[subset] use inverted set for all unicodes.

Modify the code that handles input->unicodes to be safe with possibly inverted sets. Also adds --unicodes-= and --unicodes+= flags.

src/hb-subset-plan.cc

@@ -240,7 +240,12 @@ _populate_unicodes_to_retain (const hb_set_t *unicodes,
 {
   OT::cmap::accelerator_t cmap;
   cmap.init (plan->source);
+  constexpr static const int size_threshold = 65000;
 
+  if (unicodes->get_population () < size_threshold && glyphs->is_empty ())
+  {
+    // This is the fast path if it's anticipated that size of unicodes
+    // is << then the number of codepoints in the font.
     for (hb_codepoint_t cp : *unicodes)
     {
       hb_codepoint_t gid;
@@ -253,9 +258,6 @@ _populate_unicodes_to_retain (const hb_set_t *unicodes,
       plan->codepoint_to_glyph->set (cp, gid);
       plan->_glyphset_gsub->add (gid);
     }
-
-  if (glyphs->is_empty ())
-  {
     cmap.fini ();
     return;
   }
@@ -265,17 +267,27 @@ _populate_unicodes_to_retain (const hb_set_t *unicodes,
   cmap.fini ();
 
   for (hb_pair_t<hb_codepoint_t, hb_codepoint_t> cp_gid :
-       + unicode_glyphid_map.iter () | hb_filter (glyphs, hb_second))
+       + unicode_glyphid_map.iter ())
   {
+    if (!unicodes->has (cp_gid.first) && !glyphs->has (cp_gid.second))
+      continue;
+
     plan->unicodes->add (cp_gid.first);
     plan->codepoint_to_glyph->set (cp_gid.first, cp_gid.second);
+    plan->_glyphset_gsub->add (cp_gid.second);
+  }
+
+  // Add gids which where requested, but not mapped in cmap
+  for (hb_codepoint_t gid : glyphs->iter ())
+  {
+    if (gid >= plan->source->get_num_glyphs ())
+      break;
+    plan->_glyphset_gsub->add (gid);
   }
 }
 
 static void
 _populate_gids_to_retain (hb_subset_plan_t* plan,
-			  const hb_set_t *unicodes,
-			  const hb_set_t *input_glyphs_to_retain,
 			  bool close_over_gsub,
 			  bool close_over_gpos,
 			  bool close_over_gdef)
@@ -292,7 +304,6 @@ _populate_gids_to_retain (hb_subset_plan_t* plan,
   colr.init (plan->source);
 
   plan->_glyphset_gsub->add (0); // Not-def
-  hb_set_union (plan->_glyphset_gsub, input_glyphs_to_retain);
 
   _cmap_closure (plan->source, plan->unicodes, plan->_glyphset_gsub);
 
@@ -477,8 +488,6 @@ hb_subset_plan_create (hb_face_t	 *face,
   _populate_unicodes_to_retain (input->unicodes, input->glyphs, plan);
 
   _populate_gids_to_retain (plan,
-			    input->unicodes,
-			    input->glyphs,
 			    !input->drop_tables->has (HB_OT_TAG_GSUB),
 			    !input->drop_tables->has (HB_OT_TAG_GPOS),
 			    !input->drop_tables->has (HB_OT_TAG_GDEF));

util/hb-subset.cc

@@ -265,20 +265,26 @@ parse_text (const char *name G_GNUC_UNUSED,
 	    GError    **error G_GNUC_UNUSED)
 {
   subset_main_t *subset_main = (subset_main_t *) data;
+  hb_bool_t is_remove = (name[strlen (name) - 1] == '-');
 
+  hb_set_t *unicodes = hb_subset_input_unicode_set (subset_main->input);
   if (0 == strcmp (arg, "*"))
   {
-    subset_main->add_all_unicodes ();
+    hb_set_clear (unicodes);
+    if (!is_remove)
+      hb_set_invert (unicodes);
     return true;
   }
 
-  hb_set_t *unicodes = hb_subset_input_unicode_set (subset_main->input);
   for (gchar *c = (gchar *) arg;
        *c;
        c = g_utf8_find_next_char(c, nullptr))
   {
     gunichar cp = g_utf8_get_char(c);
+    if (!is_remove)
       hb_set_add (unicodes, cp);
+    else
+      hb_set_del (unicodes, cp);
   }
   return true;
 }
@@ -674,6 +680,9 @@ subset_main_t::add_options ()
     {"text",		0, 0, G_OPTION_ARG_CALLBACK, (gpointer) &parse_text,			"Specify text to include in the subset", "string"},
     {"text-file",	0, 0, G_OPTION_ARG_CALLBACK, (gpointer) &parse_file_for<parse_text, false>,"Specify file to read text from", "filename"},
     {"unicodes",	0, 0, G_OPTION_ARG_CALLBACK, (gpointer) &parse_unicodes,		"Specify Unicode codepoints or ranges to include in the subset", "list of hex numbers/ranges"},
+    {"unicodes-",	0, 0, G_OPTION_ARG_CALLBACK, (gpointer) &parse_unicodes,		"Specify Unicode codepoints or ranges to include in the subset", "list of hex numbers/ranges"},
+    {"unicodes+",	0, 0, G_OPTION_ARG_CALLBACK, (gpointer) &parse_unicodes,		"Specify Unicode codepoints or ranges to include in the subset", "list of hex numbers/ranges"},
+
     {"unicodes-file",	0, 0, G_OPTION_ARG_CALLBACK, (gpointer) &parse_file_for<parse_unicodes>,"Specify file to read Unicode codepoints or ranges from", "filename"},
     {nullptr}
   };