Difference between revisions of "Type composition"
Jump to navigation
Jump to search
(added ref [1] and related comments) |
|||
Line 68: | Line 68: | ||
TT_T (loop (arr mergeA >>> k >>> arr splitA)) |
TT_T (loop (arr mergeA >>> k >>> arr splitA)) |
||
− | mergeA :: Applicative |
+ | mergeA :: Applicative f => (f a, f b) -> f (a,b) |
− | mergeA ~( |
+ | mergeA ~(fa,fb) = liftA2 (,) fa fb |
− | |||
⚫ | |||
⚫ | |||
⚫ | |||
⚫ | |||
Revision as of 19:44, 12 March 2007
Introduction
I'd like to get some forms of type composition into a standard library. Below is my first shot at it. I'm using these definitions in a new version of Phooey.
Comments & suggestions, please. Conal 23:16, 8 March 2007 (UTC)
Code, first draft
{-# OPTIONS -fglasgow-exts #-}
-- Various type constructor compositions and instances for them.
-- References:
-- [1] [http://www.soi.city.ac.uk/~ross/papers/Applicative.html Applicative Programming with Effects]
module Control.Compose ((:.:)(..), (:.::)(..), (::.:)(..), App(..)) where
import Control.Applicative
import Control.Arrow hiding (pure)
import Data.Monoid
-- | Composition of type constructors: unary & unary. Called "g . f"
-- in [1], section 5.
newtype (g :.: f) a = T_T { runT_T :: g (f a) }
instance (Functor g, Functor f) => Functor (g :.: f) where
fmap f (T_T m) = T_T (fmap (fmap f) m)
instance (Applicative g, Applicative f) => Applicative (g :.: f) where
pure = T_T . pure . pure
T_T getf <*> T_T getx = T_T (liftA2 (<*>) getf getx)
-- standard Monoid instance for Applicative applied to Monoid
instance (Applicative (f :.: g), Monoid a) => Monoid ((f :.: g) a) where
{ mempty = pure mempty; mappend = (*>) }
-- | Composition of type constructors: unary & binary. Called
-- "StaticArrow" in [1], section 6.
newtype (f :.:: (~>)) a b = T_TT { runT_TT :: f (a ~> b) }
instance (Applicative f, Arrow (~>)) => Arrow (f :.:: (~>)) where
arr = T_TT . pure . arr
T_TT g >>> T_TT h = T_TT (liftA2 (>>>) g h)
first (T_TT g) = T_TT (liftA first g)
-- For instance, /\ a b. f (a -> m b) =~ f :.:: Kleisli m
-- | Composition of type constructors: unary & binary.
-- Wolfgang Jeltsch pointed out a problem with these definitions: 'splitA'
-- and 'mergeA' are not inverses. The definition of 'first', e.g.,
-- violates the \"extension\" law and causes repeated execution. Look for
-- a reformulation or a clarification of required properties of the
-- applicative functor @f@.
newtype ((~>) ::.: f) a b = TT_T {runTT_T :: f a ~> f b}
instance (Arrow (~>), Applicative f) => Arrow ((~>) ::.: f) where
arr = TT_T . arr . liftA
TT_T g >>> TT_T h = TT_T (g >>> h)
first (TT_T a) =
TT_T (arr splitA >>> first a >>> arr mergeA)
instance (ArrowLoop (~>), Applicative f) => ArrowLoop ((~>) ::.: f) where
-- loop :: UI (b,d) (c,d) -> UI b c
loop (TT_T k) =
TT_T (loop (arr mergeA >>> k >>> arr splitA))
mergeA :: Applicative f => (f a, f b) -> f (a,b)
mergeA ~(fa,fb) = liftA2 (,) fa fb
splitA :: Applicative f => f (a,b) -> (f a, f b)
splitA fab = (liftA fst fab, liftA snd fab)
-- | Type application
newtype App f a = App { runApp :: f a }
-- Example: App IO ()
instance (Applicative f, Monoid m) => Monoid (App f m) where
mempty = App (pure mempty)
App a `mappend` App b = App (a *> b)
{-
-- We can also drop the App constructor, but then we overlap with many
-- other instances, like [a].
instance (Applicative f, Monoid a) => Monoid (f a) where
mempty = pure mempty
mappend = (*>)
-}