On 2021-Oct-19, Bossart, Nathan wrote: > I did consider this, but I figured it might be better to keep the lock > level consistent for a given object type no matter what the statement > type is. I don't have a strong opinion about this, though.
Yeah, the problem is that if there is a concurrent process waiting on your lock, we'll release ours and they'll grab theirs, so we'll be waiting on them afterwards, which is worse. BTW I noticed that the case of partitioned indexes was wrong too. I fixed that, added it to the tests, and pushed. -- Álvaro Herrera PostgreSQL Developer — https://www.EnterpriseDB.com/ "People get annoyed when you try to debug them." (Larry Wall)
