Summerize accoding to all leves in herarchial table

Question

I have an herarchial table that contain the folowing structure and data:
STRUCTURE_ID NUMBER(5) not null;
CATALOG_ID NUMBER(5) not null;
MASTER_ITEM_ID VARCHAR2(15) not null;
DETAILED_ITEM_ID VARCHAR2(15) not null;
4     3     0     1
4     3     1     35
4     3     35     35000
4     3     35     35010
4     3     0     3
4     3     3     50
4     3     50     50000
4     3     3     51
4     3     51     51000
4     3     51     51200
In different table, I have financial data. Structure and data:
STRUCTURE_ID NUMBER(5) not null;
CATALOG_ID NUMBER(5) not null;
ITEM_ID VARCHAR2(15) not null;
SUM NUMBER(14,4);
1     4     3     35000     -3000000
2     4     3     50000     -56886.25
3     4     3     51000     61600
4     4     3     51200     10800
I want to have a select statement, that use use the connect by prior statement (to gain the herarchial structure), and to use a group by rollup to have totals to all the hirarchial levels.
I tried the next sql statement, but it not summerise all the levels in the herarchy:
select b.master_item_id, a.item_id, sum(a.sum)
from fin_data a, fin_structure b
where b.detailed_item_id =a.item_id (+)
and a.structure_id=4
and a.structure_id=b.structure_id
connect by prior b.detailed_item_id=b.master_item_id
group by rollup (b.master_item_id,item_id);
The result looks like this :
35     35000     -3000000
35          -3000000
50     50000     -56886.25
50          -56886.25
51     51000     61600
51     51200     10800
51          72400
          -2984486.25
There are no totals to masters 1 and 3.
What I do wrong?

Answer 1

Here is my improved solution:
- Instead of a scalar subquery, the SUMS factored subquery calculates all the sums at once, then I join to it.
This technique performs better if the result set is large.
- The TOTALS factored subquery "adds" the additional lines so you don't have to insert them.
- I filter out the detail line that has no SUM because the OP did.
Here is a test with two structure/catalog combinations:insert into fin_structure select 3,2,master_item_id, detailed_item_id from fin_structure;
insert into fin_data select 3,2,item_id, fin_sum from fin_data;
WITH sums AS (
  SELECT S.STRUCTURE_ID, S.CATALOG_ID, S.MASTER_ITEM_ID ITEM_ID,
  sum(CONNECT_BY_ROOT(FIN_SUM)) FIN_SUM_SUM
  from FIN_STRUCTURE s
  LEFT JOIN FIN_DATA d
    ON (s.STRUCTURE_ID, s.CATALOG_ID, s.DETAILED_ITEM_ID)
    = ((D.STRUCTURE_ID, D.CATALOG_ID, D.ITEM_ID))
  START WITH d.fin_sum is not null
  CONNECT BY (S.STRUCTURE_ID, S.CATALOG_ID, S.DETAILED_ITEM_ID)
    = ((PRIOR S.STRUCTURE_ID, PRIOR S.CATALOG_ID, PRIOR S.MASTER_ITEM_ID))
  group by s.structure_id, s.catalog_id, s.master_item_id
  UNION ALL
  SELECT STRUCTURE_ID, CATALOG_ID, ITEM_ID, FIN_SUM FROM FIN_DATA
), TOTALS AS (
  SELECT DISTINCT STRUCTURE_ID, CATALOG_ID,
  'Total' MASTER_ITEM_ID, MASTER_ITEM_ID DETAILED_ITEM_ID
  FROM FIN_STRUCTURE
  where master_item_id not in (select detailed_item_id from fin_structure)
SELECT A.*, b.fin_sum_sum
FROM (
  SELECT * FROM TOTALS
  UNION ALL
  SELECT * FROM FIN_STRUCTURE
) a
left join sums b
      ON (a.STRUCTURE_ID, a.CATALOG_ID, a.DETAILED_ITEM_ID)
      = ((B.STRUCTURE_ID, B.CATALOG_ID, B.ITEM_ID))
WHERE B.FIN_SUM_SUM IS NOT NULL
START WITH A.MASTER_ITEM_ID = 'Total'
CONNECT BY (A.STRUCTURE_ID, A.CATALOG_ID, A.MASTER_ITEM_ID)
  = ((PRIOR A.STRUCTURE_ID, PRIOR A.CATALOG_ID, PRIOR A.DETAILED_ITEM_ID));
STRUCTURE_ID CATALOG_ID MASTER_ITEM_ID  DETAILED_ITEM_ID FIN_SUM_SUM
           3          2 Total           0                -2984486.25
           3          2 0               1                   -3000000
           3          2 1               35                  -3000000
           3          2 35              35000               -3000000
           3          2 0               3                   15513.75
           3          2 3               50                 -56886.25
           3          2 50              50000              -56886.25
           3          2 3               51                     72400
           3          2 51              51000                  61600
           3          2 51              51200                  10800
           4          3 Total           0                -2984486.25
           4          3 0               1                   -3000000
           4          3 1               35                  -3000000
           4          3 35              35000               -3000000
           4          3 0               3                   15513.75
           4          3 3               50                 -56886.25
           4          3 50              50000              -56886.25
           4          3 3               51                     72400
           4          3 51              51000                  61600
           4          3 51              51200                  10800Best regards, Stew