sql server 没有子查询的最大值计数

Question

我想编写一个 T-SQL 查询，它不仅返回最大值，而且返回具有最大值的行数。 一定有比我想出的更好的方法

 --wrong way 
 select LibraryBranchId, max(daysCheckedOut), count(daysCheckedOut) 
 from books group by LibraryBranchId

 LibraryBranchId   Expr1      Expr2
 ----------------------------------
 1                 100       398503  (WRONG!)
 2                 75         94303  (WRONG!)
 3                 120       103950  (WRONG!)

我可以通过 INNER JOINing 子查询正确地做到这一点，但这似乎很浪费

 --right way, but seems WAY too long
 select LibraryBranchId,max(daysCheckedOut),count(daysCheckedOut)
 from books inner join 
   ( select LibraryBranchId, max(daysCheckedOut) as maxDaysCheckedOut
     from books group by LibraryBranchId ) as maxbooks 
 on books.LibraryBranchId=maxbooks.LibraryBranchId
 where daysCheckedOut=maxDaysCheckedOut
 group by LibraryBranchId 

 LibraryBranchId   Expr1      Expr2
 ----------------------------------
 1                 100           17  (RIGHT!)
 2                 75            11  (RIGHT!)
 3                 120            2  (RIGHT!)

所以有没有一种方法像查询#1一样简单，但返回正确的结果，如查询#1 2？

MS SQL Server 2000

编辑：在我第一次尝试输入此内容时，我错过了上面两个重要的 GROUP BY，我已添加它们 编辑：假装凯德·鲁克斯写的版本就是我写的

Answer 1

我认为这是正确的：

SELECT maxbooks.LibraryBranchId, maxbooks.maxDaysCheckedOut, count(*)
FROM books
INNER JOIN (
    SELECT LibraryBranchId, max(daysCheckedOut) AS maxDaysCheckedOut
    FROM books
    GROUP BY LibraryBranchId
) AS maxbooks
    ON books.LibraryBranchId = maxbooks.LibraryBranchId
    AND books.daysCheckedOut = maxbooks.maxDaysCheckedOut
GROUP BY maxbooks.LibraryBranchId, maxbooks.maxDaysCheckedOut

我认为没有更简单的方法 - 从概念上讲，它是两个集合的交集。 关于分支的元组集合和满足该条件的元组集合。

Answer 2

最简单的方法是获取您的 LibraryBranchId 并选择您的签出日期
然后无论您在何处编写代码，都以编程方式对它们进行计数，并以编程方式获取最大天数。

Answer 3

这个怎么样？

select LibraryBranchId, MAX(daysCheckedOut), count(daysCheckedOut)
from books B
where daysCheckedOut = (select MAX(daysCheckedOut) from books where LibraryBranchID = B.LibraryBranchID)
group by LibraryBranchId 

Answer 4

另一种“浪费”的方式：

select LibraryBranchId, avg(daysCheckedOut) as maxDaysCheckedOut, count(*)
from
(
    select LibraryBranchId, daysCheckedOut
    from books b1
    where not exists 
    (
        select *
        from books b2
        where b2.LibraryBranchId = b1.LibraryBranchId
        and b2.daysCheckedOut > b1.daysCheckedOut
     )
) t
group by LibraryBranchId