reduce2.rb

#!/usr/bin/env ruby
class Reducer
  attr_accessor :key, :game_type, :user_id

  def initialize(key,value)
    @key = key

    #split the primary key to get user_id and game type:
    @game_type,@user_id = key.split("|")

    #to perform an on-the-fly average, we only need two variables:
    @count = 1
    @average = value.to_f
  end

  #on the fly averaging. We do NOT store the entire array !
  def accumulate(value)
    @average  = ( @count.to_f/(@count + 1) * @average ) + (value.to_f / (@count + 1) )
    @count += 1
  end

  # follow the expectations
  def output!
    puts [
      @user_id,
      @game_type,
      @average.round(1),
    ].join("\t")
  end

end

ARGF.each do |line|
  next if line.chomp.empty?

  # split the data
  new_key,value = line.chomp.split("\t").map(&:strip)

  #initialize if required
  @red ||= Reducer.new(new_key,value)

  # if the key is the same, then continue accumulating
  if new_key == @red.key
    @red.accumulate(value)

  # if the key is new, then first output current results, then instanciate a new reducer
  else
    @red.output!
    @red = Reducer.new(new_key,value)
  end
end
@red.output!