Tom Lane wrote:
> Jan Wieck <JanWieck@Yahoo.com> writes:
> > Tom Lane wrote:
> >> Good point. What do people think of making the pg_stat_activity view do
> >> COALESCE(current_query, '<unknown>')
> >> so that there's a clear flag that you're not seeing anything?
>
> > Shouldn't be done on the view. The builtin function used to extract that
> > information from the stats file can do it much better.
>
> No, I think the view is exactly the right place to do it. If we do it
> in the function then we are forcing a UI-prettiness issue onto
> applications that may not want it; but they'll have no way to bypass it.
>
> > So it could return <permission denied> or similar in that case to
> > distinguish.
>
> I don't think it's all that important to distinguish the reason why
> you're not seeing it. Returning NULL from the function is a perfectly
> good convention at the level of the function.
I think you do need to distinguish "permission denied" from "not
enabled". I think our "not enabled" status for this feature is
confusing enough for you have to report is separately.
You could return "permission denied or not enabled", I guess. In fact,
I would put the text in parens, then apps can know that strings in
quotes are errors, but of course, if they issue "(select * from
pg_class)" there are problems. Maybe "-- permission denied".
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026