Michael Hanus wrote:
> This is a good point. However, instead of a runtime error,
> one could also simply fail which expresses the view that "compare"
> could be applied to any expression but is only defined as a total
> order in ground data terms.
Yes, this would be possible as well. In fact, MCC's compare was
implemented in that way initially. I changed the behavior of
compare after tracking down a very obscure failure in a program.
Finally, this turned out to be caused by applying compare to
arguments which could include partial applications -- a fact that
had completely slipped my mind at that moment.
In order to track down such errors more quickly, I decided that
reporting an error would be more appropriate. However, I would agree
that this rather indicates a shortcoming of MCC's current support for
debugging unexpected failures. It is a pity that the declarative
debugger can help diagnosing wrong answers, but not missing answers.
Regards
Wolfgang
_______________________________________________
curry mailing list
curry_at_lists.RWTH-Aachen.DE
http://MailMan.RWTH-Aachen.DE/mailman/listinfo/curry
Received on Di Mär 21 2006 - 18:17:09 CET