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; | |
225 | uint32_t c32, c; | |
226 | const uint8_t *ss = (const uint8_t *)s; | |
227 | ||
228 | dynstr_ucs4_init(&d); | |
229 | while(ns > 0) { | |
230 | c = *ss++; | |
231 | --ns; | |
56fd389c RK |
232 | /* Acceptable UTF-8 is that which codes for Unicode Scalar Values |
233 | * (Unicode 5.0.0 s3.9 D76) | |
e5a5a138 RK |
234 | * |
235 | * 0xxxxxxx | |
236 | * 7 data bits gives 0x00 - 0x7F and all are acceptable | |
237 | * | |
238 | * 110xxxxx 10xxxxxx | |
239 | * 11 data bits gives 0x0000 - 0x07FF but only 0x0080 - 0x07FF acceptable | |
240 | * | |
241 | * 1110xxxx 10xxxxxx 10xxxxxx | |
242 | * 16 data bits gives 0x0000 - 0xFFFF but only 0x0800 - 0xFFFF acceptable | |
243 | * (and UTF-16 surrogates are not acceptable) | |
244 | * | |
245 | * 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx | |
246 | * 21 data bits gives 0x00000000 - 0x001FFFFF | |
247 | * but only 0x00010000 - 0x0010FFFF are acceptable | |
248 | * | |
56fd389c RK |
249 | * It is NOT always the case that the data bits in the first byte are |
250 | * always non-0 for the acceptable values, so we do a separate check after | |
251 | * decoding. | |
e5a5a138 RK |
252 | */ |
253 | if(c < 0x80) | |
254 | c32 = c; | |
255 | else if(c <= 0xDF) { | |
256 | if(ns < 1) goto error; | |
257 | c32 = c & 0x1F; | |
258 | c = *ss++; | |
259 | if((c & 0xC0) != 0x80) goto error; | |
260 | c32 = (c32 << 6) | (c & 0x3F); | |
261 | if(c32 < 0x80) goto error; | |
262 | } else if(c <= 0xEF) { | |
263 | if(ns < 2) goto error; | |
264 | c32 = c & 0x0F; | |
265 | c = *ss++; | |
266 | if((c & 0xC0) != 0x80) goto error; | |
267 | c32 = (c32 << 6) | (c & 0x3F); | |
268 | c = *ss++; | |
269 | if((c & 0xC0) != 0x80) goto error; | |
270 | c32 = (c32 << 6) | (c & 0x3F); | |
271 | if(c32 < 0x0800 || (c32 >= 0xD800 && c32 <= 0xDFFF)) goto error; | |
272 | } else if(c <= 0xF7) { | |
273 | if(ns < 3) goto error; | |
274 | c32 = c & 0x07; | |
275 | c = *ss++; | |
276 | if((c & 0xC0) != 0x80) goto error; | |
277 | c32 = (c32 << 6) | (c & 0x3F); | |
278 | c = *ss++; | |
279 | if((c & 0xC0) != 0x80) goto error; | |
280 | c32 = (c32 << 6) | (c & 0x3F); | |
281 | c = *ss++; | |
282 | if((c & 0xC0) != 0x80) goto error; | |
283 | c32 = (c32 << 6) | (c & 0x3F); | |
284 | if(c32 < 0x00010000 || c32 > 0x0010FFFF) goto error; | |
285 | } else | |
286 | goto error; | |
287 | dynstr_ucs4_append(&d, c32); | |
288 | } | |
289 | dynstr_ucs4_terminate(&d); | |
290 | if(ndp) | |
291 | *ndp = d.nvec; | |
292 | return d.vec; | |
293 | error: | |
294 | xfree(d.vec); | |
295 | return 0; | |
296 | } | |
297 | ||
18cda350 RK |
298 | /** @brief Test whether [s,s+ns) is valid UTF-8 |
299 | * @param s Start of string | |
300 | * @param ns Length of string | |
301 | * @return non-0 if @p s is valid UTF-8, 0 if it is not valid | |
302 | * | |
303 | * This function is intended to be much faster than calling utf8_to_utf32() and | |
304 | * throwing away the result. | |
305 | */ | |
306 | int utf8_valid(const char *s, size_t ns) { | |
307 | const uint8_t *ss = (const uint8_t *)s; | |
308 | while(ns > 0) { | |
309 | const struct unicode_utf8_row *const r = &unicode_utf8_valid[*ss]; | |
310 | if(r->count <= ns) { | |
311 | switch(r->count) { | |
312 | case 1: | |
313 | break; | |
314 | case 2: | |
315 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
316 | return 0; | |
317 | break; | |
318 | case 3: | |
319 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
320 | return 0; | |
321 | if(ss[2] < 0x80 || ss[2] > 0xBF) | |
322 | return 0; | |
323 | break; | |
324 | case 4: | |
325 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
326 | return 0; | |
327 | if(ss[2] < 0x80 || ss[2] > 0xBF) | |
328 | return 0; | |
329 | if(ss[3] < 0x80 || ss[3] > 0xBF) | |
330 | return 0; | |
331 | break; | |
332 | default: | |
333 | return 0; | |
334 | } | |
335 | } else | |
336 | return 0; | |
337 | ss += r->count; | |
338 | ns -= r->count; | |
339 | } | |
340 | return 1; | |
341 | } | |
342 | ||
092f426f RK |
343 | /*@}*/ |
344 | /** @defgroup utf32iterator UTF-32 string iterators */ | |
345 | /*@{*/ | |
346 | ||
347 | struct utf32_iterator_data { | |
348 | /** @brief Start of string */ | |
349 | const uint32_t *s; | |
350 | ||
351 | /** @brief Length of string */ | |
352 | size_t ns; | |
353 | ||
354 | /** @brief Current position */ | |
355 | size_t n; | |
356 | ||
357 | /** @brief Last two non-ignorable characters or (uint32_t)-1 | |
358 | * | |
359 | * last[1] is the non-Extend/Format character just before position @p n; | |
360 | * last[0] is the one just before that. | |
361 | * | |
362 | * Exception 1: if there is no such non-Extend/Format character then an | |
363 | * Extend/Format character is accepted instead. | |
364 | * | |
365 | * Exception 2: if there is no such character even taking that into account | |
366 | * the value is (uint32_t)-1. | |
367 | */ | |
368 | uint32_t last[2]; | |
369 | }; | |
370 | ||
371 | /** @brief Create a new iterator pointing at the start of a string | |
372 | * @param s Start of string | |
373 | * @param ns Length of string | |
374 | * @return New iterator | |
375 | */ | |
376 | utf32_iterator utf32_iterator_new(const uint32_t *s, size_t ns) { | |
377 | utf32_iterator it = xmalloc(sizeof *it); | |
378 | it->s = s; | |
379 | it->ns = ns; | |
380 | it->n = 0; | |
381 | it->last[0] = it->last[1] = -1; | |
382 | return it; | |
383 | } | |
384 | ||
385 | /** @brief Initialize an internal private iterator | |
386 | * @param it Iterator | |
387 | * @param s Start of string | |
388 | * @param ns Length of string | |
389 | * @param n Absolute position | |
390 | */ | |
391 | static void utf32__iterator_init(utf32_iterator it, | |
392 | const uint32_t *s, size_t ns, size_t n) { | |
393 | it->s = s; | |
394 | it->ns = ns; | |
395 | it->n = 0; | |
396 | it->last[0] = it->last[1] = -1; | |
397 | utf32_iterator_advance(it, n); | |
398 | } | |
399 | ||
400 | /** @brief Destroy an iterator | |
401 | * @param it Iterator | |
402 | */ | |
403 | void utf32_iterator_destroy(utf32_iterator it) { | |
404 | xfree(it); | |
405 | } | |
406 | ||
407 | /** @brief Find the current position of an interator | |
408 | * @param it Iterator | |
409 | */ | |
410 | size_t utf32_iterator_where(utf32_iterator it) { | |
411 | return it->n; | |
412 | } | |
413 | ||
414 | /** @brief Set an iterator's absolute position | |
415 | * @param it Iterator | |
416 | * @param n Absolute position | |
417 | * @return 0 on success, non-0 on error | |
418 | * | |
419 | * It is an error to position the iterator outside the string (but acceptable | |
420 | * to point it at the hypothetical post-final character). If an invalid value | |
421 | * of @p n is specified then the iterator is not changed. | |
422 | */ | |
423 | int utf32_iterator_set(utf32_iterator it, size_t n) { | |
424 | /* TODO figure out how far we must back up to be able to re-synchronize; see | |
425 | * UAX #29 s6.4. */ | |
426 | if(n > it->ns) | |
427 | return -1; | |
428 | if(n >= it->n) | |
429 | n -= it->n; | |
430 | else { | |
431 | it->n = 0; | |
432 | it->last[0] = it->last[1] = -1; | |
433 | } | |
434 | return utf32_iterator_advance(it, n); | |
435 | } | |
436 | ||
437 | /** @brief Advance an iterator | |
438 | * @param it Iterator | |
439 | * @param count Number of code points to advance by | |
440 | * @return 0 on success, non-0 on error | |
441 | * | |
442 | * It is an error to advance an iterator beyond the hypothetical post-final | |
443 | * character of the string. If an invalid value of @p n is specified then the | |
444 | * iterator is not changed. | |
445 | * | |
446 | * This function has O(n) time complexity: it works by advancing naively | |
447 | * forwards through the string. | |
448 | */ | |
449 | int utf32_iterator_advance(utf32_iterator it, size_t count) { | |
450 | if(count <= it->ns - it->n) { | |
451 | while(count > 0) { | |
452 | const uint32_t c = it->s[it->n]; | |
453 | const enum unicode_Word_Break wb = utf32__word_break(c); | |
454 | if(it->last[1] == (uint32_t)-1 | |
455 | || !utf32__boundary_ignorable(wb)) { | |
456 | it->last[0] = it->last[1]; | |
457 | it->last[1] = c; | |
458 | } | |
459 | ++it->n; | |
460 | --count; | |
461 | } | |
462 | return 0; | |
463 | } else | |
464 | return -1; | |
465 | } | |
466 | ||
467 | /** @brief Find the current code point | |
468 | * @param it Iterator | |
469 | * @return Current code point or 0 | |
470 | * | |
471 | * If the iterator points at the hypothetical post-final character of the | |
472 | * string then 0 is returned. NB that this doesn't mean that there aren't any | |
473 | * 0 code points inside the string! | |
474 | */ | |
475 | uint32_t utf32_iterator_code(utf32_iterator it) { | |
476 | if(it->n < it->ns) | |
477 | return it->s[it->n]; | |
478 | else | |
479 | return 0; | |
480 | } | |
481 | ||
482 | /** @brief Test for a grapheme boundary | |
483 | * @param it Iterator | |
484 | * @return Non-0 if pointing just after a grapheme boundary, otherwise 0 | |
485 | */ | |
486 | int utf32_iterator_grapheme_boundary(utf32_iterator it) { | |
487 | uint32_t before, after; | |
488 | enum unicode_Grapheme_Break gbbefore, gbafter; | |
489 | /* GB1 and GB2 */ | |
490 | if(it->n == 0 || it->n == it->ns) | |
491 | return 1; | |
492 | /* Now we know that s[n-1] and s[n] are safe to inspect */ | |
493 | /* GB3 */ | |
494 | before = it->s[it->n-1]; | |
495 | after = it->s[it->n]; | |
496 | if(before == 0x000D && after == 0x000A) | |
497 | return 0; | |
498 | gbbefore = utf32__grapheme_break(before); | |
499 | gbafter = utf32__grapheme_break(after); | |
500 | /* GB4 */ | |
501 | if(gbbefore == unicode_Grapheme_Break_Control | |
502 | || before == 0x000D | |
503 | || before == 0x000A) | |
504 | return 1; | |
505 | /* GB5 */ | |
506 | if(gbafter == unicode_Grapheme_Break_Control | |
507 | || after == 0x000D | |
508 | || after == 0x000A) | |
509 | return 1; | |
510 | /* GB6 */ | |
511 | if(gbbefore == unicode_Grapheme_Break_L | |
512 | && (gbafter == unicode_Grapheme_Break_L | |
513 | || gbafter == unicode_Grapheme_Break_V | |
514 | || gbafter == unicode_Grapheme_Break_LV | |
515 | || gbafter == unicode_Grapheme_Break_LVT)) | |
516 | return 0; | |
517 | /* GB7 */ | |
518 | if((gbbefore == unicode_Grapheme_Break_LV | |
519 | || gbbefore == unicode_Grapheme_Break_V) | |
520 | && (gbafter == unicode_Grapheme_Break_V | |
521 | || gbafter == unicode_Grapheme_Break_T)) | |
522 | return 0; | |
523 | /* GB8 */ | |
524 | if((gbbefore == unicode_Grapheme_Break_LVT | |
525 | || gbbefore == unicode_Grapheme_Break_T) | |
526 | && gbafter == unicode_Grapheme_Break_T) | |
527 | return 0; | |
528 | /* GB9 */ | |
529 | if(gbafter == unicode_Grapheme_Break_Extend) | |
530 | return 0; | |
531 | /* GB10 */ | |
532 | return 1; | |
533 | ||
534 | } | |
535 | ||
536 | /** @brief Test for a word boundary | |
537 | * @param it Iterator | |
538 | * @return Non-0 if pointing just after a word boundary, otherwise 0 | |
539 | */ | |
540 | int utf32_iterator_word_boundary(utf32_iterator it) { | |
541 | enum unicode_Word_Break twobefore, before, after, twoafter; | |
542 | size_t nn; | |
543 | ||
544 | /* WB1 and WB2 */ | |
545 | if(it->n == 0 || it->n == it->ns) | |
546 | return 1; | |
547 | /* WB3 */ | |
548 | if(it->s[it->n-1] == 0x000D && it->s[it->n] == 0x000A) | |
549 | return 0; | |
550 | /* WB4 */ | |
551 | /* (!Sep) x (Extend|Format) as in UAX #29 s6.2 */ | |
552 | if(utf32__sentence_break(it->s[it->n-1]) != unicode_Sentence_Break_Sep | |
553 | && utf32__boundary_ignorable(utf32__word_break(it->s[it->n]))) | |
554 | return 0; | |
555 | /* Gather the property values we'll need for the rest of the test taking the | |
556 | * s6.2 changes into account */ | |
557 | /* First we look at the code points after the proposed boundary */ | |
558 | nn = it->n; /* <it->ns */ | |
559 | after = utf32__word_break(it->s[nn++]); | |
560 | if(!utf32__boundary_ignorable(after)) { | |
561 | /* X (Extend|Format)* -> X */ | |
562 | while(nn < it->ns | |
563 | && utf32__boundary_ignorable(utf32__word_break(it->s[nn]))) | |
564 | ++nn; | |
565 | } | |
566 | /* It's possible now that nn=ns */ | |
567 | if(nn < it->ns) | |
568 | twoafter = utf32__word_break(it->s[nn]); | |
569 | else | |
570 | twoafter = unicode_Word_Break_Other; | |
571 | ||
572 | /* We've already recorded the non-ignorable code points before the proposed | |
573 | * boundary */ | |
574 | before = utf32__word_break(it->last[1]); | |
575 | twobefore = utf32__word_break(it->last[0]); | |
576 | ||
577 | /* WB5 */ | |
578 | if(before == unicode_Word_Break_ALetter | |
579 | && after == unicode_Word_Break_ALetter) | |
580 | return 0; | |
581 | /* WB6 */ | |
582 | if(before == unicode_Word_Break_ALetter | |
583 | && after == unicode_Word_Break_MidLetter | |
584 | && twoafter == unicode_Word_Break_ALetter) | |
585 | return 0; | |
586 | /* WB7 */ | |
587 | if(twobefore == unicode_Word_Break_ALetter | |
588 | && before == unicode_Word_Break_MidLetter | |
589 | && after == unicode_Word_Break_ALetter) | |
590 | return 0; | |
591 | /* WB8 */ | |
592 | if(before == unicode_Word_Break_Numeric | |
593 | && after == unicode_Word_Break_Numeric) | |
594 | return 0; | |
595 | /* WB9 */ | |
596 | if(before == unicode_Word_Break_ALetter | |
597 | && after == unicode_Word_Break_Numeric) | |
598 | return 0; | |
599 | /* WB10 */ | |
600 | if(before == unicode_Word_Break_Numeric | |
601 | && after == unicode_Word_Break_ALetter) | |
602 | return 0; | |
603 | /* WB11 */ | |
604 | if(twobefore == unicode_Word_Break_Numeric | |
605 | && before == unicode_Word_Break_MidNum | |
606 | && after == unicode_Word_Break_Numeric) | |
607 | return 0; | |
608 | /* WB12 */ | |
609 | if(before == unicode_Word_Break_Numeric | |
610 | && after == unicode_Word_Break_MidNum | |
611 | && twoafter == unicode_Word_Break_Numeric) | |
612 | return 0; | |
613 | /* WB13 */ | |
614 | if(before == unicode_Word_Break_Katakana | |
615 | && after == unicode_Word_Break_Katakana) | |
616 | return 0; | |
617 | /* WB13a */ | |
618 | if((before == unicode_Word_Break_ALetter | |
619 | || before == unicode_Word_Break_Numeric | |
620 | || before == unicode_Word_Break_Katakana | |
621 | || before == unicode_Word_Break_ExtendNumLet) | |
622 | && after == unicode_Word_Break_ExtendNumLet) | |
623 | return 0; | |
624 | /* WB13b */ | |
625 | if(before == unicode_Word_Break_ExtendNumLet | |
626 | && (after == unicode_Word_Break_ALetter | |
627 | || after == unicode_Word_Break_Numeric | |
628 | || after == unicode_Word_Break_Katakana)) | |
629 | return 0; | |
630 | /* WB14 */ | |
631 | return 1; | |
632 | } | |
633 | ||
e5a5a138 RK |
634 | /*@}*/ |
635 | /** @defgroup utf32 Functions that operate on UTF-32 strings */ | |
636 | /*@{*/ | |
637 | ||
638 | /** @brief Return the length of a 0-terminated UTF-32 string | |
639 | * @param s Pointer to 0-terminated string | |
640 | * @return Length of string in code points (excluding terminator) | |
641 | * | |
56fd389c | 642 | * Unlike the conversion functions no validity checking is done on the string. |
e5a5a138 RK |
643 | */ |
644 | size_t utf32_len(const uint32_t *s) { | |
645 | const uint32_t *t = s; | |
646 | ||
647 | while(*t) | |
648 | ++t; | |
649 | return (size_t)(t - s); | |
650 | } | |
651 | ||
e5a5a138 RK |
652 | /** @brief Stably sort [s,s+ns) into descending order of combining class |
653 | * @param s Start of array | |
654 | * @param ns Number of elements, must be at least 1 | |
655 | * @param buffer Buffer of at least @p ns elements | |
656 | */ | |
657 | static void utf32__sort_ccc(uint32_t *s, size_t ns, uint32_t *buffer) { | |
658 | uint32_t *a, *b, *bp; | |
659 | size_t na, nb; | |
660 | ||
661 | switch(ns) { | |
662 | case 1: /* 1-element array is always sorted */ | |
663 | return; | |
664 | case 2: /* 2-element arrays are trivial to sort */ | |
665 | if(utf32__combining_class(s[0]) > utf32__combining_class(s[1])) { | |
666 | uint32_t tmp = s[0]; | |
667 | s[0] = s[1]; | |
668 | s[1] = tmp; | |
669 | } | |
670 | return; | |
671 | default: | |
672 | /* Partition the array */ | |
673 | na = ns / 2; | |
674 | nb = ns - na; | |
675 | a = s; | |
676 | b = s + na; | |
677 | /* Sort the two halves of the array */ | |
678 | utf32__sort_ccc(a, na, buffer); | |
679 | utf32__sort_ccc(b, nb, buffer); | |
680 | /* Merge them back into one, via the buffer */ | |
681 | bp = buffer; | |
682 | while(na > 0 && nb > 0) { | |
683 | /* We want descending order of combining class (hence <) | |
684 | * and we want stability within combining classes (hence <=) | |
685 | */ | |
686 | if(utf32__combining_class(*a) <= utf32__combining_class(*b)) { | |
687 | *bp++ = *a++; | |
688 | --na; | |
689 | } else { | |
690 | *bp++ = *b++; | |
691 | --nb; | |
692 | } | |
693 | } | |
694 | while(na > 0) { | |
695 | *bp++ = *a++; | |
696 | --na; | |
697 | } | |
698 | while(nb > 0) { | |
699 | *bp++ = *b++; | |
700 | --nb; | |
701 | } | |
702 | memcpy(s, buffer, ns * sizeof(uint32_t)); | |
703 | return; | |
704 | } | |
705 | } | |
706 | ||
707 | /** @brief Put combining characters into canonical order | |
708 | * @param s Pointer to UTF-32 string | |
709 | * @param ns Length of @p s | |
710 | * @return 0 on success, -1 on error | |
711 | * | |
56fd389c RK |
712 | * @p s is modified in-place. See Unicode 5.0 s3.11 for details of the |
713 | * ordering. | |
e5a5a138 | 714 | * |
56fd389c RK |
715 | * Currently we only support a maximum of 1024 combining characters after each |
716 | * base character. If this limit is exceeded then -1 is returned. | |
e5a5a138 RK |
717 | */ |
718 | static int utf32__canonical_ordering(uint32_t *s, size_t ns) { | |
719 | size_t nc; | |
720 | uint32_t buffer[1024]; | |
721 | ||
722 | /* The ordering amounts to a stable sort of each contiguous group of | |
723 | * characters with non-0 combining class. */ | |
724 | while(ns > 0) { | |
725 | /* Skip non-combining characters */ | |
726 | if(utf32__combining_class(*s) == 0) { | |
727 | ++s; | |
728 | --ns; | |
729 | continue; | |
730 | } | |
731 | /* We must now have at least one combining character; see how many | |
732 | * there are */ | |
733 | for(nc = 1; nc < ns && utf32__combining_class(s[nc]) != 0; ++nc) | |
734 | ; | |
735 | if(nc > 1024) | |
736 | return -1; | |
737 | /* Sort the array */ | |
738 | utf32__sort_ccc(s, nc, buffer); | |
739 | s += nc; | |
740 | ns -= nc; | |
741 | } | |
742 | return 0; | |
743 | } | |
744 | ||
745 | /* Magic numbers from UAX #15 s16 */ | |
746 | #define SBase 0xAC00 | |
747 | #define LBase 0x1100 | |
748 | #define VBase 0x1161 | |
749 | #define TBase 0x11A7 | |
750 | #define LCount 19 | |
751 | #define VCount 21 | |
752 | #define TCount 28 | |
753 | #define NCount (VCount * TCount) | |
754 | #define SCount (LCount * NCount) | |
755 | ||
756 | /** @brief Guts of the decomposition lookup functions */ | |
757 | #define utf32__decompose_one_generic(WHICH) do { \ | |
bcf9ed7f | 758 | const uint32_t *dc = utf32__unidata(c)->WHICH; \ |
e5a5a138 RK |
759 | if(dc) { \ |
760 | /* Found a canonical decomposition in the table */ \ | |
761 | while(*dc) \ | |
762 | utf32__decompose_one_##WHICH(d, *dc++); \ | |
763 | } else if(c >= SBase && c < SBase + SCount) { \ | |
764 | /* Mechanically decomposable Hangul syllable (UAX #15 s16) */ \ | |
765 | const uint32_t SIndex = c - SBase; \ | |
766 | const uint32_t L = LBase + SIndex / NCount; \ | |
767 | const uint32_t V = VBase + (SIndex % NCount) / TCount; \ | |
768 | const uint32_t T = TBase + SIndex % TCount; \ | |
769 | dynstr_ucs4_append(d, L); \ | |
770 | dynstr_ucs4_append(d, V); \ | |
771 | if(T != TBase) \ | |
772 | dynstr_ucs4_append(d, T); \ | |
773 | } else \ | |
774 | /* Equal to own canonical decomposition */ \ | |
775 | dynstr_ucs4_append(d, c); \ | |
776 | } while(0) | |
777 | ||
778 | /** @brief Recursively compute the canonical decomposition of @p c | |
779 | * @param d Dynamic string to store decomposition in | |
780 | * @param c Code point to decompose (must be a valid!) | |
781 | * @return 0 on success, -1 on error | |
782 | */ | |
783 | static void utf32__decompose_one_canon(struct dynstr_ucs4 *d, uint32_t c) { | |
784 | utf32__decompose_one_generic(canon); | |
785 | } | |
786 | ||
787 | /** @brief Recursively compute the compatibility decomposition of @p c | |
788 | * @param d Dynamic string to store decomposition in | |
789 | * @param c Code point to decompose (must be a valid!) | |
790 | * @return 0 on success, -1 on error | |
791 | */ | |
792 | static void utf32__decompose_one_compat(struct dynstr_ucs4 *d, uint32_t c) { | |
793 | utf32__decompose_one_generic(compat); | |
794 | } | |
795 | ||
796 | /** @brief Guts of the decomposition functions */ | |
797 | #define utf32__decompose_generic(WHICH) do { \ | |
798 | struct dynstr_ucs4 d; \ | |
799 | uint32_t c; \ | |
800 | \ | |
801 | dynstr_ucs4_init(&d); \ | |
802 | while(ns) { \ | |
803 | c = *s++; \ | |
56fd389c | 804 | if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF) \ |
e5a5a138 RK |
805 | goto error; \ |
806 | utf32__decompose_one_##WHICH(&d, c); \ | |
807 | --ns; \ | |
808 | } \ | |
809 | if(utf32__canonical_ordering(d.vec, d.nvec)) \ | |
810 | goto error; \ | |
811 | dynstr_ucs4_terminate(&d); \ | |
812 | if(ndp) \ | |
813 | *ndp = d.nvec; \ | |
814 | return d.vec; \ | |
815 | error: \ | |
816 | xfree(d.vec); \ | |
817 | return 0; \ | |
818 | } while(0) | |
819 | ||
820 | /** @brief Canonically decompose @p [s,s+ns) | |
821 | * @param s Pointer to string | |
822 | * @param ns Length of string | |
823 | * @param ndp Where to store length of result | |
824 | * @return Pointer to result string, or NULL | |
825 | * | |
826 | * Computes the canonical decomposition of a string and stably sorts combining | |
827 | * characters into canonical order. The result is in Normalization Form D and | |
828 | * (at the time of writing!) passes the NFD tests defined in Unicode 5.0's | |
829 | * NormalizationTest.txt. | |
830 | * | |
56fd389c | 831 | * Returns NULL if the string is not valid for either of the following reasons: |
e5a5a138 RK |
832 | * - it codes for a UTF-16 surrogate |
833 | * - it codes for a value outside the unicode code space | |
834 | */ | |
835 | uint32_t *utf32_decompose_canon(const uint32_t *s, size_t ns, size_t *ndp) { | |
836 | utf32__decompose_generic(canon); | |
837 | } | |
838 | ||
839 | /** @brief Compatibility decompose @p [s,s+ns) | |
840 | * @param s Pointer to string | |
841 | * @param ns Length of string | |
842 | * @param ndp Where to store length of result | |
843 | * @return Pointer to result string, or NULL | |
844 | * | |
845 | * Computes the compatibility decomposition of a string and stably sorts | |
846 | * combining characters into canonical order. The result is in Normalization | |
847 | * Form KD and (at the time of writing!) passes the NFKD tests defined in | |
848 | * Unicode 5.0's NormalizationTest.txt. | |
849 | * | |
56fd389c | 850 | * Returns NULL if the string is not valid for either of the following reasons: |
e5a5a138 RK |
851 | * - it codes for a UTF-16 surrogate |
852 | * - it codes for a value outside the unicode code space | |
853 | */ | |
854 | uint32_t *utf32_decompose_compat(const uint32_t *s, size_t ns, size_t *ndp) { | |
855 | utf32__decompose_generic(compat); | |
856 | } | |
857 | ||
56fd389c RK |
858 | /** @brief Single-character case-fold and decompose operation */ |
859 | #define utf32__casefold_one(WHICH) do { \ | |
bcf9ed7f | 860 | const uint32_t *cf = utf32__unidata(c)->casefold; \ |
56fd389c RK |
861 | if(cf) { \ |
862 | /* Found a case-fold mapping in the table */ \ | |
863 | while(*cf) \ | |
864 | utf32__decompose_one_##WHICH(&d, *cf++); \ | |
865 | } else \ | |
866 | utf32__decompose_one_##WHICH(&d, c); \ | |
867 | } while(0) | |
e5a5a138 RK |
868 | |
869 | /** @brief Case-fold @p [s,s+ns) | |
870 | * @param s Pointer to string | |
871 | * @param ns Length of string | |
872 | * @param ndp Where to store length of result | |
873 | * @return Pointer to result string, or NULL | |
874 | * | |
875 | * Case-fold the string at @p s according to full default case-folding rules | |
56fd389c | 876 | * (s3.13) for caseless matching. The result will be in NFD. |
e5a5a138 | 877 | * |
56fd389c | 878 | * Returns NULL if the string is not valid for either of the following reasons: |
e5a5a138 RK |
879 | * - it codes for a UTF-16 surrogate |
880 | * - it codes for a value outside the unicode code space | |
881 | */ | |
882 | uint32_t *utf32_casefold_canon(const uint32_t *s, size_t ns, size_t *ndp) { | |
883 | struct dynstr_ucs4 d; | |
884 | uint32_t c; | |
885 | size_t n; | |
886 | uint32_t *ss = 0; | |
887 | ||
888 | /* If the canonical decomposition of the string includes any combining | |
889 | * character that case-folds to a non-combining character then we must | |
890 | * normalize before we fold. In Unicode 5.0.0 this means 0345 COMBINING | |
891 | * GREEK YPOGEGRAMMENI in its decomposition and the various characters that | |
892 | * canonically decompose to it. */ | |
bcf9ed7f RK |
893 | for(n = 0; n < ns; ++n) |
894 | if(utf32__unidata(s[n])->flags & unicode_normalize_before_casefold) | |
e5a5a138 | 895 | break; |
e5a5a138 RK |
896 | if(n < ns) { |
897 | /* We need a preliminary decomposition */ | |
898 | if(!(ss = utf32_decompose_canon(s, ns, &ns))) | |
899 | return 0; | |
900 | s = ss; | |
901 | } | |
902 | dynstr_ucs4_init(&d); | |
903 | while(ns) { | |
904 | c = *s++; | |
56fd389c | 905 | if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF) |
e5a5a138 | 906 | goto error; |
56fd389c | 907 | utf32__casefold_one(canon); |
e5a5a138 RK |
908 | --ns; |
909 | } | |
910 | if(utf32__canonical_ordering(d.vec, d.nvec)) | |
911 | goto error; | |
912 | dynstr_ucs4_terminate(&d); | |
913 | if(ndp) | |
914 | *ndp = d.nvec; | |
915 | return d.vec; | |
916 | error: | |
917 | xfree(d.vec); | |
918 | xfree(ss); | |
919 | return 0; | |
920 | } | |
921 | ||
56fd389c RK |
922 | /** @brief Compatibilit case-fold @p [s,s+ns) |
923 | * @param s Pointer to string | |
924 | * @param ns Length of string | |
925 | * @param ndp Where to store length of result | |
926 | * @return Pointer to result string, or NULL | |
927 | * | |
928 | * Case-fold the string at @p s according to full default case-folding rules | |
929 | * (s3.13) for compatibility caseless matching. The result will be in NFKD. | |
930 | * | |
931 | * Returns NULL if the string is not valid for either of the following reasons: | |
932 | * - it codes for a UTF-16 surrogate | |
933 | * - it codes for a value outside the unicode code space | |
934 | */ | |
935 | uint32_t *utf32_casefold_compat(const uint32_t *s, size_t ns, size_t *ndp) { | |
936 | struct dynstr_ucs4 d; | |
937 | uint32_t c; | |
938 | size_t n; | |
939 | uint32_t *ss = 0; | |
940 | ||
bcf9ed7f RK |
941 | for(n = 0; n < ns; ++n) |
942 | if(utf32__unidata(s[n])->flags & unicode_normalize_before_casefold) | |
56fd389c | 943 | break; |
56fd389c RK |
944 | if(n < ns) { |
945 | /* We need a preliminary _canonical_ decomposition */ | |
946 | if(!(ss = utf32_decompose_canon(s, ns, &ns))) | |
947 | return 0; | |
948 | s = ss; | |
949 | } | |
950 | /* This computes NFKD(toCaseFold(s)) */ | |
951 | #define compat_casefold_middle() do { \ | |
952 | dynstr_ucs4_init(&d); \ | |
953 | while(ns) { \ | |
954 | c = *s++; \ | |
955 | if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF) \ | |
956 | goto error; \ | |
957 | utf32__casefold_one(compat); \ | |
958 | --ns; \ | |
959 | } \ | |
960 | if(utf32__canonical_ordering(d.vec, d.nvec)) \ | |
961 | goto error; \ | |
962 | } while(0) | |
963 | /* Do the inner (NFKD o toCaseFold) */ | |
964 | compat_casefold_middle(); | |
965 | /* We can do away with the NFD'd copy of the input now */ | |
966 | xfree(ss); | |
967 | s = ss = d.vec; | |
968 | ns = d.nvec; | |
969 | /* Do the outer (NFKD o toCaseFold) */ | |
970 | compat_casefold_middle(); | |
971 | /* That's all */ | |
972 | dynstr_ucs4_terminate(&d); | |
973 | if(ndp) | |
974 | *ndp = d.nvec; | |
975 | return d.vec; | |
976 | error: | |
977 | xfree(d.vec); | |
978 | xfree(ss); | |
979 | return 0; | |
980 | } | |
981 | ||
e5a5a138 RK |
982 | /** @brief Order a pair of UTF-32 strings |
983 | * @param a First 0-terminated string | |
984 | * @param b Second 0-terminated string | |
985 | * @return -1, 0 or 1 for a less than, equal to or greater than b | |
986 | * | |
987 | * "Comparable to strcmp() at its best." | |
988 | */ | |
989 | int utf32_cmp(const uint32_t *a, const uint32_t *b) { | |
990 | while(*a && *b && *a == *b) { | |
991 | ++a; | |
992 | ++b; | |
993 | } | |
994 | return *a < *b ? -1 : (*a > *b ? 1 : 0); | |
995 | } | |
996 | ||
35b651f0 RK |
997 | /** @brief Identify a grapheme cluster boundary |
998 | * @param s Start of string (must be NFD) | |
999 | * @param ns Length of string | |
1000 | * @param n Index within string (in [0,ns].) | |
1001 | * @return 1 at a grapheme cluster boundary, 0 otherwise | |
1002 | * | |
1003 | * This function identifies default grapheme cluster boundaries as described in | |
1004 | * UAX #29 s3. It returns 1 if @p n points at the code point just after a | |
1005 | * grapheme cluster boundary (including the hypothetical code point just after | |
1006 | * the end of the string). | |
35b651f0 | 1007 | */ |
1625e11a | 1008 | int utf32_is_grapheme_boundary(const uint32_t *s, size_t ns, size_t n) { |
092f426f | 1009 | struct utf32_iterator_data it[1]; |
35b651f0 | 1010 | |
092f426f RK |
1011 | utf32__iterator_init(it, s, ns, n); |
1012 | return utf32_iterator_grapheme_boundary(it); | |
0b7052da RK |
1013 | } |
1014 | ||
1015 | /** @brief Identify a word boundary | |
1016 | * @param s Start of string (must be NFD) | |
1017 | * @param ns Length of string | |
1018 | * @param n Index within string (in [0,ns].) | |
1019 | * @return 1 at a word boundary, 0 otherwise | |
1020 | * | |
1021 | * This function identifies default word boundaries as described in UAX #29 s4. | |
1022 | * It returns 1 if @p n points at the code point just after a word boundary | |
1023 | * (including the hypothetical code point just after the end of the string). | |
1024 | */ | |
1025 | int utf32_is_word_boundary(const uint32_t *s, size_t ns, size_t n) { | |
092f426f | 1026 | struct utf32_iterator_data it[1]; |
0b7052da | 1027 | |
092f426f RK |
1028 | utf32__iterator_init(it, s, ns, n); |
1029 | return utf32_iterator_word_boundary(it); | |
0b7052da RK |
1030 | } |
1031 | ||
e5a5a138 | 1032 | /*@}*/ |
349b7b74 | 1033 | /** @defgroup utf8 Functions that operate on UTF-8 strings */ |
e5a5a138 RK |
1034 | /*@{*/ |
1035 | ||
1036 | /** @brief Wrapper to transform a UTF-8 string using the UTF-32 function */ | |
1037 | #define utf8__transform(FN) do { \ | |
1038 | uint32_t *to32 = 0, *decomp32 = 0; \ | |
1039 | size_t nto32, ndecomp32; \ | |
1040 | char *decomp8 = 0; \ | |
1041 | \ | |
1042 | if(!(to32 = utf8_to_utf32(s, ns, &nto32))) goto error; \ | |
1043 | if(!(decomp32 = FN(to32, nto32, &ndecomp32))) goto error; \ | |
1044 | decomp8 = utf32_to_utf8(decomp32, ndecomp32, ndp); \ | |
1045 | error: \ | |
1046 | xfree(to32); \ | |
1047 | xfree(decomp32); \ | |
1048 | return decomp8; \ | |
1049 | } while(0) | |
1050 | ||
1051 | /** @brief Canonically decompose @p [s,s+ns) | |
1052 | * @param s Pointer to string | |
1053 | * @param ns Length of string | |
1054 | * @param ndp Where to store length of result | |
1055 | * @return Pointer to result string, or NULL | |
1056 | * | |
1057 | * Computes the canonical decomposition of a string and stably sorts combining | |
1058 | * characters into canonical order. The result is in Normalization Form D and | |
1059 | * (at the time of writing!) passes the NFD tests defined in Unicode 5.0's | |
1060 | * NormalizationTest.txt. | |
1061 | * | |
1062 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why | |
1063 | * this might be. | |
1064 | * | |
1065 | * See also utf32_decompose_canon(). | |
1066 | */ | |
1067 | char *utf8_decompose_canon(const char *s, size_t ns, size_t *ndp) { | |
1068 | utf8__transform(utf32_decompose_canon); | |
1069 | } | |
1070 | ||
1071 | /** @brief Compatibility decompose @p [s,s+ns) | |
1072 | * @param s Pointer to string | |
1073 | * @param ns Length of string | |
1074 | * @param ndp Where to store length of result | |
1075 | * @return Pointer to result string, or NULL | |
1076 | * | |
1077 | * Computes the compatibility decomposition of a string and stably sorts | |
1078 | * combining characters into canonical order. The result is in Normalization | |
1079 | * Form KD and (at the time of writing!) passes the NFKD tests defined in | |
1080 | * Unicode 5.0's NormalizationTest.txt. | |
1081 | * | |
1082 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why | |
1083 | * this might be. | |
1084 | * | |
1085 | * See also utf32_decompose_compat(). | |
1086 | */ | |
1087 | char *utf8_decompose_compat(const char *s, size_t ns, size_t *ndp) { | |
1088 | utf8__transform(utf32_decompose_compat); | |
1089 | } | |
1090 | ||
1091 | /** @brief Case-fold @p [s,s+ns) | |
1092 | * @param s Pointer to string | |
1093 | * @param ns Length of string | |
1094 | * @param ndp Where to store length of result | |
1095 | * @return Pointer to result string, or NULL | |
1096 | * | |
1097 | * Case-fold the string at @p s according to full default case-folding rules | |
1098 | * (s3.13). The result will be in NFD. | |
1099 | * | |
1100 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why | |
1101 | * this might be. | |
1102 | */ | |
1103 | char *utf8_casefold_canon(const char *s, size_t ns, size_t *ndp) { | |
1104 | utf8__transform(utf32_casefold_canon); | |
1105 | } | |
1106 | ||
1107 | /** @brief Compatibility case-fold @p [s,s+ns) | |
1108 | * @param s Pointer to string | |
1109 | * @param ns Length of string | |
1110 | * @param ndp Where to store length of result | |
1111 | * @return Pointer to result string, or NULL | |
1112 | * | |
1113 | * Case-fold the string at @p s according to full default case-folding rules | |
1114 | * (s3.13). The result will be in NFKD. | |
1115 | * | |
1116 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why | |
1117 | * this might be. | |
1118 | */ | |
e5a5a138 RK |
1119 | char *utf8_casefold_compat(const char *s, size_t ns, size_t *ndp) { |
1120 | utf8__transform(utf32_casefold_compat); | |
1121 | } | |
e5a5a138 RK |
1122 | |
1123 | /*@}*/ | |
1124 | ||
1125 | /* | |
1126 | Local Variables: | |
1127 | c-basic-offset:2 | |
1128 | comment-column:40 | |
1129 | fill-column:79 | |
1130 | indent-tabs-mode:nil | |
1131 | End: | |
1132 | */ |