CLN: fix mypy errors in pandas\tests\indexes\interval\test_base.py #28926

[srcole]

• xref Fix mypy errors for pandas\tests\* #28926
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

Hi! I based this change on what I saw in #28947. However, when I test this with mypy pandas/tests/indexes/interval/test_base.py, I get this error, which I didn't expect:

pandas/tests/indexes/interval/test_base.py:16: error: Incompatible types in assignment (expression has type "Optional[Type[IntervalIndex]]", base class "Base" defined the type as "None")

I thought the "Optional" that was added would accept the "None." But, additionally, in my local environment, I also can't see any of the errors listed in issue #28926, so maybe I don't have something set up right.

[pep8speaks]

Hello @srcole! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2019-10-28 15:07:53 UTC

[simonjayhawkins]

Thanks @srcole for the PR.

[simonjayhawkins]

also can't see any of the errors listed in issue #28926, so maybe I don't have something set up right.

to see the errors, you would need to comment out the ignore_errors in setup.cfg

[simonjayhawkins]

@srcole generally looking good. however ci failing https://dev.azure.com/pandas-dev/pandas/_build/results?buildId=18915

can you sort imports. see https://dev.pandas.io/docs/development/contributing.html#import-formatting

[srcole]

I tested it locally, and it seems this change also fixes:

pandas\tests\indexes\test_numeric.py
pandas\tests\indexes\test_category.py

So I removed them from setup.cfg

[jbrockmendel]

@srcole can you rebase

[srcole]

Thanks for catching that @simonjayhawkins. Yup, I guess I missed that when addressing the merge conflict. Anyway, I removed that line.

[simonjayhawkins]

@srcole lgtm @WillAyd

[WillAyd]

Does this need to be Optional?

[simonjayhawkins]

IMO we should keep the Optional otherwise when we use 3.6 type annotations we'll get
pandas\tests\indexes\common.py:34: error: Incompatible types in assignment (expression has type "None", variable has type "Type[Index]")

The Optional should only be excluded where we have added something = None # type: ... for the purpose of type checking.

In this case _holder was set to None before the type annotation was added.

so when we move to 3.6, we will change to
_holder: Optional[Type[Index]] = None
whereas if the initialisation was purely added for the type annotations we would change to just
_holder: Type[Index]

[WillAyd]

Isn't the latter what we actually want here though? OK to change to 36 annotation now on this PR since 35 CI should be dropped within the next few days

[simonjayhawkins]

Isn't the latter what we actually want here though?

IMO out of scope here

but to answer your question, in general, with 3.6, we could reduce the number of assert <some_variable> is not None that are sometimes needed when class attributes are initialised to None, but we would need to evaluate on a case by case basis.

OK to change to 36 annotation now on this PR since 35 CI should be dropped within the next few days

i'm not sure a good reason to block since this is not guaranteed.

[WillAyd]

Thanks @srcole !