{-# OPTIONS_GHC -Wall -Werror #-} ---------------------------------------------------------------- -- ~ 2008.07.25 -- | -- Module : Data.List.Extras.LazyLength -- Copyright : Copyright (c) 2007--2008 wren ng thornton -- License : BSD3 -- Maintainer : wren@community.haskell.org -- Stability : stable -- Portability : portable -- -- This module provides least-strict functions for getting a list's -- length and doing natural things with it. On GHC this module also -- uses rewrite rules to convert certain calls to 'length' into our -- least-strict versions. -- -- The regular version of @length@ will traverse the entire spine -- of the list in order to return an answer. For comparing the -- length against some bound, that is by far too strict. Being too -- strict can cause a space leak by expanding a lazy list before -- necessary (or more than is ever necessary). And it can lead to -- unnecessarily non-terminating programs when trying to determine -- if an infinite list is longer or shorter than some finite bound. -- -- A nicer version of @length@ would return some lazy approximation -- of an answer which retains the proper semantics. An option for -- doing this is to return Peano integers which can be decremented -- as much as necessary and no further (i.e. at most one more than -- the bound). Of course, Peano integers are woefully inefficient -- and would wreck the cache and burn heap. This module provides -- functions with the same lazy effect as if we used Peano integers, -- but does so efficiently instead. -- -- (For Peano integers see numbers:"Data.Number.Natural" or -- non-negative:"Numeric.NonNegative.Class".) ---------------------------------------------------------------- module Data.List.Extras.LazyLength ( lengthBound, lengthCompare ) where ---------------------------------------------------------------- ---------------------------------------------------------------- -- | A variant of 'length' which is least-strict for comparing -- against a boundary length. This function is defined primarily -- for use by rewrite rules rather than for direct use (though it's -- fine for that too). -- -- @lengthBound@ is polymorphic in the return of the helper -- function so we can use 'compare' as well as '>', '>=', '==', -- '/=', '<=', '<'. If you want to use any other functions, know -- that we only preserve the ordering of the list's length vs the -- boundary length and so the function should not rely on the true -- values of either of the numbers being compared. lengthBound :: Int -> (Int -> Int -> a) -> [b] -> a lengthBound n cmp xs | n < 0 = case xs of [] -> cmp n 0 (_:_) -> cmp n 1 | otherwise = go n xs where go n' [] = cmp n' 0 go 0 (_:_) = cmp 0 1 go n' (_:xs') = (go $! n'-1) xs' {-# RULES "lengthBound/(>)" forall n xs. n > length xs = lengthBound n (>) xs "lengthBound/(>=)" forall n xs. n >= length xs = lengthBound n (>=) xs "lengthBound/(==)" forall n xs. n == length xs = lengthBound n (==) xs "lengthBound/(/=)" forall n xs. n /= length xs = lengthBound n (/=) xs "lengthBound/(<=)" forall n xs. n <= length xs = lengthBound n (<=) xs "lengthBound/(<)" forall n xs. n < length xs = lengthBound n (<) xs "lengthBound/compare" forall n xs. compare n (length xs) = lengthBound n compare xs "lengthBound\\(>)" forall n xs. length xs > n = lengthBound n (<) xs "lengthBound\\(>=)" forall n xs. length xs >= n = lengthBound n (<=) xs "lengthBound\\(==)" forall n xs. length xs == n = lengthBound n (==) xs "lengthBound\\(/=)" forall n xs. length xs /= n = lengthBound n (/=) xs "lengthBound\\(<=)" forall n xs. length xs <= n = lengthBound n (>=) xs "lengthBound\\(<)" forall n xs. length xs < n = lengthBound n (>) xs "lengthBound\\compare" forall n xs. compare (length xs) n = lengthBound n (flip compare) xs #-} ---------------------------------------------------------------- ---------------------------------------------------------------- -- | A variant of 'length' which is least-strict for comparing -- the lengths of two lists. This is as strict as the length of the -- shorter list (which allows comparing an infinite list against a -- finite list). The function itself is trivial, but again it's -- designed primarily for rewrite rules. -- -- If you're going to immediately follow this with a 'zip' function -- then see "Data.List.Extras.Pair" instead. lengthCompare :: [a] -> [b] -> Ordering lengthCompare [] [] = EQ lengthCompare (_:_) [] = GT lengthCompare [] (_:_) = LT lengthCompare (_:xs) (_:ys) = lengthCompare xs ys {-# RULES "lengthCompare/(>)" forall xs ys. length xs > length ys = lengthCompare xs ys == GT "lengthCompare/(>=)" forall xs ys. length xs >= length ys = lengthCompare xs ys /= LT "lengthCompare/(==)" forall xs ys. length xs == length ys = lengthCompare xs ys == EQ "lengthCompare/(/=)" forall xs ys. length xs /= length ys = lengthCompare xs ys /= EQ "lengthCompare/(<=)" forall xs ys. length xs <= length ys = lengthCompare xs ys /= GT "lengthCompare/(<)" forall xs ys. length xs < length ys = lengthCompare xs ys == LT "lengthCompare/compare" forall n xs. compare (length xs) (length ys) = lengthCompare xs ys #-} ---------------------------------------------------------------- ----------------------------------------------------------- fin.