0

我有一个查询问题，我想在其中累积值然后对它们进行分组 PS：这是一个简化的示例

CREATE TABLE [dbo].[OWNER](
    [personId] [int] NULL,
    [ownerId] [int] NULL,
    [firstName] [varchar](255) NULL,
    [lastName] [varchar](255) NULL,
)
GO

CREATE TABLE [dbo].[INCOME](
    [personId] [int] NULL,
    [amount] [float] NULL,
    [received] [date] NULL,
)
GO


INSERT INTO OWNER Values(1,null,   'John', 'Smith')
INSERT INTO OWNER Values(1,null,   'John', 'Smith')
INSERT INTO OWNER Values(1,null,   'John', 'Smith')
INSERT INTO OWNER Values(200,1,    'Tom', 'Lawn')
INSERT INTO OWNER Values(3,3,      'Chris', 'Hanson')
INSERT INTO OWNER Values(400,4,    'Jack', 'Man')
INSERT INTO OWNER Values(4,null,   'Donald', 'McMan')
INSERT INTO OWNER Values(5,null,   'John', 'Hanson')
INSERT INTO OWNER Values(700,5,    'Peter', 'Darcy')
INSERT INTO OWNER Values(700,5,    'Peter', 'Darcy')


Insert INTO INCOME VALUES(1, 300, '2020-01-01')
Insert INTO INCOME VALUES(200, 1000, '2020-01-01')
Insert INTO INCOME VALUES(3, 200, '2020-01-01')
Insert INTO INCOME VALUES(4,300, '2020-01-01')
Insert INTO INCOME VALUES(5,300, '2020-01-01')
Insert INTO INCOME VALUES(1,300,  '2020-01-01')
Insert INTO INCOME VALUES(3,300,  '2020-01-01')
Insert INTO INCOME VALUES(5,500,  '2020-01-01')
Insert INTO INCOME VALUES(700,500,  '2020-01-01')

在 Owner 表中可以有重复的记录。做一个聚合我可以写

select personId, 
       sum(amount) as total,
       count(*) as rows
from income
group by personid

这将导致

personId    total   rows
1           600      2
3           500      2
4           300      1
5           800      2
200         1000     1
700         500      1

问题是我想获取 Owner 的聚合并得到以下结果

personId    total  rows
1           1600   3
3           500    2
4           300    1
5           1300   3

由于 personId 200 的 ownerId 为 1，而 personId 700 的 ownerId 为 5。我将不胜感激有关如何实现这一点的帮助。

4

1 回答 1

0

您的数据模型存在真正的问题——您应该解决这个问题。有重复的行owner是很糟糕的。至少，你可以有一个生效日期，这样你就知道最近的记录是什么

也就是说，您可以使用apply来选择任意匹配记录并使用它：

select coalesce(o.ownerid, i.personid), 
       sum(amount) as total,
       count(*) as rows
from income i cross apply
     (select top (1) o.*
      from owner o
      where o.personid = i.personid
     ) o
group by coalesce(o.ownerid, i.personid);

这是一个 db<>fiddle。

于 2020-05-21T11:33:18.470 回答