MySQL 查询运行速度非常慢

Question

I have a MySQL (v5.7.26) query that runs forever. Here is the query:

SELECT
    ur.user_id      AS user_id,
    sum(r.duration) AS total_time,
    count(user_id)  AS number_of_workouts
FROM user_resource ur
INNER JOIN resource r ON r.id = ur.resource_id
WHERE 
    ur.status = 1 
    AND NOT ur.action_date IS NULL 
    AND ur.user_id IN (
        SELECT user_id
        FROM user_resource ur2
        WHERE ur2.action_date >= now() - INTERVAL 2 DAY
    )
    AND r.type = 'WORKOUT'
    GROUP BY ur.user_id;

I have played a bit with it, by trying to understand where is the problem. For the testing purposes, I tried breaking in two. So:

SELECT user_id
FROM user_resource ur2
WHERE ur2.action_date >= now() - INTERVAL 2 DAY;

That returns (very quickly) list of user user_id's. When I plug the returned result in to the first part of the query, like this:

SELECT
    ur.user_id      AS user_id,
    sum(r.duration) AS total_time,
    count(user_id)  AS number_of_workouts
FROM user_resource ur
INNER JOIN resource r ON r.id = ur.resource_id
WHERE 
    ur.status = 1 
    AND NOT ur.action_date IS NULL 
    AND ur.user_id IN (1,1,1,4,4,5,6,7,7,7);
      AND r.type = 'WORKOUT'
GROUP BY ur.user_id

It runs very fast. My assumption is the IN (Subquery) is the bottleneck.

I was thinking to extract the subquery and get the user_ids, and then used it as a variable, but I am not sure is it the good approach, and additionally I am having issues with it. this is my attempt:

-- first statement
SET @v1 = (SELECT user_id
FROM user_resource ur2
WHERE ur2.action_date >= now() - INTERVAL 2 DAY)

-- second statement
SELECT
    ur.user_id      AS user_id,
    sum(r.duration) AS total_time,
    count(user_id)  AS prefixes
FROM user_resource ur
INNER JOIN resource r ON r.id = ur.resource_id
WHERE 
    ur.status = 1 
    AND NOT ur.action_date IS NULL 
    AND ur.user_id IN (@v1);
    AND r.type = 'WORKOUT'
GROUP BY ur.user_id

Problem here is that the first statement returns an error:

Subquery returns more than 1 row.

Expected result are user_id's, that can be duplicates. And I need those duplicated for the count.

How can I fix this?

Answer 1

Try EXISTS instead of IN

...
AND EXISTS (SELECT *
                   FROM user_resource ur2
                   WHERE ur2.user_id = ur.user_id
                         AND ur2.action_date >= now() - INTERVAL 2 DAY)
...

and indices on user_resource (user_id, action_date) , user_resource (status, action_date, user_id) and/or user_resource (type) .

Answer 2

You could try:

-- first statement
SET @v1 = (SELECT GROUP_CONCAT(user_id)
FROM user_resource ur2
WHERE ur2.action_date >= now() - INTERVAL 2 DAY)

-- second statement
SELECT
ur.user_id      AS user_id,
 sum(r.duration) AS total_time,
 count(user_id)  AS prefixes
FROM user_resource ur
INNER JOIN resource r ON r.id = ur.resource_id
WHERE ur.status = 1 AND NOT ur.action_date IS NULL AND FIND_IN_SET(ur.user_id,@v1)
AND r.type = 'WORKOUT'
GROUP BY ur.user_id

Answer 3

Additional join will be faster then sub-query:

SELECT
    ur.user_id      AS user_id,
    sum(r.duration) AS total_time,
    count(user_id)  AS number_of_workouts
FROM user_resource ur
INNER JOIN resource r ON r.id = ur.resource_id
INNER JOIN (
    SELECT user_id
    FROM user_resource ur2
    WHERE ur2.action_date >= now() - INTERVAL 2 DAY
) t ON t.user_id = ur.user_id
WHERE 
    ur.status = 1 
    AND NOT ur.action_date IS NULL 
    AND r.type = 'WORKOUT'
    GROUP BY ur.user_id;