SQL Server CTE Running balance - sql-server

Question

#table1
idno | amount
-------------
1 | 700
2 | 500
#table2
idno | amount1 | amount2 | amount3 | acctno
------------------------------------------
1 | 100 | 200 | 300 | 001
1 | 100 | 200 | 300 | 002
2 | 100 | 200 | 300 | 001
What I want to happen is to distribute the amount from table2 into table 1's amount1,amount2,amount3 respectively then get the remaining balance and apply to the next row. I've tried using CTE but got stucked on passing the running balance to the next row.
Query:
Declare #table2 TABLE (idno varchar(max), amount1 decimal,amount2
decimal,amount3 decimal,acctno varchar(max))
INSERT INTO #table2 VALUES
('1',100,200,300,'001'),
('1',100,200,300,'002'),
('2',100,200,300,'001')
Declare #table1 TABLE (idno varchar(max), amount decimal)
INSERT INTO #table1 VALUES
('1',700),
('2',500);
WITH due AS (SELECT a.idno,a.amount,b.acctno,b.amount1,b.amount2,b.amount3
from #table1 a left join #table2 b on a.idno = b.idno),
payment AS (SELECT *,case when amount-amount1<0 then amount
else amount1 end as amount1pay
,case when amount-amount1<=0 then 0
when amount-amount1-amount2 <0 then amount-amount1
else amount2 end as amount2pay ,
case when amount-amount1-amount2<=0 then 0
when amount-amount1-amount2-amount3<0
then amount-amount1-amount2 else amount3 end as amount3pay
FROM due),
payment2 AS (SELECT SUM(amount-amount1pay-amount2pay-amount3pay)
OVER ( PARTITION BY idno ORDER BY acctno
ROWS UNBOUNDED PRECEDING ) as balance,* FROM payment)
select * from payment2
Current Result
balance | idno | amount | acctno | amount1 | amount2 | amount3 | amount1pay | amount2pay | amount3pay
---------------------------------------------------------------------------------------------------------
100 | 1 | 200 | 001 | 100 | 200 | 300 | 100 | 200 | 300
200 | 1 | 200 | 002 | 100 | 200 | 300 | 100 | 200 | 300
0 | 2 | 500 | 001 | 100 | 200 | 300 | 100 | 200 | 200
Expected Result
balance | idno | amount | acctno | amount1 | amount2 | amount3 | amount1pay | amount2pay | amount3pay
---------------------------------------------------------------------------------------------------------
100 | 1 | 200 | 001 | 100 | 200 | 300 | 100 | 200 | 300
100 | 1 | 200 | 002 | 100 | 200 | 300 | 100 | 0 | 0
0 | 2 | 500 | 001 | 100 | 200 | 300 | 100 | 200 | 200

Answer 1

This seems very messy, as there's a lot of different things going on here. I don't think I quite understand all of the "rules" for how you are distributing this money but this query does produce the expected results (actually it's slightly different, but I think you have a mistake in your table where you show "200" for amount in the first two rows it should be "700")?
WITH Base AS (
SELECT
t1.idno,
t2.acctno,
t1.amount,
t2.amount1,
t2.amount2,
t2.amount3,
ROW_NUMBER() OVER (ORDER BY t1.idno, t2.acctno) AS row_id
FROM
#table1 t1
INNER JOIN #table2 t2 ON t1.idno = t2.idno),
RunningBalance AS (
SELECT
*,
CASE WHEN amount > amount1 + amount2 + amount3 THEN amount - amount1 - amount2 - amount3 ELSE 0 END AS new_balance
FROM
Base),
NewIdno AS (
SELECT
idno,
MIN(row_id) AS first_row_id
FROM
Base
GROUP BY
idno),
NewBalance AS (
SELECT
n.first_row_id AS row_id,
b.amount
FROM
NewIdno n
INNER JOIN Base b ON b.row_id = n.first_row_id),
Amount1 AS (
SELECT
b.row_id,
rb1.new_balance AS balance,
b.idno,
b.amount,
b.acctno,
b.amount1,
b.amount2,
b.amount3,
CASE WHEN ISNULL(n.amount, rb2.new_balance) >= b.amount1 THEN b.amount1 ELSE b.amount1 - ISNULL(n.amount, rb2.new_balance) END AS pay_amount1,
ISNULL(n.amount, rb2.new_balance) - b.amount1 AS carried_forward_1
FROM
Base b
INNER JOIN RunningBalance rb1 ON rb1.row_id = b.row_id
LEFT JOIN RunningBalance rb2 ON rb2.row_id = b.row_id - 1
LEFT JOIN NewBalance n ON n.row_id = b.row_id),
Amount2 AS (
SELECT
*,
CASE WHEN carried_forward_1 >= amount2 THEN amount2 ELSE carried_forward_1 END AS pay_amount2,
carried_forward_1 - CASE WHEN carried_forward_1 >= amount2 THEN amount2 ELSE carried_forward_1 END AS carried_forward_2
FROM
Amount1),
Amount3 AS (
SELECT
*,
CASE WHEN carried_forward_2 >= amount3 THEN amount3 ELSE carried_forward_2 END AS pay_amount3
FROM
Amount2)
SELECT
balance,
idno,
amount,
acctno,
amount1,
amount2,
amount3,
pay_amount1,
pay_amount2,
pay_amount3
FROM
Amount3;
My results are:
balance idno amount acctno amount1 amount2 amount3 pay_amount1 pay_amount2 pay_amount3
100 1 700 001 100 200 300 100 200 300
100 1 700 002 100 200 300 100 0 0
0 2 500 001 100 200 300 100 200 200
Some of the rules I am using:
you start with the amount from table 1 to be distributed, i.e. £700 for idno #1 and £500 for idno #2;
this is assigned to table 2 by acctnos in numerical order;
where there are multiple acctnos you need to carry forward the balance from the previous one, if there's anything left to pay out;
once you start a new idno you can't carry forward any money left from the previous one.
So how does it work?
Step 1 - Order the data and add an index (row_id)
This is just the base data from the two tables, showing how much we have to distribute and the rows we are distributing it over:
idno acctno amount amount1 amount2 amount3 row_id
1 001 700 100 200 300 1
1 002 700 100 200 300 2
2 001 500 100 200 300 3
Step 2 - Work out the running balance
This works out how much money is left if we distributed the entire amount available across each row:
idno acctno amount amount1 amount2 amount3 row_id new_balance
1 001 700 100 200 300 1 100
1 002 700 100 200 300 2 100
2 001 500 100 200 300 3 0
Step 3 - (Interlude) We need to know which row we will be distributing data over first
This is just the first row_id for each idno:
idno first_row_id
1 1
2 3
Step 4 - Slightly wasteful, as we could have done this in the last step
We just need the total to be distributed over each "first" row:
row_id amount
1 700
3 500
Step 5 - This is where we deal with the running balance properly
For each row the rule is that we start with the "new balance" which only exists for the first row of each amount to be distributed. If this isn't the first row then we use the running balance instead, but we take this from the previous row (rb2.row_id = b.row_id - 1). We will always have one or the other of these:
row_id balance idno amount acctno amount1 amount2 amount3 pay_amount1 carried_forward_1
1 100 1 700 001 100 200 300 100 600
2 100 1 700 002 100 200 300 100 0
3 0 2 500 001 100 200 300 100 400
So the carried forward isn't what is to be carried to the next row, it's what is to be carried to the next amount (amount 2 in this case) to be distributed.
Note that this works for your dataset, but it wouldn't work if there were more than two rows per idno. If you were to have more than two rows per idno then you would simply need to add another stage to handle this scenario.
Step 6 - Carry Forward
For each amount we need to take it off the amount carried forward from the previous calculation, giving us the amount we can allocate to this amount, and the amount to be carried forward to the next calculation (which is redundant for amount 3 as there is no amount 4).