[
https://issues.apache.org/jira/browse/CALCITE-1681?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15904683#comment-15904683
]
Remus Rusanu commented on CALCITE-1681:
---------------------------------------
I removed the blocking HIVE-15708 since we can work around this issue by
disabling {{MATERIALIZATIONS_ENABLED}} on the {{CalciteConnectionConfig}} and
thus achieve the same behavior as we had with 1.10. Per-se the copy assert on
{{VolcanoPlanner.registerImpl}} was not coming from 1.10 vs. 1.12, but from a
different change in the HIVE-15708 changes that actually enabled
{{CalciteConnectionConfig}} and was now using the defaults, instead of having a
{{null}} config in Hive master. When I applied that change I did not
internalize the consequence when investigating the problem here.
I believe that having a way to copy plans between clusters is still a worthy
goal, but is no longer a priority for Hive.
> Provide a way to copy RelNode trees between clusters
> ----------------------------------------------------
>
> Key: CALCITE-1681
> URL: https://issues.apache.org/jira/browse/CALCITE-1681
> Project: Calcite
> Issue Type: Improvement
> Reporter: Remus Rusanu
> Assignee: Remus Rusanu
>
> In Hive we cache materialized view plans and reuse the cached plans in
> queries. For this we need a way to do deep clone copy of a RelNode tree from
> one RelOptCluster to another.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
