lib/bychar.rb in bychar-1.0.1 vs lib/bychar.rb in bychar-1.1.0

@@ -1,23 +1,26 @@
 # -*- encoding : utf-8 -*-
 require 'stringio'
 
 module Bychar
-  VERSION = '1.0.1'
+  VERSION = '1.1.0'
 
   # Default buffer size is 512k
   DEFAULT_BUFFER_SIZE = 512 * 1024
   
+  # Gets raised when you have exhausted the underlying IO
+  class EOFError < RuntimeError  #:nodoc: all
+  end
+  
   # This object helps you build parsers that parse an IO byte by byte without having to
   # read byte by byte.
   # Reading byte by byte is very inefficient, but we want to parse byte by byte since
   # this makes parser construction much easier. So what we do is cache some chunk of the
   # passed buffer and read from that. Once exhausted there will be some caching again,
   # and ad infinitum until the passed buffer is exhausted
   class Reader
     
-
     def initialize(with_io, buffer_size = DEFAULT_BUFFER_SIZE)
       @io = with_io
       @bufsize = buffer_size
       cache
     end
@@ -36,12 +39,22 @@
     # and from the internal cache buffer (checks whether there is anything that
     # can be retreived using read_one_byte)
     def eof?
       (@buf && @buf.eos?) && @io.eof?
     end
-
+    
+    # Since you parse char by char, you will be tempted to do it in a tight loop
+    # and to call eof? on each iteration. Don't. Instead. allow it to raise and do not check.
+    # This takes the profile time down from 36 seconds to 30 seconds for a large file.
+    def read_one_byte!
+      cache if @buf.eos?
+      raise EOFError if @buf.eos?
+      
+      @buf.getch
+    end
+    
     private
-
+    
     def cache
       data = @io.read(@bufsize)
       @buf = StringScanner.new(data.to_s) # Make nil become ""
     end
   end
\ No newline at end of file