Question

I have following code

SELECT MRT.sno, MRT.TypeName, MR.Adate, MAX(MRD.Value) AS Value
FROM MeterReadings MR 
INNER JOIN MeterReadingDetails MRD ON MRD.ReadingId = MR.sno
INNER JOIN MeterReadingTypes MRT ON MRT.sno = MRD.ReadingTypeId
WHERE MRT.sno IN (7,10,11)
GROUP BY MRT.sno,MRT.TypeName,MR.Adate
ORDER BY MR.Adate DESC

And result

sno TypeName                Adate                   Value

11  Toplam Kapasitif        2013-01-04 00:00:00     33,313
7   Toplam                  2013-01-04 00:00:00     7819,33
10  Toplam Reaktif          2013-01-04 00:00:00     640,492
11  Toplam Kapasitif        2013-01-03 00:00:00     33,276
7   Toplam                  2013-01-03 00:00:00     7805,934
10  Toplam Reaktif          2013-01-03 00:00:00     639,862

I want an additional column that named “OldValue”. OldValue Column shows previous day value like following (last three row for above example)

33,276
7805,934
639,862
null
null
null

How Can I do this? There may be a similar example to show me the way.

Thanks in advance…

UPDATE

In fact, I wrote something like this

SELECT MRT.sno, MRT.TypeName, MR.Adate, MAX(MRD.Value) AS Value, 

(SELECT Value From
    (SELECT TOP 1 XMR.Adate,XMRD.ReadingTypeId,MAX(XMRD.Value) AS Value 
    FROM MeterReadings XMR,MeterReadingDetails XMRD 
    WHERE XMRD.ReadingId = XMR.sno AND XMRD.ReadingTypeId = MRT.sno AND XMR.Adate<MR.Adate 
    GROUP BY XMR.Adate,XMRD.ReadingTypeId 
    ORDER BY XMR.Adate DESC) AS TBL
) AS OldValue

FROM MeterReadings MR 
INNER JOIN MeterReadingDetails MRD ON MRD.ReadingId = MR.sno
INNER JOIN MeterReadingTypes MRT ON MRT.sno = MRD.ReadingTypeId
WHERE MRT.sno IN (7,10,11)
GROUP BY MRT.sno,MRT.TypeName,MR.Adate

But I dont know, Is this best way?

Answer 1

If you’r using SQL Server 2012, you could use the new LAG analytical function

SQL Server 2012 introduces new analytical function LEAD() and LAG().
This functions accesses data from a subsequent row (for lead) and
previous row (for lag) in the same result set without the use of a
self-join .

Your statement then becomes

SELECT  *, LAG(Value) OVER (PARTITION BY sno ORDER BY Adate DESC)
FROM    (
          SELECT  MRT.sno, MRT.TypeName, MR.Adate, MAX(MRD.Value) AS Value, 
          FROM    MeterReadings MR 
                  INNER JOIN MeterReadingDetails MRD ON MRD.ReadingId = MR.sno
                  INNER JOIN MeterReadingTypes MRT ON MRT.sno = MRD.ReadingTypeId
          WHERE   MRT.sno IN (7,10,11)
          GROUP BY
                  MRT.sno,MRT.TypeName,MR.Adate
        ) q
ORDER BY
        Adate DESC

Using SQL Server 2005/2008, I would write the statement as

;WITH q AS (
  SELECT  MRT.sno, MRT.TypeName, MR.Adate, MAX(MRD.Value) AS Value, rn = ROW_NUMBER() OVER (PARTITION BY MRT.sno ORDER BY MR.Adate DESC)
  FROM    MeterReadings MR 
          INNER JOIN MeterReadingDetails MRD ON MRD.ReadingId = MR.sno
          INNER JOIN MeterReadingTypes MRT ON MRT.sno = MRD.ReadingTypeId
  WHERE   MRT.sno IN (7,10,11)
  GROUP BY
          MRT.sno,MRT.TypeName,MR.Adate
)
SELECT  q1.*, q2.Value
FROM    q q1
        LEFT OUTER JOIN q q2 ON q2.sno = q1.sno AND q2.rn = q1.rn + 1

Edit

The main difference between using either one of these solutions to yours is that your solution has to retrieve the previous result for each record, an expensive operation, while this solutions in essence can join two complete (identical) sets of data.