Safe Haskell | Safe-Inferred |
---|
You can think of Shell
as []
+ IO
+ Managed
. In fact, you can embed
all three of them within a Shell
:
select :: [a] -> Shell a liftIO :: IO a -> Shell a using :: Managed a -> Shell a
Those three embeddings obey these laws:
do { x <- select m; select (f x) } = select (do { x <- m; f x }) do { x <- liftIO m; liftIO (f x) } = liftIO (do { x <- m; f x }) do { x <- with m; using (f x) } = using (do { x <- m; f x }) select (return x) = return x liftIO (return x) = return x using (return x) = return x
... and select
obeys these additional laws:
select xs <|> select ys = select (xs <|> ys) select empty = empty
You typically won't build Shell
s using the Shell
constructor. Instead,
use these functions to generate primitive Shell
s:
-
empty
, to create aShell
that outputs nothing -
return
, to create aShell
that outputs a single value -
select
, to range over a list of values within aShell
-
liftIO
, to embed anIO
action within aShell
-
using
, to acquire aManaged
resource within aShell
Then use these classes to combine those primitive Shell
s into larger
Shell
s:
-
Alternative
, to concatenateShell
outputs using (<|>
) -
Monad
, to buildShell
comprehensions usingdo
notation
If you still insist on building your own Shell
from scratch, then the
Shell
you build must satisfy this law:
-- For every shell `s`: _foldIO s (FoldM step begin done) = do x <- begin x' <- _foldIO s (FoldM step (return x) return) done x'
... which is a fancy way of saying that your Shell
must call 'begin'
exactly once when it begins and call 'done'
exactly once when it ends.
- newtype Shell a = Shell {}
- foldIO :: MonadIO io => Shell a -> FoldM IO a r -> io r
- fold :: MonadIO io => Shell a -> Fold a b -> io b
- sh :: MonadIO io => Shell a -> io ()
- view :: (MonadIO io, Show a) => Shell a -> io ()
- select :: [a] -> Shell a
- liftIO :: MonadIO m => forall a. IO a -> m a
- using :: Managed a -> Shell a