Benchmark for determining duplicate positions

find-duplicates.rb

require 'benchmark'

REP = 100_000

dat = [1, 2, 3, 2, 4, 5, 4, 4]

Benchmark.bmbm 25 do |x|
  x.report "Jan low level" do
    REP.times do
      dups = {}

      dat.each_with_index do |value, i|
        (i + 1).upto dat.length - 1 do |j|

          if dat[j] == value
            dups[value] = [i] if dups[value].nil?
            dups[value] << j
            break
          end
        end
      end

      #p dups; exit
    end
  end

  x.report "Jan multi Hash" do
    REP.times do
      dups = dat.each_with_index.group_by(&:first).inject({}) do |result, (val, group)|
        next result if group.length == 1
        result.merge val => group.map {|pair| pair[1]}
      end

      #p dups; exit
    end
  end

  x.report "Sam" do
    REP.times do
      dups = Hash.new {|h,k| h[k] = []}

      dat.each_index {|i| dups[dat[i]] << i unless 1 == dat.count(dat[i])}

      #p dups; exit
    end
  end

  x.report "Lars" do
    REP.times do
      dups = dat.each_with_index.reduce({}) { |hash, (item, index)|
        hash[item] = (hash[item] || []) << index
        hash
      }.select { |key, value|
        value.size > 1
      }

      #p dups; exit
    end
  end

  x.report "Lars simpler" do
    REP.times do
      dups = dat.each_with_index.reduce({}) { |hash, (item, index)|
        (hash[item] ||= []) << index
        hash
      }.select { |key, value|
        value.size > 1
      }

      #p dups; exit
    end
  end

  x.report "Lars faster?" do
    REP.times do
      dups = dat.each_with_index.reduce({}) { |hash, (item, index)|
        (hash[item] ||= []) << index
        hash
      }.delete_if { |key, value|
        value.size == 1
      }

      #p dups; exit
    end
  end

  x.report "Greg" do
    REP.times do
      dups = {}

      dat.each_with_index do |key, indx|
        if !dups.has_key?(key)
          dups.store(key, [])
        end

        dups[key].push(indx)
      end

      dups.delete_if {|k,v| v.size == 1}

      #p dups; exit
    end
  end

  x.report "RK #[]=" do
    REP.times do
      dups = {}

      dat.each_with_index do |val, idx|
        (dups[val] ||= []) << idx
      end

      dups.delete_if {|k,v| v.size == 1}

      #p dups; exit
    end
  end

  x.report "RK #fetch" do
    REP.times do
      dups = {}

      dat.each_with_index do |val, idx|
        dups.fetch(val) { dups[val] = [] } << idx
      end

      dups.delete_if {|k,v| v.size == 1}

      #p dups; exit
    end
  end

  x.report "RK / Ryan #default_proc" do
    REP.times do
      dups = Hash.new {|h,k| h[k] = []}

      dat.each_with_index do |val, idx|
        dups[val] << idx
      end

      dups.delete_if {|k,v| v.size == 1}

      #p dups; exit
    end
  end

  x.report "RK 1 liner" do
    REP.times do
      dups = dat.each_with_index.group_by(&:first).delete_if {|k,v| v.size == 1}.each {|k,v| v.map! &:last}

      #p dups; exit
    end
  end
end

output.txt

$ ruby19 -v
ruby 1.9.3p125 (2012-02-16 revision 34643) [i686-linux]
$ ruby19 b.rb 
Rehearsal -------------------------------------------------------------
Jan low level               1.320000   0.000000   1.320000 (  1.373991)
Jan multi Hash              2.250000   0.000000   2.250000 (  2.395521)
Sam                         1.630000   0.000000   1.630000 (  2.000988)
Lars                        1.590000   0.000000   1.590000 (  1.675986)
Lars simpler                1.550000   0.000000   1.550000 (  1.616905)
Lars faster?                1.400000   0.000000   1.400000 (  1.467628)
Greg                        1.180000   0.000000   1.180000 (  1.210565)
RK #[]=                     0.950000   0.000000   0.950000 (  1.027279)
RK #fetch                   1.170000   0.000000   1.170000 (  1.219644)
RK / Ryan #default_proc     1.460000   0.000000   1.460000 (  1.505699)
RK 1 liner                  1.560000   0.010000   1.570000 (  1.692262)
--------------------------------------------------- total: 16.070000sec

                                user     system      total        real
Jan low level               1.360000   0.000000   1.360000 (  1.425580)
Jan multi Hash              2.340000   0.000000   2.340000 (  2.417529)
Sam                         1.600000   0.000000   1.600000 (  1.733025)
Lars                        1.560000   0.000000   1.560000 (  1.610749)
Lars simpler                1.530000   0.000000   1.530000 (  1.675452)
Lars faster?                1.390000   0.000000   1.390000 (  1.445232)
Greg                        1.170000   0.000000   1.170000 (  1.205904)
RK #[]=                     0.940000   0.000000   0.940000 (  0.984368)
RK #fetch                   1.170000   0.010000   1.180000 (  1.243567)
RK / Ryan #default_proc     1.450000   0.000000   1.450000 (  1.507328)
RK 1 liner                  1.560000   0.000000   1.560000 (  1.603388)