From: Paul Brannan <pbrannan@...>
Date: 2001-10-10T03:37:56+09:00
Subject: [ruby-talk:22332] Re: include vs. extend

On Wed, 10 Oct 2001, Yukihiro Matsumoto wrote:

> There's no case that the latter is faster, I think.
>
> 							matz.

The latter is faster in the following case (Ruby 1.6.5):

N = 1000000

module M1; def foo; end; end
module M2; def bar; end; end
module M3; def xyz; end; end
module M4; def abc; end; end

class Foo
  include M1
  include M2
  include M3
  include M4
end

class Bar
  def initialize
    extend M1
    extend M2
    extend M3
    extend M4
  end
end

start = Time.now
x = Foo.new
N.times { x.foo; x.bar; x.xyz; x.abc; }
puts Time.now - start

start = Time.now
x = Bar.new
N.times { x.foo; x.bar; x.xyz; x.abc; }
puts Time.now - start

#=> 3.197413 (include)
#=> 3.15327  (extend)

It is not much faster, but it is consistently faster.

Actually, now that I give it a second look, I am not sure that my tests
are correct.

1) The above test, with the tests in reverse order (extend is faster):
#=> 3.149055 (extend)
#=> 3.195228 (include)

2) The above test, with GC.start surrounding (outside of) each test
(include is faster):
#=> 3.141431 (include)
#=> 3.189973 (extend)

3) The above test, with GC.start surrounding each test, and the tests in
reverse order (extend is faster):
#=> 3.158805 (extend)
#=> 3.168021 (include)

So in 3/4 cases, extend is (consistently) faster.  Why is this?

Paul