Re: [Caml-list] problem with generalisation

[Markus Mottl <markus.mottl@gmail.com>]

Hi Christophe,

the following definition of "promote" works for me:

  let promote (type a) (f : a -> unit) : a iter = { f }

It seems that making the type variable locally existential within the
body of "promote" helps the compiler make the correct inference.
Using a universally quantified 'a instead won't work.

Regards,
Markus

On Sat, Sep 23, 2017 at 1:47 PM, Christophe Raffalli
<christophe@raffalli.eu> wrote:
> Hello,
>
> I am blocked by this case not working (I reported on Mantis ...
> https://caml.inria.fr/mantis/view.php?id=7636)
>
> Any hint on how to make it works, apart fomr Obj.magic ?
>
> I can not change the types, they come from a functor.
>
> ====================================================
> type ('a, 'b) elt = 'a
>
> type 'a iter = { f : 'b.('a, 'b) elt -> unit }
>
> let promote f : 'a -> unit =
>   let f : 'b.('a, 'b) elt -> unit = fun x -> f x in
>     { f }
> ====================================================
>
> Cheers,
> Christophe



-- 
Markus Mottl        http://www.ocaml.info        markus.mottl@gmail.com