ROLLUP for grand totals

Question

Hi guys.
I've been trying to summarize a bunch of rows, but I'm having a little trouble doing that.
I have few nested queries which I show below:
SELECT DATA, SAP, SADIG, DIFF
  FROM (SELECT TMPSAP.DATSAP as DATA,
               TMPSAP.VALSAP as SAP,
               TMPSADIG.VALSADIG as SADIG,
               ABS(TMPSAP.VALSAP - TMPSADIG.VALSADIG) as DIFF,
               GROUPING(TMPSAP.DATSAP) AS GRUPO
          FROM (SELECT SUM(ZVLBRT) + SUM(ZDESVP) as VALSAP,
                       TO_DATE(SPTAG, 'YYYYMMDD') as DATSAP
                  FROM SAPR3.S974@PRD
                 WHERE VKORG = '1000'
                   AND (TO_NUMBER(SPART) BETWEEN 10 AND 55 OR SPART = '99')
                   AND KUNNR <> 1004
                   AND TO_DATE(SPTAG, 'YYYYMMDD') BETWEEN
                       TRUNC(TRUNC(SYSDATE, 'DD') - 1, 'MM') AND
                       TRUNC(SYSDATE, 'DD') - 1
                 GROUP BY SPTAG) TMPSAP,
               (SELECT SUM(FATDEVVLR) as VALSADIG, DATFATDEV as DATSADIG
                  FROM SADIG_FATDEV
                 WHERE BUCDG = 01
                   AND DATFATDEV BETWEEN
                       TRUNC(TRUNC(SYSDATE, 'DD') - 1, 'MM') AND
                       TRUNC(SYSDATE, 'DD') - 1
                 GROUP BY DATFATDEV) TMPSADIG
         WHERE TRUNC(TMPSAP.DATSAP, 'DD') = TRUNC(TMPSADIG.DATSADIG, 'DD')
         GROUP BY GROUPING SETS((DATSAP, VALSAP, VALSADIG),())) ALLDATA
GROUP BY ROLLUP(ALLDATA.GRUPO, ALLDATA.DATA, SAP, SADIG, DIFF)
I'm having some trouble figuring out which GROUP BY attributes I should use for the outter query, so the results I get are like these:
DATA
SAP
SADIG
DIFF
01/08/2013
1000
2000
1000
01/08/2013
1000
2000
01/08/2013
1000
01/08/2013
02/08/2013
1000
1000
0
02/08/2013
1000
1000
02/08/2013
1000
02/08/2013
03/08/2013
2000
3000
1000
03/08/2013
2000
3000
03/08/2013
2000
03/08/2013
05/08/2013
5000
5000
0
05/08/2013
5000
5000
05/08/2013
5000
05/08/2013
06/08/2013
1000
2000
1000
06/08/2013
1000
2000
06/08/2013
1000
06/08/2013
If you look at my second select layer (the one that returns ALLDATA), my results are like these:
DATA
SAP
SADIG
DIFF
GRUPO
01/08/2013
1000
2000
1000
0
02/08/2013
1000
1000
0
0
03/08/2013
2000
3000
1000
0
05/08/2013
5000
5000
0
0
06/08/2013
1000
2000
1000
0
1
And what I want is this:
DATA
SAP
SADIG
DIFF
01/08/2013
1000
2000
1000
02/08/2013
1000
1000
0
03/08/2013
2000
3000
1000
05/08/2013
5000
5000
0
06/08/2013
1000
2000
1000
10000
13000
3000
I've tried several combinations in ROLLUP, but without success. What am I missing?
Thank you very much!
Best regards,
Vinicius

Answer 1

Hi, Frank! Thank you for your quick reply.
Just for reference, all my queries are running under Oracle 9.2.0.1.0, unless otherwise specified.
First of all, I'm sorry if I wasn't clear. I tried to provide as much information as possible, so I think I might have made things a little fuzzy.
Also, it's my very first post in this forum, so I'm not really used to oracle community standards - and I apologize for that.
Moving on, unfortunately, I'm a little way too much confused, so I don't know exactly a good way to simplify my problem, so going a little against of your advice, I'll dig a little deeper and unravel the matter in unitary steps. Also, unfortunately I can't provide CREATE TABLE or INSERT statements, since I really don't know what the whole structure behind the databases is. What I can is to provide a few CREATE TABLE and INSERT statements regarding step 2 and forward, so you could at least reproduce the situation. Here's the SQL:
CREATE TABLE TMPSAP (DATSAP DATE NOT NULL, VALSAP NUMBER (8,2) NOT NULL);
CREATE TABLE TMPSADIG (DATSADIG DATE NOT NULL, VALSADIG NUMBER (8,2) NOT NULL);
INSERT INTO TMPSAP VALUES(TO_DATE('01/08/2013','DD/MM/YYYY'), 300);
INSERT INTO TMPSAP VALUES(TO_DATE('02/08/2013','DD/MM/YYYY'), 1000);
INSERT INTO TMPSAP VALUES(TO_DATE('03/08/2013','DD/MM/YYYY'), -500);
INSERT INTO TMPSAP VALUES(TO_DATE('05/08/2013','DD/MM/YYYY'), 5000);
INSERT INTO TMPSAP VALUES(TO_DATE('06/08/2013','DD/MM/YYYY'), 1200);
INSERT INTO TMPSAP VALUES(TO_DATE('07/08/2013','DD/MM/YYYY'), 1800);
INSERT INTO TMPSADIG VALUES(TO_DATE('01/08/2013','DD/MM/YYYY'), 100);
INSERT INTO TMPSADIG VALUES(TO_DATE('02/08/2013','DD/MM/YYYY'), 1000);
INSERT INTO TMPSADIG VALUES(TO_DATE('03/08/2013','DD/MM/YYYY'), -500);
INSERT INTO TMPSADIG VALUES(TO_DATE('05/08/2013','DD/MM/YYYY'), 5000);
INSERT INTO TMPSADIG VALUES(TO_DATE('06/08/2013','DD/MM/YYYY'), 1200);
INSERT INTO TMPSADIG VALUES(TO_DATE('07/08/2013','DD/MM/YYYY'), 1800);
COMMIT;
Other than that, let me try it again:
I have two tables in two different systems which should have the same sum of values for one of their fields, given a date.
Step 1:
To check that, I have one query in one of those tables (which I'll call TMPSAP for now), which is running under Oracle 9.2.0.7.0 and accessed via DBLINK, that sums all the values and group them per date:
SELECT TO_DATE(SPTAG, 'YYYYMMDD') as DATSAP, SUM(ZVLBRT) + SUM(ZDESVP) as VALSAP                
                  FROM SAPR3.S974@PRD
                 WHERE VKORG = '1000'
                   AND (TO_NUMBER(SPART) BETWEEN 10 AND 55 OR SPART = '99')
                   AND KUNNR <> 1004
                   AND TO_DATE(SPTAG, 'YYYYMMDD') BETWEEN
                       TRUNC(TRUNC(SYSDATE, 'DD') - 1, 'MM') AND
                       TRUNC(SYSDATE, 'DD') - 1
                 GROUP BY SPTAG
                 ORDER BY DATSAP
Which returns:
DATSAP
VALSAP
01/08/2013
300
02/08/2013
1000
03/08/2013
-500
05/08/2013
5000
06/08/2013
1200
07/08/2013
1800
Step 2:
Likewise, I have another query running in the other system's tables (which I'll call TMPSADIG for now), that sums all the values and group them per date:
SELECT DATFATDEV as DATSADIG, SUM(FATDEVVLR) as VALSADIG
                  FROM SADIG_FATDEV
                 WHERE BUCDG = 01
                   AND DATFATDEV BETWEEN
                       TRUNC(TRUNC(SYSDATE, 'DD') - 1, 'MM') AND
                       TRUNC(SYSDATE, 'DD') - 1
                 GROUP BY DATFATDEV
                 ORDER BY DATSADIG
Which returns:
DATSADIG
VALSADIG
01/08/2013
100
02/08/2013
1000
03/08/2013
-500
05/08/2013
5000
06/08/2013
1200
07/08/2013
1800
Step 3:
I run a query (which I'll call ALLDATA for now) in those two tables, joining them by date and calculating the difference between the values:
SELECT TMPSAP.DATSAP as DATA,
               TMPSAP.VALSAP as SAP,
               TMPSADIG.VALSADIG as SADIG,
               ABS(TMPSAP.VALSAP - TMPSADIG.VALSADIG) as DIFF,
               GROUPING(TMPSAP.DATSAP) AS GRUPO
          FROM TMPSAP, TMPSADIG
         WHERE TRUNC(TMPSAP.DATSAP, 'DD') = TRUNC(TMPSADIG.DATSADIG, 'DD')
         GROUP BY GROUPING SETS((DATSAP, VALSAP, VALSADIG),())
         ORDER BY TMPSAP.DATSAP
Which returns:
DATA
SAP
SADIG
DIFF
GRUPO
01/08/2013
300
100
200
0
02/08/2013
1000
1000
0
0
03/08/2013
-500
-500
0
0
05/08/2013
5000
5000
0
0
06/08/2013
1200
1200
0
0
07/08/2013
1800
1800
0
0
Step 4:
And now I need to run a mysterious query in ALLDATA (results above), that will return me these exact results:
DATA
SAP
SADIG
DIFF
01/08/2013
300
100
200
02/08/2013
1000
1000
0
03/08/2013
-500
-500
0
05/08/2013
5000
5000
0
06/08/2013
1200
1200
0
07/08/2013
1800
1800
0
Total:
8800
8600
200
Which means, I need to add a row in the end of the query, containing the sum of all previous rows (the one in red). I've been messing around with ROLLUP and other grouping functions, but I couldn't really manage to achieve anything even close to that.
Of course I could run the query again and sum everything and union all, or even run this from a stored procedure and summarize with some coding, but the point of my whole thread is: Is there any way to do that using nothing more than a query?
I hope I clarified what I'm looking for and I apologize if it is still unclear or too messy, but I really can't find another way to present the information.
And I also hope someone can help me find a way to do that.
Thank you very much!
Kind regards,
Vinicius