Re: Declarative partitioning - another take

Robert Haas wrote:
> On Tue, Dec 20, 2016 at 4:51 AM, Alvaro Herrera
> <alvherre(at)2ndquadrant(dot)com> wrote:
> > Amit Langote wrote:
> >
> >> diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
> >> index 1c219b03dd..6a179596ce 100644
> >> --- a/src/backend/commands/tablecmds.c
> >> +++ b/src/backend/commands/tablecmds.c
> >> @@ -13297,8 +13297,10 @@ ATExecAttachPartition(List **wqueue, Relation rel, PartitionCmd *cmd)
> >> }
> >> }
> >>
> >> + /* It's safe to skip the validation scan after all */
> >> if (skip_validate)
> >> - elog(NOTICE, "skipping scan to validate partition constraint");
> >> + ereport(INFO,
> >> + (errmsg("skipping scan to validate partition constraint")));
> >
> > Why not just remove the message altogether?
>
> That's certainly an option. It might be noise in some situations. On
> the other hand, it affects whether attaching the partition is O(1) or
> O(n), so somebody might well want to know. Or maybe they might be
> more likely to want a message in the reverse situation, telling them
> that the partition constraint DOES need to be validated. I'm not sure
> what the best user interface is here; thoughts welcome.

Even if we decide to keep the message, I think it's not very good
wording anyhow; as a translator I disliked it on sight. Instead of
"skipping scan to validate" I would use "skipping validation scan",
except that it's not clear what it is we're validating. Mentioning
partition constraint in errcontext() doesn't like a great solution, but
I can't think of anything better.

(We have the table_rewrite event trigger, for a very similar use case.)

--
Álvaro Herrera https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services