From: Rick DeNatale <rick.denatale@...>
Date: 2008-07-04T02:41:47+09:00
Subject: Re: pseudo-randomize an array in a consistent order

------=_Part_810_1372816.1215107093518
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

On Thu, Jul 3, 2008 at 12:59 PM, Max Williams <toastkid.williams@gmail.com>
wrote:

> ThoML wrote:
> > On Jul 3, 4:56 pm, Max Williams <toastkid.willi...@gmail.com> wrote:
> >> Does anyone know how to pseudo-randomize an array (eg with a seed) so
> >> that you get the same order every time you do it?
> >
> > ruby19 has a Array#shuffle method:
> > irb(main):001:0> Kernel.srand(1); [1,2,3,4,5].shuffle
> > => [2, 4, 1, 5, 3]
> > irb(main):002:0> Kernel.srand(1); [1,2,3,4,5].shuffle
> > => [2, 4, 1, 5, 3]
>
> I'm only on 1.8.6 (and not likely to change soon since we'd want to do
> all our work servers as well), but out of interest, do you know how
> shuffle compares to .sort_by{rand} (thanks rob!) in terms of efficiency?


It looks like it should be faster for larger arrays.  There are actually two
methods Array#shuffle! and Array#shuffle

Here's a Ruby translation of the C code:

class Array
  def shuffle!
      i = length - 1
      while (i > 0)
           j = rand(i)
           # Swap self[i] and self[j]
           # This might be more elegant using a parallel assignment, but
           # parallel assignment generates a temporary array, so this is
           # a little closer to the C code
           tmp = self[i]
           self[i] = self[j]
           self[j] = self[i]
           i -= 1
      end
      self
  end

  def shuffle
      self.dup.shuffle!
  end
end

So shuffle is O(n) whereas any sort is going to be O(>n)


-- 
Rick DeNatale

My blog on Ruby
http://talklikeaduck.denhaven2.com/

------=_Part_810_1372816.1215107093518--