Redundancy in method_missing

Rails
1

Looking through the rails core code, I see method_missing performed like
this a lot:

def method_missing(method, *args)
if block_given?
current_status.send(method, *args) { |*block_args|
yield(*block_args) }
else
current_status.send(method, *args)
end
end

Why not just write:

def method_missing(method, *args, &block)
current_status.send(method, *args, &block)
end

It the above implementation much faster or do they perform differently?

2

On Sep 26, 11:53 pm, Aryk G. [email protected]
wrote:

end

Why not just write:

def method_missing(method, *args, &block)
current_status.send(method, *args, &block)
end

Building the proc object (as implied by the use of &block) is more
expensive than just passing a block:

require ‘benchmark’
include Benchmark

def dummy()
yield
end
def dummy_proc(&block)
block.call
end

bm do |x|
x.report(‘with yield’){ 100000.times { dummy {y=1}}}
x.report(‘with &block’) { 100000.times { dummy_proc {y=1}}}
end

outputs (on a fairly elderly machine):

  user     system      total        real

with yield 0.420000 0.000000 0.420000 ( 0.464097)
with &block 2.340000 0.040000 2.380000 ( 2.584321)

Fred

3

Fred, thanks for that, I still can’t believe that block.call is slower
than yield.

All this time, I thought I was speeding thing up by passing in the
&block.

thanks a bunch!

4

Sent from my iPhone

On 30 Sep 2008, at 00:07, Aryk G. <rails-mailing-list@andreas-
s.net> wrote:

Fred, thanks for that, I still can’t believe that block.call is slower
than yield.

If my memory is correct, it’s not the block call (try removing the
yield/block.call from the benchmark) that is slow, but the coercing of
the block into a proc

Fred