Phillip Cloud created ARROW-17018: ------------------------------------- Summary: Timedelta dtype metadata base unit is globally mutated by the Table.to_pandas() method Key: ARROW-17018 URL: https://issues.apache.org/jira/browse/ARROW-17018 Project: Apache Arrow Issue Type: Bug Components: C++, Python Affects Versions: 8.0.0 Reporter: Phillip Cloud Fix For: 9.0.0
Here's a failing test case to demonstrate the issue: {code} def test_mutates_dtype(): import pyarrow as pa expected = np.dtype('<m8') assert np.dtype(np.timedelta64) == expected df = pd.DataFrame({"a": [np.timedelta64()]}) t = pa.Table.from_pandas(df) t.to_pandas() assert np.dtype(np.timedelta64) == expected {code} -- This message was sent by Atlassian Jira (v8.20.10#820010)