[
https://issues.apache.org/jira/browse/FLINK-20457?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17255929#comment-17255929
]
Dian Fu commented on FLINK-20457:
---------------------------------
[~louzhengyu] My feeling is that there is no need to unpickle them at Java
side. We could just store them as byte array:
- When reading from source, it is represented as DataStream<byte[]>
- When performing an operation such as map, it will deserialize the byte array
and perform the map operation
Besides, we should not use PythonTableUtils.getCollectionInputFormat in Python
DataStream API.
> Fix the handling of timestamp in DataStream.from_collection
> -----------------------------------------------------------
>
> Key: FLINK-20457
> URL: https://issues.apache.org/jira/browse/FLINK-20457
> Project: Flink
> Issue Type: Bug
> Components: API / Python
> Affects Versions: 1.12.0
> Reporter: Dian Fu
> Assignee: zhengyu.lou
> Priority: Major
> Fix For: 1.13.0, 1.12.1
>
>
> Currently, DataStream.from_collection firstly converts date/time/dateTime
> objects to int at Python side and then construct the corresponding
> Date/Time/Timestamp object at Java side. It will lose the timezone
> information. Pickle could handle date/time/datetime properly and the
> conversion could be avoided.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
