eachvsforin.rb

require 'benchmark'
require 'forgery'

def new_ostruct_member(arg); end

def generate_data
  (0..1000).map do |i|
    { "k#{i}" => "v#{i}" }
  end.inject &:merge
end

@data = generate_data

@table = {}
@ntable = {}

Benchmark.bmbm do |x|
  x.report("old") { 1000.times {
    for k,v in @data
      @table[k.to_sym] = v
      new_ostruct_member(k)
    end
  } }

  x.report("new") { 1000.times {
    @data.each do |key, value|
      @ntable[key.to_sym] = value
      new_ostruct_member key
    end
  } }
end

output.txt

Rehearsal ---------------------------------------
old   1.660000   0.010000   1.670000 (  1.679527)
new   1.070000   0.000000   1.070000 (  1.102152)
------------------------------ total: 2.740000sec

          user     system      total        real
old   1.650000   0.010000   1.660000 (  1.667362)
new   1.040000   0.010000   1.050000 (  1.064195)

require 'benchmark'

def new_ostruct_member(arg); end

def generate_data
  (0..1000).map do |i|
    {"#{'k' * rand(100)}#{i}" => "#{'v' * rand(100)}#{i}"}
  end.inject &:merge
end

hash = generate_data


report = Benchmark.bmbm do |x|
  x.report("old") { 1000.times {
    @table = {}
    if hash
      for k,v in hash
        @table[k.to_sym] = v
        new_ostruct_member(k)
      end
    end
  } }

  x.report("new") { 1000.times {
    @ntable = {}
    hash.each do |key, value|
      @ntable[key.to_sym] = value
      new_ostruct_member key
    end if hash
  } }
end

puts "Is new result equal to old result? #{@table == @ntable}"
puts "New is #{report.map(&:to_s).map(&:split).map(&:last).map(&:to_f).inject(:/) * 100 - 100}% faster"

I added in the if and @table to best mimic the benchmarked method, and because they're relatively the same.