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 ICU library to provide Unicode data.
58  **/
59 
60 hb_script_t
hb_icu_script_to_script(UScriptCode script)61 hb_icu_script_to_script (UScriptCode script)
62 {
63   if (unlikely (script == USCRIPT_INVALID_CODE))
64     return HB_SCRIPT_INVALID;
65 
66   return hb_script_from_string (uscript_getShortName (script), -1);
67 }
68 
69 UScriptCode
hb_icu_script_from_script(hb_script_t script)70 hb_icu_script_from_script (hb_script_t script)
71 {
72   if (unlikely (script == HB_SCRIPT_INVALID))
73     return USCRIPT_INVALID_CODE;
74 
75   unsigned int numScriptCode = 1 + u_getIntPropertyMaxValue (UCHAR_SCRIPT);
76   for (unsigned int i = 0; i < numScriptCode; i++)
77     if (unlikely (hb_icu_script_to_script ((UScriptCode) i) == script))
78       return (UScriptCode) i;
79 
80   return USCRIPT_UNKNOWN;
81 }
82 
83 
84 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)85 hb_icu_unicode_combining_class (hb_unicode_funcs_t *ufuncs HB_UNUSED,
86 				hb_codepoint_t      unicode,
87 				void               *user_data HB_UNUSED)
88 
89 {
90   return (hb_unicode_combining_class_t) u_getCombiningClass (unicode);
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 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)167 hb_icu_unicode_compose (hb_unicode_funcs_t *ufuncs HB_UNUSED,
168 			hb_codepoint_t      a,
169 			hb_codepoint_t      b,
170 			hb_codepoint_t     *ab,
171 			void               *user_data HB_UNUSED)
172 {
173 #if U_ICU_VERSION_MAJOR_NUM >= 49
174   {
175     const UNormalizer2 *normalizer = (const UNormalizer2 *) user_data;
176     UChar32 ret = unorm2_composePair (normalizer, a, b);
177     if (ret < 0) return false;
178     *ab = ret;
179     return true;
180   }
181 #endif
182 
183   /* We don't ifdef-out the fallback code such that compiler always
184    * sees it and makes sure it's compilable. */
185 
186   UChar utf16[4], normalized[5];
187   unsigned int len;
188   hb_bool_t ret, err;
189   UErrorCode icu_err;
190 
191   len = 0;
192   err = false;
193   U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), a, err);
194   if (err) return false;
195   U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), b, err);
196   if (err) return false;
197 
198   icu_err = U_ZERO_ERROR;
199   len = unorm2_normalize (unorm2_getNFCInstance (&icu_err), utf16, len, normalized, ARRAY_LENGTH (normalized), &icu_err);
200   if (U_FAILURE (icu_err))
201     return false;
202   if (u_countChar32 (normalized, len) == 1) {
203     U16_GET_UNSAFE (normalized, 0, *ab);
204     ret = true;
205   } else {
206     ret = false;
207   }
208 
209   return ret;
210 }
211 
212 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)213 hb_icu_unicode_decompose (hb_unicode_funcs_t *ufuncs HB_UNUSED,
214 			  hb_codepoint_t      ab,
215 			  hb_codepoint_t     *a,
216 			  hb_codepoint_t     *b,
217 			  void               *user_data HB_UNUSED)
218 {
219 #if U_ICU_VERSION_MAJOR_NUM >= 49
220   {
221     const UNormalizer2 *normalizer = (const UNormalizer2 *) user_data;
222     UChar decomposed[4];
223     int len;
224     UErrorCode icu_err = U_ZERO_ERROR;
225     len = unorm2_getRawDecomposition (normalizer, ab, decomposed,
226 				      ARRAY_LENGTH (decomposed), &icu_err);
227     if (U_FAILURE (icu_err) || len < 0) return false;
228 
229     len = u_countChar32 (decomposed, len);
230     if (len == 1) {
231       U16_GET_UNSAFE (decomposed, 0, *a);
232       *b = 0;
233       return *a != ab;
234     } else if (len == 2) {
235       len = 0;
236       U16_NEXT_UNSAFE (decomposed, len, *a);
237       U16_NEXT_UNSAFE (decomposed, len, *b);
238     }
239     return true;
240   }
241 #endif
242 
243   /* We don't ifdef-out the fallback code such that compiler always
244    * sees it and makes sure it's compilable. */
245 
246   UChar utf16[2], normalized[2 * 19/*HB_UNICODE_MAX_DECOMPOSITION_LEN*/ + 1];
247   unsigned int len;
248   hb_bool_t ret, err;
249   UErrorCode icu_err;
250 
251   /* This function is a monster! Maybe it wasn't a good idea adding a
252    * pairwise decompose API... */
253   /* Watchout for the dragons.  Err, watchout for macros changing len. */
254 
255   len = 0;
256   err = false;
257   U16_APPEND (utf16, len, ARRAY_LENGTH (utf16), ab, err);
258   if (err) return false;
259 
260   icu_err = U_ZERO_ERROR;
261   len = unorm2_normalize (unorm2_getNFDInstance (&icu_err), utf16, len, normalized, ARRAY_LENGTH (normalized), &icu_err);
262   if (U_FAILURE (icu_err))
263     return false;
264 
265   len = u_countChar32 (normalized, len);
266 
267   if (len == 1) {
268     U16_GET_UNSAFE (normalized, 0, *a);
269     *b = 0;
270     ret = *a != ab;
271   } else if (len == 2) {
272     len = 0;
273     U16_NEXT_UNSAFE (normalized, len, *a);
274     U16_NEXT_UNSAFE (normalized, len, *b);
275 
276     /* Here's the ugly part: if ab decomposes to a single character and
277      * that character decomposes again, we have to detect that and undo
278      * the second part :-(. */
279     UChar recomposed[20];
280     icu_err = U_ZERO_ERROR;
281     unorm2_normalize (unorm2_getNFCInstance (&icu_err), normalized, len, recomposed, ARRAY_LENGTH (recomposed), &icu_err);
282     if (U_FAILURE (icu_err))
283       return false;
284     hb_codepoint_t c;
285     U16_GET_UNSAFE (recomposed, 0, c);
286     if (c != *a && c != ab) {
287       *a = c;
288       *b = 0;
289     }
290     ret = true;
291   } else {
292     /* If decomposed to more than two characters, take the last one,
293      * and recompose the rest to get the first component. */
294     U16_PREV_UNSAFE (normalized, len, *b); /* Changes len in-place. */
295     UChar recomposed[18 * 2];
296     icu_err = U_ZERO_ERROR;
297     len = unorm2_normalize (unorm2_getNFCInstance (&icu_err), normalized, len, recomposed, ARRAY_LENGTH (recomposed), &icu_err);
298     if (U_FAILURE (icu_err))
299       return false;
300     /* We expect that recomposed has exactly one character now. */
301     if (unlikely (u_countChar32 (recomposed, len) != 1))
302       return false;
303     U16_GET_UNSAFE (recomposed, 0, *a);
304     ret = true;
305   }
306 
307   return ret;
308 }
309 
310 
311 #if HB_USE_ATEXIT
312 static void free_static_icu_funcs ();
313 #endif
314 
315 static struct hb_icu_unicode_funcs_lazy_loader_t : hb_unicode_funcs_lazy_loader_t<hb_icu_unicode_funcs_lazy_loader_t>
316 {
createhb_icu_unicode_funcs_lazy_loader_t317   static hb_unicode_funcs_t *create ()
318   {
319     void *user_data = nullptr;
320 #if U_ICU_VERSION_MAJOR_NUM >= 49
321     UErrorCode icu_err = U_ZERO_ERROR;
322     user_data = (void *) unorm2_getNFCInstance (&icu_err);
323     assert (user_data);
324 #endif
325 
326     hb_unicode_funcs_t *funcs = hb_unicode_funcs_create (nullptr);
327 
328     hb_unicode_funcs_set_combining_class_func (funcs, hb_icu_unicode_combining_class, nullptr, nullptr);
329     hb_unicode_funcs_set_general_category_func (funcs, hb_icu_unicode_general_category, nullptr, nullptr);
330     hb_unicode_funcs_set_mirroring_func (funcs, hb_icu_unicode_mirroring, nullptr, nullptr);
331     hb_unicode_funcs_set_script_func (funcs, hb_icu_unicode_script, nullptr, nullptr);
332     hb_unicode_funcs_set_compose_func (funcs, hb_icu_unicode_compose, user_data, nullptr);
333     hb_unicode_funcs_set_decompose_func (funcs, hb_icu_unicode_decompose, user_data, nullptr);
334 
335     hb_unicode_funcs_make_immutable (funcs);
336 
337 #if HB_USE_ATEXIT
338     atexit (free_static_icu_funcs);
339 #endif
340 
341     return funcs;
342   }
343 } static_icu_funcs;
344 
345 #if HB_USE_ATEXIT
346 static
free_static_icu_funcs()347 void free_static_icu_funcs ()
348 {
349   static_icu_funcs.free_instance ();
350 }
351 #endif
352 
353 hb_unicode_funcs_t *
hb_icu_get_unicode_funcs()354 hb_icu_get_unicode_funcs ()
355 {
356   return static_icu_funcs.get_unconst ();
357 }
358 
359 #ifdef HB_ICU_EXTRA_SEMI_IGNORED
360 #pragma GCC diagnostic pop
361 #endif
362 
363 #endif
364