Change TrueType subsetter to avoid generating characters that need to be escaped

If text is output into a content stream, it needs to be serialized
according to the PDF spec rules. For serializing strings HexaPDF has
chosen the literal string syntax because the hexadecimal string syntax
would need twice as many characters.

For literal strings there are four characters -- \r, (, ), \ -- that
need to be escaped. If any of theses characters appears in the string,
String#gsub! has to actually do some work modifying the string.

By making sure that the encoded text using a TrueType font doesn't
contain those characters we avoid this work.

This change saves about 31% of object allocations and about 5% runtime
on the raw_text 10x TrueType benchmark.

Author: gettalong

lib/hexapdf/font/true_type/subsetter.rb

@@ -63,6 +63,12 @@ def initialize(font)
         def use_glyph(glyph_id)
           return @glyph_map[glyph_id] if @glyph_map.key?(glyph_id)
           @last_id += 1
+          # Handle codes for ASCII characters \r, (, ) and \ specially so that they never appear in
+          # the output (PDF serialization would need to escape them)
+          if @last_id == 13 || @last_id == 40 || @last_id == 41 || @last_id == 92
+            @glyph_map[:"s#{@last_id}"] = @last_id
+            @last_id += 1
+          end
           @glyph_map[glyph_id] = @last_id
         end
 
@@ -107,7 +113,7 @@ def build_glyf_table
           locations = []
 
           @glyph_map.each_key do |old_gid|
-            glyph = orig_glyf[old_gid]
+            glyph = orig_glyf[old_gid.kind_of?(Symbol) ? 0 : old_gid]
             locations << table.size
             data = glyph.raw_data
             if glyph.compound?
@@ -134,7 +140,7 @@ def build_hmtx_table
           hmtx = @font[:hmtx]
           data = ''.b
           @glyph_map.each_key do |old_gid|
-            metric = hmtx[old_gid]
+            metric = hmtx[old_gid.kind_of?(Symbol) ? 0 : old_gid]
             data << [metric.advance_width, metric.left_side_bearing].pack('n2')
           end
           data
@@ -166,7 +172,10 @@ def build_maxp_table(nr_of_glyphs)
         # Adds the components of compound glyphs to the subset.
         def add_glyph_components
           glyf = @font[:glyf]
-          @glyph_map.keys.each {|gid| glyf[gid].components&.each {|cgid| use_glyph(cgid) } }
+          @glyph_map.keys.each do |gid|
+            next if gid.kind_of?(Symbol)
+            glyf[gid].components&.each {|cgid| use_glyph(cgid) }
+          end
         end
 
       end