Re: RFC: Timing Events

Josh Berkus <josh(at)agliodbs(dot)com> writes:
> It still eliminates the main potential use of auto-explain on production
> sites, though, which is to turn it on only for specific operations.
> I've never been able to make use of auto-explain for any real diagnostic
> purpose on a production site, and I don't personally know anyone who has.

This objection seems completely off the mark. If you have a specific
query you want to test, you can just use EXPLAIN. The point of
auto-explain is to get info about queries that are unexpectedly slow,
and for that it seems to me that you'll generally want it active over a
fairly wide scope.

I'm perfectly prepared to grant that you personally may never have had a
use for auto-explain, but it does not follow that it's useless. I'm
also pretty unclear on what use-case you think there is that's
intermediate between plain EXPLAIN and cases for which auto-explain is
reasonably suitable as-is.

> To get back to the original thread, though: I'm saying that "you can do
> that with auto-explain" is zero justification to limit any of Greg's
> work, because auto-explain isn't generally useful.

I don't recall having said anything against Greg's proposal, but in any
case I don't grant your premise.

regards, tom lane