[ruby-talk:16939] Re: rb_define_class() vs Class.new()

Kero van Gelder <[e-mail]> writes:

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

Classes do not need to have names. It's a pure convenience used when
inspecting objects of that class, and when using things such as
Class#name.

When you say

   class Kero
      # ...
   end

it is not an essential part of Ruby's semantics that the class you
create is named 'Kero'. The class could equally well not have a
name. All that happens is that a constant is created, called Kero, and
that constant is set to point to the class object for the class you
just created. 'Kero' is _not_ the name of the class. 'Kero' is the
name of the constant that references the class.

When you say

  k = Kero.new

you are not using a class named Kero. Instead, you're simply sending
the message 'new' to the object referenced by the constant Kero. As
that constant references a class object, the statement does what you'd
expect.

  fred = Kero
  k = fred.new

Here we're assigned to a variable 'fred' whatever Kero referenced. As
that was a class, we can use fred.new to create new instances.

Creating anonymous classes is basically the same

  Dave = Class.new

Here Dave is a constant that points to an anonymous class.

  fred = Class.new

And here 'fred' points to a different class.


Now... to the naming issue.

Ruby classes do not need names: they are only there for output
purposes. So, matz decided to implement names by cheating: when you
ask for a class name, Ruby searches through the constants until it
finds one that references the class, and then uses it. For efficiency,
it caches the results of doing this in the class object. For anonymous
objects, the lookup and caching is done when you first try to ask for
the name. For classes defined using 'class XXX', the caching is done
when the class is defined.


Regards


Dave