|
1 /* decomp.c - Character decomposition. |
|
2 * |
|
3 * Copyright (C) 1999, 2000 Tom Tromey |
|
4 * Copyright 2000 Red Hat, Inc. |
|
5 * Portions copyright (c) 2006 Nokia Corporation. All rights reserved. |
|
6 * |
|
7 * The Gnome Library is free software; you can redistribute it and/or |
|
8 * modify it under the terms of the GNU Lesser General Public License as |
|
9 * published by the Free Software Foundation; either version 2 of the |
|
10 * License, or (at your option) any later version. |
|
11 * |
|
12 * The Gnome Library is distributed in the hope that it will be useful, |
|
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
|
15 * Lesser General Public License for more details. |
|
16 * |
|
17 * You should have received a copy of the GNU Lesser General Public |
|
18 * License along with the Gnome Library; see the file COPYING.LIB. If not, |
|
19 * write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, |
|
20 * Boston, MA 02111-1307, USA. |
|
21 */ |
|
22 |
|
23 #include "config.h" |
|
24 |
|
25 #include <stdlib.h> |
|
26 |
|
27 #include "glib.h" |
|
28 #include "gunidecomp.h" |
|
29 #include "gunicomp.h" |
|
30 #include "gunicodeprivate.h" |
|
31 #include "galias.h" |
|
32 |
|
33 |
|
34 #define CC_PART1(Page, Char) \ |
|
35 ((combining_class_table_part1[Page] >= G_UNICODE_MAX_TABLE_INDEX) \ |
|
36 ? (combining_class_table_part1[Page] - G_UNICODE_MAX_TABLE_INDEX) \ |
|
37 : (cclass_data[combining_class_table_part1[Page]][Char])) |
|
38 |
|
39 #define CC_PART2(Page, Char) \ |
|
40 ((combining_class_table_part2[Page] >= G_UNICODE_MAX_TABLE_INDEX) \ |
|
41 ? (combining_class_table_part2[Page] - G_UNICODE_MAX_TABLE_INDEX) \ |
|
42 : (cclass_data[combining_class_table_part2[Page]][Char])) |
|
43 |
|
44 #define COMBINING_CLASS(Char) \ |
|
45 (((Char) <= G_UNICODE_LAST_CHAR_PART1) \ |
|
46 ? CC_PART1 ((Char) >> 8, (Char) & 0xff) \ |
|
47 : (((Char) >= 0xe0000 && (Char) <= G_UNICODE_LAST_CHAR) \ |
|
48 ? CC_PART2 (((Char) - 0xe0000) >> 8, (Char) & 0xff) \ |
|
49 : 0)) |
|
50 |
|
51 gint |
|
52 _g_unichar_combining_class (gunichar uc) |
|
53 { |
|
54 return COMBINING_CLASS (uc); |
|
55 } |
|
56 |
|
57 /* constants for hangul syllable [de]composition */ |
|
58 #define SBase 0xAC00 |
|
59 #define LBase 0x1100 |
|
60 #define VBase 0x1161 |
|
61 #define TBase 0x11A7 |
|
62 #define LCount 19 |
|
63 #define VCount 21 |
|
64 #define TCount 28 |
|
65 #define NCount (VCount * TCount) |
|
66 #define SCount (LCount * NCount) |
|
67 |
|
68 /** |
|
69 * g_unicode_canonical_ordering: |
|
70 * @string: a UCS-4 encoded string. |
|
71 * @len: the maximum length of @string to use. |
|
72 * |
|
73 * Computes the canonical ordering of a string in-place. |
|
74 * This rearranges decomposed characters in the string |
|
75 * according to their combining classes. See the Unicode |
|
76 * manual for more information. |
|
77 **/ |
|
78 EXPORT_C void |
|
79 g_unicode_canonical_ordering (gunichar *string, |
|
80 gsize len) |
|
81 { |
|
82 gsize i; |
|
83 int swap = 1; |
|
84 |
|
85 while (swap) |
|
86 { |
|
87 int last; |
|
88 swap = 0; |
|
89 last = COMBINING_CLASS (string[0]); |
|
90 for (i = 0; i < len - 1; ++i) |
|
91 { |
|
92 int next = COMBINING_CLASS (string[i + 1]); |
|
93 if (next != 0 && last > next) |
|
94 { |
|
95 gsize j; |
|
96 /* Percolate item leftward through string. */ |
|
97 for (j = i + 1; j > 0; --j) |
|
98 { |
|
99 gunichar t; |
|
100 if (COMBINING_CLASS (string[j - 1]) <= next) |
|
101 break; |
|
102 t = string[j]; |
|
103 string[j] = string[j - 1]; |
|
104 string[j - 1] = t; |
|
105 swap = 1; |
|
106 } |
|
107 /* We're re-entering the loop looking at the old |
|
108 character again. */ |
|
109 next = last; |
|
110 } |
|
111 last = next; |
|
112 } |
|
113 } |
|
114 } |
|
115 |
|
116 /* http://www.unicode.org/unicode/reports/tr15/#Hangul |
|
117 * r should be null or have sufficient space. Calling with r == NULL will |
|
118 * only calculate the result_len; however, a buffer with space for three |
|
119 * characters will always be big enough. */ |
|
120 static void |
|
121 decompose_hangul (gunichar s, |
|
122 gunichar *r, |
|
123 gsize *result_len) |
|
124 { |
|
125 gint SIndex = s - SBase; |
|
126 |
|
127 /* not a hangul syllable */ |
|
128 if (SIndex < 0 || SIndex >= SCount) |
|
129 { |
|
130 if (r) |
|
131 r[0] = s; |
|
132 *result_len = 1; |
|
133 } |
|
134 else |
|
135 { |
|
136 gunichar L = LBase + SIndex / NCount; |
|
137 gunichar V = VBase + (SIndex % NCount) / TCount; |
|
138 gunichar T = TBase + SIndex % TCount; |
|
139 |
|
140 if (r) |
|
141 { |
|
142 r[0] = L; |
|
143 r[1] = V; |
|
144 } |
|
145 |
|
146 if (T != TBase) |
|
147 { |
|
148 if (r) |
|
149 r[2] = T; |
|
150 *result_len = 3; |
|
151 } |
|
152 else |
|
153 *result_len = 2; |
|
154 } |
|
155 } |
|
156 |
|
157 /* returns a pointer to a null-terminated UTF-8 string */ |
|
158 static const gchar * |
|
159 find_decomposition (gunichar ch, |
|
160 gboolean compat) |
|
161 { |
|
162 int start = 0; |
|
163 int end = G_N_ELEMENTS (decomp_table); |
|
164 |
|
165 if (ch >= decomp_table[start].ch && |
|
166 ch <= decomp_table[end - 1].ch) |
|
167 { |
|
168 while (TRUE) |
|
169 { |
|
170 int half = (start + end) / 2; |
|
171 if (ch == decomp_table[half].ch) |
|
172 { |
|
173 int offset; |
|
174 |
|
175 if (compat) |
|
176 { |
|
177 offset = decomp_table[half].compat_offset; |
|
178 if (offset == G_UNICODE_NOT_PRESENT_OFFSET) |
|
179 offset = decomp_table[half].canon_offset; |
|
180 } |
|
181 else |
|
182 { |
|
183 offset = decomp_table[half].canon_offset; |
|
184 if (offset == G_UNICODE_NOT_PRESENT_OFFSET) |
|
185 return NULL; |
|
186 } |
|
187 |
|
188 return &(decomp_expansion_string[offset]); |
|
189 } |
|
190 else if (half == start) |
|
191 break; |
|
192 else if (ch > decomp_table[half].ch) |
|
193 start = half; |
|
194 else |
|
195 end = half; |
|
196 } |
|
197 } |
|
198 |
|
199 return NULL; |
|
200 } |
|
201 |
|
202 /** |
|
203 * g_unicode_canonical_decomposition: |
|
204 * @ch: a Unicode character. |
|
205 * @result_len: location to store the length of the return value. |
|
206 * |
|
207 * Computes the canonical decomposition of a Unicode character. |
|
208 * |
|
209 * Return value: a newly allocated string of Unicode characters. |
|
210 * @result_len is set to the resulting length of the string. |
|
211 **/ |
|
212 EXPORT_C gunichar * |
|
213 g_unicode_canonical_decomposition (gunichar ch, |
|
214 gsize *result_len) |
|
215 { |
|
216 const gchar *decomp; |
|
217 const gchar *p; |
|
218 gunichar *r; |
|
219 |
|
220 /* Hangul syllable */ |
|
221 if (ch >= 0xac00 && ch <= 0xd7a3) |
|
222 { |
|
223 decompose_hangul (ch, NULL, result_len); |
|
224 r = g_malloc (*result_len * sizeof (gunichar)); |
|
225 decompose_hangul (ch, r, result_len); |
|
226 } |
|
227 else if ((decomp = find_decomposition (ch, FALSE)) != NULL) |
|
228 { |
|
229 /* Found it. */ |
|
230 int i; |
|
231 |
|
232 *result_len = g_utf8_strlen (decomp, -1); |
|
233 r = g_malloc (*result_len * sizeof (gunichar)); |
|
234 |
|
235 for (p = decomp, i = 0; *p != '\0'; p = g_utf8_next_char (p), i++) |
|
236 r[i] = g_utf8_get_char (p); |
|
237 } |
|
238 else |
|
239 { |
|
240 /* Not in our table. */ |
|
241 r = g_malloc (sizeof (gunichar)); |
|
242 *r = ch; |
|
243 *result_len = 1; |
|
244 } |
|
245 |
|
246 /* Supposedly following the Unicode 2.1.9 table means that the |
|
247 decompositions come out in canonical order. I haven't tested |
|
248 this, but we rely on it here. */ |
|
249 return r; |
|
250 } |
|
251 |
|
252 /* L,V => LV and LV,T => LVT */ |
|
253 static gboolean |
|
254 combine_hangul (gunichar a, |
|
255 gunichar b, |
|
256 gunichar *result) |
|
257 { |
|
258 gint LIndex = a - LBase; |
|
259 gint SIndex = a - SBase; |
|
260 |
|
261 gint VIndex = b - VBase; |
|
262 gint TIndex = b - TBase; |
|
263 |
|
264 if (0 <= LIndex && LIndex < LCount |
|
265 && 0 <= VIndex && VIndex < VCount) |
|
266 { |
|
267 *result = SBase + (LIndex * VCount + VIndex) * TCount; |
|
268 return TRUE; |
|
269 } |
|
270 else if (0 <= SIndex && SIndex < SCount && (SIndex % TCount) == 0 |
|
271 && 0 < TIndex && TIndex < TCount) |
|
272 { |
|
273 *result = a + TIndex; |
|
274 return TRUE; |
|
275 } |
|
276 |
|
277 return FALSE; |
|
278 } |
|
279 |
|
280 #define CI(Page, Char) \ |
|
281 ((compose_table[Page] >= G_UNICODE_MAX_TABLE_INDEX) \ |
|
282 ? (compose_table[Page] - G_UNICODE_MAX_TABLE_INDEX) \ |
|
283 : (compose_data[compose_table[Page]][Char])) |
|
284 |
|
285 #define COMPOSE_INDEX(Char) \ |
|
286 (((Char >> 8) > (COMPOSE_TABLE_LAST)) ? 0 : CI((Char) >> 8, (Char) & 0xff)) |
|
287 |
|
288 static gboolean |
|
289 combine (gunichar a, |
|
290 gunichar b, |
|
291 gunichar *result) |
|
292 { |
|
293 gushort index_a, index_b; |
|
294 |
|
295 if (combine_hangul (a, b, result)) |
|
296 return TRUE; |
|
297 |
|
298 index_a = COMPOSE_INDEX(a); |
|
299 |
|
300 if (index_a >= COMPOSE_FIRST_SINGLE_START && index_a < COMPOSE_SECOND_START) |
|
301 { |
|
302 if (b == compose_first_single[index_a - COMPOSE_FIRST_SINGLE_START][0]) |
|
303 { |
|
304 *result = compose_first_single[index_a - COMPOSE_FIRST_SINGLE_START][1]; |
|
305 return TRUE; |
|
306 } |
|
307 else |
|
308 return FALSE; |
|
309 } |
|
310 |
|
311 index_b = COMPOSE_INDEX(b); |
|
312 |
|
313 if (index_b >= COMPOSE_SECOND_SINGLE_START) |
|
314 { |
|
315 if (a == compose_second_single[index_b - COMPOSE_SECOND_SINGLE_START][0]) |
|
316 { |
|
317 *result = compose_second_single[index_b - COMPOSE_SECOND_SINGLE_START][1]; |
|
318 return TRUE; |
|
319 } |
|
320 else |
|
321 return FALSE; |
|
322 } |
|
323 |
|
324 if (index_a >= COMPOSE_FIRST_START && index_a < COMPOSE_FIRST_SINGLE_START && |
|
325 index_b >= COMPOSE_SECOND_START && index_b < COMPOSE_SECOND_SINGLE_START) |
|
326 { |
|
327 gunichar res = compose_array[index_a - COMPOSE_FIRST_START][index_b - COMPOSE_SECOND_START]; |
|
328 |
|
329 if (res) |
|
330 { |
|
331 *result = res; |
|
332 return TRUE; |
|
333 } |
|
334 } |
|
335 |
|
336 return FALSE; |
|
337 } |
|
338 |
|
339 gunichar * |
|
340 _g_utf8_normalize_wc (const gchar *str, |
|
341 gssize max_len, |
|
342 GNormalizeMode mode) |
|
343 { |
|
344 gsize n_wc; |
|
345 gunichar *wc_buffer; |
|
346 const char *p; |
|
347 gsize last_start; |
|
348 gboolean do_compat = (mode == G_NORMALIZE_NFKC || |
|
349 mode == G_NORMALIZE_NFKD); |
|
350 gboolean do_compose = (mode == G_NORMALIZE_NFC || |
|
351 mode == G_NORMALIZE_NFKC); |
|
352 |
|
353 n_wc = 0; |
|
354 p = str; |
|
355 while ((max_len < 0 || p < str + max_len) && *p) |
|
356 { |
|
357 const gchar *decomp; |
|
358 gunichar wc = g_utf8_get_char (p); |
|
359 |
|
360 if (wc >= 0xac00 && wc <= 0xd7a3) |
|
361 { |
|
362 gsize result_len; |
|
363 decompose_hangul (wc, NULL, &result_len); |
|
364 n_wc += result_len; |
|
365 } |
|
366 else |
|
367 { |
|
368 decomp = find_decomposition (wc, do_compat); |
|
369 |
|
370 if (decomp) |
|
371 n_wc += g_utf8_strlen (decomp, -1); |
|
372 else |
|
373 n_wc++; |
|
374 } |
|
375 |
|
376 p = g_utf8_next_char (p); |
|
377 } |
|
378 wc_buffer = g_new (gunichar, n_wc + 1); |
|
379 |
|
380 last_start = 0; |
|
381 n_wc = 0; |
|
382 p = str; |
|
383 while ((max_len < 0 || p < str + max_len) && *p) |
|
384 { |
|
385 gunichar wc = g_utf8_get_char (p); |
|
386 const gchar *decomp; |
|
387 int cc; |
|
388 gsize old_n_wc = n_wc; |
|
389 |
|
390 if (wc >= 0xac00 && wc <= 0xd7a3) |
|
391 { |
|
392 gsize result_len; |
|
393 decompose_hangul (wc, wc_buffer + n_wc, &result_len); |
|
394 n_wc += result_len; |
|
395 } |
|
396 else |
|
397 { |
|
398 decomp = find_decomposition (wc, do_compat); |
|
399 |
|
400 if (decomp) |
|
401 { |
|
402 const char *pd; |
|
403 for (pd = decomp; *pd != '\0'; pd = g_utf8_next_char (pd)) |
|
404 wc_buffer[n_wc++] = g_utf8_get_char (pd); |
|
405 } |
|
406 else |
|
407 wc_buffer[n_wc++] = wc; |
|
408 } |
|
409 |
|
410 if (n_wc > 0) |
|
411 { |
|
412 cc = COMBINING_CLASS (wc_buffer[old_n_wc]); |
|
413 |
|
414 if (cc == 0) |
|
415 { |
|
416 g_unicode_canonical_ordering (wc_buffer + last_start, n_wc - last_start); |
|
417 last_start = old_n_wc; |
|
418 } |
|
419 } |
|
420 |
|
421 p = g_utf8_next_char (p); |
|
422 } |
|
423 |
|
424 if (n_wc > 0) |
|
425 { |
|
426 g_unicode_canonical_ordering (wc_buffer + last_start, n_wc - last_start); |
|
427 last_start = n_wc; |
|
428 } |
|
429 |
|
430 wc_buffer[n_wc] = 0; |
|
431 |
|
432 /* All decomposed and reordered */ |
|
433 |
|
434 if (do_compose && n_wc > 0) |
|
435 { |
|
436 gsize i, j; |
|
437 int last_cc = 0; |
|
438 last_start = 0; |
|
439 |
|
440 for (i = 0; i < n_wc; i++) |
|
441 { |
|
442 int cc = COMBINING_CLASS (wc_buffer[i]); |
|
443 |
|
444 if (i > 0 && |
|
445 (last_cc == 0 || last_cc != cc) && |
|
446 combine (wc_buffer[last_start], wc_buffer[i], |
|
447 &wc_buffer[last_start])) |
|
448 { |
|
449 for (j = i + 1; j < n_wc; j++) |
|
450 wc_buffer[j-1] = wc_buffer[j]; |
|
451 n_wc--; |
|
452 i--; |
|
453 |
|
454 if (i == last_start) |
|
455 last_cc = 0; |
|
456 else |
|
457 last_cc = COMBINING_CLASS (wc_buffer[i-1]); |
|
458 |
|
459 continue; |
|
460 } |
|
461 |
|
462 if (cc == 0) |
|
463 last_start = i; |
|
464 |
|
465 last_cc = cc; |
|
466 } |
|
467 } |
|
468 |
|
469 wc_buffer[n_wc] = 0; |
|
470 |
|
471 return wc_buffer; |
|
472 } |
|
473 |
|
474 /** |
|
475 * g_utf8_normalize: |
|
476 * @str: a UTF-8 encoded string. |
|
477 * @len: length of @str, in bytes, or -1 if @str is nul-terminated. |
|
478 * @mode: the type of normalization to perform. |
|
479 * |
|
480 * Converts a string into canonical form, standardizing |
|
481 * such issues as whether a character with an accent |
|
482 * is represented as a base character and combining |
|
483 * accent or as a single precomposed character. You |
|
484 * should generally call g_utf8_normalize() before |
|
485 * comparing two Unicode strings. |
|
486 * |
|
487 * The normalization mode %G_NORMALIZE_DEFAULT only |
|
488 * standardizes differences that do not affect the |
|
489 * text content, such as the above-mentioned accent |
|
490 * representation. %G_NORMALIZE_ALL also standardizes |
|
491 * the "compatibility" characters in Unicode, such |
|
492 * as SUPERSCRIPT THREE to the standard forms |
|
493 * (in this case DIGIT THREE). Formatting information |
|
494 * may be lost but for most text operations such |
|
495 * characters should be considered the same. |
|
496 * For example, g_utf8_collate() normalizes |
|
497 * with %G_NORMALIZE_ALL as its first step. |
|
498 * |
|
499 * %G_NORMALIZE_DEFAULT_COMPOSE and %G_NORMALIZE_ALL_COMPOSE |
|
500 * are like %G_NORMALIZE_DEFAULT and %G_NORMALIZE_ALL, |
|
501 * but returned a result with composed forms rather |
|
502 * than a maximally decomposed form. This is often |
|
503 * useful if you intend to convert the string to |
|
504 * a legacy encoding or pass it to a system with |
|
505 * less capable Unicode handling. |
|
506 * |
|
507 * Return value: a newly allocated string, that is the |
|
508 * normalized form of @str. |
|
509 **/ |
|
510 EXPORT_C gchar * |
|
511 g_utf8_normalize (const gchar *str, |
|
512 gssize len, |
|
513 GNormalizeMode mode) |
|
514 { |
|
515 gunichar *result_wc = _g_utf8_normalize_wc (str, len, mode); |
|
516 gchar *result; |
|
517 |
|
518 result = g_ucs4_to_utf8 (result_wc, -1, NULL, NULL, NULL); |
|
519 g_free (result_wc); |
|
520 |
|
521 return result; |
|
522 } |
|
523 |
|
524 #define __G_UNIDECOMP_C__ |
|
525 #include "galiasdef.c" |