[ please keep the list cc'd ] Volker Grabsch <vog@xxxxxxxxxxxxxxxxxx> writes: > The "count(*)" in the example seems to be distracting. In fact, > it could be replaced with a simple constant value, the effect > is the same: > CREATE VIEW a AS > SELECT > (... expensive calculation ...) as expensive, > 1 as cheap > FROM > x; Well, if you try that case, you'll find that the "expensive" column *does* get thrown away. (Using EXPLAIN VERBOSE under 8.4 may help you see what's going on here.) It's only when there's some reason why the view can't get flattened that there's an issue. I've been thinking about this since your earlier mail, and I think it would probably be possible to suppress unused columns in a non-flattened subquery. I remain unconvinced that it's worth the trouble though. A real (not handwavy) example would help make the case here. As an example of why I'm not convinced, one thing we'd have to consider is whether it is okay to suppress calculation of columns containing volatile functions. I'd be inclined to think not, since that could cause side-effects to not happen that the user might be expecting to happen. (We got beat up in the past for letting the planner be cavalier about that consideration.) I suspect, though, that that case and other non-optimizable cases might account for the bulk of situations where the existing optimization doesn't happen. regards, tom lane -- Sent via pgsql-performance mailing list (pgsql-performance@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
