[Python] Round-trip type-conversion bug list object in pd.DataFrame

[MMCMA]

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

I am trying to convert a pandas DataFrame into a table and back. However the type property is lost, the list is returned as an numpy.ndarray

import pandas as pd
import pyarrow as pa

df = pd.DataFrame(dict(a=[[1,2,3],]))
type(df['a'].iloc[0])
# Out[1]: list

round_trip_df = pa.Table.from_pandas(df).to_pandas()
type(round_trip_df['a'].iloc[0])
# Out[2]: numpy.ndarray

Here is the output of pd.show_versions();

INSTALLED VERSIONS

commit : 2e218d10984e9919f0296931d92ea851c6a6faf5
python : 3.10.9.final.0
python-bits : 64
OS : Windows
OS-release : 10
Version : 10.0.19045
machine : AMD64
processor : Intel64 Family 6 Model 142 Stepping 12, GenuineIntel
byteorder : little
LC_ALL : None
LANG : None
LOCALE : English_United States.1252
pandas : 1.5.3
numpy : 1.23.5
pytz : 2022.7.1
dateutil : 2.8.2
setuptools : 64.0.2
pip : 23.0
Cython : 0.29.33
pytest : 7.2.1
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : 4.9.2
html5lib : 1.1
pymysql : None
psycopg2 : 2.9.3
jinja2 : 3.1.2
IPython : 8.10.0
pandas_datareader: 0.10.0
bs4 : 4.11.2
bottleneck : None
brotli :
fastparquet : None
fsspec : 2023.1.0
gcsfs : None
matplotlib : 3.7.0
numba : 0.56.4
numexpr : None
odfpy : None
openpyxl : 3.1.1
pandas_gbq : None
pyarrow : 9.0.0
pyreadstat : None
pyxlsb : None
s3fs : None
scipy : 1.8.1
snappy : None
sqlalchemy : 1.4.46
tables : None
tabulate : 0.9.0
xarray : None
xlrd : None
xlwt : None
zstandard : None
tzdata : None

Component(s)

Python

[AlenkaF]

If I am reading the C++ code correctly, this is the current design in the conversion of the lists

arrow/python/pyarrow/src/arrow/python/arrow_to_pandas.cc

Lines 735 to 770 in 69118b2

	Status ConvertListsLike(PandasOptions options, const ChunkedArray& data,
	PyObject** out_values) {
	// Get column of underlying value arrays
	ArrayVector value_arrays;
	for (int c = 0; c < data.num_chunks(); c++) {
	const auto& arr = checked_cast<const ListArrayT&>(*data.chunk(c));
	// values() does not account for offsets, so we need to slice into it.
	// We can't use Flatten(), because it removes the values behind a null list
	// value, and that makes the offsets into original list values and our
	// flattened_values array different.
	std::shared_ptr<Array> flattened_values = arr.values()->Slice(
	arr.value_offset(0), arr.value_offset(arr.length()) - arr.value_offset(0));
	if (arr.value_type()->id() == Type::EXTENSION) {
	const auto& arr_ext = checked_cast<const ExtensionArray&>(*flattened_values);
	value_arrays.emplace_back(arr_ext.storage());
	} else {
	value_arrays.emplace_back(flattened_values);
	}
	}
	using ListArrayType = typename ListArrayT::TypeClass;
	const auto& list_type = checked_cast<const ListArrayType&>(*data.type());
	auto value_type = list_type.value_type();
	if (value_type->id() == Type::EXTENSION) {
	value_type = checked_cast<const ExtensionType&>(*value_type).storage_type();
	}
	auto flat_column = std::make_shared<ChunkedArray>(value_arrays, value_type);
	options = MakeInnerOptions(std::move(options));
	OwnedRefNoGIL owned_numpy_array;
	RETURN_NOT_OK(ConvertChunkedArrayToPandas(options, flat_column, nullptr,
	owned_numpy_array.ref()));
	PyObject* numpy_array = owned_numpy_array.obj();
	DCHECK(PyArray_Check(numpy_array));

We do, however, store the pandas_dtype in schema metadata:

>>> table = pa.Table.from_pandas(df)
>>> table.schema.metadata
{b'pandas': b'{"index_columns": [{"kind": "range", "name": null, "start": 0, "stop": 1, "step": 1}], "column_indexes": [{"name": null, "field_name": null, "pandas_type": "unicode", "numpy_type": "object", "metadata": {"encoding": "UTF-8"}}], "columns": [{"name": "a", "field_name": "a", "pandas_type": "list[int64]", ...