一种将嵌套排名应用于几列 SUM 的更优雅的方法

Question

是否有更优雅的方法来对几个不同的列应用嵌套排名，先按最左边列的总计排序，然后再按该组内的总计排序？我提供了关于示例数据结构和所需状态的屏幕截图。

现在，我正在使用三个不同的查询（首先按总计对地区进行排名。然后按州总数排名第二，等等，第三按城市排名）维护起来很痛苦，特别是如果我需要修改排名的顺序以及日期window。（考虑按月计算的销售额、按 YTD 计算的销售额等）

使用 RANK 和 windows 函数和分区，我可以很容易地获得区域内城市排名，但要同时获得州排名和地区排名，我开始挣扎。

对于那些好奇的人，我在 SQL 而不是 Tableau 中进行排名，因为几乎不可能将排名与使用 Tableau window 函数设置的嵌套 Top N + Other 混合使用或细节层次。

CREATE TABLE RankingTest (
Region NVarChar(15),
[State] NVarChar(2),
City NVarChar(50),
Sales Int
);

INSERT INTO RankingTest (Region, [state], city, Sales) Values ('East','NY','New York',32);
INSERT INTO RankingTest (Region, [state], city, Sales) Values ('East','NY','New York',3);
INSERT INTO RankingTest (Region, [state], city, Sales) Values ('East','NY','Queens',4);
INSERT INTO RankingTest (Region, [state], city, Sales) Values ('East','MA','Boston',6);
INSERT INTO RankingTest (Region, [state], city, Sales) Values ('East','FL','Miama',7);
INSERT INTO RankingTest (Region, [state], city, Sales) Values ('West','WA','Seattle',13);
INSERT INTO RankingTest (Region, [state], city, Sales) Values ('West','WA','Bellevue',6);
INSERT INTO RankingTest (Region, [state], city, Sales) Values ('West','WA','Kirkland',8);
INSERT INTO RankingTest (Region, [state], city, Sales) Values ('West','OR','Portland',11);
INSERT INTO RankingTest (Region, [state], city, Sales) Values ('West','OR','Salem',3);
INSERT INTO RankingTest (Region, [state], city, Sales) Values ('West','CA','San Franscisco',6);
INSERT INTO RankingTest (Region, [state], city, Sales) Values ('West','CA','San Franscisco',9);

当前查询：

    WITH RegionRank AS (
        SELECT Region,
            DENSE_RANK() OVER (ORDER By SUM(Sales) DESC) AS Rank,
            SUM(Sales) AS Sales
        FROM RankingTest
        GROUP BY Region
    ),
    StateRank AS (
        SELECT RT.Region, State,
            DENSE_RANK() OVER (PARTITION BY RR.Rank ORDER BY RR.Rank, SUM(RT.Sales) DESC) AS Rank,
            SUM(RT.Sales) AS Sales
        FROM RankingTest RT LEFT JOIN RegionRank RR ON RT.Region = RR.Region
        GROUP BY RT.Region, RR.Rank, State
    ),
    CityRank AS (
        SELECT RT.Region, RR.Rank RegionRank, RR.Sales RegionSales, RT.State, 
        SR.Rank StateRank, SR.Sales StateSales, City,
        DENSE_RANK() OVER (PARTITION BY RR.Rank, SR.Rank ORDER BY RR.Rank, SR.Rank, SUM(RT.Sales) DESC) AS 
        Rank,
            SUM(RT.Sales) AS Sales
        FROM RankingTest RT
           LEFT JOIN RegionRank RR ON RT.Region = RR.Region
             LEFT JOIN StateRank SR ON RT.Region = SR.Region AND RT.State = SR.State
        GROUP BY RT.Region, RR.Rank, RR.Sales, RT.State, SR.Rank, SR.Sales, City
     )
     SELECT * FROM CityRank

SQL Fiddle

Answer 1

嵌套查询应该可以解决问题：

SELECT *,
    DENSE_RANK() OVER( ORDER BY SalesRegion DESC ) AS RankRegion,
    DENSE_RANK() OVER( PARTITION BY Region ORDER BY SalesState DESC ) AS RankState,
    DENSE_RANK() OVER( PARTITION BY Region, State ORDER BY SalesCity DESC ) AS RankCity
FROM(
    SELECT DISTINCT Region, State, City,
        SUM( Sales ) OVER( PARTITION BY city, State, Region ) as SalesCity,
        SUM( Sales ) OVER( PARTITION BY State, Region ) as SalesState,
        SUM( Sales ) OVER( PARTITION BY Region ) as SalesRegion
    FROM RankingTest ) AS DistinctSales
ORDER BY RankRegion, RankState, RankCity

请注意，DISTINCT Region, State, City 在这里充当 GROUP BY。

一种将嵌套排名应用于几列 SUM 的更优雅的方法

A more elegant way to apply Nested Ranking over SUM of a few columns

tsql

sql-server

window-functions

ranking-functions

sql-server-2014