Only keep rows within certain field based on another field - sql

Question

Having trouble with this one, can this be done in one query? Or should I delete all the N records and then use another query to get what I need? Help is appreciated!!
Only keep rows belonging to a group_id where the group_id contains YY for the same brand. If the flag is a Y/N or N/N for the same brand within a group_id, then remove those records.

Answer 1

Next time, please supply your data so that we can copy-paste it into an example and work with it. You would have saved me at least 15 min ....
I did not bother with unimportant data - they will be different from your input, but don't matter.
One way ist to use an OLAP function partitioned by grp_id,brand,flag, and a counter, and filtering by flag and counter value 2, or you could also use a grouping query, and join that grouping query back to the base table.
Here is the OLAP solution:
WITH
-- your (simplified) input, don't use in query )
indata(grp_id,secondary_id,account_id,site_id,op_id,brand,flag) AS (
SELECT 1, 1055884386, 123, 'usa', 2,'toyota', 'N' FROM dual
UNION ALL SELECT 1, 1055884386, 123, 'usa', 2,'toyota', 'N' FROM dual
UNION ALL SELECT 1, 1055884386, 123, 'usa', 2,'tesla' , 'Y' FROM dual
UNION ALL SELECT 1, 1055884386, 123, 'usa', 2,'tesla' , 'Y' FROM dual
UNION ALL SELECT 2, 1055884386, 123, 'usa', 2,'toyota', 'Y' FROM dual
UNION ALL SELECT 2, 1055884386, 123, 'usa', 2,'toyota', 'N' FROM dual
UNION ALL SELECT 2, 1055884386, 123, 'usa', 2,'tesla' , 'Y' FROM dual
UNION ALL SELECT 2, 1055884386, 123, 'usa', 2,'tesla' , 'N' FROM dual
UNION ALL SELECT 3, 1055884386, 123, 'usa', 2,'toyota', 'Y' FROM dual
UNION ALL SELECT 3, 1055884386, 123, 'usa', 2,'toyota', 'N' FROM dual
UNION ALL SELECT 3, 1055884386, 123, 'usa', 2,'tesla' , 'Y' FROM dual
UNION ALL SELECT 3, 1055884386, 123, 'usa', 2,'tesla' , 'N' FROM dual
UNION ALL SELECT 4, 1055884386, 123, 'usa', 2,'toyota', 'Y' FROM dual
UNION ALL SELECT 4, 1055884386, 123, 'usa', 2,'toyota', 'Y' FROM dual
UNION ALL SELECT 4, 1055884386, 123, 'usa', 2,'tesla' , 'N' FROM dual
UNION ALL SELECT 4, 1055884386, 123, 'usa', 2,'tesla' , 'N' FROM dual
)
-- end of your input; replace following comma with "WITH" ...
,
w_counter AS (
SELECT
*
, COUNT(*) OVER(PARTITION BY grp_id,brand,flag) AS occ_count
FROM indata
)
SELECT
*
FROM w_counter
WHERE flag='Y'
AND occ_count = 2
;
-- out grp_id | secondary_id | account_id | site_id | op_id | brand | flag | occ_count
-- out --------+--------------+------------+---------+-------+--------+------+-----------
-- out 1 | 1055884386 | 123 | usa | 2 | tesla | Y | 2
-- out 1 | 1055884386 | 123 | usa | 2 | tesla | Y | 2
-- out 4 | 1055884386 | 123 | usa | 2 | toyota | Y | 2
-- out 4 | 1055884386 | 123 | usa | 2 | toyota | Y | 2
And, for good measure, also the grouping solution - not repeating the input WITH clause ...
grp AS (
SELECT
grp_id
, brand
FROM indata
WHERE flag='Y'
GROUP BY
grp_id
, brand
HAVING COUNT(brand) = 2
)
SELECT
i.*
FROM indata i
JOIN grp USING(grp_id,brand)
ORDER BY grp_id
;
-- out grp_id | secondary_id | account_id | site_id | op_id | brand | flag
-- out --------+--------------+------------+---------+-------+--------+------
-- out 1 | 1055884386 | 123 | usa | 2 | tesla | Y
-- out 1 | 1055884386 | 123 | usa | 2 | tesla | Y
-- out 4 | 1055884386 | 123 | usa | 2 | toyota | Y
-- out 4 | 1055884386 | 123 | usa | 2 | toyota | Y