选择客户，每个地区的最高总订单，旁边两个总计

Question

我有这样的一个表：

SaleID Region  Customer  OrderAmt 
1   North  Keesha   10 
2   West  Mary   10 
3   North  Winston   10 
4   North  John   10 
5   North  Keesha   10 
6   West  John   10 
7   West  Mary   10 
8   South  John   10 

使用SQL Server 2012中，什么是选择每个地区的最高排序的客户的最佳方式，沿着为客户和区域，即总计：

Region Customer CustAmt RegAmt 
North  Keesha  20  40 
West  Mary  20  30 
South  John  10  10 

尽管多个区域可能包含相同的名称，我们要CustAmt成为总该名称仅在每个区域内，不能跨区域（即，在最后一行，约翰在南部地区的总数是10，n ot 30）。

Answer 1

的CTE：

WITH cte AS (
    SELECT Region, Customer 
     , sum(OrderAmt) AS CustAmt 
     , sum(sum(OrderAmt)) OVER (PARTITION BY Region) AS RegAmt 
     , row_number() OVER (PARTITION BY Region ORDER BY sum(OrderAmt) DESC) AS rn 
    FROM tbl 
    GROUP BY Region, Customer 
    ) 
SELECT Region, Customer, CustAmt, RegAmt 
FROM cte 
WHERE rn = 1; 

或者使用子查询相同：

SELECT Region, Customer, CustAmt, RegAmt 
FROM (
    SELECT Region, Customer 
     , sum(OrderAmt) AS CustAmt 
     , sum(sum(OrderAmt)) OVER (PARTITION BY Region) AS RegAmt 
     , row_number() OVER (PARTITION BY Region ORDER BY sum(OrderAmt) DESC) AS rn 
    FROM tbl 
    GROUP BY Region, Customer 
    ) sub 
WHERE rn = 1; 

关键特性是聚合的窗口函数，所以我们只需要一个单个的 CTE /子查询。

SQL Fiddle.

Answer 2

您可以使用多个CTE S和窗口功能：

WITH Cte AS(
    SELECT *, 
     CustAmt = SUM(OrderAmt) OVER(PARTITION BY Customer, Region), 
     RegAmt = SUM(OrderAmt) OVER(PARTITION BY Region) 
    FROM Sales 
), 
CteFinal AS(
    SELECT *, 
     rn = ROW_NUMBER() OVER(PARTITION BY Region ORDER BY CustAmt DESC) 
    FROM Cte 
) 
SELECT 
    Region, Customer, CustAmt, RegAmt 
FROM CteFinal WHERE rn = 1 

SQL Fiddle

Answer 3

在一个选择：

select top 1 with ties 
    Region , 
    Customer , 
    CustAmt = sum(OrderAmt), 
    RegAmt = sum(sum(OrderAmt)) over (partition by Region) 
from 
    your_table 
group by  
    Region, 
    Customer 
order by 
    row_number() over(partition by Region order by sum(OrderAmt) desc); 

SQL Fiddle