lirui-apache commented on issue #8205: [FLINK-12238] [TABLE/SQL] Support database related operations in GenericHiveMetastoreCatalog and setup flink-connector-hive module URL: https://github.com/apache/flink/pull/8205#issuecomment-485129366 > > I wonder whether we should differentiate Flink DB/table from Hive's own DB/table, especially if we allow Flink and Hive users to share an HMS instance? > > Do you mean why we have, for example, GenericCatalogDatabase v.s. the upcoming HiveCatalogDatabase? There are quite some differences between Flink and Hive w.r.t db and table. E.g. for db, hive always requires a valid, meaningful location uri, but flink doesn't really care about it; for column types of table, right now several types cannot be mapped 1:1 between flink and hive, and we need to wait for the rework of Flink type system finishes. > > I doubt, at least for now, that we will unify them, but we can re-evaluate later on. I think what we currently should aim for is to share code as much as possible between GenericHiveMetastoreCatalog and the upcoming HiveCatalog when implementing catalog APIs. I meant if user is using generic catalog, he/she's likely to expect to see only Flink DB/table when they call listDB/listTable. But if this generic catalog is based on a shared HMS instance, the call will return Hive DB/table as well. So I wonder whether it makes sense to do some filtering here. This is not an issue for Hive catalog, because if users use Hive catalog, they'll expect to access Hive metadata.
---------------------------------------------------------------- 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
