From: Intransition <transfire@...>
Date: 2009-11-05T23:34:42+09:00
Subject: Re: Hash#count



On Nov 4, 11:32 am, Paul Smith <p...@pollyandpaul.co.uk> wrote:
> On Wed, Nov 4, 2009 at 4:19 PM, Intransition <transf...@gmail.com> wrote:
> > I want to suggest Hash#count be defined such that it counts hash
> > values. A it stands (using Enumerable#count) it only can ever return 0
> > or 1.
>
> > Current:
>
> >  {:a=>1,:b=2,:c=>1}.count([:a,1]) #=> 1
> >  {:x=>1,:b=2,:c=>1}.count([:a,1]) #=> 0
> >  {:a=>1,:b=2,:c=>1}.count(1)      #=> 0
>
> {:a=>1,:b=2,:c=>1}.count{|x| x[1]==1}      #=> 2

Sure. And there other ways to do it too. My point is not that there is
a difficulty in achieving this behavior that must be addressed, but
that Hash#count (without using a block) is much less useful as
currently defined. My suggestion is simply to maximize it's utility
for the majority of usecases. This is not unlike what other Enumerable
Hash methods do. Consider Hash#include?. It has a Hash specific
definition whereby it checks only keys.

Let me put it another way. There is no point in counting Hash keys.
There is only ever one or none. So Hash#count should only count
values.

Now one might ask about counting all pairs with keys matching a
pattern, but such cases are going to be the exception rather then the
rule, and can be achieved almost as simply using other means, for
instance:

  hash.to_a.count{ |(k,v)|  ... }

or

  hash.keys.count{ |k| hash[k] =~ /.../ }

Oh, and b/c of enumerators, I think we can even do:

  hash.each.count{ |k,v| ... }

T.