Being instance of Monoid is not checked

[boris]

It is possible pass not only instances of Monoid to tell function and load the file with GHCi. GHCi shows type f :: (MonadWriter? ((a -> b) -> [a] -> [b]) m) => t -> m ()

f _ = tell map

However, if code is

f = tell map

GHCi shows type error(no instance) as expected.

It seems very strange that type correctness depends on number of parameters for function that is not used anywhere.

[rwbarton]

I don't think this is a bug. GHCi should accept the first declaration because later (say in another module) someone might define an instance Monoid ((a -> b) -> [a] -> [b]) and then want to call the function f. The second declaration is rejected due to the monomorphism restriction: since f is not defined via a function binding, GHCi must choose a monomorphic type for f immediately and it has no way to satisfy the Monoid instance. If you turn off the monomorphism restriction by entering :set -XNoMonomorphismRestriction, you will find that both definitions are accepted.