Sum until threshold value reached and then reset the counter

Question

user_id | date                 | distance
1       | 2019-04-09 00:00:00  | 2
1       | 2019-04-09 00:00:30  | 5
1       | 2019-04-09 00:01:00  | 3
1       | 2019-04-09 00:01:45  | 7
1       | 2019-04-09 00:02:30  | 6
1       | 2019-04-09 00:03:00  | 1

How do I sum distance over next row until threshold point is reached and reset the counter again.

For instance if the threshold value is 10 I am trying to get the following output:

1       | 2019-04-09 00:00:00  | 2
1       | 2019-04-09 00:00:30  | 7            (2 + 5)
1       | 2019-04-09 00:01:00  | 10           ( 7 + 3 )
1       | 2019-04-09 00:01:45  | 7            RESET
1       | 2019-04-09 00:02:30  | 13           (7 + 6 )
1       | 2019-04-09 00:03:00  | 1            RESET

But all I could achieve is get cumulative distance with following query:

SELECT *, sum(distance) over (order by date asc) as running_distance FROM table;

I am using PostgreSQL.

Answer 1

Use user-defined aggregate

Live test: http://sqlfiddle.com/#!17/16716/2

SELECT *, sum_with_reset(distance, 10) over (order by date asc) as running_distance 
FROM tbl;

User-defined aggregate sum_with_reset definition:

create or replace function sum_reset_accum(
    _accumulated numeric, _current numeric, _threshold numeric
)
returns numeric as
$$
    select case when _accumulated >= _threshold then
        _current
    else
        _current + _accumulated
    end    
$$ language sql;


create aggregate sum_with_reset(numeric, numeric)
(
    sfunc = sum_reset_accum,
    stype = numeric,
    initcond = 0
);

Data

CREATE TABLE tbl
    ("user_id" int, "date" timestamp, "distance" int)
;

INSERT INTO tbl
    ("user_id", "date", "distance")
VALUES
    (1, '2019-04-09 00:00:00', 2),
    (1, '2019-04-09 00:00:30', 5),
    (1, '2019-04-09 00:01:00', 3),
    (1, '2019-04-09 00:01:45', 7),
    (1, '2019-04-09 00:02:30', 6),
    (1, '2019-04-09 00:03:00', 1)
;

Output:

| user_id |                 date | distance | running_distance |
|---------|----------------------|----------|------------------|
|       1 | 2019-04-09T00:00:00Z |        2 |                2 |
|       1 | 2019-04-09T00:00:30Z |        5 |                7 |
|       1 | 2019-04-09T00:01:00Z |        3 |               10 |
|       1 | 2019-04-09T00:01:45Z |        7 |                7 |
|       1 | 2019-04-09T00:02:30Z |        6 |               13 |
|       1 | 2019-04-09T00:03:00Z |        1 |                1 |

One-liner:

create or replace function sum_reset_accum(
    _accumulated numeric, _current numeric, _threshold numeric
)
returns numeric as
$$
    select _current + _accumulated * (_accumulated < _threshold)::int
$$ language 'sql';

Postgres boolean can cast true to 1, false to 0 by using cast operator ::int.

You can use plpgsql language too:

create or replace function sum_reset_accum(
    _accumulated numeric, _current numeric, _threshold numeric
)
returns numeric as
$$begin
    return _current + _accumulated * (_accumulated < _threshold)::int;
end$$ language 'plpgsql';

Note that you cannot create plpgsql function on sqlfiddle.com, so you cannot test that plpgsql code on sqlfiddle.com. You can, on your machine though.

Answer 2

As variant you can use RECURSIVE CTE to get it

Test table:

CREATE TABLE public.table_b (
    user_id int4 NULL,
    "date" timestamp NULL,
    distance int4 NULL
);

INSERT INTO public.table_b (user_id,"date",distance) VALUES 
(1,'2019-04-09 00:00:00.000',2)
,(1,'2019-04-09 00:00:30.000',5)
,(1,'2019-04-09 00:01:00.000',3)
,(1,'2019-04-09 00:01:45.000',7)
,(1,'2019-04-09 00:02:30.000',6)
,(1,'2019-04-09 00:03:00.000',1)
,(2,'2019-04-09 00:00:00.000',2)
,(2,'2019-04-09 00:00:30.000',5)
,(2,'2019-04-09 00:01:00.000',3)
,(2,'2019-04-09 00:01:45.000',7)
,(2,'2019-04-09 00:02:30.000',6)
,(2,'2019-04-09 00:03:00.000',1);

Query:

WITH RECURSIVE cte1 AS(
    SELECT *,ROW_NUMBER()OVER(PARTITION BY user_id ORDER BY date) n
    FROM table_b
),
cte2 AS(
    SELECT user_id,date,distance,distance running_distance,n
    FROM cte1
    WHERE n=1

    UNION ALL

    SELECT c1.user_id,c1.date,c1.distance,CASE WHEN c2.running_distance<10 THEN c2.running_distance ELSE 0 END+c1.distance running_distance,c1.n
    FROM cte1 c1
    JOIN cte2 c2 ON c2.user_id=c1.user_id AND c2.n+1=c1.n
)
SELECT user_id,date,distance,running_distance
FROM cte2
ORDER BY user_id,date