1 /*
2 * Copyright © 2009 Red Hat, Inc.
3 * Copyright © 2009 Keith Stribley
4 * Copyright © 2011 Google, Inc.
5 *
6 * This is part of HarfBuzz, a text shaping library.
7 *
8 * Permission is hereby granted, without written agreement and without
9 * license or royalty fees, to use, copy, modify, and distribute this
10 * software and its documentation for any purpose, provided that the
11 * above copyright notice and the following two paragraphs appear in
12 * all copies of this software.
13 *
14 * IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE TO ANY PARTY FOR
15 * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES
16 * ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN
17 * IF THE COPYRIGHT HOLDER HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
18 * DAMAGE.
19 *
20 * THE COPYRIGHT HOLDER SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING,
21 * BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
22 * FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS
23 * ON AN "AS IS" BASIS, AND THE COPYRIGHT HOLDER HAS NO OBLIGATION TO
24 * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
25 *
26 * Red Hat Author(s): Behdad Esfahbod
27 * Google Author(s): Behdad Esfahbod
28 */
29
30 #include "hb-private.hh"
31
32 #include "hb-icu.h"
33
34 #include "hb-unicode-private.hh"
35
36 #include <unicode/uchar.h>
37 #include <unicode/unorm.h>
38 #include <unicode/ustring.h>
39 #include <unicode/uversion.h>
40
41
42 hb_script_t
hb_icu_script_to_script(UScriptCode script)43 hb_icu_script_to_script (UScriptCode script)
44 {
45 if (unlikely (script == USCRIPT_INVALID_CODE))
46 return HB_SCRIPT_INVALID;
47
48 return hb_script_from_string (uscript_getShortName (script), -1);
49 }
50
51 UScriptCode
hb_icu_script_from_script(hb_script_t script)52 hb_icu_script_from_script (hb_script_t script)
53 {
54 if (unlikely (script == HB_SCRIPT_INVALID))
55 return USCRIPT_INVALID_CODE;
56
57 for (unsigned int i = 0; i < USCRIPT_CODE_LIMIT; i++)
58 if (unlikely (hb_icu_script_to_script ((UScriptCode) i) == script))
59 return (UScriptCode) i;
60
61 return USCRIPT_UNKNOWN;
62 }
63
64
65 static hb_unicode_combining_class_t
hb_icu_unicode_combining_class(hb_unicode_funcs_t * ufuncs HB_UNUSED,hb_codepoint_t unicode,void * user_data HB_UNUSED)66 hb_icu_unicode_combining_class (hb_unicode_funcs_t *ufuncs HB_UNUSED,
67 hb_codepoint_t unicode,
68 void *user_data HB_UNUSED)
69
70 {
71 return (hb_unicode_combining_class_t) u_getCombiningClass (unicode);
72 }
73
74 static unsigned int
hb_icu_unicode_eastasian_width(hb_unicode_funcs_t * ufuncs HB_UNUSED,hb_codepoint_t unicode,void * user_data HB_UNUSED)75 hb_icu_unicode_eastasian_width (hb_unicode_funcs_t *ufuncs HB_UNUSED,
76 hb_codepoint_t unicode,
77 void *user_data HB_UNUSED)
78 {
79 switch (u_getIntPropertyValue(unicode, UCHAR_EAST_ASIAN_WIDTH))
80 {
81 case U_EA_WIDE:
82 case U_EA_FULLWIDTH:
83 return 2;
84 case U_EA_NEUTRAL:
85 case U_EA_AMBIGUOUS:
86 case U_EA_HALFWIDTH:
87 case U_EA_NARROW:
88 return 1;
89 }
90 return 1;
91 }
92
93 static hb_unicode_general_category_t
hb_icu_unicode_general_category(hb_unicode_funcs_t * ufuncs HB_UNUSED,hb_codepoint_t unicode,void * user_data HB_UNUSED)94 hb_icu_unicode_general_category (hb_unicode_funcs_t *ufuncs HB_UNUSED,
95 hb_codepoint_t unicode,
96 void *user_data HB_UNUSED)
97 {
98 switch (u_getIntPropertyValue(unicode, UCHAR_GENERAL_CATEGORY))
99 {
100 case U_UNASSIGNED: return HB_UNICODE_GENERAL_CATEGORY_UNASSIGNED;
101
102 case U_UPPERCASE_LETTER: return HB_UNICODE_GENERAL_CATEGORY_UPPERCASE_LETTER;
103 case U_LOWERCASE_LETTER: return HB_UNICODE_GENERAL_CATEGORY_LOWERCASE_LETTER;
104 case U_TITLECASE_LETTER: return HB_UNICODE_GENERAL_CATEGORY_TITLECASE_LETTER;
105 case U_MODIFIER_LETTER: return HB_UNICODE_GENERAL_CATEGORY_MODIFIER_LETTER;
106 case U_OTHER_LETTER: return HB_UNICODE_GENERAL_CATEGORY_OTHER_LETTER;
107
108 case U_NON_SPACING_MARK: return HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK;
109 case U_ENCLOSING_MARK: return HB_UNICODE_GENERAL_CATEGORY_ENCLOSING_MARK;
110 case U_COMBINING_SPACING_MARK: return HB_UNICODE_GENERAL_CATEGORY_SPACING_MARK;
111
112 case U_DECIMAL_DIGIT_NUMBER: return HB_UNICODE_GENERAL_CATEGORY_DECIMAL_NUMBER;
113 case U_LETTER_NUMBER: return HB_UNICODE_GENERAL_CATEGORY_LETTER_NUMBER;
114 case U_OTHER_NUMBER: return HB_UNICODE_GENERAL_CATEGORY_OTHER_NUMBER;
115
116 case U_SPACE_SEPARATOR: return HB_UNICODE_GENERAL_CATEGORY_SPACE_SEPARATOR;
117 case U_LINE_SEPARATOR: return HB_UNICODE_GENERAL_CATEGORY_LINE_SEPARATOR;
118 case U_PARAGRAPH_SEPARATOR: return HB_UNICODE_GENERAL_CATEGORY_PARAGRAPH_SEPARATOR;
119
120 case U_CONTROL_CHAR: return HB_UNICODE_GENERAL_CATEGORY_CONTROL;
121 case U_FORMAT_CHAR: return HB_UNICODE_GENERAL_CATEGORY_FORMAT;
122 case U_PRIVATE_USE_CHAR: return HB_UNICODE_GENERAL_CATEGORY_PRIVATE_USE;
123 case U_SURROGATE: return HB_UNICODE_GENERAL_CATEGORY_SURROGATE;
124
125
126 case U_DASH_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_DASH_PUNCTUATION;
127 case U_START_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_OPEN_PUNCTUATION;
128 case U_END_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_CLOSE_PUNCTUATION;
129 case U_CONNECTOR_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_CONNECT_PUNCTUATION;
130 case U_OTHER_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_OTHER_PUNCTUATION;
131
132 case U_MATH_SYMBOL: return HB_UNICODE_GENERAL_CATEGORY_MATH_SYMBOL;
133 case U_CURRENCY_SYMBOL: return HB_UNICODE_GENERAL_CATEGORY_CURRENCY_SYMBOL;
134 case U_MODIFIER_SYMBOL: return HB_UNICODE_GENERAL_CATEGORY_MODIFIER_SYMBOL;
135 case U_OTHER_SYMBOL: return HB_UNICODE_GENERAL_CATEGORY_OTHER_SYMBOL;
136
137 case U_INITIAL_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_INITIAL_PUNCTUATION;
138 case U_FINAL_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_FINAL_PUNCTUATION;
139 }
140
141 return HB_UNICODE_GENERAL_CATEGORY_UNASSIGNED;
142 }
143
144 static hb_codepoint_t
hb_icu_unicode_mirroring(hb_unicode_funcs_t * ufuncs HB_UNUSED,hb_codepoint_t unicode,void * user_data HB_UNUSED)145 hb_icu_unicode_mirroring (hb_unicode_funcs_t *ufuncs HB_UNUSED,
146 hb_codepoint_t unicode,
147 void *user_data HB_UNUSED)
148 {
149 return u_charMirror(unicode);
150 }
151
152 static hb_script_t
hb_icu_unicode_script(hb_unicode_funcs_t * ufuncs HB_UNUSED,hb_codepoint_t unicode,void * user_data HB_UNUSED)153 hb_icu_unicode_script (hb_unicode_funcs_t *ufuncs HB_UNUSED,
154 hb_codepoint_t unicode,
155 void *user_data HB_UNUSED)
156 {
157 UErrorCode status = U_ZERO_ERROR;
158 UScriptCode scriptCode = uscript_getScript(unicode, &status);
159
160 if (unlikely (U_FAILURE (status)))
161 return HB_SCRIPT_UNKNOWN;
162
163 return hb_icu_script_to_script (scriptCode);
164 }
165
166 #if U_ICU_VERSION_MAJOR_NUM >= 49
167 static const UNormalizer2 *normalizer;
168 #endif
169
170 static hb_bool_t
hb_icu_unicode_compose(hb_unicode_funcs_t * ufuncs HB_UNUSED,hb_codepoint_t a,hb_codepoint_t b,hb_codepoint_t * ab,void * user_data HB_UNUSED)171 hb_icu_unicode_compose (hb_unicode_funcs_t *ufuncs HB_UNUSED,
172 hb_codepoint_t a,
173 hb_codepoint_t b,
174 hb_codepoint_t *ab,
175 void *user_data HB_UNUSED)
176 {
177 #if U_ICU_VERSION_MAJOR_NUM >= 49
178 {
179 UChar32 ret = unorm2_composePair (normalizer, a, b);
180 if (ret < 0) return false;
181 *ab = ret;
182 return true;
183 }
184 #endif
185
186 /* We don't ifdef-out the fallback code such that compiler always
187 * sees it and makes sure it's compilable. */
188
189 UChar utf16[4], normalized[5];
190 unsigned int len;
191 hb_bool_t ret, err;
192 UErrorCode icu_err;
193
194 len = 0;
195 err = false;
196 U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), a, err);
197 if (err) return false;
198 U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), b, err);
199 if (err) return false;
200
201 icu_err = U_ZERO_ERROR;
202 len = unorm_normalize (utf16, len, UNORM_NFC, 0, normalized, ARRAY_LENGTH (normalized), &icu_err);
203 if (U_FAILURE (icu_err))
204 return false;
205 if (u_countChar32 (normalized, len) == 1) {
206 U16_GET_UNSAFE (normalized, 0, *ab);
207 ret = true;
208 } else {
209 ret = false;
210 }
211
212 return ret;
213 }
214
215 static hb_bool_t
hb_icu_unicode_decompose(hb_unicode_funcs_t * ufuncs HB_UNUSED,hb_codepoint_t ab,hb_codepoint_t * a,hb_codepoint_t * b,void * user_data HB_UNUSED)216 hb_icu_unicode_decompose (hb_unicode_funcs_t *ufuncs HB_UNUSED,
217 hb_codepoint_t ab,
218 hb_codepoint_t *a,
219 hb_codepoint_t *b,
220 void *user_data HB_UNUSED)
221 {
222 #if U_ICU_VERSION_MAJOR_NUM >= 49
223 {
224 UChar decomposed[4];
225 int len;
226 UErrorCode icu_err = U_ZERO_ERROR;
227 len = unorm2_getRawDecomposition (normalizer, ab, decomposed,
228 ARRAY_LENGTH (decomposed), &icu_err);
229 if (U_FAILURE (icu_err) || len < 0) return false;
230
231 len = u_countChar32 (decomposed, len);
232 if (len == 1) {
233 U16_GET_UNSAFE (decomposed, 0, *a);
234 *b = 0;
235 return *a != ab;
236 } else if (len == 2) {
237 len =0;
238 U16_NEXT_UNSAFE (decomposed, len, *a);
239 U16_NEXT_UNSAFE (decomposed, len, *b);
240 }
241 return true;
242 }
243 #endif
244
245 /* We don't ifdef-out the fallback code such that compiler always
246 * sees it and makes sure it's compilable. */
247
248 UChar utf16[2], normalized[2 * HB_UNICODE_MAX_DECOMPOSITION_LEN + 1];
249 unsigned int len;
250 hb_bool_t ret, err;
251 UErrorCode icu_err;
252
253 /* This function is a monster! Maybe it wasn't a good idea adding a
254 * pairwise decompose API... */
255 /* Watchout for the dragons. Err, watchout for macros changing len. */
256
257 len = 0;
258 err = false;
259 U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), ab, err);
260 if (err) return false;
261
262 icu_err = U_ZERO_ERROR;
263 len = unorm_normalize (utf16, len, UNORM_NFD, 0, normalized, ARRAY_LENGTH (normalized), &icu_err);
264 if (U_FAILURE (icu_err))
265 return false;
266
267 len = u_countChar32 (normalized, len);
268
269 if (len == 1) {
270 U16_GET_UNSAFE (normalized, 0, *a);
271 *b = 0;
272 ret = *a != ab;
273 } else if (len == 2) {
274 len =0;
275 U16_NEXT_UNSAFE (normalized, len, *a);
276 U16_NEXT_UNSAFE (normalized, len, *b);
277
278 /* Here's the ugly part: if ab decomposes to a single character and
279 * that character decomposes again, we have to detect that and undo
280 * the second part :-(. */
281 UChar recomposed[20];
282 icu_err = U_ZERO_ERROR;
283 unorm_normalize (normalized, len, UNORM_NFC, 0, recomposed, ARRAY_LENGTH (recomposed), &icu_err);
284 if (U_FAILURE (icu_err))
285 return false;
286 hb_codepoint_t c;
287 U16_GET_UNSAFE (recomposed, 0, c);
288 if (c != *a && c != ab) {
289 *a = c;
290 *b = 0;
291 }
292 ret = true;
293 } else {
294 /* If decomposed to more than two characters, take the last one,
295 * and recompose the rest to get the first component. */
296 U16_PREV_UNSAFE (normalized, len, *b); /* Changes len in-place. */
297 UChar recomposed[18 * 2];
298 icu_err = U_ZERO_ERROR;
299 len = unorm_normalize (normalized, len, UNORM_NFC, 0, recomposed, ARRAY_LENGTH (recomposed), &icu_err);
300 if (U_FAILURE (icu_err))
301 return false;
302 /* We expect that recomposed has exactly one character now. */
303 if (unlikely (u_countChar32 (recomposed, len) != 1))
304 return false;
305 U16_GET_UNSAFE (recomposed, 0, *a);
306 ret = true;
307 }
308
309 return ret;
310 }
311
312 static unsigned int
hb_icu_unicode_decompose_compatibility(hb_unicode_funcs_t * ufuncs HB_UNUSED,hb_codepoint_t u,hb_codepoint_t * decomposed,void * user_data HB_UNUSED)313 hb_icu_unicode_decompose_compatibility (hb_unicode_funcs_t *ufuncs HB_UNUSED,
314 hb_codepoint_t u,
315 hb_codepoint_t *decomposed,
316 void *user_data HB_UNUSED)
317 {
318 UChar utf16[2], normalized[2 * HB_UNICODE_MAX_DECOMPOSITION_LEN + 1];
319 unsigned int len;
320 int32_t utf32_len;
321 hb_bool_t err;
322 UErrorCode icu_err;
323
324 /* Copy @u into a UTF-16 array to be passed to ICU. */
325 len = 0;
326 err = false;
327 U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), u, err);
328 if (err)
329 return 0;
330
331 /* Normalise the codepoint using NFKD mode. */
332 icu_err = U_ZERO_ERROR;
333 len = unorm_normalize (utf16, len, UNORM_NFKD, 0, normalized, ARRAY_LENGTH (normalized), &icu_err);
334 if (icu_err)
335 return 0;
336
337 /* Convert the decomposed form from UTF-16 to UTF-32. */
338 icu_err = U_ZERO_ERROR;
339 u_strToUTF32 ((UChar32*) decomposed, HB_UNICODE_MAX_DECOMPOSITION_LEN, &utf32_len, normalized, len, &icu_err);
340 if (icu_err)
341 return 0;
342
343 return utf32_len;
344 }
345
346
347 hb_unicode_funcs_t *
hb_icu_get_unicode_funcs(void)348 hb_icu_get_unicode_funcs (void)
349 {
350 static const hb_unicode_funcs_t _hb_icu_unicode_funcs = {
351 HB_OBJECT_HEADER_STATIC,
352
353 NULL, /* parent */
354 true, /* immutable */
355 {
356 #define HB_UNICODE_FUNC_IMPLEMENT(name) hb_icu_unicode_##name,
357 HB_UNICODE_FUNCS_IMPLEMENT_CALLBACKS
358 #undef HB_UNICODE_FUNC_IMPLEMENT
359 }
360 };
361
362 #if U_ICU_VERSION_MAJOR_NUM >= 49
363 if (!hb_atomic_ptr_get (&normalizer)) {
364 UErrorCode icu_err = U_ZERO_ERROR;
365 /* We ignore failure in getNFCInstace(). */
366 hb_atomic_ptr_cmpexch (&normalizer, NULL, unorm2_getNFCInstance (&icu_err));
367 }
368 #endif
369 return const_cast<hb_unicode_funcs_t *> (&_hb_icu_unicode_funcs);
370 }
371
372
373