diff options
Diffstat (limited to 'src/3rdparty/harfbuzz-ng/src/hb-ot-cmap-table.hh')
| -rw-r--r-- | src/3rdparty/harfbuzz-ng/src/hb-ot-cmap-table.hh | 960 |
1 files changed, 674 insertions, 286 deletions
diff --git a/src/3rdparty/harfbuzz-ng/src/hb-ot-cmap-table.hh b/src/3rdparty/harfbuzz-ng/src/hb-ot-cmap-table.hh index 878e02ff17..e2e2581855 100644 --- a/src/3rdparty/harfbuzz-ng/src/hb-ot-cmap-table.hh +++ b/src/3rdparty/harfbuzz-ng/src/hb-ot-cmap-table.hh @@ -27,8 +27,11 @@ #ifndef HB_OT_CMAP_TABLE_HH #define HB_OT_CMAP_TABLE_HH +#include "hb-ot-os2-table.hh" +#include "hb-ot-shaper-arabic-pua.hh" #include "hb-open-type.hh" #include "hb-set.hh" +#include "hb-cache.hh" /* * cmap -- Character to Glyph Index Mapping @@ -44,11 +47,17 @@ struct CmapSubtableFormat0 bool get_glyph (hb_codepoint_t codepoint, hb_codepoint_t *glyph) const { hb_codepoint_t gid = codepoint < 256 ? glyphIdArray[codepoint] : 0; - if (!gid) + if (unlikely (!gid)) return false; *glyph = gid; return true; } + + unsigned get_language () const + { + return language; + } + void collect_unicodes (hb_set_t *out) const { for (unsigned int i = 0; i < 256; i++) @@ -87,154 +96,224 @@ struct CmapSubtableFormat0 struct CmapSubtableFormat4 { + template<typename Iterator, + typename Writer, hb_requires (hb_is_iterator (Iterator))> - HBUINT16* serialize_endcode_array (hb_serialize_context_t *c, - Iterator it) + void to_ranges (Iterator it, Writer& range_writer) { - HBUINT16 *endCode = c->start_embed<HBUINT16> (); - hb_codepoint_t prev_endcp = 0xFFFF; + hb_codepoint_t start_cp = 0, prev_run_start_cp = 0, run_start_cp = 0, end_cp = 0, last_gid = 0; + int run_length = 0 , delta = 0, prev_delta = 0; - for (const auto& _ : +it) - { - if (prev_endcp != 0xFFFF && prev_endcp + 1u != _.first) + enum { + FIRST_SUB_RANGE, + FOLLOWING_SUB_RANGE, + } mode; + + while (it) { + // Start a new range { - HBUINT16 end_code; - end_code = prev_endcp; - c->copy<HBUINT16> (end_code); + const auto& pair = *it; + start_cp = pair.first; + prev_run_start_cp = start_cp; + run_start_cp = start_cp; + end_cp = start_cp; + last_gid = pair.second; + run_length = 1; + prev_delta = 0; } - prev_endcp = _.first; - } - { - // last endCode - HBUINT16 endcode; - endcode = prev_endcp; - if (unlikely (!c->copy<HBUINT16> (endcode))) return nullptr; - // There must be a final entry with end_code == 0xFFFF. - if (prev_endcp != 0xFFFF) - { - HBUINT16 finalcode; - finalcode = 0xFFFF; - if (unlikely (!c->copy<HBUINT16> (finalcode))) return nullptr; + delta = last_gid - start_cp; + mode = FIRST_SUB_RANGE; + it++; + + while (it) { + // Process range + const auto& pair = *it; + hb_codepoint_t next_cp = pair.first; + hb_codepoint_t next_gid = pair.second; + if (next_cp != end_cp + 1) { + // Current range is over, stop processing. + break; + } + + if (next_gid == last_gid + 1) { + // The current run continues. + end_cp = next_cp; + run_length++; + last_gid = next_gid; + it++; + continue; + } + + // A new run is starting, decide if we want to commit the current run. + int split_cost = (mode == FIRST_SUB_RANGE) ? 8 : 16; + int run_cost = run_length * 2; + if (run_cost >= split_cost) { + commit_current_range(start_cp, + prev_run_start_cp, + run_start_cp, + end_cp, + delta, + prev_delta, + split_cost, + range_writer); + start_cp = next_cp; + } + + // Start the new run + mode = FOLLOWING_SUB_RANGE; + prev_run_start_cp = run_start_cp; + run_start_cp = next_cp; + end_cp = next_cp; + prev_delta = delta; + delta = next_gid - run_start_cp; + run_length = 1; + last_gid = next_gid; + it++; } + + // Finalize range + commit_current_range (start_cp, + prev_run_start_cp, + run_start_cp, + end_cp, + delta, + prev_delta, + 8, + range_writer); } - return endCode; + if (likely (end_cp != 0xFFFF)) { + range_writer (0xFFFF, 0xFFFF, 1); + } } - template<typename Iterator, - hb_requires (hb_is_iterator (Iterator))> - HBUINT16* serialize_startcode_array (hb_serialize_context_t *c, - Iterator it) - { - HBUINT16 *startCode = c->start_embed<HBUINT16> (); - hb_codepoint_t prev_cp = 0xFFFF; - - for (const auto& _ : +it) - { - if (prev_cp == 0xFFFF || prev_cp + 1u != _.first) - { - HBUINT16 start_code; - start_code = _.first; - c->copy<HBUINT16> (start_code); + /* + * Writes the current range as either one or two ranges depending on what is most efficient. + */ + template<typename Writer> + void commit_current_range (hb_codepoint_t start, + hb_codepoint_t prev_run_start, + hb_codepoint_t run_start, + hb_codepoint_t end, + int run_delta, + int previous_run_delta, + int split_cost, + Writer& range_writer) { + bool should_split = false; + if (start < run_start && run_start < end) { + int run_cost = (end - run_start + 1) * 2; + if (run_cost >= split_cost) { + should_split = true; } + } - prev_cp = _.first; + // TODO(grieger): handle case where delta is legitimately 0, mark range offset array instead? + if (should_split) { + if (start == prev_run_start) + range_writer (start, run_start - 1, previous_run_delta); + else + range_writer (start, run_start - 1, 0); + range_writer (run_start, end, run_delta); + return; } - // There must be a final entry with end_code == 0xFFFF. - if (it.len () == 0 || prev_cp != 0xFFFF) - { - HBUINT16 finalcode; - finalcode = 0xFFFF; - if (unlikely (!c->copy<HBUINT16> (finalcode))) return nullptr; + + if (start == run_start) { + // Range is only a run + range_writer (start, end, run_delta); + return; } - return startCode; + // Write only a single non-run range. + range_writer (start, end, 0); } template<typename Iterator, hb_requires (hb_is_iterator (Iterator))> - HBINT16* serialize_idDelta_array (hb_serialize_context_t *c, - Iterator it, - HBUINT16 *endCode, - HBUINT16 *startCode, - unsigned segcount) - { - unsigned i = 0; - hb_codepoint_t last_gid = 0, start_gid = 0, last_cp = 0xFFFF; - bool use_delta = true; - - HBINT16 *idDelta = c->start_embed<HBINT16> (); - if ((char *)idDelta - (char *)startCode != (int) segcount * (int) HBINT16::static_size) - return nullptr; - - for (const auto& _ : +it) - { - if (_.first == startCode[i]) - { - use_delta = true; - start_gid = _.second; + unsigned serialize_find_segcount (Iterator it) { + struct Counter { + unsigned segcount = 0; + + void operator() (hb_codepoint_t start, + hb_codepoint_t end, + int delta) { + segcount++; } - else if (_.second != last_gid + 1) use_delta = false; + } counter; + + to_ranges (+it, counter); + return counter.segcount; + } - if (_.first == endCode[i]) - { - HBINT16 delta; - if (use_delta) delta = (int)start_gid - (int)startCode[i]; - else delta = 0; - c->copy<HBINT16> (delta); - i++; + template<typename Iterator, + hb_requires (hb_is_iterator (Iterator))> + bool serialize_start_end_delta_arrays (hb_serialize_context_t *c, + Iterator it, + int segcount) + { + struct Writer { + hb_serialize_context_t *serializer_; + HBUINT16* end_code_; + HBUINT16* start_code_; + HBINT16* id_delta_; + int index_; + + Writer(hb_serialize_context_t *serializer) + : serializer_(serializer), + end_code_(nullptr), + start_code_(nullptr), + id_delta_(nullptr), + index_ (0) {} + void operator() (hb_codepoint_t start, + hb_codepoint_t end, + int delta) { + start_code_[index_] = start; + end_code_[index_] = end; + id_delta_[index_] = delta; + index_++; } + } writer(c); - last_gid = _.second; - last_cp = _.first; - } + writer.end_code_ = c->allocate_size<HBUINT16> (HBUINT16::static_size * segcount, false); + (void) c->allocate_size<HBUINT16> (2); // padding + writer.start_code_ = c->allocate_size<HBUINT16> (HBUINT16::static_size * segcount, false); + writer.id_delta_ = c->allocate_size<HBINT16> (HBINT16::static_size * segcount, false); - if (it.len () == 0 || last_cp != 0xFFFF) - { - HBINT16 delta; - delta = 1; - if (unlikely (!c->copy<HBINT16> (delta))) return nullptr; - } + if (unlikely (!writer.end_code_ || !writer.start_code_ || !writer.id_delta_)) return false; - return idDelta; + to_ranges (+it, writer); + return true; } template<typename Iterator, - hb_requires (hb_is_iterator (Iterator))> + hb_requires (hb_is_iterator (Iterator))> HBUINT16* serialize_rangeoffset_glyid (hb_serialize_context_t *c, - Iterator it, + Iterator it, HBUINT16 *endCode, HBUINT16 *startCode, HBINT16 *idDelta, unsigned segcount) { + hb_map_t cp_to_gid { it }; + HBUINT16 *idRangeOffset = c->allocate_size<HBUINT16> (HBUINT16::static_size * segcount); if (unlikely (!c->check_success (idRangeOffset))) return nullptr; if (unlikely ((char *)idRangeOffset - (char *)idDelta != (int) segcount * (int) HBINT16::static_size)) return nullptr; - + hb_range (segcount) - | hb_filter ([&] (const unsigned _) { return idDelta[_] == 0; }) - | hb_apply ([&] (const unsigned i) - { - idRangeOffset[i] = 2 * (c->start_embed<HBUINT16> () - idRangeOffset - i); - - + it - | hb_filter ([&] (const hb_item_type<Iterator> _) { return _.first >= startCode[i] && _.first <= endCode[i]; }) - | hb_apply ([&] (const hb_item_type<Iterator> _) - { - HBUINT16 glyID; - glyID = _.second; - c->copy<HBUINT16> (glyID); - }) - ; - - - }) - ; + for (unsigned i : + hb_range (segcount) + | hb_filter ([&] (const unsigned _) { return idDelta[_] == 0; })) + { + idRangeOffset[i] = 2 * (c->start_embed<HBUINT16> () - idRangeOffset - i); + for (hb_codepoint_t cp = startCode[i]; cp <= endCode[i]; cp++) + { + HBUINT16 gid; + gid = cp_to_gid[cp]; + c->copy<HBUINT16> (gid); + } + } return idRangeOffset; } @@ -246,37 +325,48 @@ struct CmapSubtableFormat4 { auto format4_iter = + it - | hb_filter ([&] (const hb_pair_t<hb_codepoint_t, hb_codepoint_t> _) + | hb_filter ([&] (const hb_codepoint_pair_t _) { return _.first <= 0xFFFF; }) ; - if (format4_iter.len () == 0) return; + if (!format4_iter) return; unsigned table_initpos = c->length (); - if (unlikely (!c->extend_min (*this))) return; + if (unlikely (!c->extend_min (this))) return; this->format = 4; - //serialize endCode[] - HBUINT16 *endCode = serialize_endcode_array (c, format4_iter); - if (unlikely (!endCode)) return; + hb_vector_t<hb_codepoint_pair_t> cp_to_gid { + format4_iter + }; - unsigned segcount = (c->length () - min_size) / HBUINT16::static_size; - - // 2 bytes of padding. - if (unlikely (!c->allocate_size<HBUINT16> (HBUINT16::static_size))) return; // 2 bytes of padding. - - // serialize startCode[] - HBUINT16 *startCode = serialize_startcode_array (c, format4_iter); - if (unlikely (!startCode)) return; + //serialize endCode[], startCode[], idDelta[] + HBUINT16* endCode = c->start_embed<HBUINT16> (); + unsigned segcount = serialize_find_segcount (cp_to_gid.iter()); + if (unlikely (!serialize_start_end_delta_arrays (c, cp_to_gid.iter(), segcount))) + return; - //serialize idDelta[] - HBINT16 *idDelta = serialize_idDelta_array (c, format4_iter, endCode, startCode, segcount); - if (unlikely (!idDelta)) return; + HBUINT16 *startCode = endCode + segcount + 1; + HBINT16 *idDelta = ((HBINT16*)startCode) + segcount; - HBUINT16 *idRangeOffset = serialize_rangeoffset_glyid (c, format4_iter, endCode, startCode, idDelta, segcount); + HBUINT16 *idRangeOffset = serialize_rangeoffset_glyid (c, + cp_to_gid.iter (), + endCode, + startCode, + idDelta, + segcount); if (unlikely (!c->check_success (idRangeOffset))) return; - if (unlikely (!c->check_assign(this->length, c->length () - table_initpos))) return; + this->length = c->length () - table_initpos; + if ((long long) this->length != (long long) c->length () - table_initpos) + { + // Length overflowed. Discard the current object before setting the error condition, otherwise + // discard is a noop which prevents the higher level code from reverting the serializer to the + // pre-error state in cmap4 overflow handling code. + c->pop_discard (); + c->err (HB_SERIALIZE_ERROR_INT_OVERFLOW); + return; + } + this->segCountX2 = segcount * 2; this->entrySelector = hb_max (1u, hb_bit_storage (segcount)) - 1; this->searchRange = 2 * (1u << this->entrySelector); @@ -285,11 +375,15 @@ struct CmapSubtableFormat4 : 0; } + unsigned get_language () const + { + return language; + } + struct accelerator_t { accelerator_t () {} accelerator_t (const CmapSubtableFormat4 *subtable) { init (subtable); } - ~accelerator_t () { fini (); } void init (const CmapSubtableFormat4 *subtable) { @@ -301,7 +395,6 @@ struct CmapSubtableFormat4 glyphIdArray = idRangeOffset + segCount; glyphIdArrayLength = (subtable->length - 16 - 8 * segCount) / 2; } - void fini () {} bool get_glyph (hb_codepoint_t codepoint, hb_codepoint_t *glyph) const { @@ -311,7 +404,7 @@ struct CmapSubtableFormat4 unsigned distance) const { if (k > last) return +1; - if (k < (&last)[distance]) return -1; + if (k < (&last)[distance]/*first*/) return -1; return 0; } HBUINT16 last; @@ -320,10 +413,10 @@ struct CmapSubtableFormat4 const HBUINT16 *found = hb_bsearch (codepoint, this->endCount, this->segCount, - 2, + sizeof (CustomRange), _hb_cmp_method<hb_codepoint_t, CustomRange, unsigned>, this->segCount + 1); - if (!found) + if (unlikely (!found)) return false; unsigned int i = found - endCount; @@ -343,7 +436,7 @@ struct CmapSubtableFormat4 gid += this->idDelta[i]; } gid &= 0xFFFFu; - if (!gid) + if (unlikely (!gid)) return false; *glyph = gid; return true; @@ -362,14 +455,14 @@ struct CmapSubtableFormat4 hb_codepoint_t start = this->startCount[i]; hb_codepoint_t end = this->endCount[i]; unsigned int rangeOffset = this->idRangeOffset[i]; + out->add_range(start, end); if (rangeOffset == 0) { for (hb_codepoint_t codepoint = start; codepoint <= end; codepoint++) { hb_codepoint_t gid = (codepoint + this->idDelta[i]) & 0xFFFFu; if (unlikely (!gid)) - continue; - out->add (codepoint); + out->del(codepoint); } } else @@ -378,11 +471,13 @@ struct CmapSubtableFormat4 { unsigned int index = rangeOffset / 2 + (codepoint - this->startCount[i]) + i - this->segCount; if (unlikely (index >= this->glyphIdArrayLength)) + { + out->del_range (codepoint, end); break; + } hb_codepoint_t gid = this->glyphIdArray[index]; if (unlikely (!gid)) - continue; - out->add (codepoint); + out->del(codepoint); } } } @@ -391,6 +486,8 @@ struct CmapSubtableFormat4 void collect_mapping (hb_set_t *unicodes, /* OUT */ hb_map_t *mapping /* OUT */) const { + // TODO(grieger): optimize similar to collect_unicodes + // (ie. use add_range()) unsigned count = this->segCount; if (count && this->startCount[count - 1] == 0xFFFFu) count--; /* Skip sentinel segment. */ @@ -459,6 +556,7 @@ struct CmapSubtableFormat4 TRACE_SANITIZE (this); if (unlikely (!c->check_struct (this))) return_trace (false); + hb_barrier (); if (unlikely (!c->check_range (this, length))) { @@ -542,11 +640,17 @@ struct CmapSubtableTrimmed { /* Rely on our implicit array bound-checking. */ hb_codepoint_t gid = glyphIdArray[codepoint - startCharCode]; - if (!gid) + if (unlikely (!gid)) return false; *glyph = gid; return true; } + + unsigned get_language () const + { + return language; + } + void collect_unicodes (hb_set_t *out) const { hb_codepoint_t start = startCharCode; @@ -582,7 +686,7 @@ struct CmapSubtableTrimmed UINT length; /* Byte length of this subtable. */ UINT language; /* Ignore. */ UINT startCharCode; /* First character code covered. */ - ArrayOf<HBGlyphID, UINT> + ArrayOf<HBGlyphID16, UINT> glyphIdArray; /* Array of glyph index values for character * codes in the range. */ public: @@ -590,7 +694,7 @@ struct CmapSubtableTrimmed }; struct CmapSubtableFormat6 : CmapSubtableTrimmed<HBUINT16> {}; -struct CmapSubtableFormat10 : CmapSubtableTrimmed<HBUINT32 > {}; +struct CmapSubtableFormat10 : CmapSubtableTrimmed<HBUINT32> {}; template <typename T> struct CmapSubtableLongSegmented @@ -600,12 +704,17 @@ struct CmapSubtableLongSegmented bool get_glyph (hb_codepoint_t codepoint, hb_codepoint_t *glyph) const { hb_codepoint_t gid = T::group_get_glyph (groups.bsearch (codepoint), codepoint); - if (!gid) + if (unlikely (!gid)) return false; *glyph = gid; return true; } + unsigned get_language () const + { + return language; + } + void collect_unicodes (hb_set_t *out, unsigned int num_glyphs) const { for (unsigned int i = 0; i < this->groups.len; i++) @@ -625,7 +734,7 @@ struct CmapSubtableLongSegmented if (unlikely ((unsigned int) (gid + end - start) >= num_glyphs)) end = start + (hb_codepoint_t) num_glyphs - gid; - out->add_range (start, end); + out->add_range (start, hb_min (end, 0x10FFFFu)); } } @@ -633,16 +742,24 @@ struct CmapSubtableLongSegmented hb_map_t *mapping, /* OUT */ unsigned num_glyphs) const { - for (unsigned i = 0; i < this->groups.len; i++) + hb_codepoint_t last_end = 0; + unsigned count = this->groups.len; + for (unsigned i = 0; i < count; i++) { - hb_codepoint_t start = this->groups[i].startCharCode; - hb_codepoint_t end = hb_min ((hb_codepoint_t) this->groups[i].endCharCode, + hb_codepoint_t start = this->groups.arrayZ[i].startCharCode; + hb_codepoint_t end = hb_min ((hb_codepoint_t) this->groups.arrayZ[i].endCharCode, (hb_codepoint_t) HB_UNICODE_MAX); - hb_codepoint_t gid = this->groups[i].glyphID; + if (unlikely (start > end || start < last_end)) { + // Range is not in order and is invalid, skip it. + continue; + } + last_end = end; + + + hb_codepoint_t gid = this->groups.arrayZ[i].glyphID; if (!gid) { - /* Intention is: if (hb_is_same (T, CmapSubtableFormat13)) continue; */ - if (! T::group_get_glyph (this->groups[i], end)) continue; + if (T::formatNumber == 13) continue; start++; gid++; } @@ -650,11 +767,13 @@ struct CmapSubtableLongSegmented if (unlikely ((unsigned int) (gid + end - start) >= num_glyphs)) end = start + (hb_codepoint_t) num_glyphs - gid; + mapping->alloc (mapping->get_population () + end - start + 1); + + unicodes->add_range (start, end); for (unsigned cp = start; cp <= end; cp++) { - unicodes->add (cp); mapping->set (cp, gid); - gid++; + gid += T::increment; } } } @@ -670,7 +789,7 @@ struct CmapSubtableLongSegmented HBUINT16 reserved; /* Reserved; set to 0. */ HBUINT32 length; /* Byte length of this subtable. */ HBUINT32 language; /* Ignore. */ - SortedArrayOf<CmapSubtableLongGroup, HBUINT32> + SortedArray32Of<CmapSubtableLongGroup> groups; /* Groupings. */ public: DEFINE_SIZE_ARRAY (16, groups); @@ -678,6 +797,9 @@ struct CmapSubtableLongSegmented struct CmapSubtableFormat12 : CmapSubtableLongSegmented<CmapSubtableFormat12> { + static constexpr int increment = 1; + static constexpr int formatNumber = 12; + static hb_codepoint_t group_get_glyph (const CmapSubtableLongGroup &group, hb_codepoint_t u) { return likely (group.startCharCode <= group.endCharCode) ? @@ -689,16 +811,16 @@ struct CmapSubtableFormat12 : CmapSubtableLongSegmented<CmapSubtableFormat12> void serialize (hb_serialize_context_t *c, Iterator it) { - if (it.len () == 0) return; + if (!it) return; unsigned table_initpos = c->length (); - if (unlikely (!c->extend_min (*this))) return; + if (unlikely (!c->extend_min (this))) return; - hb_codepoint_t startCharCode = 0xFFFF, endCharCode = 0xFFFF; + hb_codepoint_t startCharCode = (hb_codepoint_t) -1, endCharCode = (hb_codepoint_t) -1; hb_codepoint_t glyphID = 0; for (const auto& _ : +it) { - if (startCharCode == 0xFFFF) + if (startCharCode == (hb_codepoint_t) -1) { startCharCode = _.first; endCharCode = _.first; @@ -729,7 +851,7 @@ struct CmapSubtableFormat12 : CmapSubtableLongSegmented<CmapSubtableFormat12> this->format = 12; this->reserved = 0; this->length = c->length () - table_initpos; - this->groups.len = (this->length - min_size)/CmapSubtableLongGroup::static_size; + this->groups.len = (this->length - min_size) / CmapSubtableLongGroup::static_size; } static size_t get_sub_table_size (const hb_sorted_vector_t<CmapSubtableLongGroup> &groups_data) @@ -750,6 +872,9 @@ struct CmapSubtableFormat12 : CmapSubtableLongSegmented<CmapSubtableFormat12> struct CmapSubtableFormat13 : CmapSubtableLongSegmented<CmapSubtableFormat13> { + static constexpr int increment = 0; + static constexpr int formatNumber = 13; + static hb_codepoint_t group_get_glyph (const CmapSubtableLongGroup &group, hb_codepoint_t u HB_UNUSED) { return group.glyphID; } @@ -784,7 +909,7 @@ struct UnicodeValueRange DEFINE_SIZE_STATIC (4); }; -struct DefaultUVS : SortedArrayOf<UnicodeValueRange, HBUINT32> +struct DefaultUVS : SortedArray32Of<UnicodeValueRange> { void collect_unicodes (hb_set_t *out) const { @@ -801,8 +926,7 @@ struct DefaultUVS : SortedArrayOf<UnicodeValueRange, HBUINT32> DefaultUVS* copy (hb_serialize_context_t *c, const hb_set_t *unicodes) const { - DefaultUVS *out = c->start_embed<DefaultUVS> (); - if (unlikely (!out)) return nullptr; + auto *out = c->start_embed<DefaultUVS> (); auto snap = c->snapshot (); HBUINT32 len; @@ -810,37 +934,74 @@ struct DefaultUVS : SortedArrayOf<UnicodeValueRange, HBUINT32> if (unlikely (!c->copy<HBUINT32> (len))) return nullptr; unsigned init_len = c->length (); - hb_codepoint_t lastCode = HB_MAP_VALUE_INVALID; - int count = -1; - - for (const UnicodeValueRange& _ : as_array ()) + if (this->len > unicodes->get_population () * hb_bit_storage ((unsigned) this->len)) { - for (const unsigned addcnt : hb_range ((unsigned) _.additionalCount + 1)) + hb_codepoint_t start = HB_SET_VALUE_INVALID; + hb_codepoint_t end = HB_SET_VALUE_INVALID; + + for (auto u : *unicodes) { - unsigned curEntry = (unsigned) _.startUnicodeValue + addcnt; - if (!unicodes->has (curEntry)) continue; - count += 1; - if (lastCode == HB_MAP_VALUE_INVALID) - lastCode = curEntry; - else if (lastCode + count != curEntry) + if (!as_array ().bsearch (u)) + continue; + if (start == HB_SET_VALUE_INVALID) { + start = u; + end = start - 1; + } + if (end + 1 != u || end - start == 255) + { UnicodeValueRange rec; - rec.startUnicodeValue = lastCode; - rec.additionalCount = count - 1; + rec.startUnicodeValue = start; + rec.additionalCount = end - start; c->copy<UnicodeValueRange> (rec); - - lastCode = curEntry; - count = 0; + start = u; } + end = u; + } + if (start != HB_SET_VALUE_INVALID) + { + UnicodeValueRange rec; + rec.startUnicodeValue = start; + rec.additionalCount = end - start; + c->copy<UnicodeValueRange> (rec); } - } - if (lastCode != HB_MAP_VALUE_INVALID) + } + else { - UnicodeValueRange rec; - rec.startUnicodeValue = lastCode; - rec.additionalCount = count; - c->copy<UnicodeValueRange> (rec); + hb_codepoint_t lastCode = HB_SET_VALUE_INVALID; + int count = -1; + + for (const UnicodeValueRange& _ : *this) + { + hb_codepoint_t curEntry = (hb_codepoint_t) (_.startUnicodeValue - 1); + hb_codepoint_t end = curEntry + _.additionalCount + 2; + + for (; unicodes->next (&curEntry) && curEntry < end;) + { + count += 1; + if (lastCode == HB_SET_VALUE_INVALID) + lastCode = curEntry; + else if (lastCode + count != curEntry) + { + UnicodeValueRange rec; + rec.startUnicodeValue = lastCode; + rec.additionalCount = count - 1; + c->copy<UnicodeValueRange> (rec); + + lastCode = curEntry; + count = 0; + } + } + } + + if (lastCode != HB_MAP_VALUE_INVALID) + { + UnicodeValueRange rec; + rec.startUnicodeValue = lastCode; + rec.additionalCount = count; + c->copy<UnicodeValueRange> (rec); + } } if (c->length () - init_len == 0) @@ -850,7 +1011,9 @@ struct DefaultUVS : SortedArrayOf<UnicodeValueRange, HBUINT32> } else { - if (unlikely (!c->check_assign (out->len, (c->length () - init_len) / UnicodeValueRange::static_size))) return nullptr; + if (unlikely (!c->check_assign (out->len, + (c->length () - init_len) / UnicodeValueRange::static_size, + HB_SERIALIZE_ERROR_INT_OVERFLOW))) return nullptr; return out; } } @@ -871,28 +1034,26 @@ struct UVSMapping } HBUINT24 unicodeValue; /* Base Unicode value of the UVS */ - HBGlyphID glyphID; /* Glyph ID of the UVS */ + HBGlyphID16 glyphID; /* Glyph ID of the UVS */ public: DEFINE_SIZE_STATIC (5); }; -struct NonDefaultUVS : SortedArrayOf<UVSMapping, HBUINT32> +struct NonDefaultUVS : SortedArray32Of<UVSMapping> { void collect_unicodes (hb_set_t *out) const { - unsigned int count = len; - for (unsigned int i = 0; i < count; i++) - out->add (arrayZ[i].unicodeValue); + for (const auto& a : as_array ()) + out->add (a.unicodeValue); } void collect_mapping (hb_set_t *unicodes, /* OUT */ hb_map_t *mapping /* OUT */) const { - unsigned count = len; - for (unsigned i = 0; i < count; i++) + for (const auto& a : as_array ()) { - hb_codepoint_t unicode = arrayZ[i].unicodeValue; - hb_codepoint_t glyphid = arrayZ[i].glyphID; + hb_codepoint_t unicode = a.unicodeValue; + hb_codepoint_t glyphid = a.glyphID; unicodes->add (unicode); mapping->set (unicode, glyphid); } @@ -913,9 +1074,7 @@ struct NonDefaultUVS : SortedArrayOf<UVSMapping, HBUINT32> const hb_set_t *glyphs_requested, const hb_map_t *glyph_map) const { - NonDefaultUVS *out = c->start_embed<NonDefaultUVS> (); - if (unlikely (!out)) return nullptr; - + auto *out = c->start_embed<NonDefaultUVS> (); auto it = + as_array () | hb_filter ([&] (const UVSMapping& _) @@ -1041,9 +1200,9 @@ struct VariationSelectorRecord } HBUINT24 varSelector; /* Variation selector. */ - LOffsetTo<DefaultUVS> + Offset32To<DefaultUVS> defaultUVS; /* Offset to Default UVS Table. May be 0. */ - LOffsetTo<NonDefaultUVS> + Offset32To<NonDefaultUVS> nonDefaultUVS; /* Offset to Non-Default UVS Table. May be 0. */ public: DEFINE_SIZE_STATIC (11); @@ -1058,9 +1217,8 @@ struct CmapSubtableFormat14 void collect_variation_selectors (hb_set_t *out) const { - unsigned int count = record.len; - for (unsigned int i = 0; i < count; i++) - out->add (record.arrayZ[i].varSelector); + for (const auto& a : record.as_array ()) + out->add (a.varSelector); } void collect_variation_unicodes (hb_codepoint_t variation_selector, hb_set_t *out) const @@ -1076,7 +1234,7 @@ struct CmapSubtableFormat14 unsigned table_initpos = c->length (); const char* init_tail = c->tail; - if (unlikely (!c->extend_min (*this))) return; + if (unlikely (!c->extend_min (this))) return; this->format = 14; auto src_tbl = reinterpret_cast<const CmapSubtableFormat14*> (base); @@ -1112,10 +1270,12 @@ struct CmapSubtableFormat14 return; int tail_len = init_tail - c->tail; - c->check_assign (this->length, c->length () - table_initpos + tail_len); + c->check_assign (this->length, c->length () - table_initpos + tail_len, + HB_SERIALIZE_ERROR_INT_OVERFLOW); c->check_assign (this->record.len, (c->length () - table_initpos - CmapSubtableFormat14::min_size) / - VariationSelectorRecord::static_size); + VariationSelectorRecord::static_size, + HB_SERIALIZE_ERROR_INT_OVERFLOW); /* Correct the incorrect write order by reversing the order of the variation records array. */ @@ -1180,7 +1340,7 @@ struct CmapSubtableFormat14 protected: HBUINT16 format; /* Format number is set to 14. */ HBUINT32 length; /* Byte length of this subtable. */ - SortedArrayOf<VariationSelectorRecord, HBUINT32> + SortedArray32Of<VariationSelectorRecord> record; /* Variation selector records; sorted * in increasing order of `varSelector'. */ public: @@ -1235,6 +1395,20 @@ struct CmapSubtable } } + unsigned get_language () const + { + switch (u.format) { + case 0: return u.format0 .get_language (); + case 4: return u.format4 .get_language (); + case 6: return u.format6 .get_language (); + case 10: return u.format10.get_language (); + case 12: return u.format12.get_language (); + case 13: return u.format13.get_language (); + case 14: + default: return 0; + } + } + template<typename Iterator, hb_requires (hb_is_iterator (Iterator))> void serialize (hb_serialize_context_t *c, @@ -1246,7 +1420,7 @@ struct CmapSubtable switch (format) { case 4: return u.format4.serialize (c, it); case 12: return u.format12.serialize (c, it); - case 14: return u.format14.serialize (c, plan->unicodes, plan->glyphs_requested, plan->glyph_map, base); + case 14: return u.format14.serialize (c, &plan->unicodes, &plan->glyphs_requested, plan->glyph_map, base); default: return; } } @@ -1255,6 +1429,7 @@ struct CmapSubtable { TRACE_SANITIZE (this); if (!u.format.sanitize (c)) return_trace (false); + hb_barrier (); switch (u.format) { case 0: return_trace (u.format0 .sanitize (c)); case 4: return_trace (u.format4 .sanitize (c)); @@ -1338,70 +1513,248 @@ struct EncodingRecord HBUINT16 platformID; /* Platform ID. */ HBUINT16 encodingID; /* Platform-specific encoding ID. */ - LOffsetTo<CmapSubtable> + Offset32To<CmapSubtable> subtable; /* Byte offset from beginning of table to the subtable for this encoding. */ public: DEFINE_SIZE_STATIC (8); }; +struct cmap; + +struct SubtableUnicodesCache { + + private: + hb_blob_ptr_t<cmap> base_blob; + const char* base; + hb_hashmap_t<unsigned, hb::unique_ptr<hb_set_t>> cached_unicodes; + + public: + + static SubtableUnicodesCache* create (hb_blob_ptr_t<cmap> source_table) + { + SubtableUnicodesCache* cache = + (SubtableUnicodesCache*) hb_malloc (sizeof(SubtableUnicodesCache)); + new (cache) SubtableUnicodesCache (source_table); + return cache; + } + + static void destroy (void* value) { + if (!value) return; + + SubtableUnicodesCache* cache = (SubtableUnicodesCache*) value; + cache->~SubtableUnicodesCache (); + hb_free (cache); + } + + SubtableUnicodesCache(const void* cmap_base) + : base_blob(), + base ((const char*) cmap_base), + cached_unicodes () + {} + + SubtableUnicodesCache(hb_blob_ptr_t<cmap> base_blob_) + : base_blob(base_blob_), + base ((const char *) base_blob.get()), + cached_unicodes () + {} + + ~SubtableUnicodesCache() + { + base_blob.destroy (); + } + + bool same_base(const void* other) const + { + return other == (const void*) base; + } + + const hb_set_t* set_for (const EncodingRecord* record, + SubtableUnicodesCache& mutable_cache) const + { + if (cached_unicodes.has ((unsigned) ((const char *) record - base))) + return cached_unicodes.get ((unsigned) ((const char *) record - base)); + + return mutable_cache.set_for (record); + } + + const hb_set_t* set_for (const EncodingRecord* record) + { + if (!cached_unicodes.has ((unsigned) ((const char *) record - base))) + { + hb_set_t *s = hb_set_create (); + if (unlikely (s->in_error ())) + return hb_set_get_empty (); + + (base+record->subtable).collect_unicodes (s); + + if (unlikely (!cached_unicodes.set ((unsigned) ((const char *) record - base), hb::unique_ptr<hb_set_t> {s}))) + return hb_set_get_empty (); + + return s; + } + return cached_unicodes.get ((unsigned) ((const char *) record - base)); + } + +}; + +static inline uint_fast16_t +_hb_symbol_pua_map (unsigned codepoint) +{ + if (codepoint <= 0x00FFu) + { + /* For symbol-encoded OpenType fonts, we duplicate the + * U+F000..F0FF range at U+0000..U+00FF. That's what + * Windows seems to do, and that's hinted about at: + * https://docs.microsoft.com/en-us/typography/opentype/spec/recom + * under "Non-Standard (Symbol) Fonts". */ + return 0xF000u + codepoint; + } + return 0; +} + struct cmap { static constexpr hb_tag_t tableTag = HB_OT_TAG_cmap; + + static SubtableUnicodesCache* create_filled_cache(hb_blob_ptr_t<cmap> source_table) { + const cmap* cmap = source_table.get(); + auto it = + + hb_iter (cmap->encodingRecord) + | hb_filter ([&](const EncodingRecord& _) { + return cmap::filter_encoding_records_for_subset (cmap, _); + }) + ; + + SubtableUnicodesCache* cache = SubtableUnicodesCache::create(source_table); + for (const EncodingRecord& _ : it) + cache->set_for(&_); // populate the cache for this encoding record. + + return cache; + } + template<typename Iterator, typename EncodingRecIter, hb_requires (hb_is_iterator (EncodingRecIter))> - void serialize (hb_serialize_context_t *c, + bool serialize (hb_serialize_context_t *c, Iterator it, EncodingRecIter encodingrec_iter, const void *base, - const hb_subset_plan_t *plan) + hb_subset_plan_t *plan, + bool drop_format_4 = false) { - if (unlikely (!c->extend_min ((*this)))) return; + if (unlikely (!c->extend_min ((*this)))) return false; this->version = 0; unsigned format4objidx = 0, format12objidx = 0, format14objidx = 0; + auto snap = c->snapshot (); + + SubtableUnicodesCache local_unicodes_cache (base); + const SubtableUnicodesCache* unicodes_cache = &local_unicodes_cache; + + if (plan->accelerator && + plan->accelerator->cmap_cache && + plan->accelerator->cmap_cache->same_base (base)) + unicodes_cache = plan->accelerator->cmap_cache; for (const EncodingRecord& _ : encodingrec_iter) { + if (c->in_error ()) + return false; + unsigned format = (base+_.subtable).u.format; - if (!plan->glyphs_requested->is_empty ()) + if (format != 4 && format != 12 && format != 14) continue; + + const hb_set_t* unicodes_set = unicodes_cache->set_for (&_, local_unicodes_cache); + + if (!drop_format_4 && format == 4) { - hb_set_t unicodes_set; - hb_map_t cp_glyphid_map; - (base+_.subtable).collect_mapping (&unicodes_set, &cp_glyphid_map); - - auto table_iter = - + hb_zip (unicodes_set.iter(), unicodes_set.iter() | hb_map(cp_glyphid_map)) - | hb_filter (plan->_glyphset, hb_second) - | hb_filter ([plan] (const hb_pair_t<hb_codepoint_t, hb_codepoint_t>& p) - { - return plan->unicodes->has (p.first) || - plan->glyphs_requested->has (p.second); - }) - | hb_map ([plan] (const hb_pair_t<hb_codepoint_t, hb_codepoint_t>& p_org) - { - return hb_pair_t<hb_codepoint_t, hb_codepoint_t> (p_org.first, plan->glyph_map->get(p_org.second)); - }) - ; - - if (format == 4) c->copy (_, table_iter, 4u, base, plan, &format4objidx); - else if (format == 12) c->copy (_, table_iter, 12u, base, plan, &format12objidx); - else if (format == 14) c->copy (_, table_iter, 14u, base, plan, &format14objidx); + c->copy (_, + it | hb_filter (*unicodes_set, hb_first), 4u, base, plan, &format4objidx); + if (c->in_error () && c->only_overflow ()) + { + // cmap4 overflowed, reset and retry serialization without format 4 subtables. + c->revert (snap); + return serialize (c, it, + encodingrec_iter, + base, + plan, + true); + } } - /* when --gids option is not used, we iterate input unicodes instead of - * all codepoints in each subtable, which is more efficient */ - else + + else if (format == 12) { - hb_set_t unicodes_set; - (base+_.subtable).collect_unicodes (&unicodes_set); + if (_can_drop (_, + *unicodes_set, + base, + *unicodes_cache, + local_unicodes_cache, + + it | hb_map (hb_first), encodingrec_iter)) + continue; + c->copy (_, + it | hb_filter (*unicodes_set, hb_first), 12u, base, plan, &format12objidx); + } + else if (format == 14) c->copy (_, it, 14u, base, plan, &format14objidx); + } + c->check_assign(this->encodingRecord.len, + (c->length () - cmap::min_size)/EncodingRecord::static_size, + HB_SERIALIZE_ERROR_INT_OVERFLOW); + + // Fail if format 4 was dropped and there is no cmap12. + return !drop_format_4 || format12objidx; + } + + template<typename Iterator, typename EncodingRecordIterator, + hb_requires (hb_is_iterator (Iterator)), + hb_requires (hb_is_iterator (EncodingRecordIterator))> + bool _can_drop (const EncodingRecord& cmap12, + const hb_set_t& cmap12_unicodes, + const void* base, + const SubtableUnicodesCache& unicodes_cache, + SubtableUnicodesCache& local_unicodes_cache, + Iterator subset_unicodes, + EncodingRecordIterator encoding_records) + { + for (auto cp : + subset_unicodes | hb_filter (cmap12_unicodes)) + { + if (cp >= 0x10000) return false; + } + + unsigned target_platform; + unsigned target_encoding; + unsigned target_language = (base+cmap12.subtable).get_language (); - if (format == 4) c->copy (_, + it | hb_filter (unicodes_set, hb_first), 4u, base, plan, &format4objidx); - else if (format == 12) c->copy (_, + it | hb_filter (unicodes_set, hb_first), 12u, base, plan, &format12objidx); - else if (format == 14) c->copy (_, it, 14u, base, plan, &format14objidx); + if (cmap12.platformID == 0 && cmap12.encodingID == 4) + { + target_platform = 0; + target_encoding = 3; + } else if (cmap12.platformID == 3 && cmap12.encodingID == 10) { + target_platform = 3; + target_encoding = 1; + } else { + return false; + } + + for (const auto& _ : encoding_records) + { + if (_.platformID != target_platform + || _.encodingID != target_encoding + || (base+_.subtable).get_language() != target_language) + continue; + + const hb_set_t* sibling_unicodes = unicodes_cache.set_for (&_, local_unicodes_cache); + + auto cmap12 = + subset_unicodes | hb_filter (cmap12_unicodes); + auto sibling = + subset_unicodes | hb_filter (*sibling_unicodes); + for (; cmap12 && sibling; cmap12++, sibling++) + { + unsigned a = *cmap12; + unsigned b = *sibling; + if (a != b) return false; } + + return !cmap12 && !sibling; } - c->check_assign(this->encodingRecord.len, (c->length () - cmap::min_size)/EncodingRecord::static_size); + return false; } void closure_glyphs (const hb_set_t *unicodes, @@ -1420,21 +1773,12 @@ struct cmap TRACE_SUBSET (this); cmap *cmap_prime = c->serializer->start_embed<cmap> (); - if (unlikely (!c->serializer->check_success (cmap_prime))) return_trace (false); auto encodingrec_iter = + hb_iter (encodingRecord) - | hb_filter ([&] (const EncodingRecord& _) - { - if ((_.platformID == 0 && _.encodingID == 3) || - (_.platformID == 0 && _.encodingID == 4) || - (_.platformID == 3 && _.encodingID == 1) || - (_.platformID == 3 && _.encodingID == 10) || - (this + _.subtable).u.format == 14) - return true; - - return false; - }) + | hb_filter ([&](const EncodingRecord& _) { + return cmap::filter_encoding_records_for_subset (this, _); + }) ; if (unlikely (!encodingrec_iter.len ())) return_trace (false); @@ -1447,7 +1791,7 @@ struct cmap unsigned format = (this + _.subtable).u.format; if (format == 12) has_format12 = true; - const EncodingRecord *table = hb_addressof (_); + const EncodingRecord *table = std::addressof (_); if (_.platformID == 0 && _.encodingID == 3) unicode_bmp = table; else if (_.platformID == 0 && _.encodingID == 4) unicode_ucs4 = table; else if (_.platformID == 3 && _.encodingID == 1) ms_bmp = table; @@ -1458,18 +1802,16 @@ struct cmap if (unlikely (has_format12 && (!unicode_ucs4 && !ms_ucs4))) return_trace (false); auto it = - + hb_iter (c->plan->unicodes) - | hb_map ([&] (hb_codepoint_t _) - { - hb_codepoint_t new_gid = HB_MAP_VALUE_INVALID; - c->plan->new_gid_for_codepoint (_, &new_gid); - return hb_pair_t<hb_codepoint_t, hb_codepoint_t> (_, new_gid); - }) - | hb_filter ([&] (const hb_pair_t<hb_codepoint_t, hb_codepoint_t> _) + + c->plan->unicode_to_new_gid_list.iter () + | hb_filter ([&] (const hb_codepoint_pair_t _) { return (_.second != HB_MAP_VALUE_INVALID); }) ; - cmap_prime->serialize (c->serializer, it, encodingrec_iter, this, c->plan); - return_trace (true); + + return_trace (cmap_prime->serialize (c->serializer, + it, + encodingrec_iter, + this, + c->plan)); } const CmapSubtable *find_best_subtable (bool *symbol = nullptr) const @@ -1505,7 +1847,9 @@ struct cmap struct accelerator_t { - void init (hb_face_t *face) + using cache_t = hb_cache_t<21, 16, 8, true>; + + accelerator_t (hb_face_t *face) { this->table = hb_sanitize_context_t ().reference_table<cmap> (face); bool symbol; @@ -1519,7 +1863,24 @@ struct cmap this->get_glyph_data = subtable; if (unlikely (symbol)) - this->get_glyph_funcZ = get_glyph_from_symbol<CmapSubtable>; + { + switch ((unsigned) face->table.OS2->get_font_page ()) { + case OS2::font_page_t::FONT_PAGE_NONE: + this->get_glyph_funcZ = get_glyph_from_symbol<CmapSubtable, _hb_symbol_pua_map>; + break; +#ifndef HB_NO_OT_SHAPER_ARABIC_FALLBACK + case OS2::font_page_t::FONT_PAGE_SIMP_ARABIC: + this->get_glyph_funcZ = get_glyph_from_symbol<CmapSubtable, _hb_arabic_pua_simp_map>; + break; + case OS2::font_page_t::FONT_PAGE_TRAD_ARABIC: + this->get_glyph_funcZ = get_glyph_from_symbol<CmapSubtable, _hb_arabic_pua_trad_map>; + break; +#endif + default: + this->get_glyph_funcZ = get_glyph_from<CmapSubtable>; + break; + } + } else { switch (subtable->u.format) { @@ -1540,29 +1901,45 @@ struct cmap } } } + ~accelerator_t () { this->table.destroy (); } + + inline bool _cached_get (hb_codepoint_t unicode, + hb_codepoint_t *glyph, + cache_t *cache) const + { + unsigned v; + if (cache && cache->get (unicode, &v)) + { + *glyph = v; + return true; + } + bool ret = this->get_glyph_funcZ (this->get_glyph_data, unicode, glyph); - void fini () { this->table.destroy (); } + if (cache && ret) + cache->set (unicode, *glyph); + return ret; + } bool get_nominal_glyph (hb_codepoint_t unicode, - hb_codepoint_t *glyph) const + hb_codepoint_t *glyph, + cache_t *cache = nullptr) const { - if (unlikely (!this->get_glyph_funcZ)) return false; - return this->get_glyph_funcZ (this->get_glyph_data, unicode, glyph); + if (unlikely (!this->get_glyph_funcZ)) return 0; + return _cached_get (unicode, glyph, cache); } + unsigned int get_nominal_glyphs (unsigned int count, const hb_codepoint_t *first_unicode, unsigned int unicode_stride, hb_codepoint_t *first_glyph, - unsigned int glyph_stride) const + unsigned int glyph_stride, + cache_t *cache = nullptr) const { if (unlikely (!this->get_glyph_funcZ)) return 0; - hb_cmap_get_glyph_func_t get_glyph_funcZ = this->get_glyph_funcZ; - const void *get_glyph_data = this->get_glyph_data; - unsigned int done; for (done = 0; - done < count && get_glyph_funcZ (get_glyph_data, *first_unicode, first_glyph); + done < count && _cached_get (*first_unicode, first_glyph, cache); done++) { first_unicode = &StructAtOffsetUnaligned<hb_codepoint_t> (first_unicode, unicode_stride); @@ -1573,7 +1950,8 @@ struct cmap bool get_variation_glyph (hb_codepoint_t unicode, hb_codepoint_t variation_selector, - hb_codepoint_t *glyph) const + hb_codepoint_t *glyph, + cache_t *cache = nullptr) const { switch (this->subtable_uvs->get_glyph_variant (unicode, variation_selector, @@ -1584,7 +1962,7 @@ struct cmap case GLYPH_VARIANT_USE_DEFAULT: break; } - return get_nominal_glyph (unicode, glyph); + return get_nominal_glyph (unicode, glyph, cache); } void collect_unicodes (hb_set_t *out, unsigned int num_glyphs) const @@ -1602,6 +1980,7 @@ struct cmap typedef bool (*hb_cmap_get_glyph_func_t) (const void *obj, hb_codepoint_t codepoint, hb_codepoint_t *glyph); + typedef uint_fast16_t (*hb_pua_remap_func_t) (unsigned); template <typename Type> HB_INTERNAL static bool get_glyph_from (const void *obj, @@ -1612,7 +1991,7 @@ struct cmap return typed_obj->get_glyph (codepoint, glyph); } - template <typename Type> + template <typename Type, hb_pua_remap_func_t remap> HB_INTERNAL static bool get_glyph_from_symbol (const void *obj, hb_codepoint_t codepoint, hb_codepoint_t *glyph) @@ -1621,15 +2000,8 @@ struct cmap if (likely (typed_obj->get_glyph (codepoint, glyph))) return true; - if (codepoint <= 0x00FFu) - { - /* For symbol-encoded OpenType fonts, we duplicate the - * U+F000..F0FF range at U+0000..U+00FF. That's what - * Windows seems to do, and that's hinted about at: - * https://docs.microsoft.com/en-us/typography/opentype/spec/recom - * under "Non-Standard (Symbol) Fonts". */ - return typed_obj->get_glyph (0xF000u + codepoint, glyph); - } + if (hb_codepoint_t c = remap (codepoint)) + return typed_obj->get_glyph (c, glyph); return false; } @@ -1691,19 +2063,35 @@ struct cmap { TRACE_SANITIZE (this); return_trace (c->check_struct (this) && + hb_barrier () && likely (version == 0) && encodingRecord.sanitize (c, this)); } + private: + + static bool filter_encoding_records_for_subset(const cmap* cmap, + const EncodingRecord& _) + { + return + (_.platformID == 0 && _.encodingID == 3) || + (_.platformID == 0 && _.encodingID == 4) || + (_.platformID == 3 && _.encodingID == 1) || + (_.platformID == 3 && _.encodingID == 10) || + (cmap + _.subtable).u.format == 14; + } + protected: HBUINT16 version; /* Table version number (0). */ - SortedArrayOf<EncodingRecord> + SortedArray16Of<EncodingRecord> encodingRecord; /* Encoding tables. */ public: DEFINE_SIZE_ARRAY (4, encodingRecord); }; -struct cmap_accelerator_t : cmap::accelerator_t {}; +struct cmap_accelerator_t : cmap::accelerator_t { + cmap_accelerator_t (hb_face_t *face) : cmap::accelerator_t (face) {} +}; } /* namespace OT */ |