I don't quite follow how that would work in a way that doesn't violate the open world assumption.

If you pattern match in a do-binding and don't have a MonadFail instance in scope you'd get a different desugaring?

This would make the presence / absence of an instance change code generation. Adding instances shouldn't change generated code.

-Edward

On Wed, Jun 10, 2015 at 3:00 PM, Johan Tibell <johan.tibell@gmail.com> wrote:
An idea. Could the new desugaring be the same as for normal functions with a missing case? Code that uses fail would have to add the constraint still, but code that really didn't want to use fail would 1) get a warning about an incomplete pattern match and 2) get a runtime error for the failed pattern match produced by GHC. Is there anything I'm not thinking of here why this wouldn't work?

_______________________________________________
Libraries mailing list
Libraries@haskell.org
http://mail.haskell.org/cgi-bin/mailman/listinfo/libraries