Commit | Line | Data |
---|---|---|
e5a5a138 RK |
1 | /* |
2 | * This file is part of DisOrder | |
3 | * Copyright (C) 2007 Richard Kettlewell | |
4 | * | |
5 | * This program is free software; you can redistribute it and/or modify | |
6 | * it under the terms of the GNU General Public License as published by | |
7 | * the Free Software Foundation; either version 2 of the License, or | |
8 | * (at your option) any later version. | |
9 | * | |
10 | * This program is distributed in the hope that it will be useful, but | |
11 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
12 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
13 | * General Public License for more details. | |
14 | * | |
15 | * You should have received a copy of the GNU General Public License | |
16 | * along with this program; if not, write to the Free Software | |
17 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 | |
18 | * USA | |
19 | */ | |
20 | /** @file lib/unicode.c | |
21 | * @brief Unicode support functions | |
22 | * | |
23 | * Here by UTF-8 and UTF-8 we mean the encoding forms of those names (not the | |
35b651f0 RK |
24 | * encoding schemes). The primary encoding form is UTF-32 but convenience |
25 | * wrappers using UTF-8 are provided for a number of functions. | |
e5a5a138 RK |
26 | * |
27 | * The idea is that all the strings that hit the database will be in a | |
28 | * particular normalization form, and for the search and tags database | |
29 | * in case-folded form, so they can be naively compared within the | |
30 | * database code. | |
31 | * | |
32 | * As the code stands this guarantee is not well met! | |
33 | */ | |
34 | ||
35 | #include <config.h> | |
36 | #include "types.h" | |
37 | ||
38 | #include <string.h> | |
39 | #include <stdio.h> /* TODO */ | |
40 | ||
41 | #include "mem.h" | |
42 | #include "vector.h" | |
43 | #include "unicode.h" | |
44 | #include "unidata.h" | |
45 | ||
092f426f RK |
46 | /** @defgroup utf32props Unicode Code Point Properties */ |
47 | /*@{*/ | |
48 | ||
49 | static const struct unidata *utf32__unidata_hard(uint32_t c); | |
50 | ||
51 | /** @brief Find definition of code point @p c | |
52 | * @param c Code point | |
53 | * @return Pointer to @ref unidata structure for @p c | |
54 | * | |
55 | * @p c can be any 32-bit value, a sensible value will be returned regardless. | |
56 | * The returned pointer is NOT guaranteed to be unique to @p c. | |
57 | */ | |
58 | static inline const struct unidata *utf32__unidata(uint32_t c) { | |
59 | /* The bottom half of the table contains almost everything of interest | |
60 | * and we can just return the right thing straight away */ | |
61 | if(c < UNICODE_BREAK_START) | |
62 | return &unidata[c / UNICODE_MODULUS][c % UNICODE_MODULUS]; | |
63 | else | |
64 | return utf32__unidata_hard(c); | |
65 | } | |
66 | ||
67 | /** @brief Find definition of code point @p c | |
68 | * @param c Code point | |
69 | * @return Pointer to @ref unidata structure for @p c | |
70 | * | |
71 | * @p c can be any 32-bit value, a sensible value will be returned regardless. | |
72 | * The returned pointer is NOT guaranteed to be unique to @p c. | |
73 | * | |
74 | * Don't use this function (although it will work fine) - use utf32__unidata() | |
75 | * instead. | |
76 | */ | |
77 | static const struct unidata *utf32__unidata_hard(uint32_t c) { | |
78 | if(c < UNICODE_BREAK_START) | |
79 | return &unidata[c / UNICODE_MODULUS][c % UNICODE_MODULUS]; | |
80 | /* Within the break everything is unassigned */ | |
81 | if(c < UNICODE_BREAK_END) | |
82 | return utf32__unidata(0xFFFF); /* guaranteed to be Cn */ | |
83 | /* Planes 15 and 16 are (mostly) private use */ | |
84 | if((c >= 0xF0000 && c <= 0xFFFFD) | |
85 | || (c >= 0x100000 && c <= 0x10FFFD)) | |
86 | return utf32__unidata(0xE000); /* first Co code point */ | |
87 | /* Everything else above the break top is unassigned */ | |
88 | if(c >= UNICODE_BREAK_TOP) | |
89 | return utf32__unidata(0xFFFF); /* guaranteed to be Cn */ | |
90 | /* Currently the rest is language tags and variation selectors */ | |
91 | c -= (UNICODE_BREAK_END - UNICODE_BREAK_START); | |
92 | return &unidata[c / UNICODE_MODULUS][c % UNICODE_MODULUS]; | |
93 | } | |
94 | ||
95 | /** @brief Return the combining class of @p c | |
96 | * @param c Code point | |
97 | * @return Combining class of @p c | |
98 | * | |
99 | * @p c can be any 32-bit value, a sensible value will be returned regardless. | |
100 | */ | |
101 | static inline int utf32__combining_class(uint32_t c) { | |
102 | return utf32__unidata(c)->ccc; | |
103 | } | |
104 | ||
105 | /** @brief Return the General_Category value for @p c | |
106 | * @param Code point | |
107 | * @return General_Category property value | |
108 | * | |
109 | * @p c can be any 32-bit value, a sensible value will be returned regardless. | |
110 | */ | |
111 | static inline enum unicode_General_Category utf32__general_category(uint32_t c) { | |
112 | return utf32__unidata(c)->general_category; | |
113 | } | |
114 | ||
115 | /** @brief Determine Grapheme_Break property | |
116 | * @param c Code point | |
117 | * @return Grapheme_Break property value of @p c | |
118 | * | |
119 | * @p c can be any 32-bit value, a sensible value will be returned regardless. | |
120 | */ | |
121 | static inline enum unicode_Grapheme_Break utf32__grapheme_break(uint32_t c) { | |
122 | return utf32__unidata(c)->grapheme_break; | |
123 | } | |
124 | ||
125 | /** @brief Determine Word_Break property | |
126 | * @param c Code point | |
127 | * @return Word_Break property value of @p c | |
128 | * | |
129 | * @p c can be any 32-bit value, a sensible value will be returned regardless. | |
130 | */ | |
131 | static inline enum unicode_Word_Break utf32__word_break(uint32_t c) { | |
132 | return utf32__unidata(c)->word_break; | |
133 | } | |
134 | ||
135 | /** @brief Determine Sentence_Break property | |
136 | * @param c Code point | |
137 | * @return Word_Break property value of @p c | |
138 | * | |
139 | * @p c can be any 32-bit value, a sensible value will be returned regardless. | |
140 | */ | |
141 | static inline enum unicode_Sentence_Break utf32__sentence_break(uint32_t c) { | |
142 | return utf32__unidata(c)->sentence_break; | |
143 | } | |
144 | ||
145 | /** @brief Return true if @p c is ignorable for boundary specifications | |
146 | * @param wb Word break property value | |
147 | * @return non-0 if @p wb is unicode_Word_Break_Extend or unicode_Word_Break_Format | |
148 | */ | |
149 | static inline int utf32__boundary_ignorable(enum unicode_Word_Break wb) { | |
150 | return (wb == unicode_Word_Break_Extend | |
151 | || wb == unicode_Word_Break_Format); | |
152 | } | |
153 | ||
154 | /*@}*/ | |
e5a5a138 RK |
155 | /** @defgroup utftransform Functions that transform between different Unicode encoding forms */ |
156 | /*@{*/ | |
157 | ||
158 | /** @brief Convert UTF-32 to UTF-8 | |
159 | * @param s Source string | |
160 | * @param ns Length of source string in code points | |
161 | * @param ndp Where to store length of destination string (or NULL) | |
162 | * @return Newly allocated destination string or NULL on error | |
163 | * | |
56fd389c RK |
164 | * If the UTF-32 is not valid then NULL is returned. A UTF-32 code point is |
165 | * invalid if: | |
e5a5a138 RK |
166 | * - it codes for a UTF-16 surrogate |
167 | * - it codes for a value outside the unicode code space | |
168 | * | |
56fd389c RK |
169 | * The return value is always 0-terminated. The value returned via @p *ndp |
170 | * does not include the terminator. | |
e5a5a138 RK |
171 | */ |
172 | char *utf32_to_utf8(const uint32_t *s, size_t ns, size_t *ndp) { | |
173 | struct dynstr d; | |
174 | uint32_t c; | |
175 | ||
176 | dynstr_init(&d); | |
177 | while(ns > 0) { | |
178 | c = *s++; | |
179 | if(c < 0x80) | |
180 | dynstr_append(&d, c); | |
181 | else if(c < 0x0800) { | |
182 | dynstr_append(&d, 0xC0 | (c >> 6)); | |
183 | dynstr_append(&d, 0x80 | (c & 0x3F)); | |
184 | } else if(c < 0x10000) { | |
56fd389c | 185 | if(c >= 0xD800 && c <= 0xDFFF) |
e5a5a138 RK |
186 | goto error; |
187 | dynstr_append(&d, 0xE0 | (c >> 12)); | |
188 | dynstr_append(&d, 0x80 | ((c >> 6) & 0x3F)); | |
189 | dynstr_append(&d, 0x80 | (c & 0x3F)); | |
190 | } else if(c < 0x110000) { | |
191 | dynstr_append(&d, 0xF0 | (c >> 18)); | |
192 | dynstr_append(&d, 0x80 | ((c >> 12) & 0x3F)); | |
193 | dynstr_append(&d, 0x80 | ((c >> 6) & 0x3F)); | |
194 | dynstr_append(&d, 0x80 | (c & 0x3F)); | |
195 | } else | |
196 | goto error; | |
197 | --ns; | |
198 | } | |
199 | dynstr_terminate(&d); | |
200 | if(ndp) | |
201 | *ndp = d.nvec; | |
202 | return d.vec; | |
203 | error: | |
204 | xfree(d.vec); | |
205 | return 0; | |
206 | } | |
207 | ||
208 | /** @brief Convert UTF-8 to UTF-32 | |
209 | * @param s Source string | |
210 | * @param ns Length of source string in code points | |
211 | * @param ndp Where to store length of destination string (or NULL) | |
212 | * @return Newly allocated destination string or NULL | |
213 | * | |
56fd389c RK |
214 | * The return value is always 0-terminated. The value returned via @p *ndp |
215 | * does not include the terminator. | |
e5a5a138 RK |
216 | * |
217 | * If the UTF-8 is not valid then NULL is returned. A UTF-8 sequence | |
218 | * for a code point is invalid if: | |
219 | * - it is not the shortest possible sequence for the code point | |
220 | * - it codes for a UTF-16 surrogate | |
221 | * - it codes for a value outside the unicode code space | |
222 | */ | |
223 | uint32_t *utf8_to_utf32(const char *s, size_t ns, size_t *ndp) { | |
224 | struct dynstr_ucs4 d; | |
32b158f2 | 225 | uint32_t c32; |
e5a5a138 | 226 | const uint8_t *ss = (const uint8_t *)s; |
32b158f2 | 227 | int n; |
e5a5a138 RK |
228 | |
229 | dynstr_ucs4_init(&d); | |
230 | while(ns > 0) { | |
32b158f2 RK |
231 | const struct unicode_utf8_row *const r = &unicode_utf8_valid[*ss]; |
232 | if(r->count <= ns) { | |
233 | switch(r->count) { | |
234 | case 1: | |
235 | c32 = *ss; | |
236 | break; | |
237 | case 2: | |
238 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
239 | goto error; | |
240 | c32 = *ss & 0x1F; | |
241 | break; | |
242 | case 3: | |
243 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
244 | goto error; | |
245 | c32 = *ss & 0x0F; | |
246 | break; | |
247 | case 4: | |
248 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
249 | goto error; | |
250 | c32 = *ss & 0x07; | |
251 | break; | |
252 | default: | |
253 | goto error; | |
254 | } | |
e5a5a138 RK |
255 | } else |
256 | goto error; | |
32b158f2 RK |
257 | for(n = 1; n < r->count; ++n) { |
258 | if(ss[n] < 0x80 || ss[n] > 0xBF) | |
259 | goto error; | |
260 | c32 = (c32 << 6) | (ss[n] & 0x3F); | |
261 | } | |
e5a5a138 | 262 | dynstr_ucs4_append(&d, c32); |
32b158f2 RK |
263 | ss += r->count; |
264 | ns -= r->count; | |
e5a5a138 RK |
265 | } |
266 | dynstr_ucs4_terminate(&d); | |
267 | if(ndp) | |
268 | *ndp = d.nvec; | |
269 | return d.vec; | |
270 | error: | |
271 | xfree(d.vec); | |
272 | return 0; | |
273 | } | |
274 | ||
18cda350 RK |
275 | /** @brief Test whether [s,s+ns) is valid UTF-8 |
276 | * @param s Start of string | |
277 | * @param ns Length of string | |
278 | * @return non-0 if @p s is valid UTF-8, 0 if it is not valid | |
279 | * | |
280 | * This function is intended to be much faster than calling utf8_to_utf32() and | |
281 | * throwing away the result. | |
282 | */ | |
283 | int utf8_valid(const char *s, size_t ns) { | |
284 | const uint8_t *ss = (const uint8_t *)s; | |
285 | while(ns > 0) { | |
286 | const struct unicode_utf8_row *const r = &unicode_utf8_valid[*ss]; | |
287 | if(r->count <= ns) { | |
288 | switch(r->count) { | |
289 | case 1: | |
290 | break; | |
291 | case 2: | |
292 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
293 | return 0; | |
294 | break; | |
295 | case 3: | |
296 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
297 | return 0; | |
298 | if(ss[2] < 0x80 || ss[2] > 0xBF) | |
299 | return 0; | |
300 | break; | |
301 | case 4: | |
302 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
303 | return 0; | |
304 | if(ss[2] < 0x80 || ss[2] > 0xBF) | |
305 | return 0; | |
306 | if(ss[3] < 0x80 || ss[3] > 0xBF) | |
307 | return 0; | |
308 | break; | |
309 | default: | |
310 | return 0; | |
311 | } | |
312 | } else | |
313 | return 0; | |
314 | ss += r->count; | |
315 | ns -= r->count; | |
316 | } | |
317 | return 1; | |
318 | } | |
319 | ||
092f426f RK |
320 | /*@}*/ |
321 | /** @defgroup utf32iterator UTF-32 string iterators */ | |
322 | /*@{*/ | |
323 | ||
324 | struct utf32_iterator_data { | |
325 | /** @brief Start of string */ | |
326 | const uint32_t *s; | |
327 | ||
328 | /** @brief Length of string */ | |
329 | size_t ns; | |
330 | ||
331 | /** @brief Current position */ | |
332 | size_t n; | |
333 | ||
334 | /** @brief Last two non-ignorable characters or (uint32_t)-1 | |
335 | * | |
336 | * last[1] is the non-Extend/Format character just before position @p n; | |
337 | * last[0] is the one just before that. | |
338 | * | |
339 | * Exception 1: if there is no such non-Extend/Format character then an | |
340 | * Extend/Format character is accepted instead. | |
341 | * | |
342 | * Exception 2: if there is no such character even taking that into account | |
343 | * the value is (uint32_t)-1. | |
344 | */ | |
345 | uint32_t last[2]; | |
346 | }; | |
347 | ||
348 | /** @brief Create a new iterator pointing at the start of a string | |
349 | * @param s Start of string | |
350 | * @param ns Length of string | |
351 | * @return New iterator | |
352 | */ | |
353 | utf32_iterator utf32_iterator_new(const uint32_t *s, size_t ns) { | |
354 | utf32_iterator it = xmalloc(sizeof *it); | |
355 | it->s = s; | |
356 | it->ns = ns; | |
357 | it->n = 0; | |
358 | it->last[0] = it->last[1] = -1; | |
359 | return it; | |
360 | } | |
361 | ||
362 | /** @brief Initialize an internal private iterator | |
363 | * @param it Iterator | |
364 | * @param s Start of string | |
365 | * @param ns Length of string | |
366 | * @param n Absolute position | |
367 | */ | |
368 | static void utf32__iterator_init(utf32_iterator it, | |
369 | const uint32_t *s, size_t ns, size_t n) { | |
370 | it->s = s; | |
371 | it->ns = ns; | |
372 | it->n = 0; | |
373 | it->last[0] = it->last[1] = -1; | |
374 | utf32_iterator_advance(it, n); | |
375 | } | |
376 | ||
377 | /** @brief Destroy an iterator | |
378 | * @param it Iterator | |
379 | */ | |
380 | void utf32_iterator_destroy(utf32_iterator it) { | |
381 | xfree(it); | |
382 | } | |
383 | ||
384 | /** @brief Find the current position of an interator | |
385 | * @param it Iterator | |
386 | */ | |
387 | size_t utf32_iterator_where(utf32_iterator it) { | |
388 | return it->n; | |
389 | } | |
390 | ||
391 | /** @brief Set an iterator's absolute position | |
392 | * @param it Iterator | |
393 | * @param n Absolute position | |
394 | * @return 0 on success, non-0 on error | |
395 | * | |
396 | * It is an error to position the iterator outside the string (but acceptable | |
397 | * to point it at the hypothetical post-final character). If an invalid value | |
398 | * of @p n is specified then the iterator is not changed. | |
399 | */ | |
400 | int utf32_iterator_set(utf32_iterator it, size_t n) { | |
5617aaff RK |
401 | /* We can't just jump to position @p n; the @p last[] values will be wrong. |
402 | * What we need is to jump a bit behind @p n and then advance forward, | |
403 | * updating @p last[] along the way. How far back? We need to cross two | |
404 | * non-ignorable code points as we advance forwards, so we'd better pass two | |
405 | * such characters on the way back (if such are available). | |
406 | */ | |
407 | size_t m = n; | |
408 | int i; | |
409 | ||
410 | if(n > it->ns) /* range check */ | |
092f426f | 411 | return -1; |
5617aaff RK |
412 | for(i = 0; i < 2; ++i) |
413 | while(m > 0 | |
414 | && utf32__boundary_ignorable(utf32__word_break(it->s[m - 1]))) | |
415 | --m; | |
416 | it->n = m; | |
417 | return utf32_iterator_advance(it, n - m); | |
092f426f RK |
418 | } |
419 | ||
420 | /** @brief Advance an iterator | |
421 | * @param it Iterator | |
422 | * @param count Number of code points to advance by | |
423 | * @return 0 on success, non-0 on error | |
424 | * | |
425 | * It is an error to advance an iterator beyond the hypothetical post-final | |
426 | * character of the string. If an invalid value of @p n is specified then the | |
427 | * iterator is not changed. | |
428 | * | |
429 | * This function has O(n) time complexity: it works by advancing naively | |
430 | * forwards through the string. | |
431 | */ | |
432 | int utf32_iterator_advance(utf32_iterator it, size_t count) { | |
433 | if(count <= it->ns - it->n) { | |
434 | while(count > 0) { | |
435 | const uint32_t c = it->s[it->n]; | |
436 | const enum unicode_Word_Break wb = utf32__word_break(c); | |
437 | if(it->last[1] == (uint32_t)-1 | |
438 | || !utf32__boundary_ignorable(wb)) { | |
439 | it->last[0] = it->last[1]; | |
440 | it->last[1] = c; | |
441 | } | |
442 | ++it->n; | |
443 | --count; | |
444 | } | |
445 | return 0; | |
446 | } else | |
447 | return -1; | |
448 | } | |
449 | ||
450 | /** @brief Find the current code point | |
451 | * @param it Iterator | |
452 | * @return Current code point or 0 | |
453 | * | |
454 | * If the iterator points at the hypothetical post-final character of the | |
455 | * string then 0 is returned. NB that this doesn't mean that there aren't any | |
456 | * 0 code points inside the string! | |
457 | */ | |
458 | uint32_t utf32_iterator_code(utf32_iterator it) { | |
459 | if(it->n < it->ns) | |
460 | return it->s[it->n]; | |
461 | else | |
462 | return 0; | |
463 | } | |
464 | ||
465 | /** @brief Test for a grapheme boundary | |
466 | * @param it Iterator | |
467 | * @return Non-0 if pointing just after a grapheme boundary, otherwise 0 | |
468 | */ | |
469 | int utf32_iterator_grapheme_boundary(utf32_iterator it) { | |
470 | uint32_t before, after; | |
471 | enum unicode_Grapheme_Break gbbefore, gbafter; | |
472 | /* GB1 and GB2 */ | |
473 | if(it->n == 0 || it->n == it->ns) | |
474 | return 1; | |
475 | /* Now we know that s[n-1] and s[n] are safe to inspect */ | |
476 | /* GB3 */ | |
477 | before = it->s[it->n-1]; | |
478 | after = it->s[it->n]; | |
479 | if(before == 0x000D && after == 0x000A) | |
480 | return 0; | |
481 | gbbefore = utf32__grapheme_break(before); | |
482 | gbafter = utf32__grapheme_break(after); | |
483 | /* GB4 */ | |
484 | if(gbbefore == unicode_Grapheme_Break_Control | |
485 | || before == 0x000D | |
486 | || before == 0x000A) | |
487 | return 1; | |
488 | /* GB5 */ | |
489 | if(gbafter == unicode_Grapheme_Break_Control | |
490 | || after == 0x000D | |
491 | || after == 0x000A) | |
492 | return 1; | |
493 | /* GB6 */ | |
494 | if(gbbefore == unicode_Grapheme_Break_L | |
495 | && (gbafter == unicode_Grapheme_Break_L | |
496 | || gbafter == unicode_Grapheme_Break_V | |
497 | || gbafter == unicode_Grapheme_Break_LV | |
498 | || gbafter == unicode_Grapheme_Break_LVT)) | |
499 | return 0; | |
500 | /* GB7 */ | |
501 | if((gbbefore == unicode_Grapheme_Break_LV | |
502 | || gbbefore == unicode_Grapheme_Break_V) | |
503 | && (gbafter == unicode_Grapheme_Break_V | |
504 | || gbafter == unicode_Grapheme_Break_T)) | |
505 | return 0; | |
506 | /* GB8 */ | |
507 | if((gbbefore == unicode_Grapheme_Break_LVT | |
508 | || gbbefore == unicode_Grapheme_Break_T) | |
509 | && gbafter == unicode_Grapheme_Break_T) | |
510 | return 0; | |
511 | /* GB9 */ | |
512 | if(gbafter == unicode_Grapheme_Break_Extend) | |
513 | return 0; | |
514 | /* GB10 */ | |
515 | return 1; | |
516 | ||
517 | } | |
518 | ||
519 | /** @brief Test for a word boundary | |
520 | * @param it Iterator | |
521 | * @return Non-0 if pointing just after a word boundary, otherwise 0 | |
522 | */ | |
523 | int utf32_iterator_word_boundary(utf32_iterator it) { | |
524 | enum unicode_Word_Break twobefore, before, after, twoafter; | |
525 | size_t nn; | |
526 | ||
527 | /* WB1 and WB2 */ | |
528 | if(it->n == 0 || it->n == it->ns) | |
529 | return 1; | |
530 | /* WB3 */ | |
531 | if(it->s[it->n-1] == 0x000D && it->s[it->n] == 0x000A) | |
532 | return 0; | |
533 | /* WB4 */ | |
534 | /* (!Sep) x (Extend|Format) as in UAX #29 s6.2 */ | |
535 | if(utf32__sentence_break(it->s[it->n-1]) != unicode_Sentence_Break_Sep | |
536 | && utf32__boundary_ignorable(utf32__word_break(it->s[it->n]))) | |
537 | return 0; | |
538 | /* Gather the property values we'll need for the rest of the test taking the | |
539 | * s6.2 changes into account */ | |
540 | /* First we look at the code points after the proposed boundary */ | |
541 | nn = it->n; /* <it->ns */ | |
542 | after = utf32__word_break(it->s[nn++]); | |
543 | if(!utf32__boundary_ignorable(after)) { | |
544 | /* X (Extend|Format)* -> X */ | |
545 | while(nn < it->ns | |
546 | && utf32__boundary_ignorable(utf32__word_break(it->s[nn]))) | |
547 | ++nn; | |
548 | } | |
549 | /* It's possible now that nn=ns */ | |
550 | if(nn < it->ns) | |
551 | twoafter = utf32__word_break(it->s[nn]); | |
552 | else | |
553 | twoafter = unicode_Word_Break_Other; | |
554 | ||
555 | /* We've already recorded the non-ignorable code points before the proposed | |
556 | * boundary */ | |
557 | before = utf32__word_break(it->last[1]); | |
558 | twobefore = utf32__word_break(it->last[0]); | |
559 | ||
560 | /* WB5 */ | |
561 | if(before == unicode_Word_Break_ALetter | |
562 | && after == unicode_Word_Break_ALetter) | |
563 | return 0; | |
564 | /* WB6 */ | |
565 | if(before == unicode_Word_Break_ALetter | |
566 | && after == unicode_Word_Break_MidLetter | |
567 | && twoafter == unicode_Word_Break_ALetter) | |
568 | return 0; | |
569 | /* WB7 */ | |
570 | if(twobefore == unicode_Word_Break_ALetter | |
571 | && before == unicode_Word_Break_MidLetter | |
572 | && after == unicode_Word_Break_ALetter) | |
573 | return 0; | |
574 | /* WB8 */ | |
575 | if(before == unicode_Word_Break_Numeric | |
576 | && after == unicode_Word_Break_Numeric) | |
577 | return 0; | |
578 | /* WB9 */ | |
579 | if(before == unicode_Word_Break_ALetter | |
580 | && after == unicode_Word_Break_Numeric) | |
581 | return 0; | |
582 | /* WB10 */ | |
583 | if(before == unicode_Word_Break_Numeric | |
584 | && after == unicode_Word_Break_ALetter) | |
585 | return 0; | |
586 | /* WB11 */ | |
587 | if(twobefore == unicode_Word_Break_Numeric | |
588 | && before == unicode_Word_Break_MidNum | |
589 | && after == unicode_Word_Break_Numeric) | |
590 | return 0; | |
591 | /* WB12 */ | |
592 | if(before == unicode_Word_Break_Numeric | |
593 | && after == unicode_Word_Break_MidNum | |
594 | && twoafter == unicode_Word_Break_Numeric) | |
595 | return 0; | |
596 | /* WB13 */ | |
597 | if(before == unicode_Word_Break_Katakana | |
598 | && after == unicode_Word_Break_Katakana) | |
599 | return 0; | |
600 | /* WB13a */ | |
601 | if((before == unicode_Word_Break_ALetter | |
602 | || before == unicode_Word_Break_Numeric | |
603 | || before == unicode_Word_Break_Katakana | |
604 | || before == unicode_Word_Break_ExtendNumLet) | |
605 | && after == unicode_Word_Break_ExtendNumLet) | |
606 | return 0; | |
607 | /* WB13b */ | |
608 | if(before == unicode_Word_Break_ExtendNumLet | |
609 | && (after == unicode_Word_Break_ALetter | |
610 | || after == unicode_Word_Break_Numeric | |
611 | || after == unicode_Word_Break_Katakana)) | |
612 | return 0; | |
613 | /* WB14 */ | |
614 | return 1; | |
615 | } | |
616 | ||
e5a5a138 RK |
617 | /*@}*/ |
618 | /** @defgroup utf32 Functions that operate on UTF-32 strings */ | |
619 | /*@{*/ | |
620 | ||
621 | /** @brief Return the length of a 0-terminated UTF-32 string | |
622 | * @param s Pointer to 0-terminated string | |
623 | * @return Length of string in code points (excluding terminator) | |
624 | * | |
56fd389c | 625 | * Unlike the conversion functions no validity checking is done on the string. |
e5a5a138 RK |
626 | */ |
627 | size_t utf32_len(const uint32_t *s) { | |
628 | const uint32_t *t = s; | |
629 | ||
630 | while(*t) | |
631 | ++t; | |
632 | return (size_t)(t - s); | |
633 | } | |
634 | ||
e5a5a138 RK |
635 | /** @brief Stably sort [s,s+ns) into descending order of combining class |
636 | * @param s Start of array | |
637 | * @param ns Number of elements, must be at least 1 | |
638 | * @param buffer Buffer of at least @p ns elements | |
639 | */ | |
640 | static void utf32__sort_ccc(uint32_t *s, size_t ns, uint32_t *buffer) { | |
641 | uint32_t *a, *b, *bp; | |
642 | size_t na, nb; | |
643 | ||
644 | switch(ns) { | |
645 | case 1: /* 1-element array is always sorted */ | |
646 | return; | |
647 | case 2: /* 2-element arrays are trivial to sort */ | |
648 | if(utf32__combining_class(s[0]) > utf32__combining_class(s[1])) { | |
649 | uint32_t tmp = s[0]; | |
650 | s[0] = s[1]; | |
651 | s[1] = tmp; | |
652 | } | |
653 | return; | |
654 | default: | |
655 | /* Partition the array */ | |
656 | na = ns / 2; | |
657 | nb = ns - na; | |
658 | a = s; | |
659 | b = s + na; | |
660 | /* Sort the two halves of the array */ | |
661 | utf32__sort_ccc(a, na, buffer); | |
662 | utf32__sort_ccc(b, nb, buffer); | |
663 | /* Merge them back into one, via the buffer */ | |
664 | bp = buffer; | |
665 | while(na > 0 && nb > 0) { | |
666 | /* We want descending order of combining class (hence <) | |
667 | * and we want stability within combining classes (hence <=) | |
668 | */ | |
669 | if(utf32__combining_class(*a) <= utf32__combining_class(*b)) { | |
670 | *bp++ = *a++; | |
671 | --na; | |
672 | } else { | |
673 | *bp++ = *b++; | |
674 | --nb; | |
675 | } | |
676 | } | |
677 | while(na > 0) { | |
678 | *bp++ = *a++; | |
679 | --na; | |
680 | } | |
681 | while(nb > 0) { | |
682 | *bp++ = *b++; | |
683 | --nb; | |
684 | } | |
685 | memcpy(s, buffer, ns * sizeof(uint32_t)); | |
686 | return; | |
687 | } | |
688 | } | |
689 | ||
690 | /** @brief Put combining characters into canonical order | |
691 | * @param s Pointer to UTF-32 string | |
692 | * @param ns Length of @p s | |
693 | * @return 0 on success, -1 on error | |
694 | * | |
56fd389c RK |
695 | * @p s is modified in-place. See Unicode 5.0 s3.11 for details of the |
696 | * ordering. | |
e5a5a138 | 697 | * |
56fd389c RK |
698 | * Currently we only support a maximum of 1024 combining characters after each |
699 | * base character. If this limit is exceeded then -1 is returned. | |
e5a5a138 RK |
700 | */ |
701 | static int utf32__canonical_ordering(uint32_t *s, size_t ns) { | |
702 | size_t nc; | |
703 | uint32_t buffer[1024]; | |
704 | ||
705 | /* The ordering amounts to a stable sort of each contiguous group of | |
706 | * characters with non-0 combining class. */ | |
707 | while(ns > 0) { | |
708 | /* Skip non-combining characters */ | |
709 | if(utf32__combining_class(*s) == 0) { | |
710 | ++s; | |
711 | --ns; | |
712 | continue; | |
713 | } | |
714 | /* We must now have at least one combining character; see how many | |
715 | * there are */ | |
716 | for(nc = 1; nc < ns && utf32__combining_class(s[nc]) != 0; ++nc) | |
717 | ; | |
718 | if(nc > 1024) | |
719 | return -1; | |
720 | /* Sort the array */ | |
721 | utf32__sort_ccc(s, nc, buffer); | |
722 | s += nc; | |
723 | ns -= nc; | |
724 | } | |
725 | return 0; | |
726 | } | |
727 | ||
728 | /* Magic numbers from UAX #15 s16 */ | |
729 | #define SBase 0xAC00 | |
730 | #define LBase 0x1100 | |
731 | #define VBase 0x1161 | |
732 | #define TBase 0x11A7 | |
733 | #define LCount 19 | |
734 | #define VCount 21 | |
735 | #define TCount 28 | |
736 | #define NCount (VCount * TCount) | |
737 | #define SCount (LCount * NCount) | |
738 | ||
739 | /** @brief Guts of the decomposition lookup functions */ | |
740 | #define utf32__decompose_one_generic(WHICH) do { \ | |
bcf9ed7f | 741 | const uint32_t *dc = utf32__unidata(c)->WHICH; \ |
e5a5a138 RK |
742 | if(dc) { \ |
743 | /* Found a canonical decomposition in the table */ \ | |
744 | while(*dc) \ | |
745 | utf32__decompose_one_##WHICH(d, *dc++); \ | |
746 | } else if(c >= SBase && c < SBase + SCount) { \ | |
747 | /* Mechanically decomposable Hangul syllable (UAX #15 s16) */ \ | |
748 | const uint32_t SIndex = c - SBase; \ | |
749 | const uint32_t L = LBase + SIndex / NCount; \ | |
750 | const uint32_t V = VBase + (SIndex % NCount) / TCount; \ | |
751 | const uint32_t T = TBase + SIndex % TCount; \ | |
752 | dynstr_ucs4_append(d, L); \ | |
753 | dynstr_ucs4_append(d, V); \ | |
754 | if(T != TBase) \ | |
755 | dynstr_ucs4_append(d, T); \ | |
756 | } else \ | |
757 | /* Equal to own canonical decomposition */ \ | |
758 | dynstr_ucs4_append(d, c); \ | |
759 | } while(0) | |
760 | ||
761 | /** @brief Recursively compute the canonical decomposition of @p c | |
762 | * @param d Dynamic string to store decomposition in | |
763 | * @param c Code point to decompose (must be a valid!) | |
764 | * @return 0 on success, -1 on error | |
765 | */ | |
766 | static void utf32__decompose_one_canon(struct dynstr_ucs4 *d, uint32_t c) { | |
767 | utf32__decompose_one_generic(canon); | |
768 | } | |
769 | ||
770 | /** @brief Recursively compute the compatibility decomposition of @p c | |
771 | * @param d Dynamic string to store decomposition in | |
772 | * @param c Code point to decompose (must be a valid!) | |
773 | * @return 0 on success, -1 on error | |
774 | */ | |
775 | static void utf32__decompose_one_compat(struct dynstr_ucs4 *d, uint32_t c) { | |
776 | utf32__decompose_one_generic(compat); | |
777 | } | |
778 | ||
779 | /** @brief Guts of the decomposition functions */ | |
780 | #define utf32__decompose_generic(WHICH) do { \ | |
781 | struct dynstr_ucs4 d; \ | |
782 | uint32_t c; \ | |
783 | \ | |
784 | dynstr_ucs4_init(&d); \ | |
785 | while(ns) { \ | |
786 | c = *s++; \ | |
56fd389c | 787 | if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF) \ |
e5a5a138 RK |
788 | goto error; \ |
789 | utf32__decompose_one_##WHICH(&d, c); \ | |
790 | --ns; \ | |
791 | } \ | |
792 | if(utf32__canonical_ordering(d.vec, d.nvec)) \ | |
793 | goto error; \ | |
794 | dynstr_ucs4_terminate(&d); \ | |
795 | if(ndp) \ | |
796 | *ndp = d.nvec; \ | |
797 | return d.vec; \ | |
798 | error: \ | |
799 | xfree(d.vec); \ | |
800 | return 0; \ | |
801 | } while(0) | |
802 | ||
803 | /** @brief Canonically decompose @p [s,s+ns) | |
804 | * @param s Pointer to string | |
805 | * @param ns Length of string | |
806 | * @param ndp Where to store length of result | |
807 | * @return Pointer to result string, or NULL | |
808 | * | |
809 | * Computes the canonical decomposition of a string and stably sorts combining | |
810 | * characters into canonical order. The result is in Normalization Form D and | |
811 | * (at the time of writing!) passes the NFD tests defined in Unicode 5.0's | |
812 | * NormalizationTest.txt. | |
813 | * | |
56fd389c | 814 | * Returns NULL if the string is not valid for either of the following reasons: |
e5a5a138 RK |
815 | * - it codes for a UTF-16 surrogate |
816 | * - it codes for a value outside the unicode code space | |
817 | */ | |
818 | uint32_t *utf32_decompose_canon(const uint32_t *s, size_t ns, size_t *ndp) { | |
819 | utf32__decompose_generic(canon); | |
820 | } | |
821 | ||
822 | /** @brief Compatibility decompose @p [s,s+ns) | |
823 | * @param s Pointer to string | |
824 | * @param ns Length of string | |
825 | * @param ndp Where to store length of result | |
826 | * @return Pointer to result string, or NULL | |
827 | * | |
828 | * Computes the compatibility decomposition of a string and stably sorts | |
829 | * combining characters into canonical order. The result is in Normalization | |
830 | * Form KD and (at the time of writing!) passes the NFKD tests defined in | |
831 | * Unicode 5.0's NormalizationTest.txt. | |
832 | * | |
56fd389c | 833 | * Returns NULL if the string is not valid for either of the following reasons: |
e5a5a138 RK |
834 | * - it codes for a UTF-16 surrogate |
835 | * - it codes for a value outside the unicode code space | |
836 | */ | |
837 | uint32_t *utf32_decompose_compat(const uint32_t *s, size_t ns, size_t *ndp) { | |
838 | utf32__decompose_generic(compat); | |
839 | } | |
840 | ||
56fd389c RK |
841 | /** @brief Single-character case-fold and decompose operation */ |
842 | #define utf32__casefold_one(WHICH) do { \ | |
bcf9ed7f | 843 | const uint32_t *cf = utf32__unidata(c)->casefold; \ |
56fd389c RK |
844 | if(cf) { \ |
845 | /* Found a case-fold mapping in the table */ \ | |
846 | while(*cf) \ | |
847 | utf32__decompose_one_##WHICH(&d, *cf++); \ | |
848 | } else \ | |
849 | utf32__decompose_one_##WHICH(&d, c); \ | |
850 | } while(0) | |
e5a5a138 RK |
851 | |
852 | /** @brief Case-fold @p [s,s+ns) | |
853 | * @param s Pointer to string | |
854 | * @param ns Length of string | |
855 | * @param ndp Where to store length of result | |
856 | * @return Pointer to result string, or NULL | |
857 | * | |
858 | * Case-fold the string at @p s according to full default case-folding rules | |
56fd389c | 859 | * (s3.13) for caseless matching. The result will be in NFD. |
e5a5a138 | 860 | * |
56fd389c | 861 | * Returns NULL if the string is not valid for either of the following reasons: |
e5a5a138 RK |
862 | * - it codes for a UTF-16 surrogate |
863 | * - it codes for a value outside the unicode code space | |
864 | */ | |
865 | uint32_t *utf32_casefold_canon(const uint32_t *s, size_t ns, size_t *ndp) { | |
866 | struct dynstr_ucs4 d; | |
867 | uint32_t c; | |
868 | size_t n; | |
869 | uint32_t *ss = 0; | |
870 | ||
871 | /* If the canonical decomposition of the string includes any combining | |
872 | * character that case-folds to a non-combining character then we must | |
873 | * normalize before we fold. In Unicode 5.0.0 this means 0345 COMBINING | |
874 | * GREEK YPOGEGRAMMENI in its decomposition and the various characters that | |
875 | * canonically decompose to it. */ | |
bcf9ed7f RK |
876 | for(n = 0; n < ns; ++n) |
877 | if(utf32__unidata(s[n])->flags & unicode_normalize_before_casefold) | |
e5a5a138 | 878 | break; |
e5a5a138 RK |
879 | if(n < ns) { |
880 | /* We need a preliminary decomposition */ | |
881 | if(!(ss = utf32_decompose_canon(s, ns, &ns))) | |
882 | return 0; | |
883 | s = ss; | |
884 | } | |
885 | dynstr_ucs4_init(&d); | |
886 | while(ns) { | |
887 | c = *s++; | |
56fd389c | 888 | if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF) |
e5a5a138 | 889 | goto error; |
56fd389c | 890 | utf32__casefold_one(canon); |
e5a5a138 RK |
891 | --ns; |
892 | } | |
893 | if(utf32__canonical_ordering(d.vec, d.nvec)) | |
894 | goto error; | |
895 | dynstr_ucs4_terminate(&d); | |
896 | if(ndp) | |
897 | *ndp = d.nvec; | |
898 | return d.vec; | |
899 | error: | |
900 | xfree(d.vec); | |
901 | xfree(ss); | |
902 | return 0; | |
903 | } | |
904 | ||
56fd389c RK |
905 | /** @brief Compatibilit case-fold @p [s,s+ns) |
906 | * @param s Pointer to string | |
907 | * @param ns Length of string | |
908 | * @param ndp Where to store length of result | |
909 | * @return Pointer to result string, or NULL | |
910 | * | |
911 | * Case-fold the string at @p s according to full default case-folding rules | |
912 | * (s3.13) for compatibility caseless matching. The result will be in NFKD. | |
913 | * | |
914 | * Returns NULL if the string is not valid for either of the following reasons: | |
915 | * - it codes for a UTF-16 surrogate | |
916 | * - it codes for a value outside the unicode code space | |
917 | */ | |
918 | uint32_t *utf32_casefold_compat(const uint32_t *s, size_t ns, size_t *ndp) { | |
919 | struct dynstr_ucs4 d; | |
920 | uint32_t c; | |
921 | size_t n; | |
922 | uint32_t *ss = 0; | |
923 | ||
bcf9ed7f RK |
924 | for(n = 0; n < ns; ++n) |
925 | if(utf32__unidata(s[n])->flags & unicode_normalize_before_casefold) | |
56fd389c | 926 | break; |
56fd389c RK |
927 | if(n < ns) { |
928 | /* We need a preliminary _canonical_ decomposition */ | |
929 | if(!(ss = utf32_decompose_canon(s, ns, &ns))) | |
930 | return 0; | |
931 | s = ss; | |
932 | } | |
933 | /* This computes NFKD(toCaseFold(s)) */ | |
934 | #define compat_casefold_middle() do { \ | |
935 | dynstr_ucs4_init(&d); \ | |
936 | while(ns) { \ | |
937 | c = *s++; \ | |
938 | if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF) \ | |
939 | goto error; \ | |
940 | utf32__casefold_one(compat); \ | |
941 | --ns; \ | |
942 | } \ | |
943 | if(utf32__canonical_ordering(d.vec, d.nvec)) \ | |
944 | goto error; \ | |
945 | } while(0) | |
946 | /* Do the inner (NFKD o toCaseFold) */ | |
947 | compat_casefold_middle(); | |
948 | /* We can do away with the NFD'd copy of the input now */ | |
949 | xfree(ss); | |
950 | s = ss = d.vec; | |
951 | ns = d.nvec; | |
952 | /* Do the outer (NFKD o toCaseFold) */ | |
953 | compat_casefold_middle(); | |
954 | /* That's all */ | |
955 | dynstr_ucs4_terminate(&d); | |
956 | if(ndp) | |
957 | *ndp = d.nvec; | |
958 | return d.vec; | |
959 | error: | |
960 | xfree(d.vec); | |
961 | xfree(ss); | |
962 | return 0; | |
963 | } | |
964 | ||
e5a5a138 RK |
965 | /** @brief Order a pair of UTF-32 strings |
966 | * @param a First 0-terminated string | |
967 | * @param b Second 0-terminated string | |
968 | * @return -1, 0 or 1 for a less than, equal to or greater than b | |
969 | * | |
970 | * "Comparable to strcmp() at its best." | |
971 | */ | |
972 | int utf32_cmp(const uint32_t *a, const uint32_t *b) { | |
973 | while(*a && *b && *a == *b) { | |
974 | ++a; | |
975 | ++b; | |
976 | } | |
977 | return *a < *b ? -1 : (*a > *b ? 1 : 0); | |
978 | } | |
979 | ||
35b651f0 RK |
980 | /** @brief Identify a grapheme cluster boundary |
981 | * @param s Start of string (must be NFD) | |
982 | * @param ns Length of string | |
983 | * @param n Index within string (in [0,ns].) | |
984 | * @return 1 at a grapheme cluster boundary, 0 otherwise | |
985 | * | |
986 | * This function identifies default grapheme cluster boundaries as described in | |
987 | * UAX #29 s3. It returns 1 if @p n points at the code point just after a | |
988 | * grapheme cluster boundary (including the hypothetical code point just after | |
989 | * the end of the string). | |
35b651f0 | 990 | */ |
1625e11a | 991 | int utf32_is_grapheme_boundary(const uint32_t *s, size_t ns, size_t n) { |
092f426f | 992 | struct utf32_iterator_data it[1]; |
35b651f0 | 993 | |
092f426f RK |
994 | utf32__iterator_init(it, s, ns, n); |
995 | return utf32_iterator_grapheme_boundary(it); | |
0b7052da RK |
996 | } |
997 | ||
998 | /** @brief Identify a word boundary | |
999 | * @param s Start of string (must be NFD) | |
1000 | * @param ns Length of string | |
1001 | * @param n Index within string (in [0,ns].) | |
1002 | * @return 1 at a word boundary, 0 otherwise | |
1003 | * | |
1004 | * This function identifies default word boundaries as described in UAX #29 s4. | |
1005 | * It returns 1 if @p n points at the code point just after a word boundary | |
1006 | * (including the hypothetical code point just after the end of the string). | |
1007 | */ | |
1008 | int utf32_is_word_boundary(const uint32_t *s, size_t ns, size_t n) { | |
092f426f | 1009 | struct utf32_iterator_data it[1]; |
0b7052da | 1010 | |
092f426f RK |
1011 | utf32__iterator_init(it, s, ns, n); |
1012 | return utf32_iterator_word_boundary(it); | |
0b7052da RK |
1013 | } |
1014 | ||
e5a5a138 | 1015 | /*@}*/ |
349b7b74 | 1016 | /** @defgroup utf8 Functions that operate on UTF-8 strings */ |
e5a5a138 RK |
1017 | /*@{*/ |
1018 | ||
1019 | /** @brief Wrapper to transform a UTF-8 string using the UTF-32 function */ | |
1020 | #define utf8__transform(FN) do { \ | |
1021 | uint32_t *to32 = 0, *decomp32 = 0; \ | |
1022 | size_t nto32, ndecomp32; \ | |
1023 | char *decomp8 = 0; \ | |
1024 | \ | |
1025 | if(!(to32 = utf8_to_utf32(s, ns, &nto32))) goto error; \ | |
1026 | if(!(decomp32 = FN(to32, nto32, &ndecomp32))) goto error; \ | |
1027 | decomp8 = utf32_to_utf8(decomp32, ndecomp32, ndp); \ | |
1028 | error: \ | |
1029 | xfree(to32); \ | |
1030 | xfree(decomp32); \ | |
1031 | return decomp8; \ | |
1032 | } while(0) | |
1033 | ||
1034 | /** @brief Canonically decompose @p [s,s+ns) | |
1035 | * @param s Pointer to string | |
1036 | * @param ns Length of string | |
1037 | * @param ndp Where to store length of result | |
1038 | * @return Pointer to result string, or NULL | |
1039 | * | |
1040 | * Computes the canonical decomposition of a string and stably sorts combining | |
1041 | * characters into canonical order. The result is in Normalization Form D and | |
1042 | * (at the time of writing!) passes the NFD tests defined in Unicode 5.0's | |
1043 | * NormalizationTest.txt. | |
1044 | * | |
1045 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why | |
1046 | * this might be. | |
1047 | * | |
1048 | * See also utf32_decompose_canon(). | |
1049 | */ | |
1050 | char *utf8_decompose_canon(const char *s, size_t ns, size_t *ndp) { | |
1051 | utf8__transform(utf32_decompose_canon); | |
1052 | } | |
1053 | ||
1054 | /** @brief Compatibility decompose @p [s,s+ns) | |
1055 | * @param s Pointer to string | |
1056 | * @param ns Length of string | |
1057 | * @param ndp Where to store length of result | |
1058 | * @return Pointer to result string, or NULL | |
1059 | * | |
1060 | * Computes the compatibility decomposition of a string and stably sorts | |
1061 | * combining characters into canonical order. The result is in Normalization | |
1062 | * Form KD and (at the time of writing!) passes the NFKD tests defined in | |
1063 | * Unicode 5.0's NormalizationTest.txt. | |
1064 | * | |
1065 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why | |
1066 | * this might be. | |
1067 | * | |
1068 | * See also utf32_decompose_compat(). | |
1069 | */ | |
1070 | char *utf8_decompose_compat(const char *s, size_t ns, size_t *ndp) { | |
1071 | utf8__transform(utf32_decompose_compat); | |
1072 | } | |
1073 | ||
1074 | /** @brief Case-fold @p [s,s+ns) | |
1075 | * @param s Pointer to string | |
1076 | * @param ns Length of string | |
1077 | * @param ndp Where to store length of result | |
1078 | * @return Pointer to result string, or NULL | |
1079 | * | |
1080 | * Case-fold the string at @p s according to full default case-folding rules | |
1081 | * (s3.13). The result will be in NFD. | |
1082 | * | |
1083 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why | |
1084 | * this might be. | |
1085 | */ | |
1086 | char *utf8_casefold_canon(const char *s, size_t ns, size_t *ndp) { | |
1087 | utf8__transform(utf32_casefold_canon); | |
1088 | } | |
1089 | ||
1090 | /** @brief Compatibility case-fold @p [s,s+ns) | |
1091 | * @param s Pointer to string | |
1092 | * @param ns Length of string | |
1093 | * @param ndp Where to store length of result | |
1094 | * @return Pointer to result string, or NULL | |
1095 | * | |
1096 | * Case-fold the string at @p s according to full default case-folding rules | |
1097 | * (s3.13). The result will be in NFKD. | |
1098 | * | |
1099 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why | |
1100 | * this might be. | |
1101 | */ | |
e5a5a138 RK |
1102 | char *utf8_casefold_compat(const char *s, size_t ns, size_t *ndp) { |
1103 | utf8__transform(utf32_casefold_compat); | |
1104 | } | |
e5a5a138 RK |
1105 | |
1106 | /*@}*/ | |
1107 | ||
1108 | /* | |
1109 | Local Variables: | |
1110 | c-basic-offset:2 | |
1111 | comment-column:40 | |
1112 | fill-column:79 | |
1113 | indent-tabs-mode:nil | |
1114 | End: | |
1115 | */ |