Re: POC, WIP: OR-clause support for indexes

On Fri, Feb 16, 2024 at 1:32 PM Andrei Lepikhov
<a.lepikhov@postgrespro.ru> wrote:
>
> On 16/2/2024 07:00, jian he wrote:
> > On Wed, Feb 14, 2024 at 11:21 AM Andrei Lepikhov
> > <a.lepikhov@postgrespro.ru> wrote:
> > My OS: Ubuntu 22.04.3 LTS
> > I already set the max_parallel_workers_per_gather to 10.
> > So for all cases, it should use parallelism first?
> >
> > a better question would be:
> > how to make the number of OR less than 29 still faster when
> > enable_or_transformation is ON by only set parameters?
> In my test environment this example gives some subtle supremacy to ORs
> over ANY with only 3 ors and less.
> Please, provide next EXPLAIN ANALYZE results for the case you want to
> discuss here:
> 1. with enable_or_transformation enabled
> 2. with enable_or_transformation disabled
> 3. with enable_or_transformation disabled but with manual transformation
> OR -> ANY done, to check the overhead of this optimization.
>

you previously mentioned playing with parallel_tuple_cost and
parallel_setup_cost.
(https://www.postgresql.org/message-id/e3338e82-a28d-4631-9eec-b9c0984b32d5%40postgrespro.ru)

So I did by
`
SET parallel_setup_cost = 0;
SET parallel_tuple_cost = 0;
`

After setting these parameters, overall enable_or_transformation ON is
performance better.
sorry for the noise.
so now I didn't find any corner case where enable_or_transformation is
ON peforms worse than when it's OFF.

+typedef struct OrClauseGroupEntry
+{
+ OrClauseGroupKey key;
+
+ Node   *node;
+ List   *consts;
+ Oid scalar_type;
+ List   *exprs;
+} OrClauseGroupEntry;

I found that the field `scalar_type` was never used.