Hi Peter, +1 the basic idea. But I have a worry that translations for around 20 statements will be invalidated because "%s.%s" will be replaced to "%s", which needs additional work for translators (even the outcome is not changed). Not sure the conversion wins the cost, or is it acceptable for HEAD?
Anyway, I grepped source codes under the logical/ and some codes under the commands/. I found single RelationGetRelationName()/get_rel_name() call are retained in some places, but seems OK. IIUC, below four types are still retained. * replication protocol message. We have been sending them separately, and no reasons to combine. * table/sequence mactching between publication and subscription. We are now passing to queries them. * ACL checks. It's not schema-qualified even from other places, so it's OK not to do for now. * Subscription relkind checks. The function accepts the namespace and relation name separately, because the remote name can be specified. It's more natural to keep current style. > Open question: for a temporary table, should the schema be displayed > as "pg_temp" or "pg_temp_N"? The attached keeps "pg_temp_N", which is > the current status quo. Anyway, after 0001 this is a one-line change > in MsgQualifiedRelNameByRel()/...ByRelId() if the consensus is > otherwise. IIUC, for the current region, the pg_temp schema can be reported only backend processes, right? If so I prefer without suffic, pg_temp. pg_temp_xxx is usable if other processes report the temp schema, maybe like autovacuum worker. Best regards, Hayato Kuroda FUJITSU LIMITED
