Question

我有一个类似于此的数据库表（我删除了不相关的字段）

[dbo].[UserMessage]
(
    [OwnerUserId] [int] NOT NULL,
    [FromUserId] [int] NOT NULL,
    [ToUserId] [int] NOT NULL,
    [CreationDateTime] [datetime2](7) NOT NULL
)

在此表中，我有以下数据

OwnerUserId    FromUserId    ToUserId    CreationDateTime   
13             13            15          2013-06-26 13:58:01.0270000
15             13            15          2013-06-26 13:58:24.0770000
13             15            13          2013-06-26 14:08:58.0070000
15             15            13          2013-06-26 14:08:43.8570000
13             13            15          2013-06-26 14:09:15.1030000
15             13            15          2013-06-26 14:09:26.2000000
13             15            13          2013-06-26 14:09:47.0030000
15             15            13          2013-06-26 14:09:38.1330000
13             13            16          2013-06-26 14:20:27.5170000
16             13            16          2013-06-26 14:20:41.0130000

我想要的是一个SQL查询，它为不同用户之间的粒子所有者返回最新的不同消息。所以如果我想要OwnerUserId = 13的所有对话，我希望返回的结果是

OwnerUserId    FromUserId    ToUserId    CreationDateTime
13             13            16          2013-06-26 14:20:27.5170000
13             15            13          2013-06-26 14:09:47.0030000

如果我想要OwnerUserId = 15的所有会话，我希望返回的结果是

OwnerUserId    FromUserId    ToUserId    CreationDateTime
15             15            13          2013-06-26 14:09:38.1330000

如果我想要OwnerUserId = 16的所有会话，我希望返回的结果是

OwnerUserId    FromUserId    ToUserId    CreationDateTime
16             13            16          2013-06-26 14:20:41.0130000

我为[OwnerUserId] = 13

创建了一个类似于此的查询

WITH cte AS
(
   SELECT *, ROW_NUMBER() OVER (PARTITION BY OwnerUserId, fromUserId, ToUserId ORDER BY CreationDateTime DESC) AS rn
   FROM [UserMessage]
)
SELECT *
FROM cte
WHERE rn = 1
and [OwnerUserId] = 13;

但此查询返回

OwnerUserId    FromUserId    ToUserId    CreationDateTime               rn
13             13            15          2013-06-26 14:09:15.1030000    1
13             13            16          2013-06-26 14:20:27.5170000    1
13             15            13          2013-06-26 14:09:47.0030000    1

我不希望返回顶行，因为最后一行已经代表同一个对话（尽管FromUserID和ToUserId被反转）。有任何想法吗？该解决方案适用于任何[OwnerUserId]（我使用13作为示例）

!!!!!!!!!!!!!!!! ANSWER !!!!!!!!!!!!!!

WITH cte AS
(
   SELECT *, 
   ROW_NUMBER() OVER 
   (
       PARTITION BY OwnerUserId, 
       CHECKSUM(CASE WHEN ToUserId > fromUserId THEN (convert(varchar(256),ToUserId) + ':' + convert(varchar(256),fromUserId)) ELSE (convert(varchar(256),fromUserId) + ':' + convert(varchar(256),ToUserId)) END) 
       ORDER BY CreationDateTime desc
    ) AS rn
    FROM [UserMessage]
)
SELECT *
FROM cte
WHERE rn = 1
and [OwnerUserId] = 13;

我通过在查询分区中使用校验和来达到解决方案。基本上我将fromUserId和toUserId结合起来为对话创建唯一的标识符。通过对OwnerUserId和新标识符进行分区，我可以将对话组合在一起。

SQL - 通过对多个列进行分组来获取不同的最新行

0 个答案: