I would go for a split of the currying and proposed solutions. Logically y = x |> f2(a, _) y = x |> f2(_, b) are currying, but the implementation doesn't

I would go for a split of the currying and proposed solutions. Logically y = x |> f2(a, _) y = x |> f2(_, b) are currying, but the implementation doesn't need to create functions any more than if the _ was %. And fb = f2(a, _) fa = f2(_, b) are explicit currying which would be nice to have at times, but would mostly use in piped context at first, perhaps to give it a more meaningful or narrowed-scope name.