[ruby-talk:16937] Re: rb_define_class() vs Class.new()

> |Guy,
> I'm sorry that I'm not Guy.  But I believe you admit me.
no problem :)

> |What would be the semantics of
> |  a = Class.new()
> |  A, B = a, a
> |  p a.name
> |?
> |(without giving any implementation detail, pls)
> 
> Unnamed class object is referenced from two constants A and B.  When
> a.name is called (or whenever class name is required), Ruby looks for
> the class name via internal constant table.  Since constant table is a
> hash, it's hard to tell name would be either A or B.

1) I asked to leave implementation details out of this. I think a hash
   is such a detail :)
2) A, B = a, a sets a.name() to return B (on my machine?), however
   A = a; B = a sets a.name() to return A, so I probably don't
   understand which hash you mean anyway.

> |What looks nicer:
> | - Object.const_set("A", Class.new)
> | - eval("class A; end")
> | - Class.new("A")
> | - I don't use classes
>   - A = Class.new
> 
> You can access the class via A whatever its name is.  Class name is
> just a name.

The name of a class is _the_ name of that class.
I do not understand what you mean with ``just a name''.

Or is it that a class can have many names?
Like, a name is a reference to a class, instead of a property of the
class. In that case, why can't I ask for all references/names?

I still don't get /why/ this is done like it is; optimization? memory
savings?

It still feels like a side-effect
  abc = Class.new()  # abc has no name
  A = abc            # create constant A, do not touch abc in any way and
                     #   suddenly abc has a name!
  xyz = Class.new()  # xyz has no name
  A = xyz            # Warning! Already initialized constant A!
                     #   and both abc and xyz are classes with the
                     #   name A (though A.new() is still uniquely
                     #   defined, of course)

I even like the fact that you say A.new() instead of new A(); probably
one of the few useful ways to guarantee some consistency is through
the use of global references; so perhaps the concept ``name'' is not
as useful as in Java anyway. If so, why not drop the notion entirely?

Bye,
Kero.

PS: I'm rambling, I'm confused and I'm curious.

+--- Kero ------------------------------ [e-mail] ---+
|  Don't split your mentality without thinking twice       |
|                          Proud like a God -- Guano Apes  |
+--- M38c ------------------ http://huizen.dds.nl/~kero ---+