README.md

Benchmarking ruby

The problem: Create a hash from an array of objects where the key is the object's #id and the value is the #stuff method.

Known/Suspected caveats:

• Garbage collection might be happening, making the results lie.
• Not exhaustive of ways to create a hash from an array
• Not exhaustive of ways to call Benchmark.ips with code to test. (no strings yet)
• Not able to manually change the number of iterations
• No GC instrumentation/profiling/timing
• This is completely unfair, as I broke the original pattern of calling #inject to artificially 'boost' iterations.

When I started working on this benchmark, I didn't notice any GC stop-the-world time, but I'm human. The fastest was to use #each_with_object but it was only a hair faster than #inject with assign and return.

I realize there are lots of other ways to merge, but the slowest by far is to actually call #merge instead of #merge!. I intend to rewrite this to be more fair, but that might be pending a fork of benchmark/ips from evan phoenix.

benchmark.rb

require 'benchmark/ips'

User = Struct.new(:id, :stuff)
a = Array.new(1000) { |i| User.new(i, stuff: rand(1000)) }

10.times do
  Benchmark.ips do |x|
    x.report('#inject return') do
      a.inject({}) { |memo, i| memo[i.id] = i.stuff; memo }
    end
    x.report('merge') do
      a.inject({}) { |memo, i| memo.merge(i.id => i.stuff) }
    end
    x.report('merge!') do
      a.inject({}) { |memo, i| memo.merge!(i.id => i.stuff) }
    end
    x.report('each_with_object') do
      a.each_with_object({}) { |i, memo| memo[i.id] = i.stuff }
    end

    x.report('zenspider') do
      Hash[a.map(&:to_a)]
    end

    x.report('pipework') do
      a.map(&:to_a).to_h
    end

    x.compare!
  end
end