"Phil Endecott" <spam_from_postgresql_general@chezphil.org> writes:
> If I understand it correctly, it is still doing a sequential scan on
> part_tsearch that does not terminate early due to the limit clause. So
> I'm still seeing run times that are rather worse than I think should be
> possible. Can it not step through the indexes in the way that it does
> for a Merge Join until it has got enough results to satisfy the limit,
> and then terminate?
Nope, there is not that much intelligence about NOT IN.
You could possibly manually rewrite the thing as a LEFT JOIN
with a WHERE inner-join-key IS NULL clause. This would probably
lose if most of the outer relation's rows join to many inner rows,
though.
regards, tom lane