[
https://issues.apache.org/jira/browse/FLINK-31193?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
dalongliu updated FLINK-31193:
------------------------------
Description:
Sink Flink 1.17, we have implemented the native hive aggregation function,
users can enable this optimization by the option
`table.exec.hive.native-agg-function.enabled` per-job. The option works well in
Table API &
HiveServer2Endpoint, but in SQL client, this option can't work per-job, it
works at the module level.
On the SqlClient side, if we want to use the native hive aggregation function,
we need to enable the option first and then load HiveModule, only by doing that
we can load the hive native aggregation function. This behavior is inconsistent
with the definition of the option itself, and also is inconsistent with the
behavior of the Table API and HiveServerEndpoint2, which is a bug. we cannot
align the behavior of the option at the moment due to code implementation
reasons. If we want to align the behavior, we need to modify the `Module`
interface, which is a Public interface. In 1.17 we didn't have time to
complete this work, so we need to fix this in 1.18 and make the parameter
per-job effective on the SqlClient side as well.
was:Sink Flink 1.17, we have implemented the native hive aggregation
function, users can enable this optimization by
> The option `table.exec.hive.native-agg-function.enabled` should work at job
> level when using it in SqlClient side
> -----------------------------------------------------------------------------------------------------------------
>
> Key: FLINK-31193
> URL: https://issues.apache.org/jira/browse/FLINK-31193
> Project: Flink
> Issue Type: Sub-task
> Components: Connectors / Hive
> Affects Versions: 1.17.0
> Reporter: dalongliu
> Priority: Major
> Fix For: 1.18.0
>
>
> Sink Flink 1.17, we have implemented the native hive aggregation function,
> users can enable this optimization by the option
> `table.exec.hive.native-agg-function.enabled` per-job. The option works well
> in Table API &
> HiveServer2Endpoint, but in SQL client, this option can't work per-job, it
> works at the module level.
> On the SqlClient side, if we want to use the native hive aggregation
> function, we need to enable the option first and then load HiveModule, only
> by doing that we can load the hive native aggregation function. This behavior
> is inconsistent with the definition of the option itself, and also is
> inconsistent with the behavior of the Table API and HiveServerEndpoint2,
> which is a bug. we cannot align the behavior of the option at the moment due
> to code implementation reasons. If we want to align the behavior, we need to
> modify the `Module` interface, which is a Public interface. In 1.17 we
> didn't have time to complete this work, so we need to fix this in 1.18 and
> make the parameter per-job effective on the SqlClient side as well.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
