1 /*
2 * Copyright © 2011,2012 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 #include "hb.hh"
28
29 #ifndef HB_NO_OT_SHAPE
30
31 #include "hb-ot-shape-complex-khmer.hh"
32 #include "hb-ot-shape-complex-khmer-machine.hh"
33 #include "hb-ot-layout.hh"
34
35
36 /*
37 * Khmer shaper.
38 */
39
40 static const hb_ot_map_feature_t
41 khmer_features[] =
42 {
43 /*
44 * Basic features.
45 * These features are applied all at once, before reordering, constrained
46 * to the syllable.
47 */
48 {HB_TAG('p','r','e','f'), F_MANUAL_JOINERS},
49 {HB_TAG('b','l','w','f'), F_MANUAL_JOINERS},
50 {HB_TAG('a','b','v','f'), F_MANUAL_JOINERS},
51 {HB_TAG('p','s','t','f'), F_MANUAL_JOINERS},
52 {HB_TAG('c','f','a','r'), F_MANUAL_JOINERS},
53 /*
54 * Other features.
55 * These features are applied all at once after clearing syllables.
56 */
57 {HB_TAG('p','r','e','s'), F_GLOBAL_MANUAL_JOINERS},
58 {HB_TAG('a','b','v','s'), F_GLOBAL_MANUAL_JOINERS},
59 {HB_TAG('b','l','w','s'), F_GLOBAL_MANUAL_JOINERS},
60 {HB_TAG('p','s','t','s'), F_GLOBAL_MANUAL_JOINERS},
61 };
62
63 /*
64 * Must be in the same order as the khmer_features array.
65 */
66 enum {
67 KHMER_PREF,
68 KHMER_BLWF,
69 KHMER_ABVF,
70 KHMER_PSTF,
71 KHMER_CFAR,
72
73 _KHMER_PRES,
74 _KHMER_ABVS,
75 _KHMER_BLWS,
76 _KHMER_PSTS,
77
78 KHMER_NUM_FEATURES,
79 KHMER_BASIC_FEATURES = _KHMER_PRES, /* Don't forget to update this! */
80 };
81
82 static void
83 setup_syllables_khmer (const hb_ot_shape_plan_t *plan,
84 hb_font_t *font,
85 hb_buffer_t *buffer);
86 static void
87 reorder_khmer (const hb_ot_shape_plan_t *plan,
88 hb_font_t *font,
89 hb_buffer_t *buffer);
90
91 static void
collect_features_khmer(hb_ot_shape_planner_t * plan)92 collect_features_khmer (hb_ot_shape_planner_t *plan)
93 {
94 hb_ot_map_builder_t *map = &plan->map;
95
96 /* Do this before any lookups have been applied. */
97 map->add_gsub_pause (setup_syllables_khmer);
98 map->add_gsub_pause (reorder_khmer);
99
100 /* Testing suggests that Uniscribe does NOT pause between basic
101 * features. Test with KhmerUI.ttf and the following three
102 * sequences:
103 *
104 * U+1789,U+17BC
105 * U+1789,U+17D2,U+1789
106 * U+1789,U+17D2,U+1789,U+17BC
107 *
108 * https://github.com/harfbuzz/harfbuzz/issues/974
109 */
110 map->enable_feature (HB_TAG('l','o','c','l'));
111 map->enable_feature (HB_TAG('c','c','m','p'));
112
113 unsigned int i = 0;
114 for (; i < KHMER_BASIC_FEATURES; i++)
115 map->add_feature (khmer_features[i]);
116
117 map->add_gsub_pause (_hb_clear_syllables);
118
119 for (; i < KHMER_NUM_FEATURES; i++)
120 map->add_feature (khmer_features[i]);
121 }
122
123 static void
override_features_khmer(hb_ot_shape_planner_t * plan)124 override_features_khmer (hb_ot_shape_planner_t *plan)
125 {
126 hb_ot_map_builder_t *map = &plan->map;
127
128 /* Khmer spec has 'clig' as part of required shaping features:
129 * "Apply feature 'clig' to form ligatures that are desired for
130 * typographical correctness.", hence in overrides... */
131 map->enable_feature (HB_TAG('c','l','i','g'));
132
133 /* Uniscribe does not apply 'kern' in Khmer. */
134 if (hb_options ().uniscribe_bug_compatible)
135 {
136 map->disable_feature (HB_TAG('k','e','r','n'));
137 }
138
139 map->disable_feature (HB_TAG('l','i','g','a'));
140 }
141
142
143 struct khmer_shape_plan_t
144 {
145 hb_mask_t mask_array[KHMER_NUM_FEATURES];
146 };
147
148 static void *
data_create_khmer(const hb_ot_shape_plan_t * plan)149 data_create_khmer (const hb_ot_shape_plan_t *plan)
150 {
151 khmer_shape_plan_t *khmer_plan = (khmer_shape_plan_t *) hb_calloc (1, sizeof (khmer_shape_plan_t));
152 if (unlikely (!khmer_plan))
153 return nullptr;
154
155 for (unsigned int i = 0; i < ARRAY_LENGTH (khmer_plan->mask_array); i++)
156 khmer_plan->mask_array[i] = (khmer_features[i].flags & F_GLOBAL) ?
157 0 : plan->map.get_1_mask (khmer_features[i].tag);
158
159 return khmer_plan;
160 }
161
162 static void
data_destroy_khmer(void * data)163 data_destroy_khmer (void *data)
164 {
165 hb_free (data);
166 }
167
168 static void
setup_masks_khmer(const hb_ot_shape_plan_t * plan HB_UNUSED,hb_buffer_t * buffer,hb_font_t * font HB_UNUSED)169 setup_masks_khmer (const hb_ot_shape_plan_t *plan HB_UNUSED,
170 hb_buffer_t *buffer,
171 hb_font_t *font HB_UNUSED)
172 {
173 HB_BUFFER_ALLOCATE_VAR (buffer, khmer_category);
174
175 /* We cannot setup masks here. We save information about characters
176 * and setup masks later on in a pause-callback. */
177
178 unsigned int count = buffer->len;
179 hb_glyph_info_t *info = buffer->info;
180 for (unsigned int i = 0; i < count; i++)
181 set_khmer_properties (info[i]);
182 }
183
184 static void
setup_syllables_khmer(const hb_ot_shape_plan_t * plan HB_UNUSED,hb_font_t * font HB_UNUSED,hb_buffer_t * buffer)185 setup_syllables_khmer (const hb_ot_shape_plan_t *plan HB_UNUSED,
186 hb_font_t *font HB_UNUSED,
187 hb_buffer_t *buffer)
188 {
189 find_syllables_khmer (buffer);
190 foreach_syllable (buffer, start, end)
191 buffer->unsafe_to_break (start, end);
192 }
193
194
195 /* Rules from:
196 * https://docs.microsoft.com/en-us/typography/script-development/devanagari */
197
198 static void
reorder_consonant_syllable(const hb_ot_shape_plan_t * plan,hb_face_t * face HB_UNUSED,hb_buffer_t * buffer,unsigned int start,unsigned int end)199 reorder_consonant_syllable (const hb_ot_shape_plan_t *plan,
200 hb_face_t *face HB_UNUSED,
201 hb_buffer_t *buffer,
202 unsigned int start, unsigned int end)
203 {
204 const khmer_shape_plan_t *khmer_plan = (const khmer_shape_plan_t *) plan->data;
205 hb_glyph_info_t *info = buffer->info;
206
207 /* Setup masks. */
208 {
209 /* Post-base */
210 hb_mask_t mask = khmer_plan->mask_array[KHMER_BLWF] |
211 khmer_plan->mask_array[KHMER_ABVF] |
212 khmer_plan->mask_array[KHMER_PSTF];
213 for (unsigned int i = start + 1; i < end; i++)
214 info[i].mask |= mask;
215 }
216
217 unsigned int num_coengs = 0;
218 for (unsigned int i = start + 1; i < end; i++)
219 {
220 /* """
221 * When a COENG + (Cons | IndV) combination are found (and subscript count
222 * is less than two) the character combination is handled according to the
223 * subscript type of the character following the COENG.
224 *
225 * ...
226 *
227 * Subscript Type 2 - The COENG + RO characters are reordered to immediately
228 * before the base glyph. Then the COENG + RO characters are assigned to have
229 * the 'pref' OpenType feature applied to them.
230 * """
231 */
232 if (info[i].khmer_category() == OT_Coeng && num_coengs <= 2 && i + 1 < end)
233 {
234 num_coengs++;
235
236 if (info[i + 1].khmer_category() == OT_Ra)
237 {
238 for (unsigned int j = 0; j < 2; j++)
239 info[i + j].mask |= khmer_plan->mask_array[KHMER_PREF];
240
241 /* Move the Coeng,Ro sequence to the start. */
242 buffer->merge_clusters (start, i + 2);
243 hb_glyph_info_t t0 = info[i];
244 hb_glyph_info_t t1 = info[i + 1];
245 memmove (&info[start + 2], &info[start], (i - start) * sizeof (info[0]));
246 info[start] = t0;
247 info[start + 1] = t1;
248
249 /* Mark the subsequent stuff with 'cfar'. Used in Khmer.
250 * Read the feature spec.
251 * This allows distinguishing the following cases with MS Khmer fonts:
252 * U+1784,U+17D2,U+179A,U+17D2,U+1782
253 * U+1784,U+17D2,U+1782,U+17D2,U+179A
254 */
255 if (khmer_plan->mask_array[KHMER_CFAR])
256 for (unsigned int j = i + 2; j < end; j++)
257 info[j].mask |= khmer_plan->mask_array[KHMER_CFAR];
258
259 num_coengs = 2; /* Done. */
260 }
261 }
262
263 /* Reorder left matra piece. */
264 else if (info[i].khmer_category() == OT_VPre)
265 {
266 /* Move to the start. */
267 buffer->merge_clusters (start, i + 1);
268 hb_glyph_info_t t = info[i];
269 memmove (&info[start + 1], &info[start], (i - start) * sizeof (info[0]));
270 info[start] = t;
271 }
272 }
273 }
274
275 static void
reorder_syllable_khmer(const hb_ot_shape_plan_t * plan,hb_face_t * face,hb_buffer_t * buffer,unsigned int start,unsigned int end)276 reorder_syllable_khmer (const hb_ot_shape_plan_t *plan,
277 hb_face_t *face,
278 hb_buffer_t *buffer,
279 unsigned int start, unsigned int end)
280 {
281 khmer_syllable_type_t syllable_type = (khmer_syllable_type_t) (buffer->info[start].syllable() & 0x0F);
282 switch (syllable_type)
283 {
284 case khmer_broken_cluster: /* We already inserted dotted-circles, so just call the consonant_syllable. */
285 case khmer_consonant_syllable:
286 reorder_consonant_syllable (plan, face, buffer, start, end);
287 break;
288
289 case khmer_non_khmer_cluster:
290 break;
291 }
292 }
293
294 static void
reorder_khmer(const hb_ot_shape_plan_t * plan,hb_font_t * font,hb_buffer_t * buffer)295 reorder_khmer (const hb_ot_shape_plan_t *plan,
296 hb_font_t *font,
297 hb_buffer_t *buffer)
298 {
299 if (buffer->message (font, "start reordering khmer"))
300 {
301 hb_syllabic_insert_dotted_circles (font, buffer,
302 khmer_broken_cluster,
303 OT_DOTTEDCIRCLE,
304 OT_Repha);
305
306 foreach_syllable (buffer, start, end)
307 reorder_syllable_khmer (plan, font->face, buffer, start, end);
308 (void) buffer->message (font, "end reordering khmer");
309 }
310 HB_BUFFER_DEALLOCATE_VAR (buffer, khmer_category);
311 }
312
313
314 static bool
decompose_khmer(const hb_ot_shape_normalize_context_t * c,hb_codepoint_t ab,hb_codepoint_t * a,hb_codepoint_t * b)315 decompose_khmer (const hb_ot_shape_normalize_context_t *c,
316 hb_codepoint_t ab,
317 hb_codepoint_t *a,
318 hb_codepoint_t *b)
319 {
320 switch (ab)
321 {
322 /*
323 * Decompose split matras that don't have Unicode decompositions.
324 */
325
326 /* Khmer */
327 case 0x17BEu : *a = 0x17C1u; *b= 0x17BEu; return true;
328 case 0x17BFu : *a = 0x17C1u; *b= 0x17BFu; return true;
329 case 0x17C0u : *a = 0x17C1u; *b= 0x17C0u; return true;
330 case 0x17C4u : *a = 0x17C1u; *b= 0x17C4u; return true;
331 case 0x17C5u : *a = 0x17C1u; *b= 0x17C5u; return true;
332 }
333
334 return (bool) c->unicode->decompose (ab, a, b);
335 }
336
337 static bool
compose_khmer(const hb_ot_shape_normalize_context_t * c,hb_codepoint_t a,hb_codepoint_t b,hb_codepoint_t * ab)338 compose_khmer (const hb_ot_shape_normalize_context_t *c,
339 hb_codepoint_t a,
340 hb_codepoint_t b,
341 hb_codepoint_t *ab)
342 {
343 /* Avoid recomposing split matras. */
344 if (HB_UNICODE_GENERAL_CATEGORY_IS_MARK (c->unicode->general_category (a)))
345 return false;
346
347 return (bool) c->unicode->compose (a, b, ab);
348 }
349
350
351 const hb_ot_complex_shaper_t _hb_ot_complex_shaper_khmer =
352 {
353 collect_features_khmer,
354 override_features_khmer,
355 data_create_khmer,
356 data_destroy_khmer,
357 nullptr, /* preprocess_text */
358 nullptr, /* postprocess_glyphs */
359 HB_OT_SHAPE_NORMALIZATION_MODE_COMPOSED_DIACRITICS_NO_SHORT_CIRCUIT,
360 decompose_khmer,
361 compose_khmer,
362 setup_masks_khmer,
363 HB_TAG_NONE, /* gpos_tag */
364 nullptr, /* reorder_marks */
365 HB_OT_SHAPE_ZERO_WIDTH_MARKS_NONE,
366 false, /* fallback_position */
367 };
368
369
370 #endif
371