Kevin Grittner wrote:
> I'm not really complaining, since PostgreSQL is significantly faster on
> these updates than another product (for which the license agreement
> prohibits posting benchmarks without their approval); but, it does seem
> that there is room to optimize this if anyone is so inclinded. (I may
> look at it myself some day, but there are other issues that matter much
> more to me, so not soon.)
>
> (1) I would think the creation of the exception could be eliminated.
> There is no exceptional situation here.
Yeah. Patch? :)
(the only reason it gets logged is that the implementation of
java.sql.SQLException logs on construction -- i.e. we can't avoid
logging every exception we create..)
> (2) It seems like it might be possible to avoid some round trips here,
> although I haven't looked at the issues in detail.
What specifically? IIRC we're piggybacking on the metadata code which
already has lots of server-version-specific logic in it .. we don't
really want to duplicate that logic.
-O