It does seem like the existence of e.g. readily accessible constructors for std::ranges::transform_view is a mistake.
I'm not convinced it's worthwhile for the standard to insist they can be constructed from outside the views implementation at all, but certainly there's no reason to leave the constructor where programmers will use it by mistake.
I'm not convinced it's worthwhile for the standard to insist they can be constructed from outside the views implementation at all, but certainly there's no reason to leave the constructor where programmers will use it by mistake.