On Wed, 4 Nov 1998 03:36:05 +1100
Fergus Henderson <[EMAIL PROTECTED]> wrote:
> On 03-Nov-1998, Michael Hanus <[EMAIL PROTECTED]> wrote:
> > This sounds interesting but how did you implement it?
> > As far as I know, the problem of type inference with polymorphic
> > recursion is undecidable due to the undecidability of the
> > semi-unification problem.
> The practical solution is quite simple. The fixpoint calculation runs
> for a finite but user-selectable number of steps. If it doesn't reach
> a fixpoint in that time, it halts with an error message that shows the
> types it has inferred so far and that explains how to increase the limit.
As a complement, I remembered Milner himself talked about this
extension in one of his papers.
I am curious why the original poster would need this feature. To support
nested datatype, for example?
sincerely,
Shin-Cheng Mu
Academia Sinica, Taiwan.
