@Blaisorblade To take the idea further, would type projections T#X still be unsound if we looked only at the upper bound of X in T, and not at its lower bound? As in, if we had class T{ type X >: S <: String } we could manipulate values of type T#X as strings, but we could not, for example, give values of type S to functions whose domain is T#X.
I could see such restricted projections still being super useful.
4 Likes