MySQL 语法混淆 - 合并简单数据以实现优雅的单行插入

Question

我正在构建一个 MySQL 表来为客户汇总一组特定信息。我已经使用简单的INNER JOIN 命令在具有共享键/列的记录上成功组装和过滤了大部分内容，但是数据的旋转/转置，即使是固定大小，仍然导致我的查询出现一些语法混乱.这张表t_snapshots的schema如下：

+-------------+---------------------+------+-----+---------------------+----------------+
| Field       | Type                | Null | Key | Default             | Extra          |
+-------------+---------------------+------+-----+---------------------+----------------+
| id          | int(11)             | NO   | PRI | NULL                | auto_increment | 
| value_type  | tinyint(1) unsigned | YES  |     | NULL                |                |
| ch1_id      | varchar(20)         | YES  |     | NULL                |                |
| ch1_val     | float               | NO   |     | 0                   |                |
| ch2_id      | varchar(20)         | YES  |     | NULL                |                |
| ch2_val     | float               | NO   |     | 0                   |                |
| ch3_id      | varchar(20)         | YES  |     | NULL                |                |
| ch3_val     | float               | NO   |     | 0                   |                |
| ch4_id      | varchar(20)         | YES  |     | NULL                |                |
| ch4_val     | float               | NO   |     | 0                   |                |
| timestamp   | datetime            | NO   | MUL | current_timestamp() |                |
+-------------+---------------------+------+-----+---------------------+----------------+

简单地说，我想从t_other_data0.ch[n] 中选择最近更新的值发送到t_snapshots.ch[n]_val，并从t_id_pool.unit_id 中为每个唯一的t_id_pool.channel_num 选择最近更新的id 发送到t_snapshots.ch[n]_id。 t_id_pool.channel_num 与列 t_snapshots.ch[n]_val 的 n 值相关：

--EDIT--：理想情况下，来自源表t_other_data0 的样本数据从t_id_pool 中查找channel_num=1,2,3,4 的最新unit_id 值并输出到表t_snapshots：

从t_other_data0 收集最新的过程数据。在这种情况下，将选择具有 id 5-8 的行，因为它们跨越所有不同的 value_type 和最新的 timestamp。：

Table: t_other_data0
+----+------+------+------+------+------------+---------------------+
| id | ch1  | ch2  | ch3  | ch4  | value_type | timestamp           |
+----+------+------+------+------+------------+---------------------+
|  1 | 1.65 | 3.25 | 1.98 | 2.17 |          1 | 2021-07-22 16:26:40 |
|  2 | 3.12 | 2.33 | 6.42 | 3.22 |          2 | 2021-07-22 16:26:40 |
|  3 | 2.22 | 2.24 | 3.34 | 1.17 |          3 | 2021-07-22 16:26:40 |
|  4 | 1.52 | 1.34 |  1.9 | 2.01 |          4 | 2021-07-22 16:26:40 |
|  5 |  3.2 | 3.21 | 5.42 | 2.13 |          1 | 2021-07-22 16:26:50 |
|  6 | 1.55 | 1.92 | 4.32 | 4.12 |          2 | 2021-07-22 16:26:50 |
|  7 | 2.31 | 1.93 | 2.36 |  3.4 |          3 | 2021-07-22 16:26:50 |
|  8 | 1.78 | 2.17 | 5.62 | 2.34 |          4 | 2021-07-22 16:26:50 |
+----+------+------+------+------+------------+---------------------+

因为这些永久频道会更改它们所绑定的临时设备，所以我们使用来自t_id_pool 的最新unit_id 来确定每个channel_num 的当前unit_id：

Table: t_id_pool
+----+---------------------+-------------+---------+
| id | timestamp           | channel_num | unit_id |
+----+---------------------+-------------+---------+
|  1 | 2021-07-22 09:39:09 |           1 | S4251   |
|  2 | 2021-07-22 09:38:09 |           2 | S3552   |
|  3 | 2021-07-22 09:38:09 |           3 | S0001   |
|  4 | 2021-07-22 09:38:09 |           4 | S1001   |
|  5 | 2021-07-22 09:39:10 |           1 | P5251   |
|  6 | 2021-07-22 09:38:10 |           2 | P4552   |
|  7 | 2021-07-22 09:38:10 |           3 | P1001   |
|  8 | 2021-07-22 09:38:10 |           4 | P2001
+----+---------------------+-------------+---------+

输出到t_snapshots:

Table: t_snapshots
+-----+---------------------+------------+--------+---------+--------+---------+--------+---------+--------+---------+
| id  | timestamp           | value_type | ch1_id | ch1_val | ch2_id | ch2_val | ch3_id | ch3_val | ch4_id | ch4_val |
+-----+---------------------+------------+--------+---------+--------+---------+--------+---------+--------+---------+
| 211 | 2021-07-14 16:26:50 |          1 | P5251  |     3.2 | P4552  |    3.21 | P1001  |    5.42 | P2001  |    2.13 |
| 212 | 2021-07-14 16:26:50 |          2 | P5251  |    1.55 | P4552  |    1.92 | P1001  |    4.32 | P2001  |    4.12 |
| 213 | 2021-07-14 16:26:50 |          3 | P5251  |    2.31 | P4552  |    1.93 | P1001  |    2.36 | P2001  |     3.4 |
+-----+---------------------+------------+--------+---------+--------+---------+--------+---------+--------+---------+



 

Answer 1

表t_other_data0 似乎是一个数据透视表。因此，我认为第一步是将其取消旋转，然后将其与t_id_pool 表连接以获取最新的unit_id，然后再次重新旋转它。也许这样的查询可以工作：

SELECT 0 id, tod.timestamp, value_type,
       MAX(case when channel_num=1 THEN unit_id else 0 END) AS ch1_id,
       SUM(case when channel_num=1 then chan_val else 0 END) as ch1_val,
       MAX(CASE WHEN channel_num=2 THEN unit_id ELSE 0 END) AS ch2_id,
       SUM(CASE WHEN channel_num=2 THEN chan_val ELSE 0 END) AS ch2_val,
       MAX(CASE WHEN channel_num=3 THEN unit_id ELSE 0 END) AS ch3_id,
       SUM(CASE WHEN channel_num=3 THEN chan_val ELSE 0 END) AS ch3_val,
       MAX(CASE WHEN channel_num=4 THEN unit_id ELSE 0 END) AS ch4_id,
       SUM(CASE WHEN channel_num=4 THEN chan_val ELSE 0 END) AS ch4_val
FROM   (SELECT value_type, ch1 AS chan_val, 1 AS chan_num, timestamp
        FROM   (SELECT *, Row_number() OVER (partition BY value_type ORDER BY id DESC) rn
                FROM t_other_data0) AS A
        WHERE  rn = 1 UNION ALL
        SELECT value_type, ch2, 2,
               timestamp
        FROM   (SELECT *, Row_number() OVER (partition BY value_type ORDER BY id DESC) rn
                FROM t_other_data0) AS A
        WHERE  rn = 1 UNION ALL
        SELECT value_type, ch3, 3,
               timestamp
        FROM   (SELECT *, Row_number() OVER (partition BY value_type ORDER BY id DESC) rn
                FROM t_other_data0) AS A
        WHERE  rn = 1 UNION ALL
        SELECT value_type, ch4, 4,
               timestamp
        FROM   (SELECT *, Row_number() OVER (partition BY value_type ORDER BY id DESC) rn
                FROM t_other_data0) AS A
        WHERE  rn = 1) AS tod
       JOIN (SELECT id, timestamp, channel_num, unit_id, 
                    Row_number() OVER (partition BY channel_num ORDER BY timestamp DESC) rn
             FROM t_id_pool) AS tip
         ON tod.chan_num = tip.channel_num AND tip.rn = 1
    GROUP BY tod.timestamp, value_type;

这里使用的函数之一是ROW_NUMBER()，其目的是在最新的value_type 和channel_number 时间戳上分配行号1。至于表t_other_data0，我使用的是UNION ALL，在ch1, ch2, ch3 & ch4 列之后总共有4 个查询。根据我采用的列，我为每个人分配了一个硬编码的chan_num。

我不确定id 列要填充什么，但我认为由于查询的主要目的是将INSERT 放入另一个表，那么id 列可能是自动递增的。

不幸的是 dbfiddle.uk 从昨天开始就不能使用了，所以这里的小提琴是针对 MySQL v8.0 而不是 MariaDB 10.3。 https://www.db-fiddle.com/f/xf1VmfYMbnGcabJS7dS6A1/1 。小提琴中的结果将为t_other_data.id=8 增加一行（@danblack 在评论中提到）并且不包括id=4，因为您的条件是“从 t_other_table0 收集最新的流程数据”。但是从您的预期输出来看，您似乎没有包含id=4，所以您的描述中可能有一些错误。