Benchmarking if / else with kind_of? against case / in to match two values based on their classes. if / else + kind_of? appears to be 4x faster.

gistfile1.txt

ruby 3.2.4 (2024-04-23 revision af471c0e01) [x86_64-linux]
Warming up --------------------------------------
           if / else   888.000 i/100ms
           case / in   259.000 i/100ms
Calculating -------------------------------------
           if / else      8.856k (± 0.9%) i/s -     44.400k in   5.013974s
           case / in      2.560k (± 3.0%) i/s -     12.950k in   5.064976s

if_else_vs_case_in_benchmark.rb

#!/usr/bin/env ruby

require 'benchmark/ips'

class Foo
end

class Bar
end

value1 = Foo.new
value2 = Bar.new

Benchmark.ips do |b|
  n = 1_000

  b.report('if / else') do
    n.times do
      result = if (value1.kind_of?(Foo) && value2.kind_of?(Bar))
                 true
               else
                 false
               end
    end
  end

  b.report('case / in') do
    n.times do
      result = case [value1, value2]
               in [Foo, Bar]
                 true
               else
                 false
               end
    end
  end
end

It seems to be only 2.7 times in Ruby 3.3, progress! But I think with the two arrays created on every iteration in case it’s hard to match the if direct method calls…

(I dropped the n.times iterations and added a b.compare! call for the below results.)

ruby 3.3.3 (2024-06-12 revision f1c7b6f435) [arm64-darwin23]
Warming up --------------------------------------
           if / else     1.343M i/100ms
           case / in   485.319k i/100ms
Calculating -------------------------------------
           if / else     13.398M (± 0.4%) i/s -     67.140M in   5.011198s
           case / in      4.919M (± 1.5%) i/s -     24.751M in   5.032843s

Comparison:
           if / else: 13398124.2 i/s
           case / in:  4919093.6 i/s - 2.72x  slower

Ah yes, interstingly after doing

match = [Foo, Bar].freeze
# …
    case [value1, value2]
    in match
      true
    else
      false
    end

The results are almost the same:

ruby 3.3.3 (2024-06-12 revision f1c7b6f435) [arm64-darwin23]
Warming up --------------------------------------
           if / else     1.340M i/100ms
           case / in     1.307M i/100ms
Calculating -------------------------------------
           if / else     13.373M (± 0.6%) i/s -     67.014M in   5.011276s
           case / in     13.129M (± 0.5%) i/s -     66.674M in   5.078370s

Comparison:
           if / else: 13373168.5 i/s
           case / in: 13129463.4 i/s - 1.02x  slower