Nan actually has a point in floating ... point.  Here it'd not have any performance benefit and force ME and other Users to do error checking by hand in a needless painful way. Nan  on ... nats, just ain Natural. 

i want simple user friendly error messages. 

On Thu, Nov 13, 2014 at 10:34 PM, Mario Blažević <blamario@acanac.net> wrote:
On 13/11/14 05:35 PM, Brandon Allbery wrote:
On Thu, Nov 13, 2014 at 5:20 PM, Mario Blažević <mblazevic@stilo.com <mailto:mblazevic@stilo.com>> wrote:

    Regarding the partial vs. saturated negation, I'm in favour of the
    former. However, there is another option nobody mentioned so far: NaN


NaN is only defined in the context of IEEE floating point; as such, it is not available for this use.

     I'm aware of the current uses of NaN. I was just suggesting that the same concept could be used for operations on natural values that go out of band. Just like we have

   isNaN :: RealFloat a => a -> Bool

we could add a function

   isUnnatural :: Natural -> Bool

and replace exceptions by an unnatural Natural values... I'm open to an alternative terminology. The important question is whether this approach is feasible for the implementation.


_______________________________________________
Libraries mailing list
Libraries@haskell.org
http://www.haskell.org/mailman/listinfo/libraries