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#ifdef HAVE_ICU
33
34#include "hb-icu.h"
35
36#include "hb-machinery.hh"
37
38#include <unicode/uchar.h>
39#include <unicode/unorm2.h>
40#include <unicode/ustring.h>
41#include <unicode/utf16.h>
42#include <unicode/uversion.h>
43
44/* ICU extra semicolon, fixed since 65, https://github.com/unicode-org/icu/commit/480bec3 */
45#if U_ICU_VERSION_MAJOR_NUM < 65 && (defined(__GNUC__) || defined(__clang__))
46#define HB_ICU_EXTRA_SEMI_IGNORED
47#pragma GCC diagnostic push
48#pragma GCC diagnostic ignored "-Wextra-semi-stmt"
49#endif
50
51/**
52 * SECTION:hb-icu
53 * @title: hb-icu
54 * @short_description: ICU integration
55 * @include: hb-icu.h
56 *
57 * Functions for using HarfBuzz with the International Components for Unicode
58 * (ICU) library. HarfBuzz supports using ICU to provide Unicode data, by attaching
59 * ICU functions to the virtual methods in a #hb_unicode_funcs_t function
60 * structure.
61 **/
62
63/**
64 * hb_icu_script_to_script:
65 * @script: The UScriptCode identifier to query
66 *
67 * Fetches the #hb_script_t script that corresponds to the
68 * specified UScriptCode identifier.
69 *
70 * Return value: the #hb_script_t script found
71 *
72 **/
73
74hb_script_t
75hb_icu_script_to_script (UScriptCode script)
76{
77 if (unlikely (script == USCRIPT_INVALID_CODE))
78 return HB_SCRIPT_INVALID;
79
80 return hb_script_from_string (uscript_getShortName (script), -1);
81}
82
83/**
84 * hb_icu_script_from_script:
85 * @script: The #hb_script_t script to query
86 *
87 * Fetches the UScriptCode identifier that corresponds to the
88 * specified #hb_script_t script.
89 *
90 * Return value: the UScriptCode identifier found
91 *
92 **/
93UScriptCode
94hb_icu_script_from_script (hb_script_t script)
95{
96 if (unlikely (script == HB_SCRIPT_INVALID))
97 return USCRIPT_INVALID_CODE;
98
99 unsigned int numScriptCode = 1 + u_getIntPropertyMaxValue (UCHAR_SCRIPT);
100 for (unsigned int i = 0; i < numScriptCode; i++)
101 if (unlikely (hb_icu_script_to_script ((UScriptCode) i) == script))
102 return (UScriptCode) i;
103
104 return USCRIPT_UNKNOWN;
105}
106
107
108static hb_unicode_combining_class_t
109hb_icu_unicode_combining_class (hb_unicode_funcs_t *ufuncs HB_UNUSED,
110 hb_codepoint_t unicode,
111 void *user_data HB_UNUSED)
112
113{
114 return (hb_unicode_combining_class_t) u_getCombiningClass (unicode);
115}
116
117static hb_unicode_general_category_t
118hb_icu_unicode_general_category (hb_unicode_funcs_t *ufuncs HB_UNUSED,
119 hb_codepoint_t unicode,
120 void *user_data HB_UNUSED)
121{
122 switch (u_getIntPropertyValue(unicode, UCHAR_GENERAL_CATEGORY))
123 {
124 case U_UNASSIGNED: return HB_UNICODE_GENERAL_CATEGORY_UNASSIGNED;
125
126 case U_UPPERCASE_LETTER: return HB_UNICODE_GENERAL_CATEGORY_UPPERCASE_LETTER;
127 case U_LOWERCASE_LETTER: return HB_UNICODE_GENERAL_CATEGORY_LOWERCASE_LETTER;
128 case U_TITLECASE_LETTER: return HB_UNICODE_GENERAL_CATEGORY_TITLECASE_LETTER;
129 case U_MODIFIER_LETTER: return HB_UNICODE_GENERAL_CATEGORY_MODIFIER_LETTER;
130 case U_OTHER_LETTER: return HB_UNICODE_GENERAL_CATEGORY_OTHER_LETTER;
131
132 case U_NON_SPACING_MARK: return HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK;
133 case U_ENCLOSING_MARK: return HB_UNICODE_GENERAL_CATEGORY_ENCLOSING_MARK;
134 case U_COMBINING_SPACING_MARK: return HB_UNICODE_GENERAL_CATEGORY_SPACING_MARK;
135
136 case U_DECIMAL_DIGIT_NUMBER: return HB_UNICODE_GENERAL_CATEGORY_DECIMAL_NUMBER;
137 case U_LETTER_NUMBER: return HB_UNICODE_GENERAL_CATEGORY_LETTER_NUMBER;
138 case U_OTHER_NUMBER: return HB_UNICODE_GENERAL_CATEGORY_OTHER_NUMBER;
139
140 case U_SPACE_SEPARATOR: return HB_UNICODE_GENERAL_CATEGORY_SPACE_SEPARATOR;
141 case U_LINE_SEPARATOR: return HB_UNICODE_GENERAL_CATEGORY_LINE_SEPARATOR;
142 case U_PARAGRAPH_SEPARATOR: return HB_UNICODE_GENERAL_CATEGORY_PARAGRAPH_SEPARATOR;
143
144 case U_CONTROL_CHAR: return HB_UNICODE_GENERAL_CATEGORY_CONTROL;
145 case U_FORMAT_CHAR: return HB_UNICODE_GENERAL_CATEGORY_FORMAT;
146 case U_PRIVATE_USE_CHAR: return HB_UNICODE_GENERAL_CATEGORY_PRIVATE_USE;
147 case U_SURROGATE: return HB_UNICODE_GENERAL_CATEGORY_SURROGATE;
148
149
150 case U_DASH_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_DASH_PUNCTUATION;
151 case U_START_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_OPEN_PUNCTUATION;
152 case U_END_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_CLOSE_PUNCTUATION;
153 case U_CONNECTOR_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_CONNECT_PUNCTUATION;
154 case U_OTHER_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_OTHER_PUNCTUATION;
155
156 case U_MATH_SYMBOL: return HB_UNICODE_GENERAL_CATEGORY_MATH_SYMBOL;
157 case U_CURRENCY_SYMBOL: return HB_UNICODE_GENERAL_CATEGORY_CURRENCY_SYMBOL;
158 case U_MODIFIER_SYMBOL: return HB_UNICODE_GENERAL_CATEGORY_MODIFIER_SYMBOL;
159 case U_OTHER_SYMBOL: return HB_UNICODE_GENERAL_CATEGORY_OTHER_SYMBOL;
160
161 case U_INITIAL_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_INITIAL_PUNCTUATION;
162 case U_FINAL_PUNCTUATION: return HB_UNICODE_GENERAL_CATEGORY_FINAL_PUNCTUATION;
163 }
164
165 return HB_UNICODE_GENERAL_CATEGORY_UNASSIGNED;
166}
167
168static hb_codepoint_t
169hb_icu_unicode_mirroring (hb_unicode_funcs_t *ufuncs HB_UNUSED,
170 hb_codepoint_t unicode,
171 void *user_data HB_UNUSED)
172{
173 return u_charMirror(unicode);
174}
175
176static hb_script_t
177hb_icu_unicode_script (hb_unicode_funcs_t *ufuncs HB_UNUSED,
178 hb_codepoint_t unicode,
179 void *user_data HB_UNUSED)
180{
181 UErrorCode status = U_ZERO_ERROR;
182 UScriptCode scriptCode = uscript_getScript(unicode, &status);
183
184 if (unlikely (U_FAILURE (status)))
185 return HB_SCRIPT_UNKNOWN;
186
187 return hb_icu_script_to_script (scriptCode);
188}
189
190static hb_bool_t
191hb_icu_unicode_compose (hb_unicode_funcs_t *ufuncs HB_UNUSED,
192 hb_codepoint_t a,
193 hb_codepoint_t b,
194 hb_codepoint_t *ab,
195 void *user_data HB_UNUSED)
196{
197#if U_ICU_VERSION_MAJOR_NUM >= 49
198 {
199 const UNormalizer2 *normalizer = (const UNormalizer2 *) user_data;
200 UChar32 ret = unorm2_composePair (normalizer, a, b);
201 if (ret < 0) return false;
202 *ab = ret;
203 return true;
204 }
205#endif
206
207 /* We don't ifdef-out the fallback code such that compiler always
208 * sees it and makes sure it's compilable. */
209
210 UChar utf16[4], normalized[5];
211 unsigned int len;
212 hb_bool_t ret, err;
213 UErrorCode icu_err;
214
215 len = 0;
216 err = false;
217 U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), a, err);
218 if (err) return false;
219 U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), b, err);
220 if (err) return false;
221
222 icu_err = U_ZERO_ERROR;
223 len = unorm2_normalize (unorm2_getNFCInstance (&icu_err), utf16, len, normalized, ARRAY_LENGTH (normalized), &icu_err);
224 if (U_FAILURE (icu_err))
225 return false;
226 if (u_countChar32 (normalized, len) == 1) {
227 U16_GET_UNSAFE (normalized, 0, *ab);
228 ret = true;
229 } else {
230 ret = false;
231 }
232
233 return ret;
234}
235
236static hb_bool_t
237hb_icu_unicode_decompose (hb_unicode_funcs_t *ufuncs HB_UNUSED,
238 hb_codepoint_t ab,
239 hb_codepoint_t *a,
240 hb_codepoint_t *b,
241 void *user_data HB_UNUSED)
242{
243#if U_ICU_VERSION_MAJOR_NUM >= 49
244 {
245 const UNormalizer2 *normalizer = (const UNormalizer2 *) user_data;
246 UChar decomposed[4];
247 int len;
248 UErrorCode icu_err = U_ZERO_ERROR;
249 len = unorm2_getRawDecomposition (normalizer, ab, decomposed,
250 ARRAY_LENGTH (decomposed), &icu_err);
251 if (U_FAILURE (icu_err) || len < 0) return false;
252
253 len = u_countChar32 (decomposed, len);
254 if (len == 1) {
255 U16_GET_UNSAFE (decomposed, 0, *a);
256 *b = 0;
257 return *a != ab;
258 } else if (len == 2) {
259 len = 0;
260 U16_NEXT_UNSAFE (decomposed, len, *a);
261 U16_NEXT_UNSAFE (decomposed, len, *b);
262 }
263 return true;
264 }
265#endif
266
267 /* We don't ifdef-out the fallback code such that compiler always
268 * sees it and makes sure it's compilable. */
269
270 UChar utf16[2], normalized[2 * 19/*HB_UNICODE_MAX_DECOMPOSITION_LEN*/ + 1];
271 unsigned int len;
272 hb_bool_t ret, err;
273 UErrorCode icu_err;
274
275 /* This function is a monster! Maybe it wasn't a good idea adding a
276 * pairwise decompose API... */
277 /* Watchout for the dragons. Err, watchout for macros changing len. */
278
279 len = 0;
280 err = false;
281 U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), ab, err);
282 if (err) return false;
283
284 icu_err = U_ZERO_ERROR;
285 len = unorm2_normalize (unorm2_getNFDInstance (&icu_err), utf16, len, normalized, ARRAY_LENGTH (normalized), &icu_err);
286 if (U_FAILURE (icu_err))
287 return false;
288
289 len = u_countChar32 (normalized, len);
290
291 if (len == 1) {
292 U16_GET_UNSAFE (normalized, 0, *a);
293 *b = 0;
294 ret = *a != ab;
295 } else if (len == 2) {
296 len = 0;
297 U16_NEXT_UNSAFE (normalized, len, *a);
298 U16_NEXT_UNSAFE (normalized, len, *b);
299
300 /* Here's the ugly part: if ab decomposes to a single character and
301 * that character decomposes again, we have to detect that and undo
302 * the second part :-(. */
303 UChar recomposed[20];
304 icu_err = U_ZERO_ERROR;
305 unorm2_normalize (unorm2_getNFCInstance (&icu_err), normalized, len, recomposed, ARRAY_LENGTH (recomposed), &icu_err);
306 if (U_FAILURE (icu_err))
307 return false;
308 hb_codepoint_t c;
309 U16_GET_UNSAFE (recomposed, 0, c);
310 if (c != *a && c != ab) {
311 *a = c;
312 *b = 0;
313 }
314 ret = true;
315 } else {
316 /* If decomposed to more than two characters, take the last one,
317 * and recompose the rest to get the first component. */
318 U16_PREV_UNSAFE (normalized, len, *b); /* Changes len in-place. */
319 UChar recomposed[18 * 2];
320 icu_err = U_ZERO_ERROR;
321 len = unorm2_normalize (unorm2_getNFCInstance (&icu_err), normalized, len, recomposed, ARRAY_LENGTH (recomposed), &icu_err);
322 if (U_FAILURE (icu_err))
323 return false;
324 /* We expect that recomposed has exactly one character now. */
325 if (unlikely (u_countChar32 (recomposed, len) != 1))
326 return false;
327 U16_GET_UNSAFE (recomposed, 0, *a);
328 ret = true;
329 }
330
331 return ret;
332}
333
334
335#if HB_USE_ATEXIT
336static void free_static_icu_funcs ();
337#endif
338
339static struct hb_icu_unicode_funcs_lazy_loader_t : hb_unicode_funcs_lazy_loader_t<hb_icu_unicode_funcs_lazy_loader_t>
340{
341 static hb_unicode_funcs_t *create ()
342 {
343 void *user_data = nullptr;
344#if U_ICU_VERSION_MAJOR_NUM >= 49
345 UErrorCode icu_err = U_ZERO_ERROR;
346 user_data = (void *) unorm2_getNFCInstance (&icu_err);
347 assert (user_data);
348#endif
349
350 hb_unicode_funcs_t *funcs = hb_unicode_funcs_create (nullptr);
351
352 hb_unicode_funcs_set_combining_class_func (funcs, hb_icu_unicode_combining_class, nullptr, nullptr);
353 hb_unicode_funcs_set_general_category_func (funcs, hb_icu_unicode_general_category, nullptr, nullptr);
354 hb_unicode_funcs_set_mirroring_func (funcs, hb_icu_unicode_mirroring, nullptr, nullptr);
355 hb_unicode_funcs_set_script_func (funcs, hb_icu_unicode_script, nullptr, nullptr);
356 hb_unicode_funcs_set_compose_func (funcs, hb_icu_unicode_compose, user_data, nullptr);
357 hb_unicode_funcs_set_decompose_func (funcs, hb_icu_unicode_decompose, user_data, nullptr);
358
359 hb_unicode_funcs_make_immutable (funcs);
360
361#if HB_USE_ATEXIT
362 atexit (free_static_icu_funcs);
363#endif
364
365 return funcs;
366 }
367} static_icu_funcs;
368
369#if HB_USE_ATEXIT
370static
371void free_static_icu_funcs ()
372{
373 static_icu_funcs.free_instance ();
374}
375#endif
376
377/**
378 * hb_icu_get_unicode_funcs:
379 *
380 * Fetches a Unicode-functions structure that is populated
381 * with the appropriate ICU function for each method.
382 *
383 * Return value: (transfer none): a pointer to the #hb_unicode_funcs_t Unicode-functions structure
384 *
385 * Since: 0.9.38
386 **/
387hb_unicode_funcs_t *
388hb_icu_get_unicode_funcs ()
389{
390 return static_icu_funcs.get_unconst ();
391}
392
393#ifdef HB_ICU_EXTRA_SEMI_IGNORED
394#pragma GCC diagnostic pop
395#endif
396
397#endif
398