Sql grouping and summing impossible?

Question

I want to create an sql query to sum up some data but i'm starting to think it is impossible with sql alone. The data i have are of the following form :
TRAN_DT     TRAN_RS     DEBT     CRED
10-Jan     701     100     0
20-Jan     701     150     0
21-Jan     701     250     0
22-Jan     705     0     500
23-Jan     571     100     0
24-Jan     571     50     0
25-Jan     701     50     0
26-Jan     701     20     0
27-Jan     705     0     300The data are ordered by TRAN_DT and then by TRAN_RS. Tha grouping and summing of data based on tran_rs but only when it changes. So in the table above i do not want to see all 3 first recods but only one with value DEBT the sum of those 3 i.e. 100+150+250=500. So the above table after grouping would be like the one below:
TRAN_DT     TRAN_RS     DEBT     CRED
21-Jan     701     500     0
22-Jan     705     0     500
24-Jan     571     150     0
26-Jan     701     70     0
27-Jan     705     0     300The TRAN_DT is the last value of the summed records. I undestand that the tran_dt may not be selectable. What i have tried so far is the following query:
select tran_dt,
         tran_rs,
         sum(debt)over(partition by tran_rs order by tran_dt rows unbounded preceding),
         sum(cred)over(partition by tran_rs order by tran_dt rows unbounded preceding) from that_tableIs this even possible with sql alone, any thoughts?
The report i am trying to create in BI Publisher.Maybe it is possible to group the data in the template and ask my question there?

Answer 1

915218 wrote:
Is this even possible with sql alone, any thoughts?It sure is...
WITH that_table as (select to_date('10/01/2012', 'dd/mm/yyyy') tran_dt, 701 tran_rs, 100 debt, 0 cred from dual union all
                     select to_date('20/01/2012', 'dd/mm/yyyy') tran_dt, 701 tran_rs, 150 debt, 0 cred from dual union all
                     select to_date('21/01/2012', 'dd/mm/yyyy') tran_dt, 701 tran_rs, 250 debt, 0 cred from dual union all
                     select to_date('22/01/2012', 'dd/mm/yyyy') tran_dt, 705 tran_rs, 0 debt, 500 cred from dual union all
                     select to_date('23/01/2012', 'dd/mm/yyyy') tran_dt, 571 tran_rs, 100 debt, 0 cred from dual union all
                     select to_date('24/01/2012', 'dd/mm/yyyy') tran_dt, 571 tran_rs, 50 debt, 0 cred from dual union all
                     select to_date('25/01/2012', 'dd/mm/yyyy') tran_dt, 701 tran_rs, 50 debt, 0 cred from dual union all
                     select to_date('26/01/2012', 'dd/mm/yyyy') tran_dt, 701 tran_rs, 20 debt, 0 cred from dual union all
                     select to_date('27/01/2012', 'dd/mm/yyyy') tran_dt, 705 tran_rs, 0 debt, 300 cred from dual)
, brk AS (
SELECT     tran_dt,
        tran_rs,
     debt,
     cred,
        CASE WHEN Nvl (Lag (tran_rs) OVER (ORDER BY tran_dt, tran_rs), 0) != tran_rs THEN tran_rs || tran_dt END brk_tran_rs
  FROM that_table
), grp AS (
SELECT     tran_dt,
        tran_rs,
         debt,
         cred,
        Last_Value (brk_tran_rs IGNORE NULLS) OVER  (ORDER BY tran_dt, tran_rs) grp_tran_rs
  FROM brk
SELECT     Max (tran_dt),
        Max (tran_rs),
         Sum (debt),
         Sum (cred)
  FROM grp
GROUP BY grp_tran_rs
ORDER BY 1, 2
Boneist
MAX(TRAN_    TRAN_RS       DEBT       CRED
21-JAN-12        701        500          0
22-JAN-12        705          0        500
24-JAN-12        571        150          0
26-JAN-12        701         70          0
27-JAN-12        705          0        300
Me
MAX(TRAN_ MAX(TRAN_RS)  SUM(DEBT)  SUM(CRED)
21-JAN-12          701        500          0
22-JAN-12          705          0        500
24-JAN-12          571        150          0
26-JAN-12          701         70          0
27-JAN-12          705          0        300Edited by: BrendanP on 17-Feb-2012 04:05
Test data courtesy of Boneist, and fixed bug.
Edited by: BrendanP on 17-Feb-2012 04:29