From: Robert Dober <robert.dober@...>
Date: 2008-04-19T01:33:28+09:00
Subject: Re: Naming question

On Fri, Apr 18, 2008 at 6:04 PM, Sean O'Halpin <sean.ohalpin@gmail.com> wrote:
> On Fri, Apr 18, 2008 at 12:10 PM, Rick DeNatale <rick.denatale@gmail.com> wrote:
>  >
>  >  Which, I believe, first showed up in Rails sometime in 2006.
>  >
>  >  --
>  >  Rick DeNatale
>  >
>
>  Florian Gross published a version in Jan 2005:
>  http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-core/4311
module Enumerable
   def group_by
     result = Hash.new { |h, k| h[k] = Array.new }
     self.each do |item|
       group = yield(item)
       result[group] << item
     end
     return result
   end
end

This is the code of the post linked to above, as one can see it works,
not as mine above :)
but this natural way to write it is quite inefficient as the Facets
version shows. I have benchmarked them and to my surprise and that of
many others I guess the Facets implementation is almost twice as fast
- for many different keys at least

result = {}
each do | ele|
   ( result[ yield(ele) ] ||= [] ) << ele
end
result.

Does anybody have the faintest idea why Hash.new{|h,k| ... } is so inefficient?

BTW I am quite happy that #group_by is quasi a standard name for this.

Cheers
Robert
-- 
http://ruby-smalltalk.blogspot.com/

---
Whereof one cannot speak, thereof one must be silent.
Ludwig Wittgenstein