Parametric tests fail in ways that isn't visible in CI

[itamarst]

Parametric tests fail in ways that isn't visible in CI

[ritchie46]

Is main broken now? I am surprised the parameteric tests don't run. Though for CI, I must defer to @stinodego.

[itamarst]

I edited the text, it may be the issue is that my computer is faster so it runs more examples through Hypothesis. But there are failing tests on main for me. E.g.:

$ pytest tests/unit/interchange/test_roundtrip.py -m hypothesis 
==================================================================================== test session starts =====================================================================================platform linux -- Python 3.10.12, pytest-7.4.4, pluggy-1.3.0
rootdir: /home/itamarst/devel/polars/py-polars
configfile: pyproject.toml
plugins: cov-4.1.0, anyio-4.2.0, hypothesis-6.88.1, xdist-3.3.1
collected 14 items / 4 deselected / 10 selected                                                                                                                                              

tests/unit/interchange/test_roundtrip.py ......FF..                                                                                                                                    [100%]

========================================================================================== FAILURES ==========================================================================================___________________________________________________________________________ test_from_dataframe_pandas_parametric ____________________________________________________________________________tests/unit/interchange/test_roundtrip.py:135: in test_from_dataframe_pandas_parametric
    dataframes(
tests/unit/interchange/test_roundtrip.py:150: in test_from_dataframe_pandas_parametric
    result = pl.from_dataframe(df_pd)
polars/convert.py:771: in from_dataframe
    return from_dataframe(df, allow_copy=allow_copy)
polars/interchange/from_dataframe.py:46: in from_dataframe
    return _from_dataframe(
polars/interchange/from_dataframe.py:55: in _from_dataframe
    polars_chunk = _protocol_df_chunk_to_polars(chunk, allow_copy=allow_copy)
polars/interchange/from_dataframe.py:71: in _protocol_df_chunk_to_polars
    dtype = dtype_to_polars_dtype(column.dtype)
properties.pyx:36: in pandas._libs.properties.CachedProperty.__get__
    ???
../.venv/lib/python3.10/site-packages/pandas/core/interchange/column.py:128: in dtype
    return self._dtype_from_pandasdtype(dtype)
../.venv/lib/python3.10/site-packages/pandas/core/interchange/column.py:141: in _dtype_from_pandasdtype
    raise ValueError(f"Data type {dtype} not supported by interchange protocol")
E   ValueError: Data type large_string[pyarrow] not supported by interchange protocol
E   Falsifying example: test_from_dataframe_pandas_parametric(
E       df=shape: (0, 1)
E       ┌──────┐
E       │ col0 │
E       │ ---  │
E       │ str  │
E       ╞══════╡
E       └──────┘,
E   )
E   Explanation:
E       These lines were always and only run by failing examples:
E           /home/itamarst/devel/polars/.venv/lib/python3.10/site-packages/pandas/core/dtypes/dtypes.py:2152
E           /home/itamarst/devel/polars/.venv/lib/python3.10/site-packages/pandas/core/interchange/column.py:141
______________________________________________________________________ test_from_dataframe_pandas_zero_copy_parametric _______________________________________________________________________polars/testing/asserts/frame.py:114: in assert_frame_equal
    _assert_series_values_equal(
polars/testing/asserts/series.py:180: in _assert_series_values_equal
    raise_assertion_error(
E   AssertionError: Series are different (exact value mismatch)
E   [left]:  [datetime.datetime(1974, 2, 4, 21, 23, 8, 840384)]
E   [right]: [datetime.datetime(2000, 1, 1, 0, 0)]

The above exception was the direct cause of the following exception:
tests/unit/interchange/test_roundtrip.py:155: in test_from_dataframe_pandas_zero_copy_parametric
    dataframes(
tests/unit/interchange/test_roundtrip.py:177: in test_from_dataframe_pandas_zero_copy_parametric
    assert_frame_equal(result, df)
polars/testing/asserts/frame.py:123: in assert_frame_equal
    raise_assertion_error(
E   AssertionError: DataFrames are different (value mismatch for column 'col0')
E   [left]:  [datetime.datetime(1974, 2, 4, 21, 23, 8, 840384)]
E   [right]: [datetime.datetime(2000, 1, 1, 0, 0)]
E   Falsifying example: test_from_dataframe_pandas_zero_copy_parametric(
E       df=shape: (1, 1)
E       ┌─────────────────────┐
E       │ col0                │
E       │ ---                 │
E       │ datetime[μs]        │
E       ╞═════════════════════╡
E       │ 2000-01-01 00:00:00 │
E       └─────────────────────┘,
E   )
E   Explanation:
E       These lines were always and only run by failing examples:
E           /home/itamarst/devel/polars/.venv/lib/python3.10/site-packages/pandas/core/arrays/arrow/array.py:1224
E           /home/itamarst/devel/polars/.venv/lib/python3.10/site-packages/pandas/core/arrays/arrow/array.py:1238
E           /home/itamarst/devel/polars/.venv/lib/python3.10/site-packages/pandas/core/arrays/arrow/array.py:1272
E           /home/itamarst/devel/polars/.venv/lib/python3.10/site-packages/pandas/core/arrays/datetimelike.py:354
E           /home/itamarst/devel/polars/.venv/lib/python3.10/site-packages/pandas/core/construction.py:616
E           (and 54 more with settings.verbosity >= verbose)
================================================================================== short test summary info ===================================================================================FAILED tests/unit/interchange/test_roundtrip.py::test_from_dataframe_pandas_parametric - ValueError: Data type large_string[pyarrow] not supported by interchange protocol
FAILED tests/unit/interchange/test_roundtrip.py::test_from_dataframe_pandas_zero_copy_parametric - AssertionError: DataFrames are different (value mismatch for column 'col0')
========================================================================= 2 failed, 8 passed, 4 deselected in 21.59s =========================================================================

[MarcoGorelli]

I think you just need to update your pandas version

[itamarst]

Earlier today I was on 2.1.4. I just upgraded to 2.2.0, and now I'm getting just failing tests to DeprecationWarning on test_roundtrip.py, rather than what I was getting before. So, maybe. Any sense why DeprecationWarnings would be errors locally but not in CI?

[itamarst]

Anyway can't reproduce with Pandas 2.2.0 so closing, sorry for the noise.