Optimizing MySQL query across multiple rows

Question

my current query looks something like this. I've been trying to optimize it but I got no luck so far. My goal is to get the difference in stat_value between the current day and the previous day for every player and then order it. The current code works fine, but feels unoptimized to me. The ? values are filled in using Java.

SELECT t1.stat_value - 
       (SELECT stat_value 
          FROM leaderheadsplayersdata_daily t2 
         WHERE t2.player_id = t1.player_id 
           AND t2.day = ? 
           AND t2.stat_type = ? 
         LIMIT 1
       ) as sum
     , (SELECT name 
          FROM leaderheadsplayers 
         WHERE leaderheadsplayers.player_id = t1.player_id 
         LIMIT 1
       ) 
    FROM leaderheadsplayersdata_daily t1 
   WHERE day = ? 
     AND stat_type = ? 
   ORDER 
      BY sum DESC LIMIT 100

Table structure:

This is the main table containing the player id and username.

CREATE TABLE IF NOT EXISTS `leaderheadsplayers` 
(player_id INTEGER PRIMARY KEY AUTO_INCREMENT
,uuid VARCHAR(36) NOT NULL UNIQUE
,name VARCHAR(16) NOT NULL
,last_join DATETIME
) ENGINE = InnoDB

This is the table containing the daily data.

CREATE TABLE IF NOT EXISTS leaderheadsplayersdata_daily 
(player_id INTEGER NOT NULL
,stat_value DOUBLE NOT NULL
,stat_type VARCHAR(16) NOT NULL
,day INTEGER NOT NULL
,FOREIGN KEY (player_id) REFERENCES leaderheadsplayers(player_id) ON DELETE CASCADE
,PRIMARY KEY(player_id, stat_type, day)
) ENGINE = InnoDB

Thanks in advance

Answer 1

Because you only have 2 days, you can do a double join on the leaderheadsplayersdata_daily table.

It would look something like this:

SELECT
    p.player_id,
    p.name,
    (dc.stat_value-dp.statvalue) AS difference,
    dc.day,
FROM leaderheadsplayers p
JOIN leaderheadsplayersdata_daily dc ON p.player_id = dc.player_id
JOIN leaderheadsplayersdata_daily dp ON p.player_id = dp.player_id and dp.day = (dc.day-1)
WHERE dc.day = ?CURRENT_DAY?
ORDER BY difference DESC

dc stands for "date current". dp stands for "date previous". p stands for "player".

For good performance, add indexes on the columns player_id (in both tables) and day .

Sorry, I didn't test, have not ideea if I have syntax errors or stuff. Also, please replace ?CURRENT_DAY? with the actual day and add other conditions in WHERE clause, as needed

If ORDER BY difference DESC does not work, just do ORDER BY (dc.stat_value-dp.statvalue) DESC .

You have a lot of subqueries, you look like you could take a look at some documentation about JOINs . They're really useful.

Answer 2

Alternatively, you can try the following. Let me know how this version that uses implicit triple join fare with the one suggested above that uses explicit triple join. Thanks.

SELECT t3.player_id, improvement FROM 
   (SELECT t1.player_id as player_id, (t1.stat_value - t2.stat_value) as improvement 
      FROM leaderheadsplayersdata_daily t1 INNER JOIN leaderheadsplayersdata_daily t2 
         on t1.player_id = t2.player_id and t1.stat_type = t2.stat_type 
            and t2.day = ? and t1.day = ?) as t3 
   INNER JOIN leaderheadsplayers on 
      t3.player_id = leaderheadsplayers.player_id 
          ORDER BY improvement DESC LIMIT 100;