In article <16071.60350.411868.559865@tux-17.corp.peace.com>, Tom Pledger wrote:

Hal Daume III writes: | The docs claim this instance exists, but it doesn't seem to. | Moreoever, such an instance doesn't make sense to me.

The following is from Control.Monad.Error.

instance MonadPlus IO where mzero = ioError (userError "mzero") m `mplus` n = m `catch` \_ -> n

I tend to agree with Hal. This muddles up the semantics of MonadPlus IMO. I think mplus/mzero should do proper backtracking. For instance: test = do { t <- (return True) `mplus` (return False); if t then mzero else return (); } In my opinion this test should always return rather than be mzero. At the very least it's useful have separate "plus" operators that do and don't do proper backtracking. In HBase I have separate classes: class (Monad m) => MonadZero m where { mzero :: m a; }; class (MonadZero m) => MonadOr m where { mOrElse :: m a -> m a -> m a; }; class (MonadOr m) => MonadPlus m where { mplus :: m a -> m a -> m a; msum :: [m a] -> m a; returnMany :: [a] -> m a; msum [] = mzero; msum (p:ps) = mplus p (msum ps); returnMany = msum . (fmap return); }; class (MonadPlus m) => MonadFirst m where { mfirst :: m a -> m a; }; As a gloss for their meanings, these are the instances for []: instance MonadZero [] where { mzero = []; }; instance MonadOr [] where { mOrElse [] = id; mOrElse (a:_) = const [a]; }; instance MonadPlus [] where { mplus = (++); msum = concatenateList; returnMany = id; }; instance MonadFirst [] where { mfirst [] = []; mfirst (x:_) = [x]; }; -- Ashley Yakeley, Seattle WA