[
https://issues.apache.org/jira/browse/FLINK-26060?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17490743#comment-17490743
]
Francesco Guardiani commented on FLINK-26060:
---------------------------------------------
Hi [~dianfu]
So version upgrade is a feature that will come in 1.15, and the first upgrade
will be between 1.15 and 1.16 as detailed here
https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=191336489#FLIP190:SupportVersionUpgradesforTableAPI&SQLPrograms-Scoping.
So this feature was never supported up to this point.
The reason for removing the support from python exec nodes (which at the end
means removing the annotations we just added a couple of days ago with
https://issues.apache.org/jira/browse/FLINK-25387) is that once FLINK-25231 is
implemented and the Python UDF stack is reimplemented on top of the Table UDF
stack, you won't need these exec nodes anymore. So supporting them now doesn't
make sense, as we're trying to be conservative and we're not committing to
support exec nodes we know we're going to remove, like group windows.
If FLINK-25231 and the consequent reimplementation of the Python UDF stack is
done through the 1.16 development process, Python developers will be able to
upgrade between 1.16 and 1.17. Unless you perform an "intermediate" release
which features the two aforementioned features.
Does it makes sense?
> Make Python specific exec nodes unsupported
> --------------------------------------------
>
> Key: FLINK-26060
> URL: https://issues.apache.org/jira/browse/FLINK-26060
> Project: Flink
> Issue Type: Sub-task
> Components: Table SQL / Runtime
> Reporter: Francesco Guardiani
> Assignee: Francesco Guardiani
> Priority: Major
> Labels: pull-request-available
>
> These nodes are using the old type system, which is going to be removed soon.
> We should avoid supporting them in the persisted plan, as we cannot commit to
> support them. Once migrated to the new type system, PyFlink won't need these
> nodes anymore and will just rely on Table new function stack. For more
> details, also check https://issues.apache.org/jira/browse/FLINK-25231
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
