On Wed, Nov 1, 2017 at 12:25 PM, Justin Pryzby <pry...@telsasoft.com> wrote:
> On Wed, Nov 01, 2017 at 12:19:21PM -0700, David G. Johnston wrote: > > On Wed, Nov 1, 2017 at 11:59 AM, Scott Marlowe <scott.marl...@gmail.com> > > wrote: > > > > > So some of my output from an explain analyze here has a line that says > > > this: > > > > > > ex Scan using warranty_order_item_warranty_order_id_idx on > > > warranty_order_item woi_1 (cost=0.57..277.53 rows=6 width=137) (actual > > > time=0.110..0.111 rows=0 loops=1,010,844) > > > > > > > Not my strong suit but, I'm pretty sure that reads: "The index was > queried > > 1M+ times and none of those inqueries resulted in a record being found". > > IIUC I'd be wondering why some form of hash join wasn't used... > > Except that: > > https://www.postgresql.org/docs/current/static/using-explain.html > "... the loops value reports the total number of executions of the node, > and > the actual time and ROWS VALUES SHOWN ARE AVERAGES PER-EXECUTION." > I seem to recall a somewhat recent commit that dealt with this. The problem is that with 1M loops a small number of rows returned will be indistinguishable from zero when computed as an average within finite precision. Seeing entire plans, and not just a single line of one, tends to help too. David J.
