Re: Clarification of certain SQLSTATE class

On 01/25/2013 07:32 AM, gabrielle wrote:
> I'm working with some DB2 users, converting them to Pg, and I'm a bit
> confused about a certain class of SQLSTATE codes, specifically 02xxx
> "No data"
> (http://www.postgresql.org/docs/9.2/static/errcodes-appendix.html)
>
> As an example: I enable %e in log_line_prefix so I can see the
> SQLSTATE values. I run an UPDATE against a non-existent row, and find
> that my SQLSTATE value is 00000, indicating success. I understand
> that this could be considered a 'success' because the query didn't
> actually throw an error; but, based on the spec, I expected to see a
> SQLSTATE of 02000.
>
> I thought that Pg always generated an appropriate SQLSTATE code, and
> that it was just up to $client code to pick up that value, but that
> doesn't seem to be the case. Is this part of the SQL standard we
> don't implement?
If I understand it correctly, this could be emulated in client drivers -
PgJDBC, libpq, etc.

Row counts are known to be somewhat broken with the current approach to
PostgreSQL table partitioning anyway; you'll always get zero rows
affected when you UPDATE a partitioned table, and there's currently
(AFIAK) no way for a trigger/rule to aggregate the rowcounts from the
commands it executes and return the total.

--
Craig Ringer http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services