TST/CLN: remove TestData from frame-tests; replace with fixtures

[h-vetinari]

Following review in #22236:

ok, pls open a new issue that refs this, to remove use of TestData in favor of fixtures

Started the process in that PR by creating a conftest.py that translates all the current attributes of TestData to fixtures, with the following "translation guide":

• frame -> float_frame
• frame2 -> float_frame2
• intframe -> int_frame
• tsframe -> datetime_frame
• mixed_frame -> float_string_frame
• mixed_float -> mixed_float_frame
• mixed_float2 -> mixed_float_frame2
• mixed_int -> mixed_int_frame
• all_mixed -> mixed_type_frame
• tzframe -> timezone_frame
• empty -> empty_frame
• ts1 -> datetime_series
• ts2 -> datetime_series_short
• simple -> simple_frame

Need to incrementally replace their usages in pandas/tests/frame/ (example below).

• Create conftest.py and translate TestData-attributes into fixtures (TST/CLN: break up & parametrize tests for df.set_index #22236)
• test_alter_axes.py (TST/CLN: break up & parametrize tests for df.set_index #22236)
• test_analytics.py (TST/CLN: Fixturize frame/test_analytics #22733)
• test_api.py (Fixturize tests/frame/test_api and tests/sparse/frame/test_frame #22738)
• test_apply.py (Fixturize tests/frame/test_apply #22735)
• test_arithmetic.py (Fixturize tests/frame/test_arithmetic #22736)
• test_asof.py (Fixturize tests/frame/test_asof.py #25628)
• test_axis_select_reindex.py (Fixturize tests/frame/test_axis_select_reindex.py #25627)
• test_block_internals.py (TST/CLN: Fixturize tests/frame/test_block_internals.py #22926)
• test_combine_concat.py (Fixturize tests/frame/test_combine_concat.py #25634)
• test_constructors.py (Fixturize tests/frame/test_constructors.py #25635)
• test_convert_to.py
• test_dtypes.py (Fixturize tests/frame/test_dtypes.py #25636)
• test_duplicates.py
• test_indexing.py (Fixturize tests/frame/test_indexing.py #25633)
• test_join.py (Fixturize tests/frame/test_join.py #25639)
• test_missing.py (TST: Fixturize tests/frame/test_missing.py #25640)
• test_mutate_columns.py (Fixturize tests/frame/test_mutate_columns.py #25642)
• test_nonunique_indexes.py
• test_operators.py (Fixturize tests/frame/test_operators.py #25641)
• test_period.py
• test_quantile.py
• test_query_eval.py
• test_rank.py
• test_replace.py
• test_repr_info.py
• test_reshape.py
• test_sort_values_level_as_str.py
• test_sorting.py
• test_subclass.py
• test_timeseries.py
• test_timezones.py
• test_to_csv.py
• test_validate.py

Things for follow-ups:

• Remove other class-based test-methods
• Turn tests from class- to function-based

An example from #22236 - before:

def test_set_columns(self):
    cols = Index(np.arange(len(self.mixed_frame.columns)))
    self.mixed_frame.columns = cols
    with tm.assert_raises_regex(ValueError, 'Length mismatch'):
        self.mixed_frame.columns = cols[::2]

After:

def test_set_columns(self, float_string_frame):
    cols = Index(np.arange(len(float_string_frame.columns)))
    float_string_frame.columns = cols
    with tm.assert_raises_regex(ValueError, 'Length mismatch'):
        float_string_frame.columns = cols[::2]

Basically, it comes down to replacing all the occurrences of self.<name> with translation_guide[<name>] (and specifying<name> as a parameter to the function).

PS. Note that some fixtures added by #22236 have now been removed by #24885. Please check #24885 which code was removed, in case you should need it for the fixturisation. Alternatively, you can ping me, @jbrockmendel or @jreback.

[h-vetinari]

@gfyoung
I think this issue and #22550 should have a "good first issue" and potentially "difficulty: novice" tag. What do you think?

[gfyoung]

@h-vetinari : Seems reasonable to me. Done.

[akulagrawal]

Hi! I am a beginner to open source. Can I take up this issue?

[gfyoung]

@akulagrawal : Go for it!

[SaturnFromTitan]

Seems like the following files don't contain TestData (anymore) and can be checked or removed from the issue description:

• test_constructors.py
• test_dtypes.py
• test_period.py
• test_quantile.py
• test_rank.py
• test_sort_values_level_as_str.py
• test_timezones.py
• test_validate.py

I'll take a look at the others