[subset] support BASE table subsetting (#4618)

* [subset] support BASE table

* [subset] add tests for BASE table subsetting

src/hb-open-type.hh

@@ -985,6 +985,13 @@ struct SortedArrayOf : ArrayOf<Type, LenType>
     return_trace (ret);
   }
 
+  SortedArrayOf* copy (hb_serialize_context_t *c) const
+  {
+    TRACE_SERIALIZE (this);
+    SortedArrayOf* out = reinterpret_cast<SortedArrayOf *> (ArrayOf<Type, LenType>::copy (c));
+    return_trace (out);
+  }
+
   template <typename T>
   Type &bsearch (const T &x, Type &not_found = Crap (Type))
   { return *as_array ().bsearch (x, &not_found); }

src/hb-ot-layout-base-table.hh

@@ -46,6 +46,12 @@ struct BaseCoordFormat1
     return HB_DIRECTION_IS_HORIZONTAL (direction) ? font->em_scale_y (coordinate) : font->em_scale_x (coordinate);
   }
 
+  bool subset (hb_subset_context_t *c) const
+  {
+    TRACE_SUBSET (this);
+    return_trace ((bool) c->serializer->embed (*this));
+  }
+
   bool sanitize (hb_sanitize_context_t *c) const
   {
     TRACE_SANITIZE (this);
@@ -67,6 +73,17 @@ struct BaseCoordFormat2
     return HB_DIRECTION_IS_HORIZONTAL (direction) ? font->em_scale_y (coordinate) : font->em_scale_x (coordinate);
   }
 
+  bool subset (hb_subset_context_t *c) const
+  {
+    TRACE_SUBSET (this);
+    auto *out = c->serializer->embed (*this);
+    if (unlikely (!out)) return_trace (false);
+
+    return_trace (c->serializer->check_assign (out->referenceGlyph,
+                                               c->plan->glyph_map->get (referenceGlyph),
+                                               HB_SERIALIZE_ERROR_INT_OVERFLOW));
+  }
+
   bool sanitize (hb_sanitize_context_t *c) const
   {
     TRACE_SANITIZE (this);
@@ -96,6 +113,23 @@ struct BaseCoordFormat3
 	 : font->em_scale_x (coordinate) + device.get_x_delta (font, var_store);
   }
 
+  void collect_variation_indices (hb_set_t& varidx_set /* OUT */) const
+  {
+    unsigned varidx = (this+deviceTable).get_variation_index ();
+    varidx_set.add (varidx);
+  }
+
+  bool subset (hb_subset_context_t *c) const
+  {
+    TRACE_SUBSET (this);
+    auto *out = c->serializer->embed (*this);
+    if (unlikely (!out)) return_trace (false);
+
+    return_trace (out->deviceTable.serialize_copy (c->serializer, deviceTable,
+                                                   this, 0,
+                                                   hb_serialize_context_t::Head,
+                                                   &c->plan->base_variation_idx_map));
+  }
 
   bool sanitize (hb_sanitize_context_t *c) const
   {
@@ -131,6 +165,27 @@ struct BaseCoord
     }
   }
 
+  void collect_variation_indices (hb_set_t& varidx_set /* OUT */) const
+  {
+    switch (u.format) {
+    case 3: u.format3.collect_variation_indices (varidx_set);
+    default:return;
+    }
+  }
+
+  template <typename context_t, typename ...Ts>
+  typename context_t::return_t dispatch (context_t *c, Ts&&... ds) const
+  {
+    if (unlikely (!c->may_dispatch (this, &u.format))) return c->no_dispatch_return_value ();
+    TRACE_DISPATCH (this, u.format);
+    switch (u.format) {
+    case 1: return_trace (c->dispatch (u.format1, std::forward<Ts> (ds)...));
+    case 2: return_trace (c->dispatch (u.format2, std::forward<Ts> (ds)...));
+    case 3: return_trace (c->dispatch (u.format3, std::forward<Ts> (ds)...));
+    default:return_trace (c->default_return_value ());
+    }
+  }
+
   bool sanitize (hb_sanitize_context_t *c) const
   {
     TRACE_SANITIZE (this);
@@ -161,12 +216,37 @@ struct FeatMinMaxRecord
 
   bool has_data () const { return tag; }
 
+  hb_tag_t get_feature_tag () const { return tag; }
+
   void get_min_max (const BaseCoord **min, const BaseCoord **max) const
   {
     if (likely (min)) *min = &(this+minCoord);
     if (likely (max)) *max = &(this+maxCoord);
   }
 
+  void collect_variation_indices (const hb_subset_plan_t* plan,
+                                  const void *base,
+                                  hb_set_t& varidx_set /* OUT */) const
+  {
+    if (!plan->layout_features.has (tag))
+      return;
+
+    (base+minCoord).collect_variation_indices (varidx_set);
+    (base+maxCoord).collect_variation_indices (varidx_set);
+  }
+
+  bool subset (hb_subset_context_t *c,
+               const void *base) const
+  {
+    TRACE_SUBSET (this);
+    auto *out = c->serializer->embed (*this);
+    if (unlikely (!out)) return_trace (false);
+    if (!(out->minCoord.serialize_subset (c, minCoord, base)))
+      return_trace (false);
+
+    return_trace (out->maxCoord.serialize_subset (c, maxCoord, base));
+  }
+
   bool sanitize (hb_sanitize_context_t *c, const void *base) const
   {
     TRACE_SANITIZE (this);
@@ -206,6 +286,39 @@ struct MinMax
     }
   }
 
+  void collect_variation_indices (const hb_subset_plan_t* plan,
+                                  hb_set_t& varidx_set /* OUT */) const
+  {
+    (this+minCoord).collect_variation_indices (varidx_set);
+    (this+maxCoord).collect_variation_indices (varidx_set);
+    for (const FeatMinMaxRecord& record : featMinMaxRecords)
+      record.collect_variation_indices (plan, this, varidx_set);
+  }
+
+  bool subset (hb_subset_context_t *c) const
+  {
+    TRACE_SUBSET (this);
+    auto *out = c->serializer->start_embed (*this);
+    if (unlikely (!out || !c->serializer->extend_min (out))) return_trace (false);
+
+    if (!(out->minCoord.serialize_subset (c, minCoord, this)) ||
+        !(out->maxCoord.serialize_subset (c, maxCoord, this)))
+      return_trace (false);
+
+    unsigned len = 0;
+    for (const FeatMinMaxRecord& _ : featMinMaxRecords)
+    {
+      hb_tag_t feature_tag = _.get_feature_tag ();
+      if (!c->plan->layout_features.has (feature_tag))
+        continue;
+
+      if (!_.subset (c, this)) return false;
+      len++;
+    }
+    return_trace (c->serializer->check_assign (out->featMinMaxRecords.len, len,
+                                               HB_SERIALIZE_ERROR_INT_OVERFLOW));
+  }
+
   bool sanitize (hb_sanitize_context_t *c) const
   {
     TRACE_SANITIZE (this);
@@ -240,6 +353,26 @@ struct BaseValues
     return this+baseCoords[baseline_tag_index];
   }
 
+  void collect_variation_indices (hb_set_t& varidx_set /* OUT */) const
+  {
+    for (const auto& _ : baseCoords)
+      (this+_).collect_variation_indices (varidx_set);
+  }
+
+  bool subset (hb_subset_context_t *c) const
+  {
+    TRACE_SUBSET (this);
+    auto *out = c->serializer->start_embed (*this);
+    if (unlikely (!out || !c->serializer->extend_min (out))) return_trace (false);
+    out->defaultIndex = defaultIndex;
+
+    for (const auto& _ : baseCoords)
+      if (!subset_offset_array (c, out->baseCoords, this) (_))
+        return_trace (false);
+
+    return_trace (bool (out->baseCoords));
+  }
+
   bool sanitize (hb_sanitize_context_t *c) const
   {
     TRACE_SANITIZE (this);
@@ -270,6 +403,20 @@ struct BaseLangSysRecord
 
   const MinMax &get_min_max () const { return this+minMax; }
 
+  void collect_variation_indices (const hb_subset_plan_t* plan,
+                                  hb_set_t& varidx_set /* OUT */) const
+  { (this+minMax).collect_variation_indices (plan, varidx_set); }
+
+  bool subset (hb_subset_context_t *c,
+               const void *base) const
+  {
+    TRACE_SUBSET (this);
+    auto *out = c->serializer->embed (*this);
+    if (unlikely (!out)) return_trace (false);
+
+    return_trace (out->minMax.serialize_subset (c, minMax, base));
+  }
+
   bool sanitize (hb_sanitize_context_t *c, const void *base) const
   {
     TRACE_SANITIZE (this);
@@ -300,6 +447,35 @@ struct BaseScript
   bool has_values () const { return baseValues; }
   bool has_min_max () const { return defaultMinMax; /* TODO What if only per-language is present? */ }
 
+  void collect_variation_indices (const hb_subset_plan_t* plan,
+                                  hb_set_t& varidx_set /* OUT */) const
+  {
+    (this+baseValues).collect_variation_indices (varidx_set);
+    (this+defaultMinMax).collect_variation_indices (plan, varidx_set);
+    
+    for (const BaseLangSysRecord& _ : baseLangSysRecords)
+      _.collect_variation_indices (plan, varidx_set);
+  }
+
+  bool subset (hb_subset_context_t *c) const
+  {
+    TRACE_SUBSET (this);
+    auto *out = c->serializer->start_embed (*this);
+    if (unlikely (!out || !c->serializer->extend_min (out))) return_trace (false);
+
+    if (baseValues && !out->baseValues.serialize_subset (c, baseValues, this))
+      return_trace (false);
+
+    if (defaultMinMax && !out->defaultMinMax.serialize_subset (c, defaultMinMax, this))
+      return_trace (false);
+
+    for (const auto& _ : baseLangSysRecords)
+      if (!_.subset (c, this)) return_trace (false);
+
+    return_trace (c->serializer->check_assign (out->baseLangSysRecords.len, baseLangSysRecords.len,
+                                               HB_SERIALIZE_ERROR_INT_OVERFLOW));
+  }
+
   bool sanitize (hb_sanitize_context_t *c) const
   {
     TRACE_SANITIZE (this);
@@ -332,9 +508,31 @@ struct BaseScriptRecord
 
   bool has_data () const { return baseScriptTag; }
 
+  hb_tag_t get_script_tag () const { return baseScriptTag; }
+
   const BaseScript &get_base_script (const BaseScriptList *list) const
   { return list+baseScript; }
 
+  void collect_variation_indices (const hb_subset_plan_t* plan,
+                                  const void* list,
+                                  hb_set_t& varidx_set /* OUT */) const
+  {
+    if (!plan->layout_scripts.has (baseScriptTag))
+      return;
+
+    (list+baseScript).collect_variation_indices (plan, varidx_set);
+  }
+
+  bool subset (hb_subset_context_t *c,
+               const void *base) const
+  {
+    TRACE_SUBSET (this);
+    auto *out = c->serializer->embed (*this);
+    if (unlikely (!out)) return_trace (false);
+
+    return_trace (out->baseScript.serialize_subset (c, baseScript, base));
+  }
+
   bool sanitize (hb_sanitize_context_t *c, const void *base) const
   {
     TRACE_SANITIZE (this);
@@ -361,6 +559,33 @@ struct BaseScriptList
     return record->has_data () ? record->get_base_script (this) : Null (BaseScript);
   }
 
+  void collect_variation_indices (const hb_subset_plan_t* plan,
+                                  hb_set_t& varidx_set /* OUT */) const
+  {
+    for (const BaseScriptRecord& _ : baseScriptRecords)
+      _.collect_variation_indices (plan, this, varidx_set);
+  }
+
+  bool subset (hb_subset_context_t *c) const
+  {
+    TRACE_SUBSET (this);
+    auto *out = c->serializer->start_embed (*this);
+    if (unlikely (!out || !c->serializer->extend_min (out))) return_trace (false);
+
+    unsigned len = 0;
+    for (const BaseScriptRecord& _ : baseScriptRecords)
+    {
+      hb_tag_t script_tag = _.get_script_tag ();
+      if (!c->plan->layout_scripts.has (script_tag))
+        continue;
+
+      if (!_.subset (c, this)) return false;
+      len++;
+    }
+    return_trace (c->serializer->check_assign (out->baseScriptRecords.len, len,
+                                               HB_SERIALIZE_ERROR_INT_OVERFLOW));
+  }
+
   bool sanitize (hb_sanitize_context_t *c) const
   {
     TRACE_SANITIZE (this);
@@ -422,6 +647,20 @@ struct Axis
     return true;
   }
 
+  void collect_variation_indices (const hb_subset_plan_t* plan,
+                                  hb_set_t& varidx_set /* OUT */) const
+  { (this+baseScriptList).collect_variation_indices (plan, varidx_set); }
+
+  bool subset (hb_subset_context_t *c) const
+  {
+    TRACE_SUBSET (this);
+    auto *out = c->serializer->embed (*this);
+    if (unlikely (!out)) return_trace (false);
+
+    out->baseTagList.serialize_copy (c->serializer, baseTagList, this);
+    return_trace (out->baseScriptList.serialize_subset (c, baseScriptList, this));
+  }
+
   bool sanitize (hb_sanitize_context_t *c) const
   {
     TRACE_SANITIZE (this);
@@ -453,9 +692,42 @@ struct BASE
   const Axis &get_axis (hb_direction_t direction) const
   { return HB_DIRECTION_IS_VERTICAL (direction) ? this+vAxis : this+hAxis; }
 
+  bool has_var_store () const
+  { return version.to_int () >= 0x00010001u && varStore != 0; }
+
   const ItemVariationStore &get_var_store () const
   { return version.to_int () < 0x00010001u ? Null (ItemVariationStore) : this+varStore; }
 
+  void collect_variation_indices (const hb_subset_plan_t* plan,
+                                  hb_set_t& varidx_set /* OUT */) const
+  {
+    (this+hAxis).collect_variation_indices (plan, varidx_set);
+    (this+vAxis).collect_variation_indices (plan, varidx_set);
+  }
+
+  bool subset (hb_subset_context_t *c) const
+  {
+    TRACE_SUBSET (this);
+    auto *out = c->serializer->start_embed (*this);
+    if (unlikely (!out || !c->serializer->extend_min (out))) return_trace (false);
+
+    out->version = version;
+    if (hAxis && !out->hAxis.serialize_subset (c, hAxis, this))
+      return_trace (false);
+
+    if (vAxis && !out->vAxis.serialize_subset (c, vAxis, this))
+      return_trace (false);
+
+    if (has_var_store ())
+    {
+      if (!c->serializer->allocate_size<Offset32To<ItemVariationStore>> (Offset32To<ItemVariationStore>::static_size))
+        return_trace (false);
+      return_trace (out->varStore.serialize_subset (c, varStore, this, c->plan->base_varstore_inner_maps.as_array ()));
+    }
+
+    return_trace (true);
+  }
+
   bool get_baseline (hb_font_t      *font,
 		     hb_tag_t        baseline_tag,
 		     hb_direction_t  direction,

src/hb-subset-input.cc

@@ -51,7 +51,6 @@ hb_subset_input_t::hb_subset_input_t ()
     HB_TAG ('k', 'e', 'r', 'n'),
 
     // Copied from fontTools:
-    HB_TAG ('B', 'A', 'S', 'E'),
     HB_TAG ('J', 'S', 'T', 'F'),
     HB_TAG ('D', 'S', 'I', 'G'),
     HB_TAG ('E', 'B', 'D', 'T'),

src/hb-subset-plan-member-list.hh

@@ -143,6 +143,12 @@ HB_SUBSET_PLAN_MEMBER (mutable hb_hashmap_t E(<hb_codepoint_t, contour_point_vec
 //new gids set for composite glyphs
 HB_SUBSET_PLAN_MEMBER (hb_set_t, composite_new_gids)
 
+//Old BASE item variation index -> (New varidx, 0) mapping
+HB_SUBSET_PLAN_MEMBER (hb_hashmap_t E(<unsigned, hb_pair_t E(<unsigned, int>)>), base_variation_idx_map)
+
+//BASE table varstore retained varidx mapping
+HB_SUBSET_PLAN_MEMBER (hb_vector_t<hb_inc_bimap_t>, base_varstore_inner_maps)
+
 #ifdef HB_EXPERIMENTAL_API
 // name table overrides map: hb_ot_name_record_ids_t-> name string new value or
 // None to indicate should remove

src/hb-subset-plan.cc

@@ -32,6 +32,7 @@
 
 #include "hb-ot-cmap-table.hh"
 #include "hb-ot-glyf-table.hh"
+#include "hb-ot-layout-base-table.hh"
 #include "hb-ot-layout-gdef-table.hh"
 #include "hb-ot-layout-gpos-table.hh"
 #include "hb-ot-layout-gsub-table.hh"
@@ -431,6 +432,52 @@ _collect_layout_variation_indices (hb_subset_plan_t* plan)
   gdef.destroy ();
   gpos.destroy ();
 }
+
+#ifndef HB_NO_BASE
+/* used by BASE table only, delta is always set to 0 in the output map */
+static inline void
+_remap_variation_indices (const hb_set_t& indices,
+                          unsigned subtable_count,
+                          hb_hashmap_t<unsigned, hb_pair_t<unsigned, int>>& variation_idx_delta_map /* OUT */)
+{
+  unsigned new_major = 0, new_minor = 0;
+  unsigned last_major = (indices.get_min ()) >> 16;
+  for (unsigned idx : indices)
+  {
+    uint16_t major = idx >> 16;
+    if (major >= subtable_count) break;
+    if (major != last_major)
+    {
+      new_minor = 0;
+      ++new_major;
+    }
+
+    unsigned new_idx = (new_major << 16) + new_minor;
+    variation_idx_delta_map.set (idx, hb_pair_t<unsigned, int> (new_idx, 0));
+    ++new_minor;
+    last_major = major;
+  }
+}
+
+static inline void
+_collect_base_variation_indices (hb_subset_plan_t* plan)
+{
+  hb_blob_ptr_t<OT::BASE> base = plan->source_table<OT::BASE> ();
+  if (!base->has_var_store ())
+  {
+    base.destroy ();
+    return;
+  }
+
+  hb_set_t varidx_set;
+  base->collect_variation_indices (plan, varidx_set);
+  unsigned subtable_count = base->get_var_store ().get_sub_table_count ();
+  base.destroy ();
+
+  _remap_variation_indices (varidx_set, subtable_count, plan->base_variation_idx_map);
+  _generate_varstore_inner_maps (varidx_set, subtable_count, plan->base_varstore_inner_maps);
+}
+#endif
 #endif
 
 static inline void
@@ -1212,6 +1259,13 @@ hb_subset_plan_t::hb_subset_plan_t (hb_face_t *face,
   if (!drop_tables.has (HB_OT_TAG_GDEF))
     _remap_used_mark_sets (this, used_mark_sets_map);
 
+#ifndef HB_NO_VAR
+#ifndef HB_NO_BASE
+  if (!drop_tables.has (HB_OT_TAG_BASE))
+    _collect_base_variation_indices (this);
+#endif
+#endif
+
   if (unlikely (in_error ()))
     return;
 

src/hb-subset.cc

@@ -48,6 +48,7 @@
 #include "hb-ot-cff2-table.hh"
 #include "hb-ot-vorg-table.hh"
 #include "hb-ot-name-table.hh"
+#include "hb-ot-layout-base-table.hh"
 #include "hb-ot-layout-gsub-table.hh"
 #include "hb-ot-layout-gpos-table.hh"
 #include "hb-ot-var-avar-table.hh"
@@ -503,6 +504,7 @@ _subset_table (hb_subset_plan_t *plan,
   case HB_OT_TAG_CBLC: return _subset<const OT::CBLC> (plan, buf);
   case HB_OT_TAG_CBDT: return true; /* skip CBDT, handled by CBLC */
   case HB_OT_TAG_MATH: return _subset<const OT::MATH> (plan, buf);
+  case HB_OT_TAG_BASE: return _subset<const OT::BASE> (plan, buf);
 
 #ifndef HB_NO_SUBSET_CFF
   case HB_OT_TAG_CFF1: return _subset<const OT::cff1> (plan, buf);
@@ -548,6 +550,7 @@ _subset_table (hb_subset_plan_t *plan,
     }
 #endif
     return _passthrough (plan, tag);
+
   default:
     if (plan->flags & HB_SUBSET_FLAGS_PASSTHROUGH_UNRECOGNIZED)
       return _passthrough (plan, tag);

test/subset/data/Makefile.am

@@ -59,6 +59,7 @@ EXTRA_DIST += \
 	expected/glyph_map \
 	expected/math \
 	expected/math_coverage_offset \
+	expected/subset_base \
 	expected/post \
 	expected/full_instance \
 	expected/instance_feature_variations \

test/subset/data/Makefile.sources

@@ -50,6 +50,7 @@ TESTS = \
 	tests/glyph_map.tests \
 	tests/math.tests \
 	tests/math_coverage_offset.tests \
+	tests/subset_base.tests \
 	tests/post.tests \
 	tests/full_instance.tests \
 	tests/instance_feature_variations.tests \

test/subset/data/profiles/retain-base-table-filter-scripts.txt

@@ -0,0 +1,2 @@
+--drop-tables-=BASE
+--layout-scripts=grek,cyrl

test/subset/data/profiles/retain-base-table.txt

@@ -0,0 +1 @@
+--drop-tables-=BASE

test/subset/data/tests/subset_base.tests

@@ -0,0 +1,10 @@
+FONTS:
+SourceSansPro-Regular.otf
+
+PROFILES:
+retain-base-table.txt
+retain-base-table-filter-scripts.txt
+
+SUBSETS:
+abc
+

test/subset/meson.build

@@ -42,6 +42,7 @@ tests = [
   'colr_glyphs',
   'math',
   'math_coverage_offset',
+  'subset_base',
   # TODO: re-enable once colrv1 subsetting is stabilized.
   # 'colrv1.notoemoji',
   'colrv1',

test/subset/run-tests.py

@@ -52,7 +52,7 @@ def run_test (test, should_check_ots, preprocess):
 	cli_args = ["--font-file=" + test.font_path,
 		    "--output-file=" + out_file,
 		    "--unicodes=%s" % test.unicodes (),
-		    "--drop-tables+=DSIG",
+		    "--drop-tables+=DSIG,BASE",
 		    "--drop-tables-=sbix"]
 	if preprocess:
 		cli_args.extend(["--preprocess-face",])