From: Kevin Ballard <kballard@...>
Date: 2005-10-15T01:56:55+09:00
Subject: Re: Odd behaviour concerning procs.


Christophe Poucet wrote:
> Hello,
>
> I've noticed the following behaviour and I'm not certain if this is
> expected behaviour, however from the point of view of any other
> language, it certainly seems odd.
>
> Assume there is no constant with the name 'X' defined anywhere.
>
>   X => NameError: uninitialized constant X
>   proc{|X| true}.call(15) => true
>   X => 15
>   proc{|X| true}.call(15) # true, warning: already initialized constant
> X
>
> I would presume that the X would be localized to the proc.  I know that
> for normal variables, the variable in the calling scope is only changed
> if it already exists, but in this case, the constant X does not even
> exist. (And when it does exist, is it really preferred behaviour that a
> proc will change a constant?)

This doesn't seem odd to me. I would read that as the equivalent of

  proc{|a| self.class.const_set(:X, a); true}.call(15)

and that certainly defines the constant. Constants are not like local
variables - they don't have scopes. Instead, they exist on objects,
like class and instance variables. So there's no way a constant could
only exist inside the proc and remain undefined outside it.