From: "David A. Black" <dblack@...>
Date: 2005-09-22T02:16:13+09:00
Subject: Re: Set#== ain't..

Hi --

On Thu, 22 Sep 2005, Gary Shea wrote:

> While writing an acceptance test, I needed to compare two sets of sets
> of integers:
>
> require 'set'
> a = Set.new([Set.new([3,4]), Set.new([5,6])])
> b = Set.new([Set.new([3,4]), Set.new([5,6])])
> puts #{a == b}\n"
>
> prints out 'false'.
>
> Apparently the built-in Set library uses Hash#includes? to determine if
> an object is in the rhs set.  This method appears to use object id, so
> set equality is based on set member object id.
>
> A naive #==(set) that works the way I think it should is pretty simple:
>
> def Set.my_eql?(set)

You'd want that to be an instance method, not a class method:

   class Set
     def my_eql?(set)

>  ...
>  # this is the only part of my_eql? that differs from #==(set),
>  # replacing the line 'set.all? { |o| hash.include?(o) }'
>
>  set.all? { |o|
>    @hash.each_value do |my_o|
>      if my_o == o
>        last true
>      end
>      false
>    end
>  }
> end

You could just do:

     set.all? {|x| any? {|y| x == y } }

> I'm sure this has come up before... is there a better solution?  This
> one is fine for my purposes, but I would be nice there was a packaged
> solution out there.

It did come up, quite recently I think.  I can't remember what the
consensus was, if any.


David

-- 
David A. Black
dblack@wobblini.net