SQL: Difference in amounts between rows and columns - sql

Question

I have a Table
With the below data
Table X
Seq_no A Claim Payment Balance (to be calculated)
1 abc 100 10 90 (100-10)
2 abc 50 40 (90-50)
3 abc 20 20 (40-20)
1 xyz 150 10 140 (150-10)
1 qwe 200 10 190 (200-10)
I need to calculate the column Balance.
I am trying with the below Query
SQL >
Select
Seq_no, a, Claim, Payment,
CASE
When Seq_no =1
then (claim-Payment)
Else ( lag(Balance)- Payment over (order by Balance))
END as Balance
from table X
However i am getting a error
ORA-00904: "Balance": invalid identifier
00904. 00000 - "%s: invalid identifier"
I believe this is because Balance is not an existing column name.
Is there a correct way to achieve the results.?
Update:
*I missed an important part. *
The data i have is in the below format:
Table X
Seq_no A Claim Payment
1 abc 100 10
2 abc 100 50
3 abc 100 20
1 xyz 150 10
1 qwe 200 10
I need the results in the below format.
Table X
Seq_no A Claim Payment Balance (to be calculated)
1 abc 100 10 90 (100-10)
2 abc 50 40 (90-50)
3 abc 20 20 (40-20)
1 xyz 150 10 140 (150-10)
1 qwe 200 10 190 (200-10)
The Seq_no calculation has been done to make the claim column null for the cases of duplicate claims, which i had figured out already.

Answer 1

You cannot reference balance before create it.
You can use a "running sum" for what you want to achieve.
Notice that I partitioned by A because you want another balance for every A.
Select
Seq_no, a, Claim, Payment,
sum(nvl(claim,0) - payment) over (partition by A order by seq_no) as Balance
from X;
Result:
SEQ_NO A CLAIM PAYMENT BALANCE
1 abc 100 10 90
2 abc 50 40
3 abc 20 20
1 qwe 200 10 190
1 xyz 150 10 140
EDIT: With newer dataset you just need to replace nvl function with a case when seq=1:
Select
Seq_no,
a,
case when seq_no=1 then claim else 0 end as Claim,
Payment,
sum(case when seq_no=1 then claim else 0 end - payment)
over (partition by A order by seq_no) as Balance
from X;

Answer 2

In order to use balance as a column identifier, you must go one level deep, i.e. make your existing query as a sub-query.
Working demo:
SQL> WITH sample_data AS(
2 SELECT 1 Seq_no, 'abc' A, 100 Claim, 10 Payment FROM dual UNION ALL
3 SELECT 2 Seq_no, 'abc' A, NULL Claim, 50 FROM dual UNION ALL
4 SELECT 3 Seq_no, 'abc' A, NULL Claim, 20 FROM dual UNION ALL
5 SELECT 1 Seq_no, 'xyz' A, 150 Claim, 10 FROM dual UNION ALL
6 SELECT 1 Seq_no, 'qwe' A, 200 Claim, 10 FROM dual
7 )
8 -- end of sample_data mimicking real table
9 SELECT seq_no, A, claim, payment,
10 CASE
11 WHEN lag(balance) OVER(PARTITION BY A ORDER BY seq_no) IS NULL
12 THEN balance
13 ELSE lag(balance) OVER(PARTITION BY A ORDER BY seq_no) - payment
14 END balance
15 FROM
16 (SELECT seq_no, A, claim, payment,
17 CASE
18 WHEN seq_no = 1
19 THEN claim - payment
20 ELSE lag(claim - payment) OVER(PARTITION BY A ORDER BY seq_no) - payment
21 END balance
22 FROM sample_data
23 );
SEQ_NO A CLAIM PAYMENT BALANCE
---------- --- ---------- ---------- ----------
1 abc 100 10 90
2 abc 50 40
3 abc 20 20
1 qwe 200 10 190
1 xyz 150 10 140
SQL>
UPDATE : If you not-null values always for claim, then you could do a running diff:
SQL> WITH sample_data AS(
2 SELECT 1 Seq_no, 'abc' A, 100 Claim, 10 Payment FROM dual UNION ALL
3 SELECT 2 Seq_no, 'abc' A, 100 Claim, 50 FROM dual UNION ALL
4 SELECT 3 Seq_no, 'abc' A, 100 Claim, 20 FROM dual UNION ALL
5 SELECT 4 Seq_no, 'abc' A, 100 Claim, 10 FROM dual UNION ALL
6 SELECT 5 Seq_no, 'abc' A, 100 Claim, 10 FROM dual UNION ALL
7 SELECT 1 Seq_no, 'xyz' A, 150 Claim, 10 FROM dual UNION ALL
8 SELECT 1 Seq_no, 'qwe' A, 200 Claim, 10 FROM dual
9 )
10 -- end of sample_data mimicking real table
11 SELECT Seq_no,
12 a,
13 Claim,
14 Payment,
15 CASE
16 WHEN seq_no = 1
17 THEN claim - payment
18 ELSE SUM(claim - payment) over (partition BY A order by seq_no) - claim*(seq_no-1)
19 END balance
20 FROM sample_data;
SEQ_NO A CLAIM PAYMENT BALANCE
---------- --- ---------- ---------- ----------
1 abc 100 10 90
2 abc 100 50 40
3 abc 100 20 20
4 abc 100 10 10
5 abc 100 10 0
1 qwe 200 10 190
1 xyz 150 10 140
7 rows selected.
SQL>

Answer 3

You are correct that the error occurs because Balance is not a column in your table. You need to add that column.
The bigger problem is that unfortunately it's not possible to compute a running total using a single SQL statement the way you are trying to do it. SQL is not well suited to an operation like this. It can be done in a stored procedure using a cursor but the end result is procedural and not set-based, thus performance would be poor. Here's how to do it in SQL Server (as an example, not tested):
DECLARE PaymentCursor CURSOR FOR SELECT A, Claim, Payment FROM X ORDER BY A, Seq_no
OPEN PaymentCursor
DECLARE #A VARCHAR(16)
DECLARE #Claim MONEY
DECLARE #Payment MONEY
DECLARE #Balance MONEY
DECLARE #PreviousA VARCHAR(16)
SET #PreviousA = ''
FETCH NEXT FROM PaymentCursor INTO #A, #Claim, #Payment
WHILE ##FETCH_STATUS = 0
BEGIN
IF #A <> #PreviousA
BEGIN
SET #Balance = 0
SET #PreviousA = #A
END
SET #Balance = #Balance + #Claim - #Payment
UPDATE X SET Balance = #Balance WHERE CURRENT OF PaymentCursor
FETCH NEXT FROM PaymentCursor INTO #A, #Claim, #Payment
END
CLOSE PaymentCursor
DEALLOCATE PaymentCursor
If your application really needs to know the account balance associated with each transaction, the better approach is to compute the balance each time a row is inserted. Wrap the insertion logic in a stored procedure:
CREATE PROCEDURE InsertAndComputeBalance
#A VARCHAR(16),
#Claim MONEY,
#Payment MONEY
AS
DECLARE #MAX_Seq_no INTEGER = (SELECT MAX(Seq_no) FROM X WHERE A = #A)
INSERT INTO X
SELECT #MAX_Seq_no + 1, #A, #Claim, #Payment, Balance + #Claim - #Payment FROM X
WHERE A = #A AND Seq_no = #MAX_Seq_no
In theory you could call similar logic from a trigger so that it is automatically executed each time a row is inserted, but triggers can be evil so proceed very carefully with that approach.