选择具有按两列分组的最大值的行

Question

我已经看到了很多关于这类问题的解决方案（特别是这一个SQL只选择列上具有最大值的行 ），但这些似乎都不合适：

我有以下表格布局，附件的版本控制，绑定到实体：

TABLE attachments
+------+--------------+----------+----------------+---------------+
| id   | entitiy_id   | group_id | version_number | filename      |
+------+--------------+----------+----------------+---------------+
| 1    | 1            | 1        | 1              | file1-1.pdf   |
| 2    | 1            | 1        | 2              | file1-2.pdf   |
| 3    | 1            | 2        | 1              | file2-1.pdf   |
| 4    | 2            | 1        | 1              | file1-1.pdf   |
| 5    | 2            | 1        | 2              | file1-2.pdf   |
| 6    | 2            | 3        | 1              | file3-1.pdf   |
+------+--------------+----------+----------------+---------------+

输出应该是最大版本号，按group_id和entity_id分组，如果有帮助，我只需要单个entity_ids的列表：

+------+--------------+----------+----------------+---------------+
| id   | entitiy_id   | group_id | version_number | filename      |
+------+--------------+----------+----------------+---------------+
| 2    | 1            | 1        | 2              | file1-2.pdf   |
| 3    | 1            | 2        | 1              | file2-1.pdf   |
| 5    | 2            | 1        | 2              | file1-2.pdf   |
| 6    | 2            | 3        | 1              | file3-1.pdf   |
+------+--------------+----------+----------------+---------------+

我想出的是这个自我加入：

SELECT *
FROM   `attachments` `attachments`
       LEFT OUTER JOIN attachments t2
         ON ( attachments.group_id = t2.group_id
              AND attachments.version_number < t2.version_number )
WHERE  ( t2.group_id IS NULL )
   AND ( `t2`.`id` = 1 )
GROUP  BY t2.group_id

但是这个只有在不同的实体不共享相同的组号时才有效。 不幸的是，这是必要的。

我在创建视图时遇到了一个有效的解决方案，但在我当前的设置中不支持此功能。

任何想法都受到高度赞赏。 谢谢！

Answer 1

尝试这个：

select t1.* from attachments t1
left join attachments t2
on t1.entity_id = t2.entity_id and t1.group_id = t2.group_id and
   t1.version_number < t2.version_number
where t2.version_number is null

Answer 2

这可用于选择所有

SELECT attachments.*
FROM (
    SELECT entitiy_id, group_id, MAX(version_number) AS max_version
    FROM attachments
    GROUP BY entitiy_id, group_id
) AS maxVersions
INNER JOIN attachments
ON attachments.entitiy_id = maxVersions.entitiy_id
AND attachments.group_id = maxVersions.group_id
AND attachments.version_number = maxVersions.max_version

将此扩展为仅查找单个entitiy_id只需要在子查询中添加WHERE，这样就可以了

SELECT attachments.*
FROM (
    SELECT entitiy_id, group_id, MAX(version_number) AS max_version
    FROM attachments
    WHERE entitity_id = [[YOUR ENTITIY ID HERE]]
    GROUP BY entitiy_id, group_id
) AS maxVersions
INNER JOIN attachments
ON attachments.entitiy_id = maxVersions.entitiy_id
AND attachments.group_id = maxVersions.group_id
AND attachments.version_number = maxVersions.max_version

如果你希望确保随着行数的增加这种情况继续快速运行，我建议你确保在行上添加一个密钥(entitiy_id, group_id, max_version)因为子查询可以依赖因此，确保它不会锁定表。

Answer 3

这样可以解决问题：

select a1.* from attachments a1
inner join ( select entitiy_id, group_id, max(version_number) as version_number
             from attachments
             group by entitiy_id, group_id) a2 on a1.entitiy_id = a2.entitiy_id and
                                                  a1.group_id = a2.group_id and
                                                  a1.version_number = a2.version_number

Answer 4

您还可以使用高性能通用表表达式（CTE）来解决此问题。

WITH CTE AS
(
SELECT entitiy_id, group_id, version_number, filename,       
ROW_NUMBER() OVER (PARTITION BY entitiy_id, group_id ORDER BY version_number DESC) as RowNum
FROM attachments
)
SELECT entitiy_id, group_id, version_number, filename
FROM CTE 
WHERE RowNum = 1

要么

SELECT T.entitiy_id, T.group_id, T.version_number, T.filename
FROM (SELECT entitiy_id, group_id, version_number, filename,       
     ROW_NUMBER() OVER (PARTITION BY entitiy_id, group_id ORDER BY version_number DESC) as RowNum
     FROM attachments
     ) as T 
WHERE RowNum = 1