Data.Lub

Compute least upper bound (lub) of two values, with respect to information content. I.e., merge the information available in each. For flat types (in which all values are either bottom or fully defined), lub is equivalent to unamb.

Least upper bounds

class HasLub a where

Types that support information merging (lub)

Methods

lub :: a -> a -> a

Least upper information bound. Combines information available from each argument. The arguments must be consistent, i.e., must have a common upper bound.

lubs :: [a] -> a

n-ary lub. Defaults to foldr lub undefined

Instances

HasLub Bool

HasLub Char

HasLub Double

HasLub Float

HasLub Int

HasLub Integer

HasLub ()

HasLub a => HasLub [a]

HasLub a => HasLub (Maybe a)

HasLub b => HasLub (a -> b)

(HasLub a, HasLub b) => HasLub (Either a b)

(HasLub a, HasLub b) => HasLub (a, b)

flatLub :: a -> a -> a

A lub for flat domains. Equivalent to unamb. Handy for defining HasLub instances, e.g.,

   instance HasLub Integer where lub = flatLub

Some useful special applications of lub

parCommute :: HasLub b => (a -> a -> b) -> a -> a -> b

Turn a binary commutative operation into that tries both orders in parallel, lub-merging the results. Useful when there are special cases that don't require evaluating both arguments.

Similar to parCommute from Unamb, but uses lub instead of unamb.

ptimes :: (HasLub a, Num a) => a -> a -> a

Multiplication optimized for either argument being zero or one, where the other might be expensive/delayed.