Vladimir Sitnikov

@avinashyn , could you please clarify why do you think `getCatalogs` should return a single entry only?

> expectation is that getCatalogs() should return only the one passed in JDBC URL If you want something "passed to JDBC URL", then you should consider `String java.sql.Connection#getCatalog()` Have you...

At the same time, in JMeter we could probably drop support for IE

@ns7381, please rebase and add a test that reproduces the issue

It looks like we should "round" with saturation. That is avoid wrapping to year 10000 What do you think?

LocalDateTime has nanosecond precision, and PostgreSQL has microsecond precision only. That is why we round 2019-02-10 03:04:05.123456789 to 2019-02-10 03:04:05.123457 However LocalDateTime.MAX should be truncated from ... 23:59:59.999999999 to ......

We should probably consider the expected behaviour when the user supplies invalid schema name. Should `setSchema` fail immediately or is it allowed to defer "set search_path" till the next prepare/execute...

Let's skip creating transactions for `setSchema` calls

Do you think it would be a user-noticeable change? I can't imagine a case when users would be able to tell if we execute `setSchema` within a transaction or not.

Ah, we do not commit after `setSchema`. That would indeed be noticeable as they might check "if there's a transaction or not".