Running total of calculated field in pivot

Question

VERSION: ORACLE 11
TABLE:
create table chart_detail (
DIS        NUMBER,
BLD_MO VARCHAR2(7),
BLD        NUMBER(10),
RPLC      NUMBER(10));DATA:
insert into chart_detail values (60,'2011-03',0,2);
insert into chart_detail values (150,'2011-04',10572,0);
insert into chart_detail values (120,'2011-04',26449,5);
insert into chart_detail values (30,'2011-04',0,1);
insert into chart_detail values (60,'2011-04',0,7);
insert into chart_detail values (90,'2011-04',0,9);
insert into chart_detail values (120,'2011-05',5714,0);
insert into chart_detail values (90,'2011-05',24557,1);
insert into chart_detail values (60,'2011-05',0,4);
insert into chart_detail values (30,'2011-05',0,0);
COMMIT;EXPECTED RESULTS:
     2011-04                    2011-05               
DIS     RPLC     BLD     TBLD     IPTV     RPLC     BLD     TBLD     IPTV
30     1     0     37021     0.03     0     0     30271     0.00
60     7     0     37021     0.22     4     0     30271     0.13
90     9     0     37021     0.46     1     24557     30271     0.17
120     5     26449     37021     0.59     0     5714     5714     0.17
150     0     10572     10572     0.59               0     
180               0                    0     
TOTAL     22     37021               5     30271          PROBLEM: I need to have a running total of IPTV like in the above example. I can get the IPTV for each DIS/bld_mo but I don't know how to get the running total of it. In the script below I just used an example where I tried summing the IPTV like was done for build. I know it can't be done that way because IPTV is a calculated field in the query but if I substitute "APR_IPTV" with the formula for IPTV I get an error that window functions aren't allowed here. I do not know a way around this. I commented out the bad piece of code.
PROBLEM SCRIPT:
WITH  pivot_results  AS
  SELECT    dis
  ,    NVL (apr11_rep,  0)  AS apr11_rep
  ,    NVL (apr11_bld,   0)  AS apr11_bld
  ,    NVL ( SUM (apr11_bld)
                  OVER (ORDER BY dis DESC)
                ,                 0
                )        AS apr11_tbld
,      DECODE(NVL ( SUM (apr11_bld)
                  OVER (ORDER BY dis DESC)
                ,                 0),0,0,ROUND(NVL(apr11_rep, 0)*1000/ NVL ( SUM (apr11_bld)
                  OVER (ORDER BY dis DESC)
                ,                 0),2
                ))        AS apr11_iptv               
,      NVL ( SUM (apr11_iptv)
                  OVER (ORDER BY dis DESC)
                ,                 0
                )        AS apr11_tiptv  
  ,    NVL (may11_rep,  0)  AS may11_rep
  ,    NVL (may11_bld,   0)  AS may11_bld
  ,    NVL ( SUM (may11_bld)
                  OVER (ORDER BY dis DESC)
                ,                 0
                )        AS may11_tbld
,      DECODE(NVL ( SUM (may11_bld)
                  OVER (ORDER BY dis DESC)
                ,                 0),0,0,ROUND(NVL(may11_rep, 0)*1000/ NVL ( SUM (may11_bld)
                  OVER (ORDER BY dis DESC)
                ,                 0),2
                ))        AS may11_iptv
,      DECODE(NVL ( SUM (may11_bld)
                  OVER (ORDER BY dis DESC)
                ,                 0),0,0,ROUND(NVL(may11_rep, 0)*1000/ NVL ( SUM (may11_bld)
                  OVER (ORDER BY dis DESC)
                ,                 0),2
                ))        AS may11_tiptv               
  FROM     chart_detail
  PIVOT     (  MAX (rplc)  AS rep
          ,  MAX (bld)  AS bld
          FOR  bld_mo   IN ( '2011-04'  AS apr11
                                         , '2011-05'  AS may11
SELECT    CASE
        WHEN  GROUPING (dis) = 0
        THEN  TO_CHAR (dis)
        ELSE  'Total'
    END      AS dis
,    SUM (apr11_rep)  AS apr11_rep
,    SUM (apr11_bld)  AS apr11_bld
,    SUM (apr11_tbld) AS apr11_tbld
,    CASE
        WHEN  GROUPING (dis) = 0
        THEN  SUM (apr11_iptv)
    END      AS apr11_iptv
,    SUM (apr11_tiptv) AS apr11_tiptv
,    CASE
        WHEN  GROUPING (dis) = 0
        THEN  SUM (apr11_tpiptv)
    END      AS apr11_tiptv
,    SUM (may11_rep)  AS may11_rep
,    SUM (may11_bld)  AS may11_bld
,    SUM (may11_tbld) AS may11_tbld
,    CASE
        WHEN  GROUPING (dis) = 0
        THEN  SUM (may11_iptv)
    END      AS may11_iptv   
FROM    pivot_results
GROUP BY  ROLLUP (dis)
ORDER BY  pivot_results.dis
;Thank you,

Answer 1

Hi,
So you know how to compute iptv for an individual row; the problem now is that you want to get a running total of iptv; is that it?
The problem there is that computing iptv requires an analytic function, and analytic functions can't be nested. To get the results of nesting f (g (x)), where f and g are analytic funtions, you have to compute g in a sub-query, and then use the results as the argument to f in a super-query.
Here's how to apply that to your situation:
WITH  pivot_results  AS
     SELECT    dis
-- April, 2011
       ,           NVL (apr11_rep,   0)  AS apr11_rep
       ,           NVL (apr11_bld,   0)  AS apr11_bld
       ,           NVL ( SUM (apr11_bld)
                                   OVER (ORDER BY dis DESC)
                      , 0
                      )               AS apr11_tbld
     ,      NVL ( 1000 * apr11_rep
                          / NULLIF ( SUM (apr11_bld) OVER (ORDER BY dis DESC)
                                      , 0
               , 0
               )               AS apr11_iptv
-- May, 2011
       ,           NVL (may11_rep,   0)  AS may11_rep
       ,           NVL (may11_bld,   0)  AS may11_bld
       ,           NVL ( SUM (may11_bld)
                                   OVER (ORDER BY dis DESC)
                      , 0
                      )               AS may11_tbld
     ,      NVL ( 1000 * may11_rep
                          / NULLIF ( SUM (may11_bld) OVER (ORDER BY dis DESC)
                                      , 0
               , 0
               )               AS may11_iptv
       FROM     chart_detail
       PIVOT    (    MAX (rplc)  AS rep
                ,    MAX (bld)   AS bld
                FOR  bld_mo   IN ( '2011-04'  AS apr11
                                  , '2011-05'  AS may11
SELECT    CASE
              WHEN  GROUPING (dis) = 0
              THEN  TO_CHAR (dis)
              ELSE  'Total'
           END      AS dis
-- April 2011
,           SUM (apr11_rep)  AS apr11_rep
,           SUM (apr11_bld)  AS apr11_bld
,           SUM (apr11_tbld) AS apr11_tbld
,           CASE
              WHEN  GROUPING (dis) = 0
              THEN  ROUND ( SUM (SUM (apr11_iptv))
                                      OVER (ORDER BY  dis)
                  , 2
           END      AS apr11_iptv
-- May 2011
,           SUM (may11_rep)  AS may11_rep
,           SUM (may11_bld)  AS may11_bld
,           SUM (may11_tbld) AS may11_tbld
,           CASE
              WHEN  GROUPING (dis) = 0
              THEN  ROUND ( SUM (SUM (may11_iptv))
                                      OVER (ORDER BY  dis)
                  , 2
           END      AS may11_iptv
FROM      pivot_results
GROUP BY  ROLLUP (dis)
ORDER BY  pivot_results.dis
;Output:
      APR11  APR11   APR11  APR11 MAY11  MAY11   MAY11  MAY11
DIS    _REP   _BLD   _TBLD  _IPTV  _REP   _BLD   _TBLD  _IPTV
30        1      0   37021    .03     0      0   30271    .00
60        7      0   37021    .22     4      0   30271    .13
90        9      0   37021    .46     1  24557   30271    .17
120       5  26449   37021    .59     0   5714    5714    .17
150       0  10572   10572    .59     0      0       0    .17
Total    22  37021  158656            5  30271   96527As you can see, this is not quite what you wanted on the row where dis='150'. You asked for NULLS in the may11_rep, may11_bld and may11_iptv columns. You can get those results if you need them; just explain the rules that govern whether to display the values and when to display NULL.
The way you posted the sample data and results, and the quantity of sample data were all excellent; it really helped me find a solution. Thanks.
It would have also helped it you had explained how iptv is computed. Basically, iptv = 1000 * rep / tbld, right?
It looks like most of this code:
,      DECODE(NVL ( SUM (may11_bld)
                  OVER (ORDER BY dis DESC)
                ,                 0),0,0,ROUND(NVL(may11_rep, 0)*1000/ NVL ( SUM (may11_bld)
                  OVER (ORDER BY dis DESC)
                ,                 0),2
                ))        AS may11_iptvwas a way of avoiding divide by 0 errors; it would have been helpful if you had explained that.