Zeugswetter Andreas IZ5 <Andreas.Zeugswetter@telecom.at> writes:
> 5. Actually postgresql has behaved in this manner because of certain
> "bugs" in the optimizer. Recently a lot of those "bugs" have been
> identified and "fixed", thus destroying the defacto rule based
> behavior.
That's a real good point --- I think we've already heard a couple of
complaints about the new optimizer doing "silly" things that it didn't
use to do.
I repeat my proposal: CREATE TABLE should insert a default size (say
about 1000 tuples) into pg_class.reltuples, rather than inserting 0.
That way, the optimizer will only choose small-table-oriented plans
if the table has actually been verified to be small by vacuum.
regards, tom lane