From: Robert Klemme <shortcutter@...>
Date: 2009-06-20T01:00:30+09:00
Subject: Re: hash of hash?

On 19.06.2009 17:42, Jeroen Lodewijks wrote:
> Thanks all for the replies, very clear.
> 
> I think I will implement an accessor to the hash, I come from a Perl 
> world and would build the hash in a seperate routine (to hide the 
> initialisation) that returns a reference to the hash.

If you create a new class then the constructor is just sufficient - you 
do not need another method that hides away invocation of YourClass.new.

> Would it be possible to make Hashhash inherit from Hash (by using < 
> Hash) and be able to use the [][] operator immediately in a way?

Actually, you do not need to create another class (see my earlier 
example).  For that it might make sense to create a method that returns 
a new instance with the initialization I have shown.

If you lack explanation: the construct Hash.new {|h,k| ...} creates a 
new Hash instance with its default_proc set to the block you provide. 
This block is invoked whenever there is a key miss.  In my case I just 
built it in a way as to create another Hash on the fly with the same 
default_proc and stuff it in the other Hash.  The default_proc 
propagation ensures you can arbitrarily nest Hashes.  Of course, you can 
also do something like this if you know you just need one level of 
nested Hashes

ha = Hash.new {|h,k| h[k] = {}}

Question is whether you want to bother to create a class for this.  IMHO 
this makes only sense if you want to put additional logic into it (e.g. 
checking elements that are placed in the Hash).

Kind regards

	robert

-- 
remember.guy do |as, often| as.you_can - without end
http://blog.rubybestpractices.com/