Moises Lopez <moylop...@vauxoo.com> writes: > -> Seq Scan on public.link l1 (cost=0.00..10.40 rows=40 width=516) > (actual time=0.010..0.011 rows=1 loops=1) > * Output: l1.id <http://l1.id>, l1.url, l1.name > <http://l1.name>, l1.description, l1.rel*
This is normal; it is not a bug, and it is not a source of performance issues either. The planner is choosing to do that to avoid a projection step in this plan node, because there's no need for one. On the other scan, where it *is* important to project out just the required columns to minimize the size of the hash table above the scan, it does do so: > -> Seq Scan on public.link l2 (cost=0.00..10.40 rows=40 > width=516) (actual time=0.010..0.011 rows=1 loops=1) > Output: l2.url > Is there a manner to avoid returning all columns in order to get a better > performance? You have not shown us anything about what your actual performance issue is, but this isn't it. https://wiki.postgresql.org/wiki/Slow_Query_Questions regards, tom lane
