) which is 0xCD 0x8F in
1588 * UTF-8 and turn this into a Stream-Safe
1589 * Text. This will be extremely rare but
1592 * The following will also guarantee that
1593 * we are not writing more than 32 characters
1594 * plus a NULL at u8s[].
1596 if (last
>= U8_UPPER_LIMIT_IN_A_SEQ
) {
1598 *state
= U8_STATE_START
;
1599 comb_class
[last
] = 0;
1600 start
[last
] = saved_sz
;
1604 u8s
[saved_sz
++] = 0xCD;
1605 u8s
[saved_sz
++] = 0x8F;
1611 * Some combining marks also do decompose into
1612 * another combining mark or marks.
1614 if (*state
== U8_STATE_COMBINING_MARK
) {
1617 i
= do_decomp(uv
, uts
, s
, sz
,
1618 canonical_decomposition
, state
);
1619 for (j
= 0; j
< i
; ) {
1620 sz
= u8_number_of_bytes
[uts
[j
]];
1625 start
[last
] = saved_sz
+ j
;
1630 U8_UPPER_LIMIT_IN_A_SEQ
) {
1632 goto TURN_STREAM_SAFE
;
1637 *state
= U8_STATE_COMBINING_MARK
;
1641 for (i
= 0; i
< sz
; i
++)
1642 u8s
[saved_sz
++] = uts
[i
];
1644 comb_class
[last
] = i
;
1645 start
[last
] = saved_sz
;
1649 for (i
= 0; i
< sz
; i
++)
1650 u8s
[saved_sz
++] = *s
++;
1654 * If this is U+0345 COMBINING GREEK
1655 * YPOGEGRAMMENI (0xCD 0x85 in UTF-8), a.k.a.,
1656 * iota subscript, and need to be converted to
1657 * uppercase letter, convert it to U+0399 GREEK
1658 * CAPITAL LETTER IOTA (0xCE 0x99 in UTF-8),
1659 * i.e., convert to capital adscript form as
1660 * specified in the Unicode standard.
1662 * This is the only special case of (ambiguous)
1663 * case conversion at combining marks and
1664 * probably the standard will never have
1665 * anything similar like this in future.
1667 if (is_it_toupper
&& sz
>= 2 &&
1668 u8s
[saved_sz
- 2] == 0xCD &&
1669 u8s
[saved_sz
- 1] == 0x85) {
1670 u8s
[saved_sz
- 2] = 0xCE;
1671 u8s
[saved_sz
- 1] = 0x99;
1677 * Let's try to ensure a canonical ordering for the collected
1678 * combining marks. We do this only if we have collected
1679 * at least one more non-Starter. (The decomposition mapping
1680 * data tables have fully (and recursively) expanded and
1681 * canonically ordered decompositions.)
1683 * The U8_SWAP_COMB_MARKS() convenience macro has some
1684 * assumptions and we are meeting the assumptions.
1687 if (last
>= saved_last
) {
1688 for (i
= 0; i
< last
; i
++)
1689 for (j
= last
; j
> i
; j
--)
1690 if (comb_class
[j
] &&
1691 comb_class
[j
- 1] > comb_class
[j
]) {
1692 U8_SWAP_COMB_MARKS(j
- 1, j
);
1698 if (! canonical_composition
) {
1699 u8s
[saved_sz
] = '\0';
1704 * Now do the canonical composition. Note that we do this
1705 * only after a canonical or compatibility decomposition to
1706 * finish up NFC or NFKC.
1708 sz
= do_composition(uv
, u8s
, comb_class
, start
, disp
, last
,
1714 return ((size_t)sz
);
1718 * The do_norm_compare() function does string comparion based on Unicode
1719 * simple case mappings and Unicode Normalization definitions.
1721 * It does so by collecting a sequence of character at a time and comparing
1722 * the collected sequences from the strings.
1724 * The meanings on the return values are the same as the usual strcmp().
1727 do_norm_compare(size_t uv
, uchar_t
*s1
, uchar_t
*s2
, size_t n1
, size_t n2
,
1728 int flag
, int *errnum
)
1733 uchar_t u8s1
[U8_STREAM_SAFE_TEXT_MAX
+ 1];
1734 uchar_t u8s2
[U8_STREAM_SAFE_TEXT_MAX
+ 1];
1737 boolean_t is_it_toupper
;
1738 boolean_t is_it_tolower
;
1739 boolean_t canonical_decomposition
;
1740 boolean_t compatibility_decomposition
;
1741 boolean_t canonical_composition
;
1742 u8_normalization_states_t state
;
1747 is_it_toupper
= flag
& U8_TEXTPREP_TOUPPER
;
1748 is_it_tolower
= flag
& U8_TEXTPREP_TOLOWER
;
1749 canonical_decomposition
= flag
& U8_CANON_DECOMP
;
1750 compatibility_decomposition
= flag
& U8_COMPAT_DECOMP
;
1751 canonical_composition
= flag
& U8_CANON_COMP
;
1753 while (s1
< s1last
&& s2
< s2last
) {
1755 * If the current character is a 7-bit ASCII and the last
1756 * character, or, if the current character and the next
1757 * character are both some 7-bit ASCII characters then
1758 * we treat the current character as a sequence.
1760 * In any other cases, we need to call collect_a_seq().
1763 if (U8_ISASCII(*s1
) && ((s1
+ 1) >= s1last
||
1764 ((s1
+ 1) < s1last
&& U8_ISASCII(*(s1
+ 1))))) {
1766 u8s1
[0] = U8_ASCII_TOUPPER(*s1
);
1767 else if (is_it_tolower
)
1768 u8s1
[0] = U8_ASCII_TOLOWER(*s1
);
1775 state
= U8_STATE_START
;
1776 sz1
= collect_a_seq(uv
, u8s1
, &s1
, s1last
,
1777 is_it_toupper
, is_it_tolower
,
1778 canonical_decomposition
,
1779 compatibility_decomposition
,
1780 canonical_composition
, errnum
, &state
);
1783 if (U8_ISASCII(*s2
) && ((s2
+ 1) >= s2last
||
1784 ((s2
+ 1) < s2last
&& U8_ISASCII(*(s2
+ 1))))) {
1786 u8s2
[0] = U8_ASCII_TOUPPER(*s2
);
1787 else if (is_it_tolower
)
1788 u8s2
[0] = U8_ASCII_TOLOWER(*s2
);
1795 state
= U8_STATE_START
;
1796 sz2
= collect_a_seq(uv
, u8s2
, &s2
, s2last
,
1797 is_it_toupper
, is_it_tolower
,
1798 canonical_decomposition
,
1799 compatibility_decomposition
,
1800 canonical_composition
, errnum
, &state
);
1804 * Now compare the two characters. If they are the same,
1805 * we move on to the next character sequences.
1807 if (sz1
== 1 && sz2
== 1) {
1813 result
= strcmp((const char *)u8s1
, (const char *)u8s2
);
1820 * We compared until the end of either or both strings.
1822 * If we reached to or went over the ends for the both, that means
1823 * they are the same.
1825 * If we reached only one end, that means the other string has
1826 * something which then can be used to determine the return value.
1837 * The u8_strcmp() function compares two UTF-8 strings quite similar to
1838 * the strcmp(). For the comparison, however, Unicode Normalization specific
1839 * equivalency and Unicode simple case conversion mappings based equivalency
1840 * can be requested and checked against.
1843 u8_strcmp(const char *s1
, const char *s2
, size_t n
, int flag
, size_t uv
,
1853 * Check on the requested Unicode version, case conversion, and
1854 * normalization flag values.
1857 if (uv
> U8_UNICODE_LATEST
) {
1859 uv
= U8_UNICODE_LATEST
;
1863 flag
= U8_STRCMP_CS
;
1865 f
= flag
& (U8_STRCMP_CS
| U8_STRCMP_CI_UPPER
|
1866 U8_STRCMP_CI_LOWER
);
1868 flag
|= U8_STRCMP_CS
;
1869 } else if (f
!= U8_STRCMP_CS
&& f
!= U8_STRCMP_CI_UPPER
&&
1870 f
!= U8_STRCMP_CI_LOWER
) {
1872 flag
= U8_STRCMP_CS
;
1875 f
= flag
& (U8_CANON_DECOMP
| U8_COMPAT_DECOMP
| U8_CANON_COMP
);
1876 if (f
&& f
!= U8_STRCMP_NFD
&& f
!= U8_STRCMP_NFC
&&
1877 f
!= U8_STRCMP_NFKD
&& f
!= U8_STRCMP_NFKC
) {
1879 flag
= U8_STRCMP_CS
;
1883 if (flag
== U8_STRCMP_CS
) {
1884 return (n
== 0 ? strcmp(s1
, s2
) : strncmp(s1
, s2
, n
));
1897 * Simple case conversion can be done much faster and so we do
1898 * them separately here.
1900 if (flag
== U8_STRCMP_CI_UPPER
) {
1901 return (do_case_compare(uv
, (uchar_t
*)s1
, (uchar_t
*)s2
,
1902 n1
, n2
, B_TRUE
, errnum
));
1903 } else if (flag
== U8_STRCMP_CI_LOWER
) {
1904 return (do_case_compare(uv
, (uchar_t
*)s1
, (uchar_t
*)s2
,
1905 n1
, n2
, B_FALSE
, errnum
));
1908 return (do_norm_compare(uv
, (uchar_t
*)s1
, (uchar_t
*)s2
, n1
, n2
,
1913 u8_textprep_str(char *inarray
, size_t *inlen
, char *outarray
, size_t *outlen
,
1914 int flag
, size_t unicode_version
, int *errnum
)
1922 boolean_t do_not_ignore_null
;
1923 boolean_t do_not_ignore_invalid
;
1924 boolean_t is_it_toupper
;
1925 boolean_t is_it_tolower
;
1926 boolean_t canonical_decomposition
;
1927 boolean_t compatibility_decomposition
;
1928 boolean_t canonical_composition
;
1932 uchar_t u8s
[U8_STREAM_SAFE_TEXT_MAX
+ 1];
1933 u8_normalization_states_t state
;
1935 if (unicode_version
> U8_UNICODE_LATEST
) {
1937 return ((size_t)-1);
1940 f
= flag
& (U8_TEXTPREP_TOUPPER
| U8_TEXTPREP_TOLOWER
);
1941 if (f
== (U8_TEXTPREP_TOUPPER
| U8_TEXTPREP_TOLOWER
)) {
1943 return ((size_t)-1);
1946 f
= flag
& (U8_CANON_DECOMP
| U8_COMPAT_DECOMP
| U8_CANON_COMP
);
1947 if (f
&& f
!= U8_TEXTPREP_NFD
&& f
!= U8_TEXTPREP_NFC
&&
1948 f
!= U8_TEXTPREP_NFKD
&& f
!= U8_TEXTPREP_NFKC
) {
1950 return ((size_t)-1);
1953 if (inarray
== NULL
|| *inlen
== 0)
1956 if (outarray
== NULL
) {
1958 return ((size_t)-1);
1961 ib
= (uchar_t
*)inarray
;
1962 ob
= (uchar_t
*)outarray
;
1963 ibtail
= ib
+ *inlen
;
1964 obtail
= ob
+ *outlen
;
1966 do_not_ignore_null
= !(flag
& U8_TEXTPREP_IGNORE_NULL
);
1967 do_not_ignore_invalid
= !(flag
& U8_TEXTPREP_IGNORE_INVALID
);
1968 is_it_toupper
= flag
& U8_TEXTPREP_TOUPPER
;
1969 is_it_tolower
= flag
& U8_TEXTPREP_TOLOWER
;
1974 * If we don't have a normalization flag set, we do the simple case
1975 * conversion based text preparation separately below. Text
1976 * preparation involving Normalization will be done in the false task
1977 * block, again, separately since it will take much more time and
1978 * resource than doing simple case conversions.
1981 while (ib
< ibtail
) {
1982 if (*ib
== '\0' && do_not_ignore_null
)
1985 sz
= u8_number_of_bytes
[*ib
];
1988 if (do_not_ignore_invalid
) {
1990 ret_val
= (size_t)-1;
2001 ret_val
= (size_t)-1;
2006 *ob
= U8_ASCII_TOUPPER(*ib
);
2007 else if (is_it_tolower
)
2008 *ob
= U8_ASCII_TOLOWER(*ib
);
2013 } else if ((ib
+ sz
) > ibtail
) {
2014 if (do_not_ignore_invalid
) {
2016 ret_val
= (size_t)-1;
2020 if ((obtail
- ob
) < (ibtail
- ib
)) {
2022 ret_val
= (size_t)-1;
2027 * We treat the remaining incomplete character
2028 * bytes as a character.
2035 if (is_it_toupper
|| is_it_tolower
) {
2036 i
= do_case_conv(unicode_version
, u8s
,
2037 ib
, sz
, is_it_toupper
);
2039 if ((obtail
- ob
) < i
) {
2041 ret_val
= (size_t)-1;
2047 for (sz
= 0; sz
< i
; sz
++)
2050 if ((obtail
- ob
) < sz
) {
2052 ret_val
= (size_t)-1;
2056 for (i
= 0; i
< sz
; i
++)
2062 canonical_decomposition
= flag
& U8_CANON_DECOMP
;
2063 compatibility_decomposition
= flag
& U8_COMPAT_DECOMP
;
2064 canonical_composition
= flag
& U8_CANON_COMP
;
2066 while (ib
< ibtail
) {
2067 if (*ib
== '\0' && do_not_ignore_null
)
2071 * If the current character is a 7-bit ASCII
2072 * character and it is the last character, or,
2073 * if the current character is a 7-bit ASCII
2074 * character and the next character is also a 7-bit
2075 * ASCII character, then, we copy over this
2076 * character without going through collect_a_seq().
2078 * In any other cases, we need to look further with
2079 * the collect_a_seq() function.
2081 if (U8_ISASCII(*ib
) && ((ib
+ 1) >= ibtail
||
2082 ((ib
+ 1) < ibtail
&& U8_ISASCII(*(ib
+ 1))))) {
2085 ret_val
= (size_t)-1;
2090 *ob
= U8_ASCII_TOUPPER(*ib
);
2091 else if (is_it_tolower
)
2092 *ob
= U8_ASCII_TOLOWER(*ib
);
2099 state
= U8_STATE_START
;
2101 j
= collect_a_seq(unicode_version
, u8s
,
2105 canonical_decomposition
,
2106 compatibility_decomposition
,
2107 canonical_composition
,
2110 if (*errnum
&& do_not_ignore_invalid
) {
2111 ret_val
= (size_t)-1;
2115 if ((obtail
- ob
) < j
) {
2117 ret_val
= (size_t)-1;
2121 for (i
= 0; i
< j
; i
++)
2127 *inlen
= ibtail
- ib
;
2128 *outlen
= obtail
- ob
;
2133 #if defined(_KERNEL) && defined(HAVE_SPL)
2145 module_init(unicode_init
);
2146 module_exit(unicode_fini
);
2148 MODULE_DESCRIPTION("Unicode implementation");
2149 MODULE_AUTHOR(ZFS_META_AUTHOR
);
2150 MODULE_LICENSE(ZFS_META_LICENSE
);
2151 MODULE_VERSION(ZFS_META_VERSION
"-" ZFS_META_RELEASE
);
2153 EXPORT_SYMBOL(u8_validate
);
2154 EXPORT_SYMBOL(u8_strcmp
);
2155 EXPORT_SYMBOL(u8_textprep_str
);