[
https://issues.apache.org/jira/browse/ARROW-370?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15647890#comment-15647890
]
Wes McKinney commented on ARROW-370:
------------------------------------
I think we can do optimistic type inference on dtype=object columns like we are
doing for strings and booleans -- see {{ArrowSerializer<NPY_OBJECT>::Convert}}
-- we can add a case for datetime.date objects
> Python: Pandas conversion from `datetime.date` columns
> ------------------------------------------------------
>
> Key: ARROW-370
> URL: https://issues.apache.org/jira/browse/ARROW-370
> Project: Apache Arrow
> Issue Type: New Feature
> Components: Python
> Reporter: Uwe L. Korn
>
> It seems to be a common practice to store some columns as Python
> {{datetime.date}} to avoid issues with far future/past dates in Pandas. We
> can natively store this data in Arrow as well in Parquet but there is no
> conversion available yet. A simple way could be to provide a path
> pandas.Series -> numpy.ndarray(datetime64[D]) -> arrow.Array(DATE))
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
