Arrow CSV bench

arrow-csv.rb

require 'bundler/inline'

gemfile do
  source 'https://rubygems.org'

  gem 'benchmark-ips'
  gem 'red-arrow'
end

require 'arrow'
require 'benchmark/ips'

def benchmark_sum(filename)
  f = Arrow::Function.find('sum')

  Benchmark.ips do |ips|
    ips.report('Arrow') { f.execute([Arrow::Table.load(filename)['revenue'].data]).value.value }
    ips.report('File')  { c = 0; File.foreach(filename) { |row| c += row.split(',').last.to_f } }
    ips.report('CSV')   { c = 0; CSV.foreach(filename) { |row| c += row[1].to_f } }
    ips.compare!
  end
end

benchmark_sum('data10k.csv')
# Warming up --------------------------------------
#                Arrow   132.000  i/100ms
#                 File    46.000  i/100ms
#                  CSV    22.000  i/100ms
# Calculating -------------------------------------
#                Arrow      1.373k (±12.8%) i/s -      6.732k in   5.024702s
#                 File    722.441  (± 2.6%) i/s -      3.634k in   5.033618s
#                  CSV    211.504  (± 5.7%) i/s -      1.078k in   5.114960s
#
# Comparison:
#                Arrow:     1372.6 i/s
#                 File:      722.4 i/s - 1.90x  (± 0.00) slower
#                  CSV:      211.5 i/s - 6.49x  (± 0.00) slower


benchmark_sum('data10M.csv')
# Warming up --------------------------------------
#                Arrow     1.000  i/100ms
#                 File     1.000  i/100ms
#                  CSV     1.000  i/100ms
# Calculating -------------------------------------
#                Arrow      1.769  (±56.5%) i/s -      8.000  in   5.238670s
#                 File      0.069  (± 0.0%) i/s -      1.000  in  14.407080s
#                  CSV      0.016  (± 0.0%) i/s -      1.000  in  62.368109s
#
# Comparison:
#                Arrow:        1.8 i/s
#                 File:        0.1 i/s - 25.48x  (± 0.00) slower
#                  CSV:        0.0 i/s - 110.31x  (± 0.00) slower

thanks