Error message for __all__ operations

[r-priyam]

Environment data

• Language Server version: v2021.12.1-pre.1
• OS and version: Windows 11 Home Single Language 22000.348
• Python version (& distribution if applicable, e.g. Anaconda): Python 3.9.9

Expected behaviour

Shouldn't error with the __all__ operations.

Code Snippet / Additional information

test.py

class AllBug:
    ...

__init__.py

from test import *

__all__ = (AllBug) # errors here

Error message - Operation on "all" is not supported, so exported symbol list may be incorrect

[r-priyam]

It appears that this was fixed previously in this release of pyright Point 4

[erictraut]

This is an incorrect usage of __all__. You need to assign a tuple or list of symbol names, not actual symbols. You can fix this bug by changing your code above to one of the following:

__all__ = ("AllBug",)

or

__all__ = ["AllBug"]

[r-priyam]

Sorry and thanks eric

[mangelozzi]

This is an incorrect usage of __all__. You need to assign a tuple or list of symbol names, not actual symbols. You can fix this bug by changing your code above to one of the following:

__all__ = ("AllBug",)

or

__all__ = ["AllBug"]

I get the same error even though it is a tuple (also tried a list same error):

[erictraut]

You need to include a list of strings in the tuple, as in __all__ = ("DevHw", ). Your code is including a list of symbol references, not strings.

[mangelozzi]

You need to include a list of strings in the tuple, as in __all__ = ("DevHw", ). Your code is including a list of symbol references, not strings.

Oh my goodness sorry about that, I kept checking and couldnt see the difference, thank you.

[xeroc]

Funny how this works:

import abc


class AbstractTradingStrategy(abc.ABC):
    pass


__all__ = ["AbstractTradingStrategy"]

but this fails:

import abc


class AbstractTradingStrategy(abc.ABC):
    pass


__all__ = [AbstractTradingStrategy.__name__]    ## WARNING HERE!

[erictraut]

@xeroc, that's correct. The first example is supported by type checkers and the second is not. Refer to this documentation for a full list of supported expression forms supported for __all__.

[xeroc]

@xeroc, that's correct. The first example is supported by type checkers and the second is not. Refer to this documentation for a full list of supported expression forms supported for __all__.

Thanks for the explanation and the link.
Do you have a proposal how I should do do this when using autoflake? It remives unused imports. So either i use .name, then the import is used, or i have to eun an assert on all imports. I prefer .name, as its cleaner code.

[erictraut]

Do you have a proposal how I should do do this when using autoflake

I presume autoflake is a linter and code formatter? If so, this sounds like a bug in the linter. It should treat a symbol as referenced if it appears in the __all__ list. Consider reporting it as a bug to the maintainer.

[xeroc]

Thanks for pointing me to the right direction. Turns out autoflake has a parameter for that already:
PyCQA/autoflake#35

Sorry for bothering you for nothing.