> On 30 Dec 2015, at 17:44, Tom Lane <[email protected]> wrote:
>
> Oleksii Kliukin <[email protected]> writes:
>>> On 30 Dec 2015, at 17:02, Tom Lane <[email protected]> wrote:
>>> Another idea would be to use the heap's row density as calculated
>>> by the last ANALYZE (ie, reltuples/relpages), with a fallback to 100
>>> if relpages=0. This'd only be convenient if the bitmap scan node has
>>> the parent heap rel open, which it might not.
>
>> +1
>
> Any objections to the attached?
Looks good to me. On my sample system with 100K rows, the new version gives me:
— CREATE TABLE test AS SELECT id FROM generate_series(1,100000) id;
— CREATE INDEX ON test USING brin(id);
postgres=# explain analyze select 1 from test where id = 500;
QUERY PLAN
-------------------------------------------------------------------------------------------------------------------------
Bitmap Heap Scan on test (cost=12.01..16.02 rows=1 width=0) (actual
time=0.199..4.220 rows=1 loops=1)
Recheck Cond: (id = 500)
Rows Removed by Index Recheck: 28927
Heap Blocks: lossy=128
-> Bitmap Index Scan on test_id_idx (cost=0.00..12.01 rows=1 width=0)
(actual time=0.072..0.072 rows=28800 loops=1)
Index Cond: (id = 500)
Planning time: 0.433 ms
Execution time: 4.323 ms
(8 rows)
which is much closer to the actual number of rows removed by the index recheck
+ the one left.
--
Oleksii
