1 /*
2 * Copyright © 1998-2004 David Turner and Werner Lemberg
3 * Copyright © 2004,2007,2009,2010 Red Hat, Inc.
4 * Copyright © 2011,2012 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): Owen Taylor, Behdad Esfahbod
27 * Google Author(s): Behdad Esfahbod
28 */
29
30 #ifndef HB_BUFFER_HH
31 #define HB_BUFFER_HH
32
33 #include "hb.hh"
34 #include "hb-unicode.hh"
35
36
37 #ifndef HB_BUFFER_MAX_LEN_FACTOR
38 #define HB_BUFFER_MAX_LEN_FACTOR 32
39 #endif
40 #ifndef HB_BUFFER_MAX_LEN_MIN
41 #define HB_BUFFER_MAX_LEN_MIN 8192
42 #endif
43 #ifndef HB_BUFFER_MAX_LEN_DEFAULT
44 #define HB_BUFFER_MAX_LEN_DEFAULT 0x3FFFFFFF /* Shaping more than a billion chars? Let us know! */
45 #endif
46
47 #ifndef HB_BUFFER_MAX_OPS_FACTOR
48 #define HB_BUFFER_MAX_OPS_FACTOR 64
49 #endif
50 #ifndef HB_BUFFER_MAX_OPS_MIN
51 #define HB_BUFFER_MAX_OPS_MIN 1024
52 #endif
53 #ifndef HB_BUFFER_MAX_OPS_DEFAULT
54 #define HB_BUFFER_MAX_OPS_DEFAULT 0x1FFFFFFF /* Shaping more than a billion operations? Let us know! */
55 #endif
56
57 static_assert ((sizeof (hb_glyph_info_t) == 20), "");
58 static_assert ((sizeof (hb_glyph_info_t) == sizeof (hb_glyph_position_t)), "");
59
60 HB_MARK_AS_FLAG_T (hb_buffer_flags_t);
61 HB_MARK_AS_FLAG_T (hb_buffer_serialize_flags_t);
62 HB_MARK_AS_FLAG_T (hb_buffer_diff_flags_t);
63
64 enum hb_buffer_scratch_flags_t {
65 HB_BUFFER_SCRATCH_FLAG_DEFAULT = 0x00000000u,
66 HB_BUFFER_SCRATCH_FLAG_HAS_NON_ASCII = 0x00000001u,
67 HB_BUFFER_SCRATCH_FLAG_HAS_DEFAULT_IGNORABLES = 0x00000002u,
68 HB_BUFFER_SCRATCH_FLAG_HAS_SPACE_FALLBACK = 0x00000004u,
69 HB_BUFFER_SCRATCH_FLAG_HAS_GPOS_ATTACHMENT = 0x00000008u,
70 HB_BUFFER_SCRATCH_FLAG_HAS_UNSAFE_TO_BREAK = 0x00000010u,
71 HB_BUFFER_SCRATCH_FLAG_HAS_CGJ = 0x00000020u,
72
73 /* Reserved for complex shapers' internal use. */
74 HB_BUFFER_SCRATCH_FLAG_COMPLEX0 = 0x01000000u,
75 HB_BUFFER_SCRATCH_FLAG_COMPLEX1 = 0x02000000u,
76 HB_BUFFER_SCRATCH_FLAG_COMPLEX2 = 0x04000000u,
77 HB_BUFFER_SCRATCH_FLAG_COMPLEX3 = 0x08000000u,
78 };
79 HB_MARK_AS_FLAG_T (hb_buffer_scratch_flags_t);
80
81
82 /*
83 * hb_buffer_t
84 */
85
86 struct hb_buffer_t
87 {
88 hb_object_header_t header;
89
90 /* Information about how the text in the buffer should be treated */
91 hb_unicode_funcs_t *unicode; /* Unicode functions */
92 hb_buffer_flags_t flags; /* BOT / EOT / etc. */
93 hb_buffer_cluster_level_t cluster_level;
94 hb_codepoint_t replacement; /* U+FFFD or something else. */
95 hb_codepoint_t invisible; /* 0 or something else. */
96 hb_buffer_scratch_flags_t scratch_flags; /* Have space-fallback, etc. */
97 unsigned int max_len; /* Maximum allowed len. */
98 int max_ops; /* Maximum allowed operations. */
99
100 /* Buffer contents */
101 hb_buffer_content_type_t content_type;
102 hb_segment_properties_t props; /* Script, language, direction */
103
104 bool successful; /* Allocations successful */
105 bool have_output; /* Whether we have an output buffer going on */
106 bool have_positions; /* Whether we have positions */
107
108 unsigned int idx; /* Cursor into ->info and ->pos arrays */
109 unsigned int len; /* Length of ->info and ->pos arrays */
110 unsigned int out_len; /* Length of ->out array if have_output */
111
112 unsigned int allocated; /* Length of allocated arrays */
113 hb_glyph_info_t *info;
114 hb_glyph_info_t *out_info;
115 hb_glyph_position_t *pos;
116
117 unsigned int serial;
118
119 /* Text before / after the main buffer contents.
120 * Always in Unicode, and ordered outward.
121 * Index 0 is for "pre-context", 1 for "post-context". */
122 static constexpr unsigned CONTEXT_LENGTH = 5u;
123 hb_codepoint_t context[2][CONTEXT_LENGTH];
124 unsigned int context_len[2];
125
126 /* Debugging API */
127 hb_buffer_message_func_t message_func;
128 void *message_data;
129 hb_destroy_func_t message_destroy;
130
131 /* Internal debugging. */
132 /* The bits here reflect current allocations of the bytes in glyph_info_t's var1 and var2. */
133 #ifndef HB_NDEBUG
134 uint8_t allocated_var_bits;
135 #endif
136
137
138 /* Methods */
139
140 bool in_error () const { return !successful; }
141
142 void allocate_var (unsigned int start, unsigned int count)
143 {
144 #ifndef HB_NDEBUG
145 unsigned int end = start + count;
146 assert (end <= 8);
147 unsigned int bits = (1u<<end) - (1u<<start);
148 assert (0 == (allocated_var_bits & bits));
149 allocated_var_bits |= bits;
150 #endif
151 }
152 void deallocate_var (unsigned int start, unsigned int count)
153 {
154 #ifndef HB_NDEBUG
155 unsigned int end = start + count;
156 assert (end <= 8);
157 unsigned int bits = (1u<<end) - (1u<<start);
158 assert (bits == (allocated_var_bits & bits));
159 allocated_var_bits &= ~bits;
160 #endif
161 }
162 void assert_var (unsigned int start, unsigned int count)
163 {
164 #ifndef HB_NDEBUG
165 unsigned int end = start + count;
166 assert (end <= 8);
167 unsigned int bits = (1u<<end) - (1u<<start);
168 assert (bits == (allocated_var_bits & bits));
169 #endif
170 }
171 void deallocate_var_all ()
172 {
173 #ifndef HB_NDEBUG
174 allocated_var_bits = 0;
175 #endif
176 }
177
178 hb_glyph_info_t &cur (unsigned int i = 0) { return info[idx + i]; }
179 hb_glyph_info_t cur (unsigned int i = 0) const { return info[idx + i]; }
180
181 hb_glyph_position_t &cur_pos (unsigned int i = 0) { return pos[idx + i]; }
182 hb_glyph_position_t cur_pos (unsigned int i = 0) const { return pos[idx + i]; }
183
184 hb_glyph_info_t &prev () { return out_info[out_len ? out_len - 1 : 0]; }
185 hb_glyph_info_t prev () const { return out_info[out_len ? out_len - 1 : 0]; }
186
187 bool has_separate_output () const { return info != out_info; }
188
189
190 HB_INTERNAL void reset ();
191 HB_INTERNAL void clear ();
192
193 unsigned int backtrack_len () const { return have_output? out_len : idx; }
194 unsigned int lookahead_len () const { return len - idx; }
195 unsigned int next_serial () { return serial++; }
196
197 HB_INTERNAL void add (hb_codepoint_t codepoint,
198 unsigned int cluster);
199 HB_INTERNAL void add_info (const hb_glyph_info_t &glyph_info);
200
201 HB_INTERNAL void reverse_range (unsigned int start, unsigned int end);
202 HB_INTERNAL void reverse ();
203 HB_INTERNAL void reverse_clusters ();
204 HB_INTERNAL void guess_segment_properties ();
205
206 HB_INTERNAL void swap_buffers ();
207 HB_INTERNAL void remove_output ();
208 HB_INTERNAL void clear_output ();
209 HB_INTERNAL void clear_positions ();
210
211 HB_INTERNAL void replace_glyphs (unsigned int num_in,
212 unsigned int num_out,
213 const hb_codepoint_t *glyph_data);
214
215 void replace_glyph (hb_codepoint_t glyph_index)
216 {
217 if (unlikely (out_info != info || out_len != idx)) {
218 if (unlikely (!make_room_for (1, 1))) return;
219 out_info[out_len] = info[idx];
220 }
221 out_info[out_len].codepoint = glyph_index;
222
223 idx++;
224 out_len++;
225 }
226 /* Makes a copy of the glyph at idx to output and replace glyph_index */
227 hb_glyph_info_t & output_glyph (hb_codepoint_t glyph_index)
228 {
229 if (unlikely (!make_room_for (0, 1))) return Crap(hb_glyph_info_t);
230
231 if (unlikely (idx == len && !out_len))
232 return Crap(hb_glyph_info_t);
233
234 out_info[out_len] = idx < len ? info[idx] : out_info[out_len - 1];
235 out_info[out_len].codepoint = glyph_index;
236
237 out_len++;
238
239 return out_info[out_len - 1];
240 }
241 void output_info (const hb_glyph_info_t &glyph_info)
242 {
243 if (unlikely (!make_room_for (0, 1))) return;
244
245 out_info[out_len] = glyph_info;
246
247 out_len++;
248 }
249 /* Copies glyph at idx to output but doesn't advance idx */
250 void copy_glyph ()
251 {
252 if (unlikely (!make_room_for (0, 1))) return;
253
254 out_info[out_len] = info[idx];
255
256 out_len++;
257 }
258 /* Copies glyph at idx to output and advance idx.
259 * If there's no output, just advance idx. */
260 void
261 next_glyph ()
262 {
263 if (have_output)
264 {
265 if (out_info != info || out_len != idx)
266 {
267 if (unlikely (!make_room_for (1, 1))) return;
268 out_info[out_len] = info[idx];
269 }
270 out_len++;
271 }
272
273 idx++;
274 }
275 /* Copies n glyphs at idx to output and advance idx.
276 * If there's no output, just advance idx. */
277 void
278 next_glyphs (unsigned int n)
279 {
280 if (have_output)
281 {
282 if (out_info != info || out_len != idx)
283 {
284 if (unlikely (!make_room_for (n, n))) return;
285 memmove (out_info + out_len, info + idx, n * sizeof (out_info[0]));
286 }
287 out_len += n;
288 }
289
290 idx += n;
291 }
292 /* Advance idx without copying to output. */
293 void skip_glyph () { idx++; }
294 void reset_masks (hb_mask_t mask)
295 {
296 for (unsigned int j = 0; j < len; j++)
297 info[j].mask = mask;
298 }
299 void add_masks (hb_mask_t mask)
300 {
301 for (unsigned int j = 0; j < len; j++)
302 info[j].mask |= mask;
303 }
304 HB_INTERNAL void set_masks (hb_mask_t value, hb_mask_t mask,
305 unsigned int cluster_start, unsigned int cluster_end);
306
307 void merge_clusters (unsigned int start, unsigned int end)
308 {
309 if (end - start < 2)
310 return;
311 merge_clusters_impl (start, end);
312 }
313 HB_INTERNAL void merge_clusters_impl (unsigned int start, unsigned int end);
314 HB_INTERNAL void merge_out_clusters (unsigned int start, unsigned int end);
315 /* Merge clusters for deleting current glyph, and skip it. */
316 HB_INTERNAL void delete_glyph ();
317
318 void unsafe_to_break (unsigned int start,
319 unsigned int end)
320 {
321 if (end - start < 2)
322 return;
323 unsafe_to_break_impl (start, end);
324 }
325 HB_INTERNAL void unsafe_to_break_impl (unsigned int start, unsigned int end);
326 HB_INTERNAL void unsafe_to_break_from_outbuffer (unsigned int start, unsigned int end);
327
328
329 /* Internal methods */
330 HB_INTERNAL bool move_to (unsigned int i); /* i is output-buffer index. */
331
332 HB_INTERNAL bool enlarge (unsigned int size);
333
334 bool ensure (unsigned int size)
335 { return likely (!size || size < allocated) ? true : enlarge (size); }
336
337 bool ensure_inplace (unsigned int size)
338 { return likely (!size || size < allocated); }
339
340 HB_INTERNAL bool make_room_for (unsigned int num_in, unsigned int num_out);
341 HB_INTERNAL bool shift_forward (unsigned int count);
342
343 typedef long scratch_buffer_t;
344 HB_INTERNAL scratch_buffer_t *get_scratch_buffer (unsigned int *size);
345
346 void clear_context (unsigned int side) { context_len[side] = 0; }
347
348 HB_INTERNAL void sort (unsigned int start, unsigned int end, int(*compar)(const hb_glyph_info_t *, const hb_glyph_info_t *));
349
350 bool messaging () { return unlikely (message_func); }
351 bool message (hb_font_t *font, const char *fmt, ...) HB_PRINTF_FUNC(3, 4)
352 {
353 if (!messaging ())
354 return true;
355 va_list ap;
356 va_start (ap, fmt);
357 bool ret = message_impl (font, fmt, ap);
358 va_end (ap);
359 return ret;
360 }
361 HB_INTERNAL bool message_impl (hb_font_t *font, const char *fmt, va_list ap) HB_PRINTF_FUNC(3, 0);
362
363 static void
364 set_cluster (hb_glyph_info_t &inf, unsigned int cluster, unsigned int mask = 0)
365 {
366 if (inf.cluster != cluster)
367 {
368 if (mask & HB_GLYPH_FLAG_UNSAFE_TO_BREAK)
369 inf.mask |= HB_GLYPH_FLAG_UNSAFE_TO_BREAK;
370 else
371 inf.mask &= ~HB_GLYPH_FLAG_UNSAFE_TO_BREAK;
372 }
373 inf.cluster = cluster;
374 }
375
376 int
377 _unsafe_to_break_find_min_cluster (const hb_glyph_info_t *infos,
378 unsigned int start, unsigned int end,
379 unsigned int cluster) const
380 {
381 for (unsigned int i = start; i < end; i++)
382 cluster = MIN<unsigned int> (cluster, infos[i].cluster);
383 return cluster;
384 }
385 void
386 _unsafe_to_break_set_mask (hb_glyph_info_t *infos,
387 unsigned int start, unsigned int end,
388 unsigned int cluster)
389 {
390 for (unsigned int i = start; i < end; i++)
391 if (cluster != infos[i].cluster)
392 {
393 scratch_flags |= HB_BUFFER_SCRATCH_FLAG_HAS_UNSAFE_TO_BREAK;
394 infos[i].mask |= HB_GLYPH_FLAG_UNSAFE_TO_BREAK;
395 }
396 }
397
398 void unsafe_to_break_all ()
399 { unsafe_to_break_impl (0, len); }
400 void safe_to_break_all ()
401 {
402 for (unsigned int i = 0; i < len; i++)
403 info[i].mask &= ~HB_GLYPH_FLAG_UNSAFE_TO_BREAK;
404 }
405 };
406 DECLARE_NULL_INSTANCE (hb_buffer_t);
407
408
409 /* Loop over clusters. Duplicated in foreach_syllable(). */
410 #define foreach_cluster(buffer, start, end) \
411 for (unsigned int \
412 _count = buffer->len, \
413 start = 0, end = _count ? _next_cluster (buffer, 0) : 0; \
414 start < _count; \
415 start = end, end = _next_cluster (buffer, start))
416
417 static inline unsigned int
418 _next_cluster (hb_buffer_t *buffer, unsigned int start)
419 {
420 hb_glyph_info_t *info = buffer->info;
421 unsigned int count = buffer->len;
422
423 unsigned int cluster = info[start].cluster;
424 while (++start < count && cluster == info[start].cluster)
425 ;
426
427 return start;
428 }
429
430
431 #define HB_BUFFER_XALLOCATE_VAR(b, func, var) \
432 b->func (offsetof (hb_glyph_info_t, var) - offsetof(hb_glyph_info_t, var1), \
433 sizeof (b->info[0].var))
434 #define HB_BUFFER_ALLOCATE_VAR(b, var) HB_BUFFER_XALLOCATE_VAR (b, allocate_var, var ())
435 #define HB_BUFFER_DEALLOCATE_VAR(b, var) HB_BUFFER_XALLOCATE_VAR (b, deallocate_var, var ())
436 #define HB_BUFFER_ASSERT_VAR(b, var) HB_BUFFER_XALLOCATE_VAR (b, assert_var, var ())
437
438
439 #endif /* HB_BUFFER_HH */