Selftype with TupleType #2436
Selftype with TupleType #2436
Conversation
|
So the real bug was not actually in #2408, but in my original implementation of selftype. |
|
Tested, no problems. I'll come back later with a review. |
| @@ -32,9 +32,9 @@ def analyze_member_access(name: str, | |||
| is_operator: bool, | |||
| builtin_type: Callable[[str], Instance], | |||
| not_ready_callback: Callable[[str, Context], None], | |||
| msg: MessageBuilder, | |||
| msg: MessageBuilder, *, | |||
| original_type: Type, | |||
There was a problem hiding this comment.
Now that you've made this mandatory and passed it explicitly everywhere, do you still need line 53 below? (original_type = original_type or typ -- I can't put a comment there.)
There was a problem hiding this comment.
No, I was about to remove it
| @@ -391,7 +391,7 @@ def f(x: Union[List[int], List[str], int]) -> None: | |||
| else: | |||
| x[0] # E: Value of type "int" is not indexable | |||
| x + 1 | |||
| x[0] # E: Value of type "int" is not indexable | |||
| x[0] # E: Value of type "Union[List[int], List[str], int]" is not indexable | |||
There was a problem hiding this comment.
I'm guessing this is because of the better determination of original_type right? I approve of the new error message.
|
(The last commit also makes some cosmetic changes) |
|
Thanks! This is a real improvement. |
|
BTW Can you write up some docs for self-type? I think it would be nice if we had some examples for this in mypy's chapter on generics. |
|
We are probably going to have a mypy release this week. It would be good to have the docs so that we can ask users to give it a spin. |
|
#2450. I am not confident with the examples, the phrasing, or the English grammar :| |
(Reopening #2408 after the revert #2414. This fixes parts of #2090)
This PR does three things:
original_typerecursively)_replace()and_make()return selftype, serving as test case for (1) and (2)As a consequence of (1), some error messages are changed, as exemplified in
check-isinstance.test. I think it's better now, but if it isn't, perhaps we should separateoriginal_typeandreport_typeas discussed in #2193.