Get rows from table whose sum is less than value - mysql

Question

Consider I have the following rows in the table
| id | user_id | amount | date |
------------------------------------------------
| 1 | 1 | 100 | 2019-09-30 |
------------------------------------------------
| 2 | 2 | 100 | 2019-09-30 |
------------------------------------------------
| 3 | 1 | 100 | 2019-09-30 |
------------------------------------------------
| 4 | 3 | 100 | 2019-10-01 |
------------------------------------------------
| 5 | 1 | 75 | 2019-10-01 |
------------------------------------------------
| 6 | 3 | 100 | 2019-10-01 |
------------------------------------------------
| 7 | 1 | 35 | 2019-10-01 |
------------------------------------------------
I am trying find a way to get all the rows with user_id = 1 where the sum(amount) < 300 and date <= '2019-10-01'.
What I am trying to do is to only process records that meet a certain threshold sum. I am not quite sure where to start.
Expected Result
| id | user_id | amount | date |
------------------------------------------------
| 1 | 1 | 100 | 2019-09-30 |
------------------------------------------------
| 3 | 1 | 100 | 2019-09-30 |
------------------------------------------------
| 5 | 1 | 75 | 2019-10-01 |
------------------------------------------------
Here is what I have tried so far
SELECT id, SUM(amount) as total_sum
FROM table
WHERE date <= '2019-10-01' AND user_id = 1
ORDER BY date ASC
HAVING total_sum <= 300
I don't get the desired output based on the above query.
MySQL Version currently using: 5.7.25
I did look at this question MySQL select records with sum greater than threshold assuming they are trying to do the same thing, but this isn't what I am looking at

Answer 1

It is a Rolling Sum problem. In MySQL 8.0.2 and above, you can solve this using Window functions with Frames. In older versions, we can do the same using User-defined Session variables.
We first calculate the rolling sum using Session variables.
Then, use the result-set in a Derived table, and find the id where total sum crosses the "barrier" of 300. Barrier is reached when the New rolling Sum is greater than 300. We set the barrier value to 1 at this point, 0 for rows before it, and 2 and more, for the rows afterwards.
We will only consider the rows where barrier is 0.
Try (works for all MySQL versions):
Query #1
SELECT dt.id,
dt.user_id,
dt.amount,
dt.date
FROM
(
SELECT
t.id,
t.user_id,
t.amount,
t.date,
#barrier := CASE
WHEN
(#tot_qty := #tot_qty + t.amount) > 300
THEN (#barrier + 1)
ELSE 0
END AS barrier
FROM
your_table AS t
CROSS JOIN (SELECT #tot_qty := 0,
#barrier := 0) AS user_init
WHERE t.user_id = 1
AND t.date <= '2019-10-01'
ORDER BY t.user_id, t.date, t.id
) AS dt
WHERE dt.barrier = 0
ORDER BY dt.user_id, dt.date, dt.id;
Result
| id | user_id | amount | date |
| --- | ------- | ------ | ---------- |
| 1 | 1 | 100 | 2019-09-30 |
| 3 | 1 | 100 | 2019-09-30 |
| 5 | 1 | 75 | 2019-10-01 |
View on DB Fiddle

Answer 2

If you don't like to use Session Variables (some experienced SO users dislike them vehemently), you can utilize a technique based on "Self-Join" and then use GROUP BY with HAVING to filter out.
General idea is that we left join to get previous rows for the specific user_id, and then aggregate to get the rolling sum, and then filtering using Having clause.
Query
SELECT
t1.*
FROM
your_table AS t1
LEFT JOIN your_table AS t2
ON t2.user_id = t1.user_id
AND t2.date <= t1.date
AND t2.id <= t1.id
WHERE t1.user_id = 1
AND t1.date <= '2019-10-31'
GROUP BY t1.user_id, t1.date, t1.id, t1.amount
HAVING COALESCE(SUM(t2.amount),0) < 300;
Result
| id | user_id | amount | date |
| --- | ------- | ------ | ---------- |
| 1 | 1 | 100 | 2019-09-30 |
| 3 | 1 | 100 | 2019-09-30 |
| 5 | 1 | 75 | 2019-10-01 |
View on DB Fiddle
You can benchmark both the approaches and decide which one is suitable.
For this query, you will need the composite index: (user_id, date)