[ty] Clarify behavior of constraint sets for gradual upper bounds and constraints #21287
Conversation
dcreager
requested review from
AlexWaygood,
carljm and
sharkdp
as code owners
dcreager
added
internal
Diagnostic diff on typing conformance testsNo changes detected when running ty on typing conformance tests ✅ |
|
| must satisfy the constraint set. (That means we will almost always choose `Never` — or more | ||
| precisely, the bottom specialization — as the upper bound.) In inferable positions, the opposite is |
There was a problem hiding this comment.
I think that only holds true if the upper bound of a typevar is a gradual specialization of an invariant type? The bottom materialization of list[Any] is Never, but the bottom materialization of Sequence[Any] is Sequence[Never], since Sequence[Never] is a subtype of Sequence[int] and Sequence[str], and is still an inhabited type ([], () can both inhabit Sequence[Never], for example). Similarly for contravariance, the bottom materialization of Callable[[Any], int] is Callable[[object], int].
There was a problem hiding this comment.
Which part doesn't hold true? This might be poor wording on my part — I am not trying to claim that the bottom materialization is always Never. I am (attempting to) say that for T: Any, we pick T: Never, but it's more precise to say that we pick the bottom specialization, because for T: list[Any], we pick T: Bottom[list[Any]].
There was a problem hiding this comment.
might be poor wording on my part
I think I understood it the way you explained in your comment now, but I also stumbled over it.
There was a problem hiding this comment.
This might be poor wording on my part — I am not trying to claim that the bottom materialization is always
Never.
Ah, okay -- yes, I think the wording can be improved there a little!
When checking whether a constraint set is satisfied, if a typevar has a non-fully-static upper bound or constraint, we are free to choose any materialization that makes the check succeed.
In non-inferable positions, we have to show that the constraint set is satisfied for all valid specializations, so it's best to choose the most restrictive materialization, since that minimizes the set of valid specializations that have to pass.
In inferable positions, we only have to show that the constraint set is satisfied for some valid specializations, so it's best to choose the most permissive materialization, since that maximizes our chances of finding a specialization that passes.