Bryan Cutler created ARROW-10512:
------------------------------------
Summary: [Python] Arrow to Pandas conversion promotes child array
to float for NULL values
Key: ARROW-10512
URL: https://issues.apache.org/jira/browse/ARROW-10512
Project: Apache Arrow
Issue Type: Improvement
Components: Python
Reporter: Bryan Cutler
When converting a nested Arrow array to Pandas, if a child array is an integer
type with NULL values, it gets promoted to floating point and NULL values are
replaced with NaNs. Since the Pandas conversion for these types results in
Python objects, it is not necessary to use NaN and `None` values could be
inserted instead. This is the case for ListType, MapType and StructType, etc.
{code}
In [4]: s = pd.Series([[1, 2, 3], [4, 5, None]])
In [5]: arr = pa.Array.from_pandas(s)
In [6]: arr.type
Out[6]: ListType(list<item: int64>)
In [7]: arr.to_pandas()
Out[7]:
0 [1.0, 2.0, 3.0]
1 [4.0, 5.0, nan]
dtype: object {code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
