Extending the span of typeclasses with mono-traversable Abstract Data Types.
The standard Haskell typeclass library is based on classes for types parameterized on the element type, e.g. [a], Seq a. This excludes from its use the collection types with a fixed element type, eg. ByteString, Text, ....
The mono-traversable library uses class associated types and type families to bring type classes with a mono-kinded type index (kind: *) suited to all sort of collection types, whether parameterized or not.
-- see https://hackage.haskell.org/package/mono-traversable/docs/Data-Sequences.html#t:SemiSequence
class SemiSequence t -- Semigroup based sequence operations
type Index t -- the type of the index as a class associated one
-- type Element t -- possible Element associated type
-- detacching the Element type from the classes avoids having Foldable as a derived class dependent on the structure
type family Element t
-- Element for sequences
type instance Element [a] = a
type instance Element ByteString = Word8
type instance Element Text = Char
-- Element for sets
type instance Element IntSet = Int
type instance Element (Set a) = a
-- Element for dictionaries
type instance Element (Map k v) = v
type instance Element (IntMap v) = v
-- a MonoFoldable class for all sort of structures, since Element is not class dependent
class MonoFoldable t where
ofoldr :: (Element t -> b -> b) -> b -> t -> b
ofoldl' :: (a -> Element t -> a) -> a -> t -> a
ofoldMap :: Monoid m => (Element t -> m) -> t -> m
...
The function names which collide with names predefined in Prelude, like null, add an 'o' prefix, as onull to avoid continuos desambiguation.
Example of use:
{-| chunks.hs
It implements the function chunks of bounded length of a sequence,
returning the sequence of sequences, abstracting over any sequence type
-}
{-# LANGUAGE PackageImports, TypeFamilies #-}
import "mono-traversable" Data.Sequences as S
import "mono-traversable" Data.MonoTraversable as M
import Data.Function ((&)) -- (&): backwards application
import Control.Monad (mfilter)
-- IsSequence implies Monoid and MonoFoldable to which belongs 'onull'
chunks :: (IsSequence t, IsSequence t', Element t' ~ t) => Index t -> t -> t'
chunks n seq = unfoldr (splitAndValidate n) seq
splitAndValidate :: IsSequence t => Index t -> t -> Maybe (t, t)
splitAndValidate n seq = seq & S.splitAt n
& Just
& mfilter (not . M.onull . fst)
unfoldr :: (IsSequence t, Element t ~ a) => (b -> Maybe (a, b)) -> b -> t
unfoldr f state = case f state of
Just (x, state') -> S.cons x $ unfoldr f state'
Nothing -> mempty Using it:
$ ghci
GHCi, version 7.10.3: http://www.haskell.org/ghc/ :? for help
Prelude> :load chunks
[1 of 1] Compiling Main ( chunks.hs, interpreted )
Ok, modules loaded: Main.
*Main> chunks 2 "abcdef" :: [String]
["ab","cd","ef"]
*Main> import Data.Vector as Vec
*Main Vec> chunks 2 (Vec.fromList [1..4]) :: Vector (Vector Int)
[[1,2],[3,4]]Sequences ADTs
- SemiSequence: Semigroup based sequence operations.
- IsSequence: extends SemiSequence with Monoid based operations.
- Textual: A sequence splittable in lines and words with letter case operations.
- LazySequence: A generic class (multiparameter) parameterized with a strict sequence companion.
- Utf8: A generic class for Textual data which can be encoded to and decoded from UTF8, parameterized with the binary counterpart.
Container ADTs
- SetContainer: (key, value) pairs containers as monoids (union and empty ops.), foldable on its elements (MonoFoldable)
- IsMap: A traversable SetContainer
- IsSet: A SetContainer whose elements are ContainerKeys
ADTs for all
The NonNull type wrapper grants a use of MonoFoldable without exceptions.
See NonNull.
-- Safely convert from an unsafe container to a safe non-empty container.
fromNullable :: MonoFoldable mono => mono -> Maybe (NonNull mono)
-- Safe foldings over ordered domains
maximum, minimum :: (MonoFoldable mono, Ord (Element mono)) => NonNull mono -> Element mono
