Copyright | (C) 2011-2012 Edward Kmett |
---|---|
License | BSD-style (see the file LICENSE) |
Maintainer | Edward Kmett <ekmett@gmail.com> |
Stability | provisional |
Portability | non-portable (rank-2 polymorphism) |
Safe Haskell | Safe-Inferred |
Language | Haskell2010 |
"Free Monads for Less"
The most straightforward way of implementing free monads is as a recursive
datatype that allows for arbitrarily deep nesting of the base functor. This is
akin to a tree, with the leaves containing the values, and the nodes being a
level of Functor
over subtrees.
For each time that the fmap
or >>=
operations is used, the old tree is
traversed up to the leaves, a new set of nodes is allocated, and
the old ones are garbage collected. Even if the Haskell runtime
optimizes some of the overhead through laziness and generational garbage
collection, the asymptotic runtime is still quadratic.
On the other hand, if the Church encoding is used, the tree only needs to be constructed once, because:
- All uses of
fmap
are collapsed into a single one, so that the values on the _leaves_ are transformed in one pass.
fmap f . fmap g == fmap (f . g)
- All uses of
>>=
are right associated, so that every new subtree created is final.
(m >>= f) >>= g == m >>= (\x -> f x >>= g)
Asymptotically, the Church encoding supports the monadic operations more
efficiently than the naïve Free
.
This is based on the "Free Monads for Less" series of articles by Edward Kmett:
- newtype F f a = F {
- runF :: forall r. (a -> r) -> (f r -> r) -> r
- improve :: Functor f => (forall m. MonadFree f m => m a) -> Free f a
- fromF :: MonadFree f m => F f a -> m a
- iterM :: (Monad m, Functor f) => (f (m a) -> m a) -> F f a -> m a
- toF :: Functor f => Free f a -> F f a
- retract :: Monad m => F m a -> m a
- hoistF :: (forall x. f x -> g x) -> F f a -> F g a
- class Monad m => MonadFree f m | m -> f where
- wrap :: f (m a) -> m a
- liftF :: (Functor f, MonadFree f m) => f a -> m a
Documentation
The Church-encoded free monad for a functor f
.
It is asymptotically more efficient to use (>>=
) for F
than it is to (>>=
) with Free
.
MonadTrans F | |
MonadReader e m => MonadReader e (F m) | |
MonadState s m => MonadState s (F m) | |
MonadWriter w m => MonadWriter w (F m) | |
Functor f => MonadFree f (F f) | |
Alternative f => Alternative (F f) | |
Monad (F f) | |
Functor (F f) | |
MonadFix (F f) | |
MonadPlus f => MonadPlus (F f) | |
Applicative (F f) | |
(Foldable f, Functor f) => Foldable (F f) | |
MonadCont m => MonadCont (F m) | |
Apply (F f) | |
Bind (F f) |
improve :: Functor f => (forall m. MonadFree f m => m a) -> Free f a Source
Improve the asymptotic performance of code that builds a free monad with only binds and returns by using F
behind the scenes.
This is based on the "Free Monads for Less" series of articles by Edward Kmett:
and "Asymptotic Improvement of Computations over Free Monads" by Janis Voightländer.
iterM :: (Monad m, Functor f) => (f (m a) -> m a) -> F f a -> m a Source
Like iter for monadic values.
hoistF :: (forall x. f x -> g x) -> F f a -> F g a Source
Lift a natural transformation from f
to g
into a natural transformation from F f
to F g
.
class Monad m => MonadFree f m | m -> f where Source
Monads provide substitution (fmap
) and renormalization (join
):
m>>=
f =join
(fmap
f m)
A free Monad
is one that does no work during the normalization step beyond simply grafting the two monadic values together.
[]
is not a free Monad
(in this sense) because
smashes the lists flat.join
[[a]]
On the other hand, consider:
data Tree a = Bin (Tree a) (Tree a) | Tip a
instanceMonad
Tree wherereturn
= Tip Tip a>>=
f = f a Bin l r>>=
f = Bin (l>>=
f) (r>>=
f)
This Monad
is the free Monad
of Pair:
data Pair a = Pair a a
And we could make an instance of MonadFree
for it directly:
instanceMonadFree
Pair Tree wherewrap
(Pair l r) = Bin l r
Or we could choose to program with
instead of Free
PairTree
and thereby avoid having to define our own Monad
instance.
Moreover, Control.Monad.Free.Church provides a MonadFree
instance that can improve the asymptotic complexity of code that
constructs free monads by effectively reassociating the use of
(>>=
). You may also want to take a look at the kan-extensions
package (http://hackage.haskell.org/package/kan-extensions).
See Free
for a more formal definition of the free Monad
for a Functor
.
Nothing
(Functor f, MonadFree f m) => MonadFree f (ListT m) | |
(Functor f, MonadFree f m) => MonadFree f (IdentityT m) | |
(Functor f, MonadFree f m) => MonadFree f (MaybeT m) | |
Functor f => MonadFree f (Free f) | |
Functor f => MonadFree f (F f) | |
Monad m => MonadFree Identity (IterT m) | |
(Functor f, MonadFree f m, Error e) => MonadFree f (ErrorT e m) | |
(Functor f, MonadFree f m, Monoid w) => MonadFree f (WriterT w m) | |
(Functor f, MonadFree f m, Monoid w) => MonadFree f (WriterT w m) | |
(Functor f, MonadFree f m) => MonadFree f (ContT r m) | |
(Functor f, MonadFree f m) => MonadFree f (StateT s m) | |
(Functor f, MonadFree f m) => MonadFree f (StateT s m) | |
(Functor f, MonadFree f m) => MonadFree f (ReaderT e m) | |
(Functor f, Monad m) => MonadFree f (FreeT f m) | |
Functor f => MonadFree f (FT f m) | |
(Functor f, MonadFree f m, Monoid w) => MonadFree f (RWST r w s m) | |
(Functor f, MonadFree f m, Monoid w) => MonadFree f (RWST r w s m) |