Aggregate values by parents recursively in Oracle - sql

Question

Consider the the following example structure:
DEPARTMENT
ID
PARENT_ID
NAME
DEPTH
PROJECT
ID
NAME
COST
DEPARTMENT_ID
Some data, just for the sake of the examples bellow:
| ID | PARENT_ID | NAME | DEPTH |
|----|-----------|-------|-------|
| 1 | NULL | DEPT1 | 1 |
| 2 | 1 | DEPT2 | 2 |
| 3 | 1 | DEPT3 | 2 |
| 4 | 2 | DEPT4 | 3 |
| 5 | 3 | DEPT5 | 3 |
| 6 | NULL | DEPT6 | 1 |
| 7 | 6 | DEPT7 | 2 |
| ID | NAME | COST | DEPARTMENT_ID |
|------|--------|-------|---------------|
| 1 | PRJ1 | 100 | 1 |
| 2 | PRJ2 | 200 | 2 |
| 3 | PRJ3 | 300 | 3 |
| 4 | PRJ4 | 400 | 4 |
| 5 | PRJ5 | 500 | 5 |
| 6 | PRJ6 | 600 | 6 |
| 7 | PRJ7 | 700 | 7 |
Now, I need to somehow aggregate the costs of the, projects by one department and then by its direct children.
If the choosen filter is DEPT1, the intented result is:
| LINE | DEPARTMENT_ID | PARENT_ID | NAME | AGGREGATE_COST |
|------|----------------|-----------|--------|----------------|
| 1 | 1 | NULL | DEPT1 | 1500 |
| 2 | 2 | 1 | DEPT2 | 600 |
| 3 | 3 | 1 | DEPT3 | 800 |
Where:
Line 3 aggregate is PRJ5 (of DEPT5, which is child of DEPT3) + PRJ3 (of DEPT3) cost
Line 2 aggregate is PRJ4 (of DEPT4, which is child of DEPT2) + PRJ2 (of DEPT2) cost
Line 1 aggregate is the sum of his childrens aggregates.
PRJ6 and PRJ7 costs are ignored because the are from DEPT6 and DEPT7, and those are not in the hierachy of DEPT1 (DEPT6 would be his sibling, not child)
EDIT:
| ID | NAME | COST | DEPARTMENT_ID |
|------|--------|-------|---------------|
| 1 | PRJ1 | 1 | 1 |
| 2 | PRJ2 | 1 | 1 |
| 3 | PRJ3 | 1 | 2 |
| 4 | PRJ4 | 1 | 2 |
| 5 | PRJ5 | 1 | 4 |
In this scenario, the solution ivanzg presented, doesn't seem to work.
I get doubled results for the projects in the highers ranks
If I get the aggregate for DEPT1, it returns something similar to this:
| LINE | DEPARTMENT_ID | PARENT_ID | NAME | AGGREGATE_COST |
|------|----------------|-----------|--------|----------------|
| 1 | 1 | NULL | DEPT1 | 8 |
| 2 | 2 | NULL | DEPT1 | 4 |

Answer 1

You can tag rows in a hierarchy query (to later create groups) by using CONNECT_BY_ROOT hierarchy operator. In the hierarchy query, by making all rows root rows you create every hierarchy combination, later only specified combinations are taken and aggregated. For your test data this returns what you specified.
SELECT ROOT_DEPT AS DEPARTMENT_ID
,ROOT_PARENT AS PARENT_ID
,ROOT_NAME AS NAME
,SUM(COST) AS AGGREGATE_COST
FROM (SELECT COST
,CONNECT_BY_ROOT DEPARTMENT_ID ROOT_DEPT
,CONNECT_BY_ROOT PARENT_ID ROOT_PARENT
,CONNECT_BY_ROOT NAME ROOT_NAME
FROM (SELECT B.DEPARTMENT_ID
,NVL(A.PARENT_ID,'0') PARENT_ID
,A.NAME
,SUM(B.COST) COST
FROM DEPARTMENT A
JOIN PROJECT B
ON A.ID = B.DEPARTMENT_ID
--> GROUP COST OF PROJECTS IN THE SAME DEPARTMENT IF THERE ARE ANY
GROUP BY B.DEPARTMENT_ID
,NVL(A.PARENT_ID,'0')
,A.NAME
)
--> MAKE ALL ROWS ROOT ROWS
CONNECT BY PRIOR DEPARTMENT_ID = PARENT_ID
)
WHERE ROOT_DEPT = 1 OR ROOT_PARENT = 1
GROUP BY ROOT_DEPT
,ROOT_PARENT
,ROOT_NAME