如何在 MariaDB 中选择每个组中的最新成员？答案

【问题标题】：How to select latest member in each group in MariaDB?如何在 MariaDB 中选择每个组中的最新成员？
【发布时间】：2017-02-17 08:09:23
【问题描述】：

我有 3 张桌子：

帐户 - 帐户信息
机器 - 机器信息
account_machine - 将帐户映射到某个日期的机器

每个帐户由一台机器处理。随着时间的推移，一个帐户可以迁移到不同的机器上，但在某一天它只能由一台机器处理。如果一个帐户不再有效，则对应的 machine_id 为 0。给定一个日期，我想查找所有活动帐户，所以我想出了这个查询：

SELECT account.id 
FROM account JOIN account_machine m 
ON m.account_id=account.id && m.machine_id && m.machine_id=
(SELECT machine_id 
FROM account_machine 
WHERE account_id=account.id && date<=20170215 
ORDER BY date DESC LIMIT 1) 
GROUP BY account.id;

这适用于 MySQL，但不适用于 MariaDB。

MariaDB [db]> select * from account_machine;
+------------+------------+------------+
| date       | account_id | machine_id |
+------------+------------+------------+
| 2013-01-01 |          1 |          1 |
| 2013-01-01 |          8 |          1 |
| 2013-01-01 |          2 |          2 |
| 2013-01-01 |          3 |          2 |
| 2013-01-01 |          4 |          3 |
| 2013-01-01 |         12 |          3 |
| 2016-04-01 |         24 |          3 |
| 2013-01-01 |          5 |          5 |
| 2013-01-01 |          6 |          8 |
| 2013-01-01 |          7 |          6 |
| 2014-01-01 |          9 |          6 |
| 2013-01-01 |         10 |          4 |
| 2014-07-01 |         11 |         10 |
| 2014-01-01 |         13 |          7 |
| 2014-01-01 |         14 |          7 |
| 2014-07-01 |         15 |         11 |
| 2014-07-01 |         16 |         14 |
| 2014-07-01 |         17 |         12 |
| 2015-01-01 |         18 |         13 |
| 2015-01-01 |         19 |         13 |
| 2015-04-01 |         20 |         13 |
| 2015-04-01 |         21 |          7 |
| 2015-04-01 |         22 |         13 |
| 2016-04-01 |         23 |         15 |
| 2016-05-01 |         25 |          9 |
| 2016-05-19 |         26 |          4 |
| 2014-08-06 |          1 |          0 |
| 2016-01-15 |         12 |          0 |
| 2015-11-04 |         19 |         12 |
| 2016-05-23 |         10 |          0 |
| 2016-05-26 |          2 |         18 |
| 2016-05-27 |         13 |         16 |
| 2016-06-02 |         27 |          3 |
| 2016-06-02 |          4 |          0 |
| 2016-06-08 |         28 |         17 |
| 2016-06-21 |         29 |         19 |
| 2016-07-11 |         30 |         20 |
| 2016-08-15 |         13 |          0 |
| 2016-08-19 |          2 |         18 |
| 2016-08-25 |         31 |         21 |
| 2016-09-08 |         32 |         20 |
| 2016-11-30 |         19 |         12 |
| 2016-11-30 |         22 |         13 |
| 2017-01-20 |         33 |         15 |
+------------+------------+------------+

MariaDB [db]> select account.id from account join account_machine m on m.account_id=account.id && m.machine_id && m.machine_id=(select a.machine_id from account_machine a where a.account_id=account.id && a.date<=20170215 order by a.date desc limit 1) group by account.id;
+----+
| id |
+----+
| 23 |
| 33 |
+----+

mysql> select account.id from account join account_machine m on m.account_id=account.id && m.machine_id && m.machine_id=(select a.machine_id from account_machine a where a.account_id=account.id && a.date<=20170215 order by a.date desc limit 1) group by account.id;
+----+
| id |
+----+
|  2 |
|  3 |
|  5 |
|  6 |
|  7 |
|  8 |
|  9 |
| 11 |
| 14 |
| 15 |
| 16 |
| 17 |
| 18 |
| 19 |
| 20 |
| 21 |
| 22 |
| 23 |
| 24 |
| 25 |
| 26 |
| 27 |
| 28 |
| 29 |
| 30 |
| 31 |
| 32 |
| 33 |
+----+

附：这里有 3 个表格供您复制：

CREATE TABLE `account` (
  `id` smallint(5) unsigned NOT NULL AUTO_INCREMENT,
  PRIMARY KEY (`id`)
) ENGINE=MyISAM;
INSERT INTO `account` VALUES (1),(2),(3),(4),(5),(6),(7),(8),(9),(10),(11),(12),(13),(14),(15),(16),(17),(18),(19),(20),(21),(22),(23),(24),(25),(26),(27),(28),(29),(30),(31),(32),(33);

CREATE TABLE `account_machine` (
  `date` date NOT NULL,
  `account_id` smallint(5) unsigned NOT NULL,
  `machine_id` smallint(5) unsigned NOT NULL,
  PRIMARY KEY (`date`,`account_id`)
) ENGINE=MyISAM;
INSERT INTO `account_machine` VALUES ('2013-01-01',1,1),('2013-01-01',8,1),('2013-01-01',2,2),('2013-01-01',3,2),('2013-01-01',4,3),('2013-01-01',12,3),('2016-04-01',24,3),('2013-01-01',5,5),('2013-01-01',6,8),('2013-01-01',7,6),('2014-01-01',9,6),('2013-01-01',10,4),('2014-07-01',11,10),('2014-01-01',13,7),('2014-01-01',14,7),('2014-07-01',15,11),('2014-07-01',16,14),('2014-07-01',17,12),('2015-01-01',18,13),('2015-01-01',19,13),('2015-04-01',20,13),('2015-04-01',21,7),('2015-04-01',22,13),('2016-04-01',23,15),('2016-05-01',25,9),('2016-05-19',26,4),('2014-08-06',1,0),('2016-01-15',12,0),('2015-11-04',19,12),('2016-05-23',10,0),('2016-05-26',2,18),('2016-05-27',13,16),('2016-06-02',27,3),('2016-06-02',4,0),('2016-06-08',28,17),('2016-06-21',29,19),('2016-07-11',30,20),('2016-08-15',13,0),('2016-08-19',2,18),('2016-08-25',31,21),('2016-09-08',32,20),('2016-11-30',19,12),('2016-11-30',22,13),('2017-01-20',33,15);

CREATE TABLE `machine` (
  `id` smallint(5) unsigned NOT NULL AUTO_INCREMENT,
  PRIMARY KEY (`id`)
) ENGINE=MyISAM;
INSERT INTO `machine` VALUES (1),(2),(3),(4),(5),(6),(7),(8),(9),(10),(11),(12),(13),(14),(15),(16),(17),(18),(19),(20),(21),(22);

【问题讨论】：

两个模式中的表是否具有相同的数据？
顺便说一句，And m.machine_id 总是正确的。如果您费心格式化查询，这对您来说也很明显
见meta.stackoverflow.com/questions/333952/…
@Strawberry 感谢您进行编辑以使其清晰。 && m.machine_id 不是错字。正如我所解释的， machine_id=0 意味着该帐户不再有效。因此，&& m.machine_id 我只会获得活跃帐户。
我没有建议它是。再次阅读您的查询。小心点。

标签： mysql group-by subquery mariadb

【解决方案1】：

这样的事情呢？

SELECT am1.account_id AS id
FROM account_machine am1
JOIN (
    SELECT account_id, MAX(date) AS date
    FROM account_machine
    GROUP BY account_id
    ) am2
ON am1.account_id = am2.account_id
AND am1.date = am2.date
AND am1.machine_id != 0
ORDER BY am1.account_id;

+----+
| id |
+----+
|  2 |
|  3 |
|  5 |
|  6 |
|  7 |
|  8 |
|  9 |
| 11 |
| 14 |
| 15 |
| 16 |
| 17 |
| 18 |
| 19 |
| 20 |
| 21 |
| 22 |
| 23 |
| 24 |
| 25 |
| 26 |
| 27 |
| 28 |
| 29 |
| 30 |
| 31 |
| 32 |
| 33 |
+----+
28 rows in set (0.00 sec)

我很想看看 MySQL 和 MariaDB 的 EXPLAIN EXTENDED / SHOW WARNINGS 的输出。这将准确地向您展示查询优化器是如何重写查询的。例如：

root@localhost [stack]> EXPLAIN EXTENDED SELECT am1.account_id AS id
    -> FROM account_machine am1
    -> JOIN (
    ->     SELECT account_id, MAX(date) AS date
    ->     FROM account_machine
    ->     GROUP BY account_id
    -> ) am2
    -> ON am1.account_id = am2.account_id
    -> AND am1.date = am2.date
    -> AND am1.machine_id != 0
    -> ORDER BY am1.account_id\G
*************************** 1. row ***************************
           id: 1
  select_type: PRIMARY
        table: <derived2>
         type: ALL
possible_keys: NULL
          key: NULL
      key_len: NULL
          ref: NULL
         rows: 44
     filtered: 100.00
        Extra: Using where; Using temporary; Using filesort
*************************** 2. row ***************************
           id: 1
  select_type: PRIMARY
        table: am1
         type: eq_ref
possible_keys: PRIMARY
          key: PRIMARY
      key_len: 5
          ref: am2.date,am2.account_id
         rows: 1
     filtered: 100.00
        Extra: Using where
*************************** 3. row ***************************
           id: 2
  select_type: DERIVED
        table: account_machine
         type: index
possible_keys: NULL
          key: PRIMARY
      key_len: 5
          ref: NULL
         rows: 44
     filtered: 100.00
        Extra: Using index; Using temporary; Using filesort
3 rows in set, 1 warning (0.00 sec)

root@localhost [stack]> SHOW WARNINGS\G
*************************** 1. row ***************************
  Level: Note
   Code: 1003
Message: select `stack`.`am1`.`account_id` AS `id` from
`stack`.`account_machine` `am1` join (select 
`stack`.`account_machine`.`account_id` AS 
`account_id`,max(`stack`.`account_machine`.`date`) AS `date` from 
`stack`.`account_machine` group by 
`stack`.`account_machine`.`account_id`) `am2` where 
((`stack`.`am1`.`account_id` = `am2`.`account_id`) and 
(`stack`.`am1`.`date` = `am2`.`date`) and (`stack`.`am1`.`machine_id` 
<> 0)) order by `stack`.`am1`.`account_id`
1 row in set (0.00 sec)

显然不是没有索引的高性能查询，但对于有限的数据集来说它很好。

【讨论】：

【解决方案2】：

我怀疑您的查询存在设计缺陷 - 如果子查询返回 account_id 为 machine_id = 0。之后就不会再看下去了。

当使用JOIN...ON 时，最好将加入信息放在ON 子句中，而不是过滤信息；进入WHERE。

看起来这样会更简单更快：

SELECT  account_id
    FROM  account_machine AS m
    WHERE  machine_id != 0
      AND  date <= 20170215
      AND  EXISTS (
        SELECT  *
            FROM  account
            WHERE  id = m.account_id 
                  )
    ORDER BY  date DESC
    LIMIT  1;

也许EXISTS() 测试是多余的，可以删除？

INDEX(date) 可能有助于提高性能。

（不，我没有发现为什么两台服务器的工作方式可能不同。看看我的版本是否有效。）

【讨论】：

您的查询显然是错误的——我想返回所有活动帐户，但您的帐户仅限于 1 个。此外，您对 machine_id=0 的猜测也是错误的。我试图删除该子句，但 MariaDB 中仍然只有 2 行。