From: Albert Wagner <alwagner@...>
Date: 2002-09-06T12:52:28+09:00
Subject: Re: Ruby aesthetics

On Thursday 05 September 2002 09:45 pm, dblack@candle.superlink.net wrote:
> Hi --
>
> On Fri, 6 Sep 2002, Albert Wagner wrote:
> > On Thursday 05 September 2002 12:08 pm, Christian Szegedy wrote:
> > <snip>
> >
> > > What I really dislike in Ruby, that the data members are seen
> > > by the child classes. This is a very serious issue, and a claer
> > > violation of object-orientation.
> > >
> > > I think Ruby would be a much better language if the instance
> > > variables would only live per class and not per hierarchy...
> >
> > I don't understand what is bad about this.  I typically depend on
> > instance variables in the parent(s) being directly accesssible.
>
> I think I understand the objection (thanks in large part to Holden's
> explaining it to me on irc).  As I understand it, the problem is that
> if you do this:
>
>   class A
>     def initialize(x)
>       @thing = x
>     end
>
>     def talk
>       puts "My thing is #{@thing}."
>     end
>   end
>
>   class B < A
>     def talk
>       puts "I have a thing which is #{@thing}."
>     end
>   end
>
> and then class A is rewritten so that @thing is renamed to @item, B
> will break.
>
> I've tended to see instance variables as part of the... well, I was
> going to say the interface, but that's wrong... part of the stuff
> you're allowed to know about.  But I'm seeing that that's pretty
> non-robust.

I suppose what is needed is a good refactoring browser for Ruby.  Then, in the 
example above, all references to @thing would be automaticallly renamed in 
progeny.  However, speaking for myself, I don't trust the parent's accessors 
to not cause side effects.  I view instance variables, like methods, as what 
a subclass inherits.  I design and maintain with this is mind.  When I change 
the name of a superclass instance variable, I know to change it in it's 
subclasses also.  Standard operating procedure. 


>
>
> David