improve text extraction of built in fonts

* if the font includes a difference table (allowing it to use non ASCII
  chars) then we need to do some extra work to extract the glyph width
  from the relevant AFM file

lib/pdf/reader/encoding.rb

@@ -111,6 +111,23 @@ def int_to_utf8_string(glyph_code)
       @string_cache[glyph_code] ||= internal_int_to_utf8_string(glyph_code)
     end
 
+    # convert an integer glyph code into an Adobe glyph name.
+    #
+    #     int_to_name(65)
+    #     => :A
+    #
+    # TODO: this needs to be expanded to return the appropriate name for standard
+    #       glyph codes in the encoding. 65 to :A, etc. At the moment it only
+    #       handles glyphs in the difference table
+    #
+    def int_to_name(glyph_code)
+      if @enc_name == "Identity-H" || @enc_name == "Identity-V"
+        nil
+      else
+        @differences[glyph_code]
+      end
+    end
+
     private
 
     def internal_int_to_utf8_string(glyph_code)

lib/pdf/reader/width_calculator/built_in.rb

@@ -3,6 +3,16 @@
 require 'afm'
 require 'pdf/reader/synchronized_cache'
 
+module AFM
+  # this is a monkey patch for the AFM gem. hopefully my patch will be accepted
+  # upstream and I can drop this
+  class Font
+    def metrics_for_name(name)
+      @char_metrics[name.to_s]
+    end
+  end
+end
+
 class PDF::Reader
   module WidthCalculator
 
@@ -28,6 +38,10 @@ def glyph_width(code_point)
         return 0 if code_point.nil? || code_point < 0
 
         m = @metrics.metrics_for(code_point)
+        if m.nil?
+          name = @font.encoding.int_to_name(code_point)
+          m = @metrics.metrics_for_name(name)
+        end
         m[:wx]
       end