{-# OPTIONS_GHC -fno-warn-incomplete-patterns -fno-warn-name-shadowing -fno-warn-orphans #-} ----------------------------------------------------------------------------- -- | -- Module : Data.Collections -- Copyright : (c) Jean-Philippe Bernardy, 2006 -- License : BSD3 -- Maintainer : jeanphilippe.bernardy; google mail. -- Stability : experimental -- -- This module defines a class framework for collection types. It provides: -- -- * Classes for the most common type of collections -- -- * /View/ types to change the type of a collection, so it implements other classes. -- This allows to use types for purposes that they are not originally designed for. (eg. 'ElemsView') -- -- * A few generic functions for handling collections. -- -- * Infix (operator) version of common functions. -- -- Should you need a more precise documentation, "Data.Collections.Properties" lists laws that -- implementations are entitled to assume. -- -- The classes defined in this module are intended to give hints about performance. -- eg. if a function has a @'Map' c k v@ context, this indicates that the function -- will perform better if @c@ has an efficitent lookup function. -- -- This class framework is based on ideas found in Simon Peyton Jones, \"/Bulk types with class/\". -- -- -- Another inspiration source are the examples of MPTC and fuctional dependencies in Oleg Kiselyov's -- many articles posted to the haskell mailing list. -- -- -- This module name-clashes with a lot of Prelude functions, subsuming those. -- The user is encouraged to import Prelude hiding the clashing functions. -- Alternatively, it can be imported @qualified@. -- {- Selling points: * Unification of Map and Set (required by the below) * inclusion of Arrays * Good integration with existing base libraries * Relative simplicity: few classes, not too many methods, very little redundancy. * Reuses the same identifiers as other standard hierarchy modules. Conversion from the module-based API to this class-based one should be easy. * Comprehensive set of properties that define the behaviour of the classes. * Compatibility with GHC and Hugs. Bad points * Extra complexity due to heavy usage of MTPC (although imho it's a matter of getting used to it) TODO: * test with nhc98/hugs * add missing functions (partition, ..., ?) * optimizations (rules pragmas) * see how multimap/multiset fits this scheme. * Think about class Map' :: (* -> *) -> * -> $ * Fix infelicity about null map test; (== mempty). -} module Data.Collections ( -- * Classes -- ** Unfoldable Unfoldable(..), -- ** Collection Collection(..), SortingCollection(..), -- ** Map Map(..), lookupWithDefault, unionsWith, intersectionWith', differenceWith', mapWithKey', (!), -- ** Set Set(..), unions, notMember, (\\), -- ** Sequence Sequence(..), head, tail, append, concat, concatMap, -- length, (<|), (|>), (><), -- ** Others Array(..), Indexed(..), -- * Conversions fromFoldable, fromAscFoldable, fromList, fromListWith, fromAscList, -- * Views KeysView(..), ElemsView(..), withKeys, withElems, -- * Foldable module Data.Collections.Foldable, ) where -- import Prelude (Bool(..), Int, Maybe(..), -- (==), (.), (+), ($), (-), (&&), (||), -- Eq, Ord, -- error, const, not, fst, snd, maybe, head, otherwise, curry, uncurry, flip, -- min, max, Show) import Prelude hiding (sum,concat,lookup,map,filter,foldr,foldr1,foldl,null,reverse,(++),minimum,maximum,all,elem,concatMap,drop,head,tail,init) import Data.Monoid import Data.Collections.Foldable import qualified Data.Array as Array import qualified Data.List as List import qualified Data.Maybe as Maybe infixl 9 ! infixl 9 \\ -- infixr 5 >< infixr 5 <| infixl 5 |> ------------------------------------------------------------------------ -- * Type classes -- | Class of collection types. class (Foldable c a, Unfoldable c a) => Collection c a | c -> a where -- | @filter f c@ returns the collection of those elements that satisfy the predicate @f@. filter :: (a -> Bool) -> c -> c -- | Class of collection with unobservable elements. It is the dual of the 'Foldable' class. class Unfoldable c i | c -> i where -- | \'natural\' insertion of an element into a collection. insert :: i -> c -> c --insert i c = cofold (\Right c -> Right c; Left (i,c) -> Left (i,Right c)) (Left (i,c)) -- | The empty collection. empty :: c empty = unfold (const Nothing) undefined -- | Creates a collection with a single element. singleton :: i -> c singleton i = insert i empty -- | Insert all the elements of a foldable. insertMany :: Foldable c' i => c' -> c -> c insertMany c' c = foldr insert c c' -- At first sight, it looks like the above could just use List instead of any Foldable. -- However, it would then be more difficult to ensure that the conversion could be made -- very efficient between certain types. -- | Same as insertMany, but with the unchecked precondition that the input 'Foldable' is sorted. insertManySorted :: Foldable c' i => c' -> c -> c insertManySorted = insertMany unfold :: Unfoldable c a => (b -> Maybe (a, b)) -> b -> c unfold f b = insertMany (List.unfoldr f b) empty class Collection c o => SortingCollection c o where minView :: c -> Maybe (o,c) -- isSorted :: (Ord a, Foldable c a) => c -> Bool -- isSorted = fst . foldr cmp (True, Nothing) -- where curr `cmp` (acc, prev) = (acc && maybe True (curr <=) prev, Just curr) -- | Conversion from a Foldable to a Collection. fromFoldable :: (Foldable f a, Collection c' a) => f -> c' fromFoldable = flip insertMany empty -- TODO: Should be specialized (RULE pragmas) so it's efficient when converting from/to set/maps -- | Conversion from a Foldable to a Collection, with the /unchecked/ precondition that the input is sorted fromAscFoldable :: (Foldable f a, Collection c' a) => f -> c' fromAscFoldable = flip insertManySorted empty -- | Converts a list into a collection. fromList :: Collection c a => [a] -> c fromList = fromFoldable -- | Converts a list into a collection, with the precondition that the input is sorted. fromAscList :: Collection c a => [a] -> c fromAscList = fromAscFoldable -- | Class of sequential-access types. -- In addition of the 'Collection' services, it provides deconstruction and concatenation. class (Monoid c, Collection c a) => Sequence c a where -- | The first @i@ elements of a sequence. take :: Int -> c -> c -- | Elements of a sequence after the first @i@. drop :: Int -> c -> c -- | Split a sequence at a given index. splitAt :: Int -> c -> (c,c) -- | Reverse a sequence. reverse :: c -> c -- | Analyse the left end of a sequence. front :: c -> Maybe (a,c) -- | Analyse the right end of a sequence. back :: c -> Maybe (c,a) -- | Add an element to the left end of a sequence. cons :: a -> c -> c -- | Add an element to the right end of a sequence. snoc :: c -> a -> c -- | The 'isPrefix' function takes two seqences and returns True iff -- the first is a prefix of the second. isPrefix :: Eq a => c -> c -> Bool cons = insert isPrefix s1 s2 = case front s1 of Nothing -> True Just (x,xs) -> case front s2 of Nothing -> False Just (y,ys) -> x == y && isPrefix xs ys -- -- | Length of a sequence -- length :: Sequence c i o => c -> Int -- length = size -- | Concatenate two sequences. append :: Sequence c a => c -> c -> c append = mappend -- TODO: span ? -- | Infix version of 'cons': add an element to the left end of a sequence. -- Mnemonic: a triangle with the single element at the pointy end. (<|) :: Sequence c i => i -> c -> c (<|) = cons -- | Infix version of 'snoc': add an element to the right end of a sequence. -- Mnemonic: a triangle with the single element at the pointy end. (|>) :: Sequence c i => c -> i -> c (|>) = snoc -- | Infix verion of 'append'. Concatenate two sequences. (><) :: Sequence c a => c -> c -> c (><) = append -- | The concatenation of all the elements of a container of sequences. concat :: (Sequence s a, Foldable t s) => t -> s concat = fold -- | Map a function over all the elements of a container and concatenate -- the resulting sequences. concatMap :: (Sequence s b, Foldable t a) => (a -> s) -> t -> s concatMap = foldMap head :: Sequence s a => s -> a head = fst . Maybe.fromJust . front tail :: Sequence s a => s -> s tail = drop 1 -- | Class of indexed types. -- The collection is 'dense': there is no way to /remove/ an element nor for lookup -- to return "not found". -- -- In practice however, most shallow collection types will instanciate this -- class in addition of 'Map', and leave the responsibility of failure to the caller. class Indexed c k v | c -> k v where -- | @index c k@ returns element associated to @k@ index :: k -> c -> v -- | @adjust f k c@ applies @f@ to element associated to @k@ and returns the resulting collection. adjust :: (v -> v) -> k -> c -> c -- | if @inDomain k c@, then @index c k@ is guaranteed not to fail. inDomain :: k -> c -> Bool -- | Constructs a collection identical to the first argument except that it has -- been updated by the associations in the right argument. -- For example, if @m@ is a 1-origin, @n@ by @n@ matrix, then -- -- > m//[((i,i), 0) | i <- [1..n]] -- -- is the same matrix, except with the diagonal zeroed. (//) :: Foldable l (k,v) => c -> l -> c (//) = foldr replace where replace (k,v) = adjust (const v) k -- | @'accum' f@ takes an array and an association list and accumulates -- pairs from the list into the array with the accumulating function @f@. -- Thus 'accumArray' can be defined using 'accum': accum :: Foldable l (k,v') => (v -> v' -> v) -> c -> l -> c accum f = foldr adjust' where adjust' (k,v') = adjust (\v->f v v') k -- | Infix version of 'index', with arguments swapped. (!) :: Indexed c k v => c -> k -> v (!) = flip index class (Array.Ix k, Foldable c (k,v), Indexed c k v) => Array c k v | c -> k v where -- | if @(l,r) = bounds c@, then @inDomain k c <==> l <= k <= r@ bounds :: c -> (k,k) -- | Construct an array with the specified bounds and containing values -- for given indices within these bounds. -- -- The array is undefined (i.e. bottom) if any index in the list is -- out of bounds. The Haskell 98 Report further specifies that if any -- two associations in the list have the same index, the value at that -- index is undefined (i.e. bottom). However in GHC's implementation, -- the value at such an index is the value part of the last association -- with that index in the list. -- -- Because the indices must be checked for these errors, 'array' is -- strict in the bounds argument and in the indices of the association -- list, but nonstrict in the values. Thus, recurrences such as the -- following are possible: -- -- > a = array (1,100) ((1,1) : [(i, i * a!(i-1)) | i <- [2..100]]) -- -- Not every index within the bounds of the array need appear in the -- association list, but the values associated with indices that do not -- appear will be undefined (i.e. bottom). -- -- If, in any dimension, the lower bound is greater than the upper bound, -- then the array is legal, but empty. Indexing an empty array always -- gives an array-bounds error, but 'bounds' still yields the bounds -- with which the array was constructed. array :: Foldable l (k,v) => (k,k) -> l -> c -- | Class of map-like types. (aka. for sparse associative types). -- -- In opposition of Indexed, Map supports unexisting value for some indices. class Monoid c => Map c k a | c -> k a where -- | Remove a key from the keySet (and therefore the associated value in the Map). delete :: k -> c -> c delete = alter (const Nothing) -- | Tells whether an key is member of the keySet. member :: k -> c -> Bool member k = Maybe.isJust . lookup k -- | Union of two keySets. -- When duplicates are encountered, the keys may come from any of the two input sets. -- Values come from the map given as first arguement. union :: c -> c -> c union = unionWith const -- | Intersection of two keySets. -- -- When duplicates are encountered, the keys may come from any of the two input sets. -- Values come from the map given as first arguement. intersection :: c -> c -> c intersection = intersectionWith const -- | Difference of two keySets. -- Difference is to be read infix: @a `difference` b@ returns a set containing the -- elements of @a@ that are also absent from @b@. difference :: c -> c -> c difference = differenceWith (\_ _-> Nothing) -- | @s1 `isSubset` s2@ returns True iff. the keys in s1 form a subset of the keys in s2. isSubset :: c -> c -> Bool isSubset = isSubmapBy (\_ _->True) -- | @s1 `isProperSubset` s2@ returns True iff. @s1 `isProperSubset` s2@ and @s1 /= s2@ isProperSubset :: c -> c -> Bool isProperSubset = isProperSubmapBy (\_ _->True) -- Follows functions for fully-fledged maps. -- | Lookup the value at a given key. lookup :: k -> c -> Maybe a -- | Change the value associated to a given key. 'Nothing' represents no associated value. alter :: (Maybe a -> Maybe a) -> k -> c -> c alter f k m = case lookup k m of j@(Just _) -> case f j of Just a' -> insertWith (\a _ -> a) k a' m Nothing -> delete k m Nothing -> case f Nothing of Just a' -> insertWith (\a _ -> a) k a' m Nothing -> m -- | Insert with a combining function. -- -- @insertWith f key value m@ -- will insert the pair @(key, value)@ into @m@ if @key@ does -- not exist in the map. If the key does exist, the function will -- insert the pair @(key, f new_value old_value)@. insertWith :: (a -> a -> a) -> k -> a -> c -> c insertWith f k a c = alter (\x -> Just $ case x of {Nothing->a;Just a' -> f a a'}) k c -- | Convert a 'Foldable' to a 'Map', with a combining function. -- Note the applications of the combining function: -- @fromFoldableWith (+) [(k,x1), (k,x2), ..., (k,xn)] = fromFoldable [(k, xn + (... + (x2 + x1)))]@ -- or more generally @fromFoldableWith f [(k,x) | x <- l] = fromFoldable [(k,foldl1 (flip f) l)]@ -- 'foldGroups' is probably less surprising, so use it. fromFoldableWith :: Foldable l (k,a) => (a -> a -> a) -> l -> c fromFoldableWith f = foldr (uncurry (insertWith f)) mempty -- | Convert a 'Foldable' to a 'Map', with a combining function. -- @foldGroups f a l = let mkGroup g = (fst $ head g, foldr f a (map snd g)) in fromList . map mkGroup . groupBy ((==) `on` fst)) . toList@ foldGroups :: Foldable l (k,b) => (b -> a -> a) -> a -> l -> c foldGroups f a = foldr' (\(k,b) c -> (alter (\x -> Just $ case x of {Nothing->f b a;Just a' -> f b a'}) k c)) mempty -- | Apply a function over all values in the map. mapWithKey :: (k -> a -> a) -> c -> c -- | Union with a combining function. unionWith :: (a -> a -> a) -> c -> c -> c -- | Intersection with a combining function. intersectionWith :: (a -> a -> a) -> c -> c -> c -- | Difference with a combining function. differenceWith :: (a -> a -> Maybe a) -> c -> c -> c -- | isSubmapBy isSubmapBy :: (a -> a -> Bool) -> c -> c -> Bool -- | isProperSubmapBy isProperSubmapBy :: (a -> a -> Bool) -> c -> c -> Bool -- isProperSubmapBy f m1 m2 = isSubmapBy f m1 m2 && not (isEmpty (differenceWith (\_ _->Nothing) m1 m2)) -- | Tells whether a key is not a member of the keySet. notMember :: (Map c k a) => k -> c -> Bool notMember k s = not $ member k s -- | The expression @('lookupWithDefault' def k map)@ returns -- the value at key @k@ or returns @def@ when the key is not in the map. lookupWithDefault :: (Map c k a) => a -> k -> c -> a lookupWithDefault a k c = Maybe.fromMaybe a (lookup k c) -- | Specialized version of fromFoldableWith for lists. fromListWith :: (Map c k a) => (a -> a -> a) -> [(k,a)] -> c fromListWith = fromFoldableWith data O a b c = L !a | R !b | O !c -- | Same as 'intersectionWith', but with a more general type. intersectionWith' :: (Functor m, Map (m (O a b c)) k (O a b c)) => (a->b->c) -> m a -> m b -> m c intersectionWith' f m1 m2 = fmap extract (intersectionWith combine (fmap L m1) (fmap R m2)) where combine (L l) (R r) = O (f l r) extract (O a) = a -- | Same as 'differenceWith', but with a more general type. differenceWith' :: (Functor m, Map (m (O a b c)) k (O a b c)) => (a->b->Maybe c) -> m a -> m b -> m c differenceWith' f m1 m2 = fmap extract (differenceWith combine (fmap L m1) (fmap R m2)) where combine (L l) (R r) = fmap O (f l r) extract (O a) = a mapWithKey' :: (Functor m, Map (m (Either a b)) k (Either a b)) => (k -> a -> b) -> m a -> m b mapWithKey' f = fmap (either (error "mapWithKey': bug.") id) . mapWithKey f' . fmap Left where f' k (Left x) = Right (f k x) -- | Class for set-like collection types. A set is really a map -- with no value associated to the keys, -- so Set just states so. -- Note that this should be a class alias, if it existed. -- See: http://repetae.net/john/recent/out/classalias.html class Map c k () => Set c k | c -> k where -- | Dummy method for haddock to accept the class. haddock_candy :: c -> k -- | Infix version of 'difference'. Difference of two (key) sets. (\\) :: Map c k a => c -> c -> c (\\) = difference -- NOTE: the following two are only tentative, and thus not exported. -- | Infix version of 'union'. Union of two (key) sets. (\/) :: Map c k a => c -> c -> c (\/) = union -- | Infix version of 'intersection'. Intersection of two (key) sets. (/\) :: Map c k a => c -> c -> c (/\) = intersection {- Maybe it would be a good idea to bite the bullet and use a Lattice class for intersection and union. Maybe leave it unrelated to the Map class. In a separate module/package? Something like: class Lattice a where (/\) :: a -> a -> a (\/) :: a -> a -> a instance Lattice () where _ /\ _ = () _ \/ _ = () instance Lattice Bool where (/\) = (&&) (\/) = (||) instance (Lattice a, Map c k a) => Lattice c where (/\) = intersectionWith (/\) (\/) = unionWith (\/) -} -- | Union of many (key) sets. unions :: (Unfoldable s i, Map s k a, Foldable cs s) => cs -> s unions sets = foldl' union empty sets -- | Union of many (key) sets, with combining function unionsWith :: (Unfoldable s i, Map s k a, Foldable cs s) => (a->a->a) -> cs -> s unionsWith f sets = foldl' (unionWith f) empty sets ------------------------------------------------------------------------ -- Trickier stuff for alternate dictionnary usages -- | "View" to the keys of a dictionnary newtype KeysView m k v = KeysView {fromKeysView :: m} -- | "View" to the elements of a dictionnary newtype ElemsView m k v = ElemsView {fromElemsView :: m} -- The following requires undecidable instances. An alternative -- implementation is to define these instances directly on the -- concrete map types and drop the requirement for the aforementioned -- extension. type T a = a->a withKeys :: Collection m (k,v) => T (KeysView m k v) -> T m withKeys f c = fromKeysView $ f (KeysView c) withElems :: Collection m (k,v) => T (ElemsView m k v) -> T m withElems f c = fromElemsView $ f (ElemsView c) instance Foldable m (k,v) => Foldable (KeysView m k v) k where foldr f i (KeysView c) = foldr (f . fst) i c null (KeysView c) = null c instance Unfoldable m (k,v) => Unfoldable (KeysView m k v) (k,v) where empty = KeysView empty insert x (KeysView m) = KeysView $ insert x m singleton x = KeysView (singleton x) instance Foldable m (k,v) => Foldable (ElemsView m k v) v where foldr f i (ElemsView c) = foldr (f . snd) i c null (ElemsView c) = null c instance Unfoldable m (k,v) => Unfoldable (ElemsView m k v) (k,v) where empty = ElemsView empty insert x (ElemsView m) = ElemsView $ insert x m singleton x = ElemsView (singleton x) instance (Monoid m, Map m k v) => Monoid (KeysView m k v) where mempty = KeysView mempty mappend = union instance Map m k v => Map (KeysView m k v) k v where isSubmapBy f (KeysView m) (KeysView m') = isSubmapBy f m m' isProperSubmapBy f (KeysView m) (KeysView m') = isProperSubmapBy f m m' member k (KeysView m) = Maybe.isJust $ lookup k m union (KeysView m) (KeysView m') = KeysView $ union m m' difference (KeysView m) (KeysView m') = KeysView $ difference m m' intersection (KeysView m) (KeysView m') = KeysView $ intersection m m' delete k (KeysView m) = KeysView $ delete k m insertWith f k a (KeysView m) = KeysView $ insertWith f k a m lookup k (KeysView m) = lookup k m alter f k (KeysView m) = KeysView $ alter f k m unionWith f (KeysView m) (KeysView m') = KeysView $ unionWith f m m' differenceWith f (KeysView m) (KeysView m') = KeysView $ differenceWith f m m' intersectionWith f (KeysView m) (KeysView m') = KeysView $ intersectionWith f m m' mapWithKey f (KeysView m) = KeysView $ mapWithKey f m