From: Robert Feldt <feldt@...>
Date: 2001-03-06T01:02:00+09:00
Subject: [ruby-talk:12033] Re: ANN: Memoize 0.1.2

On Tue, 6 Mar 2001, Avi Bryant wrote:

> It could be written in Ruby, in the same style as the current incarnation
> of AspectR - doing a lot of fragile, special-cased syntax generation and
> calling instance_eval.  What doing it in C allows is to have one,
> generalized dispatcher that can still easily find out
> a) which method was called to get to the dispatcher, and
> b) which object it was called on.
> 
Ok. But this also gives us a hint that something might be "missing" from
current Ruby (as you said below).

Maybe this info should be accessible by calling 'caller'? Or 'call_info'?

> In Ruby, you could have a generalized dispatcher using define_method with
> a block, but then the block has the wrong context (no knowledge of "self"
> or of which method was called).
> 
> I don't see having a very short C extension of the runtime on which to
> build more complex AOP structures as being that evil, personally... after
> all (for now) we still depend on a C interpreter.
> 
No, I agree. Just want to be a critic to probe this issue so that I fully
understands it, sorry... ;-)

> The ideal, really, would be to have built in support for alternate
> dispatchers - that is, to truly be able to redefine send, whether in C or
> in Ruby.
>
Yes, maybe this is the way to go. With your knowledge from advice.c what
would be a sensible API/naming?

Maybe this can grow into a RCR?

/Robert