-Wredundant-constraints warns about constraints introduced via type synonyms.

I ran into this when using lens, here's a simplied example:

l :: Getter [a] Int
l = to length

In this case, Getter introduces constraints (Functor f, Contravariant f), but the Functor constraint is not used. This leads to a warning from -Wredundant-constraints:

<interactive>:1:6: warning: [-Wredundant-constraints]
    • Redundant constraint: Functor f
    • In the type signature for:
           l :: (Contravariant f, Functor f) => (Int -> f Int) -> [a] -> f [a]

This shouldn't warn, because the constraints were introduced by type synonym and so may well be opaque to the user.

Trac metadata

Trac field	Value
Version	8.0.1
Type	Bug
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Compiler
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture

changed weight to 5

added Tbug Trac import labels

I can see that's annoying, but it's hard to avoid. And it does tell you something useful, namely that l is less polymorphic, and therefore less usable, than it could be.

It's tantalising -- -Wredundant-constraints really is jolly useful but it is also sometimes annoying.

In the interest of not getting drowned in complaints by the users of lens on GHC 8, we'll likely just add the constraint to to and rely on {-# INLINE #-} dropping it out in practice. This is a common enough scenario to warrant us eating the noise on our side.

Of course, you'll get the same sort of thing if you go to declare a non-law abiding traversal as a fold, etc. There i'm afraid the only answer will be to squelch the warning. GHC can't see 'laws'.

added error messages label

I just got bit by this in quite a confusing way: there's no indication that the redundant constraint is coming from the type synonym, rather than the use site. Concretely, my code looked like this:

somethingLensy :: Functor f => Getter (f a) (f b)
somethingLensy = to etc

Note that Functor f is the same as the constraint in the definition of Getter. When GHC started complaining about a redundant Functor f, I was very confused, because somethingLensy didn't compile if I removed its constraint!

If there was some indication that the constraint being warned about was coming from a type synonym, I think the situation would be much improved. The warning itself was correct and eventually actionable once I figured out what was going on.

Ha! There are two f's here: one in your siganture and one in the (expansion of) the type synonym.

Can you give the exact program and the exact error message?

I agree this is confusing, but I don't think there'll be a quick fix. I'd be happy to discuss with anyone who wanted to put cycles into fixing it, though.

added Pnormal label