如何为条件日期查询构建高效的 PostgreSQL 索引？

Question

I'm trying to query a large database of ~10 million rows, to find all rows with a timestamp above a certain date and with a boolean flag set to false, like:

SELECT COUNT(id) FROM mytable WHERE (NOT fresh OR created IS NULL OR created <= '2021-1-1');

This query was taking an hour to run, but isn't too complex, so I tried creating an index for it like:

CREATE INDEX CONCURRENTLY mytable_fresh_created ON mytable (fresh, created ASC) WHERE fresh = false;

but this has made virtually no improvement on performance. And running EXPLAIN shows a sequential scan:

 Finalize Aggregate  (cost=18641642.51..18641642.52 rows=1 width=8)
   ->  Gather  (cost=18641642.29..18641642.50 rows=2 width=8)
         Workers Planned: 2
         ->  Partial Aggregate  (cost=18640642.29..18640642.30 rows=1 width=8)
               ->  Parallel Seq Scan on mytable  (cost=0.00..18629381.98 rows=4504126 width=4)
                     Filter: ((NOT fresh) OR (created IS NULL) OR (created <= '2021-1-1'::timestamp with time zone))

so it doesn't even look like the index is even being used.

How do I construct a better index for this type of query?

Answer 1

When you are using OR then indexes not using. I recommended using UNION ALL . Example:

SELECT *
FROM mytable 
WHERE (created <= '2021-01-01'::date)
union all 
SELECT *
FROM mytable 
WHERE (created IS NULL)