heaps-0.4: src/Data/Heap.hs
{-# LANGUAGE CPP #-}
{-# LANGUAGE DeriveDataTypeable #-}
#if __GLASGOW_HASKELL__ >= 707
{-# LANGUAGE RoleAnnotations #-}
#endif
-----------------------------------------------------------------------------
-- |
-- Copyright : (c) Edward Kmett 2010-2015
-- License : BSD-style
-- Maintainer : ekmett@gmail.com
-- Stability : experimental
-- Portability : portable
--
-- An efficient, asymptotically optimal, implementation of a priority queues
-- extended with support for efficient size, and `Data.Foldable`
--
-- /Note/: Since many function names (but not the type name) clash with
-- "Prelude" names, this module is usually imported @qualified@, e.g.
--
-- > import Data.Heap (Heap)
-- > import qualified Data.Heap as Heap
--
-- The implementation of 'Heap' is based on /bootstrapped skew binomial heaps/
-- as described by:
--
-- * G. Brodal and C. Okasaki , <http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.48.973 "Optimal Purely Functional Priority Queues">,
-- /Journal of Functional Programming/ 6:839-857 (1996)
--
-- All time bounds are worst-case.
-----------------------------------------------------------------------------
module Data.Heap
(
-- * Heap Type
Heap -- instance Eq,Ord,Show,Read,Data,Typeable
-- * Entry type
, Entry(..) -- instance Eq,Ord,Show,Read,Data,Typeable
-- * Basic functions
, empty -- O(1) :: Heap a
, null -- O(1) :: Heap a -> Bool
, size -- O(1) :: Heap a -> Int
, singleton -- O(1) :: Ord a => a -> Heap a
, insert -- O(1) :: Ord a => a -> Heap a -> Heap a
, minimum -- O(1) (/partial/) :: Ord a => Heap a -> a
, deleteMin -- O(log n) :: Heap a -> Heap a
, adjustMin -- O(log n) :: (a -> a) -> Heap a -> Heap a
, union -- O(1) :: Heap a -> Heap a -> Heap a
, uncons, viewMin -- O(1)\/O(log n) :: Heap a -> Maybe (a, Heap a)
-- * Transformations
, mapMonotonic -- O(n) :: Ord b => (a -> b) -> Heap a -> Heap b
, map -- O(n) :: Ord b => (a -> b) -> Heap a -> Heap b
-- * To/From Lists
, toUnsortedList -- O(n) :: Heap a -> [a]
, fromList -- O(n) :: Ord a => [a] -> Heap a
, sort -- O(n log n) :: Ord a => [a] -> [a]
, traverse -- O(n log n) :: (Applicative t, Ord b) => (a -> t b) -> Heap a -> t (Heap b)
, mapM -- O(n log n) :: (Monad m, Ord b) => (a -> m b) -> Heap a -> m (Heap b)
, concatMap -- O(n) :: Ord b => Heap a -> (a -> Heap b) -> Heap b
-- * Filtering
, filter -- O(n) :: (a -> Bool) -> Heap a -> Heap a
, partition -- O(n) :: (a -> Bool) -> Heap a -> (Heap a, Heap a)
, split -- O(n) :: a -> Heap a -> (Heap a, Heap a, Heap a)
, break -- O(n log n) :: (a -> Bool) -> Heap a -> (Heap a, Heap a)
, span -- O(n log n) :: (a -> Bool) -> Heap a -> (Heap a, Heap a)
, take -- O(n log n) :: Int -> Heap a -> Heap a
, drop -- O(n log n) :: Int -> Heap a -> Heap a
, splitAt -- O(n log n) :: Int -> Heap a -> (Heap a, Heap a)
, takeWhile -- O(n log n) :: (a -> Bool) -> Heap a -> Heap a
, dropWhile -- O(n log n) :: (a -> Bool) -> Heap a -> Heap a
-- * Grouping
, group -- O(n log n) :: Heap a -> Heap (Heap a)
, groupBy -- O(n log n) :: (a -> a -> Bool) -> Heap a -> Heap (Heap a)
, nub -- O(n log n) :: Heap a -> Heap a
-- * Intersection
, intersect -- O(n log n + m log m) :: Heap a -> Heap a -> Heap a
, intersectWith -- O(n log n + m log m) :: Ord b => (a -> a -> b) -> Heap a -> Heap a -> Heap b
-- * Duplication
, replicate -- O(log n) :: Ord a => a -> Int -> Heap a
) where
import Prelude hiding
( map
, span, dropWhile, takeWhile, break, filter, take, drop, splitAt
, foldr, minimum, replicate, mapM
, concatMap, null
#if MIN_VERSION_base(4,8,0)
, traverse
#endif
)
import Control.Monad (liftM)
import Data.Data (DataType, Constr, mkConstr, mkDataType, Fixity(Prefix), Data(..), constrIndex)
import qualified Data.Foldable as F
import Data.Function (on)
import qualified Data.List as L
import qualified Data.Traversable as T
import Data.Typeable (Typeable)
import Text.Read
#if MIN_VERSION_base(4,8,0)
import Data.Bifunctor
#else
import Control.Applicative (Applicative)
import Data.Foldable (Foldable)
import Data.Monoid (Monoid(mappend, mempty))
import Data.Traversable (Traversable)
#endif
#if MIN_VERSION_base(4,9,0) && !(MIN_VERSION_base(4,11,0))
import Data.Semigroup (Semigroup(..))
#endif
-- $setup
-- >>> let break = Data.Heap.break
-- >>> let concatMap = Data.Heap.concatMap
-- >>> let dropWhile = Data.Heap.dropWhile
-- >>> let filter = Data.Heap.filter
-- >>> let minimum = Data.Heap.minimum
-- >>> let null = Data.Heap.null
-- >>> let span = Data.Heap.span
-- >>> let take = Data.Heap.take
-- >>> let takeWhile = Data.Heap.takeWhile
--
-- -- GHC 7.0 and 7.2 will default the `Ord` constraints to () in the types of
-- -- the following functions unless we give them explicit type signatures.
-- >>> let { map :: Ord b => (a -> b) -> Heap a -> Heap b; map = Data.Heap.map }
-- >>> let { replicate :: Ord a => a -> Int -> Heap a ; replicate = Data.Heap.replicate }
-- The implementation of 'Heap' must internally hold onto the dictionary entry for ('<='),
-- so that it can be made 'Foldable'. Confluence in the absence of incoherent instances
-- is provided by the fact that we only ever build these from instances of 'Ord' a (except in the case of 'groupBy')
-- | A min-heap of values of type @a@.
data Heap a
= Empty
| Heap {-# UNPACK #-} !Int (a -> a -> Bool) {-# UNPACK #-} !(Tree a)
deriving Typeable
#if __GLASGOW_HASKELL__ >= 707
type role Heap nominal
#endif
instance Show a => Show (Heap a) where
showsPrec _ Empty = showString "fromList []"
showsPrec d (Heap _ _ t) = showParen (d > 10) $
showString "fromList " . showsPrec 11 (F.toList t)
instance (Ord a, Read a) => Read (Heap a) where
readPrec = parens $ prec 10 $ do
Ident "fromList" <- lexP
fromList `fmap` step readPrec
instance (Ord a, Data a) => Data (Heap a) where
gfoldl k z h = z fromList `k` toUnsortedList h
toConstr _ = fromListConstr
dataTypeOf _ = heapDataType
gunfold k z c = case constrIndex c of
1 -> k (z fromList)
_ -> error "gunfold"
heapDataType :: DataType
heapDataType = mkDataType "Data.Heap.Heap" [fromListConstr]
fromListConstr :: Constr
fromListConstr = mkConstr heapDataType "fromList" [] Prefix
instance Eq (Heap a) where
Empty == Empty = True
Empty == Heap{} = False
Heap{} == Empty = False
a@(Heap s1 leq _) == b@(Heap s2 _ _) = s1 == s2 && go leq (F.toList a) (F.toList b)
where
go f (x:xs) (y:ys) = f x y && f y x && go f xs ys
go _ [] [] = True
go _ _ _ = False
instance Ord (Heap a) where
Empty `compare` Empty = EQ
Empty `compare` Heap{} = LT
Heap{} `compare` Empty = GT
a@(Heap _ leq _) `compare` b = go leq (F.toList a) (F.toList b)
where
go f (x:xs) (y:ys) =
if f x y
then if f y x
then go f xs ys
else LT
else GT
go _ [] [] = EQ
go _ [] (_:_) = LT
go _ (_:_) [] = GT
-- | /O(1)/. The empty heap
--
-- @'empty' ≡ 'fromList' []@
--
-- >>> size empty
-- 0
empty :: Heap a
empty = Empty
{-# INLINE empty #-}
-- | /O(1)/. A heap with a single element
--
-- @
-- 'singleton' x ≡ 'fromList' [x]
-- 'singleton' x ≡ 'insert' x 'empty'
-- @
--
-- >>> size (singleton "hello")
-- 1
singleton :: Ord a => a -> Heap a
singleton = singletonWith (<=)
{-# INLINE singleton #-}
singletonWith :: (a -> a -> Bool) -> a -> Heap a
singletonWith f a = Heap 1 f (Node 0 a Nil)
{-# INLINE singletonWith #-}
-- | /O(1)/. Insert a new value into the heap.
--
-- >>> insert 2 (fromList [1,3])
-- fromList [1,2,3]
--
-- @
-- 'insert' x 'empty' ≡ 'singleton' x
-- 'size' ('insert' x xs) ≡ 1 + 'size' xs
-- @
insert :: Ord a => a -> Heap a -> Heap a
insert = insertWith (<=)
{-# INLINE insert #-}
insertWith :: (a -> a -> Bool) -> a -> Heap a -> Heap a
insertWith leq x Empty = singletonWith leq x
insertWith leq x (Heap s _ t@(Node _ y f))
| leq x y = Heap (s+1) leq (Node 0 x (t `Cons` Nil))
| otherwise = Heap (s+1) leq (Node 0 y (skewInsert leq (Node 0 x Nil) f))
{-# INLINE insertWith #-}
-- | /O(1)/. Meld the values from two heaps into one heap.
--
-- >>> union (fromList [1,3,5]) (fromList [6,4,2])
-- fromList [1,2,6,4,3,5]
-- >>> union (fromList [1,1,1]) (fromList [1,2,1])
-- fromList [1,1,1,2,1,1]
union :: Heap a -> Heap a -> Heap a
union Empty q = q
union q Empty = q
union (Heap s1 leq t1@(Node _ x1 f1)) (Heap s2 _ t2@(Node _ x2 f2))
| leq x1 x2 = Heap (s1 + s2) leq (Node 0 x1 (skewInsert leq t2 f1))
| otherwise = Heap (s1 + s2) leq (Node 0 x2 (skewInsert leq t1 f2))
{-# INLINE union #-}
-- | /O(log n)/. Create a heap consisting of multiple copies of the same value.
--
-- >>> replicate 'a' 10
-- fromList "aaaaaaaaaa"
replicate :: Ord a => a -> Int -> Heap a
replicate x0 y0
| y0 < 0 = error "Heap.replicate: negative length"
| y0 == 0 = mempty
| otherwise = f (singleton x0) y0
where
f x y
| even y = f (union x x) (quot y 2)
| y == 1 = x
| otherwise = g (union x x) (quot (y - 1) 2) x
g x y z
| even y = g (union x x) (quot y 2) z
| y == 1 = union x z
| otherwise = g (union x x) (quot (y - 1) 2) (union x z)
{-# INLINE replicate #-}
-- | Provides both /O(1)/ access to the minimum element and /O(log n)/ access to the remainder of the heap.
-- This is the same operation as 'viewMin'
--
-- >>> uncons (fromList [2,1,3])
-- Just (1,fromList [2,3])
uncons :: Heap a -> Maybe (a, Heap a)
uncons Empty = Nothing
uncons l@(Heap _ _ t) = Just (root t, deleteMin l)
{-# INLINE uncons #-}
-- | Same as 'uncons'
viewMin :: Heap a -> Maybe (a, Heap a)
viewMin = uncons
{-# INLINE viewMin #-}
-- | /O(1)/. Assumes the argument is a non-'null' heap.
--
-- >>> minimum (fromList [3,1,2])
-- 1
minimum :: Heap a -> a
minimum Empty = error "Heap.minimum: empty heap"
minimum (Heap _ _ t) = root t
{-# INLINE minimum #-}
trees :: Forest a -> [Tree a]
trees (a `Cons` as) = a : trees as
trees Nil = []
-- | /O(log n)/. Delete the minimum key from the heap and return the resulting heap.
--
-- >>> deleteMin (fromList [3,1,2])
-- fromList [2,3]
deleteMin :: Heap a -> Heap a
deleteMin Empty = Empty
deleteMin (Heap _ _ (Node _ _ Nil)) = Empty
deleteMin (Heap s leq (Node _ _ f0)) = Heap (s - 1) leq (Node 0 x f3)
where
(Node r x cf, ts2) = getMin leq f0
(zs, ts1, f1) = splitForest r Nil Nil cf
f2 = skewMeld leq (skewMeld leq ts1 ts2) f1
f3 = F.foldr (skewInsert leq) f2 (trees zs)
{-# INLINE deleteMin #-}
-- | /O(log n)/. Adjust the minimum key in the heap and return the resulting heap.
--
-- >>> adjustMin (+1) (fromList [1,2,3])
-- fromList [2,2,3]
adjustMin :: (a -> a) -> Heap a -> Heap a
adjustMin _ Empty = Empty
adjustMin f (Heap s leq (Node r x xs)) = Heap s leq (heapify leq (Node r (f x) xs))
{-# INLINE adjustMin #-}
type ForestZipper a = (Forest a, Forest a)
zipper :: Forest a -> ForestZipper a
zipper xs = (Nil, xs)
{-# INLINE zipper #-}
emptyZ :: ForestZipper a
emptyZ = (Nil, Nil)
{-# INLINE emptyZ #-}
-- leftZ :: ForestZipper a -> ForestZipper a
-- leftZ (x :> path, xs) = (path, x :> xs)
rightZ :: ForestZipper a -> ForestZipper a
rightZ (path, x `Cons` xs) = (x `Cons` path, xs)
rightZ _ = error "Heap.rightZ: empty zipper"
{-# INLINE rightZ #-}
-- adjustZ :: (Tree a -> Tree a) -> ForestZipper a -> ForestZipper a
-- adjustZ f (path, x `Cons` xs) = (path, f x `Cons` xs)
-- adjustZ _ z = z
-- {-# INLINE adjustZ #-}
rezip :: ForestZipper a -> Forest a
rezip (Nil, xs) = xs
rezip (x `Cons` path, xs) = rezip (path, x `Cons` xs)
-- assumes non-empty zipper
rootZ :: ForestZipper a -> a
rootZ (_ , x `Cons` _) = root x
rootZ _ = error "Heap.rootZ: empty zipper"
{-# INLINE rootZ #-}
minZ :: (a -> a -> Bool) -> Forest a -> ForestZipper a
minZ _ Nil = emptyZ
minZ f xs = minZ' f z z
where z = zipper xs
{-# INLINE minZ #-}
minZ' :: (a -> a -> Bool) -> ForestZipper a -> ForestZipper a -> ForestZipper a
minZ' _ lo (_, Nil) = lo
minZ' leq lo z = minZ' leq (if leq (rootZ lo) (rootZ z) then lo else z) (rightZ z)
heapify :: (a -> a -> Bool) -> Tree a -> Tree a
heapify _ n@(Node _ _ Nil) = n
heapify leq n@(Node r a as)
| leq a a' = n
| otherwise = Node r a' (rezip (left, heapify leq (Node r' a as') `Cons` right))
where
(left, Node r' a' as' `Cons` right) = minZ leq as
-- | /O(n)/. Build a heap from a list of values.
--
-- @
-- 'fromList' '.' 'toList' ≡ 'id'
-- 'toList' '.' 'fromList' ≡ 'sort'
-- @
-- >>> size (fromList [1,5,3])
-- 3
fromList :: Ord a => [a] -> Heap a
fromList = F.foldr insert mempty
{-# INLINE fromList #-}
fromListWith :: (a -> a -> Bool) -> [a] -> Heap a
fromListWith f = F.foldr (insertWith f) mempty
{-# INLINE fromListWith #-}
-- | /O(n log n)/. Perform a heap sort
sort :: Ord a => [a] -> [a]
sort = F.toList . fromList
{-# INLINE sort #-}
#if MIN_VERSION_base(4,9,0)
instance Semigroup (Heap a) where
(<>) = union
{-# INLINE (<>) #-}
#endif
instance Monoid (Heap a) where
mempty = empty
{-# INLINE mempty #-}
#if !(MIN_VERSION_base(4,11,0))
mappend = union
{-# INLINE mappend #-}
#endif
-- | /O(n)/. Returns the elements in the heap in some arbitrary, very likely unsorted, order.
--
-- >>> toUnsortedList (fromList [3,1,2])
-- [1,3,2]
--
-- @'fromList' '.' 'toUnsortedList' ≡ 'id'@
toUnsortedList :: Heap a -> [a]
toUnsortedList Empty = []
toUnsortedList (Heap _ _ t) = F.foldMap return t
{-# INLINE toUnsortedList #-}
instance Foldable Heap where
foldMap _ Empty = mempty
foldMap f l@(Heap _ _ t) = f (root t) `mappend` F.foldMap f (deleteMin l)
#if MIN_VERSION_base(4,8,0)
null = null
length = size
#endif
-- | /O(1)/. Is the heap empty?
--
-- >>> null empty
-- True
--
-- >>> null (singleton "hello")
-- False
null :: Heap a -> Bool
null Empty = True
null _ = False
{-# INLINE null #-}
-- | /O(1)/. The number of elements in the heap.
--
-- >>> size empty
-- 0
-- >>> size (singleton "hello")
-- 1
-- >>> size (fromList [4,1,2])
-- 3
size :: Heap a -> Int
size Empty = 0
size (Heap s _ _) = s
{-# INLINE size #-}
-- | /O(n)/. Map a function over the heap, returning a new heap ordered appropriately for its fresh contents
--
-- >>> map negate (fromList [3,1,2])
-- fromList [-3,-1,-2]
map :: Ord b => (a -> b) -> Heap a -> Heap b
map _ Empty = Empty
map f (Heap _ _ t) = F.foldMap (singleton . f) t
{-# INLINE map #-}
-- | /O(n)/. Map a monotone increasing function over the heap.
-- Provides a better constant factor for performance than 'map', but no checking is performed that the function provided is monotone increasing. Misuse of this function can cause a Heap to violate the heap property.
--
-- >>> mapMonotonic (+1) (fromList [1,2,3])
-- fromList [2,3,4]
-- >>> mapMonotonic (*2) (fromList [1,2,3])
-- fromList [2,4,6]
mapMonotonic :: Ord b => (a -> b) -> Heap a -> Heap b
mapMonotonic _ Empty = Empty
mapMonotonic f (Heap s _ t) = Heap s (<=) (fmap f t)
{-# INLINE mapMonotonic #-}
-- * Filter
-- | /O(n)/. Filter the heap, retaining only values that satisfy the predicate.
--
-- >>> filter (>'a') (fromList "ab")
-- fromList "b"
-- >>> filter (>'x') (fromList "ab")
-- fromList []
-- >>> filter (<'a') (fromList "ab")
-- fromList []
filter :: (a -> Bool) -> Heap a -> Heap a
filter _ Empty = Empty
filter p (Heap _ leq t) = F.foldMap f t
where
f x | p x = singletonWith leq x
| otherwise = Empty
{-# INLINE filter #-}
-- | /O(n)/. Partition the heap according to a predicate. The first heap contains all elements that satisfy the predicate, the second all elements that fail the predicate. See also 'split'.
--
-- >>> partition (>'a') (fromList "ab")
-- (fromList "b",fromList "a")
partition :: (a -> Bool) -> Heap a -> (Heap a, Heap a)
partition _ Empty = (Empty, Empty)
partition p (Heap _ leq t) = F.foldMap f t
where
f x | p x = (singletonWith leq x, mempty)
| otherwise = (mempty, singletonWith leq x)
{-# INLINE partition #-}
-- | /O(n)/. Partition the heap into heaps of the elements that are less than, equal to, and greater than a given value.
--
-- >>> split 'h' (fromList "hello")
-- (fromList "e",fromList "h",fromList "llo")
split :: a -> Heap a -> (Heap a, Heap a, Heap a)
split _ Empty = (Empty, Empty, Empty)
split a (Heap _ leq t) = F.foldMap f t
where
f x = if leq x a
then if leq a x
then (mempty, singletonWith leq x, mempty)
else (singletonWith leq x, mempty, mempty)
else (mempty, mempty, singletonWith leq x)
{-# INLINE split #-}
-- * Subranges
-- | /O(n log n)/. Return a heap consisting of the least @n@ elements of a given heap.
--
-- >>> take 3 (fromList [10,2,4,1,9,8,2])
-- fromList [1,2,2]
take :: Int -> Heap a -> Heap a
take = withList . L.take
{-# INLINE take #-}
-- | /O(n log n)/. Return a heap consisting of all members of given heap except for the @n@ least elements.
drop :: Int -> Heap a -> Heap a
drop = withList . L.drop
{-# INLINE drop #-}
-- | /O(n log n)/. Split a heap into two heaps, the first containing the @n@ least elements, the latter consisting of all members of the heap except for those elements.
splitAt :: Int -> Heap a -> (Heap a, Heap a)
splitAt = splitWithList . L.splitAt
{-# INLINE splitAt #-}
-- | /O(n log n)/. 'break' applied to a predicate @p@ and a heap @xs@ returns a tuple where the first element is a heap consisting of the
-- longest prefix the least elements of @xs@ that /do not satisfy/ p and the second element is the remainder of the elements in the heap.
--
-- >>> break (\x -> x `mod` 4 == 0) (fromList [3,5,7,12,13,16])
-- (fromList [3,5,7],fromList [12,13,16])
--
-- 'break' @p@ is equivalent to @'span' ('not' . p)@.
break :: (a -> Bool) -> Heap a -> (Heap a, Heap a)
break = splitWithList . L.break
{-# INLINE break #-}
-- | /O(n log n)/. 'span' applied to a predicate @p@ and a heap @xs@ returns a tuple where the first element is a heap consisting of the
-- longest prefix the least elements of xs that satisfy @p@ and the second element is the remainder of the elements in the heap.
--
-- >>> span (\x -> x `mod` 4 == 0) (fromList [4,8,12,14,16])
-- (fromList [4,8,12],fromList [14,16])
--
-- 'span' @p xs@ is equivalent to @('takeWhile' p xs, 'dropWhile' p xs)@
span :: (a -> Bool) -> Heap a -> (Heap a, Heap a)
span = splitWithList . L.span
{-# INLINE span #-}
-- | /O(n log n)/. 'takeWhile' applied to a predicate @p@ and a heap @xs@ returns a heap consisting of the
-- longest prefix the least elements of @xs@ that satisfy @p@.
--
-- >>> takeWhile (\x -> x `mod` 4 == 0) (fromList [4,8,12,14,16])
-- fromList [4,8,12]
takeWhile :: (a -> Bool) -> Heap a -> Heap a
takeWhile = withList . L.takeWhile
{-# INLINE takeWhile #-}
-- | /O(n log n)/. 'dropWhile' @p xs@ returns the suffix of the heap remaining after 'takeWhile' @p xs@.
--
-- >>> dropWhile (\x -> x `mod` 4 == 0) (fromList [4,8,12,14,16])
-- fromList [14,16]
dropWhile :: (a -> Bool) -> Heap a -> Heap a
dropWhile = withList . L.dropWhile
{-# INLINE dropWhile #-}
-- | /O(n log n)/. Remove duplicate entries from the heap.
--
-- >>> nub (fromList [1,1,2,6,6])
-- fromList [1,2,6]
nub :: Heap a -> Heap a
nub Empty = Empty
nub h@(Heap _ leq t) = insertWith leq x (nub zs)
where
x = root t
xs = deleteMin h
zs = dropWhile (`leq` x) xs
{-# INLINE nub #-}
-- | /O(n)/. Construct heaps from each element in another heap, and union them together.
--
-- >>> concatMap (\a -> fromList [a,a+1]) (fromList [1,4])
-- fromList [1,4,5,2]
concatMap :: (a -> Heap b) -> Heap a -> Heap b
concatMap _ Empty = Empty
concatMap f (Heap _ _ t) = F.foldMap f t
{-# INLINE concatMap #-}
-- | /O(n log n)/. Group a heap into a heap of heaps, by unioning together duplicates.
--
-- >>> group (fromList "hello")
-- fromList [fromList "e",fromList "h",fromList "ll",fromList "o"]
group :: Heap a -> Heap (Heap a)
group Empty = Empty
group h@(Heap _ leq _) = groupBy (flip leq) h
{-# INLINE group #-}
-- | /O(n log n)/. Group using a user supplied function.
groupBy :: (a -> a -> Bool) -> Heap a -> Heap (Heap a)
groupBy _ Empty = Empty
groupBy f h@(Heap _ leq t) = insert (insertWith leq x ys) (groupBy f zs)
where
x = root t
xs = deleteMin h
(ys,zs) = span (f x) xs
{-# INLINE groupBy #-}
-- | /O(n log n + m log m)/. Intersect the values in two heaps, returning the value in the left heap that compares as equal
intersect :: Heap a -> Heap a -> Heap a
intersect Empty _ = Empty
intersect _ Empty = Empty
intersect a@(Heap _ leq _) b = go leq (F.toList a) (F.toList b)
where
go leq' xxs@(x:xs) yys@(y:ys) =
if leq' x y
then if leq' y x
then insertWith leq' x (go leq' xs ys)
else go leq' xs yys
else go leq' xxs ys
go _ [] _ = empty
go _ _ [] = empty
{-# INLINE intersect #-}
-- | /O(n log n + m log m)/. Intersect the values in two heaps using a function to generate the elements in the right heap.
intersectWith :: Ord b => (a -> a -> b) -> Heap a -> Heap a -> Heap b
intersectWith _ Empty _ = Empty
intersectWith _ _ Empty = Empty
intersectWith f a@(Heap _ leq _) b = go leq f (F.toList a) (F.toList b)
where
go :: Ord b => (a -> a -> Bool) -> (a -> a -> b) -> [a] -> [a] -> Heap b
go leq' f' xxs@(x:xs) yys@(y:ys)
| leq' x y =
if leq' y x
then insert (f' x y) (go leq' f' xs ys)
else go leq' f' xs yys
| otherwise = go leq' f' xxs ys
go _ _ [] _ = empty
go _ _ _ [] = empty
{-# INLINE intersectWith #-}
-- | /O(n log n)/. Traverse the elements of the heap in sorted order and produce a new heap using 'Applicative' side-effects.
traverse :: (Applicative t, Ord b) => (a -> t b) -> Heap a -> t (Heap b)
traverse f = fmap fromList . T.traverse f . F.toList
{-# INLINE traverse #-}
-- | /O(n log n)/. Traverse the elements of the heap in sorted order and produce a new heap using 'Monad'ic side-effects.
mapM :: (Monad m, Ord b) => (a -> m b) -> Heap a -> m (Heap b)
mapM f = liftM fromList . T.mapM f . F.toList
{-# INLINE mapM #-}
both :: (a -> b) -> (a, a) -> (b, b)
both f (a,b) = (f a, f b)
{-# INLINE both #-}
-- we hold onto the children counts in the nodes for /O(1)/ 'size'
data Tree a = Node
{ rank :: {-# UNPACK #-} !Int
, root :: a
, _forest :: !(Forest a)
} deriving (Show,Read,Typeable)
data Forest a = !(Tree a) `Cons` !(Forest a) | Nil
deriving (Show,Read,Typeable)
infixr 5 `Cons`
instance Functor Tree where
fmap f (Node r a as) = Node r (f a) (fmap f as)
instance Functor Forest where
fmap f (a `Cons` as) = fmap f a `Cons` fmap f as
fmap _ Nil = Nil
-- internal foldable instances that should only be used over commutative monoids
instance Foldable Tree where
foldMap f (Node _ a as) = f a `mappend` F.foldMap f as
-- internal foldable instances that should only be used over commutative monoids
instance Foldable Forest where
foldMap f (a `Cons` as) = F.foldMap f a `mappend` F.foldMap f as
foldMap _ Nil = mempty
link :: (a -> a -> Bool) -> Tree a -> Tree a -> Tree a
link f t1@(Node r1 x1 cf1) t2@(Node r2 x2 cf2) -- assumes r1 == r2
| f x1 x2 = Node (r1+1) x1 (t2 `Cons` cf1)
| otherwise = Node (r2+1) x2 (t1 `Cons` cf2)
skewLink :: (a -> a -> Bool) -> Tree a -> Tree a -> Tree a -> Tree a
skewLink f t0@(Node _ x0 cf0) t1@(Node r1 x1 cf1) t2@(Node r2 x2 cf2)
| f x1 x0 && f x1 x2 = Node (r1+1) x1 (t0 `Cons` t2 `Cons` cf1)
| f x2 x0 && f x2 x1 = Node (r2+1) x2 (t0 `Cons` t1 `Cons` cf2)
| otherwise = Node (r1+1) x0 (t1 `Cons` t2 `Cons` cf0)
ins :: (a -> a -> Bool) -> Tree a -> Forest a -> Forest a
ins _ t Nil = t `Cons` Nil
ins f t (t' `Cons` ts) -- assumes rank t <= rank t'
| rank t < rank t' = t `Cons` t' `Cons` ts
| otherwise = ins f (link f t t') ts
uniqify :: (a -> a -> Bool) -> Forest a -> Forest a
uniqify _ Nil = Nil
uniqify f (t `Cons` ts) = ins f t ts
unionUniq :: (a -> a -> Bool) -> Forest a -> Forest a -> Forest a
unionUniq _ Nil ts = ts
unionUniq _ ts Nil = ts
unionUniq f tts1@(t1 `Cons` ts1) tts2@(t2 `Cons` ts2) = case compare (rank t1) (rank t2) of
LT -> t1 `Cons` unionUniq f ts1 tts2
EQ -> ins f (link f t1 t2) (unionUniq f ts1 ts2)
GT -> t2 `Cons` unionUniq f tts1 ts2
skewInsert :: (a -> a -> Bool) -> Tree a -> Forest a -> Forest a
skewInsert f t ts@(t1 `Cons` t2 `Cons`rest)
| rank t1 == rank t2 = skewLink f t t1 t2 `Cons` rest
| otherwise = t `Cons` ts
skewInsert _ t ts = t `Cons` ts
{-# INLINE skewInsert #-}
skewMeld :: (a -> a -> Bool) -> Forest a -> Forest a -> Forest a
skewMeld f ts ts' = unionUniq f (uniqify f ts) (uniqify f ts')
{-# INLINE skewMeld #-}
getMin :: (a -> a -> Bool) -> Forest a -> (Tree a, Forest a)
getMin _ (t `Cons` Nil) = (t, Nil)
getMin f (t `Cons` ts)
| f (root t) (root t') = (t, ts)
| otherwise = (t', t `Cons` ts')
where (t',ts') = getMin f ts
getMin _ Nil = error "Heap.getMin: empty forest"
splitForest :: Int -> Forest a -> Forest a -> Forest a -> (Forest a, Forest a, Forest a)
splitForest a b c d | a `seq` b `seq` c `seq` d `seq` False = undefined
splitForest 0 zs ts f = (zs, ts, f)
splitForest 1 zs ts (t `Cons` Nil) = (zs, t `Cons` ts, Nil)
splitForest 1 zs ts (t1 `Cons` t2 `Cons` f)
-- rank t1 == 0
| rank t2 == 0 = (t1 `Cons` zs, t2 `Cons` ts, f)
| otherwise = (zs, t1 `Cons` ts, t2 `Cons` f)
splitForest r zs ts (t1 `Cons` t2 `Cons` cf)
-- r1 = r - 1 or r1 == 0
| r1 == r2 = (zs, t1 `Cons` t2 `Cons` ts, cf)
| r1 == 0 = splitForest (r-1) (t1 `Cons` zs) (t2 `Cons` ts) cf
| otherwise = splitForest (r-1) zs (t1 `Cons` ts) (t2 `Cons` cf)
where
r1 = rank t1
r2 = rank t2
splitForest _ _ _ _ = error "Heap.splitForest: invalid arguments"
withList :: ([a] -> [a]) -> Heap a -> Heap a
withList _ Empty = Empty
withList f hp@(Heap _ leq _) = fromListWith leq (f (F.toList hp))
{-# INLINE withList #-}
splitWithList :: ([a] -> ([a],[a])) -> Heap a -> (Heap a, Heap a)
splitWithList _ Empty = (Empty, Empty)
splitWithList f hp@(Heap _ leq _) = both (fromListWith leq) (f (F.toList hp))
{-# INLINE splitWithList #-}
-- | Explicit priority/payload tuples. Useful to build a priority queue using
-- a 'Heap', since the payload is ignored in the Eq/Ord instances.
--
-- @
-- myHeap = 'fromList' ['Entry' 2 \"World", 'Entry' 1 \"Hello", 'Entry' 3 "!"]
--
-- ==> 'foldMap' 'payload' myHeap ≡ "HelloWorld!"
-- @
data Entry p a = Entry { priority :: p, payload :: a }
deriving (Read,Show,Data,Typeable)
instance Functor (Entry p) where
fmap f (Entry p a) = Entry p (f a)
{-# INLINE fmap #-}
#if MIN_VERSION_base(4,8,0)
instance Bifunctor Entry where
bimap f g (Entry p a) = Entry (f p) (g a)
#endif
instance Foldable (Entry p) where
foldMap f (Entry _ a) = f a
{-# INLINE foldMap #-}
instance Traversable (Entry p) where
traverse f (Entry p a) = Entry p `fmap` f a
{-# INLINE traverse #-}
-- instance Comonad (Entry p) where
-- extract (Entry _ a) = a
-- extend f pa@(Entry p _) Entry p (f pa)
instance Eq p => Eq (Entry p a) where
(==) = (==) `on` priority
{-# INLINE (==) #-}
instance Ord p => Ord (Entry p a) where
compare = compare `on` priority
{-# INLINE compare #-}