updating records by putting sum(column) value - sql

Question

I have table1 like this:
+-----+-----+------+
| cat | val | type |
+-----+-----+------+
| A | 100 | c1 |
| H | 25 | c2 |
| H | 50 | c3 |
| H | 30 | c2 |
| A | 15 | c3 |
| H | 10 | c1 |
| H | 15 | c1 |
| B | 10 | c4 |
| H | 20 | c4 |
| H | 15 | c3 |
+-----+-----+------+
I need to add the sum(val) group by type to only one H belonging to each type
So I have after grouping by type we have say table2 :
+-----+-----+
| cat | val |
+-----+-----+
| c1 | 125 |
| c2 | 55 |
| c3 | 80 |
| c4 | 30 |
+-----+-----+
I need 125 added to any one H values with type c1, 55 added to any one H values with c2, and so on..If there is no H with c1,then it should create that record.
So finally we get:
+-----+-----+------+
| cat | val | type |
+-----+-----+------+
| A | 100 | c1 |
| H | 25 | c2 |
| H | 130 | c3 |
| H | 85 | c2 |
| A | 15 | c3 |
| H | 135 | c1 |
| H | 15 | c1 |
| B | 10 | c4 |
| H | 50 | c4 |
| H | 15 | c3 |
+-----+-----+------+
How do I do it without doing table1 union table2 (with 'H' as cat) group by type? Also I don't have update privileges and cannot use stored procedures. I also have to keep in mind that table1 is a result of a query involving multiple inner joins that I don't want to use over and over again for select statements.

Answer 1

See whether this makes sense. I've added the ID column just to display the final result in the same order as the input (for easier reading).
SQL> -- T1 is what you currently have; it can/could be your current query
SQL> with t1 (id, cat, val, type) as
2 (select 1, 'A', 100, 'C1' from dual union all
3 select 2, 'H', 25 , 'C2' from dual union all
4 select 3, 'H', 50 , 'C3' from dual union all
5 select 4, 'H', 30 , 'C2' from dual union all
6 select 5, 'A', 15 , 'C3' from dual union all
7 select 6, 'H', 10 , 'C1' from dual union all
8 select 7, 'H', 15 , 'C1' from dual union all
9 select 8, 'B', 10 , 'C4' from dual union all
10 select 9, 'H', 20 , 'C4' from dual union all
11 select 10,'H', 15 , 'C3' from dual
12 ),
13 -- sum VAL per type
14 t1_sum as
15 (select type, sum(val) sum_val
16 from t1
17 group by type
18 ),
19 -- find row number; let any H be number 1
20 t1_rn as
21 (select id, cat, val, type,
22 row_number() over (partition by type
23 order by case when cat = 'H' then 1 end) rn
24 from t1
25 )
26 -- the final result; add SUM_VAL to the first H row per type
27 select r.cat, r.val + case when r.rn = 1 then s.sum_val else 0 end val,
28 r.type
29 From t1_rn r join t1_sum s on s.type = r.type
30 order by r.id;
CAT VAL TYPE
--- ---------- ----
A 100 C1
H 80 C2
H 130 C3
H 30 C2
A 15 C3
H 135 C1
H 15 C1
B 10 C4
H 50 C4
H 15 C3
10 rows selected.
SQL>
[EDIT: trying to clarify how to use your large query]
Suppose that this is that very large and complex query of yours:
select a.cat,
case when a.cat = 'A' then b.val
when a.cat = 'Z' then c.val
else 'H'
end val,
c.type
from a join b on a.id = b.id and a.x = b.y
join c on c.id = b.idx
where a.date_column < sysdate
and c.type = 'X';
As I've said, create a view based on it as
create or replace view v_view as
select a.cat,
case when a.cat = 'A' then b.val
when a.cat = 'Z' then c.val
else 'H'
end val,
c.type
from a join b on a.id = b.id and a.x = b.y
join c on c.id = b.idx
where a.date_column < sysdate
and c.type = 'X';
and use it as a source for "my" query (from line 14 onwards):
with t1_sum as
(select type, sum(val) sum_val
from v_view --> here's the view
group by type
), etc.
Or, use the "huge" query itself as the initial CTE:
with t1 as
-- this is your "huge" query
(select a.cat,
case when a.cat = 'A' then b.val
when a.cat = 'Z' then c.val
else 'H'
end val,
c.type
from a join b on a.id = b.id and a.x = b.y
join c on c.id = b.idx
where a.date_column < sysdate
and c.type = 'X'
),
-- sum VAL per type
t1_sum as
(select type, sum(val) sum_val
from t1
group by type
), etc.