1 | /* |
2 | * Copyright © 2012 Google, Inc. |
3 | * |
4 | * This is part of HarfBuzz, a text shaping library. |
5 | * |
6 | * Permission is hereby granted, without written agreement and without |
7 | * license or royalty fees, to use, copy, modify, and distribute this |
8 | * software and its documentation for any purpose, provided that the |
9 | * above copyright notice and the following two paragraphs appear in |
10 | * all copies of this software. |
11 | * |
12 | * IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE TO ANY PARTY FOR |
13 | * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES |
14 | * ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN |
15 | * IF THE COPYRIGHT HOLDER HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH |
16 | * DAMAGE. |
17 | * |
18 | * THE COPYRIGHT HOLDER SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING, |
19 | * BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND |
20 | * FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS |
21 | * ON AN "AS IS" BASIS, AND THE COPYRIGHT HOLDER HAS NO OBLIGATION TO |
22 | * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS. |
23 | * |
24 | * Google Author(s): Behdad Esfahbod |
25 | */ |
26 | |
27 | #ifndef HB_OT_SHAPE_COMPLEX_INDIC_HH |
28 | #define HB_OT_SHAPE_COMPLEX_INDIC_HH |
29 | |
30 | #include "hb.hh" |
31 | |
32 | |
33 | #include "hb-ot-shape-complex.hh" |
34 | #include "hb-ot-shape.hh" /* XXX Remove */ |
35 | |
36 | |
37 | /* buffer var allocations */ |
38 | #define indic_category() complex_var_u8_0() /* indic_category_t */ |
39 | #define indic_position() complex_var_u8_1() /* indic_position_t */ |
40 | |
41 | |
42 | #define INDIC_TABLE_ELEMENT_TYPE uint16_t |
43 | |
44 | /* Cateories used in the OpenType spec: |
45 | * https://docs.microsoft.com/en-us/typography/script-development/devanagari |
46 | */ |
47 | /* Note: This enum is duplicated in the -machine.rl source file. |
48 | * Not sure how to avoid duplication. */ |
49 | enum indic_category_t { |
50 | OT_X = 0, |
51 | OT_C = 1, |
52 | OT_V = 2, |
53 | OT_N = 3, |
54 | OT_H = 4, |
55 | OT_ZWNJ = 5, |
56 | OT_ZWJ = 6, |
57 | OT_M = 7, |
58 | OT_SM = 8, |
59 | /* OT_VD = 9, UNUSED; we use OT_A instead. */ |
60 | OT_A = 10, |
61 | OT_PLACEHOLDER = 11, |
62 | OT_DOTTEDCIRCLE = 12, |
63 | OT_RS = 13, /* Register Shifter, used in Khmer OT spec. */ |
64 | OT_Coeng = 14, /* Khmer-style Virama. */ |
65 | OT_Repha = 15, /* Atomically-encoded logical or visual repha. */ |
66 | OT_Ra = 16, |
67 | OT_CM = 17, /* Consonant-Medial. */ |
68 | OT_Symbol = 18, /* Avagraha, etc that take marks (SM,A,VD). */ |
69 | OT_CS = 19 |
70 | }; |
71 | |
72 | #define MEDIAL_FLAGS (FLAG (OT_CM)) |
73 | |
74 | /* Note: |
75 | * |
76 | * We treat Vowels and placeholders as if they were consonants. This is safe because Vowels |
77 | * cannot happen in a consonant syllable. The plus side however is, we can call the |
78 | * consonant syllable logic from the vowel syllable function and get it all right! */ |
79 | #define CONSONANT_FLAGS (FLAG (OT_C) | FLAG (OT_CS) | FLAG (OT_Ra) | MEDIAL_FLAGS | FLAG (OT_V) | FLAG (OT_PLACEHOLDER) | FLAG (OT_DOTTEDCIRCLE)) |
80 | #define JOINER_FLAGS (FLAG (OT_ZWJ) | FLAG (OT_ZWNJ)) |
81 | |
82 | |
83 | /* Visual positions in a syllable from left to right. */ |
84 | enum indic_position_t { |
85 | POS_START = 0, |
86 | |
87 | POS_RA_TO_BECOME_REPH = 1, |
88 | POS_PRE_M = 2, |
89 | POS_PRE_C = 3, |
90 | |
91 | POS_BASE_C = 4, |
92 | POS_AFTER_MAIN = 5, |
93 | |
94 | POS_ABOVE_C = 6, |
95 | |
96 | POS_BEFORE_SUB = 7, |
97 | POS_BELOW_C = 8, |
98 | POS_AFTER_SUB = 9, |
99 | |
100 | POS_BEFORE_POST = 10, |
101 | POS_POST_C = 11, |
102 | POS_AFTER_POST = 12, |
103 | |
104 | POS_FINAL_C = 13, |
105 | POS_SMVD = 14, |
106 | |
107 | POS_END = 15 |
108 | }; |
109 | |
110 | /* Categories used in IndicSyllabicCategory.txt from UCD. */ |
111 | enum indic_syllabic_category_t { |
112 | INDIC_SYLLABIC_CATEGORY_OTHER = OT_X, |
113 | |
114 | INDIC_SYLLABIC_CATEGORY_AVAGRAHA = OT_Symbol, |
115 | INDIC_SYLLABIC_CATEGORY_BINDU = OT_SM, |
116 | INDIC_SYLLABIC_CATEGORY_BRAHMI_JOINING_NUMBER = OT_PLACEHOLDER, /* Don't care. */ |
117 | INDIC_SYLLABIC_CATEGORY_CANTILLATION_MARK = OT_A, |
118 | INDIC_SYLLABIC_CATEGORY_CONSONANT = OT_C, |
119 | INDIC_SYLLABIC_CATEGORY_CONSONANT_DEAD = OT_C, |
120 | INDIC_SYLLABIC_CATEGORY_CONSONANT_FINAL = OT_CM, |
121 | INDIC_SYLLABIC_CATEGORY_CONSONANT_HEAD_LETTER = OT_C, |
122 | INDIC_SYLLABIC_CATEGORY_CONSONANT_KILLER = OT_M, /* U+17CD only. */ |
123 | INDIC_SYLLABIC_CATEGORY_CONSONANT_MEDIAL = OT_CM, |
124 | INDIC_SYLLABIC_CATEGORY_CONSONANT_PLACEHOLDER = OT_PLACEHOLDER, |
125 | INDIC_SYLLABIC_CATEGORY_CONSONANT_PRECEDING_REPHA = OT_Repha, |
126 | INDIC_SYLLABIC_CATEGORY_CONSONANT_PREFIXED = OT_X, /* Don't care. */ |
127 | INDIC_SYLLABIC_CATEGORY_CONSONANT_SUBJOINED = OT_CM, |
128 | INDIC_SYLLABIC_CATEGORY_CONSONANT_SUCCEEDING_REPHA = OT_N, |
129 | INDIC_SYLLABIC_CATEGORY_CONSONANT_WITH_STACKER = OT_CS, |
130 | INDIC_SYLLABIC_CATEGORY_GEMINATION_MARK = OT_SM, /* https://github.com/harfbuzz/harfbuzz/issues/552 */ |
131 | INDIC_SYLLABIC_CATEGORY_INVISIBLE_STACKER = OT_Coeng, |
132 | INDIC_SYLLABIC_CATEGORY_JOINER = OT_ZWJ, |
133 | INDIC_SYLLABIC_CATEGORY_MODIFYING_LETTER = OT_X, |
134 | INDIC_SYLLABIC_CATEGORY_NON_JOINER = OT_ZWNJ, |
135 | INDIC_SYLLABIC_CATEGORY_NUKTA = OT_N, |
136 | INDIC_SYLLABIC_CATEGORY_NUMBER = OT_PLACEHOLDER, |
137 | INDIC_SYLLABIC_CATEGORY_NUMBER_JOINER = OT_PLACEHOLDER, /* Don't care. */ |
138 | INDIC_SYLLABIC_CATEGORY_PURE_KILLER = OT_M, /* Is like a vowel matra. */ |
139 | INDIC_SYLLABIC_CATEGORY_REGISTER_SHIFTER = OT_RS, |
140 | INDIC_SYLLABIC_CATEGORY_SYLLABLE_MODIFIER = OT_SM, |
141 | INDIC_SYLLABIC_CATEGORY_TONE_LETTER = OT_X, |
142 | INDIC_SYLLABIC_CATEGORY_TONE_MARK = OT_N, |
143 | INDIC_SYLLABIC_CATEGORY_VIRAMA = OT_H, |
144 | INDIC_SYLLABIC_CATEGORY_VISARGA = OT_SM, |
145 | INDIC_SYLLABIC_CATEGORY_VOWEL = OT_V, |
146 | INDIC_SYLLABIC_CATEGORY_VOWEL_DEPENDENT = OT_M, |
147 | INDIC_SYLLABIC_CATEGORY_VOWEL_INDEPENDENT = OT_V |
148 | }; |
149 | |
150 | /* Categories used in IndicSMatraCategory.txt from UCD */ |
151 | enum indic_matra_category_t { |
152 | INDIC_MATRA_CATEGORY_NOT_APPLICABLE = POS_END, |
153 | |
154 | INDIC_MATRA_CATEGORY_LEFT = POS_PRE_C, |
155 | INDIC_MATRA_CATEGORY_TOP = POS_ABOVE_C, |
156 | INDIC_MATRA_CATEGORY_BOTTOM = POS_BELOW_C, |
157 | INDIC_MATRA_CATEGORY_RIGHT = POS_POST_C, |
158 | |
159 | /* These should resolve to the position of the last part of the split sequence. */ |
160 | INDIC_MATRA_CATEGORY_BOTTOM_AND_RIGHT = INDIC_MATRA_CATEGORY_RIGHT, |
161 | INDIC_MATRA_CATEGORY_LEFT_AND_RIGHT = INDIC_MATRA_CATEGORY_RIGHT, |
162 | INDIC_MATRA_CATEGORY_TOP_AND_BOTTOM = INDIC_MATRA_CATEGORY_BOTTOM, |
163 | INDIC_MATRA_CATEGORY_TOP_AND_BOTTOM_AND_RIGHT = INDIC_MATRA_CATEGORY_RIGHT, |
164 | INDIC_MATRA_CATEGORY_TOP_AND_LEFT = INDIC_MATRA_CATEGORY_TOP, |
165 | INDIC_MATRA_CATEGORY_TOP_AND_LEFT_AND_RIGHT = INDIC_MATRA_CATEGORY_RIGHT, |
166 | INDIC_MATRA_CATEGORY_TOP_AND_RIGHT = INDIC_MATRA_CATEGORY_RIGHT, |
167 | |
168 | INDIC_MATRA_CATEGORY_OVERSTRUCK = POS_AFTER_MAIN, |
169 | INDIC_MATRA_CATEGORY_VISUAL_ORDER_LEFT = POS_PRE_M |
170 | }; |
171 | |
172 | #define INDIC_COMBINE_CATEGORIES(S,M) \ |
173 | ( \ |
174 | ASSERT_STATIC_EXPR_ZERO (S < 255 && M < 255) + \ |
175 | ( S | \ |
176 | ( \ |
177 | ( \ |
178 | S == INDIC_SYLLABIC_CATEGORY_CONSONANT_MEDIAL || \ |
179 | S == INDIC_SYLLABIC_CATEGORY_GEMINATION_MARK || \ |
180 | S == INDIC_SYLLABIC_CATEGORY_REGISTER_SHIFTER || \ |
181 | S == INDIC_SYLLABIC_CATEGORY_CONSONANT_SUCCEEDING_REPHA || \ |
182 | S == INDIC_SYLLABIC_CATEGORY_VIRAMA || \ |
183 | S == INDIC_SYLLABIC_CATEGORY_VOWEL_DEPENDENT || \ |
184 | false \ |
185 | ? M : INDIC_MATRA_CATEGORY_NOT_APPLICABLE \ |
186 | ) << 8 \ |
187 | ) \ |
188 | ) \ |
189 | ) |
190 | |
191 | HB_INTERNAL INDIC_TABLE_ELEMENT_TYPE |
192 | hb_indic_get_categories (hb_codepoint_t u); |
193 | |
194 | |
195 | static inline bool |
196 | is_one_of (const hb_glyph_info_t &info, unsigned int flags) |
197 | { |
198 | /* If it ligated, all bets are off. */ |
199 | if (_hb_glyph_info_ligated (&info)) return false; |
200 | return !!(FLAG_UNSAFE (info.indic_category()) & flags); |
201 | } |
202 | |
203 | static inline bool |
204 | is_joiner (const hb_glyph_info_t &info) |
205 | { |
206 | return is_one_of (info, JOINER_FLAGS); |
207 | } |
208 | |
209 | static inline bool |
210 | is_consonant (const hb_glyph_info_t &info) |
211 | { |
212 | return is_one_of (info, CONSONANT_FLAGS); |
213 | } |
214 | |
215 | static inline bool |
216 | is_halant (const hb_glyph_info_t &info) |
217 | { |
218 | return is_one_of (info, FLAG (OT_H)); |
219 | } |
220 | |
221 | #define IN_HALF_BLOCK(u, Base) (((u) & ~0x7Fu) == (Base)) |
222 | |
223 | #define IS_DEVA(u) (IN_HALF_BLOCK (u, 0x0900u)) |
224 | #define IS_BENG(u) (IN_HALF_BLOCK (u, 0x0980u)) |
225 | #define IS_GURU(u) (IN_HALF_BLOCK (u, 0x0A00u)) |
226 | #define IS_GUJR(u) (IN_HALF_BLOCK (u, 0x0A80u)) |
227 | #define IS_ORYA(u) (IN_HALF_BLOCK (u, 0x0B00u)) |
228 | #define IS_TAML(u) (IN_HALF_BLOCK (u, 0x0B80u)) |
229 | #define IS_TELU(u) (IN_HALF_BLOCK (u, 0x0C00u)) |
230 | #define IS_KNDA(u) (IN_HALF_BLOCK (u, 0x0C80u)) |
231 | #define IS_MLYM(u) (IN_HALF_BLOCK (u, 0x0D00u)) |
232 | #define IS_SINH(u) (IN_HALF_BLOCK (u, 0x0D80u)) |
233 | |
234 | |
235 | #define MATRA_POS_LEFT(u) POS_PRE_M |
236 | #define MATRA_POS_RIGHT(u) ( \ |
237 | IS_DEVA(u) ? POS_AFTER_SUB : \ |
238 | IS_BENG(u) ? POS_AFTER_POST : \ |
239 | IS_GURU(u) ? POS_AFTER_POST : \ |
240 | IS_GUJR(u) ? POS_AFTER_POST : \ |
241 | IS_ORYA(u) ? POS_AFTER_POST : \ |
242 | IS_TAML(u) ? POS_AFTER_POST : \ |
243 | IS_TELU(u) ? (u <= 0x0C42u ? POS_BEFORE_SUB : POS_AFTER_SUB) : \ |
244 | IS_KNDA(u) ? (u < 0x0CC3u || u > 0xCD6u ? POS_BEFORE_SUB : POS_AFTER_SUB) : \ |
245 | IS_MLYM(u) ? POS_AFTER_POST : \ |
246 | IS_SINH(u) ? POS_AFTER_SUB : \ |
247 | /*default*/ POS_AFTER_SUB \ |
248 | ) |
249 | #define MATRA_POS_TOP(u) ( /* BENG and MLYM don't have top matras. */ \ |
250 | IS_DEVA(u) ? POS_AFTER_SUB : \ |
251 | IS_GURU(u) ? POS_AFTER_POST : /* Deviate from spec */ \ |
252 | IS_GUJR(u) ? POS_AFTER_SUB : \ |
253 | IS_ORYA(u) ? POS_AFTER_MAIN : \ |
254 | IS_TAML(u) ? POS_AFTER_SUB : \ |
255 | IS_TELU(u) ? POS_BEFORE_SUB : \ |
256 | IS_KNDA(u) ? POS_BEFORE_SUB : \ |
257 | IS_SINH(u) ? POS_AFTER_SUB : \ |
258 | /*default*/ POS_AFTER_SUB \ |
259 | ) |
260 | #define MATRA_POS_BOTTOM(u) ( \ |
261 | IS_DEVA(u) ? POS_AFTER_SUB : \ |
262 | IS_BENG(u) ? POS_AFTER_SUB : \ |
263 | IS_GURU(u) ? POS_AFTER_POST : \ |
264 | IS_GUJR(u) ? POS_AFTER_POST : \ |
265 | IS_ORYA(u) ? POS_AFTER_SUB : \ |
266 | IS_TAML(u) ? POS_AFTER_POST : \ |
267 | IS_TELU(u) ? POS_BEFORE_SUB : \ |
268 | IS_KNDA(u) ? POS_BEFORE_SUB : \ |
269 | IS_MLYM(u) ? POS_AFTER_POST : \ |
270 | IS_SINH(u) ? POS_AFTER_SUB : \ |
271 | /*default*/ POS_AFTER_SUB \ |
272 | ) |
273 | |
274 | static inline indic_position_t |
275 | matra_position_indic (hb_codepoint_t u, indic_position_t side) |
276 | { |
277 | switch ((int) side) |
278 | { |
279 | case POS_PRE_C: return MATRA_POS_LEFT (u); |
280 | case POS_POST_C: return MATRA_POS_RIGHT (u); |
281 | case POS_ABOVE_C: return MATRA_POS_TOP (u); |
282 | case POS_BELOW_C: return MATRA_POS_BOTTOM (u); |
283 | }; |
284 | return side; |
285 | } |
286 | |
287 | /* XXX |
288 | * This is a hack for now. We should move this data into the main Indic table. |
289 | * Or completely remove it and just check in the tables. |
290 | */ |
291 | static const hb_codepoint_t ra_chars[] = { |
292 | 0x0930u, /* Devanagari */ |
293 | 0x09B0u, /* Bengali */ |
294 | 0x09F0u, /* Bengali */ |
295 | 0x0A30u, /* Gurmukhi */ /* No Reph */ |
296 | 0x0AB0u, /* Gujarati */ |
297 | 0x0B30u, /* Oriya */ |
298 | 0x0BB0u, /* Tamil */ /* No Reph */ |
299 | 0x0C30u, /* Telugu */ /* Reph formed only with ZWJ */ |
300 | 0x0CB0u, /* Kannada */ |
301 | 0x0D30u, /* Malayalam */ /* No Reph, Logical Repha */ |
302 | |
303 | 0x0DBBu, /* Sinhala */ /* Reph formed only with ZWJ */ |
304 | |
305 | 0x179Au, /* Khmer */ |
306 | }; |
307 | |
308 | static inline bool |
309 | is_ra (hb_codepoint_t u) |
310 | { |
311 | for (unsigned int i = 0; i < ARRAY_LENGTH (ra_chars); i++) |
312 | if (u == ra_chars[i]) |
313 | return true; |
314 | return false; |
315 | } |
316 | |
317 | static inline void |
318 | set_indic_properties (hb_glyph_info_t &info) |
319 | { |
320 | hb_codepoint_t u = info.codepoint; |
321 | unsigned int type = hb_indic_get_categories (u); |
322 | indic_category_t cat = (indic_category_t) (type & 0x7Fu); |
323 | indic_position_t pos = (indic_position_t) (type >> 8); |
324 | |
325 | |
326 | /* |
327 | * Re-assign category |
328 | */ |
329 | |
330 | /* The following act more like the Bindus. */ |
331 | if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x0953u, 0x0954u))) |
332 | cat = OT_SM; |
333 | /* The following act like consonants. */ |
334 | else if (unlikely (hb_in_ranges<hb_codepoint_t> (u, 0x0A72u, 0x0A73u, |
335 | 0x1CF5u, 0x1CF6u))) |
336 | cat = OT_C; |
337 | /* TODO: The following should only be allowed after a Visarga. |
338 | * For now, just treat them like regular tone marks. */ |
339 | else if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x1CE2u, 0x1CE8u))) |
340 | cat = OT_A; |
341 | /* TODO: The following should only be allowed after some of |
342 | * the nasalization marks, maybe only for U+1CE9..U+1CF1. |
343 | * For now, just treat them like tone marks. */ |
344 | else if (unlikely (u == 0x1CEDu)) |
345 | cat = OT_A; |
346 | /* The following take marks in standalone clusters, similar to Avagraha. */ |
347 | else if (unlikely (hb_in_ranges<hb_codepoint_t> (u, 0xA8F2u, 0xA8F7u, |
348 | 0x1CE9u, 0x1CECu, |
349 | 0x1CEEu, 0x1CF1u))) |
350 | { |
351 | cat = OT_Symbol; |
352 | static_assert (((int) INDIC_SYLLABIC_CATEGORY_AVAGRAHA == OT_Symbol), "" ); |
353 | } |
354 | else if (unlikely (u == 0x0A51u)) |
355 | { |
356 | /* https://github.com/harfbuzz/harfbuzz/issues/524 */ |
357 | cat = OT_M; |
358 | pos = POS_BELOW_C; |
359 | } |
360 | |
361 | /* According to ScriptExtensions.txt, these Grantha marks may also be used in Tamil, |
362 | * so the Indic shaper needs to know their categories. */ |
363 | else if (unlikely (u == 0x11301u || u == 0x11303u)) cat = OT_SM; |
364 | else if (unlikely (u == 0x1133cu)) cat = OT_N; |
365 | |
366 | else if (unlikely (u == 0x0AFBu)) cat = OT_N; /* https://github.com/harfbuzz/harfbuzz/issues/552 */ |
367 | |
368 | else if (unlikely (u == 0x0980u)) cat = OT_PLACEHOLDER; /* https://github.com/harfbuzz/harfbuzz/issues/538 */ |
369 | else if (unlikely (u == 0x0C80u)) cat = OT_PLACEHOLDER; /* https://github.com/harfbuzz/harfbuzz/pull/623 */ |
370 | else if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x2010u, 0x2011u))) |
371 | cat = OT_PLACEHOLDER; |
372 | else if (unlikely (u == 0x25CCu)) cat = OT_DOTTEDCIRCLE; |
373 | |
374 | |
375 | /* |
376 | * Re-assign position. |
377 | */ |
378 | |
379 | if ((FLAG_UNSAFE (cat) & CONSONANT_FLAGS)) |
380 | { |
381 | pos = POS_BASE_C; |
382 | if (is_ra (u)) |
383 | cat = OT_Ra; |
384 | } |
385 | else if (cat == OT_M) |
386 | { |
387 | pos = matra_position_indic (u, pos); |
388 | } |
389 | else if ((FLAG_UNSAFE (cat) & (FLAG (OT_SM) /* | FLAG (OT_VD) */ | FLAG (OT_A) | FLAG (OT_Symbol)))) |
390 | { |
391 | pos = POS_SMVD; |
392 | } |
393 | |
394 | if (unlikely (u == 0x0B01u)) pos = POS_BEFORE_SUB; /* Oriya Bindu is BeforeSub in the spec. */ |
395 | |
396 | |
397 | |
398 | info.indic_category() = cat; |
399 | info.indic_position() = pos; |
400 | } |
401 | |
402 | |
403 | #endif /* HB_OT_SHAPE_COMPLEX_INDIC_HH */ |
404 | |