SQL optimization GROUP BY same table, same column - sql

Question

I have this GROUP BY query where I need to select some records from TABLE_1 and aggregate them.
The SELECTs are similar but I need to aggregate LON and BHAM separately as they are two different concepts but reside in same table.
My question is can I write the below differently in oracle that will optimize the performance of the query?
SELECT *
FROM (
( SELECT /*+ full(t1) */
t3.custId AS ID,
t2.secID AS SEC_ID,
t1.org_date AS SETT_DATE,
SUM(t1.amount) AS TOTAL
FROM test.TABLE_1 t1
INNER JOIN test.TABLE_2 t2 on t2.a_code = t1.a_code and t2.c_code = t1.c_code and t2.expiry_date > trunc(sysdate)
INNER JOIN test.TABLE_3 t3 on t3.account_id = t1.account_id
WHERE t1.city = 'LON'
AND t1.amount < 50000 and t1.amount > -50000
GROUP BY t3.custId, t2.secID, t1.org_date
)
UNION ALL
( SELECT /*+ full(t1) */
t3.custId AS ID,
t2.secID AS SEC_ID,
t1.org_date AS SETT_DATE,
SUM(t1.amount) AS TOTAL
FROM test.TABLE_1 t1
INNER JOIN test.TABLE_2 t2 on t2.a_code = t1.a_code and t2.c_code = t1.c_code and t2.expiry_date > trunc(sysdate)
INNER JOIN test.TABLE_3 t3 on t3.account_id = t1.account_id
WHERE t1.city = 'BHAM'
AND t3.alias = 'ABC'
AND t1.amount < 50000 and t1.amount > -50000
GROUP BY t3.custId, t2.secID, t1.org_date
)
)
ORDER BY ID, SEC_ID,
CASE WHEN SETT_DATE < TRUNC(sysdate) THEN trunc(sysdate) ELSE TRUNC(SETT_DATE) end

Answer 1

Remove union all and everything after, remove outer select, write where clause like here:
where -50000 < t1.amount and t1.amount < 50000
and (t1.city = 'LON' or (t1.city = 'BHAM' and t3.alias = 'ABC'))

Answer 2

You need to add the city column into the group by, and update the where clause to get both sets of rows, e.g.:
SELECT custid,
sec_id,
sett_date,
total
FROM (SELECT t3.custid AS id,
t2.secid AS sec_id,
CASE
WHEN t1.org_date < trunc(SYSDATE) THEN
trunc(SYSDATE)
ELSE
trunc(t1.org_date)
END AS sett_date,
t1.city,
SUM(t1.amount) AS total
FROM test.table_1 t1
INNER JOIN test.table_2 t2
ON t2.a_code = t1.a_code
AND t2.c_code = t1.c_code
AND t2.expiry_date > trunc(SYSDATE)
INNER JOIN test.table_3 t3
ON t3.account_id = t1.account_id
WHERE (t1.city = 'LON' OR (t1.city = 'BHAM' AND t3.alias = 'ABC'))
AND t1.amount < 50000
AND t1.amount > -50000
GROUP BY t3.custid,
t2.secid,
CASE
WHEN t1.org_date < trunc(SYSDATE) THEN
trunc(SYSDATE)
ELSE
trunc(t1.org_date)
END)
ORDER BY id,
sec_id,
sett_date;
However, I'm surprised you need both rows, since you can't tell which belongs to which city. I suspect you need to include the city (and maybe alias) columns into the final results.