Benchmarking `respond_to?` and `rescue NoMethodError` for method delegation in Ruby.

gistfile1.rb

require 'benchmark'

# Say you have a Foo class and you want instances of this class to delegate unknown
# methods to two objects, first to @object_1, then to @object_2.

class Foo
  def initialize
    @object_1 = nil
    @object_2 = ''
  end
end

@foo = Foo.new

# Which is faster, first calling respond_to? on @object_1, or calling the method
# directly, then rescue NoMethodError before calling it on @object_2?

Benchmark.benchmark Benchmark::CAPTION, 11 do |bm|
  bm.report 'respond_to?' do
    class Foo
      def method_missing(*args, &block)
        (@object_1.respond_to?(args[0]) ? @object_1 : @object_2).send(*args, &block)
      end
    end

    1_000_000.times { @foo.split }
  end

  bm.report 'rescue' do
    class Foo
      def method_missing(*args, &block)
        @object_1.send(*args, &block)
      rescue NoMethodError
        @object_2.send(*args, &block)
      end
    end

    1_000_000.times { @foo.split }
  end
end

__END__

Results:
                  user     system      total        real
respond_to?   0.870000   0.000000   0.870000 (  0.873963)
rescue       13.190000   0.590000  13.780000 ( 13.779372)

respond_to? FTW!