From: Adriano Ferreira <a.r.ferreira@...>
Date: 2005-02-22T03:52:23+09:00
Subject: Re: YAML and hash

> You're right.  This should throw an error or a warning (according the
> YAML specification).  This a simple fix, but cannot be done in Ruby.

I've spent some time trying to write code to make such a thing work.
I've just found that I don't know nothing about Ruby.

I wrote the following Hash subclass that I thought could be forced
into the YAML parser (maybe through a simple explicit header like '---
!ruby/hash:AngryHash' or via a directive to the YAML parser locally
use AngryHash'es rather than primitive Hash'es).

class AngryHash < Hash

  def []=(key, value)
    if has_key?(key)
      raise "not again: duplicate key '" + key.to_s + "'"
    end
    super(key, value)      
  end

  def AngryHash.[](pairs)
    h = new
    pairs.each_pair {|k,v| h[k] = v}
    h
  end

end

Well, the rationale is not to allow pairs which keys that already
appeared (AngryHash#[]= in a pair-by-pair basis) and I guessed that
something like Hash[k1 => v1, ... ] is used by YAML parser when
constructing map-like objects. This is a dead-end. If the parser uses
this form, pairs with repeated keys were already trashed (except for
the last one) and this code is good for nothing.

The fix why_ mentioned has to do with this behaviour? If I tried to
subclass YAML::Pairs, would it work?

Regards,
Adriano.