Question

EDIT

Here is an example of my data:

 create TABLE #Table 
(
    [Market] VARCHAR(100),
    [Operator] VARCHAR(100),
    [Date] DATE,
    [Num Registrations] INT,
    [Num Active] INT,
    [Num Accepted Purchases] INT
)

INSERT INTO #Table VALUES
('Market1','Operator1','2012-12-01',2,4,7),
('Market1','Operator2','2012-12-01',3,5,7),
('Market1','Operator3','2012-12-01',1,2,7),
('Market2','Operator4','2012-12-01',2,1,7),
('Market2','Operator5','2012-12-01',0,4,7),
('Market3','Operator6','2012-12-01',2,44,7)

I have the following script:

SELECT 
    a.[Market],
    a.[Operator],
    a.[Date],
    a.[Num Registrations],
    a.[Num Active],
    a.[Num Accepted Purchases],
    [rnk] = b.rnk
FROM 
    #Table a
    INNER JOIN 
        (
        SELECT 
            [Market],
            [rnk] = RANK() OVER (ORDER BY SUM([Num Registrations] + [Num Active]))
        FROM #Table
        GROUP BY [Market]
        ) b
        ON 
          a.[Market] = b.[Market]
GROUP BY
    a.[Market],
    a.[Operator],
    a.[Date],
    a.[Num Registrations],
    a.[Num Active],
    a.[Num Accepted Purchases],
    b.rnk

The above works ok but have I over-complicated things with the sub-query; could I have used the RANK function directly in the main query’s SELECT clause?

Answer 1

Yes, you don’t need to do an explicit join at all:

SELECT distinct
    a.[Market],
    a.[Operator],
    a.[Date],
    a.[Num Registrations],
    a.[Num Active],
    a.[Num Accepted Purchases],
    [rnk] = dense_rank() over (order by tot, market)
FROM (select a.*,
             sum(a.[Num Registrations] + a.[Num Active]) over (partition by market) as tot
      from xxx.dbo.tb_r12044dxx_yyyy a
     ) a

If you don’t have duplicates, then you can remove the distinct from the select.