follow up from #237 & #239

[mutricyl]

This is a follow up from pandas nightly build CI job introduction. The following tests have been skipped to get green light workflow for #239. They need to be fixed.

Looks like also that unique function was inducing errors in tests but this was unexpectedly solved. It might be worth to investigate this a little bit.

skipped tests:

• test_pandas_extensiontests.py::TestPintArray::test_arith_frame_with_scalar : tests with "radd", "rsub", "rmul", "rtruediv", "rpow" are not working with pandas 3
• test_issues.py::TestIssue80::test_div : pandas 3 is so fast that _timeit returns 0
• test_pandas_extensiontests.py : all test including complex are skipped due to NumpyExtensionArray issue in pandas take function BUG: pd.api.extensions.take does not accept NumpyExtensionArray pandas-dev/pandas#59177

all temporary skip are marked by GH #239

Originally posted by @mutricyl in #239 (comment)

Looks like there's a unique function issue similar to the take issue

pint_pandas/testsuite/test_pandas_extensiontests.py::TestPintArray::test_value_counts_with_normalize[float] - TypeError: unique requires a Series, Index, ExtensionArray, or np.ndarray, got NumpyExtensionArray. FAILED

Originally posted by @andrewgsavage in #239 (comment)

[mutricyl]

pandas-dev/pandas#59177 was resolved so complex may be allowed again for pandas >= 3

[mutricyl]

The two pandas issues with take and unique being solved, I add a look at the r operator issue in test_pandas_extensiontests.py::TestPintArray::test_arith_frame_with_scalar

I have identified the change in behavior is pandas 3 that induce pint and pint-pandas to fail with right operator. pint.compat.is_upcast_type behavior changed between pandas version due to type(df) that used to return <class 'pandas.core.frame.DataFrame'> which is now returning <class 'pandas.DataFrame'>.

@andrewgsavage do you think we should tackle this on the pandas side or on pint side ?

[andrewgsavage]

nice find, just need to add pandas.DataFrame to
https://github.com/hgrecco/pint/blob/b2fc74a7d6038d206d158804daa32e1698f89c26/pint/compat.py#L242

[mutricyl]

for the record the pandas' change is a features : pandas-dev/pandas#55171

[mutricyl]

shall we make CI-pandas-pre use also pint pre version or update the tests to xfail while pint is not strictly above 0.24.1 ?

[andrewgsavage]

I'll release a new pint version after pandas.DataFrame and pandas.Series are added to upcast types