On Thu, Dec 05, 2002 at 07:04:34PM +0100, Ralf Welter wrote:
I think I have found a bug in the hugs prelude: Bug: definition of recip in the instance declaration of breaks silent assumptions about the internal data in Ratio types.
Description: data Integral a => Ratio a = !a :% !a deriving (Eq) by deriving Eq a normal form is assumed. Further places do at least assume a positive denominator:
instance Integral a => Num (Ratio a) where ... abs (x :% y) = abs x :% y signum (x :% y) = signum x :% 1 silently assume the denominator is positive
unforturtunately the definition of recip
instance Integral a => Fractional (Ratio a) where ... recip (x:%y) = y :% x will break these properties. fix: remove the colon on the right hand side of the recip definition.
Thanks for the report and fix -- you're right, this is what the Report specifies (and what GHC does). It's now fixed in CVS.