[ketil@ii.uib.no: Re: Enum class]

[Ketil Malde]

Dylan Thurston <dpt@math.harvard.edu> writes:

> On Thu, Oct 25, 2001 at 10:01:57AM +0200, Ketil Malde wrote:

> I think it's clear from the thread that the language designers did
> _not_ intend this, and that one should not use the default instances
> for the standard numeric classes.  See Simon Peyton-Jones' recent
> post.

Yeah, I just read it.  Unfortunately, I got this thread first :-)

> I agree that Enum instances for Float/Double are not likely to be
> useful.

>From a gut feeling, I could see a use for expressions like

        [1.5, 1.6..] and [1.5, 1.6..2.0]

(i.e. enumFromThen and enumFromThenTo) but enumFrom and enumFromTo
making list of rounded integers seems strange to me.

It seems to me that enumFromThen and -To could be implemented
something like:

        enumFromThen beg next = beg : enumFromThen next (next+delta)
                where delta = next-beg

similarly for enumFromThenTo, of course.

i.e. depending only on functionality found in Num.  Why not put these
functions there, and remove Float and Double as Enum instances?

What am I missing?

-kzm
-- 
If I haven't seen further, it is by standing in the footprints of giants