Commit | Line | Data |
---|---|---|
e5a5a138 RK |
1 | /* |
2 | * This file is part of DisOrder | |
bb5c7798 | 3 | * Copyright (C) 2007, 2009, 2013 Richard Kettlewell |
e5a5a138 | 4 | * |
e7eb3a27 | 5 | * This program is free software: you can redistribute it and/or modify |
e5a5a138 | 6 | * it under the terms of the GNU General Public License as published by |
e7eb3a27 | 7 | * the Free Software Foundation, either version 3 of the License, or |
e5a5a138 | 8 | * (at your option) any later version. |
e7eb3a27 RK |
9 | * |
10 | * This program is distributed in the hope that it will be useful, | |
11 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
12 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
13 | * GNU General Public License for more details. | |
14 | * | |
e5a5a138 | 15 | * You should have received a copy of the GNU General Public License |
e7eb3a27 | 16 | * along with this program. If not, see <http://www.gnu.org/licenses/>. |
e5a5a138 RK |
17 | */ |
18 | /** @file lib/unicode.c | |
19 | * @brief Unicode support functions | |
20 | * | |
21 | * Here by UTF-8 and UTF-8 we mean the encoding forms of those names (not the | |
35b651f0 RK |
22 | * encoding schemes). The primary encoding form is UTF-32 but convenience |
23 | * wrappers using UTF-8 are provided for a number of functions. | |
e5a5a138 RK |
24 | * |
25 | * The idea is that all the strings that hit the database will be in a | |
26 | * particular normalization form, and for the search and tags database | |
27 | * in case-folded form, so they can be naively compared within the | |
28 | * database code. | |
29 | * | |
30 | * As the code stands this guarantee is not well met! | |
0ae60b83 RK |
31 | * |
32 | * Subpages: | |
33 | * - @ref utf32props | |
34 | * - @ref utftransform | |
35 | * - @ref utf32iterator | |
36 | * - @ref utf32 | |
37 | * - @ref utf8 | |
e5a5a138 RK |
38 | */ |
39 | ||
05b75f8d | 40 | #include "common.h" |
e5a5a138 RK |
41 | |
42 | #include "mem.h" | |
43 | #include "vector.h" | |
44 | #include "unicode.h" | |
45 | #include "unidata.h" | |
46 | ||
092f426f RK |
47 | /** @defgroup utf32props Unicode Code Point Properties */ |
48 | /*@{*/ | |
49 | ||
50 | static const struct unidata *utf32__unidata_hard(uint32_t c); | |
51 | ||
52 | /** @brief Find definition of code point @p c | |
53 | * @param c Code point | |
54 | * @return Pointer to @ref unidata structure for @p c | |
55 | * | |
56 | * @p c can be any 32-bit value, a sensible value will be returned regardless. | |
57 | * The returned pointer is NOT guaranteed to be unique to @p c. | |
58 | */ | |
59 | static inline const struct unidata *utf32__unidata(uint32_t c) { | |
60 | /* The bottom half of the table contains almost everything of interest | |
61 | * and we can just return the right thing straight away */ | |
62 | if(c < UNICODE_BREAK_START) | |
63 | return &unidata[c / UNICODE_MODULUS][c % UNICODE_MODULUS]; | |
64 | else | |
65 | return utf32__unidata_hard(c); | |
66 | } | |
67 | ||
68 | /** @brief Find definition of code point @p c | |
69 | * @param c Code point | |
70 | * @return Pointer to @ref unidata structure for @p c | |
71 | * | |
72 | * @p c can be any 32-bit value, a sensible value will be returned regardless. | |
73 | * The returned pointer is NOT guaranteed to be unique to @p c. | |
74 | * | |
75 | * Don't use this function (although it will work fine) - use utf32__unidata() | |
76 | * instead. | |
77 | */ | |
78 | static const struct unidata *utf32__unidata_hard(uint32_t c) { | |
79 | if(c < UNICODE_BREAK_START) | |
80 | return &unidata[c / UNICODE_MODULUS][c % UNICODE_MODULUS]; | |
81 | /* Within the break everything is unassigned */ | |
82 | if(c < UNICODE_BREAK_END) | |
83 | return utf32__unidata(0xFFFF); /* guaranteed to be Cn */ | |
84 | /* Planes 15 and 16 are (mostly) private use */ | |
85 | if((c >= 0xF0000 && c <= 0xFFFFD) | |
86 | || (c >= 0x100000 && c <= 0x10FFFD)) | |
87 | return utf32__unidata(0xE000); /* first Co code point */ | |
88 | /* Everything else above the break top is unassigned */ | |
89 | if(c >= UNICODE_BREAK_TOP) | |
90 | return utf32__unidata(0xFFFF); /* guaranteed to be Cn */ | |
91 | /* Currently the rest is language tags and variation selectors */ | |
92 | c -= (UNICODE_BREAK_END - UNICODE_BREAK_START); | |
93 | return &unidata[c / UNICODE_MODULUS][c % UNICODE_MODULUS]; | |
94 | } | |
95 | ||
96 | /** @brief Return the combining class of @p c | |
97 | * @param c Code point | |
98 | * @return Combining class of @p c | |
99 | * | |
100 | * @p c can be any 32-bit value, a sensible value will be returned regardless. | |
101 | */ | |
102 | static inline int utf32__combining_class(uint32_t c) { | |
103 | return utf32__unidata(c)->ccc; | |
104 | } | |
105 | ||
3c82b504 RK |
106 | /** @brief Return the combining class of @p c |
107 | * @param c Code point | |
108 | * @return Combining class of @p c | |
109 | * | |
110 | * @p c can be any 32-bit value, a sensible value will be returned regardless. | |
111 | */ | |
112 | int utf32_combining_class(uint32_t c) { | |
113 | return utf32__combining_class(c); | |
114 | } | |
115 | ||
092f426f | 116 | /** @brief Return the General_Category value for @p c |
0ae60b83 | 117 | * @param c Code point |
092f426f RK |
118 | * @return General_Category property value |
119 | * | |
120 | * @p c can be any 32-bit value, a sensible value will be returned regardless. | |
121 | */ | |
122 | static inline enum unicode_General_Category utf32__general_category(uint32_t c) { | |
123 | return utf32__unidata(c)->general_category; | |
124 | } | |
125 | ||
126 | /** @brief Determine Grapheme_Break property | |
127 | * @param c Code point | |
128 | * @return Grapheme_Break property value of @p c | |
129 | * | |
130 | * @p c can be any 32-bit value, a sensible value will be returned regardless. | |
131 | */ | |
132 | static inline enum unicode_Grapheme_Break utf32__grapheme_break(uint32_t c) { | |
133 | return utf32__unidata(c)->grapheme_break; | |
134 | } | |
135 | ||
136 | /** @brief Determine Word_Break property | |
137 | * @param c Code point | |
138 | * @return Word_Break property value of @p c | |
139 | * | |
140 | * @p c can be any 32-bit value, a sensible value will be returned regardless. | |
141 | */ | |
142 | static inline enum unicode_Word_Break utf32__word_break(uint32_t c) { | |
143 | return utf32__unidata(c)->word_break; | |
144 | } | |
145 | ||
146 | /** @brief Determine Sentence_Break property | |
147 | * @param c Code point | |
148 | * @return Word_Break property value of @p c | |
149 | * | |
150 | * @p c can be any 32-bit value, a sensible value will be returned regardless. | |
151 | */ | |
152 | static inline enum unicode_Sentence_Break utf32__sentence_break(uint32_t c) { | |
153 | return utf32__unidata(c)->sentence_break; | |
154 | } | |
155 | ||
156 | /** @brief Return true if @p c is ignorable for boundary specifications | |
157 | * @param wb Word break property value | |
158 | * @return non-0 if @p wb is unicode_Word_Break_Extend or unicode_Word_Break_Format | |
159 | */ | |
160 | static inline int utf32__boundary_ignorable(enum unicode_Word_Break wb) { | |
161 | return (wb == unicode_Word_Break_Extend | |
162 | || wb == unicode_Word_Break_Format); | |
163 | } | |
164 | ||
f98fcddb RK |
165 | /** @brief Return the canonical decomposition of @p c |
166 | * @param c Code point | |
167 | * @return 0-terminated canonical decomposition, or 0 | |
168 | */ | |
169 | static inline const uint32_t *utf32__decomposition_canon(uint32_t c) { | |
170 | const struct unidata *const data = utf32__unidata(c); | |
171 | const uint32_t *const decomp = data->decomp; | |
172 | ||
173 | if(decomp && !(data->flags & unicode_compatibility_decomposition)) | |
174 | return decomp; | |
175 | else | |
176 | return 0; | |
177 | } | |
178 | ||
179 | /** @brief Return the compatibility decomposition of @p c | |
180 | * @param c Code point | |
181 | * @return 0-terminated decomposition, or 0 | |
182 | */ | |
183 | static inline const uint32_t *utf32__decomposition_compat(uint32_t c) { | |
184 | return utf32__unidata(c)->decomp; | |
185 | } | |
186 | ||
092f426f | 187 | /*@}*/ |
e5a5a138 RK |
188 | /** @defgroup utftransform Functions that transform between different Unicode encoding forms */ |
189 | /*@{*/ | |
190 | ||
191 | /** @brief Convert UTF-32 to UTF-8 | |
192 | * @param s Source string | |
193 | * @param ns Length of source string in code points | |
194 | * @param ndp Where to store length of destination string (or NULL) | |
195 | * @return Newly allocated destination string or NULL on error | |
196 | * | |
56fd389c RK |
197 | * If the UTF-32 is not valid then NULL is returned. A UTF-32 code point is |
198 | * invalid if: | |
e5a5a138 RK |
199 | * - it codes for a UTF-16 surrogate |
200 | * - it codes for a value outside the unicode code space | |
201 | * | |
56fd389c RK |
202 | * The return value is always 0-terminated. The value returned via @p *ndp |
203 | * does not include the terminator. | |
e5a5a138 RK |
204 | */ |
205 | char *utf32_to_utf8(const uint32_t *s, size_t ns, size_t *ndp) { | |
206 | struct dynstr d; | |
207 | uint32_t c; | |
208 | ||
209 | dynstr_init(&d); | |
210 | while(ns > 0) { | |
211 | c = *s++; | |
212 | if(c < 0x80) | |
213 | dynstr_append(&d, c); | |
214 | else if(c < 0x0800) { | |
215 | dynstr_append(&d, 0xC0 | (c >> 6)); | |
216 | dynstr_append(&d, 0x80 | (c & 0x3F)); | |
217 | } else if(c < 0x10000) { | |
56fd389c | 218 | if(c >= 0xD800 && c <= 0xDFFF) |
e5a5a138 RK |
219 | goto error; |
220 | dynstr_append(&d, 0xE0 | (c >> 12)); | |
221 | dynstr_append(&d, 0x80 | ((c >> 6) & 0x3F)); | |
222 | dynstr_append(&d, 0x80 | (c & 0x3F)); | |
223 | } else if(c < 0x110000) { | |
224 | dynstr_append(&d, 0xF0 | (c >> 18)); | |
225 | dynstr_append(&d, 0x80 | ((c >> 12) & 0x3F)); | |
226 | dynstr_append(&d, 0x80 | ((c >> 6) & 0x3F)); | |
227 | dynstr_append(&d, 0x80 | (c & 0x3F)); | |
228 | } else | |
229 | goto error; | |
230 | --ns; | |
231 | } | |
232 | dynstr_terminate(&d); | |
233 | if(ndp) | |
234 | *ndp = d.nvec; | |
235 | return d.vec; | |
236 | error: | |
237 | xfree(d.vec); | |
238 | return 0; | |
239 | } | |
240 | ||
241 | /** @brief Convert UTF-8 to UTF-32 | |
242 | * @param s Source string | |
243 | * @param ns Length of source string in code points | |
244 | * @param ndp Where to store length of destination string (or NULL) | |
f98fcddb | 245 | * @return Newly allocated destination string or NULL on error |
e5a5a138 | 246 | * |
56fd389c RK |
247 | * The return value is always 0-terminated. The value returned via @p *ndp |
248 | * does not include the terminator. | |
e5a5a138 RK |
249 | * |
250 | * If the UTF-8 is not valid then NULL is returned. A UTF-8 sequence | |
251 | * for a code point is invalid if: | |
252 | * - it is not the shortest possible sequence for the code point | |
253 | * - it codes for a UTF-16 surrogate | |
254 | * - it codes for a value outside the unicode code space | |
255 | */ | |
256 | uint32_t *utf8_to_utf32(const char *s, size_t ns, size_t *ndp) { | |
257 | struct dynstr_ucs4 d; | |
32b158f2 | 258 | uint32_t c32; |
e5a5a138 | 259 | const uint8_t *ss = (const uint8_t *)s; |
32b158f2 | 260 | int n; |
e5a5a138 RK |
261 | |
262 | dynstr_ucs4_init(&d); | |
263 | while(ns > 0) { | |
32b158f2 RK |
264 | const struct unicode_utf8_row *const r = &unicode_utf8_valid[*ss]; |
265 | if(r->count <= ns) { | |
266 | switch(r->count) { | |
267 | case 1: | |
268 | c32 = *ss; | |
269 | break; | |
270 | case 2: | |
271 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
272 | goto error; | |
273 | c32 = *ss & 0x1F; | |
274 | break; | |
275 | case 3: | |
276 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
277 | goto error; | |
278 | c32 = *ss & 0x0F; | |
279 | break; | |
280 | case 4: | |
281 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
282 | goto error; | |
283 | c32 = *ss & 0x07; | |
284 | break; | |
285 | default: | |
286 | goto error; | |
287 | } | |
e5a5a138 RK |
288 | } else |
289 | goto error; | |
32b158f2 RK |
290 | for(n = 1; n < r->count; ++n) { |
291 | if(ss[n] < 0x80 || ss[n] > 0xBF) | |
292 | goto error; | |
293 | c32 = (c32 << 6) | (ss[n] & 0x3F); | |
294 | } | |
e5a5a138 | 295 | dynstr_ucs4_append(&d, c32); |
32b158f2 RK |
296 | ss += r->count; |
297 | ns -= r->count; | |
e5a5a138 RK |
298 | } |
299 | dynstr_ucs4_terminate(&d); | |
300 | if(ndp) | |
301 | *ndp = d.nvec; | |
302 | return d.vec; | |
303 | error: | |
304 | xfree(d.vec); | |
305 | return 0; | |
306 | } | |
307 | ||
bb5c7798 RK |
308 | /** @brief Convert UTF-16 to UTF-8 |
309 | * @param s Source string | |
310 | * @param ns Length of source string in code points | |
311 | * @param ndp Where to store length of destination string (or NULL) | |
312 | * @return Newly allocated destination string or NULL on error | |
313 | * | |
314 | * If the UTF-16 is not valid then NULL is returned. A UTF-16 sequence t is | |
315 | * invalid if it contains an incomplete surrogate. | |
316 | * | |
317 | * The return value is always 0-terminated. The value returned via @p *ndp | |
318 | * does not include the terminator. | |
319 | */ | |
320 | char *utf16_to_utf8(const uint16_t *s, size_t ns, size_t *ndp) { | |
321 | struct dynstr d; | |
322 | uint32_t c; | |
323 | ||
324 | dynstr_init(&d); | |
325 | while(ns > 0) { | |
326 | c = *s++; | |
327 | --ns; | |
328 | if(c >= 0xD800 && c <= 0xDBFF) { | |
329 | if(ns && *s >= 0xDC00 && c <= 0xDFFF) | |
330 | c = ((c - 0xD800) << 10) + (*s++ - 0xDC00) + 0x10000; | |
331 | else | |
332 | goto error; | |
333 | } else if(c >= 0xDC00 && c <= 0xDFFF) | |
334 | goto error; | |
335 | if(c < 0x80) | |
336 | dynstr_append(&d, c); | |
337 | else if(c < 0x0800) { | |
338 | dynstr_append(&d, 0xC0 | (c >> 6)); | |
339 | dynstr_append(&d, 0x80 | (c & 0x3F)); | |
340 | } else if(c < 0x10000) { | |
341 | if(c >= 0xD800 && c <= 0xDFFF) | |
342 | goto error; | |
343 | dynstr_append(&d, 0xE0 | (c >> 12)); | |
344 | dynstr_append(&d, 0x80 | ((c >> 6) & 0x3F)); | |
345 | dynstr_append(&d, 0x80 | (c & 0x3F)); | |
346 | } else if(c < 0x110000) { | |
347 | dynstr_append(&d, 0xF0 | (c >> 18)); | |
348 | dynstr_append(&d, 0x80 | ((c >> 12) & 0x3F)); | |
349 | dynstr_append(&d, 0x80 | ((c >> 6) & 0x3F)); | |
350 | dynstr_append(&d, 0x80 | (c & 0x3F)); | |
351 | } else | |
352 | goto error; | |
353 | } | |
354 | dynstr_terminate(&d); | |
355 | if(ndp) | |
356 | *ndp = d.nvec; | |
357 | return d.vec; | |
358 | error: | |
359 | xfree(d.vec); | |
360 | return 0; | |
361 | } | |
362 | ||
363 | /** @brief Convert UTF-8 to UTF-16 | |
364 | * @param s Source string | |
365 | * @param ns Length of source string in code points | |
366 | * @param ndp Where to store length of destination string (or NULL) | |
367 | * @return Newly allocated destination string or NULL on error | |
368 | * | |
369 | * The return value is always 0-terminated. The value returned via @p *ndp | |
370 | * does not include the terminator. | |
371 | * | |
372 | * If the UTF-8 is not valid then NULL is returned. A UTF-8 sequence | |
373 | * for a code point is invalid if: | |
374 | * - it is not the shortest possible sequence for the code point | |
375 | * - it codes for a UTF-16 surrogate | |
376 | * - it codes for a value outside the unicode code space | |
377 | */ | |
378 | uint16_t *utf8_to_utf16(const char *s, size_t ns, size_t *ndp) { | |
379 | struct dynstr_utf16 d; | |
380 | uint32_t c32; | |
381 | const uint8_t *ss = (const uint8_t *)s; | |
382 | int n; | |
383 | ||
384 | dynstr_utf16_init(&d); | |
385 | while(ns > 0) { | |
386 | const struct unicode_utf8_row *const r = &unicode_utf8_valid[*ss]; | |
387 | if(r->count <= ns) { | |
388 | switch(r->count) { | |
389 | case 1: | |
390 | c32 = *ss; | |
391 | break; | |
392 | case 2: | |
393 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
394 | goto error; | |
395 | c32 = *ss & 0x1F; | |
396 | break; | |
397 | case 3: | |
398 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
399 | goto error; | |
400 | c32 = *ss & 0x0F; | |
401 | break; | |
402 | case 4: | |
403 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
404 | goto error; | |
405 | c32 = *ss & 0x07; | |
406 | break; | |
407 | default: | |
408 | goto error; | |
409 | } | |
410 | } else | |
411 | goto error; | |
412 | for(n = 1; n < r->count; ++n) { | |
413 | if(ss[n] < 0x80 || ss[n] > 0xBF) | |
414 | goto error; | |
415 | c32 = (c32 << 6) | (ss[n] & 0x3F); | |
416 | } | |
417 | if(c32 >= 0x10000) { | |
418 | c32 -= 0x10000; | |
419 | dynstr_utf16_append(&d, 0xD800 + (c32 >> 10)); | |
420 | dynstr_utf16_append(&d, 0xDC00 + (c32 & 0x03FF)); | |
421 | } else | |
422 | dynstr_utf16_append(&d, c32); | |
423 | ss += r->count; | |
424 | ns -= r->count; | |
425 | } | |
426 | dynstr_utf16_terminate(&d); | |
427 | if(ndp) | |
428 | *ndp = d.nvec; | |
429 | return d.vec; | |
430 | error: | |
431 | xfree(d.vec); | |
432 | return 0; | |
433 | } | |
434 | ||
18cda350 RK |
435 | /** @brief Test whether [s,s+ns) is valid UTF-8 |
436 | * @param s Start of string | |
437 | * @param ns Length of string | |
438 | * @return non-0 if @p s is valid UTF-8, 0 if it is not valid | |
439 | * | |
440 | * This function is intended to be much faster than calling utf8_to_utf32() and | |
441 | * throwing away the result. | |
442 | */ | |
443 | int utf8_valid(const char *s, size_t ns) { | |
444 | const uint8_t *ss = (const uint8_t *)s; | |
445 | while(ns > 0) { | |
446 | const struct unicode_utf8_row *const r = &unicode_utf8_valid[*ss]; | |
447 | if(r->count <= ns) { | |
448 | switch(r->count) { | |
449 | case 1: | |
450 | break; | |
451 | case 2: | |
452 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
453 | return 0; | |
454 | break; | |
455 | case 3: | |
456 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
457 | return 0; | |
458 | if(ss[2] < 0x80 || ss[2] > 0xBF) | |
459 | return 0; | |
460 | break; | |
461 | case 4: | |
462 | if(ss[1] < r->min2 || ss[1] > r->max2) | |
463 | return 0; | |
464 | if(ss[2] < 0x80 || ss[2] > 0xBF) | |
465 | return 0; | |
466 | if(ss[3] < 0x80 || ss[3] > 0xBF) | |
467 | return 0; | |
468 | break; | |
469 | default: | |
470 | return 0; | |
471 | } | |
472 | } else | |
473 | return 0; | |
474 | ss += r->count; | |
475 | ns -= r->count; | |
476 | } | |
477 | return 1; | |
478 | } | |
479 | ||
e5a5a138 | 480 | /*@}*/ |
092f426f RK |
481 | /** @defgroup utf32iterator UTF-32 string iterators */ |
482 | /*@{*/ | |
483 | ||
484 | struct utf32_iterator_data { | |
485 | /** @brief Start of string */ | |
486 | const uint32_t *s; | |
487 | ||
488 | /** @brief Length of string */ | |
489 | size_t ns; | |
490 | ||
491 | /** @brief Current position */ | |
492 | size_t n; | |
493 | ||
494 | /** @brief Last two non-ignorable characters or (uint32_t)-1 | |
495 | * | |
496 | * last[1] is the non-Extend/Format character just before position @p n; | |
497 | * last[0] is the one just before that. | |
498 | * | |
499 | * Exception 1: if there is no such non-Extend/Format character then an | |
500 | * Extend/Format character is accepted instead. | |
501 | * | |
502 | * Exception 2: if there is no such character even taking that into account | |
503 | * the value is (uint32_t)-1. | |
504 | */ | |
505 | uint32_t last[2]; | |
092f426f | 506 | |
c85b7022 RK |
507 | /** @brief Tailoring for Word_Break */ |
508 | unicode_property_tailor *word_break; | |
509 | }; | |
092f426f RK |
510 | |
511 | /** @brief Initialize an internal private iterator | |
512 | * @param it Iterator | |
513 | * @param s Start of string | |
514 | * @param ns Length of string | |
515 | * @param n Absolute position | |
516 | */ | |
517 | static void utf32__iterator_init(utf32_iterator it, | |
518 | const uint32_t *s, size_t ns, size_t n) { | |
519 | it->s = s; | |
520 | it->ns = ns; | |
521 | it->n = 0; | |
522 | it->last[0] = it->last[1] = -1; | |
c85b7022 | 523 | it->word_break = 0; |
b21a155c | 524 | utf32_iterator_set(it, n); |
092f426f RK |
525 | } |
526 | ||
c85b7022 RK |
527 | /** @brief Create a new iterator pointing at the start of a string |
528 | * @param s Start of string | |
529 | * @param ns Length of string | |
530 | * @return New iterator | |
531 | */ | |
532 | utf32_iterator utf32_iterator_new(const uint32_t *s, size_t ns) { | |
533 | utf32_iterator it = xmalloc(sizeof *it); | |
534 | utf32__iterator_init(it, s, ns, 0); | |
535 | return it; | |
536 | } | |
537 | ||
538 | /** @brief Tailor this iterator's interpretation of the Word_Break property. | |
539 | * @param it Iterator | |
540 | * @param pt Property tailor function or NULL | |
541 | * | |
542 | * After calling this the iterator will call @p pt to determine the Word_Break | |
543 | * property of each code point. If it returns -1 the default value will be | |
544 | * used otherwise the returned value will be used. | |
545 | * | |
546 | * @p pt can be NULL to revert to the default value of the property. | |
547 | * | |
548 | * It is safe to call this function at any time; the iterator's internal state | |
549 | * will be reset to suit the new tailoring. | |
550 | */ | |
551 | void utf32_iterator_tailor_word_break(utf32_iterator it, | |
552 | unicode_property_tailor *pt) { | |
553 | it->word_break = pt; | |
554 | utf32_iterator_set(it, it->n); | |
555 | } | |
556 | ||
557 | static inline enum unicode_Word_Break utf32__iterator_word_break(utf32_iterator it, | |
558 | uint32_t c) { | |
559 | if(!it->word_break) | |
560 | return utf32__word_break(c); | |
561 | else { | |
562 | const int t = it->word_break(c); | |
563 | ||
564 | if(t < 0) | |
565 | return utf32__word_break(c); | |
566 | else | |
567 | return t; | |
568 | } | |
569 | } | |
570 | ||
092f426f RK |
571 | /** @brief Destroy an iterator |
572 | * @param it Iterator | |
573 | */ | |
574 | void utf32_iterator_destroy(utf32_iterator it) { | |
575 | xfree(it); | |
576 | } | |
577 | ||
578 | /** @brief Find the current position of an interator | |
579 | * @param it Iterator | |
580 | */ | |
581 | size_t utf32_iterator_where(utf32_iterator it) { | |
582 | return it->n; | |
583 | } | |
584 | ||
585 | /** @brief Set an iterator's absolute position | |
586 | * @param it Iterator | |
587 | * @param n Absolute position | |
588 | * @return 0 on success, non-0 on error | |
589 | * | |
590 | * It is an error to position the iterator outside the string (but acceptable | |
591 | * to point it at the hypothetical post-final character). If an invalid value | |
592 | * of @p n is specified then the iterator is not changed. | |
f98fcddb RK |
593 | * |
594 | * This function works by backing up and then advancing to reconstruct the | |
595 | * iterator's internal state for position @p n. The worst case will be O(n) | |
596 | * time complexity (with a worse constant factor that utf32_iterator_advance()) | |
597 | * but the typical case is essentially constant-time. | |
092f426f RK |
598 | */ |
599 | int utf32_iterator_set(utf32_iterator it, size_t n) { | |
5617aaff RK |
600 | /* We can't just jump to position @p n; the @p last[] values will be wrong. |
601 | * What we need is to jump a bit behind @p n and then advance forward, | |
602 | * updating @p last[] along the way. How far back? We need to cross two | |
603 | * non-ignorable code points as we advance forwards, so we'd better pass two | |
604 | * such characters on the way back (if such are available). | |
605 | */ | |
b21a155c | 606 | size_t m; |
5617aaff RK |
607 | |
608 | if(n > it->ns) /* range check */ | |
092f426f | 609 | return -1; |
b21a155c RK |
610 | /* Walk backwards skipping ignorable code points */ |
611 | m = n; | |
c85b7022 RK |
612 | while(m > 0 |
613 | && (utf32__boundary_ignorable(utf32__iterator_word_break(it, | |
614 | it->s[m-1])))) | |
b21a155c RK |
615 | --m; |
616 | /* Either m=0 or s[m-1] is not ignorable */ | |
617 | if(m > 0) { | |
618 | --m; | |
619 | /* s[m] is our first non-ignorable code; look for a second in the same | |
620 | way **/ | |
c85b7022 RK |
621 | while(m > 0 |
622 | && (utf32__boundary_ignorable(utf32__iterator_word_break(it, | |
623 | it->s[m-1])))) | |
5617aaff | 624 | --m; |
b21a155c RK |
625 | /* Either m=0 or s[m-1] is not ignorable */ |
626 | if(m > 0) | |
627 | --m; | |
628 | } | |
629 | it->last[0] = it->last[1] = -1; | |
5617aaff RK |
630 | it->n = m; |
631 | return utf32_iterator_advance(it, n - m); | |
092f426f RK |
632 | } |
633 | ||
634 | /** @brief Advance an iterator | |
635 | * @param it Iterator | |
636 | * @param count Number of code points to advance by | |
637 | * @return 0 on success, non-0 on error | |
638 | * | |
639 | * It is an error to advance an iterator beyond the hypothetical post-final | |
640 | * character of the string. If an invalid value of @p n is specified then the | |
641 | * iterator is not changed. | |
642 | * | |
643 | * This function has O(n) time complexity: it works by advancing naively | |
644 | * forwards through the string. | |
645 | */ | |
646 | int utf32_iterator_advance(utf32_iterator it, size_t count) { | |
647 | if(count <= it->ns - it->n) { | |
648 | while(count > 0) { | |
649 | const uint32_t c = it->s[it->n]; | |
c85b7022 | 650 | const enum unicode_Word_Break wb = utf32__iterator_word_break(it, c); |
092f426f RK |
651 | if(it->last[1] == (uint32_t)-1 |
652 | || !utf32__boundary_ignorable(wb)) { | |
653 | it->last[0] = it->last[1]; | |
654 | it->last[1] = c; | |
655 | } | |
656 | ++it->n; | |
657 | --count; | |
658 | } | |
659 | return 0; | |
660 | } else | |
661 | return -1; | |
662 | } | |
663 | ||
664 | /** @brief Find the current code point | |
665 | * @param it Iterator | |
666 | * @return Current code point or 0 | |
667 | * | |
668 | * If the iterator points at the hypothetical post-final character of the | |
669 | * string then 0 is returned. NB that this doesn't mean that there aren't any | |
670 | * 0 code points inside the string! | |
671 | */ | |
672 | uint32_t utf32_iterator_code(utf32_iterator it) { | |
673 | if(it->n < it->ns) | |
674 | return it->s[it->n]; | |
675 | else | |
676 | return 0; | |
677 | } | |
678 | ||
679 | /** @brief Test for a grapheme boundary | |
680 | * @param it Iterator | |
681 | * @return Non-0 if pointing just after a grapheme boundary, otherwise 0 | |
f98fcddb RK |
682 | * |
683 | * This function identifies default grapheme cluster boundaries as described in | |
684 | * UAX #29 s3. It returns non-0 if @p it points at the code point just after a | |
685 | * grapheme cluster boundary (including the hypothetical code point just after | |
686 | * the end of the string). | |
092f426f RK |
687 | */ |
688 | int utf32_iterator_grapheme_boundary(utf32_iterator it) { | |
689 | uint32_t before, after; | |
690 | enum unicode_Grapheme_Break gbbefore, gbafter; | |
691 | /* GB1 and GB2 */ | |
692 | if(it->n == 0 || it->n == it->ns) | |
693 | return 1; | |
694 | /* Now we know that s[n-1] and s[n] are safe to inspect */ | |
695 | /* GB3 */ | |
696 | before = it->s[it->n-1]; | |
697 | after = it->s[it->n]; | |
698 | if(before == 0x000D && after == 0x000A) | |
699 | return 0; | |
700 | gbbefore = utf32__grapheme_break(before); | |
701 | gbafter = utf32__grapheme_break(after); | |
702 | /* GB4 */ | |
703 | if(gbbefore == unicode_Grapheme_Break_Control | |
704 | || before == 0x000D | |
705 | || before == 0x000A) | |
706 | return 1; | |
707 | /* GB5 */ | |
708 | if(gbafter == unicode_Grapheme_Break_Control | |
709 | || after == 0x000D | |
710 | || after == 0x000A) | |
711 | return 1; | |
712 | /* GB6 */ | |
713 | if(gbbefore == unicode_Grapheme_Break_L | |
714 | && (gbafter == unicode_Grapheme_Break_L | |
715 | || gbafter == unicode_Grapheme_Break_V | |
716 | || gbafter == unicode_Grapheme_Break_LV | |
717 | || gbafter == unicode_Grapheme_Break_LVT)) | |
718 | return 0; | |
719 | /* GB7 */ | |
720 | if((gbbefore == unicode_Grapheme_Break_LV | |
721 | || gbbefore == unicode_Grapheme_Break_V) | |
722 | && (gbafter == unicode_Grapheme_Break_V | |
723 | || gbafter == unicode_Grapheme_Break_T)) | |
724 | return 0; | |
725 | /* GB8 */ | |
726 | if((gbbefore == unicode_Grapheme_Break_LVT | |
727 | || gbbefore == unicode_Grapheme_Break_T) | |
728 | && gbafter == unicode_Grapheme_Break_T) | |
729 | return 0; | |
730 | /* GB9 */ | |
731 | if(gbafter == unicode_Grapheme_Break_Extend) | |
732 | return 0; | |
e2e88ad8 RK |
733 | /* GB9a */ |
734 | if(gbafter == unicode_Grapheme_Break_SpacingMark) | |
735 | return 0; | |
736 | /* GB9b */ | |
737 | if(gbbefore == unicode_Grapheme_Break_Prepend) | |
738 | return 0; | |
092f426f RK |
739 | /* GB10 */ |
740 | return 1; | |
741 | ||
742 | } | |
743 | ||
744 | /** @brief Test for a word boundary | |
745 | * @param it Iterator | |
746 | * @return Non-0 if pointing just after a word boundary, otherwise 0 | |
f98fcddb RK |
747 | * |
748 | * This function identifies default word boundaries as described in UAX #29 s4. | |
749 | * It returns non-0 if @p it points at the code point just after a word | |
750 | * boundary (including the hypothetical code point just after the end of the | |
751 | * string) and 0 otherwise. | |
092f426f RK |
752 | */ |
753 | int utf32_iterator_word_boundary(utf32_iterator it) { | |
36f522a4 | 754 | uint32_t before, after; |
2dc0bc24 | 755 | enum unicode_Word_Break wbtwobefore, wbbefore, wbafter, wbtwoafter; |
092f426f RK |
756 | size_t nn; |
757 | ||
758 | /* WB1 and WB2 */ | |
759 | if(it->n == 0 || it->n == it->ns) | |
760 | return 1; | |
36f522a4 RK |
761 | before = it->s[it->n-1]; |
762 | after = it->s[it->n]; | |
092f426f | 763 | /* WB3 */ |
36f522a4 | 764 | if(before == 0x000D && after == 0x000A) |
092f426f | 765 | return 0; |
fb4c61da | 766 | /* WB3a */ |
36f522a4 RK |
767 | if(utf32__iterator_word_break(it, before) == unicode_Word_Break_Newline |
768 | || before == 0x000D | |
769 | || before == 0x000A) | |
fb4c61da RK |
770 | return 1; |
771 | /* WB3b */ | |
36f522a4 RK |
772 | if(utf32__iterator_word_break(it, after) == unicode_Word_Break_Newline |
773 | || after == 0x000D | |
774 | || after == 0x000A) | |
fb4c61da | 775 | return 1; |
092f426f RK |
776 | /* WB4 */ |
777 | /* (!Sep) x (Extend|Format) as in UAX #29 s6.2 */ | |
36f522a4 RK |
778 | if(utf32__sentence_break(before) != unicode_Sentence_Break_Sep |
779 | && utf32__boundary_ignorable(utf32__iterator_word_break(it, after))) | |
092f426f RK |
780 | return 0; |
781 | /* Gather the property values we'll need for the rest of the test taking the | |
782 | * s6.2 changes into account */ | |
783 | /* First we look at the code points after the proposed boundary */ | |
784 | nn = it->n; /* <it->ns */ | |
2dc0bc24 RK |
785 | wbafter = utf32__iterator_word_break(it, it->s[nn++]); |
786 | if(!utf32__boundary_ignorable(wbafter)) { | |
092f426f RK |
787 | /* X (Extend|Format)* -> X */ |
788 | while(nn < it->ns | |
c85b7022 RK |
789 | && utf32__boundary_ignorable(utf32__iterator_word_break(it, |
790 | it->s[nn]))) | |
092f426f RK |
791 | ++nn; |
792 | } | |
793 | /* It's possible now that nn=ns */ | |
794 | if(nn < it->ns) | |
2dc0bc24 | 795 | wbtwoafter = utf32__iterator_word_break(it, it->s[nn]); |
092f426f | 796 | else |
2dc0bc24 | 797 | wbtwoafter = unicode_Word_Break_Other; |
092f426f RK |
798 | |
799 | /* We've already recorded the non-ignorable code points before the proposed | |
800 | * boundary */ | |
2dc0bc24 RK |
801 | wbbefore = utf32__iterator_word_break(it, it->last[1]); |
802 | wbtwobefore = utf32__iterator_word_break(it, it->last[0]); | |
092f426f RK |
803 | |
804 | /* WB5 */ | |
2dc0bc24 RK |
805 | if(wbbefore == unicode_Word_Break_ALetter |
806 | && wbafter == unicode_Word_Break_ALetter) | |
092f426f RK |
807 | return 0; |
808 | /* WB6 */ | |
2dc0bc24 RK |
809 | if(wbbefore == unicode_Word_Break_ALetter |
810 | && (wbafter == unicode_Word_Break_MidLetter | |
811 | || wbafter == unicode_Word_Break_MidNumLet) | |
812 | && wbtwoafter == unicode_Word_Break_ALetter) | |
092f426f RK |
813 | return 0; |
814 | /* WB7 */ | |
2dc0bc24 RK |
815 | if(wbtwobefore == unicode_Word_Break_ALetter |
816 | && (wbbefore == unicode_Word_Break_MidLetter | |
817 | || wbbefore == unicode_Word_Break_MidNumLet) | |
818 | && wbafter == unicode_Word_Break_ALetter) | |
092f426f | 819 | return 0; |
c85b7022 | 820 | /* WB8 */ |
2dc0bc24 RK |
821 | if(wbbefore == unicode_Word_Break_Numeric |
822 | && wbafter == unicode_Word_Break_Numeric) | |
092f426f RK |
823 | return 0; |
824 | /* WB9 */ | |
2dc0bc24 RK |
825 | if(wbbefore == unicode_Word_Break_ALetter |
826 | && wbafter == unicode_Word_Break_Numeric) | |
092f426f RK |
827 | return 0; |
828 | /* WB10 */ | |
2dc0bc24 RK |
829 | if(wbbefore == unicode_Word_Break_Numeric |
830 | && wbafter == unicode_Word_Break_ALetter) | |
092f426f RK |
831 | return 0; |
832 | /* WB11 */ | |
2dc0bc24 RK |
833 | if(wbtwobefore == unicode_Word_Break_Numeric |
834 | && (wbbefore == unicode_Word_Break_MidNum | |
835 | || wbbefore == unicode_Word_Break_MidNumLet) | |
836 | && wbafter == unicode_Word_Break_Numeric) | |
092f426f RK |
837 | return 0; |
838 | /* WB12 */ | |
2dc0bc24 RK |
839 | if(wbbefore == unicode_Word_Break_Numeric |
840 | && (wbafter == unicode_Word_Break_MidNum | |
841 | || wbafter == unicode_Word_Break_MidNumLet) | |
842 | && wbtwoafter == unicode_Word_Break_Numeric) | |
092f426f RK |
843 | return 0; |
844 | /* WB13 */ | |
2dc0bc24 RK |
845 | if(wbbefore == unicode_Word_Break_Katakana |
846 | && wbafter == unicode_Word_Break_Katakana) | |
092f426f RK |
847 | return 0; |
848 | /* WB13a */ | |
2dc0bc24 RK |
849 | if((wbbefore == unicode_Word_Break_ALetter |
850 | || wbbefore == unicode_Word_Break_Numeric | |
851 | || wbbefore == unicode_Word_Break_Katakana | |
852 | || wbbefore == unicode_Word_Break_ExtendNumLet) | |
853 | && wbafter == unicode_Word_Break_ExtendNumLet) | |
092f426f RK |
854 | return 0; |
855 | /* WB13b */ | |
2dc0bc24 RK |
856 | if(wbbefore == unicode_Word_Break_ExtendNumLet |
857 | && (wbafter == unicode_Word_Break_ALetter | |
858 | || wbafter == unicode_Word_Break_Numeric | |
859 | || wbafter == unicode_Word_Break_Katakana)) | |
092f426f RK |
860 | return 0; |
861 | /* WB14 */ | |
862 | return 1; | |
863 | } | |
864 | ||
865 | /*@}*/ | |
e5a5a138 RK |
866 | /** @defgroup utf32 Functions that operate on UTF-32 strings */ |
867 | /*@{*/ | |
868 | ||
869 | /** @brief Return the length of a 0-terminated UTF-32 string | |
870 | * @param s Pointer to 0-terminated string | |
871 | * @return Length of string in code points (excluding terminator) | |
872 | * | |
56fd389c | 873 | * Unlike the conversion functions no validity checking is done on the string. |
e5a5a138 RK |
874 | */ |
875 | size_t utf32_len(const uint32_t *s) { | |
876 | const uint32_t *t = s; | |
877 | ||
878 | while(*t) | |
879 | ++t; | |
880 | return (size_t)(t - s); | |
881 | } | |
882 | ||
e5a5a138 RK |
883 | /** @brief Stably sort [s,s+ns) into descending order of combining class |
884 | * @param s Start of array | |
885 | * @param ns Number of elements, must be at least 1 | |
886 | * @param buffer Buffer of at least @p ns elements | |
887 | */ | |
888 | static void utf32__sort_ccc(uint32_t *s, size_t ns, uint32_t *buffer) { | |
889 | uint32_t *a, *b, *bp; | |
890 | size_t na, nb; | |
891 | ||
892 | switch(ns) { | |
893 | case 1: /* 1-element array is always sorted */ | |
894 | return; | |
895 | case 2: /* 2-element arrays are trivial to sort */ | |
896 | if(utf32__combining_class(s[0]) > utf32__combining_class(s[1])) { | |
897 | uint32_t tmp = s[0]; | |
898 | s[0] = s[1]; | |
899 | s[1] = tmp; | |
900 | } | |
901 | return; | |
902 | default: | |
903 | /* Partition the array */ | |
904 | na = ns / 2; | |
905 | nb = ns - na; | |
906 | a = s; | |
907 | b = s + na; | |
908 | /* Sort the two halves of the array */ | |
909 | utf32__sort_ccc(a, na, buffer); | |
910 | utf32__sort_ccc(b, nb, buffer); | |
911 | /* Merge them back into one, via the buffer */ | |
912 | bp = buffer; | |
913 | while(na > 0 && nb > 0) { | |
16506c9d | 914 | /* We want ascending order of combining class (hence <) |
e5a5a138 RK |
915 | * and we want stability within combining classes (hence <=) |
916 | */ | |
917 | if(utf32__combining_class(*a) <= utf32__combining_class(*b)) { | |
918 | *bp++ = *a++; | |
919 | --na; | |
920 | } else { | |
921 | *bp++ = *b++; | |
922 | --nb; | |
923 | } | |
924 | } | |
925 | while(na > 0) { | |
926 | *bp++ = *a++; | |
927 | --na; | |
928 | } | |
929 | while(nb > 0) { | |
930 | *bp++ = *b++; | |
931 | --nb; | |
932 | } | |
933 | memcpy(s, buffer, ns * sizeof(uint32_t)); | |
934 | return; | |
935 | } | |
936 | } | |
937 | ||
938 | /** @brief Put combining characters into canonical order | |
939 | * @param s Pointer to UTF-32 string | |
940 | * @param ns Length of @p s | |
f98fcddb | 941 | * @return 0 on success, non-0 on error |
e5a5a138 | 942 | * |
56fd389c RK |
943 | * @p s is modified in-place. See Unicode 5.0 s3.11 for details of the |
944 | * ordering. | |
e5a5a138 | 945 | * |
56fd389c | 946 | * Currently we only support a maximum of 1024 combining characters after each |
f98fcddb | 947 | * base character. If this limit is exceeded then a non-0 value is returned. |
e5a5a138 RK |
948 | */ |
949 | static int utf32__canonical_ordering(uint32_t *s, size_t ns) { | |
950 | size_t nc; | |
951 | uint32_t buffer[1024]; | |
952 | ||
953 | /* The ordering amounts to a stable sort of each contiguous group of | |
954 | * characters with non-0 combining class. */ | |
955 | while(ns > 0) { | |
956 | /* Skip non-combining characters */ | |
957 | if(utf32__combining_class(*s) == 0) { | |
958 | ++s; | |
959 | --ns; | |
960 | continue; | |
961 | } | |
962 | /* We must now have at least one combining character; see how many | |
963 | * there are */ | |
964 | for(nc = 1; nc < ns && utf32__combining_class(s[nc]) != 0; ++nc) | |
965 | ; | |
966 | if(nc > 1024) | |
967 | return -1; | |
968 | /* Sort the array */ | |
969 | utf32__sort_ccc(s, nc, buffer); | |
970 | s += nc; | |
971 | ns -= nc; | |
972 | } | |
973 | return 0; | |
974 | } | |
975 | ||
976 | /* Magic numbers from UAX #15 s16 */ | |
977 | #define SBase 0xAC00 | |
978 | #define LBase 0x1100 | |
979 | #define VBase 0x1161 | |
980 | #define TBase 0x11A7 | |
981 | #define LCount 19 | |
982 | #define VCount 21 | |
983 | #define TCount 28 | |
984 | #define NCount (VCount * TCount) | |
985 | #define SCount (LCount * NCount) | |
986 | ||
987 | /** @brief Guts of the decomposition lookup functions */ | |
988 | #define utf32__decompose_one_generic(WHICH) do { \ | |
f98fcddb | 989 | const uint32_t *dc = utf32__decomposition_##WHICH(c); \ |
e5a5a138 RK |
990 | if(dc) { \ |
991 | /* Found a canonical decomposition in the table */ \ | |
992 | while(*dc) \ | |
993 | utf32__decompose_one_##WHICH(d, *dc++); \ | |
994 | } else if(c >= SBase && c < SBase + SCount) { \ | |
995 | /* Mechanically decomposable Hangul syllable (UAX #15 s16) */ \ | |
996 | const uint32_t SIndex = c - SBase; \ | |
997 | const uint32_t L = LBase + SIndex / NCount; \ | |
998 | const uint32_t V = VBase + (SIndex % NCount) / TCount; \ | |
999 | const uint32_t T = TBase + SIndex % TCount; \ | |
1000 | dynstr_ucs4_append(d, L); \ | |
1001 | dynstr_ucs4_append(d, V); \ | |
1002 | if(T != TBase) \ | |
1003 | dynstr_ucs4_append(d, T); \ | |
1004 | } else \ | |
1005 | /* Equal to own canonical decomposition */ \ | |
1006 | dynstr_ucs4_append(d, c); \ | |
1007 | } while(0) | |
1008 | ||
1009 | /** @brief Recursively compute the canonical decomposition of @p c | |
1010 | * @param d Dynamic string to store decomposition in | |
1011 | * @param c Code point to decompose (must be a valid!) | |
f98fcddb | 1012 | * @return 0 on success, non-0 on error |
e5a5a138 RK |
1013 | */ |
1014 | static void utf32__decompose_one_canon(struct dynstr_ucs4 *d, uint32_t c) { | |
1015 | utf32__decompose_one_generic(canon); | |
1016 | } | |
1017 | ||
1018 | /** @brief Recursively compute the compatibility decomposition of @p c | |
1019 | * @param d Dynamic string to store decomposition in | |
1020 | * @param c Code point to decompose (must be a valid!) | |
f98fcddb | 1021 | * @return 0 on success, non-0 on error |
e5a5a138 RK |
1022 | */ |
1023 | static void utf32__decompose_one_compat(struct dynstr_ucs4 *d, uint32_t c) { | |
1024 | utf32__decompose_one_generic(compat); | |
1025 | } | |
1026 | ||
16506c9d RK |
1027 | /** @brief Magic utf32__compositions() return value for Hangul Choseong */ |
1028 | static const uint32_t utf32__hangul_L[1]; | |
1029 | ||
1030 | /** @brief Return the list of compositions that @p c starts | |
1031 | * @param c Starter code point | |
1032 | * @return Composition list or NULL | |
1033 | * | |
1034 | * For Hangul leading (Choseong) jamo we return the special value | |
1035 | * utf32__hangul_L. These code points are not listed as the targets of | |
1036 | * canonical decompositions (make-unidata checks) so there is no confusion with | |
1037 | * real decompositions here. | |
1038 | */ | |
1039 | static const uint32_t *utf32__compositions(uint32_t c) { | |
1040 | const uint32_t *compositions = utf32__unidata(c)->composed; | |
1041 | ||
1042 | if(compositions) | |
1043 | return compositions; | |
1044 | /* Special-casing for Hangul */ | |
1045 | switch(utf32__grapheme_break(c)) { | |
1046 | default: | |
1047 | return 0; | |
1048 | case unicode_Grapheme_Break_L: | |
1049 | return utf32__hangul_L; | |
1050 | } | |
1051 | } | |
1052 | ||
1053 | /** @brief Composition step | |
1054 | * @param s Start of string | |
1055 | * @param ns Length of string | |
1056 | * @return New length of string | |
1057 | * | |
1058 | * This is called from utf32__decompose_generic() to compose the result string | |
1059 | * in place. | |
1060 | */ | |
1061 | static size_t utf32__compose(uint32_t *s, size_t ns) { | |
1062 | const uint32_t *compositions; | |
1063 | uint32_t *start = s, *t = s, *tt, cc; | |
1064 | ||
1065 | while(ns > 0) { | |
1066 | uint32_t starter = *s++; | |
1067 | int block_starters = 0; | |
1068 | --ns; | |
1069 | /* We don't attempt to compose the following things: | |
1070 | * - final characters whatever kind they are | |
1071 | * - non-starter characters | |
1072 | * - starters that don't take part in a canonical decomposition mapping | |
1073 | */ | |
1074 | if(ns == 0 | |
1075 | || utf32__combining_class(starter) | |
1076 | || !(compositions = utf32__compositions(starter))) { | |
1077 | *t++ = starter; | |
1078 | continue; | |
1079 | } | |
1080 | if(compositions != utf32__hangul_L) { | |
1081 | /* Where we'll put the eventual starter */ | |
1082 | tt = t++; | |
1083 | do { | |
1084 | /* See if we can find composition of starter+*s */ | |
1085 | const uint32_t cchar = *s, *cp = compositions; | |
1086 | while((cc = *cp++)) { | |
1087 | const uint32_t *decomp = utf32__decomposition_canon(cc); | |
1088 | /* We know decomp[0] == starter */ | |
1089 | if(decomp[1] == cchar) | |
1090 | break; | |
1091 | } | |
1092 | if(cc) { | |
1093 | /* Found a composition: cc decomposes to starter,*s */ | |
1094 | starter = cc; | |
1095 | compositions = utf32__compositions(starter); | |
1096 | ++s; | |
1097 | --ns; | |
1098 | } else { | |
1099 | /* No composition found. */ | |
1100 | const int class = utf32__combining_class(*s); | |
1101 | if(class) { | |
1102 | /* Transfer the uncomposable combining character to the output */ | |
1103 | *t++ = *s++; | |
1104 | --ns; | |
1105 | /* All the combining characters of the same class of the | |
1106 | * uncomposable character are blocked by it, but there may be | |
1107 | * others of higher class later. We eat the uncomposable and | |
1108 | * blocked characters and go back round the loop for that higher | |
1109 | * class. */ | |
1110 | while(ns > 0 && utf32__combining_class(*s) == class) { | |
1111 | *t++ = *s++; | |
1112 | --ns; | |
1113 | } | |
1114 | /* Block any subsequent starters */ | |
1115 | block_starters = 1; | |
1116 | } else { | |
1117 | /* The uncombinable character is itself a starter, so we don't | |
1118 | * transfer it to the output but instead go back round the main | |
1119 | * loop. */ | |
1120 | break; | |
1121 | } | |
1122 | } | |
1123 | /* Keep going while there are still characters and the starter takes | |
1124 | * part in some composition */ | |
1125 | } while(ns > 0 && compositions | |
1126 | && (!block_starters || utf32__combining_class(*s))); | |
1127 | /* Store any remaining combining characters */ | |
1128 | while(ns > 0 && utf32__combining_class(*s)) { | |
1129 | *t++ = *s++; | |
1130 | --ns; | |
1131 | } | |
1132 | /* Store the resulting starter */ | |
1133 | *tt = starter; | |
1134 | } else { | |
1135 | /* Special-casing for Hangul | |
1136 | * | |
1137 | * If there are combining characters between the L and the V then they | |
1138 | * will block the V and so no composition happens. Similarly combining | |
1139 | * characters between V and T will block the T and so we only get as far | |
1140 | * as LV. | |
1141 | */ | |
1142 | if(utf32__grapheme_break(*s) == unicode_Grapheme_Break_V) { | |
1143 | const uint32_t V = *s++; | |
1144 | const uint32_t LIndex = starter - LBase; | |
1145 | const uint32_t VIndex = V - VBase; | |
1146 | uint32_t TIndex; | |
1147 | --ns; | |
1148 | if(ns > 0 | |
1149 | && utf32__grapheme_break(*s) == unicode_Grapheme_Break_T) { | |
1150 | /* We have an L V T sequence */ | |
1151 | const uint32_t T = *s++; | |
1152 | TIndex = T - TBase; | |
1153 | --ns; | |
1154 | } else | |
1155 | /* It's just L V */ | |
1156 | TIndex = 0; | |
1157 | /* Compose to LVT or LV as appropriate */ | |
1158 | starter = (LIndex * VCount + VIndex) * TCount + TIndex + SBase; | |
1159 | } /* else we only have L or LV and no V or T */ | |
1160 | *t++ = starter; | |
1161 | /* There could be some combining characters that belong to the V or T. | |
1162 | * These will be treated as non-starter characters at the top of the loop | |
1163 | * and thuss transferred to the output. */ | |
1164 | } | |
1165 | } | |
1166 | return t - start; | |
1167 | } | |
1168 | ||
1169 | /** @brief Guts of the composition and decomposition functions | |
1170 | * @param WHICH @c canon or @c compat to choose decomposition | |
1171 | * @param COMPOSE @c 0 or @c 1 to compose | |
1172 | */ | |
1173 | #define utf32__decompose_generic(WHICH, COMPOSE) do { \ | |
e5a5a138 RK |
1174 | struct dynstr_ucs4 d; \ |
1175 | uint32_t c; \ | |
1176 | \ | |
1177 | dynstr_ucs4_init(&d); \ | |
1178 | while(ns) { \ | |
1179 | c = *s++; \ | |
56fd389c | 1180 | if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF) \ |
e5a5a138 RK |
1181 | goto error; \ |
1182 | utf32__decompose_one_##WHICH(&d, c); \ | |
1183 | --ns; \ | |
1184 | } \ | |
1185 | if(utf32__canonical_ordering(d.vec, d.nvec)) \ | |
1186 | goto error; \ | |
16506c9d RK |
1187 | if(COMPOSE) \ |
1188 | d.nvec = utf32__compose(d.vec, d.nvec); \ | |
e5a5a138 RK |
1189 | dynstr_ucs4_terminate(&d); \ |
1190 | if(ndp) \ | |
1191 | *ndp = d.nvec; \ | |
1192 | return d.vec; \ | |
1193 | error: \ | |
1194 | xfree(d.vec); \ | |
1195 | return 0; \ | |
1196 | } while(0) | |
1197 | ||
1198 | /** @brief Canonically decompose @p [s,s+ns) | |
1199 | * @param s Pointer to string | |
1200 | * @param ns Length of string | |
1201 | * @param ndp Where to store length of result | |
f98fcddb | 1202 | * @return Pointer to result string, or NULL on error |
e5a5a138 | 1203 | * |
16506c9d RK |
1204 | * Computes NFD (Normalization Form D) of the string at @p s. This implies |
1205 | * performing all canonical decompositions and then normalizing the order of | |
1206 | * combining characters. | |
e5a5a138 | 1207 | * |
56fd389c | 1208 | * Returns NULL if the string is not valid for either of the following reasons: |
e5a5a138 RK |
1209 | * - it codes for a UTF-16 surrogate |
1210 | * - it codes for a value outside the unicode code space | |
16506c9d RK |
1211 | * |
1212 | * See also: | |
1213 | * - utf32_decompose_compat() | |
1214 | * - utf32_compose_canon() | |
e5a5a138 RK |
1215 | */ |
1216 | uint32_t *utf32_decompose_canon(const uint32_t *s, size_t ns, size_t *ndp) { | |
16506c9d | 1217 | utf32__decompose_generic(canon, 0); |
e5a5a138 RK |
1218 | } |
1219 | ||
1220 | /** @brief Compatibility decompose @p [s,s+ns) | |
1221 | * @param s Pointer to string | |
1222 | * @param ns Length of string | |
1223 | * @param ndp Where to store length of result | |
f98fcddb | 1224 | * @return Pointer to result string, or NULL on error |
e5a5a138 | 1225 | * |
16506c9d RK |
1226 | * Computes NFKD (Normalization Form KD) of the string at @p s. This implies |
1227 | * performing all canonical and compatibility decompositions and then | |
1228 | * normalizing the order of combining characters. | |
e5a5a138 | 1229 | * |
56fd389c | 1230 | * Returns NULL if the string is not valid for either of the following reasons: |
e5a5a138 RK |
1231 | * - it codes for a UTF-16 surrogate |
1232 | * - it codes for a value outside the unicode code space | |
16506c9d RK |
1233 | * |
1234 | * See also: | |
1235 | * - utf32_decompose_canon() | |
1236 | * - utf32_compose_compat() | |
e5a5a138 RK |
1237 | */ |
1238 | uint32_t *utf32_decompose_compat(const uint32_t *s, size_t ns, size_t *ndp) { | |
16506c9d RK |
1239 | utf32__decompose_generic(compat, 0); |
1240 | } | |
1241 | ||
1242 | /** @brief Canonically compose @p [s,s+ns) | |
1243 | * @param s Pointer to string | |
1244 | * @param ns Length of string | |
1245 | * @param ndp Where to store length of result | |
1246 | * @return Pointer to result string, or NULL on error | |
1247 | * | |
1248 | * Computes NFC (Normalization Form C) of the string at @p s. This implies | |
1249 | * performing all canonical decompositions, normalizing the order of combining | |
1250 | * characters and then composing all unblocked primary compositables. | |
1251 | * | |
1252 | * Returns NULL if the string is not valid for either of the following reasons: | |
1253 | * - it codes for a UTF-16 surrogate | |
1254 | * - it codes for a value outside the unicode code space | |
1255 | * | |
1256 | * See also: | |
1257 | * - utf32_compose_compat() | |
1258 | * - utf32_decompose_canon() | |
1259 | */ | |
1260 | uint32_t *utf32_compose_canon(const uint32_t *s, size_t ns, size_t *ndp) { | |
1261 | utf32__decompose_generic(canon, 1); | |
1262 | } | |
1263 | ||
1264 | /** @brief Compatibility compose @p [s,s+ns) | |
1265 | * @param s Pointer to string | |
1266 | * @param ns Length of string | |
1267 | * @param ndp Where to store length of result | |
1268 | * @return Pointer to result string, or NULL on error | |
1269 | * | |
1270 | * Computes NFKC (Normalization Form KC) of the string at @p s. This implies | |
1271 | * performing all canonical and compatibility decompositions, normalizing the | |
1272 | * order of combining characters and then composing all unblocked primary | |
1273 | * compositables. | |
1274 | * | |
1275 | * Returns NULL if the string is not valid for either of the following reasons: | |
1276 | * - it codes for a UTF-16 surrogate | |
1277 | * - it codes for a value outside the unicode code space | |
1278 | * | |
1279 | * See also: | |
1280 | * - utf32_compose_canon() | |
1281 | * - utf32_decompose_compat() | |
1282 | */ | |
1283 | uint32_t *utf32_compose_compat(const uint32_t *s, size_t ns, size_t *ndp) { | |
1284 | utf32__decompose_generic(compat, 1); | |
e5a5a138 RK |
1285 | } |
1286 | ||
56fd389c RK |
1287 | /** @brief Single-character case-fold and decompose operation */ |
1288 | #define utf32__casefold_one(WHICH) do { \ | |
bcf9ed7f | 1289 | const uint32_t *cf = utf32__unidata(c)->casefold; \ |
56fd389c RK |
1290 | if(cf) { \ |
1291 | /* Found a case-fold mapping in the table */ \ | |
1292 | while(*cf) \ | |
1293 | utf32__decompose_one_##WHICH(&d, *cf++); \ | |
1294 | } else \ | |
1295 | utf32__decompose_one_##WHICH(&d, c); \ | |
1296 | } while(0) | |
e5a5a138 RK |
1297 | |
1298 | /** @brief Case-fold @p [s,s+ns) | |
1299 | * @param s Pointer to string | |
1300 | * @param ns Length of string | |
1301 | * @param ndp Where to store length of result | |
f98fcddb | 1302 | * @return Pointer to result string, or NULL on error |
e5a5a138 RK |
1303 | * |
1304 | * Case-fold the string at @p s according to full default case-folding rules | |
56fd389c | 1305 | * (s3.13) for caseless matching. The result will be in NFD. |
e5a5a138 | 1306 | * |
56fd389c | 1307 | * Returns NULL if the string is not valid for either of the following reasons: |
e5a5a138 RK |
1308 | * - it codes for a UTF-16 surrogate |
1309 | * - it codes for a value outside the unicode code space | |
1310 | */ | |
1311 | uint32_t *utf32_casefold_canon(const uint32_t *s, size_t ns, size_t *ndp) { | |
1312 | struct dynstr_ucs4 d; | |
1313 | uint32_t c; | |
1314 | size_t n; | |
1315 | uint32_t *ss = 0; | |
1316 | ||
1317 | /* If the canonical decomposition of the string includes any combining | |
1318 | * character that case-folds to a non-combining character then we must | |
1319 | * normalize before we fold. In Unicode 5.0.0 this means 0345 COMBINING | |
1320 | * GREEK YPOGEGRAMMENI in its decomposition and the various characters that | |
1321 | * canonically decompose to it. */ | |
bcf9ed7f RK |
1322 | for(n = 0; n < ns; ++n) |
1323 | if(utf32__unidata(s[n])->flags & unicode_normalize_before_casefold) | |
e5a5a138 | 1324 | break; |
e5a5a138 RK |
1325 | if(n < ns) { |
1326 | /* We need a preliminary decomposition */ | |
1327 | if(!(ss = utf32_decompose_canon(s, ns, &ns))) | |
1328 | return 0; | |
1329 | s = ss; | |
1330 | } | |
1331 | dynstr_ucs4_init(&d); | |
1332 | while(ns) { | |
1333 | c = *s++; | |
56fd389c | 1334 | if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF) |
e5a5a138 | 1335 | goto error; |
56fd389c | 1336 | utf32__casefold_one(canon); |
e5a5a138 RK |
1337 | --ns; |
1338 | } | |
1339 | if(utf32__canonical_ordering(d.vec, d.nvec)) | |
1340 | goto error; | |
1341 | dynstr_ucs4_terminate(&d); | |
1342 | if(ndp) | |
1343 | *ndp = d.nvec; | |
1344 | return d.vec; | |
1345 | error: | |
1346 | xfree(d.vec); | |
1347 | xfree(ss); | |
1348 | return 0; | |
1349 | } | |
1350 | ||
f98fcddb | 1351 | /** @brief Compatibility case-fold @p [s,s+ns) |
56fd389c RK |
1352 | * @param s Pointer to string |
1353 | * @param ns Length of string | |
1354 | * @param ndp Where to store length of result | |
f98fcddb | 1355 | * @return Pointer to result string, or NULL on error |
56fd389c RK |
1356 | * |
1357 | * Case-fold the string at @p s according to full default case-folding rules | |
1358 | * (s3.13) for compatibility caseless matching. The result will be in NFKD. | |
1359 | * | |
1360 | * Returns NULL if the string is not valid for either of the following reasons: | |
1361 | * - it codes for a UTF-16 surrogate | |
1362 | * - it codes for a value outside the unicode code space | |
1363 | */ | |
1364 | uint32_t *utf32_casefold_compat(const uint32_t *s, size_t ns, size_t *ndp) { | |
1365 | struct dynstr_ucs4 d; | |
1366 | uint32_t c; | |
1367 | size_t n; | |
1368 | uint32_t *ss = 0; | |
1369 | ||
bcf9ed7f RK |
1370 | for(n = 0; n < ns; ++n) |
1371 | if(utf32__unidata(s[n])->flags & unicode_normalize_before_casefold) | |
56fd389c | 1372 | break; |
56fd389c RK |
1373 | if(n < ns) { |
1374 | /* We need a preliminary _canonical_ decomposition */ | |
1375 | if(!(ss = utf32_decompose_canon(s, ns, &ns))) | |
1376 | return 0; | |
1377 | s = ss; | |
1378 | } | |
1379 | /* This computes NFKD(toCaseFold(s)) */ | |
1380 | #define compat_casefold_middle() do { \ | |
1381 | dynstr_ucs4_init(&d); \ | |
1382 | while(ns) { \ | |
1383 | c = *s++; \ | |
1384 | if((c >= 0xD800 && c <= 0xDFFF) || c > 0x10FFFF) \ | |
1385 | goto error; \ | |
1386 | utf32__casefold_one(compat); \ | |
1387 | --ns; \ | |
1388 | } \ | |
1389 | if(utf32__canonical_ordering(d.vec, d.nvec)) \ | |
1390 | goto error; \ | |
1391 | } while(0) | |
1392 | /* Do the inner (NFKD o toCaseFold) */ | |
1393 | compat_casefold_middle(); | |
1394 | /* We can do away with the NFD'd copy of the input now */ | |
1395 | xfree(ss); | |
1396 | s = ss = d.vec; | |
1397 | ns = d.nvec; | |
1398 | /* Do the outer (NFKD o toCaseFold) */ | |
1399 | compat_casefold_middle(); | |
1400 | /* That's all */ | |
1401 | dynstr_ucs4_terminate(&d); | |
1402 | if(ndp) | |
1403 | *ndp = d.nvec; | |
1404 | return d.vec; | |
1405 | error: | |
1406 | xfree(d.vec); | |
1407 | xfree(ss); | |
1408 | return 0; | |
1409 | } | |
1410 | ||
e5a5a138 RK |
1411 | /** @brief Order a pair of UTF-32 strings |
1412 | * @param a First 0-terminated string | |
1413 | * @param b Second 0-terminated string | |
1414 | * @return -1, 0 or 1 for a less than, equal to or greater than b | |
1415 | * | |
1416 | * "Comparable to strcmp() at its best." | |
1417 | */ | |
1418 | int utf32_cmp(const uint32_t *a, const uint32_t *b) { | |
1419 | while(*a && *b && *a == *b) { | |
1420 | ++a; | |
1421 | ++b; | |
1422 | } | |
1423 | return *a < *b ? -1 : (*a > *b ? 1 : 0); | |
1424 | } | |
1425 | ||
35b651f0 RK |
1426 | /** @brief Identify a grapheme cluster boundary |
1427 | * @param s Start of string (must be NFD) | |
1428 | * @param ns Length of string | |
1429 | * @param n Index within string (in [0,ns].) | |
1430 | * @return 1 at a grapheme cluster boundary, 0 otherwise | |
1431 | * | |
1432 | * This function identifies default grapheme cluster boundaries as described in | |
f98fcddb | 1433 | * UAX #29 s3. It returns non-0 if @p n points at the code point just after a |
35b651f0 RK |
1434 | * grapheme cluster boundary (including the hypothetical code point just after |
1435 | * the end of the string). | |
f98fcddb RK |
1436 | * |
1437 | * This function uses utf32_iterator_set() internally; see that function for | |
1438 | * remarks on performance. | |
35b651f0 | 1439 | */ |
1625e11a | 1440 | int utf32_is_grapheme_boundary(const uint32_t *s, size_t ns, size_t n) { |
092f426f | 1441 | struct utf32_iterator_data it[1]; |
35b651f0 | 1442 | |
092f426f RK |
1443 | utf32__iterator_init(it, s, ns, n); |
1444 | return utf32_iterator_grapheme_boundary(it); | |
0b7052da RK |
1445 | } |
1446 | ||
1447 | /** @brief Identify a word boundary | |
1448 | * @param s Start of string (must be NFD) | |
1449 | * @param ns Length of string | |
1450 | * @param n Index within string (in [0,ns].) | |
1451 | * @return 1 at a word boundary, 0 otherwise | |
1452 | * | |
1453 | * This function identifies default word boundaries as described in UAX #29 s4. | |
f98fcddb | 1454 | * It returns non-0 if @p n points at the code point just after a word boundary |
0b7052da | 1455 | * (including the hypothetical code point just after the end of the string). |
f98fcddb RK |
1456 | * |
1457 | * This function uses utf32_iterator_set() internally; see that function for | |
1458 | * remarks on performance. | |
0b7052da RK |
1459 | */ |
1460 | int utf32_is_word_boundary(const uint32_t *s, size_t ns, size_t n) { | |
092f426f | 1461 | struct utf32_iterator_data it[1]; |
0b7052da | 1462 | |
092f426f RK |
1463 | utf32__iterator_init(it, s, ns, n); |
1464 | return utf32_iterator_word_boundary(it); | |
0b7052da RK |
1465 | } |
1466 | ||
8818b7fc RK |
1467 | /** @brief Split [s,ns) into multiple words |
1468 | * @param s Pointer to start of string | |
1469 | * @param ns Length of string | |
1470 | * @param nwp Where to store word count, or NULL | |
c85b7022 | 1471 | * @param wbreak Word_Break property tailor, or NULL |
8818b7fc RK |
1472 | * @return Pointer to array of pointers to words |
1473 | * | |
1474 | * The returned array is terminated by a NULL pointer and individual | |
1475 | * strings are 0-terminated. | |
1476 | */ | |
c85b7022 RK |
1477 | uint32_t **utf32_word_split(const uint32_t *s, size_t ns, size_t *nwp, |
1478 | unicode_property_tailor *wbreak) { | |
8818b7fc RK |
1479 | struct utf32_iterator_data it[1]; |
1480 | size_t b1 = 0, b2 = 0 ,i; | |
1481 | int isword; | |
1482 | struct vector32 v32[1]; | |
1483 | uint32_t *w; | |
1484 | ||
1485 | vector32_init(v32); | |
1486 | utf32__iterator_init(it, s, ns, 0); | |
c85b7022 | 1487 | it->word_break = wbreak; |
8818b7fc RK |
1488 | /* Work our way through the string stopping at each word break. */ |
1489 | do { | |
1490 | if(utf32_iterator_word_boundary(it)) { | |
1491 | /* We've found a new boundary */ | |
1492 | b1 = b2; | |
1493 | b2 = it->n; | |
1494 | /*fprintf(stderr, "[%zu, %zu) is a candidate word\n", b1, b2);*/ | |
1495 | /* Inspect the characters between the boundary and form an opinion as to | |
1496 | * whether they are a word or not */ | |
1497 | isword = 0; | |
1498 | for(i = b1; i < b2; ++i) { | |
c85b7022 | 1499 | switch(utf32__iterator_word_break(it, it->s[i])) { |
8818b7fc RK |
1500 | case unicode_Word_Break_ALetter: |
1501 | case unicode_Word_Break_Numeric: | |
1502 | case unicode_Word_Break_Katakana: | |
1503 | isword = 1; | |
1504 | break; | |
1505 | default: | |
1506 | break; | |
1507 | } | |
1508 | } | |
1509 | /* If it's a word add it to the list of results */ | |
1510 | if(isword) { | |
8e93ddd1 RK |
1511 | const size_t len = b2 - b1; |
1512 | w = xcalloc_noptr(len + 1, sizeof(uint32_t)); | |
1513 | memcpy(w, it->s + b1, len * sizeof (uint32_t)); | |
1514 | w[len] = 0; | |
8818b7fc RK |
1515 | vector32_append(v32, w); |
1516 | } | |
1517 | } | |
1518 | } while(!utf32_iterator_advance(it, 1)); | |
1519 | vector32_terminate(v32); | |
1520 | if(nwp) | |
1521 | *nwp = v32->nvec; | |
1522 | return v32->vec; | |
1523 | } | |
1524 | ||
e5a5a138 | 1525 | /*@}*/ |
349b7b74 | 1526 | /** @defgroup utf8 Functions that operate on UTF-8 strings */ |
e5a5a138 RK |
1527 | /*@{*/ |
1528 | ||
1529 | /** @brief Wrapper to transform a UTF-8 string using the UTF-32 function */ | |
1530 | #define utf8__transform(FN) do { \ | |
1531 | uint32_t *to32 = 0, *decomp32 = 0; \ | |
1532 | size_t nto32, ndecomp32; \ | |
1533 | char *decomp8 = 0; \ | |
1534 | \ | |
1535 | if(!(to32 = utf8_to_utf32(s, ns, &nto32))) goto error; \ | |
1536 | if(!(decomp32 = FN(to32, nto32, &ndecomp32))) goto error; \ | |
1537 | decomp8 = utf32_to_utf8(decomp32, ndecomp32, ndp); \ | |
1538 | error: \ | |
1539 | xfree(to32); \ | |
1540 | xfree(decomp32); \ | |
1541 | return decomp8; \ | |
1542 | } while(0) | |
1543 | ||
1544 | /** @brief Canonically decompose @p [s,s+ns) | |
1545 | * @param s Pointer to string | |
1546 | * @param ns Length of string | |
1547 | * @param ndp Where to store length of result | |
f98fcddb | 1548 | * @return Pointer to result string, or NULL on error |
e5a5a138 | 1549 | * |
0ae60b83 RK |
1550 | * Computes NFD (Normalization Form D) of the string at @p s. This implies |
1551 | * performing all canonical decompositions and then normalizing the order of | |
1552 | * combining characters. | |
e5a5a138 RK |
1553 | * |
1554 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why | |
1555 | * this might be. | |
1556 | * | |
0ae60b83 RK |
1557 | * See also: |
1558 | * - utf32_decompose_canon(). | |
1559 | * - utf8_decompose_compat() | |
1560 | * - utf8_compose_canon() | |
e5a5a138 RK |
1561 | */ |
1562 | char *utf8_decompose_canon(const char *s, size_t ns, size_t *ndp) { | |
1563 | utf8__transform(utf32_decompose_canon); | |
1564 | } | |
1565 | ||
1566 | /** @brief Compatibility decompose @p [s,s+ns) | |
1567 | * @param s Pointer to string | |
1568 | * @param ns Length of string | |
1569 | * @param ndp Where to store length of result | |
f98fcddb | 1570 | * @return Pointer to result string, or NULL on error |
e5a5a138 | 1571 | * |
0ae60b83 RK |
1572 | * Computes NFKD (Normalization Form KD) of the string at @p s. This implies |
1573 | * performing all canonical and compatibility decompositions and then | |
1574 | * normalizing the order of combining characters. | |
e5a5a138 RK |
1575 | * |
1576 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why | |
1577 | * this might be. | |
1578 | * | |
0ae60b83 RK |
1579 | * See also: |
1580 | * - utf32_decompose_compat(). | |
1581 | * - utf8_decompose_canon() | |
1582 | * - utf8_compose_compat() | |
e5a5a138 RK |
1583 | */ |
1584 | char *utf8_decompose_compat(const char *s, size_t ns, size_t *ndp) { | |
1585 | utf8__transform(utf32_decompose_compat); | |
1586 | } | |
1587 | ||
0ae60b83 RK |
1588 | /** @brief Canonically compose @p [s,s+ns) |
1589 | * @param s Pointer to string | |
1590 | * @param ns Length of string | |
1591 | * @param ndp Where to store length of result | |
1592 | * @return Pointer to result string, or NULL on error | |
1593 | * | |
1594 | * Computes NFC (Normalization Form C) of the string at @p s. This implies | |
1595 | * performing all canonical decompositions, normalizing the order of combining | |
1596 | * characters and then composing all unblocked primary compositables. | |
1597 | * | |
1598 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why | |
1599 | * this might be. | |
1600 | * | |
1601 | * See also: | |
1602 | * - utf32_compose_canon() | |
1603 | * - utf8_compose_compat() | |
1604 | * - utf8_decompose_canon() | |
1605 | */ | |
1606 | char *utf8_compose_canon(const char *s, size_t ns, size_t *ndp) { | |
1607 | utf8__transform(utf32_compose_canon); | |
1608 | } | |
1609 | ||
1610 | /** @brief Compatibility compose @p [s,s+ns) | |
1611 | * @param s Pointer to string | |
1612 | * @param ns Length of string | |
1613 | * @param ndp Where to store length of result | |
1614 | * @return Pointer to result string, or NULL on error | |
1615 | * | |
1616 | * Computes NFKC (Normalization Form KC) of the string at @p s. This implies | |
1617 | * performing all canonical and compatibility decompositions, normalizing the | |
1618 | * order of combining characters and then composing all unblocked primary | |
1619 | * compositables. | |
1620 | * | |
1621 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why | |
1622 | * this might be. | |
1623 | * | |
1624 | * See also: | |
1625 | * - utf32_compose_compat() | |
1626 | * - utf8_compose_canon() | |
1627 | * - utf8_decompose_compat() | |
1628 | */ | |
1629 | char *utf8_compose_compat(const char *s, size_t ns, size_t *ndp) { | |
1630 | utf8__transform(utf32_compose_compat); | |
1631 | } | |
1632 | ||
e5a5a138 RK |
1633 | /** @brief Case-fold @p [s,s+ns) |
1634 | * @param s Pointer to string | |
1635 | * @param ns Length of string | |
1636 | * @param ndp Where to store length of result | |
f98fcddb | 1637 | * @return Pointer to result string, or NULL on error |
e5a5a138 RK |
1638 | * |
1639 | * Case-fold the string at @p s according to full default case-folding rules | |
1640 | * (s3.13). The result will be in NFD. | |
1641 | * | |
1642 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why | |
1643 | * this might be. | |
1644 | */ | |
1645 | char *utf8_casefold_canon(const char *s, size_t ns, size_t *ndp) { | |
1646 | utf8__transform(utf32_casefold_canon); | |
1647 | } | |
1648 | ||
1649 | /** @brief Compatibility case-fold @p [s,s+ns) | |
1650 | * @param s Pointer to string | |
1651 | * @param ns Length of string | |
1652 | * @param ndp Where to store length of result | |
f98fcddb | 1653 | * @return Pointer to result string, or NULL on error |
e5a5a138 RK |
1654 | * |
1655 | * Case-fold the string at @p s according to full default case-folding rules | |
1656 | * (s3.13). The result will be in NFKD. | |
1657 | * | |
1658 | * Returns NULL if the string is not valid; see utf8_to_utf32() for reasons why | |
1659 | * this might be. | |
1660 | */ | |
e5a5a138 RK |
1661 | char *utf8_casefold_compat(const char *s, size_t ns, size_t *ndp) { |
1662 | utf8__transform(utf32_casefold_compat); | |
1663 | } | |
e5a5a138 | 1664 | |
8818b7fc RK |
1665 | /** @brief Split [s,ns) into multiple words |
1666 | * @param s Pointer to start of string | |
1667 | * @param ns Length of string | |
1668 | * @param nwp Where to store word count, or NULL | |
c85b7022 | 1669 | * @param wbreak Word_Break property tailor, or NULL |
8818b7fc RK |
1670 | * @return Pointer to array of pointers to words |
1671 | * | |
1672 | * The returned array is terminated by a NULL pointer and individual | |
1673 | * strings are 0-terminated. | |
1674 | */ | |
c85b7022 RK |
1675 | char **utf8_word_split(const char *s, size_t ns, size_t *nwp, |
1676 | unicode_property_tailor *wbreak) { | |
8818b7fc RK |
1677 | uint32_t *to32 = 0, **v32 = 0; |
1678 | size_t nto32, nv, n; | |
1679 | char **v8 = 0, **ret = 0; | |
c85b7022 | 1680 | |
8818b7fc | 1681 | if(!(to32 = utf8_to_utf32(s, ns, &nto32))) goto error; |
c85b7022 | 1682 | if(!(v32 = utf32_word_split(to32, nto32, &nv, wbreak))) goto error; |
8818b7fc RK |
1683 | v8 = xcalloc(sizeof (char *), nv + 1); |
1684 | for(n = 0; n < nv; ++n) | |
1685 | if(!(v8[n] = utf32_to_utf8(v32[n], utf32_len(v32[n]), 0))) | |
1686 | goto error; | |
1687 | ret = v8; | |
1688 | *nwp = nv; | |
1689 | v8 = 0; /* don't free */ | |
c85b7022 | 1690 | error: |
8818b7fc RK |
1691 | if(v8) { |
1692 | for(n = 0; n < nv; ++n) | |
1693 | xfree(v8[n]); | |
1694 | xfree(v8); | |
1695 | } | |
1696 | if(v32) { | |
1697 | for(n = 0; n < nv; ++n) | |
1698 | xfree(v32[n]); | |
1699 | xfree(v32); | |
1700 | } | |
1701 | xfree(to32); | |
1702 | return ret; | |
1703 | } | |
1704 | ||
1705 | ||
e5a5a138 RK |
1706 | /*@}*/ |
1707 | ||
bb5c7798 RK |
1708 | /** @brief Return the length of a 0-terminated UTF-16 string |
1709 | * @param s Pointer to 0-terminated string | |
1710 | * @return Length of string in code points (excluding terminator) | |
1711 | * | |
1712 | * Unlike the conversion functions no validity checking is done on the string. | |
1713 | */ | |
1714 | size_t utf16_len(const uint16_t *s) { | |
1715 | const uint16_t *t = s; | |
1716 | ||
1717 | while(*t) | |
1718 | ++t; | |
1719 | return (size_t)(t - s); | |
1720 | } | |
1721 | ||
e5a5a138 RK |
1722 | /* |
1723 | Local Variables: | |
1724 | c-basic-offset:2 | |
1725 | comment-column:40 | |
1726 | fill-column:79 | |
1727 | indent-tabs-mode:nil | |
1728 | End: | |
1729 | */ |