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.hh"
31
32 #include "hb-icu.h"
33
34 #include "hb-machinery.hh"
35
36 #include <unicode/uchar.h>
37 #include <unicode/unorm2.h>
38 #include <unicode/ustring.h>
39 #include <unicode/utf16.h>
40 #include <unicode/uversion.h>
41
42
43 /**
44 * SECTION:hb-icu
45 * @title: hb-icu
46 * @short_description: ICU integration
47 * @include: hb-icu.h
48 *
49 * Functions for using HarfBuzz with the ICU library to provide Unicode data.
50 **/
51
52
53 hb_script_t
hb_icu_script_to_script(UScriptCode script)54 hb_icu_script_to_script (UScriptCode script)
55 {
56 if (unlikely (script == USCRIPT_INVALID_CODE))
57 return HB_SCRIPT_INVALID;
58
59 return hb_script_from_string (uscript_getShortName (script), -1);
60 }
61
62 UScriptCode
hb_icu_script_from_script(hb_script_t script)63 hb_icu_script_from_script (hb_script_t script)
64 {
65 if (unlikely (script == HB_SCRIPT_INVALID))
66 return USCRIPT_INVALID_CODE;
67
68 unsigned int numScriptCode = 1 + u_getIntPropertyMaxValue (UCHAR_SCRIPT);
69 for (unsigned int i = 0; i < numScriptCode; i++)
70 if (unlikely (hb_icu_script_to_script ((UScriptCode) i) == script))
71 return (UScriptCode) i;
72
73 return USCRIPT_UNKNOWN;
74 }
75
76
77 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)78 hb_icu_unicode_combining_class (hb_unicode_funcs_t *ufuncs HB_UNUSED,
79 hb_codepoint_t unicode,
80 void *user_data HB_UNUSED)
81
82 {
83 return (hb_unicode_combining_class_t) u_getCombiningClass (unicode);
84 }
85
86 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)87 hb_icu_unicode_general_category (hb_unicode_funcs_t *ufuncs HB_UNUSED,
88 hb_codepoint_t unicode,
89 void *user_data HB_UNUSED)
90 {
91 switch (u_getIntPropertyValue(unicode, UCHAR_GENERAL_CATEGORY))
92 {
93 case U_UNASSIGNED: return HB_UNICODE_GENERAL_CATEGORY_UNASSIGNED;
94
95 case U_UPPERCASE_LETTER: return HB_UNICODE_GENERAL_CATEGORY_UPPERCASE_LETTER;
96 case U_LOWERCASE_LETTER: return HB_UNICODE_GENERAL_CATEGORY_LOWERCASE_LETTER;
97 case U_TITLECASE_LETTER: return HB_UNICODE_GENERAL_CATEGORY_TITLECASE_LETTER;
98 case U_MODIFIER_LETTER: return HB_UNICODE_GENERAL_CATEGORY_MODIFIER_LETTER;
99 case U_OTHER_LETTER: return HB_UNICODE_GENERAL_CATEGORY_OTHER_LETTER;
100
101 case U_NON_SPACING_MARK: return HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK;
102 case U_ENCLOSING_MARK: return HB_UNICODE_GENERAL_CATEGORY_ENCLOSING_MARK;
103 case U_COMBINING_SPACING_MARK: return HB_UNICODE_GENERAL_CATEGORY_SPACING_MARK;
104
105 case U_DECIMAL_DIGIT_NUMBER: return HB_UNICODE_GENERAL_CATEGORY_DECIMAL_NUMBER;
106 case U_LETTER_NUMBER: return HB_UNICODE_GENERAL_CATEGORY_LETTER_NUMBER;
107 case U_OTHER_NUMBER: return HB_UNICODE_GENERAL_CATEGORY_OTHER_NUMBER;
108
109 case U_SPACE_SEPARATOR: return HB_UNICODE_GENERAL_CATEGORY_SPACE_SEPARATOR;
110 case U_LINE_SEPARATOR: return HB_UNICODE_GENERAL_CATEGORY_LINE_SEPARATOR;
111 case U_PARAGRAPH_SEPARATOR: return HB_UNICODE_GENERAL_CATEGORY_PARAGRAPH_SEPARATOR;
112
113 case U_CONTROL_CHAR: return HB_UNICODE_GENERAL_CATEGORY_CONTROL;
114 case U_FORMAT_CHAR: return HB_UNICODE_GENERAL_CATEGORY_FORMAT;
115 case U_PRIVATE_USE_CHAR: return HB_UNICODE_GENERAL_CATEGORY_PRIVATE_USE;
116 case U_SURROGATE: return HB_UNICODE_GENERAL_CATEGORY_SURROGATE;
117
118
119 case U_DASH_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_DASH_PUNCTUATION;
120 case U_START_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_OPEN_PUNCTUATION;
121 case U_END_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_CLOSE_PUNCTUATION;
122 case U_CONNECTOR_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_CONNECT_PUNCTUATION;
123 case U_OTHER_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_OTHER_PUNCTUATION;
124
125 case U_MATH_SYMBOL: return HB_UNICODE_GENERAL_CATEGORY_MATH_SYMBOL;
126 case U_CURRENCY_SYMBOL: return HB_UNICODE_GENERAL_CATEGORY_CURRENCY_SYMBOL;
127 case U_MODIFIER_SYMBOL: return HB_UNICODE_GENERAL_CATEGORY_MODIFIER_SYMBOL;
128 case U_OTHER_SYMBOL: return HB_UNICODE_GENERAL_CATEGORY_OTHER_SYMBOL;
129
130 case U_INITIAL_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_INITIAL_PUNCTUATION;
131 case U_FINAL_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_FINAL_PUNCTUATION;
132 }
133
134 return HB_UNICODE_GENERAL_CATEGORY_UNASSIGNED;
135 }
136
137 static hb_codepoint_t
hb_icu_unicode_mirroring(hb_unicode_funcs_t * ufuncs HB_UNUSED,hb_codepoint_t unicode,void * user_data HB_UNUSED)138 hb_icu_unicode_mirroring (hb_unicode_funcs_t *ufuncs HB_UNUSED,
139 hb_codepoint_t unicode,
140 void *user_data HB_UNUSED)
141 {
142 return u_charMirror(unicode);
143 }
144
145 static hb_script_t
hb_icu_unicode_script(hb_unicode_funcs_t * ufuncs HB_UNUSED,hb_codepoint_t unicode,void * user_data HB_UNUSED)146 hb_icu_unicode_script (hb_unicode_funcs_t *ufuncs HB_UNUSED,
147 hb_codepoint_t unicode,
148 void *user_data HB_UNUSED)
149 {
150 UErrorCode status = U_ZERO_ERROR;
151 UScriptCode scriptCode = uscript_getScript(unicode, &status);
152
153 if (unlikely (U_FAILURE (status)))
154 return HB_SCRIPT_UNKNOWN;
155
156 return hb_icu_script_to_script (scriptCode);
157 }
158
159 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)160 hb_icu_unicode_compose (hb_unicode_funcs_t *ufuncs HB_UNUSED,
161 hb_codepoint_t a,
162 hb_codepoint_t b,
163 hb_codepoint_t *ab,
164 void *user_data HB_UNUSED)
165 {
166 #if U_ICU_VERSION_MAJOR_NUM >= 49
167 {
168 const UNormalizer2 *normalizer = (const UNormalizer2 *) user_data;
169 UChar32 ret = unorm2_composePair (normalizer, a, b);
170 if (ret < 0) return false;
171 *ab = ret;
172 return true;
173 }
174 #endif
175
176 /* We don't ifdef-out the fallback code such that compiler always
177 * sees it and makes sure it's compilable. */
178
179 UChar utf16[4], normalized[5];
180 unsigned int len;
181 hb_bool_t ret, err;
182 UErrorCode icu_err;
183
184 len = 0;
185 err = false;
186 U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), a, err);
187 if (err) return false;
188 U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), b, err);
189 if (err) return false;
190
191 icu_err = U_ZERO_ERROR;
192 len = unorm2_normalize (unorm2_getNFCInstance (&icu_err), utf16, len, normalized, ARRAY_LENGTH (normalized), &icu_err);
193 if (U_FAILURE (icu_err))
194 return false;
195 if (u_countChar32 (normalized, len) == 1) {
196 U16_GET_UNSAFE (normalized, 0, *ab);
197 ret = true;
198 } else {
199 ret = false;
200 }
201
202 return ret;
203 }
204
205 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)206 hb_icu_unicode_decompose (hb_unicode_funcs_t *ufuncs HB_UNUSED,
207 hb_codepoint_t ab,
208 hb_codepoint_t *a,
209 hb_codepoint_t *b,
210 void *user_data HB_UNUSED)
211 {
212 #if U_ICU_VERSION_MAJOR_NUM >= 49
213 {
214 const UNormalizer2 *normalizer = (const UNormalizer2 *) user_data;
215 UChar decomposed[4];
216 int len;
217 UErrorCode icu_err = U_ZERO_ERROR;
218 len = unorm2_getRawDecomposition (normalizer, ab, decomposed,
219 ARRAY_LENGTH (decomposed), &icu_err);
220 if (U_FAILURE (icu_err) || len < 0) return false;
221
222 len = u_countChar32 (decomposed, len);
223 if (len == 1) {
224 U16_GET_UNSAFE (decomposed, 0, *a);
225 *b = 0;
226 return *a != ab;
227 } else if (len == 2) {
228 len =0;
229 U16_NEXT_UNSAFE (decomposed, len, *a);
230 U16_NEXT_UNSAFE (decomposed, len, *b);
231 }
232 return true;
233 }
234 #endif
235
236 /* We don't ifdef-out the fallback code such that compiler always
237 * sees it and makes sure it's compilable. */
238
239 UChar utf16[2], normalized[2 * HB_UNICODE_MAX_DECOMPOSITION_LEN + 1];
240 unsigned int len;
241 hb_bool_t ret, err;
242 UErrorCode icu_err;
243
244 /* This function is a monster! Maybe it wasn't a good idea adding a
245 * pairwise decompose API... */
246 /* Watchout for the dragons. Err, watchout for macros changing len. */
247
248 len = 0;
249 err = false;
250 U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), ab, err);
251 if (err) return false;
252
253 icu_err = U_ZERO_ERROR;
254 len = unorm2_normalize (unorm2_getNFDInstance (&icu_err), utf16, len, normalized, ARRAY_LENGTH (normalized), &icu_err);
255 if (U_FAILURE (icu_err))
256 return false;
257
258 len = u_countChar32 (normalized, len);
259
260 if (len == 1) {
261 U16_GET_UNSAFE (normalized, 0, *a);
262 *b = 0;
263 ret = *a != ab;
264 } else if (len == 2) {
265 len =0;
266 U16_NEXT_UNSAFE (normalized, len, *a);
267 U16_NEXT_UNSAFE (normalized, len, *b);
268
269 /* Here's the ugly part: if ab decomposes to a single character and
270 * that character decomposes again, we have to detect that and undo
271 * the second part :-(. */
272 UChar recomposed[20];
273 icu_err = U_ZERO_ERROR;
274 unorm2_normalize (unorm2_getNFCInstance (&icu_err), normalized, len, recomposed, ARRAY_LENGTH (recomposed), &icu_err);
275 if (U_FAILURE (icu_err))
276 return false;
277 hb_codepoint_t c;
278 U16_GET_UNSAFE (recomposed, 0, c);
279 if (c != *a && c != ab) {
280 *a = c;
281 *b = 0;
282 }
283 ret = true;
284 } else {
285 /* If decomposed to more than two characters, take the last one,
286 * and recompose the rest to get the first component. */
287 U16_PREV_UNSAFE (normalized, len, *b); /* Changes len in-place. */
288 UChar recomposed[18 * 2];
289 icu_err = U_ZERO_ERROR;
290 len = unorm2_normalize (unorm2_getNFCInstance (&icu_err), normalized, len, recomposed, ARRAY_LENGTH (recomposed), &icu_err);
291 if (U_FAILURE (icu_err))
292 return false;
293 /* We expect that recomposed has exactly one character now. */
294 if (unlikely (u_countChar32 (recomposed, len) != 1))
295 return false;
296 U16_GET_UNSAFE (recomposed, 0, *a);
297 ret = true;
298 }
299
300 return ret;
301 }
302
303
304 #if HB_USE_ATEXIT
305 static void free_static_icu_funcs ();
306 #endif
307
308 static struct hb_icu_unicode_funcs_lazy_loader_t : hb_unicode_funcs_lazy_loader_t<hb_icu_unicode_funcs_lazy_loader_t>
309 {
createhb_icu_unicode_funcs_lazy_loader_t310 static hb_unicode_funcs_t *create ()
311 {
312 void *user_data = nullptr;
313 #if U_ICU_VERSION_MAJOR_NUM >= 49
314 UErrorCode icu_err = U_ZERO_ERROR;
315 user_data = (void *) unorm2_getNFCInstance (&icu_err);
316 assert (user_data);
317 #endif
318
319 hb_unicode_funcs_t *funcs = hb_unicode_funcs_create (nullptr);
320
321 hb_unicode_funcs_set_combining_class_func (funcs, hb_icu_unicode_combining_class, nullptr, nullptr);
322 hb_unicode_funcs_set_general_category_func (funcs, hb_icu_unicode_general_category, nullptr, nullptr);
323 hb_unicode_funcs_set_mirroring_func (funcs, hb_icu_unicode_mirroring, nullptr, nullptr);
324 hb_unicode_funcs_set_script_func (funcs, hb_icu_unicode_script, nullptr, nullptr);
325 hb_unicode_funcs_set_compose_func (funcs, hb_icu_unicode_compose, user_data, nullptr);
326 hb_unicode_funcs_set_decompose_func (funcs, hb_icu_unicode_decompose, user_data, nullptr);
327
328 hb_unicode_funcs_make_immutable (funcs);
329
330 #if HB_USE_ATEXIT
331 atexit (free_static_icu_funcs);
332 #endif
333
334 return funcs;
335 }
336 } static_icu_funcs;
337
338 #if HB_USE_ATEXIT
339 static
free_static_icu_funcs()340 void free_static_icu_funcs ()
341 {
342 static_icu_funcs.free_instance ();
343 }
344 #endif
345
346 hb_unicode_funcs_t *
hb_icu_get_unicode_funcs()347 hb_icu_get_unicode_funcs ()
348 {
349 return static_icu_funcs.get_unconst ();
350 }
351