SQL Query using Cross Apply to get Sum Conditionally

问题

Output to be produced

using this as reference but now with different scenario SQL Server query : get the sum conditionally

explanation:

Item, Sales, and remarks columns are given column from a database, New Sales column is a formulated column where in, it is getting the sum of items with the same keyword remarks except the default n/a remarks.

(regardless the remarks is not fully identical, at least there's a common similarity like what is on the image above - item 5 has "new" in it, still it sums up with item 6 because of their similar keyword found "small")

code used

FIRST OPTION- using partition - This doesn't work because when the remarks is not identical to each other it will not get the sum properly (for item5 and item6)

 CASE
     WHEN ([remarks] not  like '%big%') AND ([remarks] not  like '%PAENGS%') 


     THEN sales 
     ELSE SUM(sales) OVER(PARTITION BY [remarks])
     END as 'New Sales'

SECOND OPTION -using Cross Apply - So it leave me to this, but I was lost as it is not getting the desired output.

 CROSS APPLY
     SELECT
        d.*, 
        NewSales = 
        CASE
        WHEN ([remarks] not like '%big%') or ([remarks] not like '%small%')
        THEN Sales 
        ELSE x.NewSales
        END


        FROM #MSRSuperFinal3  d
        CROSS APPLY(SELECT NewSales = SUM(Sales)
                    FROM #MSRSuperFinal3 
                    WHERE ([remarks] like  '%big%') or ([remarks] like  '%small%')
        )x

Any help will be highly appreciated

回答1:

Using CROSS APPLY

SELECT *
FROM temp t
CROSS APPLY(
    SELECT SUM(sales)
    FROM temp
    WHERE
        remarks LIKE '%' + t.remarks + '%'
        OR t.remarks LIKE '%' + remarks + '%'
)x(NewSales)
WHERE remarks <> 'n/a'

UNION ALL

SELECT *, 
    NewSales = sales
FROM temp
WHERE remarks = 'n/a'
ORDER BY item

Based on your comment, this should be your final query:

SELECT * 
FROM #MSRSuperFinal3 t
CROSS APPLY( 
    SELECT 
        SUM(CurrentMonth)
    FROM #MSRSuperFinal3 
    WHERE 
        t.misc LIKE '%' + misc + '%'
        OR misc LIKE '%' + t.misc + '%'
)x(NewSales) 
WHERE 
    ([misc] LIKE '%BIGKAHUNA%') 
    or ([misc] LIKE '%PAENGS%') 

UNION ALL 
SELECT *, 
    NewSales = CurrentMonth 
FROM #MSRSuperFinal3 
WHERE 
    ([misc] not like '%BIGKAHUNA%')
    AND ([misc] not like '%PAENGS%')
    AND ([misc] not like '%ROBINSONS%')
ORDER BY location, name

回答2:

Try Left Join clause instead of Cross Apply:

SELECT a.item,
       a.sales,
       a.remarks,
       CASE
           WHEN a.remarks = 'n/a' 
           THEN a.sales
           ELSE SUM( b.sales )
       END AS [New Sales]
FROM query_using_cross_apply_to_get_sum_conditionally a
LEFT JOIN query_using_cross_apply_to_get_sum_conditionally b 
      ON(a.remarks LIKE '%' + b.remarks + '%'
         AND b.remarks != 'n/a')
GROUP BY a.item, a.sales, a.remarks;

回答3:

CASE
    WHEN (a.[remarks] <> 'n/a') THEN a.sales 
    ELSE 
    (
        SELECT SUM(b.sales) 
        FROM #MSRSuperFinal3 b
        WHERE b.[remark] LIKE '%' + a.[remarks] + '%'
    )
END as 'New Sales'

Note that i just added a table alias so that the query knows which remark to use where. Otherwise something that might be faster is doing a Common table expression and firstly doing a sum for sales that contain each other, so for each unique remark sum all sales that contain in, and then in your last select you cna merely join on your common table expression table.

来源：https://stackoverflow.com/questions/31157651/sql-query-using-cross-apply-to-get-sum-conditionally

标签

sql

sql-server

sql-server-2008

cross-apply