Jacques Garrigue wrote:
> From: Alessandro Baretta <alex@baretta.com>
> 
>>At any rate, why does my technique not work?
>>
>>class salutation : object
>>   method greet : 'a. (#parent as 'a) -> string
>>end = object
>>   method greet obj -> obj # foo ^ obj # bar
>>end
> 
> 
> You can see the above (and it is more or less type the same way) as
> 
> class salutation0 = object
>    method greet obj -> obj # foo ^ obj # bar
> end
> 
> class salutation : object
>    method greet : 'a. (#parent as 'a) -> string
> end = salutation0
> 
> So you see that your constraint is applied too late: salutation0
> itself is not typable.

Oh! This is curious. You are telling me that the type 
constraints defined in the class type are applied once and 
for all *after* the class implementation has been typed? 
This would make sense after all, given that the class type 
and the class implementation are two distinct syntactic and 
semantic entities in O'Caml. This is not something I would 
have thought of on my own.

> I though at some time of attempting to make this work, since something
> like that is already done for recursive functions for instance, but
> that's not regular. In fact, you would probably end up wanting to
> write the implementation in the .ml and the interface in the .mli, and
> this would not work anyway.

This is not really a problem. I can live with locally 
applied constraints. I just did not realize why the my 
attempt did not work.

Let me suggest another idea where you might better spend 
your time. I can write
 > let f x : 'a -> 'a = x
but I cannot write
method f x : 'a . 'a -> 'a = x
I have to use the more cumbersome
method f : 'a .'a -> 'a = fun x -> x

The ability to type check correctly the first form of method 
declaration is an improvement I would appreciate. While not 
terriblly important, this is something that might be worth 
your time for some future release.

> So, there is a deep problem about which constraints apply
> where. The examples in the tutorial give you the way to do that
> without ambiguity.
> 
>         Jacques

Thank you very much for your explanations, Jacques. It is a 
pleasure to work with a language where I continually get to 
learn new, conceptually interesting ideas. I'd get bored to 
death otherwise.

Alex

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners