From: Matthew Smillie <M.B.Smillie@...>
Date: 2006-06-20T01:36:29+09:00
Subject: Re: Is it more convenient to have a random_each in ruby stdlib?

> I'm not too keen on #shuffle having a dual function as both a  
> shuffler and an iterator -- I'd be more comfortable using  
> `ary.shuffle.each'. That would also allow a major simplification of  
> the implementation:
>
>   class Array
>     def shuffle
>       values_at(*(0..size-1).to_a.shuffle!)
>     end
>   end


I agree with you about 'arr.shuffle.each' being better semantically,  
but I was worried about efficiency; since the thread started out with  
concerns about shuffling really big arrays, I wanted to avoid  
creating an intermediate copy of the array.  I didn't really think  
about it that hard, though, because thinking about it now, I realise  
an intermediate copy isn't all that bad (being just references), so  
I'm not sure the savings would matter all that much.

Anyway, Array's just a special case: I'd say that it belongs in  
Enumerable along with sort, for which I'd grab your implementation.

module Enumerable
   def shuffle
     entries.values_at(*(0...entries.size).entries.shuffle!)
   end
end

Though this still relies on Array#shuffle!, which I can't say is a  
good or bad thing.  Bit of a moot question unless someone does decide  
to put it into core or stdlib, though.

matthew smillie.