Question

Long time reader, first time poster. Hoping some SQL guru can lend me some processing power to solve a SQL problem elegantly and without cursors.

I am trying to create a time line type result set from a data set with money and a Date, where the order in which they appear is important. See sample data and expectation below.

DECLARE @OrderData TABLE(
      ID                      INT IDENTITY,
      ProductId               INT,
      WarehouseId             INT,
      Cost MONEY,
      SaleDate          DATETIME
)

 

INSERT INTO @OrderData
VALUES
      (1, 1, 2.71, '2012-02-23 10:01')
      ,(1, 2, 2.71, '2012-02-23 10:02')
      ,(1, 1, 2.71, '2012-02-23 10:03')
      ,(1, 1, 2.71, '2012-02-23 10:04')
      ,(1, 1, 2.71, '2012-02-23 10:05')
      ,(1, 1, 2.8, '2012-02-23 10:06')
      ,(1, 1, 2.9, '2012-02-23 10:07')
      ,(1, 1, 2.71, '2012-02-23 10:08')
      ,(1, 1, 2.71, '2012-02-23 10:09')

The result I am looking for is a time line of sales with from and to dates for a product and warehouse combination within it own set of sales. As soon as the sale price changes in the dataset this should spawn a new row in the result set.

Result structure for sample data.

Product, warehouse, minsale date, Max sale date
1,1,2.71, 2012-02-23 10.01, 2012-02-23 10.05
1,1,2.80, 2012-02-23 10.06, 2012-02-23 10.06
1,1,2.90, 2012-02-23 10.07, 2012-02-23 10.07
1,1,2.71, 2012-02-23 10.08, 2012-02-23 10.09
1,2,2.71, 2012-02-23 10.02, 2012-02-23 10.02

Cheers for any help available.

Answer 1

This is a “gaps and islands” problem.

select ProductID, WarehouseId, cost, MIN(saledate), MAX(saledate) from
(      
    select *,
        ROW_NUMBER() over (order by saledate)-
        ROW_NUMBER() over (partition by cost order by saledate) grp
    from @OrderData
) v
group by ProductID, WarehouseId, grp, cost
order by WarehouseID, MIN(SaleDate)

For further information, search for “gaps and islands by itzik ben-gan”