chore: pandas-nightly and duckdb-nightly fixes

[MarcoGorelli]

What type of PR is this? (check all applicable)

• 💾 Refactor
• ✨ Feature
• 🐛 Bug Fix
• 🔧 Optimization
• 📝 Documentation
• ✅ Test
• 🐳 Other

Related issues

• Related issue #<issue number>
• Closes #<issue number>

Checklist

• Code follows style guide (ruff)
• Tests added
• Documented the changes

If you have comments or can explain your changes, please do so below

[MarcoGorelli]

Alright only 5 test failures left

FAILED tests/expr_and_series/is_nan_test.py::test_nan_series[pandas[nullable]] - AssertionError: Mismatch at index 0, key float_na: <NA> != True
Expected: {'int': [False, False, None], 'float': [False, False, None], 'float_na': [True, False, None]}
Got: {'int': [False, False, <NA>], 'float': [False, False, <NA>], 'float_na': [<NA>, False, <NA>]}
FAILED tests/expr_and_series/is_nan_test.py::test_nan_series[pandas[pyarrow]] - AssertionError: Mismatch at index 0, key float_na: <NA> != True
Expected: {'int': [False, False, None], 'float': [False, False, None], 'float_na': [True, False, None]}
Got: {'int': [False, False, <NA>], 'float': [False, False, <NA>], 'float_na': [<NA>, False, <NA>]}
FAILED tests/expr_and_series/str/zfill_test.py::test_str_zfill[pandas[pyarrow]]
FAILED tests/expr_and_series/str/zfill_test.py::test_str_zfill_series[pandas[pyarrow]]
FAILED tests/expr_and_series/reduction_test.py::test_empty_scalar_reduction_with_columns[duckdb]
==== 5 failed, 10891 passed, 231 skipped, 599 xfailed in 115.30s (0:01:55) =====

[FBruzzesi]

Should we use

Suggested change

	dtypes.Int128: duckdb_dtypes.DuckDBPyType("INT128"),
	dtypes.Int128: duckdb_dtypes.HUGEINT,

(and dtypes.UInt128: duckdb_dtypes.UHUGEINT, below)?

[MarcoGorelli]

😭

[FBruzzesi]

Thanks @MarcoGorelli - I am pretty onboarded with the duckdb changes.

Regarding pandas I can see we are adding yet another different behavior from polars and pandas pre v3. Is it really not feasible to remap one of the two behavior? e.g. pandas v2 to pandas v3 or pandas v3 to polars? Apologies if the answer to this is obvious, but I didn't even try it myself. And if the answer is negative, should we document in the null vs nan section the pandas v3 behavior?

[MarcoGorelli]

thanks for your review!

there's a whole discussion going on at pandas-dev/pandas#62040 😩 not sure where things will end up, but i think we should wait until the next release before documenting the mode.nan_is_na behaviour (in particular, in case it doesn't land)