table.delete()/overwrite() with null values in table and with non-null filter will delete null rows

[jqin61]

Apache Iceberg version

0.7.0

Please describe the bug 🐞

Hi I added this test which breaks:

def test_delete_overwrite_with_null(session_catalog: RestCatalog) -> None:
    arrow_schema = pa.schema([pa.field("ints", pa.int32())])
    arrow_tbl = pa.Table.from_pylist(
        [
            {"ints": 1},
            {"ints": 2},
            {"ints": None}
        ],
        schema=arrow_schema,
    )

    iceberg_schema = Schema(NestedField(1, "ints", IntegerType()))

    tbl_identifier = "default.test_delete_overwrite_with_null"

    try:
        session_catalog.drop_table(tbl_identifier)
    except NoSuchTableError:
        pass

    tbl = session_catalog.create_table(tbl_identifier, iceberg_schema)
    tbl.append(arrow_tbl)

    assert [snapshot.summary.operation for snapshot in tbl.snapshots()] == [Operation.APPEND]

    arrow_tbl_overwrite = pa.Table.from_pylist(
        [
            {"ints": 3},
            {"ints": 4},
        ],
        schema=arrow_schema,
    )
    tbl.overwrite(arrow_tbl_overwrite, "ints == 2")  # Should rewrite one file

    assert [snapshot.summary.operation for snapshot in tbl.snapshots()] == [
        Operation.APPEND,
        Operation.OVERWRITE,
        Operation.APPEND,
    ]

    assert tbl.scan().to_arrow()["ints"].to_pylist() == [3, 4, 1, None]

FAILED tests/integration/test_deletes.py::test_delete_overwrite_with_null - assert [3, 4, 1] == [3, 4, 1, None]

The bug is from this line: https://github.com/apache/iceberg-python/blob/main/pyiceberg/table/__init__.py#L596 which does not build a predicate including the null because the revert operator of pyarrow.compute underneath only handles null correctly when the raw operator mentions the field is with pc.is_null or pc.is_valid

Working on a fix now.

[sungwy]

Hi @jqin61 - this looks like a critical issue that should be fixed for the 0.7.0.

Thank you very much for flagging this issue and starting to work on the fix!!