benchmark_hashes.rb

require 'benchmark'
 

module Helper1
  def twitterized_type(type)
    case type
      when :alert
        "warning"
      when :error
        "error"
      when :notice
        "info"
      when :success
        "success"
      else
        type.to_s
    end
  end
  module_function :twitterized_type
end

module Helper2
  def twitterized_type(type)
    {
      alert:   'warning',
      error:   'error',
      notice:  'info',
      success: 'success'
    }.fetch(type, type.to_s)
  end  
  module_function :twitterized_type
end

module Helper3
  def twitterized_type(type)
    {
      alert:   'warning',
      error:   'error',
      notice:  'info',
      success: 'success'
    }[type] || type.to_s
  end  
  module_function :twitterized_type
end


n = 500000

Benchmark.bm(7) do |x|
  x.report('case statement')  { n.times { |i| Helper1.twitterized_type(:success) } }
  x.report('hash lookup   ')  { n.times { |i| Helper2.twitterized_type(:success) } }
  x.report('hash lookup 2 ')  { n.times { |i| Helper3.twitterized_type(:success) } }

  x.report('when not exist: case statement')  { n.times { |i| Helper1.twitterized_type(:non_existent) } }
  x.report('when not exist: hash lookup   ')  { n.times { |i| Helper2.twitterized_type(:non_existent) } }
  x.report('when not exist: hash lookup 2 ')  { n.times { |i| Helper3.twitterized_type(:non_existent) } }
end

You need to create the hash just once, check out my fork, it is almost the same speed, but surprisingly still not as fast as the case.

@amalagaura Thanks. Of course, I should have memoized the hash (or used a constant). It makes a significant difference when I run the benchmark again. I think case statements will always be faster – there is still one fewer Hash object being created.