How to get most recent consecutive records with the same value

Question

Hi,
Can someone please help me to solve my problem?
Below is my table
Prod_Code-----Purchase_date---Inv_number-------Inv_amt
MS_AV16850------18/01/2005-----------6575947----------------7.93
MS_AV16850------22/07/2005-----------7034012----------------51.82
MS_AV16850------04/01/2006-----------8719618----------------51.82
MS_AV16850------20/06/2006-----------9515864----------------104.69
MS_AV16850------16/04/2007-----------10353759----------------189.29
MS_AV16850------30/05/2007-----------10689899----------------189.29
MS_AV16850------06/01/2008-----------1653821----------------65.49
MS_AV16850------22/02/2009-----------10866311----------------189.29
I want my query to show the rows that has most recent purchase dates with same amount in consecutive purchase date.
So from the table above, the query should display:
Prod_Code-----Purchase_date---Inv_number-------Inv_amt
MS_AV16850------16/04/2007-----------10353759----------------189.29
MS_AV16850------30/05/2007-----------10689899----------------189.29
It should not get
MS_AV16850------16/04/2007-----------10353759----------------189.29
MS_AV16850------30/05/2007-----------10689899----------------189.29
MS_AV16850------22/02/2009-----------10866311----------------189.29
because inv_number 10866311 has a prvevious inv_amount of 65.49.
and not get this
MS_AV16850------22/07/2005-----------7034012----------------51.82
MS_AV16850------04/01/2006-----------8719618----------------51.82
because they are not the most recent purchase date even if they have the same inv_amount.
Thanks in advance.

Answer 1

Hi,
You're right; thanks for catching my mistake.
I changed the WHERE clause of the main query (including subquery there) to deal with that situation:
WITH     got_grp          AS
     SELECT      x.*
     ,      ROW_NUMBER () OVER ( ORDER BY      purchase_date )
            - ROW_NUMBER () OVER ( PARTITION BY  inv_amt
                                         ORDER BY         purchase_date )     AS grp
--     ,      ROW_NUMBER () OVER ( ORDER BY      purchase_date )     AS r1
--     ,        ROW_NUMBER () OVER ( PARTITION BY  inv_amt
--                                         ORDER BY         purchase_date )     AS r2
     FROM     table_x     x
SELECT     *     -- Or list all columns except grp
FROM     got_grp
WHERE     (inv_amt, grp)  IN  (
                           SELECT    MAX (inv_amt) KEEP (DENSE_RANK LAST ORDER BY MAX (purchase_date))
                    ,       MAX (MAX (grp))
                            FROM      got_grp
                            GROUP BY  grp
                            ,       inv_amt
                            HAVING    COUNT (*)     > 1
;Thanks, too, for posting the sample data. Apparantly, you're more interested in solving this problem than OP is.
Edited by: Frank Kulash on Nov 22, 2010 1:36 PM
The r1 and r2 columns are not needed for the solution. You may want to display them, just to help see how the query works.