0

我正在使用一个看起来像这样的表：

|   |  Name | CaseID | UsrID | DL_NO |   SSN   | Address     | DateSeen   |
|---|:-----:|:------:|:-----:|:-----:|:-------:|-------------|------------|
| 1 | Smith |  AB190 | 88885 |       | 1234567 | 222 Side Rd | 01/01/2020 |
| 2 | Smith |  AB186 | 88885 | B0938 |         |             | 10/01/2019 |
| 3 | Smith |  AB170 | 88885 |       | 1234567 | 123 Side Rd | 09/01/2019 |
| 4 | Smith |  AB168 | 88885 | B0938 |         | 123 Road St | 03/05/2019 |
| 5 | Smith |  AB132 | 88885 | B0938 | 1234567 |             | 03/01/2019 |
| 6 | Smith |  AB102 | 88885 | B0938 | 1234567 | 123 Road St | 02/01/2019 |

我无法理解如何正确包含更新/偶尔丢失位的数据。

我想看到的是每列最新的非空值：

|   |  Name | NumOfCases | UsrID | DL_NO |   SSN   | Address     |
|---|:-----:|:----------:|:-----:|:-----:|:-------:|-------------|
| 1 | Smith |      6     | 88885 | B0938 | 1234567 | 222 Side Rd |

我正在使用这个：

SELECT TOP 50 Name, UsrID, COUNT(DISTINCT CaseID) as NumofCases
FROM People
WHERE DateSeen between 01/31/2019 and 10/02/2019
GROUP BY Name, UsrID
ORDER BY DateSeen desc

退货

|   |  Name | UsrID | NumofCases |
|---|:-----:|-------|:----------:|
| 1 | Smith | 88885 |      6     |

在我意识到我对其他领域有用之前，这一切正常。

当我尝试对附加列进行类似查询时，强制 GROUP BY 显然会破坏“NumofCases”计数。

即使是我“SELECT TOP 1”的 CROSS APPLY 也需要分组依据。

有什么想法吗？

4

3 回答 3

0

您可以尝试以下查询 -

SELECT Name
      ,COUNT(DISTINCT CaseID) OVER(PARTITION BY Name)
      ,UsrID
      ,MAX(DL_NO) DL_NO
      ,MAX(SSN) SSN
      ,MAX(Address) Address
      ,MAX(DateSeen) DateSeen
FROM People
WHERE DateSeen BETWEEN 01/31/2019 AND 10/02/2019
GROUP BY Name, UsrID
ORDER BY DateSeen desc
于 2020-01-24T04:50:41.933 回答
0

这是可以帮助的：

SELECT TOP 50 Name, UsrID, COUNT(DISTINCT CaseID) as NumofCases,
(select top 1 b.DL_NO FROM People b where a.UsrID = b.UsrID and ltrim(rtrim(b.DL_NO)) <> '' and b.DL_NO is not null order by b.DateSeen desc) as DL_NO,
(select top 1 b.SSN  FROM People b where a.UsrID = b.UsrID and ltrim(rtrim(b.SSN)) <> '' and b.SSN is not null order by b.DateSeen desc) as SSN,
(select top 1 b.Address FROM People b where a.UsrID = b.UsrID and ltrim(rtrim(b.Address)) <> '' and b.Address is not null order by b.DateSeen desc) as Address,
FROM People a
WHERE DateSeen between '01/31/2019' and '10/02/2019'
GROUP BY Name, UsrID
于 2020-01-24T00:10:07.323 回答
0

如果您控制数据模型，规范化您的表将使问题变得更简单。它还可以防止数据不一致，例如示例数据中的地址不一致

create table People ( name, usrid, dl_no,ssn, address); --add data types
create table Case (usrid, dateseen, caseid);--add data types

create case_view as
select name,usrid,dl_no,ssn,address,dateseen,caseid
from people p join case c on p.usrid=c.usrid;

那么你的查询是微不足道的

select name,usrid,dl_no,ssn,address,COUNT(DISTINCT CaseID) as NumofCases
from case_view
group by name,usrid,dl_no,ssn,address

您可以添加日期或计数过滤器。

于 2020-01-24T14:15:07.017 回答