Re: Inheritance in Moto

On Saturday, January 25, 2003, at 05:14  PM, Stefano Corsi wrote:

> For inherited classes, do you think is better to copy the parent 
> methods and
> members definition in the new class definition or to have the various
> interface functions perform recursively on the parent chain to find 
> the right
> method and/or member?
>
> Stefano

Hmm, I haven't thought too much about this.

For the interpreter either way should be fine although if you copy 
methods into the new class definitions you will probably have to add 
some nasty checks when someone overloads a method in a child class so 
that a 'method redefinition error' doesn't occur. Also, in the compiler 
you will have to make sure the parent method definition is only written 
out once . Sooo after thinking about it I guess a recursive function 
that discovers the nearest ancestor that defines the method is the way 
to go

With regard to compiled code we want to avoid any non-fixed-cost 
operations. What I had in mind for this was the pre-computed function 
pointer array in the Class objects I mentioned in a previous email e.g.

If

A defined methods a(),b(), and c()
B extends A and defines its own b()
C extends B and defines its own c() and d()

then the class definitions for the above three classes would have a 
pre-computed array of function pointers that look like:

A Definition Methods { A::a, A::b, A::c }
B Definition Methods { A::a, B::b, A::c }
C Definition Methods { A::a, B::b, C::c, C::d }

Does this help ?

-Dave