2 * Copyright (C) 1998-2004 David Turner and Werner Lemberg
3 * Copyright (C) 2004,2007,2009,2010 Red Hat, Inc.
5 * This is part of HarfBuzz, a text shaping library.
7 * Permission is hereby granted, without written agreement and without
8 * license or royalty fees, to use, copy, modify, and distribute this
9 * software and its documentation for any purpose, provided that the
10 * above copyright notice and the following two paragraphs appear in
11 * all copies of this software.
13 * IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE TO ANY PARTY FOR
14 * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES
15 * ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN
16 * IF THE COPYRIGHT HOLDER HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
19 * THE COPYRIGHT HOLDER SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING,
20 * BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
21 * FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS
22 * ON AN "AS IS" BASIS, AND THE COPYRIGHT HOLDER HAS NO OBLIGATION TO
23 * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
25 * Red Hat Author(s): Owen Taylor, Behdad Esfahbod
28 #include "hb-buffer-private.hh"
35 static hb_buffer_t _hb_buffer_nil = {
36 HB_REFERENCE_COUNT_INVALID, /* ref_count */
38 &_hb_unicode_funcs_nil /* unicode */
41 /* Here is how the buffer works internally:
43 * There are two info pointers: info and out_info. They always have
44 * the same allocated size, but different lengths.
46 * As an optimization, both info and out_info may point to the
47 * same piece of memory, which is owned by info. This remains the
48 * case as long as out_len doesn't exceed len at any time.
49 * In that case, swap() is no-op and the glyph operations operate
52 * As soon as out_info gets longer than info, out_info is moved over
53 * to an alternate buffer (which we reuse the pos buffer for!), and its
54 * current contents (out_len entries) are copied to the new place.
55 * This should all remain transparent to the user. swap() then
56 * switches info and out_info.
61 _hb_buffer_enlarge (hb_buffer_t *buffer, unsigned int size)
63 if (unlikely (buffer->in_error))
66 unsigned int new_allocated = buffer->allocated;
67 hb_internal_glyph_position_t *new_pos;
68 hb_internal_glyph_info_t *new_info;
71 separate_out = buffer->out_info != buffer->info;
73 while (size > new_allocated)
74 new_allocated += (new_allocated >> 1) + 8;
76 new_pos = (hb_internal_glyph_position_t *) realloc (buffer->pos, new_allocated * sizeof (buffer->pos[0]));
77 new_info = (hb_internal_glyph_info_t *) realloc (buffer->info, new_allocated * sizeof (buffer->info[0]));
79 if (unlikely (!new_pos || !new_info))
80 buffer->in_error = TRUE;
83 buffer->pos = new_pos;
85 if (likely (new_info))
86 buffer->info = new_info;
88 buffer->out_info = separate_out ? (hb_internal_glyph_info_t *) buffer->pos : buffer->info;
89 if (likely (!buffer->in_error))
90 buffer->allocated = new_allocated;
92 return likely (!buffer->in_error);
95 static inline hb_bool_t
96 _hb_buffer_ensure (hb_buffer_t *buffer, unsigned int size)
98 return likely (size <= buffer->allocated) ? TRUE : _hb_buffer_enlarge (buffer, size);
101 static inline hb_bool_t
102 _hb_buffer_ensure_separate (hb_buffer_t *buffer, unsigned int size)
104 if (unlikely (!_hb_buffer_ensure (buffer, size))) return FALSE;
106 if (buffer->out_info == buffer->info)
108 assert (buffer->have_output);
110 buffer->out_info = (hb_internal_glyph_info_t *) buffer->pos;
111 memcpy (buffer->out_info, buffer->info, buffer->out_len * sizeof (buffer->out_info[0]));
121 hb_buffer_create (unsigned int pre_alloc_size)
125 if (!HB_OBJECT_DO_CREATE (hb_buffer_t, buffer))
126 return &_hb_buffer_nil;
129 _hb_buffer_ensure (buffer, pre_alloc_size);
131 buffer->unicode = &_hb_unicode_funcs_nil;
137 hb_buffer_reference (hb_buffer_t *buffer)
139 HB_OBJECT_DO_REFERENCE (buffer);
143 hb_buffer_get_reference_count (hb_buffer_t *buffer)
145 HB_OBJECT_DO_GET_REFERENCE_COUNT (buffer);
149 hb_buffer_destroy (hb_buffer_t *buffer)
151 HB_OBJECT_DO_DESTROY (buffer);
153 hb_unicode_funcs_destroy (buffer->unicode);
163 hb_buffer_set_unicode_funcs (hb_buffer_t *buffer,
164 hb_unicode_funcs_t *unicode)
167 unicode = &_hb_unicode_funcs_nil;
169 hb_unicode_funcs_reference (unicode);
170 hb_unicode_funcs_destroy (buffer->unicode);
171 buffer->unicode = unicode;
175 hb_buffer_get_unicode_funcs (hb_buffer_t *buffer)
177 return buffer->unicode;
181 hb_buffer_set_direction (hb_buffer_t *buffer,
182 hb_direction_t direction)
185 buffer->props.direction = direction;
189 hb_buffer_get_direction (hb_buffer_t *buffer)
191 return buffer->props.direction;
195 hb_buffer_set_script (hb_buffer_t *buffer,
198 buffer->props.script = script;
202 hb_buffer_get_script (hb_buffer_t *buffer)
204 return buffer->props.script;
208 hb_buffer_set_language (hb_buffer_t *buffer,
209 hb_language_t language)
211 buffer->props.language = language;
215 hb_buffer_get_language (hb_buffer_t *buffer)
217 return buffer->props.language;
222 hb_buffer_clear (hb_buffer_t *buffer)
224 buffer->have_output = FALSE;
225 buffer->have_positions = FALSE;
226 buffer->in_error = FALSE;
230 buffer->out_info = buffer->info;
231 buffer->max_lig_id = 0;
235 hb_buffer_ensure (hb_buffer_t *buffer, unsigned int size)
237 return _hb_buffer_ensure (buffer, size);
241 hb_buffer_add_glyph (hb_buffer_t *buffer,
242 hb_codepoint_t codepoint,
244 unsigned int cluster)
246 hb_internal_glyph_info_t *glyph;
248 if (unlikely (!_hb_buffer_ensure (buffer, buffer->len + 1))) return;
250 glyph = &buffer->info[buffer->len];
251 glyph->codepoint = codepoint;
253 glyph->cluster = cluster;
254 glyph->component = 0;
256 glyph->gproperty = HB_BUFFER_GLYPH_PROPERTIES_UNKNOWN;
262 hb_buffer_clear_positions (hb_buffer_t *buffer)
264 _hb_buffer_clear_output (buffer);
265 buffer->have_output = FALSE;
266 buffer->have_positions = TRUE;
268 if (unlikely (!buffer->pos))
270 buffer->pos = (hb_internal_glyph_position_t *) calloc (buffer->allocated, sizeof (buffer->pos[0]));
274 memset (buffer->pos, 0, sizeof (buffer->pos[0]) * buffer->len);
277 /* HarfBuzz-Internal API */
280 _hb_buffer_clear_output (hb_buffer_t *buffer)
282 buffer->have_output = TRUE;
283 buffer->have_positions = FALSE;
285 buffer->out_info = buffer->info;
289 _hb_buffer_swap (hb_buffer_t *buffer)
293 assert (buffer->have_output);
295 if (unlikely (buffer->in_error)) return;
297 if (buffer->out_info != buffer->info)
299 hb_internal_glyph_info_t *tmp_string;
300 tmp_string = buffer->info;
301 buffer->info = buffer->out_info;
302 buffer->out_info = tmp_string;
303 buffer->pos = (hb_internal_glyph_position_t *) buffer->out_info;
307 buffer->len = buffer->out_len;
308 buffer->out_len = tmp;
313 /* The following function copies `num_out' elements from `glyph_data'
314 to `buffer->out_info', advancing the in array pointer in the structure
315 by `num_in' elements, and the out array pointer by `num_out' elements.
316 Finally, it sets the `length' field of `out' equal to
317 `pos' of the `out' structure.
319 If `component' is 0xFFFF, the component value from buffer->i
320 will copied `num_out' times, otherwise `component' itself will
321 be used to fill the `component' fields.
323 If `lig_id' is 0xFFFF, the lig_id value from buffer->i
324 will copied `num_out' times, otherwise `lig_id' itself will
325 be used to fill the `lig_id' fields.
327 The mask for all replacement glyphs are taken
328 from the glyph at position `buffer->i'.
330 The cluster value for the glyph at position buffer->i is used
331 for all replacement glyphs */
334 _hb_buffer_add_output_glyphs (hb_buffer_t *buffer,
336 unsigned int num_out,
337 const hb_codepoint_t *glyph_data,
338 unsigned short component,
339 unsigned short lig_id)
343 unsigned int cluster;
345 if (buffer->out_info != buffer->info ||
346 buffer->out_len + num_out > buffer->i + num_in)
348 if (unlikely (!_hb_buffer_ensure_separate (buffer, buffer->out_len + num_out)))
352 mask = buffer->info[buffer->i].mask;
353 cluster = buffer->info[buffer->i].cluster;
354 if (component == 0xFFFF)
355 component = buffer->info[buffer->i].component;
356 if (lig_id == 0xFFFF)
357 lig_id = buffer->info[buffer->i].lig_id;
359 for (i = 0; i < num_out; i++)
361 hb_internal_glyph_info_t *info = &buffer->out_info[buffer->out_len + i];
362 info->codepoint = glyph_data[i];
364 info->cluster = cluster;
365 info->component = component;
366 info->lig_id = lig_id;
367 info->gproperty = HB_BUFFER_GLYPH_PROPERTIES_UNKNOWN;
371 buffer->out_len += num_out;
375 _hb_buffer_add_output_glyphs_be16 (hb_buffer_t *buffer,
377 unsigned int num_out,
378 const uint16_t *glyph_data_be,
379 unsigned short component,
380 unsigned short lig_id)
384 unsigned int cluster;
386 if (buffer->out_info != buffer->info ||
387 buffer->out_len + num_out > buffer->i + num_in)
389 if (unlikely (!_hb_buffer_ensure_separate (buffer, buffer->out_len + num_out)))
393 mask = buffer->info[buffer->i].mask;
394 cluster = buffer->info[buffer->i].cluster;
395 if (component == 0xFFFF)
396 component = buffer->info[buffer->i].component;
397 if (lig_id == 0xFFFF)
398 lig_id = buffer->info[buffer->i].lig_id;
400 for (i = 0; i < num_out; i++)
402 hb_internal_glyph_info_t *info = &buffer->out_info[buffer->out_len + i];
403 info->codepoint = hb_be_uint16 (glyph_data_be[i]);
405 info->cluster = cluster;
406 info->component = component;
407 info->lig_id = lig_id;
408 info->gproperty = HB_BUFFER_GLYPH_PROPERTIES_UNKNOWN;
412 buffer->out_len += num_out;
416 _hb_buffer_add_output_glyph (hb_buffer_t *buffer,
417 hb_codepoint_t glyph_index,
418 unsigned short component,
419 unsigned short lig_id)
421 hb_internal_glyph_info_t *info;
423 if (buffer->out_info != buffer->info)
425 if (unlikely (!_hb_buffer_ensure (buffer, buffer->out_len + 1))) return;
426 buffer->out_info[buffer->out_len] = buffer->info[buffer->i];
428 else if (buffer->out_len != buffer->i)
429 buffer->out_info[buffer->out_len] = buffer->info[buffer->i];
431 info = &buffer->out_info[buffer->out_len];
432 info->codepoint = glyph_index;
433 if (component != 0xFFFF)
434 info->component = component;
435 if (lig_id != 0xFFFF)
436 info->lig_id = lig_id;
437 info->gproperty = HB_BUFFER_GLYPH_PROPERTIES_UNKNOWN;
444 _hb_buffer_next_glyph (hb_buffer_t *buffer)
446 if (buffer->have_output)
448 if (buffer->out_info != buffer->info)
450 if (unlikely (!_hb_buffer_ensure (buffer, buffer->out_len + 1))) return;
451 buffer->out_info[buffer->out_len] = buffer->info[buffer->i];
453 else if (buffer->out_len != buffer->i)
454 buffer->out_info[buffer->out_len] = buffer->info[buffer->i];
463 _hb_buffer_clear_masks (hb_buffer_t *buffer)
465 unsigned int count = buffer->len;
466 for (unsigned int i = 0; i < count; i++)
467 buffer->info[i].mask = 1;
471 _hb_buffer_set_masks (hb_buffer_t *buffer,
474 unsigned int cluster_start,
475 unsigned int cluster_end)
477 hb_mask_t not_mask = ~mask;
479 if (cluster_start == 0 && cluster_end == (unsigned int)-1) {
480 unsigned int count = buffer->len;
481 for (unsigned int i = 0; i < count; i++)
482 buffer->info[i].mask = (buffer->info[i].mask & not_mask) | value;
486 /* Binary search to find the start position and go from there. */
487 unsigned int min = 0, max = buffer->len;
490 unsigned int mid = min + ((max - min) / 2);
491 if (buffer->info[mid].cluster < cluster_start)
496 unsigned int count = buffer->len;
497 for (unsigned int i = min; i < count && buffer->info[i].cluster < cluster_end; i++)
498 buffer->info[i].mask = (buffer->info[i].mask & not_mask) | value;
502 /* Public API again */
505 hb_buffer_get_length (hb_buffer_t *buffer)
510 /* Return value valid as long as buffer not modified */
512 hb_buffer_get_glyph_infos (hb_buffer_t *buffer)
514 return (hb_glyph_info_t *) buffer->info;
517 /* Return value valid as long as buffer not modified */
518 hb_glyph_position_t *
519 hb_buffer_get_glyph_positions (hb_buffer_t *buffer)
521 if (!buffer->have_positions)
522 hb_buffer_clear_positions (buffer);
524 return (hb_glyph_position_t *) buffer->pos;
529 reverse_range (hb_buffer_t *buffer,
535 for (i = start, j = end - 1; i < j; i++, j--) {
536 hb_internal_glyph_info_t t;
539 buffer->info[i] = buffer->info[j];
544 for (i = 0, j = end - 1; i < j; i++, j--) {
545 hb_internal_glyph_position_t t;
548 buffer->pos[i] = buffer->pos[j];
555 hb_buffer_reverse (hb_buffer_t *buffer)
557 if (unlikely (!buffer->len))
560 reverse_range (buffer, 0, buffer->len);
564 hb_buffer_reverse_clusters (hb_buffer_t *buffer)
566 unsigned int i, start, count, last_cluster;
568 if (unlikely (!buffer->len))
571 hb_buffer_reverse (buffer);
575 last_cluster = buffer->info[0].cluster;
576 for (i = 1; i < count; i++) {
577 if (last_cluster != buffer->info[i].cluster) {
578 reverse_range (buffer, start, i);
580 last_cluster = buffer->info[i].cluster;
583 reverse_range (buffer, start, i);
589 const T *next = (const T *) text + item_offset; \
590 const T *end = next + item_length; \
591 while (next < end) { \
593 const T *old_next = next; \
594 next = UTF_NEXT (next, end, u); \
595 hb_buffer_add_glyph (buffer, u, 1, old_next - (const T *) text); \
600 #define UTF8_COMPUTE(Char, Mask, Len) \
601 if (Char < 128) { Len = 1; Mask = 0x7f; } \
602 else if ((Char & 0xe0) == 0xc0) { Len = 2; Mask = 0x1f; } \
603 else if ((Char & 0xf0) == 0xe0) { Len = 3; Mask = 0x0f; } \
604 else if ((Char & 0xf8) == 0xf0) { Len = 4; Mask = 0x07; } \
607 static inline const uint8_t *
608 hb_utf8_next (const uint8_t *text,
610 hb_codepoint_t *unicode)
613 unsigned int mask, len;
615 /* TODO check for overlong sequences? also: optimize? */
617 UTF8_COMPUTE (c, mask, len);
618 if (unlikely (!len || (unsigned int) (end - text) < len)) {
622 hb_codepoint_t result;
625 for (i = 1; i < len; i++)
627 if (unlikely ((text[i] & 0xc0) != 0x80))
633 result |= (text[i] & 0x3f);
641 hb_buffer_add_utf8 (hb_buffer_t *buffer,
643 unsigned int text_length HB_UNUSED,
644 unsigned int item_offset,
645 unsigned int item_length)
647 #define UTF_NEXT(S, E, U) hb_utf8_next (S, E, &(U))
652 static inline const uint16_t *
653 hb_utf16_next (const uint16_t *text,
655 hb_codepoint_t *unicode)
657 uint16_t c = *text++;
659 if (unlikely (c >= 0xd800 && c < 0xdc00)) {
662 if (text < end && ((l = *text), unlikely (l >= 0xdc00 && l < 0xe000))) {
664 *unicode = ((hb_codepoint_t) ((c) - 0xd800) * 0x400 + (l) - 0xdc00 + 0x10000);
675 hb_buffer_add_utf16 (hb_buffer_t *buffer,
676 const uint16_t *text,
677 unsigned int text_length HB_UNUSED,
678 unsigned int item_offset,
679 unsigned int item_length)
681 #define UTF_NEXT(S, E, U) hb_utf16_next (S, E, &(U))
687 hb_buffer_add_utf32 (hb_buffer_t *buffer,
688 const uint32_t *text,
689 unsigned int text_length HB_UNUSED,
690 unsigned int item_offset,
691 unsigned int item_length)
693 #define UTF_NEXT(S, E, U) ((U) = *(S), (S)+1)