On Thu, Mar 18, 2021 at 07:29:56AM +0800, Julien Rouhaud wrote:
> On Wed, Mar 17, 2021 at 06:32:16PM -0400, Bruce Momjian wrote:
> > OK, is that what everyone wants? I think that is what the patch already
> > does.
>
> Note exactly. Right now a custom queryid can be computed even if
> compute_queryid is off, if some extension does that in post_parse_analyze_hook.
>
> I'm assuming that what Robert was thinking was more like:
>
> if (compute_queryid)
> {
> if (queryid_hook)
> queryId = queryid_hook(...);
> else
> queryId = JumbeQuery(...);
> }
> else
> queryId = 0;
>
> And that should be done *after* post_parse_analyse_hook so that it's clear that
> this hook is no longer the place to compute queryid.
>
> Is that what should be done?
No, I don't think so. I think having extensions change behavior
controlled by GUCs is a bad interface.
The docs are going to say that you have to enable compute_queryid to see
the query id in pg_stat_activity and log_line_prefix, but if you install
an extension, the query id will be visible even if you don't have
compute_queryid enabled. I think you need to only honor the hook if
compute_queryid is enabled, and update the pg_stat_statements docs to
say you have to enable compute_queryid for pg_stat_statements to work.
Also, should it be compute_queryid or compute_query_id?
Also, the overhead of computing the query id was reported as 2% --- that
seems quite high for what it does. Do we know why it is so high?
--
Bruce Momjian <bruce@momjian.us> https://momjian.us
EDB https://enterprisedb.com
If only the physical world exists, free will is an illusion.