Thomas Zaksek <zaksek@ptt.uni-due.de> writes:
> Nested Loop Left Join (cost=0.00..32604.48 rows=3204 width=14) (actual
> time=11.991..2223.227 rows=2950 loops=1)
> -> Index Scan using
> messungen_v_dat_2007_11_12_messpunkt_minute_tag_idx on
> messungen_v_dat_2007_11_12 m (cost=0.00..5371.09 rows=3204 width=4)
> (actual time=0.152..12.385 rows=2950 loops=1)
> Index Cond: ((ganglinientyp = 'M'::bpchar) AND (992 = minute_tag))
> -> Index Scan using messwerte_mv_nr_idx on messwerte_mv w
> (cost=0.00..8.49 rows=1 width=18) (actual time=0.730..0.734 rows=1
> loops=2950)
> Index Cond: (w.nr = m.messpunkt)
> Total runtime: 2234.143 ms
> (6 rows)
> To me this plan looks very clean and nearly optimal,
For so many rows I'm surprised it's not using a bitmap indexscan.
What PG version is this? How big are these tables?
regards, tom lane