ne 30. 9. 2018 v 18:49 odesílatel Arup Rakshit <ar@xxxxxxx> napsal:
I just added it as you said, but I am getting same plan.Sort (cost=62842.16..62846.91 rows=1897 width=35) (actual time=1845.831..1845.950 rows=1229 loops=1)Sort Key: projects.idSort Method: quicksort Memory: 145kB-> HashAggregate (cost=62710.42..62738.88 rows=1897 width=35) (actual time=1844.178..1845.060 rows=1229 loops=1)Group Key: projects.id-> Hash Right Join (cost=159.68..45382.09 rows=364807 width=35) (actual time=1.534..618.717 rows=365784 loops=1)Hash Cond: (workitems.project_id = projects.id)Filter: (workitems.deleted_at IS NULL)Rows Removed by Filter: 257457-> Seq Scan on workitems (cost=0.00..36653.75 rows=623175 width=43) (actual time=0.047..213.842 rows=623175 loops=1)-> Hash (cost=135.97..135.97 rows=1897 width=16) (actual time=1.478..1.478 rows=1897 loops=1)Buckets: 2048 Batches: 1 Memory Usage: 105kB-> Seq Scan on projects (cost=0.00..135.97 rows=1897 width=16) (actual time=0.006..0.914 rows=1897 loops=1)Planning time: 0.498 msExecution time: 1846.100 ms
Then there is not too much what can be done better - maybe you can try PostgreSQL 11 with paralel hash join -- it is process about 6M rows, the time about 2 sec is good
——————Indexes:"workitems_pkey" PRIMARY KEY, btree (id)"index_workitems_on_company_id" btree (company_id)"index_workitems_on_deleted_at" btree (deleted_at)"index_workitems_on_parent_workitem_id" btree (parent_workitem_id)"index_workitems_on_project_id" btree (project_id)"index_workitems_on_standard_workitem_id" btree (standard_workitem_id)"index_workitems_on_workitem_category_id" btree (workitem_category_id)"patrial_index_workitems_200_1" btree (project_id) WHERE deleted_at IS NULL
On 30-Sep-2018, at 10:15 PM, Pavel Stehule <pavel.stehule@xxxxxxxxx> wrote:CREATE INDEX ON workitems(project_id) WHERE deleted_at is null
