Suggest less liftM

[ndmitchell]

From #131 @watashi asked "In there any plan for hlint to suggest fmap or <$> instead of liftM in future? As ghc 7.10 is released, Functor is a supperclass of Monad now according to Functor-Applicative-Monad Proposal". I replied:

Yes, I'd be happy to do that now (specifically a ticket for me to do in future, or a patch from someone else doing it now). Taking a quick look, where I have patterns that match on liftM they should probably match on both fmap and <$> as well as liftM. For the liftM fusion rule, there should probably be fmap and <$> versions which stick to producing the same. Where I suggest something it should never be liftM. The only debate is if to suggest liftM ==> fmap, and I think my inclination is that's now a good idea too - I certainly never use liftM.

[watashi]

I love the idea of deprecating liftM. It's usually a good suggestion to follow and it will also reduce the number of "equivalent" or "duplicate" rules in hlint.

[ndmitchell]

True - if we have the rule to avoid liftM, we don't need to match on it - I hadn't thought of that.

[ndmitchell]

All done now, one hint to replace liftM with fmap, and all other liftM hints now suggest fmap.