Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix type narrowing of == None and in (None,) conditions #15760

Merged
merged 7 commits into from
Aug 2, 2023

Conversation

intgr
Copy link
Contributor

@intgr intgr commented Jul 25, 2023

Fixes #15759.

The type narrowing logic internally used is_optional() to check whether an operand type might be None. is_optional() only considers UnionType, and ignored an immediate NoneType type, which clearly is None-able.

I have created a new helper can_be_none(), which does also consider immediate NoneType.

The narrowing logic here could be further improved, but that's out of scope for the bugfix.

# Old behavior
def func(val: str | None) -> None:
    if val == None:
        reveal_type(val)  # ❌ Revealed type is "builtins.str"
    if val in (None,):
        reveal_type(val)  # ❌ Revealed type is "builtins.str"

# New behavior
def func(val: str | None) -> None:
    if val == None:
        reveal_type(val)  # ✅ Revealed type is "Union[builtins.str, None]"
    if val in (None,):
        reveal_type(val)  # ✅ Revealed type is "Union[builtins.str, None]"

TODO:

  • Add tests

@@ -141,7 +141,7 @@ def find_shallow_matching_overload_item(overload: Overloaded, call: CallExpr) ->
break
elif (
arg_none
and not is_optional(arg_type)
and not can_be_none(arg_type)
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It seems to me that this logic should also use the new can_be_none(), but I'm not sure how to test this code path.

@github-actions

This comment has been minimized.

@intgr intgr marked this pull request as ready for review July 25, 2023 12:31
@intgr intgr changed the title Fix type narrowing in == None and in (None,) conditions Fix type narrowing of == None and in (None,) conditions Jul 25, 2023
@github-actions

This comment has been minimized.

Comment on lines +1273 to +1289
if val == None:
reveal_type(val) # N: Revealed type is "Union[__main__.A, None]"
else:
reveal_type(val) # N: Revealed type is "Union[__main__.A, None]"
if val != None:
reveal_type(val) # N: Revealed type is "Union[__main__.A, None]"
else:
reveal_type(val) # N: Revealed type is "Union[__main__.A, None]"

if val in (None,):
reveal_type(val) # N: Revealed type is "Union[__main__.A, None]"
else:
reveal_type(val) # N: Revealed type is "Union[__main__.A, None]"
if val not in (None,):
reveal_type(val) # N: Revealed type is "Union[__main__.A, None]"
else:
reveal_type(val) # N: Revealed type is "Union[__main__.A, None]"
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

meant to fail?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm not sure what you're asking. This test passes.

Mypy doesn't support type narrowing for these cases yet, I'm only fixing its buggy behavior here in this PR.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ahh, got it now! 👌

Copy link
Contributor

@ikonst ikonst Jul 25, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

(I assume actually narrowing is much more work, or simply unrelated and could be done incrementally?)

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It's more tricky for sure, and quite rare in Python code. Not as easy as supporting is None, since classes (even a subclas of str) can override __eq__ and return True when compared to None.

I looked into this code because I want to solve #9718, I'd rather put my effort there for now :)

Copy link
Collaborator

@hauntsaninja hauntsaninja left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for fixing this... Can we just fold this logic into is_optional? I think everything works out (although maybe remove_optional should return UninhabitedType if fed None), and it seems less footgun-y

[case testNarrowingOptionalEqualsNone]
from typing import Optional

class A: ...
Copy link
Member

@sobolevn sobolevn Jul 26, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What about the difference between classes with and without custom __eq__?

In the original issue str instances can never be equal to None.

But, if some type has custom __eq__ - we cannot be sure about this anymore.

Copy link
Contributor Author

@intgr intgr Jul 26, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This comment is quite open-ended, there are so many nuances here that I'm left guessing what I should do.

This particular bug or fix isn't currently affected by __eq__ method.

Are you asking that I should add tests with str also? And add __eq__ method to A? It miiiight be helpful in the future when narrowing logic changes.

In the original issue str intances can never be equal to None.

One could possibly have a subclass instance of str that overrides __eq__. But is that considered unsound usage? I dunno.


There is another slightly related unsoundness case, when objects consider themselves equal to None. But that would be a separate bug and a harder discussion whether a fix would reduce the usefulness of narrowing.

https://mypy-play.net/?mypy=1.4.1&python=3.11&gist=d6199eb69f69c933ced7da5c51c25a50

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This comment is good enough for me, thanks! :)

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Opened another issue for the described case: #15764

@intgr
Copy link
Contributor Author

intgr commented Jul 26, 2023

@hauntsaninja

Can we just fold this logic into is_optional?

Makes sense, but the name seems potentially misleading. IMO "optional" implies "Union[T, None]". But NoneType by itself doesn't seem to fit the description "optional".

I could merge them under another name. So bikeshed time: can_be_none()? is_noneable()? is_possibly_none()? intersects_none()? overlaps_none()... None of these sound very good :)

I'm OK with keeping is_optional() too, I'd rather leave this decision up to mypy maintainers.

although maybe remove_optional should return UninhabitedType if fed None

👍 Makes sense and seems quite straightforward. I'm new to mypy development and I'd prefer moving in baby steps, I will do that in a follow-up PR.

@intgr
Copy link
Contributor Author

intgr commented Jul 27, 2023

I merged the two functions as suggested by hauntsaninja.

Due to the logic change, I opted to rename is_optional() -> is_overlapping_none(), because there is already precedent for similarly named functions (is_overlapping_types, is_overlapping_erased_types, is_enum_overlapping_union etc).

@github-actions

This comment has been minimized.

@github-actions
Copy link
Contributor

According to mypy_primer, this change doesn't affect type check results on a corpus of open source code. ✅

@hauntsaninja hauntsaninja merged commit 2b613e5 into python:master Aug 2, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Mypy performs wrong narrowing with == None and in (None,) conditions
4 participants