The MaybeT monad. See http://www.haskell.org/haskellwiki/New_monads/MaybeT for more widely-used version. Our Functor instance and our implementation of >>= are borrowed from there.

Computation type:

Computations which may fail or return nothing.

Binding strategy:

Failure returns the value Nothing, bypassing any bound functions which follow. Success returns a value wrapped in Just.

Useful for:

Building computations from steps which may fail. No error information is returned. (If error information is required, see Error.)

The instance MonadPlus is not provided, because it has ambiguous semantics. It could refer to either

instance MonadPlus m => MonadPlus (MaybeT m)

...lifting the semantics of an underlying MaybeT monad, or

instance MonadPlus (MaybeT m)

...with semantics similar to MonadPlus Maybe.

Here is an example that shows how to use MaybeT to propagate an end-of-file condition in the IO monad. In the example below, both maybeReadLine and failIfQuit may cause a failure, which will propagate out to main without further intervention.

import System.Console.Readline
import Data.Maybe
import Control.Monad
import Control.Monad.Trans
import Control.Monad.Maybe

-- 'MaybeIO' is the type of computations which do IO, and which may fail.
type MaybeIO = MaybeT IO

-- 'readline' already has type 'String -> IO (Maybe String)'; we just need
-- to wrap it.
maybeReadLine :: String -> MaybeIO String
maybeReadLine prompt = MaybeT (readline prompt)

-- Fail if 'str' equals "quit".
failIfQuit :: (Monad m) => String -> m ()
failIfQuit str = when (str == "quit") (fail "Quitting")

-- This task may fail in several places.  Try typing Control-D or "quit" at
-- any prompt.
concatTwoInputs :: MaybeIO ()
concatTwoInputs = do
  s1 <- maybeReadLine "String 1> "
  failIfQuit s1
  s2 <- maybeReadLine "String 2> "
  failIfQuit s2
  liftIO (putStrLn ("Concatenated: " ++ s1 ++ s2))

-- Loop until failure.
main :: IO ()
main = do
  result <- runMaybeT concatTwoInputs
  if isNothing result
    then putStrLn "Bye!"
    else main