Functional lenses

A lens consists of two functions, a getter and a setter:

data Lens a b = Lens { getter :: a -> b, setter :: b -> a -> a }

For example, we might have lenses for the first and second parts of a pair:

fstLens :: Lens (a, b) a
fstLens = Lens fst $ \x (a, b) -> (x, b)

sndLens :: Lens (a, b) b
sndLens = Lens snd $ \x (a, b) -> (a, x)

The real convenience of lenses is that they compose:

compose :: Lens b c -> Lens a b -> Lens a c
compose f g = Lens (getter f . getter g) $
                   \c a -> setter g (setter f c (getter g a)) a

And they mechanically convert to State transitions:

lensGet :: MonadState s m => Lens s a -> m a
lensGet = gets . getter

lensSet :: MonadState s m => Lens s b -> b -> m ()
lensSet f = modify . setter f

lensMod :: MonadState s m => Lens s b -> (b -> b) -> m ()
lensMod f g = modify $ setter f =<< g . getter f

(+=) :: (MonadState s m, Num b) => Lens s b -> b -> m ()
f += x = lensMod f (+ x)

See the answer to question lenses, fclabels, data-accessor - which library for structure access and mutation is better - it has a very clear explanation on lenses.

Also, the documentation for the Data.Lenses and fclabel libraries give some good examples of them being used.