在Select和where子句中优化MYSQL查询TIMEDIFF

Question

I have the following query which runs over a large database.

select TIME_TO_SEC(TIMEDIFF(walkStartTime,walkEndTime)) 
from users 
where categoryType='1' 
and TIME_TO_SEC(TIMEDIFF(walkStartTime,walkEndTime)) < 1000

If you notice the TIME_TO_SEC is used in the query as well as in the where clause.

I tried for alias but, as alias can't be use so, don't know what could be a better solution for it.

Answer 1

One way to do this is

SELECT * FROM (
    SELECT TIME_TO_SEC(TIMEDIFF(walkStartTime,walkEndTime)) diff FROM users WHERE categoryType=1
) a
WHERE a.diff < 1000

Update

assuming you have a column named user_id which contains the user's id

SELECT user_id, AVG(diff) as 'avg_diff', MAX(diff) as 'max_diff' FROM (
    SELECT user_id, TIME_TO_SEC(TIMEDIFF(walkStartTime,walkEndTime)) diff FROM users WHERE categoryType=1
) a
WHERE a.diff < 1000
GROUP BY 1

This will return each user's id, average diff and max diff

Answer 2

i find only one optimized way to query your problem. When you calculate a value from your fields in the WHERE clause MySQL must calculate each row. So it is every time a full table scan and it can use much time and is not very performant.

Spend a new field with the calculated timediff and a index. you not must calculated it, you can use a VIRTUAL PERSISTENT field which calculated it automatic when you insert or change a field

I have made same Samples to clarify my idea:

First create a new table:

CREATE TABLE `users` (
  `id` INT(11) UNSIGNED NOT NULL AUTO_INCREMENT,
  `walkStartTime` TIMESTAMP NULL DEFAULT NULL,
  `walkEndTime` TIMESTAMP NULL DEFAULT NULL,
  `categoryType` INT(11) DEFAULT NULL,
  `diffp` INT(11) AS (TIME_TO_SEC(TIMEDIFF(walkEndTime,walkStartTime))) PERSISTENT,
  PRIMARY KEY (`id`),
  KEY `diffs` (`diffp`)
) ENGINE=INNODB AUTO_INCREMENT=10 DEFAULT CHARSET=utf8;

Insert some stuff:

INSERT INTO `users` (`id`, `walkStartTime`, `walkEndTime`, `categoryType`)
VALUES
    (1, '2015-09-27 07:00:00', '2015-09-27 07:30:00', 1),
    (2, '2015-09-27 07:00:01', '2015-09-27 07:31:00', 1),
    (3, '2015-09-27 07:00:02', '2015-09-27 07:32:00', 0),
    (4, '2015-09-27 07:00:10', '2015-09-27 07:15:00', 1),
    (5, '2015-09-27 07:00:20', '2015-09-27 07:16:10', 1),
    (6, '2015-09-27 07:00:30', '2015-09-27 07:17:20', 0),
    (7, '2015-09-27 07:01:00', '2015-09-27 07:10:00', 1),
    (8, '2015-09-27 07:02:00', '2015-09-27 07:09:33', 1),
    (9, '2015-09-27 07:03:00', '2015-09-27 08:12:00', 1);

try your query and EXPLAIN it: see that they read all 9 rows

MariaDB [tmp]> select TIME_TO_SEC(TIMEDIFF(walkEndTime,walkStartTime))
    -> from users
    -> where categoryType='1'
    -> and TIME_TO_SEC(TIMEDIFF(walkEndTime,walkStartTime)) < 1000;
+--------------------------------------------------+
| TIME_TO_SEC(TIMEDIFF(walkEndTime,walkStartTime)) |
+--------------------------------------------------+
|                                              890 |
|                                              950 |
|                                              540 |
|                                              453 |
+--------------------------------------------------+
4 rows in set (0.00 sec)

MariaDB [tmp]> EXPLAIN select TIME_TO_SEC(TIMEDIFF(walkEndTime,walkStartTime))
    -> from users
    -> where categoryType='1'
    -> and TIME_TO_SEC(TIMEDIFF(walkEndTime,walkStartTime)) < 1000;
+------+-------------+-------+------+---------------+------+---------+------+------+-------------+
| id   | select_type | table | type | possible_keys | key  | key_len | ref  | rows | Extra       |
+------+-------------+-------+------+---------------+------+---------+------+------+-------------+
|    1 | SIMPLE      | users | ALL  | NULL          | NULL | NULL    | NULL |    9 | Using where |
+------+-------------+-------+------+---------------+------+---------+------+------+-------------+
1 row in set (0.00 sec)

MariaDB [tmp]>

run my optimized query with the new VIRTUAL field (diffp):

MariaDB [tmp]> SELECT diffp
    -> FROM users
    -> WHERE diffp < 1000;
+-------+
| diffp |
+-------+
|   453 |
|   540 |
|   890 |
|   950 |
+-------+
4 rows in set (0.00 sec)

MariaDB [tmp]> EXPLAIN SELECT diffp
    -> FROM users
    -> WHERE diffp < 1000;
+------+-------------+-------+-------+---------------+-------+---------+------+------+--------------------------+
| id   | select_type | table | type  | possible_keys | key   | key_len | ref  | rows | Extra                    |
+------+-------------+-------+-------+---------------+-------+---------+------+------+--------------------------+
|    1 | SIMPLE      | users | range | diffs         | diffs | 5       | NULL |    3 | Using where; Using index |
+------+-------------+-------+-------+---------------+-------+---------+------+------+--------------------------+
1 row in set (0.00 sec)

MariaDB [tmp]>

so you can see MySQL only reads 3 ROWS and use the Index. It is also possible that can can more speedup when you using a composite index over both fields (diffp and CategoryType) and the order of the both fields in the index can also change the speed.