Bug 1883518 - Remove a bunch of unused ServoBindings.toml entries. r=firefox-style...
[gecko.git] / gfx / harfbuzz / src / hb-ot-cmap-table.hh
blobe2e258185572ded7c051d9e627cd35849c1fdcc2
1 /*
2 * Copyright © 2014 Google, Inc.
4 * This is part of HarfBuzz, a text shaping library.
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.
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.
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.
24 * Google Author(s): Behdad Esfahbod
27 #ifndef HB_OT_CMAP_TABLE_HH
28 #define HB_OT_CMAP_TABLE_HH
30 #include "hb-ot-os2-table.hh"
31 #include "hb-ot-shaper-arabic-pua.hh"
32 #include "hb-open-type.hh"
33 #include "hb-set.hh"
34 #include "hb-cache.hh"
37 * cmap -- Character to Glyph Index Mapping
38 * https://docs.microsoft.com/en-us/typography/opentype/spec/cmap
40 #define HB_OT_TAG_cmap HB_TAG('c','m','a','p')
42 namespace OT {
45 struct CmapSubtableFormat0
47 bool get_glyph (hb_codepoint_t codepoint, hb_codepoint_t *glyph) const
49 hb_codepoint_t gid = codepoint < 256 ? glyphIdArray[codepoint] : 0;
50 if (unlikely (!gid))
51 return false;
52 *glyph = gid;
53 return true;
56 unsigned get_language () const
58 return language;
61 void collect_unicodes (hb_set_t *out) const
63 for (unsigned int i = 0; i < 256; i++)
64 if (glyphIdArray[i])
65 out->add (i);
68 void collect_mapping (hb_set_t *unicodes, /* OUT */
69 hb_map_t *mapping /* OUT */) const
71 for (unsigned i = 0; i < 256; i++)
72 if (glyphIdArray[i])
74 hb_codepoint_t glyph = glyphIdArray[i];
75 unicodes->add (i);
76 mapping->set (i, glyph);
80 bool sanitize (hb_sanitize_context_t *c) const
82 TRACE_SANITIZE (this);
83 return_trace (c->check_struct (this));
86 protected:
87 HBUINT16 format; /* Format number is set to 0. */
88 HBUINT16 length; /* Byte length of this subtable. */
89 HBUINT16 language; /* Ignore. */
90 HBUINT8 glyphIdArray[256];/* An array that maps character
91 * code to glyph index values. */
92 public:
93 DEFINE_SIZE_STATIC (6 + 256);
96 struct CmapSubtableFormat4
100 template<typename Iterator,
101 typename Writer,
102 hb_requires (hb_is_iterator (Iterator))>
103 void to_ranges (Iterator it, Writer& range_writer)
105 hb_codepoint_t start_cp = 0, prev_run_start_cp = 0, run_start_cp = 0, end_cp = 0, last_gid = 0;
106 int run_length = 0 , delta = 0, prev_delta = 0;
108 enum {
109 FIRST_SUB_RANGE,
110 FOLLOWING_SUB_RANGE,
111 } mode;
113 while (it) {
114 // Start a new range
116 const auto& pair = *it;
117 start_cp = pair.first;
118 prev_run_start_cp = start_cp;
119 run_start_cp = start_cp;
120 end_cp = start_cp;
121 last_gid = pair.second;
122 run_length = 1;
123 prev_delta = 0;
126 delta = last_gid - start_cp;
127 mode = FIRST_SUB_RANGE;
128 it++;
130 while (it) {
131 // Process range
132 const auto& pair = *it;
133 hb_codepoint_t next_cp = pair.first;
134 hb_codepoint_t next_gid = pair.second;
135 if (next_cp != end_cp + 1) {
136 // Current range is over, stop processing.
137 break;
140 if (next_gid == last_gid + 1) {
141 // The current run continues.
142 end_cp = next_cp;
143 run_length++;
144 last_gid = next_gid;
145 it++;
146 continue;
149 // A new run is starting, decide if we want to commit the current run.
150 int split_cost = (mode == FIRST_SUB_RANGE) ? 8 : 16;
151 int run_cost = run_length * 2;
152 if (run_cost >= split_cost) {
153 commit_current_range(start_cp,
154 prev_run_start_cp,
155 run_start_cp,
156 end_cp,
157 delta,
158 prev_delta,
159 split_cost,
160 range_writer);
161 start_cp = next_cp;
164 // Start the new run
165 mode = FOLLOWING_SUB_RANGE;
166 prev_run_start_cp = run_start_cp;
167 run_start_cp = next_cp;
168 end_cp = next_cp;
169 prev_delta = delta;
170 delta = next_gid - run_start_cp;
171 run_length = 1;
172 last_gid = next_gid;
173 it++;
176 // Finalize range
177 commit_current_range (start_cp,
178 prev_run_start_cp,
179 run_start_cp,
180 end_cp,
181 delta,
182 prev_delta,
184 range_writer);
187 if (likely (end_cp != 0xFFFF)) {
188 range_writer (0xFFFF, 0xFFFF, 1);
193 * Writes the current range as either one or two ranges depending on what is most efficient.
195 template<typename Writer>
196 void commit_current_range (hb_codepoint_t start,
197 hb_codepoint_t prev_run_start,
198 hb_codepoint_t run_start,
199 hb_codepoint_t end,
200 int run_delta,
201 int previous_run_delta,
202 int split_cost,
203 Writer& range_writer) {
204 bool should_split = false;
205 if (start < run_start && run_start < end) {
206 int run_cost = (end - run_start + 1) * 2;
207 if (run_cost >= split_cost) {
208 should_split = true;
212 // TODO(grieger): handle case where delta is legitimately 0, mark range offset array instead?
213 if (should_split) {
214 if (start == prev_run_start)
215 range_writer (start, run_start - 1, previous_run_delta);
216 else
217 range_writer (start, run_start - 1, 0);
218 range_writer (run_start, end, run_delta);
219 return;
223 if (start == run_start) {
224 // Range is only a run
225 range_writer (start, end, run_delta);
226 return;
229 // Write only a single non-run range.
230 range_writer (start, end, 0);
233 template<typename Iterator,
234 hb_requires (hb_is_iterator (Iterator))>
235 unsigned serialize_find_segcount (Iterator it) {
236 struct Counter {
237 unsigned segcount = 0;
239 void operator() (hb_codepoint_t start,
240 hb_codepoint_t end,
241 int delta) {
242 segcount++;
244 } counter;
246 to_ranges (+it, counter);
247 return counter.segcount;
251 template<typename Iterator,
252 hb_requires (hb_is_iterator (Iterator))>
253 bool serialize_start_end_delta_arrays (hb_serialize_context_t *c,
254 Iterator it,
255 int segcount)
257 struct Writer {
258 hb_serialize_context_t *serializer_;
259 HBUINT16* end_code_;
260 HBUINT16* start_code_;
261 HBINT16* id_delta_;
262 int index_;
264 Writer(hb_serialize_context_t *serializer)
265 : serializer_(serializer),
266 end_code_(nullptr),
267 start_code_(nullptr),
268 id_delta_(nullptr),
269 index_ (0) {}
270 void operator() (hb_codepoint_t start,
271 hb_codepoint_t end,
272 int delta) {
273 start_code_[index_] = start;
274 end_code_[index_] = end;
275 id_delta_[index_] = delta;
276 index_++;
278 } writer(c);
280 writer.end_code_ = c->allocate_size<HBUINT16> (HBUINT16::static_size * segcount, false);
281 (void) c->allocate_size<HBUINT16> (2); // padding
282 writer.start_code_ = c->allocate_size<HBUINT16> (HBUINT16::static_size * segcount, false);
283 writer.id_delta_ = c->allocate_size<HBINT16> (HBINT16::static_size * segcount, false);
285 if (unlikely (!writer.end_code_ || !writer.start_code_ || !writer.id_delta_)) return false;
287 to_ranges (+it, writer);
288 return true;
291 template<typename Iterator,
292 hb_requires (hb_is_iterator (Iterator))>
293 HBUINT16* serialize_rangeoffset_glyid (hb_serialize_context_t *c,
294 Iterator it,
295 HBUINT16 *endCode,
296 HBUINT16 *startCode,
297 HBINT16 *idDelta,
298 unsigned segcount)
300 hb_map_t cp_to_gid { it };
302 HBUINT16 *idRangeOffset = c->allocate_size<HBUINT16> (HBUINT16::static_size * segcount);
303 if (unlikely (!c->check_success (idRangeOffset))) return nullptr;
304 if (unlikely ((char *)idRangeOffset - (char *)idDelta != (int) segcount * (int) HBINT16::static_size)) return nullptr;
306 for (unsigned i : + hb_range (segcount)
307 | hb_filter ([&] (const unsigned _) { return idDelta[_] == 0; }))
309 idRangeOffset[i] = 2 * (c->start_embed<HBUINT16> () - idRangeOffset - i);
310 for (hb_codepoint_t cp = startCode[i]; cp <= endCode[i]; cp++)
312 HBUINT16 gid;
313 gid = cp_to_gid[cp];
314 c->copy<HBUINT16> (gid);
318 return idRangeOffset;
321 template<typename Iterator,
322 hb_requires (hb_is_iterator (Iterator))>
323 void serialize (hb_serialize_context_t *c,
324 Iterator it)
326 auto format4_iter =
327 + it
328 | hb_filter ([&] (const hb_codepoint_pair_t _)
329 { return _.first <= 0xFFFF; })
332 if (!format4_iter) return;
334 unsigned table_initpos = c->length ();
335 if (unlikely (!c->extend_min (this))) return;
336 this->format = 4;
338 hb_vector_t<hb_codepoint_pair_t> cp_to_gid {
339 format4_iter
342 //serialize endCode[], startCode[], idDelta[]
343 HBUINT16* endCode = c->start_embed<HBUINT16> ();
344 unsigned segcount = serialize_find_segcount (cp_to_gid.iter());
345 if (unlikely (!serialize_start_end_delta_arrays (c, cp_to_gid.iter(), segcount)))
346 return;
348 HBUINT16 *startCode = endCode + segcount + 1;
349 HBINT16 *idDelta = ((HBINT16*)startCode) + segcount;
351 HBUINT16 *idRangeOffset = serialize_rangeoffset_glyid (c,
352 cp_to_gid.iter (),
353 endCode,
354 startCode,
355 idDelta,
356 segcount);
357 if (unlikely (!c->check_success (idRangeOffset))) return;
359 this->length = c->length () - table_initpos;
360 if ((long long) this->length != (long long) c->length () - table_initpos)
362 // Length overflowed. Discard the current object before setting the error condition, otherwise
363 // discard is a noop which prevents the higher level code from reverting the serializer to the
364 // pre-error state in cmap4 overflow handling code.
365 c->pop_discard ();
366 c->err (HB_SERIALIZE_ERROR_INT_OVERFLOW);
367 return;
370 this->segCountX2 = segcount * 2;
371 this->entrySelector = hb_max (1u, hb_bit_storage (segcount)) - 1;
372 this->searchRange = 2 * (1u << this->entrySelector);
373 this->rangeShift = segcount * 2 > this->searchRange
374 ? 2 * segcount - this->searchRange
375 : 0;
378 unsigned get_language () const
380 return language;
383 struct accelerator_t
385 accelerator_t () {}
386 accelerator_t (const CmapSubtableFormat4 *subtable) { init (subtable); }
388 void init (const CmapSubtableFormat4 *subtable)
390 segCount = subtable->segCountX2 / 2;
391 endCount = subtable->values.arrayZ;
392 startCount = endCount + segCount + 1;
393 idDelta = startCount + segCount;
394 idRangeOffset = idDelta + segCount;
395 glyphIdArray = idRangeOffset + segCount;
396 glyphIdArrayLength = (subtable->length - 16 - 8 * segCount) / 2;
399 bool get_glyph (hb_codepoint_t codepoint, hb_codepoint_t *glyph) const
401 struct CustomRange
403 int cmp (hb_codepoint_t k,
404 unsigned distance) const
406 if (k > last) return +1;
407 if (k < (&last)[distance]/*first*/) return -1;
408 return 0;
410 HBUINT16 last;
413 const HBUINT16 *found = hb_bsearch (codepoint,
414 this->endCount,
415 this->segCount,
416 sizeof (CustomRange),
417 _hb_cmp_method<hb_codepoint_t, CustomRange, unsigned>,
418 this->segCount + 1);
419 if (unlikely (!found))
420 return false;
421 unsigned int i = found - endCount;
423 hb_codepoint_t gid;
424 unsigned int rangeOffset = this->idRangeOffset[i];
425 if (rangeOffset == 0)
426 gid = codepoint + this->idDelta[i];
427 else
429 /* Somebody has been smoking... */
430 unsigned int index = rangeOffset / 2 + (codepoint - this->startCount[i]) + i - this->segCount;
431 if (unlikely (index >= this->glyphIdArrayLength))
432 return false;
433 gid = this->glyphIdArray[index];
434 if (unlikely (!gid))
435 return false;
436 gid += this->idDelta[i];
438 gid &= 0xFFFFu;
439 if (unlikely (!gid))
440 return false;
441 *glyph = gid;
442 return true;
445 HB_INTERNAL static bool get_glyph_func (const void *obj, hb_codepoint_t codepoint, hb_codepoint_t *glyph)
446 { return ((const accelerator_t *) obj)->get_glyph (codepoint, glyph); }
448 void collect_unicodes (hb_set_t *out) const
450 unsigned int count = this->segCount;
451 if (count && this->startCount[count - 1] == 0xFFFFu)
452 count--; /* Skip sentinel segment. */
453 for (unsigned int i = 0; i < count; i++)
455 hb_codepoint_t start = this->startCount[i];
456 hb_codepoint_t end = this->endCount[i];
457 unsigned int rangeOffset = this->idRangeOffset[i];
458 out->add_range(start, end);
459 if (rangeOffset == 0)
461 for (hb_codepoint_t codepoint = start; codepoint <= end; codepoint++)
463 hb_codepoint_t gid = (codepoint + this->idDelta[i]) & 0xFFFFu;
464 if (unlikely (!gid))
465 out->del(codepoint);
468 else
470 for (hb_codepoint_t codepoint = start; codepoint <= end; codepoint++)
472 unsigned int index = rangeOffset / 2 + (codepoint - this->startCount[i]) + i - this->segCount;
473 if (unlikely (index >= this->glyphIdArrayLength))
475 out->del_range (codepoint, end);
476 break;
478 hb_codepoint_t gid = this->glyphIdArray[index];
479 if (unlikely (!gid))
480 out->del(codepoint);
486 void collect_mapping (hb_set_t *unicodes, /* OUT */
487 hb_map_t *mapping /* OUT */) const
489 // TODO(grieger): optimize similar to collect_unicodes
490 // (ie. use add_range())
491 unsigned count = this->segCount;
492 if (count && this->startCount[count - 1] == 0xFFFFu)
493 count--; /* Skip sentinel segment. */
494 for (unsigned i = 0; i < count; i++)
496 hb_codepoint_t start = this->startCount[i];
497 hb_codepoint_t end = this->endCount[i];
498 unsigned rangeOffset = this->idRangeOffset[i];
499 if (rangeOffset == 0)
501 for (hb_codepoint_t codepoint = start; codepoint <= end; codepoint++)
503 hb_codepoint_t gid = (codepoint + this->idDelta[i]) & 0xFFFFu;
504 if (unlikely (!gid))
505 continue;
506 unicodes->add (codepoint);
507 mapping->set (codepoint, gid);
510 else
512 for (hb_codepoint_t codepoint = start; codepoint <= end; codepoint++)
514 unsigned index = rangeOffset / 2 + (codepoint - this->startCount[i]) + i - this->segCount;
515 if (unlikely (index >= this->glyphIdArrayLength))
516 break;
517 hb_codepoint_t gid = this->glyphIdArray[index];
518 if (unlikely (!gid))
519 continue;
520 unicodes->add (codepoint);
521 mapping->set (codepoint, gid);
527 const HBUINT16 *endCount;
528 const HBUINT16 *startCount;
529 const HBUINT16 *idDelta;
530 const HBUINT16 *idRangeOffset;
531 const HBUINT16 *glyphIdArray;
532 unsigned int segCount;
533 unsigned int glyphIdArrayLength;
536 bool get_glyph (hb_codepoint_t codepoint, hb_codepoint_t *glyph) const
538 accelerator_t accel (this);
539 return accel.get_glyph_func (&accel, codepoint, glyph);
541 void collect_unicodes (hb_set_t *out) const
543 accelerator_t accel (this);
544 accel.collect_unicodes (out);
547 void collect_mapping (hb_set_t *unicodes, /* OUT */
548 hb_map_t *mapping /* OUT */) const
550 accelerator_t accel (this);
551 accel.collect_mapping (unicodes, mapping);
554 bool sanitize (hb_sanitize_context_t *c) const
556 TRACE_SANITIZE (this);
557 if (unlikely (!c->check_struct (this)))
558 return_trace (false);
559 hb_barrier ();
561 if (unlikely (!c->check_range (this, length)))
563 /* Some broken fonts have too long of a "length" value.
564 * If that is the case, just change the value to truncate
565 * the subtable at the end of the blob. */
566 uint16_t new_length = (uint16_t) hb_min ((uintptr_t) 65535,
567 (uintptr_t) (c->end -
568 (char *) this));
569 if (!c->try_set (&length, new_length))
570 return_trace (false);
573 return_trace (16 + 4 * (unsigned int) segCountX2 <= length);
578 protected:
579 HBUINT16 format; /* Format number is set to 4. */
580 HBUINT16 length; /* This is the length in bytes of the
581 * subtable. */
582 HBUINT16 language; /* Ignore. */
583 HBUINT16 segCountX2; /* 2 x segCount. */
584 HBUINT16 searchRange; /* 2 * (2**floor(log2(segCount))) */
585 HBUINT16 entrySelector; /* log2(searchRange/2) */
586 HBUINT16 rangeShift; /* 2 x segCount - searchRange */
588 UnsizedArrayOf<HBUINT16>
589 values;
590 #if 0
591 HBUINT16 endCount[segCount]; /* End characterCode for each segment,
592 * last=0xFFFFu. */
593 HBUINT16 reservedPad; /* Set to 0. */
594 HBUINT16 startCount[segCount]; /* Start character code for each segment. */
595 HBINT16 idDelta[segCount]; /* Delta for all character codes in segment. */
596 HBUINT16 idRangeOffset[segCount];/* Offsets into glyphIdArray or 0 */
597 UnsizedArrayOf<HBUINT16>
598 glyphIdArray; /* Glyph index array (arbitrary length) */
599 #endif
601 public:
602 DEFINE_SIZE_ARRAY (14, values);
605 struct CmapSubtableLongGroup
607 friend struct CmapSubtableFormat12;
608 friend struct CmapSubtableFormat13;
609 template<typename U>
610 friend struct CmapSubtableLongSegmented;
611 friend struct cmap;
613 int cmp (hb_codepoint_t codepoint) const
615 if (codepoint < startCharCode) return -1;
616 if (codepoint > endCharCode) return +1;
617 return 0;
620 bool sanitize (hb_sanitize_context_t *c) const
622 TRACE_SANITIZE (this);
623 return_trace (c->check_struct (this));
626 private:
627 HBUINT32 startCharCode; /* First character code in this group. */
628 HBUINT32 endCharCode; /* Last character code in this group. */
629 HBUINT32 glyphID; /* Glyph index; interpretation depends on
630 * subtable format. */
631 public:
632 DEFINE_SIZE_STATIC (12);
634 DECLARE_NULL_NAMESPACE_BYTES (OT, CmapSubtableLongGroup);
636 template <typename UINT>
637 struct CmapSubtableTrimmed
639 bool get_glyph (hb_codepoint_t codepoint, hb_codepoint_t *glyph) const
641 /* Rely on our implicit array bound-checking. */
642 hb_codepoint_t gid = glyphIdArray[codepoint - startCharCode];
643 if (unlikely (!gid))
644 return false;
645 *glyph = gid;
646 return true;
649 unsigned get_language () const
651 return language;
654 void collect_unicodes (hb_set_t *out) const
656 hb_codepoint_t start = startCharCode;
657 unsigned int count = glyphIdArray.len;
658 for (unsigned int i = 0; i < count; i++)
659 if (glyphIdArray[i])
660 out->add (start + i);
663 void collect_mapping (hb_set_t *unicodes, /* OUT */
664 hb_map_t *mapping /* OUT */) const
666 hb_codepoint_t start_cp = startCharCode;
667 unsigned count = glyphIdArray.len;
668 for (unsigned i = 0; i < count; i++)
669 if (glyphIdArray[i])
671 hb_codepoint_t unicode = start_cp + i;
672 hb_codepoint_t glyphid = glyphIdArray[i];
673 unicodes->add (unicode);
674 mapping->set (unicode, glyphid);
678 bool sanitize (hb_sanitize_context_t *c) const
680 TRACE_SANITIZE (this);
681 return_trace (c->check_struct (this) && glyphIdArray.sanitize (c));
684 protected:
685 UINT formatReserved; /* Subtable format and (maybe) padding. */
686 UINT length; /* Byte length of this subtable. */
687 UINT language; /* Ignore. */
688 UINT startCharCode; /* First character code covered. */
689 ArrayOf<HBGlyphID16, UINT>
690 glyphIdArray; /* Array of glyph index values for character
691 * codes in the range. */
692 public:
693 DEFINE_SIZE_ARRAY (5 * sizeof (UINT), glyphIdArray);
696 struct CmapSubtableFormat6 : CmapSubtableTrimmed<HBUINT16> {};
697 struct CmapSubtableFormat10 : CmapSubtableTrimmed<HBUINT32> {};
699 template <typename T>
700 struct CmapSubtableLongSegmented
702 friend struct cmap;
704 bool get_glyph (hb_codepoint_t codepoint, hb_codepoint_t *glyph) const
706 hb_codepoint_t gid = T::group_get_glyph (groups.bsearch (codepoint), codepoint);
707 if (unlikely (!gid))
708 return false;
709 *glyph = gid;
710 return true;
713 unsigned get_language () const
715 return language;
718 void collect_unicodes (hb_set_t *out, unsigned int num_glyphs) const
720 for (unsigned int i = 0; i < this->groups.len; i++)
722 hb_codepoint_t start = this->groups[i].startCharCode;
723 hb_codepoint_t end = hb_min ((hb_codepoint_t) this->groups[i].endCharCode,
724 (hb_codepoint_t) HB_UNICODE_MAX);
725 hb_codepoint_t gid = this->groups[i].glyphID;
726 if (!gid)
728 /* Intention is: if (hb_is_same (T, CmapSubtableFormat13)) continue; */
729 if (! T::group_get_glyph (this->groups[i], end)) continue;
730 start++;
731 gid++;
733 if (unlikely ((unsigned int) gid >= num_glyphs)) continue;
734 if (unlikely ((unsigned int) (gid + end - start) >= num_glyphs))
735 end = start + (hb_codepoint_t) num_glyphs - gid;
737 out->add_range (start, hb_min (end, 0x10FFFFu));
741 void collect_mapping (hb_set_t *unicodes, /* OUT */
742 hb_map_t *mapping, /* OUT */
743 unsigned num_glyphs) const
745 hb_codepoint_t last_end = 0;
746 unsigned count = this->groups.len;
747 for (unsigned i = 0; i < count; i++)
749 hb_codepoint_t start = this->groups.arrayZ[i].startCharCode;
750 hb_codepoint_t end = hb_min ((hb_codepoint_t) this->groups.arrayZ[i].endCharCode,
751 (hb_codepoint_t) HB_UNICODE_MAX);
752 if (unlikely (start > end || start < last_end)) {
753 // Range is not in order and is invalid, skip it.
754 continue;
756 last_end = end;
759 hb_codepoint_t gid = this->groups.arrayZ[i].glyphID;
760 if (!gid)
762 if (T::formatNumber == 13) continue;
763 start++;
764 gid++;
766 if (unlikely ((unsigned int) gid >= num_glyphs)) continue;
767 if (unlikely ((unsigned int) (gid + end - start) >= num_glyphs))
768 end = start + (hb_codepoint_t) num_glyphs - gid;
770 mapping->alloc (mapping->get_population () + end - start + 1);
772 unicodes->add_range (start, end);
773 for (unsigned cp = start; cp <= end; cp++)
775 mapping->set (cp, gid);
776 gid += T::increment;
781 bool sanitize (hb_sanitize_context_t *c) const
783 TRACE_SANITIZE (this);
784 return_trace (c->check_struct (this) && groups.sanitize (c));
787 protected:
788 HBUINT16 format; /* Subtable format; set to 12. */
789 HBUINT16 reserved; /* Reserved; set to 0. */
790 HBUINT32 length; /* Byte length of this subtable. */
791 HBUINT32 language; /* Ignore. */
792 SortedArray32Of<CmapSubtableLongGroup>
793 groups; /* Groupings. */
794 public:
795 DEFINE_SIZE_ARRAY (16, groups);
798 struct CmapSubtableFormat12 : CmapSubtableLongSegmented<CmapSubtableFormat12>
800 static constexpr int increment = 1;
801 static constexpr int formatNumber = 12;
803 static hb_codepoint_t group_get_glyph (const CmapSubtableLongGroup &group,
804 hb_codepoint_t u)
805 { return likely (group.startCharCode <= group.endCharCode) ?
806 group.glyphID + (u - group.startCharCode) : 0; }
809 template<typename Iterator,
810 hb_requires (hb_is_iterator (Iterator))>
811 void serialize (hb_serialize_context_t *c,
812 Iterator it)
814 if (!it) return;
815 unsigned table_initpos = c->length ();
816 if (unlikely (!c->extend_min (this))) return;
818 hb_codepoint_t startCharCode = (hb_codepoint_t) -1, endCharCode = (hb_codepoint_t) -1;
819 hb_codepoint_t glyphID = 0;
821 for (const auto& _ : +it)
823 if (startCharCode == (hb_codepoint_t) -1)
825 startCharCode = _.first;
826 endCharCode = _.first;
827 glyphID = _.second;
829 else if (!_is_gid_consecutive (endCharCode, startCharCode, glyphID, _.first, _.second))
831 CmapSubtableLongGroup grouprecord;
832 grouprecord.startCharCode = startCharCode;
833 grouprecord.endCharCode = endCharCode;
834 grouprecord.glyphID = glyphID;
835 c->copy<CmapSubtableLongGroup> (grouprecord);
837 startCharCode = _.first;
838 endCharCode = _.first;
839 glyphID = _.second;
841 else
842 endCharCode = _.first;
845 CmapSubtableLongGroup record;
846 record.startCharCode = startCharCode;
847 record.endCharCode = endCharCode;
848 record.glyphID = glyphID;
849 c->copy<CmapSubtableLongGroup> (record);
851 this->format = 12;
852 this->reserved = 0;
853 this->length = c->length () - table_initpos;
854 this->groups.len = (this->length - min_size) / CmapSubtableLongGroup::static_size;
857 static size_t get_sub_table_size (const hb_sorted_vector_t<CmapSubtableLongGroup> &groups_data)
858 { return 16 + 12 * groups_data.length; }
860 private:
861 static bool _is_gid_consecutive (hb_codepoint_t endCharCode,
862 hb_codepoint_t startCharCode,
863 hb_codepoint_t glyphID,
864 hb_codepoint_t cp,
865 hb_codepoint_t new_gid)
867 return (cp - 1 == endCharCode) &&
868 new_gid == glyphID + (cp - startCharCode);
873 struct CmapSubtableFormat13 : CmapSubtableLongSegmented<CmapSubtableFormat13>
875 static constexpr int increment = 0;
876 static constexpr int formatNumber = 13;
878 static hb_codepoint_t group_get_glyph (const CmapSubtableLongGroup &group,
879 hb_codepoint_t u HB_UNUSED)
880 { return group.glyphID; }
883 typedef enum
885 GLYPH_VARIANT_NOT_FOUND = 0,
886 GLYPH_VARIANT_FOUND = 1,
887 GLYPH_VARIANT_USE_DEFAULT = 2
888 } glyph_variant_t;
890 struct UnicodeValueRange
892 int cmp (const hb_codepoint_t &codepoint) const
894 if (codepoint < startUnicodeValue) return -1;
895 if (codepoint > startUnicodeValue + additionalCount) return +1;
896 return 0;
899 bool sanitize (hb_sanitize_context_t *c) const
901 TRACE_SANITIZE (this);
902 return_trace (c->check_struct (this));
905 HBUINT24 startUnicodeValue; /* First value in this range. */
906 HBUINT8 additionalCount; /* Number of additional values in this
907 * range. */
908 public:
909 DEFINE_SIZE_STATIC (4);
912 struct DefaultUVS : SortedArray32Of<UnicodeValueRange>
914 void collect_unicodes (hb_set_t *out) const
916 unsigned int count = len;
917 for (unsigned int i = 0; i < count; i++)
919 hb_codepoint_t first = arrayZ[i].startUnicodeValue;
920 hb_codepoint_t last = hb_min ((hb_codepoint_t) (first + arrayZ[i].additionalCount),
921 (hb_codepoint_t) HB_UNICODE_MAX);
922 out->add_range (first, last);
926 DefaultUVS* copy (hb_serialize_context_t *c,
927 const hb_set_t *unicodes) const
929 auto *out = c->start_embed<DefaultUVS> ();
930 auto snap = c->snapshot ();
932 HBUINT32 len;
933 len = 0;
934 if (unlikely (!c->copy<HBUINT32> (len))) return nullptr;
935 unsigned init_len = c->length ();
937 if (this->len > unicodes->get_population () * hb_bit_storage ((unsigned) this->len))
939 hb_codepoint_t start = HB_SET_VALUE_INVALID;
940 hb_codepoint_t end = HB_SET_VALUE_INVALID;
942 for (auto u : *unicodes)
944 if (!as_array ().bsearch (u))
945 continue;
946 if (start == HB_SET_VALUE_INVALID)
948 start = u;
949 end = start - 1;
951 if (end + 1 != u || end - start == 255)
953 UnicodeValueRange rec;
954 rec.startUnicodeValue = start;
955 rec.additionalCount = end - start;
956 c->copy<UnicodeValueRange> (rec);
957 start = u;
959 end = u;
961 if (start != HB_SET_VALUE_INVALID)
963 UnicodeValueRange rec;
964 rec.startUnicodeValue = start;
965 rec.additionalCount = end - start;
966 c->copy<UnicodeValueRange> (rec);
970 else
972 hb_codepoint_t lastCode = HB_SET_VALUE_INVALID;
973 int count = -1;
975 for (const UnicodeValueRange& _ : *this)
977 hb_codepoint_t curEntry = (hb_codepoint_t) (_.startUnicodeValue - 1);
978 hb_codepoint_t end = curEntry + _.additionalCount + 2;
980 for (; unicodes->next (&curEntry) && curEntry < end;)
982 count += 1;
983 if (lastCode == HB_SET_VALUE_INVALID)
984 lastCode = curEntry;
985 else if (lastCode + count != curEntry)
987 UnicodeValueRange rec;
988 rec.startUnicodeValue = lastCode;
989 rec.additionalCount = count - 1;
990 c->copy<UnicodeValueRange> (rec);
992 lastCode = curEntry;
993 count = 0;
998 if (lastCode != HB_MAP_VALUE_INVALID)
1000 UnicodeValueRange rec;
1001 rec.startUnicodeValue = lastCode;
1002 rec.additionalCount = count;
1003 c->copy<UnicodeValueRange> (rec);
1007 if (c->length () - init_len == 0)
1009 c->revert (snap);
1010 return nullptr;
1012 else
1014 if (unlikely (!c->check_assign (out->len,
1015 (c->length () - init_len) / UnicodeValueRange::static_size,
1016 HB_SERIALIZE_ERROR_INT_OVERFLOW))) return nullptr;
1017 return out;
1021 public:
1022 DEFINE_SIZE_ARRAY (4, *this);
1025 struct UVSMapping
1027 int cmp (const hb_codepoint_t &codepoint) const
1028 { return unicodeValue.cmp (codepoint); }
1030 bool sanitize (hb_sanitize_context_t *c) const
1032 TRACE_SANITIZE (this);
1033 return_trace (c->check_struct (this));
1036 HBUINT24 unicodeValue; /* Base Unicode value of the UVS */
1037 HBGlyphID16 glyphID; /* Glyph ID of the UVS */
1038 public:
1039 DEFINE_SIZE_STATIC (5);
1042 struct NonDefaultUVS : SortedArray32Of<UVSMapping>
1044 void collect_unicodes (hb_set_t *out) const
1046 for (const auto& a : as_array ())
1047 out->add (a.unicodeValue);
1050 void collect_mapping (hb_set_t *unicodes, /* OUT */
1051 hb_map_t *mapping /* OUT */) const
1053 for (const auto& a : as_array ())
1055 hb_codepoint_t unicode = a.unicodeValue;
1056 hb_codepoint_t glyphid = a.glyphID;
1057 unicodes->add (unicode);
1058 mapping->set (unicode, glyphid);
1062 void closure_glyphs (const hb_set_t *unicodes,
1063 hb_set_t *glyphset) const
1065 + as_array ()
1066 | hb_filter (unicodes, &UVSMapping::unicodeValue)
1067 | hb_map (&UVSMapping::glyphID)
1068 | hb_sink (glyphset)
1072 NonDefaultUVS* copy (hb_serialize_context_t *c,
1073 const hb_set_t *unicodes,
1074 const hb_set_t *glyphs_requested,
1075 const hb_map_t *glyph_map) const
1077 auto *out = c->start_embed<NonDefaultUVS> ();
1078 auto it =
1079 + as_array ()
1080 | hb_filter ([&] (const UVSMapping& _)
1082 return unicodes->has (_.unicodeValue) || glyphs_requested->has (_.glyphID);
1086 if (!it) return nullptr;
1088 HBUINT32 len;
1089 len = it.len ();
1090 if (unlikely (!c->copy<HBUINT32> (len))) return nullptr;
1092 for (const UVSMapping& _ : it)
1094 UVSMapping mapping;
1095 mapping.unicodeValue = _.unicodeValue;
1096 mapping.glyphID = glyph_map->get (_.glyphID);
1097 c->copy<UVSMapping> (mapping);
1100 return out;
1103 public:
1104 DEFINE_SIZE_ARRAY (4, *this);
1107 struct VariationSelectorRecord
1109 glyph_variant_t get_glyph (hb_codepoint_t codepoint,
1110 hb_codepoint_t *glyph,
1111 const void *base) const
1113 if ((base+defaultUVS).bfind (codepoint))
1114 return GLYPH_VARIANT_USE_DEFAULT;
1115 const UVSMapping &nonDefault = (base+nonDefaultUVS).bsearch (codepoint);
1116 if (nonDefault.glyphID)
1118 *glyph = nonDefault.glyphID;
1119 return GLYPH_VARIANT_FOUND;
1121 return GLYPH_VARIANT_NOT_FOUND;
1124 VariationSelectorRecord(const VariationSelectorRecord& other)
1126 *this = other;
1129 void operator= (const VariationSelectorRecord& other)
1131 varSelector = other.varSelector;
1132 HBUINT32 offset = other.defaultUVS;
1133 defaultUVS = offset;
1134 offset = other.nonDefaultUVS;
1135 nonDefaultUVS = offset;
1138 void collect_unicodes (hb_set_t *out, const void *base) const
1140 (base+defaultUVS).collect_unicodes (out);
1141 (base+nonDefaultUVS).collect_unicodes (out);
1144 void collect_mapping (const void *base,
1145 hb_set_t *unicodes, /* OUT */
1146 hb_map_t *mapping /* OUT */) const
1148 (base+defaultUVS).collect_unicodes (unicodes);
1149 (base+nonDefaultUVS).collect_mapping (unicodes, mapping);
1152 int cmp (const hb_codepoint_t &variation_selector) const
1153 { return varSelector.cmp (variation_selector); }
1155 bool sanitize (hb_sanitize_context_t *c, const void *base) const
1157 TRACE_SANITIZE (this);
1158 return_trace (c->check_struct (this) &&
1159 defaultUVS.sanitize (c, base) &&
1160 nonDefaultUVS.sanitize (c, base));
1163 hb_pair_t<unsigned, unsigned>
1164 copy (hb_serialize_context_t *c,
1165 const hb_set_t *unicodes,
1166 const hb_set_t *glyphs_requested,
1167 const hb_map_t *glyph_map,
1168 const void *base) const
1170 auto snap = c->snapshot ();
1171 auto *out = c->embed<VariationSelectorRecord> (*this);
1172 if (unlikely (!out)) return hb_pair (0, 0);
1174 out->defaultUVS = 0;
1175 out->nonDefaultUVS = 0;
1177 unsigned non_default_uvs_objidx = 0;
1178 if (nonDefaultUVS != 0)
1180 c->push ();
1181 if (c->copy (base+nonDefaultUVS, unicodes, glyphs_requested, glyph_map))
1182 non_default_uvs_objidx = c->pop_pack ();
1183 else c->pop_discard ();
1186 unsigned default_uvs_objidx = 0;
1187 if (defaultUVS != 0)
1189 c->push ();
1190 if (c->copy (base+defaultUVS, unicodes))
1191 default_uvs_objidx = c->pop_pack ();
1192 else c->pop_discard ();
1196 if (!default_uvs_objidx && !non_default_uvs_objidx)
1197 c->revert (snap);
1199 return hb_pair (default_uvs_objidx, non_default_uvs_objidx);
1202 HBUINT24 varSelector; /* Variation selector. */
1203 Offset32To<DefaultUVS>
1204 defaultUVS; /* Offset to Default UVS Table. May be 0. */
1205 Offset32To<NonDefaultUVS>
1206 nonDefaultUVS; /* Offset to Non-Default UVS Table. May be 0. */
1207 public:
1208 DEFINE_SIZE_STATIC (11);
1211 struct CmapSubtableFormat14
1213 glyph_variant_t get_glyph_variant (hb_codepoint_t codepoint,
1214 hb_codepoint_t variation_selector,
1215 hb_codepoint_t *glyph) const
1216 { return record.bsearch (variation_selector).get_glyph (codepoint, glyph, this); }
1218 void collect_variation_selectors (hb_set_t *out) const
1220 for (const auto& a : record.as_array ())
1221 out->add (a.varSelector);
1223 void collect_variation_unicodes (hb_codepoint_t variation_selector,
1224 hb_set_t *out) const
1225 { record.bsearch (variation_selector).collect_unicodes (out, this); }
1227 void serialize (hb_serialize_context_t *c,
1228 const hb_set_t *unicodes,
1229 const hb_set_t *glyphs_requested,
1230 const hb_map_t *glyph_map,
1231 const void *base)
1233 auto snap = c->snapshot ();
1234 unsigned table_initpos = c->length ();
1235 const char* init_tail = c->tail;
1237 if (unlikely (!c->extend_min (this))) return;
1238 this->format = 14;
1240 auto src_tbl = reinterpret_cast<const CmapSubtableFormat14*> (base);
1243 * Some versions of OTS require that offsets are in order. Due to the use
1244 * of push()/pop_pack() serializing the variation records in order results
1245 * in the offsets being in reverse order (first record has the largest
1246 * offset). While this is perfectly valid, it will cause some versions of
1247 * OTS to consider this table bad.
1249 * So to prevent this issue we serialize the variation records in reverse
1250 * order, so that the offsets are ordered from small to large. Since
1251 * variation records are supposed to be in increasing order of varSelector
1252 * we then have to reverse the order of the written variation selector
1253 * records after everything is finalized.
1255 hb_vector_t<hb_pair_t<unsigned, unsigned>> obj_indices;
1256 for (int i = src_tbl->record.len - 1; i >= 0; i--)
1258 hb_pair_t<unsigned, unsigned> result = src_tbl->record[i].copy (c, unicodes, glyphs_requested, glyph_map, base);
1259 if (result.first || result.second)
1260 obj_indices.push (result);
1263 if (c->length () - table_initpos == CmapSubtableFormat14::min_size)
1265 c->revert (snap);
1266 return;
1269 if (unlikely (!c->check_success (!obj_indices.in_error ())))
1270 return;
1272 int tail_len = init_tail - c->tail;
1273 c->check_assign (this->length, c->length () - table_initpos + tail_len,
1274 HB_SERIALIZE_ERROR_INT_OVERFLOW);
1275 c->check_assign (this->record.len,
1276 (c->length () - table_initpos - CmapSubtableFormat14::min_size) /
1277 VariationSelectorRecord::static_size,
1278 HB_SERIALIZE_ERROR_INT_OVERFLOW);
1280 /* Correct the incorrect write order by reversing the order of the variation
1281 records array. */
1282 _reverse_variation_records ();
1284 /* Now that records are in the right order, we can set up the offsets. */
1285 _add_links_to_variation_records (c, obj_indices);
1288 void _reverse_variation_records ()
1290 record.as_array ().reverse ();
1293 void _add_links_to_variation_records (hb_serialize_context_t *c,
1294 const hb_vector_t<hb_pair_t<unsigned, unsigned>>& obj_indices)
1296 for (unsigned i = 0; i < obj_indices.length; i++)
1299 * Since the record array has been reversed (see comments in copy())
1300 * but obj_indices has not been, the indices at obj_indices[i]
1301 * are for the variation record at record[j].
1303 int j = obj_indices.length - 1 - i;
1304 c->add_link (record[j].defaultUVS, obj_indices[i].first);
1305 c->add_link (record[j].nonDefaultUVS, obj_indices[i].second);
1309 void closure_glyphs (const hb_set_t *unicodes,
1310 hb_set_t *glyphset) const
1312 + hb_iter (record)
1313 | hb_filter (hb_bool, &VariationSelectorRecord::nonDefaultUVS)
1314 | hb_map (&VariationSelectorRecord::nonDefaultUVS)
1315 | hb_map (hb_add (this))
1316 | hb_apply ([=] (const NonDefaultUVS& _) { _.closure_glyphs (unicodes, glyphset); })
1320 void collect_unicodes (hb_set_t *out) const
1322 for (const VariationSelectorRecord& _ : record)
1323 _.collect_unicodes (out, this);
1326 void collect_mapping (hb_set_t *unicodes, /* OUT */
1327 hb_map_t *mapping /* OUT */) const
1329 for (const VariationSelectorRecord& _ : record)
1330 _.collect_mapping (this, unicodes, mapping);
1333 bool sanitize (hb_sanitize_context_t *c) const
1335 TRACE_SANITIZE (this);
1336 return_trace (c->check_struct (this) &&
1337 record.sanitize (c, this));
1340 protected:
1341 HBUINT16 format; /* Format number is set to 14. */
1342 HBUINT32 length; /* Byte length of this subtable. */
1343 SortedArray32Of<VariationSelectorRecord>
1344 record; /* Variation selector records; sorted
1345 * in increasing order of `varSelector'. */
1346 public:
1347 DEFINE_SIZE_ARRAY (10, record);
1350 struct CmapSubtable
1352 /* Note: We intentionally do NOT implement subtable formats 2 and 8. */
1354 bool get_glyph (hb_codepoint_t codepoint,
1355 hb_codepoint_t *glyph) const
1357 switch (u.format) {
1358 case 0: return u.format0 .get_glyph (codepoint, glyph);
1359 case 4: return u.format4 .get_glyph (codepoint, glyph);
1360 case 6: return u.format6 .get_glyph (codepoint, glyph);
1361 case 10: return u.format10.get_glyph (codepoint, glyph);
1362 case 12: return u.format12.get_glyph (codepoint, glyph);
1363 case 13: return u.format13.get_glyph (codepoint, glyph);
1364 case 14:
1365 default: return false;
1368 void collect_unicodes (hb_set_t *out, unsigned int num_glyphs = UINT_MAX) const
1370 switch (u.format) {
1371 case 0: u.format0 .collect_unicodes (out); return;
1372 case 4: u.format4 .collect_unicodes (out); return;
1373 case 6: u.format6 .collect_unicodes (out); return;
1374 case 10: u.format10.collect_unicodes (out); return;
1375 case 12: u.format12.collect_unicodes (out, num_glyphs); return;
1376 case 13: u.format13.collect_unicodes (out, num_glyphs); return;
1377 case 14:
1378 default: return;
1382 void collect_mapping (hb_set_t *unicodes, /* OUT */
1383 hb_map_t *mapping, /* OUT */
1384 unsigned num_glyphs = UINT_MAX) const
1386 switch (u.format) {
1387 case 0: u.format0 .collect_mapping (unicodes, mapping); return;
1388 case 4: u.format4 .collect_mapping (unicodes, mapping); return;
1389 case 6: u.format6 .collect_mapping (unicodes, mapping); return;
1390 case 10: u.format10.collect_mapping (unicodes, mapping); return;
1391 case 12: u.format12.collect_mapping (unicodes, mapping, num_glyphs); return;
1392 case 13: u.format13.collect_mapping (unicodes, mapping, num_glyphs); return;
1393 case 14:
1394 default: return;
1398 unsigned get_language () const
1400 switch (u.format) {
1401 case 0: return u.format0 .get_language ();
1402 case 4: return u.format4 .get_language ();
1403 case 6: return u.format6 .get_language ();
1404 case 10: return u.format10.get_language ();
1405 case 12: return u.format12.get_language ();
1406 case 13: return u.format13.get_language ();
1407 case 14:
1408 default: return 0;
1412 template<typename Iterator,
1413 hb_requires (hb_is_iterator (Iterator))>
1414 void serialize (hb_serialize_context_t *c,
1415 Iterator it,
1416 unsigned format,
1417 const hb_subset_plan_t *plan,
1418 const void *base)
1420 switch (format) {
1421 case 4: return u.format4.serialize (c, it);
1422 case 12: return u.format12.serialize (c, it);
1423 case 14: return u.format14.serialize (c, &plan->unicodes, &plan->glyphs_requested, plan->glyph_map, base);
1424 default: return;
1428 bool sanitize (hb_sanitize_context_t *c) const
1430 TRACE_SANITIZE (this);
1431 if (!u.format.sanitize (c)) return_trace (false);
1432 hb_barrier ();
1433 switch (u.format) {
1434 case 0: return_trace (u.format0 .sanitize (c));
1435 case 4: return_trace (u.format4 .sanitize (c));
1436 case 6: return_trace (u.format6 .sanitize (c));
1437 case 10: return_trace (u.format10.sanitize (c));
1438 case 12: return_trace (u.format12.sanitize (c));
1439 case 13: return_trace (u.format13.sanitize (c));
1440 case 14: return_trace (u.format14.sanitize (c));
1441 default:return_trace (true);
1445 public:
1446 union {
1447 HBUINT16 format; /* Format identifier */
1448 CmapSubtableFormat0 format0;
1449 CmapSubtableFormat4 format4;
1450 CmapSubtableFormat6 format6;
1451 CmapSubtableFormat10 format10;
1452 CmapSubtableFormat12 format12;
1453 CmapSubtableFormat13 format13;
1454 CmapSubtableFormat14 format14;
1455 } u;
1456 public:
1457 DEFINE_SIZE_UNION (2, format);
1461 struct EncodingRecord
1463 int cmp (const EncodingRecord &other) const
1465 int ret;
1466 ret = platformID.cmp (other.platformID);
1467 if (ret) return ret;
1468 ret = encodingID.cmp (other.encodingID);
1469 if (ret) return ret;
1470 return 0;
1473 bool sanitize (hb_sanitize_context_t *c, const void *base) const
1475 TRACE_SANITIZE (this);
1476 return_trace (c->check_struct (this) &&
1477 subtable.sanitize (c, base));
1480 template<typename Iterator,
1481 hb_requires (hb_is_iterator (Iterator))>
1482 EncodingRecord* copy (hb_serialize_context_t *c,
1483 Iterator it,
1484 unsigned format,
1485 const void *base,
1486 const hb_subset_plan_t *plan,
1487 /* INOUT */ unsigned *objidx) const
1489 TRACE_SERIALIZE (this);
1490 auto snap = c->snapshot ();
1491 auto *out = c->embed (this);
1492 if (unlikely (!out)) return_trace (nullptr);
1493 out->subtable = 0;
1495 if (*objidx == 0)
1497 CmapSubtable *cmapsubtable = c->push<CmapSubtable> ();
1498 unsigned origin_length = c->length ();
1499 cmapsubtable->serialize (c, it, format, plan, &(base+subtable));
1500 if (c->length () - origin_length > 0) *objidx = c->pop_pack ();
1501 else c->pop_discard ();
1504 if (*objidx == 0)
1506 c->revert (snap);
1507 return_trace (nullptr);
1510 c->add_link (out->subtable, *objidx);
1511 return_trace (out);
1514 HBUINT16 platformID; /* Platform ID. */
1515 HBUINT16 encodingID; /* Platform-specific encoding ID. */
1516 Offset32To<CmapSubtable>
1517 subtable; /* Byte offset from beginning of table to the subtable for this encoding. */
1518 public:
1519 DEFINE_SIZE_STATIC (8);
1522 struct cmap;
1524 struct SubtableUnicodesCache {
1526 private:
1527 hb_blob_ptr_t<cmap> base_blob;
1528 const char* base;
1529 hb_hashmap_t<unsigned, hb::unique_ptr<hb_set_t>> cached_unicodes;
1531 public:
1533 static SubtableUnicodesCache* create (hb_blob_ptr_t<cmap> source_table)
1535 SubtableUnicodesCache* cache =
1536 (SubtableUnicodesCache*) hb_malloc (sizeof(SubtableUnicodesCache));
1537 new (cache) SubtableUnicodesCache (source_table);
1538 return cache;
1541 static void destroy (void* value) {
1542 if (!value) return;
1544 SubtableUnicodesCache* cache = (SubtableUnicodesCache*) value;
1545 cache->~SubtableUnicodesCache ();
1546 hb_free (cache);
1549 SubtableUnicodesCache(const void* cmap_base)
1550 : base_blob(),
1551 base ((const char*) cmap_base),
1552 cached_unicodes ()
1555 SubtableUnicodesCache(hb_blob_ptr_t<cmap> base_blob_)
1556 : base_blob(base_blob_),
1557 base ((const char *) base_blob.get()),
1558 cached_unicodes ()
1561 ~SubtableUnicodesCache()
1563 base_blob.destroy ();
1566 bool same_base(const void* other) const
1568 return other == (const void*) base;
1571 const hb_set_t* set_for (const EncodingRecord* record,
1572 SubtableUnicodesCache& mutable_cache) const
1574 if (cached_unicodes.has ((unsigned) ((const char *) record - base)))
1575 return cached_unicodes.get ((unsigned) ((const char *) record - base));
1577 return mutable_cache.set_for (record);
1580 const hb_set_t* set_for (const EncodingRecord* record)
1582 if (!cached_unicodes.has ((unsigned) ((const char *) record - base)))
1584 hb_set_t *s = hb_set_create ();
1585 if (unlikely (s->in_error ()))
1586 return hb_set_get_empty ();
1588 (base+record->subtable).collect_unicodes (s);
1590 if (unlikely (!cached_unicodes.set ((unsigned) ((const char *) record - base), hb::unique_ptr<hb_set_t> {s})))
1591 return hb_set_get_empty ();
1593 return s;
1595 return cached_unicodes.get ((unsigned) ((const char *) record - base));
1600 static inline uint_fast16_t
1601 _hb_symbol_pua_map (unsigned codepoint)
1603 if (codepoint <= 0x00FFu)
1605 /* For symbol-encoded OpenType fonts, we duplicate the
1606 * U+F000..F0FF range at U+0000..U+00FF. That's what
1607 * Windows seems to do, and that's hinted about at:
1608 * https://docs.microsoft.com/en-us/typography/opentype/spec/recom
1609 * under "Non-Standard (Symbol) Fonts". */
1610 return 0xF000u + codepoint;
1612 return 0;
1615 struct cmap
1617 static constexpr hb_tag_t tableTag = HB_OT_TAG_cmap;
1620 static SubtableUnicodesCache* create_filled_cache(hb_blob_ptr_t<cmap> source_table) {
1621 const cmap* cmap = source_table.get();
1622 auto it =
1623 + hb_iter (cmap->encodingRecord)
1624 | hb_filter ([&](const EncodingRecord& _) {
1625 return cmap::filter_encoding_records_for_subset (cmap, _);
1629 SubtableUnicodesCache* cache = SubtableUnicodesCache::create(source_table);
1630 for (const EncodingRecord& _ : it)
1631 cache->set_for(&_); // populate the cache for this encoding record.
1633 return cache;
1636 template<typename Iterator, typename EncodingRecIter,
1637 hb_requires (hb_is_iterator (EncodingRecIter))>
1638 bool serialize (hb_serialize_context_t *c,
1639 Iterator it,
1640 EncodingRecIter encodingrec_iter,
1641 const void *base,
1642 hb_subset_plan_t *plan,
1643 bool drop_format_4 = false)
1645 if (unlikely (!c->extend_min ((*this)))) return false;
1646 this->version = 0;
1648 unsigned format4objidx = 0, format12objidx = 0, format14objidx = 0;
1649 auto snap = c->snapshot ();
1651 SubtableUnicodesCache local_unicodes_cache (base);
1652 const SubtableUnicodesCache* unicodes_cache = &local_unicodes_cache;
1654 if (plan->accelerator &&
1655 plan->accelerator->cmap_cache &&
1656 plan->accelerator->cmap_cache->same_base (base))
1657 unicodes_cache = plan->accelerator->cmap_cache;
1659 for (const EncodingRecord& _ : encodingrec_iter)
1661 if (c->in_error ())
1662 return false;
1664 unsigned format = (base+_.subtable).u.format;
1665 if (format != 4 && format != 12 && format != 14) continue;
1667 const hb_set_t* unicodes_set = unicodes_cache->set_for (&_, local_unicodes_cache);
1669 if (!drop_format_4 && format == 4)
1671 c->copy (_, + it | hb_filter (*unicodes_set, hb_first), 4u, base, plan, &format4objidx);
1672 if (c->in_error () && c->only_overflow ())
1674 // cmap4 overflowed, reset and retry serialization without format 4 subtables.
1675 c->revert (snap);
1676 return serialize (c, it,
1677 encodingrec_iter,
1678 base,
1679 plan,
1680 true);
1684 else if (format == 12)
1686 if (_can_drop (_,
1687 *unicodes_set,
1688 base,
1689 *unicodes_cache,
1690 local_unicodes_cache,
1691 + it | hb_map (hb_first), encodingrec_iter))
1692 continue;
1693 c->copy (_, + it | hb_filter (*unicodes_set, hb_first), 12u, base, plan, &format12objidx);
1695 else if (format == 14) c->copy (_, it, 14u, base, plan, &format14objidx);
1697 c->check_assign(this->encodingRecord.len,
1698 (c->length () - cmap::min_size)/EncodingRecord::static_size,
1699 HB_SERIALIZE_ERROR_INT_OVERFLOW);
1701 // Fail if format 4 was dropped and there is no cmap12.
1702 return !drop_format_4 || format12objidx;
1705 template<typename Iterator, typename EncodingRecordIterator,
1706 hb_requires (hb_is_iterator (Iterator)),
1707 hb_requires (hb_is_iterator (EncodingRecordIterator))>
1708 bool _can_drop (const EncodingRecord& cmap12,
1709 const hb_set_t& cmap12_unicodes,
1710 const void* base,
1711 const SubtableUnicodesCache& unicodes_cache,
1712 SubtableUnicodesCache& local_unicodes_cache,
1713 Iterator subset_unicodes,
1714 EncodingRecordIterator encoding_records)
1716 for (auto cp : + subset_unicodes | hb_filter (cmap12_unicodes))
1718 if (cp >= 0x10000) return false;
1721 unsigned target_platform;
1722 unsigned target_encoding;
1723 unsigned target_language = (base+cmap12.subtable).get_language ();
1725 if (cmap12.platformID == 0 && cmap12.encodingID == 4)
1727 target_platform = 0;
1728 target_encoding = 3;
1729 } else if (cmap12.platformID == 3 && cmap12.encodingID == 10) {
1730 target_platform = 3;
1731 target_encoding = 1;
1732 } else {
1733 return false;
1736 for (const auto& _ : encoding_records)
1738 if (_.platformID != target_platform
1739 || _.encodingID != target_encoding
1740 || (base+_.subtable).get_language() != target_language)
1741 continue;
1743 const hb_set_t* sibling_unicodes = unicodes_cache.set_for (&_, local_unicodes_cache);
1745 auto cmap12 = + subset_unicodes | hb_filter (cmap12_unicodes);
1746 auto sibling = + subset_unicodes | hb_filter (*sibling_unicodes);
1747 for (; cmap12 && sibling; cmap12++, sibling++)
1749 unsigned a = *cmap12;
1750 unsigned b = *sibling;
1751 if (a != b) return false;
1754 return !cmap12 && !sibling;
1757 return false;
1760 void closure_glyphs (const hb_set_t *unicodes,
1761 hb_set_t *glyphset) const
1763 + hb_iter (encodingRecord)
1764 | hb_map (&EncodingRecord::subtable)
1765 | hb_map (hb_add (this))
1766 | hb_filter ([&] (const CmapSubtable& _) { return _.u.format == 14; })
1767 | hb_apply ([=] (const CmapSubtable& _) { _.u.format14.closure_glyphs (unicodes, glyphset); })
1771 bool subset (hb_subset_context_t *c) const
1773 TRACE_SUBSET (this);
1775 cmap *cmap_prime = c->serializer->start_embed<cmap> ();
1777 auto encodingrec_iter =
1778 + hb_iter (encodingRecord)
1779 | hb_filter ([&](const EncodingRecord& _) {
1780 return cmap::filter_encoding_records_for_subset (this, _);
1784 if (unlikely (!encodingrec_iter.len ())) return_trace (false);
1786 const EncodingRecord *unicode_bmp= nullptr, *unicode_ucs4 = nullptr, *ms_bmp = nullptr, *ms_ucs4 = nullptr;
1787 bool has_format12 = false;
1789 for (const EncodingRecord& _ : encodingrec_iter)
1791 unsigned format = (this + _.subtable).u.format;
1792 if (format == 12) has_format12 = true;
1794 const EncodingRecord *table = std::addressof (_);
1795 if (_.platformID == 0 && _.encodingID == 3) unicode_bmp = table;
1796 else if (_.platformID == 0 && _.encodingID == 4) unicode_ucs4 = table;
1797 else if (_.platformID == 3 && _.encodingID == 1) ms_bmp = table;
1798 else if (_.platformID == 3 && _.encodingID == 10) ms_ucs4 = table;
1801 if (unlikely (!has_format12 && !unicode_bmp && !ms_bmp)) return_trace (false);
1802 if (unlikely (has_format12 && (!unicode_ucs4 && !ms_ucs4))) return_trace (false);
1804 auto it =
1805 + c->plan->unicode_to_new_gid_list.iter ()
1806 | hb_filter ([&] (const hb_codepoint_pair_t _)
1807 { return (_.second != HB_MAP_VALUE_INVALID); })
1810 return_trace (cmap_prime->serialize (c->serializer,
1812 encodingrec_iter,
1813 this,
1814 c->plan));
1817 const CmapSubtable *find_best_subtable (bool *symbol = nullptr) const
1819 if (symbol) *symbol = false;
1821 const CmapSubtable *subtable;
1823 /* Symbol subtable.
1824 * Prefer symbol if available.
1825 * https://github.com/harfbuzz/harfbuzz/issues/1918 */
1826 if ((subtable = this->find_subtable (3, 0)))
1828 if (symbol) *symbol = true;
1829 return subtable;
1832 /* 32-bit subtables. */
1833 if ((subtable = this->find_subtable (3, 10))) return subtable;
1834 if ((subtable = this->find_subtable (0, 6))) return subtable;
1835 if ((subtable = this->find_subtable (0, 4))) return subtable;
1837 /* 16-bit subtables. */
1838 if ((subtable = this->find_subtable (3, 1))) return subtable;
1839 if ((subtable = this->find_subtable (0, 3))) return subtable;
1840 if ((subtable = this->find_subtable (0, 2))) return subtable;
1841 if ((subtable = this->find_subtable (0, 1))) return subtable;
1842 if ((subtable = this->find_subtable (0, 0))) return subtable;
1844 /* Meh. */
1845 return &Null (CmapSubtable);
1848 struct accelerator_t
1850 using cache_t = hb_cache_t<21, 16, 8, true>;
1852 accelerator_t (hb_face_t *face)
1854 this->table = hb_sanitize_context_t ().reference_table<cmap> (face);
1855 bool symbol;
1856 this->subtable = table->find_best_subtable (&symbol);
1857 this->subtable_uvs = &Null (CmapSubtableFormat14);
1859 const CmapSubtable *st = table->find_subtable (0, 5);
1860 if (st && st->u.format == 14)
1861 subtable_uvs = &st->u.format14;
1864 this->get_glyph_data = subtable;
1865 if (unlikely (symbol))
1867 switch ((unsigned) face->table.OS2->get_font_page ()) {
1868 case OS2::font_page_t::FONT_PAGE_NONE:
1869 this->get_glyph_funcZ = get_glyph_from_symbol<CmapSubtable, _hb_symbol_pua_map>;
1870 break;
1871 #ifndef HB_NO_OT_SHAPER_ARABIC_FALLBACK
1872 case OS2::font_page_t::FONT_PAGE_SIMP_ARABIC:
1873 this->get_glyph_funcZ = get_glyph_from_symbol<CmapSubtable, _hb_arabic_pua_simp_map>;
1874 break;
1875 case OS2::font_page_t::FONT_PAGE_TRAD_ARABIC:
1876 this->get_glyph_funcZ = get_glyph_from_symbol<CmapSubtable, _hb_arabic_pua_trad_map>;
1877 break;
1878 #endif
1879 default:
1880 this->get_glyph_funcZ = get_glyph_from<CmapSubtable>;
1881 break;
1884 else
1886 switch (subtable->u.format) {
1887 /* Accelerate format 4 and format 12. */
1888 default:
1889 this->get_glyph_funcZ = get_glyph_from<CmapSubtable>;
1890 break;
1891 case 12:
1892 this->get_glyph_funcZ = get_glyph_from<CmapSubtableFormat12>;
1893 break;
1894 case 4:
1896 this->format4_accel.init (&subtable->u.format4);
1897 this->get_glyph_data = &this->format4_accel;
1898 this->get_glyph_funcZ = this->format4_accel.get_glyph_func;
1899 break;
1904 ~accelerator_t () { this->table.destroy (); }
1906 inline bool _cached_get (hb_codepoint_t unicode,
1907 hb_codepoint_t *glyph,
1908 cache_t *cache) const
1910 unsigned v;
1911 if (cache && cache->get (unicode, &v))
1913 *glyph = v;
1914 return true;
1916 bool ret = this->get_glyph_funcZ (this->get_glyph_data, unicode, glyph);
1918 if (cache && ret)
1919 cache->set (unicode, *glyph);
1920 return ret;
1923 bool get_nominal_glyph (hb_codepoint_t unicode,
1924 hb_codepoint_t *glyph,
1925 cache_t *cache = nullptr) const
1927 if (unlikely (!this->get_glyph_funcZ)) return 0;
1928 return _cached_get (unicode, glyph, cache);
1931 unsigned int get_nominal_glyphs (unsigned int count,
1932 const hb_codepoint_t *first_unicode,
1933 unsigned int unicode_stride,
1934 hb_codepoint_t *first_glyph,
1935 unsigned int glyph_stride,
1936 cache_t *cache = nullptr) const
1938 if (unlikely (!this->get_glyph_funcZ)) return 0;
1940 unsigned int done;
1941 for (done = 0;
1942 done < count && _cached_get (*first_unicode, first_glyph, cache);
1943 done++)
1945 first_unicode = &StructAtOffsetUnaligned<hb_codepoint_t> (first_unicode, unicode_stride);
1946 first_glyph = &StructAtOffsetUnaligned<hb_codepoint_t> (first_glyph, glyph_stride);
1948 return done;
1951 bool get_variation_glyph (hb_codepoint_t unicode,
1952 hb_codepoint_t variation_selector,
1953 hb_codepoint_t *glyph,
1954 cache_t *cache = nullptr) const
1956 switch (this->subtable_uvs->get_glyph_variant (unicode,
1957 variation_selector,
1958 glyph))
1960 case GLYPH_VARIANT_NOT_FOUND: return false;
1961 case GLYPH_VARIANT_FOUND: return true;
1962 case GLYPH_VARIANT_USE_DEFAULT: break;
1965 return get_nominal_glyph (unicode, glyph, cache);
1968 void collect_unicodes (hb_set_t *out, unsigned int num_glyphs) const
1969 { subtable->collect_unicodes (out, num_glyphs); }
1970 void collect_mapping (hb_set_t *unicodes, hb_map_t *mapping,
1971 unsigned num_glyphs = UINT_MAX) const
1972 { subtable->collect_mapping (unicodes, mapping, num_glyphs); }
1973 void collect_variation_selectors (hb_set_t *out) const
1974 { subtable_uvs->collect_variation_selectors (out); }
1975 void collect_variation_unicodes (hb_codepoint_t variation_selector,
1976 hb_set_t *out) const
1977 { subtable_uvs->collect_variation_unicodes (variation_selector, out); }
1979 protected:
1980 typedef bool (*hb_cmap_get_glyph_func_t) (const void *obj,
1981 hb_codepoint_t codepoint,
1982 hb_codepoint_t *glyph);
1983 typedef uint_fast16_t (*hb_pua_remap_func_t) (unsigned);
1985 template <typename Type>
1986 HB_INTERNAL static bool get_glyph_from (const void *obj,
1987 hb_codepoint_t codepoint,
1988 hb_codepoint_t *glyph)
1990 const Type *typed_obj = (const Type *) obj;
1991 return typed_obj->get_glyph (codepoint, glyph);
1994 template <typename Type, hb_pua_remap_func_t remap>
1995 HB_INTERNAL static bool get_glyph_from_symbol (const void *obj,
1996 hb_codepoint_t codepoint,
1997 hb_codepoint_t *glyph)
1999 const Type *typed_obj = (const Type *) obj;
2000 if (likely (typed_obj->get_glyph (codepoint, glyph)))
2001 return true;
2003 if (hb_codepoint_t c = remap (codepoint))
2004 return typed_obj->get_glyph (c, glyph);
2006 return false;
2009 private:
2010 hb_nonnull_ptr_t<const CmapSubtable> subtable;
2011 hb_nonnull_ptr_t<const CmapSubtableFormat14> subtable_uvs;
2013 hb_cmap_get_glyph_func_t get_glyph_funcZ;
2014 const void *get_glyph_data;
2016 CmapSubtableFormat4::accelerator_t format4_accel;
2018 public:
2019 hb_blob_ptr_t<cmap> table;
2022 protected:
2024 const CmapSubtable *find_subtable (unsigned int platform_id,
2025 unsigned int encoding_id) const
2027 EncodingRecord key;
2028 key.platformID = platform_id;
2029 key.encodingID = encoding_id;
2031 const EncodingRecord &result = encodingRecord.bsearch (key);
2032 if (!result.subtable)
2033 return nullptr;
2035 return &(this+result.subtable);
2038 const EncodingRecord *find_encodingrec (unsigned int platform_id,
2039 unsigned int encoding_id) const
2041 EncodingRecord key;
2042 key.platformID = platform_id;
2043 key.encodingID = encoding_id;
2045 return encodingRecord.as_array ().bsearch (key);
2048 bool find_subtable (unsigned format) const
2050 auto it =
2051 + hb_iter (encodingRecord)
2052 | hb_map (&EncodingRecord::subtable)
2053 | hb_map (hb_add (this))
2054 | hb_filter ([&] (const CmapSubtable& _) { return _.u.format == format; })
2057 return it.len ();
2060 public:
2062 bool sanitize (hb_sanitize_context_t *c) const
2064 TRACE_SANITIZE (this);
2065 return_trace (c->check_struct (this) &&
2066 hb_barrier () &&
2067 likely (version == 0) &&
2068 encodingRecord.sanitize (c, this));
2071 private:
2073 static bool filter_encoding_records_for_subset(const cmap* cmap,
2074 const EncodingRecord& _)
2076 return
2077 (_.platformID == 0 && _.encodingID == 3) ||
2078 (_.platformID == 0 && _.encodingID == 4) ||
2079 (_.platformID == 3 && _.encodingID == 1) ||
2080 (_.platformID == 3 && _.encodingID == 10) ||
2081 (cmap + _.subtable).u.format == 14;
2084 protected:
2085 HBUINT16 version; /* Table version number (0). */
2086 SortedArray16Of<EncodingRecord>
2087 encodingRecord; /* Encoding tables. */
2088 public:
2089 DEFINE_SIZE_ARRAY (4, encodingRecord);
2092 struct cmap_accelerator_t : cmap::accelerator_t {
2093 cmap_accelerator_t (hb_face_t *face) : cmap::accelerator_t (face) {}
2096 } /* namespace OT */
2099 #endif /* HB_OT_CMAP_TABLE_HH */