At Tue, 20 Dec 2016 22:42:48 -0500, Robert Haas <robertmhaas@gmail.com> wrote in
<CA+TgmoZT94brLAGK7gCmxB4mO=C-Cdz1N8KN8Xen4sexHozN=Q@mail.gmail.com>
> On Tue, Dec 20, 2016 at 6:18 AM, Fabien COELHO <coelho@cri.ensmp.fr> wrote:
> > Would this approach be acceptable, or is modifying Nodes a no-go area?
> >
> > If it is acceptable, I can probably put together a patch and submit it.
> >
> > If not, I suggest to update the documentation to tell that
> > pg_stat_statements does not work properly with combined queries.
>
> I think you've found a bug, but I'm a little doubtful about your
> proposed fix. However, I haven't studied the code, so I don't know
> what other approach might be better.
That will work and doable, but the more fundamental issue here
seems to be that post_parse_analyze_hook or other similar hooks
are called with a Query incosistent with query_debug_string. It
is very conveniently used but the name seems to me to be
suggesting that such usage is out of purpose. I'm not sure,
though.
A similar behavior is shown in error messages but this harms far
less than the case of pg_stat_statements.
ERROR: column "b" does not exist
LINE 1: ...elect * from t where a = 1; select * from t where b = 2; com...
^
1. Let all of the parse node have location in debug_query_string. (The original proposal)
2. Let the parser return once for each statement (like psql parser) and corresponding query string is stored in a
varialbleother than debug_query_string.
...
regards,
--
Kyotaro Horiguchi
NTT Open Source Software Center