根据 SQL Server 中其他列的值在列内进行划分

Question

Having some trouble solving this.

store key  promotion key  sales year  average weekly sales          *Desired Output Column*
1               1           2007             100                        1 (100 / by itself)
1               1           2008             120                        1 (120 / by itself)
1               2           2008             240                        2 (240 / 120)
1               3           2007              50                        0.5 (50 / 100)
1               4           2007              70                        0.7 (70 / 100)
1               5           2007              80                        0.8 (80 / 100)
1               3           2008              180                       1.5 (180 / 120)
2               1           2009              150                       1 (150 / by itself)
2               2           2009              200                       1.3 (200 / 150)
2               3           2007              300                       Null (no baseline value in 2007)

Each store key and sales year has a baseline value indicated by promotion key 1. All other promotion keys reflect entries I want divided into the baseline value for the same store key and sales year.

So in the above example, I want all entries with store key 1 and year 2007 divided by the baseline value for that year and store (ie, promotion key 1 for the same store key of 1 and sales year of 2007). Similarly, all entries with store key 1 and year 2008 divided by the baseline value for that year (promotion key 1 for store 1 and sales year 2008). Likewise for store 2, I want store key 2 and year 2009 for promotion key 2 divided by the baseline value (promotion key 1, store key 2, and year 2009). As there is no baseline value (promotion key 1) for store key 2 in 2007, that entry will produce a null.

Answer 1

I think you want:

select t.*,
       (avgweeklysales  * 1.0 /
        max(case when promotionkey = 1 then avgweeklysales end) over (partition by storekey, salesyear)
       ) as desired_output
from t;

Answer 2

This is a very similar idea to Gordon's, but I used FIRST_VALUE :

WITH YourTable AS(
    SELECT V.StoreKey,
           V.PromotionKey,
           V.SalesYear,
           V.AvgSales
    FROM (VALUES(1,1,2007,100),
                (1,1,2008,120),
                (1,2,2008,240),
                (1,3,2007,50),
                (1,4,2007,70),
                (1,5,2007,80),
                (1,3,2008,180),
                (2,1,2009,150),
                (2,2,2009,200),
                (2,3,2007,300))V(StoreKey,PromotionKey,SalesYear,AvgSales))
SELECT YT.StoreKey,
       YT.PromotionKey,
       YT.SalesYear,
       YT.AvgSales,
       (YT.AvgSales * 1.) / FIRST_VALUE(CASE YT.PromotionKey WHEN 1 THEN YT.AvgSales END) OVER (PARTITION BY YT.StoreKey, YT.SalesYear ORDER BY YT.PromotionKey) AS DesiredColumn
FROM YourTable YT
ORDER BY YT.StoreKey,
         YT.PromotionKey,
         YT.SalesYear;