Search code examples
sqlpostgresqloptimizationindexingsql-execution-plan

Why does creating an unrelated index make my query faster?

I have a table:

debts (
    name       text,
    inv_no     integer,
    inv_type   text,
    status     text,
);

I have a following select:

SELECT COUNT(*) FROM debts WHERE name = '...' AND inv_no = 100 AND inv_type = '...';

To optimize something else, I added:

CREATE INDEX ON debt (status);

Nowhere in SELECT do I refer to the status, yet running...

EXPLAIN SELECT COUNT(*)... (as above)

...before and after creating index gives me cost 16.65..16.66 changing into 1.25..1.26. Why?

Full explain (analyze, verbose) before/after:

Before:

QUERY PLAN
----------
 Aggregate  (cost=16.65..16.66 rows=1 width=0) (actual time=0.126..0.128 rows=1 loops=1)
   Output: count(*)
   ->  Seq Scan on ab123456.debts  (cost=0.00..16.65 rows=1 width=0) (actual time=0.106..0.106 rows=0 loops=1)
         Output: name, inv_no, inv_type, status
         Filter: ((debts.name = '...'::text) AND (debts.inv_type = '...'::text) AND (debts.inv_no = 100))
 Total runtime: 0.387 ms

After:

QUERY PLAN
----------
 Aggregate  (cost=1.25..1.26 rows=1 width=0) (actual time=0.031..0.033 rows=1 loops=1)
   Output: count(*)
   ->  Seq Scan on ab123456.debts  (cost=0.00..1.25 rows=1 width=0) (actual time=0.024..0.024 rows=0 loops=1)
         Output: name, inv_no, inv_type, status
         Filter: ((debts.name = '...'::text) AND (debts.inv_type = '...'::text) AND (debts.inv_no = 100))
 Total runtime: 0.118 ms
Solution

  • Some utility statements (including CREATE INDEX!) update table statistics while being at it. The manual:

    For efficiency reasons, reltuples and relpages are not updated on-the-fly, and so they usually contain somewhat out-of-date values. They are updated by VACUUM, ANALYZE, and a few DDL commands such as CREATE INDEX.

    Bold emphasis mine. So even if your index seems completely unrelated, the updated table statistics can have an impact - especially on count() which primarily depends on the two statistics mentioned.

    Related: