int64() is converted to float for nullable field when None is provided + nan returned when None is expected

Apache Iceberg version

0.6.0 (latest release)

Please describe the bug 🐞

For reproduction using https://github.com/apache/iceberg-python/blob/main/tests/catalog/test_glue.py

Here is failing test

@mock_aws
def test_create_table_with_pyarrow_schema(
    _bucket_initialize: None,
    moto_endpoint_url: str,
    database_name: str,
    table_name: str,
) -> None:
    catalog_name = "glue"
    identifier = (database_name, table_name)
    test_catalog = GlueCatalog(catalog_name, **{"s3.endpoint": moto_endpoint_url})
    test_catalog.create_namespace(namespace=database_name)

    pa_schema = pa.schema([
        pa.field('year', pa.int64(), nullable=False),
        pa.field('n_legs', pa.int64(), nullable=True),
        pa.field('animals', pa.string(), nullable=True)
    ])

    table = test_catalog.create_table(
        identifier=identifier,
        schema=pa_schema,
        location=f"s3://{BUCKET_NAME}/{database_name}.db/{table_name}",
    )
    assert table.identifier == (catalog_name,) + identifier
    assert TABLE_METADATA_LOCATION_REGEX.match(table.metadata_location)
    assert test_catalog._parse_metadata_version(table.metadata_location) == 0

    table.append(
        pa.Table.from_pylist(
            [
                {"year": 2001, "n_legs": 2, "animals": None},
                {"year": 2002, "n_legs": None, "animals": "Horse"},
            ], schema=pa_schema
        )
    )

    assert len(table.scan().to_arrow()) == 2

    table.append(
        pa.Table.from_pylist(
            [
                {"year": 2003, "n_legs": 6, "animals": "Cicada"},
                {"year": 2004, "n_legs": 8, "animals": "Spider"},
            ], schema=pa_schema
        )
    )

    assert len(table.scan().to_arrow()) == 4

    assert table.scan().to_pandas().to_dict("records") == [
        {"animals": "Cicada", "n_legs": 6, "year": 2003},
        {"animals": "Spider", "n_legs": 8, "year": 2004},
        {"animals": None, "n_legs": 2, "year": 2001},
        {"animals": "Horse", "n_legs": None, "year": 2002},
    ]

Error part:

E         Full diff:
E           [
E         -  {'animals': 'Cicada', 'n_legs': 6, 'year': 2003},
E         +  {'animals': 'Cicada', 'n_legs': 6.0, 'year': 2003},
E         ?                                   ++
E         -  {'animals': 'Spider', 'n_legs': 8, 'year': 2004},
E         +  {'animals': 'Spider', 'n_legs': 8.0, 'year': 2004},
E         ?                                   ++
E         -  {'animals': None, 'n_legs': 2, 'year': 2001},
E         +  {'animals': None, 'n_legs': 2.0, 'year': 2001},
E         ?                               ++
E         -  {'animals': 'Horse', 'n_legs': None, 'year': 2002},
E         ?                                 -- ^
E         +  {'animals': 'Horse', 'n_legs': nan, 'year': 2002},
E         ?                                  ^^
E           ]

It looks like an issue of pyarrow. scan() returns correct data and type.

pyiceberg.io.pyarrow.project_table() does the conversion...

This issue has been automatically marked as stale because it has been open for 180 days with no activity. It will be closed in next 14 days if no further activity occurs. To permanently prevent this issue from being considered stale, add the label 'not-stale', but commenting on the issue is preferred when possible.

This issue has been closed because it has not received any activity in the last 14 days since being marked as 'stale'