Fabien COELHO <coelho@cri.ensmp.fr> writes:
> I'm not sure it is worth bothering with setting dbname which seems not to
> be used afterwards anyway. I would suggest to coldly call PQdb on the
> connection for the error message and thus simplify the code, as attached.
I looked at this a bit and have a few comments:
* I think it might be best to restrict the hint to appear only in the
case that the SQLSTATE is ERRCODE_UNDEFINED_TABLE. There are a lot of
possible failures here, and I'm not sure this hint is apropos for any
but that one.
* I really don't care for "HINT pgbench: " as that's got pretty much
nothing to do with any other message formatting we have anywhere,
least of all the existing hint-like messages in pgbench. I've started
a different thread at
http://www.postgresql.org/message-id/22280.1436139973@sss.pgh.pa.us
to discuss what the general message formatting in pgbench ought to be,
and probably this needs to wait for the outcome of that discussion.
* I'd be a bit inclined to reduce the message to something like
Perhaps you need to do initialization ("pgbench -i") in database "foo"
so that we can avoid the question of formatting of multi-sentence hints.
regards, tom lane