TST: parametrize test_info

[ivanovmg]

• closes #xxxx
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

Precursor for #37320

• Parametrize tests
• Rename some of test functions

[jbrockmendel]

"just_works" -> "smoke_test" maybe?

the existing test is a pattern i really dislike, where we have two unrelated fixtures for what should be separate tests (that you've separated, which i like). but it would be nice to find a way to parametrize over float_frame/datetime_frame. i think indirect might be related, but never fully got the hang of that

[ivanovmg]

I managed to do so using the approach described here: https://stackoverflow.com/a/64246323
For that, however, I needed to explicitly import float_frame. Without the import it would not work.

[ivanovmg]

Now pre-commit check complains about the non-standard imports because I explicitly import fixture functions.
Please suggest

[ivanovmg]

Fixed that as @jreback suggested #37887 (comment)

[jbrockmendel]

separating these i like, but im trying to push back against parametrization that creates DataFrame etc objects at test collection time. it increases the memory footprint which is breaking the windows builds.

[ivanovmg]

Right, I remember your suggestion to reduce footprint.

[ivanovmg]

Reverted back.

[jreback]

can you do result = io.getvalue() and assert somthing about the result (not the whole thing)

[jreback]

e.g. assert len(result.splitlines()) > 100 or something

[ivanovmg]

Done.

[ivanovmg]

Can anyone suggest on dealing with pre-commit check?
It complains that I used non-standard imports. In particular I imported fixtures from pandas/conftest.py.
Without this import the parametrization I used will not work (import error).
@jreback @jbrockmendel

[jbrockmendel]

I think we need to abandon the idea of using the indirect fixture parametrization

[ivanovmg]

I think we need to abandon the idea of using the indirect fixture parametrization

So I should go ahead and split test_info_smoke_test into four different test functions (for int_frame, float_frame, datatime_frame and duplicate_columns_frame)? Much like it was there before.

[jbrockmendel]

bottom line i dont think this is worth futzing over all that much. if you really want, you could try something like

@pytest.mark.parametrize("tmfunc", [tm.getSeriesData, tm.getTimeSeriesData, lambda: tm.getSeriesData()).astype("int64"),  duplicate_columns_frame]
def test_thing(tmfunc):
    frame = DataFrame(tmfunc())
    [...]

[ivanovmg]

It's greenish.

[jbrockmendel]

LGTM

[jreback]

thanks @ivanovmg