Optimize query for for cumulative result - sql

Question

I have this query to find out item count for every month of year. But I am looking for a optimized query for cumulative result
SELECT
COUNT(ITM.ID) AS ItemCount,
Month(ITM.ItemProcureDate),
Year(ITM.ItemProcureDate)
FROM
Rpt_Item ITM
WHERE
ITM.ItemProcureDate IS NOT NULL
AND
ITM.ItemStatusID = 2 -- Item sold, Item Rejected
AND
ITM.ItemProcureDate >= CONVERT(DATETIME,'02/01/2014',1) --#Beg_Date
AND
ITM.ItemProcureDate <= CONVERT(DATETIME,'04/12/2014',1) --#End_Date
GROUP BY
Month(ITM.ItemProcureDate),
Year(ITM.ItemProcureDate)
Query result should be like this:
Item sold In month 2
Item Sold Till Month 2
Item Rejected 1
Item Rejected Till Month 1
Year 2014
Month Feb
Last Date of Month 02/28/2014
-----------------------------------------------
Item sold In month 2
Item Sold Till Month 4
Item Rejected 1
Item Rejected Till Month 2
Year 2014
Month March
LastDate of Month 03/31/2014
-----------------------------------------------
Item sold In month 2
Item Sold Till Month 6
Item Rejected 1
Item Rejected Till Month 3
Year 2014
Month April
Last Date of Month 04/30/2014
I have to find out Item_Sold, Item_Rejected, Item_Added for last three months where every next month it should be cumulative of all previous months values of Item_Sold, Item_Rejected, Item_Added

Answer 1

In SQL Server 2008, you can do this using a correlated subquery or using a non-equijoin. SQL Server 2012 supports a cumulative sum function. Here is a way to do it with a correlated subquery:
with ym as (
SELECT COUNT(ITM.ID) AS ItemCount,
Month(ITM.ItemProcureDate) as mon, Year(ITM.ItemProcureDate) as yr,
Month(ITM.ItemProcureDate) + 100*Year(ITM.ItemProcureDate) as yyyymm
FROM Rpt_Item ITM
WHERE ITM.ItemProcureDate IS NOT NULL AND
ITM.ItemStatusID = 2 AND
ITM.ItemProcureDate >= CONVERT(DATETIME,'02/01/2014',1) AND
ITM.ItemProcureDate <= CONVERT(DATETIME,'04/12/2014',1)
GROUP BY Month(ITM.ItemProcureDate), Year(ITM.ItemProcureDate)
)
select ym.*,
(select sum(ItemCount)
from ym ym2
where ym.yyyymm <= ym.yyyy.mm
) as cumsum
from ym;
Note that this puts the year-month into a YYYYMM format. This is just a convenience so the comparison on the time period uses only one column.
Also, if the ITM table is really big or is a view, then this might not perform as well as one would like. If performance is an issue, use a temporary table instead of a CTE. (SQL Server tends not to materialize CTEs so it is likely to run the code twice.)