numeric-prelude- An experimental alternative hierarchy of numeric type classes

PortabilityHaskell 98
Safe HaskellNone



A type class for non-negative numbers. Prominent instances are T and T numbers. This class cannot do any checks, but it let you show to the user what arguments your function expects. Thus you must define class instances with care. In fact many standard functions (take, '(!!)', ...) should have this type class constraint.



class (Ord a, C a) => C a whereSource

Instances of this class must ensure non-negative values. We cannot enforce this by types, but the type class constraint NonNegative.C avoids accidental usage of types which allow for negative numbers.

The Monoid superclass contributes a zero and an addition.


split :: a -> a -> (a, (Bool, a))Source

split x y == (m,(b,d)) means that b == (x<=y), m == min x y, d == max x y - min x y, that is d == abs(x-y).

We have chosen this function as base function, since it provides comparison and subtraction in one go, which is important for replacing common structures like

 if x<=y
   then f(x-y)
   else g(y-x)

that lead to a memory leak for peano numbers. We have choosen the simple check x<=y instead of a full-blown compare, since we want Zero <= undefined for peano numbers. Because of undefined values split is in general not commutative in the sense

 let (m0,(b0,d0)) = split x y
     (m1,(b1,d1)) = split y x
 in  m0==m1 && d0==d1

The result values are in the order in which they are generated for Peano numbers. We have chosen the nested pair instead of a triple in order to prevent a memory leak that occurs if you only use b and d and ignore m. This is demonstrated by test cases Chunky.splitSpaceLeak3 and Chunky.splitSpaceLeak4.


C T 
(Ord a, C a) => C (T a) 
C a => C (T a) 

splitDefault :: (Ord b, C b) => (a -> b) -> (b -> a) -> a -> a -> (a, (Bool, a))Source

Default implementation for wrapped types of Ord and Num class.

(-|) :: C a => a -> a -> aSource

x -| y == max 0 (x-y)

The default implementation is not efficient, because it compares the values and then subtracts, again, if safe. max 0 (x-y) is more elegant and efficient but not possible in the general case, since x-y may already yield a negative number.

zero :: C a => aSource

add :: C a => a -> a -> aSource

sum :: C a => [a] -> aSource