On 3 December 2010 00:52, Nala Ginrut <nalaginrut@gmail.com> wrote:

> thanks, but my question is something like this:
> ...
> (cache ... #:slot-ref (lambda (o) (slot-ref o 'cache)) ...) ;; ERROR
> ...
> (sunday ... #:slot-ref (lambda (o) (slot-ref o 'cache)) ...) ;; That's OK
>
> I could call "(slot-ref o 'cache)" in other slot except "cache", but what
> should I do if I need to use "(slot-ref o 'cache)" in the "cache"
> definition?
> I used it directly and got "stack overflow". I think it may cause infinite
> recursive.
> Actually my question can be described more explicitly: "Can I just
> re-define "slot-ref" or "slot-set!" any one of them but NOT both?"
> Sometimes I may need a re-defined "slot-set!" but I expect to let alone
> "slot-ref".
> Anybody catch my mind?
>

Yes, I think so.

The thing about #:allocation #:virtual is that there then isn't any storage
for that slot name, at all, in instances of the relevant class.  So clearly
a normal slot-ref can't work.

But I think I understand what you want: you don't actually want #:allocation
#:virtual, but you do want some kind of customised processing when setting a
slot's value; and you don't need to do anything special when reading the
value.  Is that right?

If so, the simplest solution is to define your slot-setting call as
something other than plain slot-set!, but which uses slot-set! internally
after it has done the custom processing.

You could also define a new kind of #:allocation to implement the behaviour
you're looking for, and a metaclass for classes that can contain that kind
of slots, and write a compute-get-n-set method for that metaclass ... but
that's quite a bit more complex.

Regards,
        Neil