Обсуждение: log_min_cost_statement

Hello!

Would it be possible to have an option "log_min_cost_statement", similar to
"log_min_duration_statement" except that it logs all statement that has
an minimum cost estimated by the planner? I think it would be easier to
find unoptimized queries that way.

Just a thought!

/Nichlas

Nichlas L�fdahl wrote:
> Hello!
>
> Would it be possible to have an option "log_min_cost_statement", similar to
> "log_min_duration_statement" except that it logs all statement that has
> an minimum cost estimated by the planner? I think it would be easier to
> find unoptimized queries that way.

Uh, yea, I guess we could, but the costs are more comparison with other
costs, not really with other queries, I think, so I am not sure it would
be meaningful.

--
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

Nichlas =?iso-8859-1?Q?L=F6fdahl?= <crotalus@acc.umu.se> writes:
> Would it be possible to have an option "log_min_cost_statement", similar to
> "log_min_duration_statement" except that it logs all statement that has
> an minimum cost estimated by the planner? I think it would be easier to
> find unoptimized queries that way.

Why would you want to filter on the planner's cost estimate rather than
the true runtime?

            regards, tom lane

Nichlas Löfdahl wrote:
> Would it be possible to have an option "log_min_cost_statement",
> similar to "log_min_duration_statement" except that it logs all
> statement that has an minimum cost estimated by the planner? I think
> it would be easier to find unoptimized queries that way.

If you believe that the costs have any absolute meaning at all then they
should be proportional to the execution time, so the existing setting
should be enough.