Line data Source code
1 : /*
2 : * Copyright © 2010,2012 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-ot-shape-complex-private.hh"
28 :
29 :
30 : /* Thai / Lao shaper */
31 :
32 :
33 : /* PUA shaping */
34 :
35 :
36 : enum thai_consonant_type_t
37 : {
38 : NC,
39 : AC,
40 : RC,
41 : DC,
42 : NOT_CONSONANT,
43 : NUM_CONSONANT_TYPES = NOT_CONSONANT
44 : };
45 :
46 : static thai_consonant_type_t
47 0 : get_consonant_type (hb_codepoint_t u)
48 : {
49 0 : if (u == 0x0E1Bu || u == 0x0E1Du || u == 0x0E1Fu/* || u == 0x0E2Cu*/)
50 0 : return AC;
51 0 : if (u == 0x0E0Du || u == 0x0E10u)
52 0 : return RC;
53 0 : if (u == 0x0E0Eu || u == 0x0E0Fu)
54 0 : return DC;
55 0 : if (hb_in_range (u, 0x0E01u, 0x0E2Eu))
56 0 : return NC;
57 0 : return NOT_CONSONANT;
58 : }
59 :
60 :
61 : enum thai_mark_type_t
62 : {
63 : AV,
64 : BV,
65 : T,
66 : NOT_MARK,
67 : NUM_MARK_TYPES = NOT_MARK
68 : };
69 :
70 : static thai_mark_type_t
71 0 : get_mark_type (hb_codepoint_t u)
72 : {
73 0 : if (u == 0x0E31u || hb_in_range (u, 0x0E34u, 0x0E37u) ||
74 0 : u == 0x0E47u || hb_in_range (u, 0x0E4Du, 0x0E4Eu))
75 0 : return AV;
76 0 : if (hb_in_range (u, 0x0E38u, 0x0E3Au))
77 0 : return BV;
78 0 : if (hb_in_range (u, 0x0E48u, 0x0E4Cu))
79 0 : return T;
80 0 : return NOT_MARK;
81 : }
82 :
83 :
84 : enum thai_action_t
85 : {
86 : NOP,
87 : SD, /* Shift combining-mark down */
88 : SL, /* Shift combining-mark left */
89 : SDL, /* Shift combining-mark down-left */
90 : RD /* Remove descender from base */
91 : };
92 :
93 : static hb_codepoint_t
94 0 : thai_pua_shape (hb_codepoint_t u, thai_action_t action, hb_font_t *font)
95 : {
96 : struct thai_pua_mapping_t {
97 : hb_codepoint_t u;
98 : hb_codepoint_t win_pua;
99 : hb_codepoint_t mac_pua;
100 0 : } const *pua_mappings = NULL;
101 : static const thai_pua_mapping_t SD_mappings[] = {
102 : {0x0E48u, 0xF70Au, 0xF88Bu}, /* MAI EK */
103 : {0x0E49u, 0xF70Bu, 0xF88Eu}, /* MAI THO */
104 : {0x0E4Au, 0xF70Cu, 0xF891u}, /* MAI TRI */
105 : {0x0E4Bu, 0xF70Du, 0xF894u}, /* MAI CHATTAWA */
106 : {0x0E4Cu, 0xF70Eu, 0xF897u}, /* THANTHAKHAT */
107 : {0x0E38u, 0xF718u, 0xF89Bu}, /* SARA U */
108 : {0x0E39u, 0xF719u, 0xF89Cu}, /* SARA UU */
109 : {0x0E3Au, 0xF71Au, 0xF89Du}, /* PHINTHU */
110 : {0x0000u, 0x0000u, 0x0000u}
111 : };
112 : static const thai_pua_mapping_t SDL_mappings[] = {
113 : {0x0E48u, 0xF705u, 0xF88Cu}, /* MAI EK */
114 : {0x0E49u, 0xF706u, 0xF88Fu}, /* MAI THO */
115 : {0x0E4Au, 0xF707u, 0xF892u}, /* MAI TRI */
116 : {0x0E4Bu, 0xF708u, 0xF895u}, /* MAI CHATTAWA */
117 : {0x0E4Cu, 0xF709u, 0xF898u}, /* THANTHAKHAT */
118 : {0x0000u, 0x0000u, 0x0000u}
119 : };
120 : static const thai_pua_mapping_t SL_mappings[] = {
121 : {0x0E48u, 0xF713u, 0xF88Au}, /* MAI EK */
122 : {0x0E49u, 0xF714u, 0xF88Du}, /* MAI THO */
123 : {0x0E4Au, 0xF715u, 0xF890u}, /* MAI TRI */
124 : {0x0E4Bu, 0xF716u, 0xF893u}, /* MAI CHATTAWA */
125 : {0x0E4Cu, 0xF717u, 0xF896u}, /* THANTHAKHAT */
126 : {0x0E31u, 0xF710u, 0xF884u}, /* MAI HAN-AKAT */
127 : {0x0E34u, 0xF701u, 0xF885u}, /* SARA I */
128 : {0x0E35u, 0xF702u, 0xF886u}, /* SARA II */
129 : {0x0E36u, 0xF703u, 0xF887u}, /* SARA UE */
130 : {0x0E37u, 0xF704u, 0xF888u}, /* SARA UEE */
131 : {0x0E47u, 0xF712u, 0xF889u}, /* MAITAIKHU */
132 : {0x0E4Du, 0xF711u, 0xF899u}, /* NIKHAHIT */
133 : {0x0000u, 0x0000u, 0x0000u}
134 : };
135 : static const thai_pua_mapping_t RD_mappings[] = {
136 : {0x0E0Du, 0xF70Fu, 0xF89Au}, /* YO YING */
137 : {0x0E10u, 0xF700u, 0xF89Eu}, /* THO THAN */
138 : {0x0000u, 0x0000u, 0x0000u}
139 : };
140 :
141 0 : switch (action) {
142 0 : case NOP: return u;
143 0 : case SD: pua_mappings = SD_mappings; break;
144 0 : case SDL: pua_mappings = SDL_mappings; break;
145 0 : case SL: pua_mappings = SL_mappings; break;
146 0 : case RD: pua_mappings = RD_mappings; break;
147 : }
148 0 : for (; pua_mappings->u; pua_mappings++)
149 0 : if (pua_mappings->u == u)
150 : {
151 : hb_codepoint_t glyph;
152 0 : if (hb_font_get_glyph (font, pua_mappings->win_pua, 0, &glyph))
153 0 : return pua_mappings->win_pua;
154 0 : if (hb_font_get_glyph (font, pua_mappings->mac_pua, 0, &glyph))
155 0 : return pua_mappings->mac_pua;
156 0 : break;
157 : }
158 0 : return u;
159 : }
160 :
161 :
162 : static enum thai_above_state_t
163 : { /* Cluster above looks like: */
164 : T0, /* ⣤ */
165 : T1, /* ⣼ */
166 : T2, /* ⣾ */
167 : T3, /* ⣿ */
168 : NUM_ABOVE_STATES
169 : } thai_above_start_state[NUM_CONSONANT_TYPES + 1/* For NOT_CONSONANT */] =
170 : {
171 : T0, /* NC */
172 : T1, /* AC */
173 : T0, /* RC */
174 : T0, /* DC */
175 : T3, /* NOT_CONSONANT */
176 : };
177 :
178 : static const struct thai_above_state_machine_edge_t {
179 : thai_action_t action;
180 : thai_above_state_t next_state;
181 : } thai_above_state_machine[NUM_ABOVE_STATES][NUM_MARK_TYPES] =
182 : { /*AV*/ /*BV*/ /*T*/
183 : /*T0*/ {{NOP,T3}, {NOP,T0}, {SD, T3}},
184 : /*T1*/ {{SL, T2}, {NOP,T1}, {SDL,T2}},
185 : /*T2*/ {{NOP,T3}, {NOP,T2}, {SL, T3}},
186 : /*T3*/ {{NOP,T3}, {NOP,T3}, {NOP,T3}},
187 : };
188 :
189 :
190 : static enum thai_below_state_t
191 : {
192 : B0, /* No descender */
193 : B1, /* Removable descender */
194 : B2, /* Strict descender */
195 : NUM_BELOW_STATES
196 : } thai_below_start_state[NUM_CONSONANT_TYPES + 1/* For NOT_CONSONANT */] =
197 : {
198 : B0, /* NC */
199 : B0, /* AC */
200 : B1, /* RC */
201 : B2, /* DC */
202 : B2, /* NOT_CONSONANT */
203 : };
204 :
205 : static const struct thai_below_state_machine_edge_t {
206 : thai_action_t action;
207 : thai_below_state_t next_state;
208 : } thai_below_state_machine[NUM_BELOW_STATES][NUM_MARK_TYPES] =
209 : { /*AV*/ /*BV*/ /*T*/
210 : /*B0*/ {{NOP,B0}, {NOP,B2}, {NOP, B0}},
211 : /*B1*/ {{NOP,B1}, {RD, B2}, {NOP, B1}},
212 : /*B2*/ {{NOP,B2}, {SD, B2}, {NOP, B2}},
213 : };
214 :
215 :
216 : static void
217 0 : do_thai_pua_shaping (const hb_ot_shape_plan_t *plan HB_UNUSED,
218 : hb_buffer_t *buffer,
219 : hb_font_t *font)
220 : {
221 0 : thai_above_state_t above_state = thai_above_start_state[NOT_CONSONANT];
222 0 : thai_below_state_t below_state = thai_below_start_state[NOT_CONSONANT];
223 0 : unsigned int base = 0;
224 :
225 0 : hb_glyph_info_t *info = buffer->info;
226 0 : unsigned int count = buffer->len;
227 0 : for (unsigned int i = 0; i < count; i++)
228 : {
229 0 : thai_mark_type_t mt = get_mark_type (info[i].codepoint);
230 :
231 0 : if (mt == NOT_MARK) {
232 0 : thai_consonant_type_t ct = get_consonant_type (info[i].codepoint);
233 0 : above_state = thai_above_start_state[ct];
234 0 : below_state = thai_below_start_state[ct];
235 0 : base = i;
236 0 : continue;
237 : }
238 :
239 0 : const thai_above_state_machine_edge_t &above_edge = thai_above_state_machine[above_state][mt];
240 0 : const thai_below_state_machine_edge_t &below_edge = thai_below_state_machine[below_state][mt];
241 0 : above_state = above_edge.next_state;
242 0 : below_state = below_edge.next_state;
243 :
244 : /* At least one of the above/below actions is NOP. */
245 0 : thai_action_t action = above_edge.action != NOP ? above_edge.action : below_edge.action;
246 :
247 0 : if (action == RD)
248 0 : info[base].codepoint = thai_pua_shape (info[base].codepoint, action, font);
249 : else
250 0 : info[i].codepoint = thai_pua_shape (info[i].codepoint, action, font);
251 : }
252 0 : }
253 :
254 :
255 : static void
256 0 : preprocess_text_thai (const hb_ot_shape_plan_t *plan,
257 : hb_buffer_t *buffer,
258 : hb_font_t *font)
259 : {
260 : /* This function implements the shaping logic documented here:
261 : *
262 : * http://linux.thai.net/~thep/th-otf/shaping.html
263 : *
264 : * The first shaping rule listed there is needed even if the font has Thai
265 : * OpenType tables. The rest do fallback positioning based on PUA codepoints.
266 : * We implement that only if there exist no Thai GSUB in the font.
267 : */
268 :
269 : /* The following is NOT specified in the MS OT Thai spec, however, it seems
270 : * to be what Uniscribe and other engines implement. According to Eric Muller:
271 : *
272 : * When you have a SARA AM, decompose it in NIKHAHIT + SARA AA, *and* move the
273 : * NIKHAHIT backwards over any tone mark (0E48-0E4B).
274 : *
275 : * <0E14, 0E4B, 0E33> -> <0E14, 0E4D, 0E4B, 0E32>
276 : *
277 : * This reordering is legit only when the NIKHAHIT comes from a SARA AM, not
278 : * when it's there to start with. The string <0E14, 0E4B, 0E4D> is probably
279 : * not what a user wanted, but the rendering is nevertheless nikhahit above
280 : * chattawa.
281 : *
282 : * Same for Lao.
283 : *
284 : * Note:
285 : *
286 : * Uniscribe also does some below-marks reordering. Namely, it positions U+0E3A
287 : * after U+0E38 and U+0E39. We do that by modifying the ccc for U+0E3A.
288 : * See unicode->modified_combining_class (). Lao does NOT have a U+0E3A
289 : * equivalent.
290 : */
291 :
292 :
293 : /*
294 : * Here are the characters of significance:
295 : *
296 : * Thai Lao
297 : * SARA AM: U+0E33 U+0EB3
298 : * SARA AA: U+0E32 U+0EB2
299 : * Nikhahit: U+0E4D U+0ECD
300 : *
301 : * Testing shows that Uniscribe reorder the following marks:
302 : * Thai: <0E31,0E34..0E37,0E47..0E4E>
303 : * Lao: <0EB1,0EB4..0EB7,0EC7..0ECE>
304 : *
305 : * Note how the Lao versions are the same as Thai + 0x80.
306 : */
307 :
308 : /* We only get one script at a time, so a script-agnostic implementation
309 : * is adequate here. */
310 : #define IS_SARA_AM(x) (((x) & ~0x0080u) == 0x0E33u)
311 : #define NIKHAHIT_FROM_SARA_AM(x) ((x) - 0x0E33u + 0x0E4Du)
312 : #define SARA_AA_FROM_SARA_AM(x) ((x) - 1)
313 : #define IS_TONE_MARK(x) (hb_in_ranges ((x) & ~0x0080u, 0x0E34u, 0x0E37u, 0x0E47u, 0x0E4Eu, 0x0E31u, 0x0E31u))
314 :
315 0 : buffer->clear_output ();
316 0 : unsigned int count = buffer->len;
317 0 : for (buffer->idx = 0; buffer->idx < count && !buffer->in_error;)
318 : {
319 0 : hb_codepoint_t u = buffer->cur().codepoint;
320 0 : if (likely (!IS_SARA_AM (u))) {
321 0 : buffer->next_glyph ();
322 0 : continue;
323 : }
324 :
325 : /* Is SARA AM. Decompose and reorder. */
326 0 : hb_codepoint_t decomposed[2] = {hb_codepoint_t (NIKHAHIT_FROM_SARA_AM (u)),
327 0 : hb_codepoint_t (SARA_AA_FROM_SARA_AM (u))};
328 0 : buffer->replace_glyphs (1, 2, decomposed);
329 0 : if (unlikely (buffer->in_error))
330 0 : return;
331 :
332 : /* Make Nikhahit be recognized as a ccc=0 mark when zeroing widths. */
333 0 : unsigned int end = buffer->out_len;
334 0 : _hb_glyph_info_set_general_category (&buffer->out_info[end - 2], HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK);
335 :
336 : /* Ok, let's see... */
337 0 : unsigned int start = end - 2;
338 0 : while (start > 0 && IS_TONE_MARK (buffer->out_info[start - 1].codepoint))
339 0 : start--;
340 :
341 0 : if (start + 2 < end)
342 : {
343 : /* Move Nikhahit (end-2) to the beginning */
344 0 : buffer->merge_out_clusters (start, end);
345 0 : hb_glyph_info_t t = buffer->out_info[end - 2];
346 0 : memmove (buffer->out_info + start + 1,
347 0 : buffer->out_info + start,
348 0 : sizeof (buffer->out_info[0]) * (end - start - 2));
349 0 : buffer->out_info[start] = t;
350 : }
351 : else
352 : {
353 : /* Since we decomposed, and NIKHAHIT is combining, merge clusters with the
354 : * previous cluster. */
355 0 : if (start && buffer->cluster_level == HB_BUFFER_CLUSTER_LEVEL_MONOTONE_GRAPHEMES)
356 0 : buffer->merge_out_clusters (start - 1, end);
357 : }
358 : }
359 0 : buffer->swap_buffers ();
360 :
361 : /* If font has Thai GSUB, we are done. */
362 0 : if (plan->props.script == HB_SCRIPT_THAI && !plan->map.found_script[0])
363 0 : do_thai_pua_shaping (plan, buffer, font);
364 : }
365 :
366 : const hb_ot_complex_shaper_t _hb_ot_complex_shaper_thai =
367 : {
368 : "thai",
369 : NULL, /* collect_features */
370 : NULL, /* override_features */
371 : NULL, /* data_create */
372 : NULL, /* data_destroy */
373 : preprocess_text_thai,
374 : NULL, /* postprocess_glyphs */
375 : HB_OT_SHAPE_NORMALIZATION_MODE_DEFAULT,
376 : NULL, /* decompose */
377 : NULL, /* compose */
378 : NULL, /* setup_masks */
379 : NULL, /* disable_otl */
380 : HB_OT_SHAPE_ZERO_WIDTH_MARKS_BY_GDEF_LATE,
381 : false,/* fallback_position */
382 : };
|
