Re: improve transparency of bitmap-only heap scans - Mailing list pgsql-hackers

On Thu, Mar 19, 2020 at 9:26 PM Justin Pryzby <pryzby@telsasoft.com> wrote:
>
> On Mon, Mar 16, 2020 at 09:08:36AM -0400, James Coleman wrote:
> > Does the original optimization cover parallel bitmap heap scans like this?
>
> It works for parallel bitmap only scans.
>
> template1=# explain analyze select count(*) from exp where a between 25 and 35 and d between 5 and 10;
>  Finalize Aggregate  (cost=78391.68..78391.69 rows=1 width=8) (actual time=525.972..525.972 rows=1 loops=1)
>    ->  Gather  (cost=78391.47..78391.68 rows=2 width=8) (actual time=525.416..533.133 rows=3 loops=1)
>          Workers Planned: 2
>          Workers Launched: 2
>          ->  Partial Aggregate  (cost=77391.47..77391.48 rows=1 width=8) (actual time=518.406..518.406 rows=1
loops=3)
>                ->  Parallel Bitmap Heap Scan on exp  (cost=31825.37..77245.01 rows=58582 width=0) (actual
time=296.309..508.440rows=43887 loops=3)
 
>                      Recheck Cond: ((a >= 25) AND (a <= 35) AND (d >= 5) AND (d <= 10))
>                      Heap Blocks: unfetched=4701 exact=9650
>                      ->  BitmapAnd  (cost=31825.37..31825.37 rows=140597 width=0) (actual time=282.590..282.590
rows=0loops=1)
 
>                            ->  Bitmap Index Scan on index_exp_a  (cost=0.00..15616.99 rows=1166456 width=0) (actual
time=147.036..147.036rows=1099872 loops=1)
 
>                                  Index Cond: ((a >= 25) AND (a <= 35))
>                            ->  Bitmap Index Scan on index_exp_d  (cost=0.00..16137.82 rows=1205339 width=0) (actual
time=130.366..130.366rows=1200000 loops=1)
 
>                                  Index Cond: ((d >= 5) AND (d <= 10))
>
>
> > +++ b/src/backend/commands/explain.c
> > @@ -2777,6 +2777,8 @@ show_tidbitmap_info(BitmapHeapScanState *planstate, ExplainState *es)
> >  {
> >       if (es->format != EXPLAIN_FORMAT_TEXT)
> >       {
> > +             ExplainPropertyInteger("Unfetched Heap Blocks", NULL,
> > +                                                        planstate->unfetched_pages, es);
> >               ExplainPropertyInteger("Exact Heap Blocks", NULL,
> >                                                          planstate->exact_pages, es);
> >               ExplainPropertyInteger("Lossy Heap Blocks", NULL,
> > @@ -2784,10 +2786,14 @@ show_tidbitmap_info(BitmapHeapScanState *planstate, ExplainState *es)
> >       }
> >       else
> >       {
> > -             if (planstate->exact_pages > 0 || planstate->lossy_pages > 0)
> > +             if (planstate->exact_pages > 0 || planstate->lossy_pages > 0
> > +                     || planstate->unfetched_pages > 0)
> >               {
> >                       ExplainIndentText(es);
> >                       appendStringInfoString(es->str, "Heap Blocks:");
> > +                     if (planstate->unfetched_pages > 0)
> > +                             appendStringInfo(es->str, " unfetched=%ld",
> > +                                                              planstate->unfetched_pages);
> >                       if (planstate->exact_pages > 0)
> >                               appendStringInfo(es->str, " exact=%ld", planstate->exact_pages);
> >                       if (planstate->lossy_pages > 0)

Awesome, thanks for confirming with an actual plan.

> I don't think it matters in nontext mode, but at least in text mode, I think
> maybe the Unfetched blocks should be output after the exact and lossy blocks,
> in case someone is parsing it, and because bitmap-only is a relatively new
> feature.  Its output is probably less common than exact/lossy.

I tweaked that (and a comment that didn't reference the change); see attached.

James