On Tue, Nov 24, 2020 at 10:20:58AM +0100, Henning Thielemann wrote:
Rather than waiting, it is perhaps more pragmatic to go with a custom Prelude, which one can have now, even for versions of GHC/base, that won't have the feature in question:
I have such a Prelude: https://hackage.haskell.org/package/prelude-compat
It helps a bit, but does not fully solve the problem.
If you use Fold.all on a Map and then switch from Map k a to (Map k a, b) you will again not encounter a warning nor a type error.
Yes, for that one might have to be disciplined and use a monomorphised variant: allMapValues :: (a -> Bool) -> Map k a -> Bool allMapValues = Data.Foldable.all although the risk is only present when (a ~ b), or the predicate is sufficiently polymorphic (over some superclass of `a` and `b`). And of course, when refactoring, instead of moving to a 2-tuple, that brings in possibly unwanted instances, one can instead choose a custom product type, that does not have a Foldable instance: -- coercible to a 2-tuple if/as needed newtype T2 a b = T2 { _unT2 :: (a, b) } deriving (Eq, Ord, Show) to give one's 2-tuples exactly the desired instances and no more. This admittedly is not terribly ergonomic. -- Viktor.