使用 `COUNT` 优化 SQL 查询

Question

I have the SQL query (I am using MariaDB/MySQL) below which works fine. But I am wondering if I can simplify/optimize it. The difficulty I am facing is because of COUNT with INNER JOIN . Do I really need the subquery IN or you think there is a way I can eliminate it?

SELECT a.* 
FROM aaa a 
INNER JOIN bbb b ON a.field1 = b.field1 
WHERE a.yyy = 'yyy' AND b.zzz = 'zzz' 
AND a.field1 IN (
    SELECT field1 
    FROM bbb 
    WHERE xxx = 'x' 
    GROUP BY field1 
    HAVING COUNT(field1) <= 5
)

Thank you!

Answer 1

Your query may be as good as possible, but you can also try joining with the subquery instead of using IN .

SELECT DISTINCT a.*
FROM aaa AS a
INNER JOIN bbb b ON a.field1 = b.field1 
INNER JOIN (
    SELECT field1 
    FROM bbb 
    WHERE xxx = 'x' 
    GROUP BY field1 
    HAVING COUNT(*) <= 5
) AS bc ON a.field1 = bc.field1
WHERE a.yyy = 'yyy' AND b.zzz = 'zzz' 

Answer 2

I like Barmar's answer, but because of the filtering on b , I suggest a small(?) change. (I rearrange the tables primarily to follow the Optimizer's likely order.)

SELECT DISTINCT a.*
FROM  (
    SELECT field1 
    FROM bbb 
    WHERE xxx = 'x' 
    GROUP BY field1 
    HAVING COUNT(*) <= 5
) AS bc
INNER JOIN bbb AS b ON b.field1 = bc.field1 
INNER JOIN aaa AS a ON a.field1 = bc.field1
WHERE b.zzz = 'zzz' 
  AND a.yyy = 'yyy'

Plus these indexes:

    bbb:  INDEX(xxx, field1)  -- this order
    b:    INDEX(field1, zzz)  -- either order
    a:    INDEX(field1, yyy)  -- either order

The first two composite indexes will be "covering".

The main change from Barmar's answer was to leave a.* for last.

If you want to discuss this Question further, it may be useful to provide SHOW CREATE TABLE and EXPLAIN .