MySQL 优化平均查询

Question

我有这样的数据库（实际上有超过30个不同的sKey）：

+----+------+------+---------------------+
|ID  | sKey | sVal |      timestamp      |
+----+------+------+---------------------+
| 1  | temp |   19 | 2023-07-14 20:32:06 |
| 2  | humi |   60 | 2023-07-14 20:33:06 |
| 3  | temp |   20 | 2023-07-14 20:34:06 |
| 4  | humi |   65 | 2023-07-14 20:35:06 |
| 5  | pres | 1023 | 2023-07-14 20:36:06 |
| 6  | temp |   22 | 2023-07-14 20:37:06 |
| 7  | temp |   21 | 2023-07-14 20:38:06 |
| 8  | pres | 1028 | 2023-07-14 20:39:06 |
| 9  | temp |   20 | 2023-07-14 20:40:06 |
|10  | temp |   19 | 2023-07-14 20:43:06 |  <-time glitch
|11  | pres | 1022 | 2023-07-14 20:44:06 |
|12  | temp |   19 | 2023-07-14 20:45:06 |
|13  | humi |   66 | 2023-07-14 20:46:06 |
|14  | humi |   63 | 2023-07-14 20:47:06 |
|15  | temp |   19 | 2023-07-14 20:48:06 |
|16  | pres | 1029 | 2023-07-14 20:49:06 |
|20  | temp |   19 | 2023-07-14 20:50:06 | <- ID not consecutive (deleted records)
|21  | pres | 1022 | 2023-07-14 20:61:06 |
|22  | temp |   19 | 2023-07-14 20:62:06 |
|23  | pres | 1029 | 2023-07-14 20:63:06 |
+----+------+------+---------------------+

现在我希望获得每个 sKey 的平均值（值以 sVal 为单位）。 现在我有可行的解决方案，但速度很慢。实际上我有 3 个独立的查询，如下所示：

SELECT AVG(`sVal`), `timestamp` FROM `Test` WHERE sKey='temp'  AND timestamp between '2023-07-14 20:34:06' and '2023-07-14 20:51:06' GROUP BY FLOOR(TO_SECONDS(`timestamp`)/180)
SELECT AVG(`sVal`), `timestamp` FROM `Test` WHERE sKey='humi'  AND timestamp between '2023-07-14 20:34:06' and '2023-07-14 20:51:06' GROUP BY FLOOR(TO_SECONDS(`timestamp`)/180)
SELECT AVG(`sVal`), `timestamp` FROM `Test` WHERE sKey='pres'  AND timestamp between '2023-07-14 20:34:06' and '2023-07-14 20:51:06' GROUP BY FLOOR(TO_SECONDS(`timestamp`)/180)

由于数据库目前已经有超过2000000条记录，单次查询大约需要3秒。我假设如果我可以以某种方式将查询连接到仅一个 WHERE sKey=... 结果应该会更快。那么如何改进呢。

或者也许我的方法根本就是错误的。想要的结果是获得每个 sKey 的平均 sVal。平均间隔不能以记录编号（ID）为单位，因为某些记录可能会被删除。即使 ID 在行中，也可能会丢失一些记录。所以我认为只能使用时间戳本身的间隔。但我是 SQL 新手，我可能会错过一些东西。

Answer 1

您可以使用条件平均值

SELECT 
    AVG(IF(sKey = 'temp', `sVal`, 0)) AS `temp`,
    AVG(IF(sKey = 'humi', `sVal`, 0)) AS `humi`,
    AVG(IF(sKey = 'pres', `sVal`, 0)) AS `pres`,
    MIN(`timestamp`) As `timestamp`
FROM
    `Test`
WHERE
    timestamp BETWEEN '2023-07-14 20:34:06' AND '2023-07-14 20:51:06'
GROUP BY FLOOR(TO_SECONDS(`timestamp`) / 180);