data Tree t = Empty | Node t (Tree t) (Tree t)
We can create Functor instance and use
fmap :: (t -> a) -> Tree t -> Tree a
But what if instead of (t -> a) I want (Tree t -> a) so I could have access to a whole (Node t) not just t
treeMap :: (Tree t -> a) -> Tree t -> Tree a
treeMap f Empty = Empty
treeMap f n@(Node _ l r) = Node (f n) (treeMap f l) (treeMap f r)
Same with fold
treeFold :: (Tree t -> a -> a) -> a -> Tree t -> a
Is there any generalization over functions like these?
map :: (f t -> a) -> f t -> f a
fold :: (f t -> a -> a) -> a -> f t -> a
You've just discovered comonads! Well, almost.
class Functor f => Comonad f where
extract :: f a -> a
duplicate :: f a -> f (f a)
instance Comonad Tree where
extract (Node x _ _) = x -- this one only works if your trees are guaranteed non-empty
duplicate t@(Node n b1 b2) = Node t (duplicate b1) (duplicate b2)
With duplicate
you can implement your functions:
treeMap f = fmap f . duplicate
freeFold f i = foldr f i . duplicate
To do it properly, you should enforce non-emptyness by the type system:
type Tree' a = Maybe (Tree'' a)
data Tree'' t = Node' t (Tree' t) (Tree' t)
deriving (Functor)
instance Comonad Tree'' where
extract (Node' x _ _) = x
duplicate t@(Node' _ b1 b2) = Node' t (fmap duplicate b1) (fmap duplicate b2)
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With