From lemming at henning-thielemann.de  Sat Aug 27 06:30:37 2016
From: lemming at henning-thielemann.de (Henning Thielemann)
Date: Sat, 27 Aug 2016 05:30:37 -0000
Subject: [NumericPrelude] NP bug reports
In-Reply-To: <CAApao_n8riHgP_6qte7c-wU-bk+fs6B+qS5Lc7Y2=6XYnsKrwQ@mail.gmail.com>
References: <CAApao_n8riHgP_6qte7c-wU-bk+fs6B+qS5Lc7Y2=6XYnsKrwQ@mail.gmail.com>
Message-ID: <alpine.DEB.2.02.1608260956590.18049@sputnik>


On Thu, 25 Aug 2016, Eric Crockett wrote:

> Is there a github (or similar) repo for this library where I can file bug reports and feature requests?

E-mail to me is ok, you may also mail to 
numeric-prelude at projects.haskell.org. However, the mailing list had 
problems. I could move to darcs-hub. It includes a bug tracker, but 
requires conversion to darcs-2 format. So I did not perform the move so 
far because of low volume of issues.

> With the NP equivalent:
> 
> {-# LANGUAGE NoImplicitPrelude, RebindableSyntax #-}

Btw. you may omit NoImplicitPrelude. It is implied by RebindableSyntax.

> module Main where
> 
> import NumericPrelude
> 
> main :: IO ()
> main = do
> ? let x = sqrt $ -1.0 :: Double
> ? print x -- NaN
> ? let y = round x :: Int
> ? print y -- crashes
> 
> When compiled with -O0, I get NaN and 0, respectively (same as Prelude). When compiled with -O1, I get NaN and
> "Prelude.Enum.pred{Int}: tried to take `pred' of minBound".


I think the behavior is better than that of Prelude, yet, the error 
message is confusing. I think it is caused by Algebra.RealRing.roundInt. I 
wrote 'pred n' instead of 'n-1'. In general I did such things in order to 
avoid working with number literals because they are not available in all 
numeric-prelude modules. However, in this case 'n-1' should also work.