Removing duplicates with aggregated values that cancel out - duplicates

Question

I am attempting to remove duplicates that are found in the serial number, but only if the amount bill of the duplicate serial number cancels out when aggregating the amount bill. For example, there are five serial number as "abc-321", but have five different amount dues ($500, $250, $-250, $30, $-30). The four duplicates would be $250, $-250, $30, $-30, because adding them together would cancel them or be 0.
I was only able to come up with a query to identify what the duplicate serial numbers are, but not sure how to aggregate them to cancel them out.
SELECT a.Serial, a.BillAmt, a.Date, a.Code
FROM TableA a
WHERE (a.Serial) in
(SELECT Serial
FROM TableA a
GROUP BY Serial
HAVING COUNT(Serial)>1)
GROUP BY a.Serial, a.BillAmt, a.Date, a.Code
ORDER BY a.Serial ASC;
Sample Output:
+--------+---------+----------+------+
|Serial | BillAmt | Date | Code |
+--------+---------+----------+------+
|abc-112 | $240 | 20200720 | MPO |
|abc-112 | -$400 | 20200527 | CPP |
|acc-130 | $300 | 20200515 | CPP |
|acc-130 | $300 | 20200420 | DUB |
|acc-130 | -$300 | 20200515 | CPP |
|bcc-111 | $500 | 20200701 | MPO |
|bcc-111 | -$500 | 20200701 | MPO |
|caa-321 | $700 | 20200805 | DUB |
|caa-321 | $700 | 20200805 | MPO |
+--------+---------+----------+------+
Desire Results:
+--------+---------+----------+------+
|Serial | BillAmt | Date | Code |
+--------+---------+----------+------+
|abc-112 | $240 | 20200720 | MPO |
|abc-112 | -$400 | 20200527 | CPP |
|acc-130 | $300 | 20200420 | DUB |
|caa-321 | $700 | 20200805 | DUB |
|caa-321 | $700 | 20200805 | MPO |
+--------+---------+----------+------+

Answer 1

You are trying to compare the data to itself. You can do this by using a JOIN statement from TableA to TableA.
Here's what worked for me, with some test data. I did it by doing a partial OUTER JOIN (specifically a LEFT JOIN), with one of the JOIN conditions being the BillAmt's cancelling. Then in the WHERE, I look for where the JOIN failed. This is will be where the RIGHT side is NULL.
CREATE TABLE TableA (
Serial VARCHAR(10),
BillAmt DOUBLE,
`Date` DATE,
Code INT
);
INSERT INTO TableA (Serial, BillAmt, Date, Code) VALUES
('abc-321',500,'2021-06-01',1),
('abc-321',250,'2021-06-02',1),
('abc-321',-250,'2021-06-03',1),
('abc-321',50,'2021-06-04',1),
('abc-321',-50,'2021-06-05',1),
('abc-321',123,'2021-06-06',1);
SELECT a.Serial, a.BillAmt, a.Date, a.Code
FROM TableA AS a
LEFT JOIN TableA AS b
ON a.Serial = b.Serial
AND a.BillAmt = (b.BillAmt * -1)
WHERE b.Serial IS NULL;
It gives this output:
+---------+---------+------------+------+
| Serial | BillAmt | Date | Code |
+---------+---------+------------+------+
| abc-321 | 500 | 2021-06-01 | 1 |
| abc-321 | 123 | 2021-06-06 | 1 |
+---------+---------+------------+------+
2 rows in set (0.00 sec)

Answer 2

This solution will give you the result without a self-join, and a single pass over the table.
We calculate the total number of positive numbers per Serial per absolute value of BillAmt (ignoring the negative sign)
Same for the number of negatives
We calculate a row number over the same grouping, per sign
We filter only where the BillAmt is positive and current row number < = number of positives minus negatives, or it's negative and number of current row number <= negatives - positives
SELECT a.Serial, a.BillAmt, a.Date, a.Code
FROM (
SELECT a.*,
positives = COUNT(CASE WHEN a.BillAmt > 0 THEN 1 END) OVER (PARTITION BY a.Serial, ABS(a.BillAmt)),
negatives = COUNT(CASE WHEN a.BillAmt < 0 THEN 1 END) OVER (PARTITION BY a.Serial, ABS(a.BillAmt)),
rn = ROW_NUMBER() OVER (PARTITION BY a.Serial, ABS(a.BillAmt), SIGN(a.BillAmt) ORDER BY a.Date)
FROM TableA a
) a
WHERE a.BillAmt > 0 AND rn <= positives - negatives
OR a.BillAmt < 0 AND rn <= negatives - positives
ORDER BY a.Serial ASC;
db<>fiddle