Start factoring out the logic that computes the mapping from one tuple type to another. We're only using it in one place now (the tuple-to-tuple coercion), but it should be picked up by the constra...