Calculate new value in MariaDB based on values in the previous row

Question

In MariaDB 10.3, I have a table like this:

id	date	sum
1	2020-01-01	120
1	2020-02-01	130
1	2020-03-01	140
1	2020-04-01	150
1	2020-05-01	160
1	2020-06-01	170

I need to calculate the remaining sum after every date. Something like this. I need the calculation to happen as a query in the MariaDB.

id	date	sum	remaining total sum before pay
1	2020-01-01	120	870
1	2020-02-01	130	750
1	2020-03-01	140	620
1	2020-04-01	150	480
1	2020-05-01	160	330
1	2020-06-01	170	170

I found a few close solutions but can't alter them in a proper way to suit my need. Any ideas would be appreciated.

The logic for the last column is:

1. The first value is the whole sum owed (120+130+140...)
2. Every next value is calculated from that number decreased by the value in the sum column on the same row ie

• 870-120=750;
• 750-130=620;
• 620-140=480; etc.

My attempt was:

1. First queries which were close, but didn't work:

    SELECT 
    id,s.sum,s.date,
    @b := @b + s.sum   AS balance
    FROM
    (SELECT @b := 0.0) AS dummy 
      CROSS JOIN
        tpp AS s
        where id=1
    ORDER BY
        s.date ;

The result was:

id	date	sum	remaining total sum before pay
1	2020-01-01	120	120
1	2020-02-01	130	250
1	2020-03-01	140	380
1	2020-04-01	150	520
1	2020-05-01	160	670
1	2020-06-01	170	840

Ie it kind of reversed the result. And increased the value with the value on the next row.

2. The other query was with the LAG function but the subtracting part was disappointing

    SELECT 
    id,date,
    sum(sum)-LAG(sum) OVER (ORDER BY date) AS l
    FROM tpp
    where id=1
    group by date,id
    ORDER BY date

And the result:

id	date	remaining total sum before pay
1	2020-01-01	null
1	2020-02-01	10
1	2020-03-01	10
1	2020-04-01	10
1	2020-05-01	10
1	2020-06-01	10

It subtracted:

• 130-120=10;
• 140-130=10;
• 150-140=10; etc.

Answer 1

You need only SUM() window function:

SELECT id, date,
       SUM(sum) OVER (ORDER BY date DESC) AS l
FROM tpp
WHERE id = 1
ORDER BY date

See the demo .
Results:

> id | date       |   l
> -: | :--------- | --:
>  1 | 2020-01-01 | 870
>  1 | 2020-02-01 | 750
>  1 | 2020-03-01 | 620
>  1 | 2020-04-01 | 480
>  1 | 2020-05-01 | 330
>  1 | 2020-06-01 | 170