jruby solution to histogram coding challenge

bigdata.rb

require 'thread'

class Chunk
  def initialize(file, size, offset)
    @handle = File.open(file)
    @handle.pos = offset
    @string = @handle.read(size)
    if @string.index("\n") < @string.index(',')
      look_back(4)
    end
  end

  def each_age
    offset = 0
    while pos = @string.index(',', offset)
      if eol = @string.index("\n", pos)
        yield @string[pos + 1..eol - 1]
        offset = eol + 1
      else
        offset = pos + 1
      end
    end
  end

  def update_results(results)
    offset = 0
    while pos = @string.index(',', offset)
      if eol = @string.index("\n", pos)
        age = @string[pos + 1..eol - 1]
        results[age] += 1
        offset = eol + 1
      else
        offset = pos + 1
      end
    end
  end
    
  def look_back(bytes)
    @handle.seek(- @size - bytes, IO::SEEK_CUR)
    @string = @handle.read(bytes) + @string
  end
end

THREAD_COUNT = 10
chunk_size = File.size(ARGV[0]) / THREAD_COUNT

threads = []
results = []
THREAD_COUNT.times do |i|
  results[i] = Hash.new.tap{|h| h.default = 0 }
  threads << Thread.new do
    chunk = Chunk.new(ARGV[0], chunk_size, i * chunk_size)
    
    # this slows it down to 19s for some infernal reason
#     chunk.each_age do |age|
#       results[i][age] += 1
#     end
    
    chunk.update_results(results[i])
  end
end
threads.each{|t| t.join}

histogram = results.inject{|memo,hash| memo.merge(hash) {|_,a,b| a + b}}

p histogram