ArrowIndexError when converting PyArrow Table with nested null values to Pandas DataFrame

[emanueledomingo]

Describe the bug, including details regarding any error messages, version, and platform.

The code below produces an ArrowIndexError when attempting to convert a PyArrow Table with nested null values into a Pandas DataFram. Everything works fine if i use to_pydict instead.

import pyarrow as pa
import pandas as pd
import json

json_data = json.loads("""
[
    {
        "key1": [
            {
                "key2": null
            },
            {
                "key2": null
            }
        ]
    }
]
""")

pa_table = pa.Table.from_pylist(json_data)
my_dict = pa_table.to_pydict()  # This works as expected: {'key1': [[{'key2': None}, {'key2': None}]]}
df = pa_table.to_pandas(types_mapper=pd.ArrowDtype).to_dict()

I'm using pyarrow 17 (pandas 2.2.3)

Component(s)

Python