Andrew - Supernews <andrew+nonews@supernews.com> writes:
> It turns out that the scenario above is trivial to hit in 8.0 using
> referential constraints; RI triggers cache their plans, and on 8.0 the RI
> query is planned as a seqscan if the tables are freshly created. (On 7.4
> the plan is an index scan, thanks to the default 1000 rows / 10 pages stats.)
Hm. One thing we could do is to throw in some default values when we
see the table has exactly zero pages --- perhaps ye olde traditional
1000/10, or possibly something else, but anyway not exactly 0/0.
The reason I thought we didn't need to do this sort of hack anymore
is that pg_dump loads the tables first and then creates the RI
constraints. What exactly is the common case where the wrong thing
happens?
regards, tom lane