JingsongLi commented on issue #9796: [FLINK-14253][table-planner-blink] Add hash distribution and sort grouping only when dynamic partition insert URL: https://github.com/apache/flink/pull/9796#issuecomment-541021987 > Before this can actually merged, I would raise another issue i have just observed, which is we might should revisit the interface of `PartitionableTableSink:getPartitionFieldNames`. I think such information should be held by framework. CatalogTable should definitely gets this, maybe also relational node `LogicalSink`, but not `TableSink` and `PartitionableTableSink`. `PartitionableTableSink` should only be fed with such information but not get them out of it. What do you think? Thanks for your review. I couldn't agree more. The point is the partition support for temporary table. The temporary table uses ConnectorDescriptor. The ConnectorDescriptor should have the partition fields information, and then the planner gets the corresponding CatalogTable through the CatalogManager. But now: 1. CatalogManager now has no temporary table-related interface, which is bypassed by default Catalog and default database. 2. ConnectorCatalogTable does not transmit partition-related information, which should also be modified. But we can remove partitioning support for temporary tables and wait until FLIP-64 is complete. For non-temporary tables, I can go the "perfect" and "final" way to improve. I'll create PRs ASAP.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
