SQL，max（），分组依据

Question

我的问题-应用group by时，我需要对某列的某些条件进行MAX，但选择该列（而不是我用来评估max的条件）。需要帮助。

例如，假设我有一个名为RuleUID的列，其值类似70_rule，110_rule，120_rule等。现在我想做这样的事情。

select firstname, lastname, MAX( CAST(SUBSTRING(RuleUID,1,CHARINDEX('_',RuleUID)-1) AS INT) as **?????**
from employee
group by firstname, lastname

在上面的代码片段中，我需要能够选择对应于MAX标准的RuleUid（在??????中）。我该怎么办。

Answer 1

您可以使用order by并以降序获得top 1。

sql服务器

select firstname
    , lastname
    , RuleUID
from employee
order by substring(RuleUID,1,charindex('_',RuleUID)-1) desc

mysql

select top 1 firstname
    , lastname
    , RuleUID
from employee
order by substring(RuleUID,1,charindex('_',RuleUID)-1) desc
limit 1

Answer 2

看来您需要将数字从RuleUID中抽出很多。因此，使一个函数可以重用：

CREATE FUNCTION dbo.getRuleNo(@ruleUID varchar(50))  
RETURNS int   
AS   
BEGIN  

  RETURN CAST(SUBSTRING(@ruleUID, 1, CHARINDEX('_',@ruleUID) - 1) AS INT)

END;
GO

然后将该函数用于此问题和其他问题：

select firstname, 
lastname, 
CAST(MAX(dbo.getRuleNo(RuleUID)) AS VARCHAR) + '_rule' as MaxRule
from employee
group by firstname, lastname

Answer 3

    select firstname, 
    lastname, 
    replace(ltrim(replace(MAX(RIGHT('0000000000' + RuleUID, 10)), '0', ' ')), ' ', '0')
        as MaxRule
    from employee
    group by firstname, lastname

Answer 4

有多种解决方法。一种是将RuleUID的相同子字符串与MAX值进行比较：

SELECT firstname, lastname, RuleUID
FROM employee
WHERE CAST(SUBSTRING(RuleUID,1,CHARINDEX('_',RuleUID)-1) AS INT) =
          (SELECT MAX(CAST(SUBSTRING(RuleUID,1,CHARINDEX('_',RuleUID)-1) AS INT)) 
           FROM employee)

[另一个是将值的_rule部分加回MAX值，然后直接进行比较。这应该更有效，因为您可以在RuleUID上使用索引：

SELECT firstname, lastname, RuleUID
FROM employee
WHERE RuleUID = (SELECT CONCAT(MAX(CAST(SUBSTRING(RuleUID,1,CHARINDEX('_',RuleUID)-1) AS INT)), '_rule') 
                 FROM employee)

如果您不想使用子查询，则可以将CTE与ROW_NUMBER()一起使用：

WITH CTE AS (
SELECT firstname, lastname, RuleUID,
       ROW_NUMBER() OVER (ORDER BY CAST(SUBSTRING(RuleUID,1,CHARINDEX('_',RuleUID)-1) AS INT) DESC) AS rn
FROM employee)
SELECT firstname, lastname, RuleUID
FROM CTE
WHERE rn = 1

或RANK()：

WITH CTE AS (
SELECT firstname, lastname, RuleUID,
       ROW_NUMBER() OVER (ORDER BY CAST(SUBSTRING(RuleUID,1,CHARINDEX('_',RuleUID)-1) AS INT) DESC) AS rn
FROM employee)
SELECT firstname, lastname, RuleUID
FROM CTE
WHERE rn = 1

最后，您可以使用简单的TOP查询：

SELECT TOP 1 firstname
    , lastname
    , RuleUID
FROM employee
ORDER BY CAST(SUBSTRING(RuleUID,1,CHARINDEX('_',RuleUID)-1) AS INT) DESC

在所有情况下，对于我的演示，输出都是相同的：

firstname   lastname    RuleUID
someone     else        120_rule

Demo on dbfiddle

Answer 5

首先，我们按LEN（RuleUID）DESC排序。然后通过RuleUID DESC。第一个是您的最高现在您不需要GROUP BY。

    ;WITH cte AS
    (
       SELECT LastName, FirstName, RuleUID,
             ROW_NUMBER() OVER (PARTITION BY LastName, FirstName 
             ORDER BY LEN(RuleUID) DESC, RuleUID DESC) AS rn
       FROM Employees
    )


    SELECT *
    FROM cte
    WHERE rn = 1
    ORDER BY LastName, FirstName

这应该为每个名称提供1行以及它们的最大RuleUID。