
On 27 December 2011 17:47, Bas van Dijk
I fixed a similar breakage in the hmatrix library:
https://github.com/AlbertoRuiz/hmatrix/commit/a4f38eb196209436f72b938f6355f6...
GHC-7.4.1-rc1 also reported another type error in code that was accepted by GHC <= 7.2.2. These were the type errors I got: [24 of 36] Compiling Numeric.LinearAlgebra.Algorithms ( lib/Numeric/LinearAlgebra/Algorithms.hs, dist/build/Numeric/LinearAlgebra/Algorithms.o ) lib/Numeric/LinearAlgebra/Algorithms.hs:576:23: No instance for (RealFrac (RealOf t0)) arising from a use of `floor' Possible fix: add an instance declaration for (RealFrac (RealOf t0)) In the expression: floor In the second argument of `($)', namely `floor $ logBase 2 $ pnorm Infinity m' In the expression: max 0 $ floor $ logBase 2 $ pnorm Infinity m lib/Numeric/LinearAlgebra/Algorithms.hs:576:31: No instance for (Floating (RealOf t0)) arising from a use of `logBase' Possible fix: add an instance declaration for (Floating (RealOf t0)) In the expression: logBase 2 In the second argument of `($)', namely `logBase 2 $ pnorm Infinity m' In the second argument of `($)', namely `floor $ logBase 2 $ pnorm Infinity m' lib/Numeric/LinearAlgebra/Algorithms.hs:576:39: No instance for (Num (RealOf t0)) arising from the literal `2' Possible fix: add an instance declaration for (Num (RealOf t0)) In the first argument of `logBase', namely `2' In the expression: logBase 2 In the second argument of `($)', namely `logBase 2 $ pnorm Infinity m' lib/Numeric/LinearAlgebra/Algorithms.hs:576:43: No instance for (Normed Matrix t0) arising from a use of `pnorm' Possible fix: add an instance declaration for (Normed Matrix t0) In the second argument of `($)', namely `pnorm Infinity m' In the second argument of `($)', namely `logBase 2 $ pnorm Infinity m' In the second argument of `($)', namely `floor $ logBase 2 $ pnorm Infinity m' lib/Numeric/LinearAlgebra/Algorithms.hs:593:19: No instance for (Container Vector t0) arising from a use of `add' Possible fix: add an instance declaration for (Container Vector t0) In the expression: add In an equation for `|+|': |+| = add In an equation for `expGolub': expGolub m = iterate msq f !! j where j = max 0 $ floor $ logBase 2 $ pnorm Infinity m a = m */ fromIntegral ((2 :: Int) ^ j) q = geps eps eye = ident (rows m) .... lib/Numeric/LinearAlgebra/Algorithms.hs:599:1: Couldn't match type `t0' with `t' because type variable `t' would escape its scope This (rigid, skolem) type variable is bound by the type signature for expm :: Field t => Matrix t -> Matrix t The following variables have types that mention t0 expGolub :: Matrix t0 -> Matrix t0 (bound at lib/Numeric/LinearAlgebra/Algorithms.hs:575:1) Note that RealOf is a type family: type family RealOf x type instance RealOf Double = Double type instance RealOf (Complex Double) = Double type instance RealOf Float = Float type instance RealOf (Complex Float) = Float Adding the following explicit type signature fixed it: expGolub :: ( Fractional t, Element t, Field t , Normed Matrix t , RealFrac (RealOf t) , Floating (RealOf t) ) => Matrix t -> Matrix t I have no idea if this should be considered a bug. Regards, Bas