[
https://issues.apache.org/jira/browse/ARROW-842?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16833703#comment-16833703
]
Joris Van den Bossche commented on ARROW-842:
---------------------------------------------
{{pd.NaT}} is one that is not recognized, but, only in lists / object dtype:
{code}
In [25]: pa.array([pd.Timestamp("2012-01-01"), pd.NaT, None, np.nan,
float('nan')])
Out[25]:
<pyarrow.lib.TimestampArray object at 0x7f27703b82c8>
[
1325376000000000,
-62135596800000000,
null,
null,
null
]
{code}
If you actually pass a pandas Series object with the proper dtype such as
datetime64[ns], then those missing values are handled correctly:
{code}
In [32]: pa.array(pd.Series([pd.Timestamp("2012-01-01"), pd.NaT]))
Out[32]:
<pyarrow.lib.TimestampArray object at 0x7f2752662b38>
[
1325376000000000000,
null
]
{code}
> [Python] Handle more kinds of null sentinel objects from pandas 0.x
> -------------------------------------------------------------------
>
> Key: ARROW-842
> URL: https://issues.apache.org/jira/browse/ARROW-842
> Project: Apache Arrow
> Issue Type: Improvement
> Components: Python
> Reporter: Wes McKinney
> Priority: Major
>
> Follow-on work to ARROW-707. See
> https://github.com/pandas-dev/pandas/blob/master/pandas/_libs/lib.pyx#L193
> and discussion in https://github.com/apache/arrow/pull/554
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
