The best way to perform such calculation logic in T-SQL - sql

Question

My program require to pass in an amount into the query to perform such calculation, but in my case it loop through row by row and deduct the correct amount, i know this is not a efficient way to implement. so i am here to seeking a better way.
PS: It is just my draft code, i am sorry about i cannot post the complete source code for some reason. Now i had re-structure my code to make it more complete and reasonable.
--- the amount column is just for reference.
insert into tbl1 (idx,amount,balance) values (1, 50, 50)
insert into tbl1 (idx,amount,balance) values (2, 30, 30)
insert into tbl1 (idx,amount,balance) values (3, 20, 20)
insert into tbl1 (idx,amount,balance) values (4, 50, 50)
insert into tbl1 (idx,amount,balance) values (5, 60, 60)
declare #total_value_to_deduct int
declare #cs_index int, #cs_balance int, #deduct_amount int
set #total_value_to_deduct = 130
declare csDeduct Cursor for select idx, balance from tbl1 where balance > 0
open csDeduct fetch next from csDeduct into #cs_index, #cs_balance
while ##FETCH_STATUS = 0 and #total_value_to_deduct > 0
begin
if #cs_balance >= #total_value_to_deduct
set #deduct_amount = #total_value_to_deduct
else
set #deduct_amount = #cs_balance
-- contine deduct row by row if the total_value_to_deduct is not 0
set #total_value_to_deduct = #total_value_to_deduct - #deduct_amount
update tbl1 set balance = balance - #deduct_amount where idx = #cs_index
fetch next from csDeduct into #cs_index, #cs_balance
end
close csDeduct
deallocate csDeduct
Expected Result :
idx amount balance
1 50 0
2 30 0
3 20 0
4 50 20
5 60 60
Your help is must appreciate. thank

Answer 1

Revision 1: I have added a third solution
First solution (SQL2005+; online query)
DECLARE #tbl1 TABLE
(
idx INT IDENTITY(2,2) PRIMARY KEY,
amount INT NOT NULL,
balance INT NOT NULL
);
INSERT INTO #tbl1 (amount,balance) VALUES (50, 50);
INSERT INTO #tbl1 (amount,balance) VALUES (30, 30);
INSERT INTO #tbl1 (amount,balance) VALUES (20, 20);
INSERT INTO #tbl1 (amount,balance) VALUES (50, 50);
INSERT INTO #tbl1 (amount,balance) VALUES (60, 60);
DECLARE #total_value_to_deduct INT;
SET #total_value_to_deduct = 130;
WITH CteRowNumber
AS
(
SELECT *, ROW_NUMBER() OVER(ORDER BY idx) AS RowNum
FROM #tbl1 a
), CteRecursive
AS
(
SELECT a.idx,
a.amount,
a.amount AS running_total,
CASE
WHEN a.amount <= #total_value_to_deduct THEN 0
ELSE a.amount - #total_value_to_deduct
END AS new_balance,
a.RowNum
FROM CteRowNumber a
WHERE a.RowNum = 1
--AND a.amount < #total_value_to_deduct
UNION ALL
SELECT crt.idx,
crt.amount,
crt.amount + prev.running_total AS running_total,
CASE
WHEN crt.amount + prev.running_total <= #total_value_to_deduct THEN 0
WHEN prev.running_total < #total_value_to_deduct AND crt.amount + prev.running_total > #total_value_to_deduct THEN crt.amount + prev.running_total - #total_value_to_deduct
ELSE crt.amount
END AS new_balance,
crt.RowNum
FROM CteRowNumber crt
INNER JOIN CteRecursive prev ON crt.RowNum = prev.RowNum + 1
--WHERE prev.running_total < #total_value_to_deduct
)
UPDATE #tbl1
SET balance = b.new_balance
FROM #tbl1 a
Second solution (SQL2012)
UPDATE #tbl1
SET balance = b.new_balance
FROM #tbl1 a
INNER JOIN
(
SELECT x.idx,
SUM(x.amount) OVER(ORDER BY x.idx) AS running_total,
CASE
WHEN SUM(x.amount) OVER(ORDER BY x.idx) <= #total_value_to_deduct THEN 0
WHEN SUM(x.amount) OVER(ORDER BY x.idx) - x.amount < #total_value_to_deduct --prev_running_total < #total_value_to_deduct
AND SUM(x.amount) OVER(ORDER BY x.idx) > #total_value_to_deduct THEN SUM(x.amount) OVER(ORDER BY x.idx) - #total_value_to_deduct
ELSE x.amount
END AS new_balance
FROM #tbl1 x
) b ON a.idx = b.idx;
Third solution (SQ2000+) uses triangular join:
UPDATE #tbl1
SET balance = d.new_balance
FROM #tbl1 e
INNER JOIN
(
SELECT c.idx,
CASE
WHEN c.running_total <= #total_value_to_deduct THEN 0
WHEN c.running_total - c.amount < #total_value_to_deduct --prev_running_total < #total_value_to_deduct
AND c.running_total > #total_value_to_deduct THEN c.running_total - #total_value_to_deduct
ELSE c.amount
END AS new_balance
FROM
(
SELECT a.idx,
a.amount,
(SELECT SUM(b.amount) FROM #tbl1 b WHERE b.idx <= a.idx) AS running_total
FROM #tbl1 a
) c
)d ON d.idx = e.idx;

Answer 2

Here is one of the ways to do it. It finds first running sum greater than or equal to requested amount and then updates all records participating in this sum. This should probably be written differently in a sense that a column "toDeduct" should be introduced and would initially have the value of amount. This would allow this update to work over previously used data sets, because toDeduct = 0 would mean that nothing can be deducted from this row. Furthermore, an index on toDeduct, idx will allow for quick toDeduct <> 0 filter you would use to lessen number of meaningless searches/updates.
declare #total_value_to_deduct int
set #total_value_to_deduct = 130
update tbl1
set balance = case when balance.idx = tbl1.idx
then balance.sumamount - #total_value_to_deduct
else 0
end
from tbl1 inner join
(
select top 1 *
from
(
select idx, (select sum (a.amount)
from tbl1 a
where a.idx <= tbl1.idx) sumAmount
from tbl1
) balance
where balance.sumamount >= #total_value_to_deduct
order by sumamount
) balance
on tbl1.idx <= balance.idx
Now on to your cursor. One would gain performance by simply declaring cursor fast_forward:
declare csDeduct Cursor local fast_forward
for select idx, balance
from tbl1
where balance > 0
order by idx
And you might rewrite fetch loop to avoid repeating fetch statement:
open csDeduct
while 1 = 1
begin
fetch next from csDeduct into #cs_index, #cs_balance
if ##fetch_status <> 0
break
if #cs_balance >= #total_value_to_deduct
set #deduct_amount = #total_value_to_deduct
else
set #deduct_amount = #cs_balance
-- contine deduct row by row if the total_value_to_deduct is not 0
set #total_value_to_deduct = #total_value_to_deduct - #deduct_amount
update tbl1 set balance = balance - #deduct_amount where idx = #cs_index
end
close csDeduct
deallocate csDeduct
Makes changing select part of a cursor a bit easier.

Answer 3

I'm pretty sure this query won't work anyway, as "index" is a keyword and so should be wrapped in square brackets to indicate otherwise.
In general it's not a good idea to do anything on a row-by-row basis for performance.
If I'm reading it right, you're setting each balance column to the amount column minus the #total_value_to_deduct variable, or setting it to 0 if the deductions would result in a negative amount. If that's true then why not just do calculations on that directly? Without you posting any expected results I can't double check my logic, but please correct me if I'm wrong and it's more complicated than this.
UPDATE tbl1
SET balance = CASE
WHEN amount < #total_value_to_deduct THEN 0
ELSE amount - #total_value_to_deduct
END
Edit: OK thanks for the edit to the question it's more clear now. You're trying to take the total amount over all the accounts sequentially. I'll see if I can come up with a script to do this and edit my answer further.
Edit #2: OK, I couldn't find a way of doing it without interating through all of the rows (I tried a recursive CTE, but couldn't get it to work) so I've done it with a while loop like you did originally. It's effectively doing 3 data accesses per row though - I tried to knock this down to 2 but again no luck. I'm posting it anyway in case it's faster than what you have now. This should be all the code you need (apart from the table create/populate).
DECLARE #id INT
SELECT #id = Min([index])
FROM tbl1
WHILE #id IS NOT NULL
BEGIN
UPDATE tbl1
SET balance = CASE
WHEN amount < #total_value_to_deduct THEN 0
ELSE amount - #total_value_to_deduct
END
FROM tbl1
WHERE [index] = #id
SELECT #total_value_to_deduct = CASE
WHEN #total_value_to_deduct < amount THEN 0
ELSE #total_value_to_deduct - amount
END
FROM tbl1
WHERE [index] = #id
SELECT #id = Min([index])
FROM tbl1
WHERE [index] > #id
END

Answer 4

If your indexes don't have gaps, the simplest solution would be to
Create a recursive CTE, starting with the value to deduct and decrementing it in the recursive part.
Use the results of the CTE to update your actual table
SQL Statement
;WITH q AS (
SELECT idx, amount, balance, 130 AS Deduct
FROM tbl1
WHERE idx = 1
UNION ALL
SELECT t.idx, t.amount, t.balance, q.Deduct - q.balance
FROM q
INNER JOIN #tbl1 t ON t.idx = q.idx + 1
WHERE q.Deduct - q.balance > 0
)
UPDATE #tbl1
SET Balance = CASE WHEN q.Balance - q.Deduct > 0 THEN q.Balance - q.Deduct ELSE 0 END
FROM q
INNER JOIN tbl1 t ON t.idx = q.idx
Using ROW_NUMBERyou can alleviate the gap problem but it complicates the query a bit.
;WITH r AS (
SELECT idx, amount, balance, rn = ROW_NUMBER() OVER (ORDER BY idx)
FROM tbl1
), q AS (
SELECT rn, amount, balance, 130 AS Deduct, idx
FROM r
WHERE rn = 1
UNION ALL
SELECT r.rn, r.amount, r.balance, q.Deduct - q.balance, r.idx
FROM q
INNER JOIN r ON r.rn = q.rn + 1
WHERE q.Deduct - q.balance > 0
)
UPDATE tbl1
SET Balance = CASE WHEN q.Balance - q.Deduct > 0 THEN q.Balance - q.Deduct ELSE 0 END
FROM q
INNER JOIN #tbl1 t ON t.idx = q.idx
Test script
DECLARE #tbl1 TABLE (idx INTEGER, Amount INTEGER, Balance INTEGER)
INSERT INTO #tbl1 (idx,amount,balance) VALUES (1, 50, 50)
INSERT INTO #tbl1 (idx,amount,balance) VALUES (2, 30, 30)
INSERT INTO #tbl1 (idx,amount,balance) VALUES (3, 20, 20)
INSERT INTO #tbl1 (idx,amount,balance) VALUES (4, 50, 50)
INSERT INTO #tbl1 (idx,amount,balance) VALUES (5, 60, 60)
;WITH q AS (
SELECT idx, amount, balance, 130 AS Deduct
FROM #tbl1
WHERE idx = 1
UNION ALL
SELECT t.idx, t.amount, t.balance, q.Deduct - q.balance
FROM q
INNER JOIN #tbl1 t ON t.idx = q.idx + 1
WHERE q.Deduct - q.balance > 0
)
UPDATE #tbl1
SET Balance = CASE WHEN q.Balance - q.Deduct > 0 THEN q.Balance - q.Deduct ELSE 0 END
FROM q
INNER JOIN #tbl1 t ON t.idx = q.idx
SELECT *
FROM #tbl1
Output
idx Amount Balance
1 50 0
2 30 0
3 20 0
4 50 20
5 60 60

Answer 5

Create a new column in the table with the previous balance for each row, then you could use a trigger on INSERT/UPDATE to create the balance for the newly inserted row.