Get streaming hash aggregate back

[adam8157]

trafficstars

In multi-phase aggregate, this feature streams entries of the bottom phase when out of memory instead of spilling to disk, it avoids the disk I/O and roughly deduplicates to save the network I/O.

The planner always supports streaming aggregate and there are lots of cases, but the executor part was lost while merging PostgreSQL 12, this commit gets it back.

[adam8157]

The planner always supports streaming aggregate and there are lots of cases, but the executor just did nothing special.

[adam8157]

create table dqa_t1 (d int, i int, c char, dt date);
insert into dqa_t1 select i%23, i%12, (i%10) || '', '2009-06-10'::date + ( (i%34) || ' days')::interval
from generate_series(0, 99999999) i;

SET statement_mem='1000kB';
explain analyze select count(distinct i) from dqa_t1 group by c;

gpadmin=# explain select count(distinct i) from dqa_t1 group by c;
                                                     QUERY PLAN
--------------------------------------------------------------------------------------------------------------------
 Gather Motion 3:1  (slice1; segments: 3)  (cost=0.00..9710.67 rows=10 width=8)
   ->  HashAggregate  (cost=0.00..9710.67 rows=4 width=8)
         Group Key: c
         ->  Redistribute Motion 3:3  (slice2; segments: 3)  (cost=0.00..9710.67 rows=23 width=6)
               Hash Key: c
               ->  GroupAggregate  (cost=0.00..9710.67 rows=23 width=6)
                     Group Key: c, i
                     ->  Sort  (cost=0.00..9710.67 rows=23 width=6)
                           Sort Key: c, i
                           ->  Redistribute Motion 3:3  (slice3; segments: 3)  (cost=0.00..9710.66 rows=23 width=6)
                                 Hash Key: c, i
                                 ->  Streaming HashAggregate  (cost=0.00..9710.66 rows=23 width=6)
                                       Group Key: c, i
                                       ->  Seq Scan on dqa_t1  (cost=0.00..1237.68 rows=33333680 width=6)
 Optimizer: Pivotal Optimizer (GPORCA)
(15 rows)

The above case generates a streaming plan, but has no big duration difference between streaming execution and not.