TST: Regression testing for fixed issues

[mroeschke]

• closes BUG: Applying function on column of Groupby object with as_index=False does not select column #5764
• closes Unable to retrieve same-index dataframe with the last item per group #7883
• closes Index([1,2,.np.nan]).get_indexer([np.nan]) returns wrong value? #7820
• closes BUG/ERR: adding column with float label to Int64Index gives ValueError #8720
• closes smaller than microsecond timedelta64 Series are not saved correctly with to_csv #6783
• closes truncation issue with pd.read_csv #7072
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff

[simonjayhawkins]

Suggested change

	expected = np.array([2])
	expected = np.array([2], dtype=np.intp)

[simonjayhawkins]

Suggested change

	expected = ",0\n0,0 days 00:00:00.000000001\n1,0 days 00:00:00.000000001\n"
	expected_rows = [
	",0",
	"0,0 days 00:00:00.000000001",
	"1,0 days 00:00:00.000000001",
	]
	expected = tm.convert_rows_list_to_csv_str(expected_rows)

[jreback]

lgtm. merge on green.

[TomAugspurger]

Fixed one of the 32-bit failures. We may have a remaining one that I don't immediately see a fix for.

    def test_to_csv_timedelta_precision(self):
        # GH 6783
        s = pd.Series([1, 1]).astype("timedelta64[ns]")
        buf = io.StringIO()
        s.to_csv(buf)
        result = buf.getvalue()
        expected = ",0\n0,0 days 00:00:00.000000001\n1,0 days 00:00:00.000000001\n"
>       assert result == expected
E       AssertionError: assert ',0\r\n0,0 da...000000001\r\n' == ',0\n0,0 days...0.000000001\n'
E         - ,0

E         ?   -
E         + ,0
E         - 0,0 days 00:00:00.000000001

E         ?                            -
E         + 0,0 days 00:00:00.000000001
E         - 1,0 days 00:00:00.000000001
...
E         
E         ...Full output truncated (3 lines hidden), use '-vv' to show

[jbrockmendel]

at least one of the failures looks like a np.intp thing

[TomAugspurger]

Strange, on the 32-bit linux build we have

self = <pandas.tests.indexes.test_numeric.TestInt64Index object at 0xd35cbbcc>

    def test_get_indexer_nan(self):
        # GH 7820
        result = Index([1, 2, np.nan]).get_indexer([np.nan])
        expected = np.array([2], dtype="int64")
>       tm.assert_numpy_array_equal(result, expected)
E       AssertionError: numpy array are different
E       
E       Attribute "dtype" are different
E       [left]:  int32
E       [right]: int64

Does that make sense?

Still not sure about the to_csv one. We might want to drop that test / issue from the list and merge the rest.

[simonjayhawkins]

@TomAugspurger the original 2 Windows failures are fixed with the suggestions above (tested locally)

[TomAugspurger]

Sorry, I missed that! Will push that up.

[TomAugspurger]

Thanks!