How to get distinct closest date in SQL Server?

Question

I don't know how to format?!, but I believe it's easy to understand. I have the following table, lets call it "sales"

|Item|        |Price|        |PriceDate|  
ItemA          801.36         09/23/2011  
ItemA          800.64         09/23/2011  
ItemA          803.55         09/22/2011  
ItemB          4701.36         09/22/2011  
ItemB          1101.36         09/22/2011  
ItemB          4801.36         09/20/2011  
ItemB          401.36         09/22/2011    
ItemC          9601.36         09/21/2011  
ItemC          201.36         09/19/2011  
ItemC          301.36         09/17/2011  

I'm given a date and I need to retrieve the records with the closest date, and only those, for example, if 09/24/2011 is the input, the output should be only the records from the 23rd for item A, 22nd for itemB, and 21st for itemC.

Using SQL Server 2012.

Answer 1

One way to go about it is to calculate the difference between the row's data and the given date by using datediff , assign a rank to each row accordingly and filter by it.

Here I'm using ? as a placeholder for the required date, just switch it with the correct syntax of the language you're using:

SELECT item, price, pricedate
FROM   (SELECT item, price, pricedate,
               RANK() OVER (PARTITION BY item
                            ORDER BY ABS(DATEDIFF(day, pricedate, ?))) AS rk
        FROM   salse) t
WHERE  rk = 1

Answer 2

DECLARE @theTable TABLE (Item VARCHAR(10), price DECIMAL(10,2), priceDate DATE)
INSERT @theTable ( Item, price, priceDate )
VALUES  
('ItemA',801.36,'2011-09-23'),
('ItemA',800.64,'2011-09-23'),
('ItemA',803.55,'2011-09-22'),
('ItemB',4701.36,'2011-09-22'),
('ItemB',1101.36,'2011-09-22'),
('ItemB',4801.36,'2011-09-20'),
('ItemB',401.36,'2011-09-22'),
('ItemC',9601.36,'2011-09-21'),
('ItemC',201.36,'2011-09-19'),
('ItemC',301.36,'2011-09-17')

DECLARE @inputDate DATE
SET @inputDate = '2011-09-24'

SELECT X.Item, X.price, X.priceDate FROM (
SELECT TT.Item, TT.price, TT.priceDate, 
       RANK() OVER (PARTITION BY [Item] 
                          ORDER BY ABS(DATEDIFF(DAY, @inputDate, TT.priceDate))) AS RN
FROM @theTable TT 
) AS X
WHERE RN = 1

---

 (10 row(s) affected) Item price priceDate ---------- --------------------------------------- ---------- ItemA 801.36 2011-09-23 ItemA 800.64 2011-09-23 ItemB 4701.36 2011-09-22 ItemB 1101.36 2011-09-22 ItemB 401.36 2011-09-22 ItemC 9601.36 2011-09-21

Answer 3

i have the same problem but i found that i can solve it on front end by easier way

/*back end*/
/*(DistinctItem)   select distinct Item from sales*/
/*(saleRecords)    select * from sales order by PriceDate desc*/

/*front end*/
for(var j = 0; j < DistinctItem.length;j++){
  for(var i = 0; i < saleRecords.length;i++){
    if(DistinctItem[j].Item==saleRecords[i].Item){
      if(anyDate>saleRecords[i].PriceDate){
        use saleRecords[i];
        break;
      }
    }
  }
}