Help with over by clause in query.

Question

guys i have a table like this
create table fgbtrnh ( fgbtrnh_doc_code varchar2(9),
                                         fgbtrnh_trans_date date,
                                         fgbtrnh_trans_amt number(17,2),
                                         fgbtrnh_acct varchar2(6) ,
                                         fgbtrnh_fund_code varchar2(6),
                                         fgbtrnh_rucl_code varchar2(6) );
with data like this.
   insert into fgbtrnh( fgbtrnh_doc_code,fgbtrnh_trans_date,fgbtrnh_trans_amt,fgbtrnh_acct,fgbtrnh_rucl_code,fgbtrnh_fund_code) values('J0005445','31-MAY-10','38491','6001','BD01','360098');
insert into fgbtrnh( fgbtrnh_doc_code,fgbtrnh_trans_date,fgbtrnh_trans_amt,fgbtrnh_acct,fgbtrnh_rucl_code,fgbtrnh_fund_code) values('L0000005','01-JUL-08','38260','6001','BD01','360098');
insert into fgbtrnh( fgbtrnh_doc_code,fgbtrnh_trans_date,fgbtrnh_trans_amt,fgbtrnh_acct,fgbtrnh_rucl_code,fgbtrnh_fund_code) values('L0000002','30-JUN-08','24425.29','6001','BD01','360125');
insert into fgbtrnh( fgbtrnh_doc_code,fgbtrnh_trans_date,fgbtrnh_trans_amt,fgbtrnh_acct,fgbtrnh_rucl_code,fgbtrnh_fund_code) values('L0000002','30-JUN-08','48057.71','6001','BD01','360125');
insert into fgbtrnh( fgbtrnh_doc_code,fgbtrnh_trans_date,fgbtrnh_trans_amt,fgbtrnh_acct,fgbtrnh_rucl_code,fgbtrnh_fund_code) values('M0000002','30-JUN-08','90','7200','BD01','360098');i would like to get a running total of these items so i tried something like this.
select  f.fgbtrnh_doc_code,f.fgbtrnh_trans_date,f.fgbtrnh_trans_amt, sum(f.fgbtrnh_trans_amt)
over
(--partition by  f.fgbtrnh_doc_code
order by fgbtrnh_trans_date desc  ROWS UNBOUNDED PRECEDING
--group by  f.fgbtrnh_doc_code
)total
From fgbtrnh f
where f.fgbtrnh_fund_code in ('360098', '360125')
and f.fgbtrnh_rucl_code = 'BD01'
and f.fgbtrnh_acct = '6001'
order by  f.fgbtrnh_trans_date desc,  f.fgbtrnh_doc_codebut i end up with a result set like
"FGBTRNH_DOC_CODE", "FGBTRNH_TRANS_DATE", "FGBTRNH_TRANS_AMT", "TOTAL"
"J0005445", 31-MAY-10, 38491, 38491
"L0000005", 01-JUL-08, 38260, 76751
"L0000002", 30-JUN-08, 24425.29, 101176.29
"L0000002", 30-JUN-08, 48057.71, 149234
i would like to thave the running total to start from the bottom in other word is my total column i would like to end up with the 149234 at the top
so it would look something like so.
"FGBTRNH_DOC_CODE", "FGBTRNH_TRANS_DATE", "FGBTRNH_TRANS_AMT", "TOTAL"
"J0005445", 31-MAY-10, 38491, 149234
"L0000005", 01-JUL-08, 38260, 110743
"L0000002", 30-JUN-08, 24425.29, 72483
"L0000002", 30-JUN-08, 48057.71, 48057.71
i have tried everything and just cant seem to make this work can someone please point me in the rigth direction.
I would really appreciate the help.
Thanks
Miguel

Answer 1

Hi, Miguel,
mlov83 wrote:
... Also, if you uniquely order the rows, you won't need the windowing clause ("ROWS UNBOUNDED PRECEEDING"); the default ("RANGE UNBOUNDED PRECEDING") will produce exactly what you want, so you don;'t need to say it.
I dont really understand what you mean by this ? but if i take a gander are you saying that all my rows would have to be unique and then i wont have to use ("ROWS UNBOUNDED PRECEEDING")I think you got it right.
The analytic ORDER BY clause doesn't have to result in a unique ordering; there are good reasons for having a unique oprdering, and there are good reasons for not having a unique ordering.
I'm saying that if the analytic ORDER BY is unique, then you don't need to give a widnowing clause, such as "ROWS UNBOUNDED PRECEEDING".
Frank sorry if im asking some really stupid questions but i have tried and tried to read and understand "partion by" and "over" work but im not quite sure I understand yet. It's not stupid at all! Analytic functions can be very subtle and confusing.
Let's use a query based on the scott.emp table, which has seveal rows for each deptno.
-- Tell SQL*Plus to put a blank line between deptnos, just to make the output easier to read
BREAK     ON deptno   SKIP 1     DUPLICATES
SELECT       deptno
,       ename
,       sal
,       SUM (sal) OVER ( PARTITION BY  deptno
                            ORDER BY        sal
                  ROWS            UNBOUNDED PRECEDING
                )     AS running_total
FROM       scott.emp
ORDER BY  deptno
,            sal          DESC
,       ename
;Output:
`   DEPTNO ENAME             SAL RUNNING_TOTAL
        10 KING             5000          8750
        10 CLARK            2450          3750
        10 MILLER           1300          1300
        20 FORD             3000         10875
        20 SCOTT            3000          7875
        20 JONES            2975          4875
        20 ADAMS            1100          1900
        20 SMITH             800           800
        30 BLAKE            2850          9400
        30 ALLEN            1600          6550
        30 TURNER           1500          4950
        30 MARTIN           1250          2200
        30 WARD             1250          3450
        30 JAMES             950           950PARTITION BY deptno" means do a separate calculation for each distinct value of deptno. Rows with deptno=10 don't effect the results on rows where deptno=20 or deptno=30. Since there are 3 distinct values of deptno, there are 3 distinct running totals.
Notice that the aNalytic ORDER BY clause results only in a partial ordering. If there are two or more rows in the same deptno that happen to have the same sal, look what can happen:
{code}
` DEPTNO ENAME SAL RUNNING_TOTAL
... 30 TURNER 1500 4950
30 MARTIN 1250 2200
30 WARD 1250 3450
30 JAMES 950 950
{code}
MARTIN and WARD are in the same partition (deptno=30), and they both have the same sal (1250), so there is no reason why one of those rows would be considered "before" the other one. When you use a windowing clause based on ROWS, as above, and there is a tie for whcih row comes first (as there is a tie between MARTIN and WARD), then one of the rows will arbitrarily be condidered to be before the other one. In this example, it happened to chose MARTIN as the 2nd lowest sal, so running_total=2200 (= 950 + 1250) on the row for MARTIN, and running_total=3450 ( = 950 + 1250 + 1250) on the row for WARD. There's no particular reason for that; it's completely arbitrary. I might do the exact same query tomorrow, or in 10 minutes, and get running_total=2200 on WARD's row, and 3450 on MARTIN's.
However, it is no accident that MARTIN comes before WARD in the output; the *query* ORDER BY clause (which has nothing to do with the analytic ORDER BY clause) guarantees that, when two rows have the same deptno and sal, then the one with the earlier ename will come first.
Now, what's the difference between a window based on ROWS and a window bnased on RANGE?
One difference is that, when a tie occurs in the ORDER BY clause, all rows with the same value of sal get the same value for SUM (sal):
{code}
SELECT     deptno
,     ename
,     sal
,     SUM (sal) OVER ( PARTITION BY deptno
               ORDER BY      sal
               )     AS running_total
FROM     scott.emp
ORDER BY deptno
,      sal          DESC
,     ename
{code}
Notice that the only difference between the first query above and this one is that this one does not have an analytic windowing clause, so the default window, *RANGE* UNBOUNDED PRECEDING is used.
Output:
{code}
` DEPTNO ENAME SAL RUNNING_TOTAL
10 KING 5000 8750
10 CLARK 2450 3750
10 MILLER 1300 1300
20 FORD 3000 10875
20 SCOTT 3000 10875
20 JONES 2975 4875
20 ADAMS 1100 1900
20 SMITH 800 800
30 BLAKE 2850 9400
30 ALLEN 1600 6550
30 TURNER 1500 4950
30 MARTIN 1250 3450
30 WARD 1250 3450
30 JAMES 950 950
{code}
Again, look at MARTIN and WARD near the end. They both have the ame sal, so they both have the same running_total=3450 (= 950 + 1250 + 1250). This is often a desireable result, but, in your case, it seems not to be. If you want separate running_totals for MARTIN and WARD, then you eigher have to use a ROW-based window, like we did earlier, or add a tie-breaker to the ORDER BY clause, like this:
{code}
SELECT     deptno
,     ename
,     sal
,     SUM (sal) OVER ( PARTITION BY deptno
               ORDER BY      sal
               ,          ename          DESC     -- Changed (this is the only change)
               )     AS running_total
FROM     scott.emp
ORDER BY deptno
,      sal          DESC
,     ename
{code}
Output:
{code}
` DEPTNO ENAME SAL RUNNING_TOTAL
10 KING 5000 8750
10 CLARK 2450 3750
10 MILLER 1300 1300
20 FORD 3000 10875
20 SCOTT 3000 7875
20 JONES 2975 4875
20 ADAMS 1100 1900
20 SMITH 800 800
30 BLAKE 2850 9400
30 ALLEN 1600 6550
30 TURNER 1500 4950
30 MARTIN 1250 3450
30 WARD 1250 2200
30 JAMES 950 950
{code}