Bernd Brassel wrote:
> In my point of view the lesson is clear: there should not be
> unification, free variables nor any kind of narrowing for type A -> B.
> With Wolfgang's proposal to add type classes like "Narrow" or
> "Unifiable" to the definition of the Curry language, we could omit this
> problem and all unsafe features would be in Unsafe again.
> For function patterns like in PAKCS giving up patterns of functional
> type would not be problematic. But HO-Patterns seem to be much closer to
> the essence of the problem.
I think this is a serious argument and I agree with it.
In particular, it has the consequence that function application
is no longer rigid in its first argument but should produce
a run-time error since there would be no way to instantiate
an unbound variable of functional type.
I'll have to check whether any of my applications uses
such kinds of "features". If you know of any important
application which requires it, please let us know.
Best regards,
Michael
_______________________________________________
curry mailing list
curry_at_lists.RWTH-Aachen.DE
http://MailMan.RWTH-Aachen.DE/mailman/listinfo/curry
Received on Do Mai 15 2008 - 18:59:13 CEST