From: Ben Tilly <ben_tilly@...>
Date: 2001-01-09T20:52:14+09:00
Subject: [ruby-talk:8917] Re: Modules and mixins

"Christoph Rippel" <crippel@primenet.com> wrote:
>
> > -----Original Message-----
> > From: Ben Tilly [mailto:ben_tilly@hotmail.com]
> > Sent: Monday, January 08, 2001 08:26 PM
> > To: ruby-talk ML
> > Subject: [ruby-talk:8897] Re: Modules and mixins
>...
> > If length is the primary issue (yes, I renamed the method):
>
>The length was not the issue on the python ml. The original author of the 
>MRange
>class - he called  it the ``Dice class '' if I remember correctly - just 
>wanted to
>demonstrate that the block & closure implementation are definite highlights 
>of ruby.
>I was a little amused about a response on the python ml, arguing that a 
>rather
>beautiful but also long winded iterator implementation which explicitly 
>calculates
>the position of an individual element is a conceptually superior solution. 
>Even so
>``block  iterators'' probably take a while to get used to they are 
>preferable over
>the ``position  math'' of something like ...
>
Say no more.  Explicit calculations open up room for all
sorts of potential mistakes.
>
>class Combs
>     def initialize(m,n,p=1)
>         raise ArgumentError if m - n < p
>         @m = m
>  	  @n = n
>         @p = p
>      end
>      def each
>        __e(@m,@n) { |l| yield l }
>      end
>private
>      def __e(i,j)
>          if j  <=  @p
>             i.downto (1) { |l|  yield [l] }
>           else
>             i.downto (j) \
>             { |l|  __e(l-@p,j-@p) { |r|  yield [l]+r }}
>           end
>       end
>end
>
>Combs.new (8, 5, 2).each \
>{| l|  puts l.join " "}
>
>
>The implementation of the original Dice version was actually closer to your 
>solution
>and if you throw in your argument separation trick

Not much of a trick.  I just intended to use argument
processsing to do what would otherwise be mechanical in
my original Perl version.

> >def bind_loops(fn, my_range, *more_args)
>                         ^^^^^^^^^^^
>...
> >        bind_loops(bound_fn, *more_args) :
>                                ^^^^^^^^
> >end
>
>could be rewritten in a rather compact form
>
>def  each (arg,*more_args)
>      ( [] != args ) ?
>       	arg.each {|l| each(*more_args) {|r| yield [l]+ r }} :
>       	arg.each { |r| yield [r] }
>end
>
>each( 1..2, 'a'..'c', 'A'..'D') \
>{ |args|   puts args.join " " }
>
Huh.  This is what I said I had not played with.

Looking at this "feels" like an inside-out twist on
my solution.  I think I like the pattern of this one
better, just a tad more direct.  However I would
prefer different names:

    def  each_comb (arg,*more_args)
      ( [] != more_args ) ?
        arg.each {|l| each_comb(*more_args) {|r| yield [l]+ r }} :
        arg.each { |r| yield [r] }
    end

    each_comb( 1..2, 'a'..'c', 'A'..'D') \
      { |args|   puts args.join " " }

Why?  Because to me "each" is an already well-defined
concept and each combination doesn't not seem to me to
be at all the same.  So the subtly different names
make that intent a lot clearer.  (Well for me at least.)

Cheers,
Ben
_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com