A way to have a rolling summation - mysql

Question

I have the below dataset. In the below example records for the year 1993. The Tgrowth column is start - end. Started is the number of employees that joined on a specific month and ended is the number of employees that left for the same month.
SELECT
r.Tgrowth,
CASE
WHEN t.mon_num = 1 THEN 'JAN'
WHEN t.mon_num = 2 THEN 'FEB'
WHEN t.mon_num = 3 THEN 'MAR'
WHEN t.mon_num = 4 THEN 'APR'
WHEN t.mon_num = 5 THEN 'MAY'
WHEN t.mon_num = 6 THEN 'JUN'
WHEN t.mon_num = 7 THEN 'JUL'
WHEN t.mon_num = 8 THEN 'AUG'
WHEN t.mon_num = 9 THEN 'SEP'
WHEN t.mon_num = 10 THEN 'OCT'
WHEN t.mon_num = 11 THEN 'NOV'
WHEN t.mon_num = 12 THEN 'DEC'
END AS myMONTH
FROM
(SELECT 1 mon_num UNION SELECT 2 UNION SELECT 3 UNION SELECT 4 UNION SELECT 5 UNION SELECT 6 UNION SELECT 7 UNION SELECT 8 UNION SELECT 9 UNION SELECT 10 UNION SELECT 11 UNION SELECT 12) t
LEFT JOIN Reports r ON t.mon_num = r.theMONTH
AND r.Tyear = 1993
GROUP BY r.Tgrowth , myMONTH
ORDER BY t.mon_num ASC
The result set for the above is as follows,
Tgrowth Month
1 JAN
0 FEB
2 MAR
0 APR
0 MAY
0 JUN
0 JUL
0 AUG
0 SEP
0 OCT
0 NOV
0 DEC
Instead I would like the result to show a rolling sum i.e. add to the Tgrowth field. Something like the below,
growth Emp_Count myMONTH
1 1 JAN
0 1 FEB
2 3 MAR
0 3 APR
0 3 MAY
0 3 JUN
0 3 JUL
0 3 AUG
0 3 SEP
0 3 OCT
0 3 NOV
0 3 DEC

Answer 1

There are 2 options:
use join
use variables
The method of using join is as following:
SELECT
t1.Tgrowth,
sum(t2.Tgrowth) as Emp_Count,
CASE
WHEN t1.Month = 1 THEN 'JAN'
WHEN t1.Month = 2 THEN 'FEB'
WHEN t1.Month = 3 THEN 'MAR'
WHEN t1.Month = 4 THEN 'APR'
WHEN t1.Month = 5 THEN 'MAY'
WHEN t1.Month = 6 THEN 'JUN'
WHEN t1.Month = 7 THEN 'JUL'
WHEN t1.Month = 8 THEN 'AUG'
WHEN t1.Month = 9 THEN 'SEP'
WHEN t1.Month = 10 THEN 'OCT'
WHEN t1.Month = 11 THEN 'NOV'
WHEN t1.Month = 12 THEN 'DEC'
END AS myMONTH
FROM (
SELECT
case
when r.growth is not null then r.growth
when r.growth is null then 0
END as Tgrowth,
t.mon_num AS Month
FROM
(SELECT 1 mon_num UNION SELECT 2 UNION SELECT 3 UNION SELECT 4
UNION SELECT 5 UNION SELECT 6 UNION SELECT 7 UNION SELECT 8
UNION SELECT 9 UNION SELECT 10 UNION SELECT 11 UNION SELECT 12) t
LEFT JOIN Reports r ON t.mon_num = r.themonth
AND r.theYear = 1993
GROUP BY r.growth , Month
ORDER BY t.mon_num ASC
) as t1 join (
SELECT
case
when r.growth is not null then r.growth
when r.growth is null then 0
END as Tgrowth,
t.mon_num AS Month
FROM
(SELECT 1 mon_num UNION SELECT 2 UNION SELECT 3 UNION SELECT 4
UNION SELECT 5 UNION SELECT 6 UNION SELECT 7 UNION SELECT 8
UNION SELECT 9 UNION SELECT 10 UNION SELECT 11 UNION SELECT 12) t
LEFT JOIN Reports r ON t.mon_num = r.themonth
AND r.theYear = 1993
GROUP BY r.growth , Month
ORDER BY t.mon_num ASC
) as t2 on t1.Month >= t2.Month group by t1.Month;
Use variables solution is as following:
SET #num := 0;
select
Tgrowth,
#num := #num + Tgrowth as Emp_Count,
CASE
WHEN t1.Month = 1 THEN 'JAN'
WHEN t1.Month = 2 THEN 'FEB'
WHEN t1.Month = 3 THEN 'MAR'
WHEN t1.Month = 4 THEN 'APR'
WHEN t1.Month = 5 THEN 'MAY'
WHEN t1.Month = 6 THEN 'JUN'
WHEN t1.Month = 7 THEN 'JUL'
WHEN t1.Month = 8 THEN 'AUG'
WHEN t1.Month = 9 THEN 'SEP'
WHEN t1.Month = 10 THEN 'OCT'
WHEN t1.Month = 11 THEN 'NOV'
WHEN t1.Month = 12 THEN 'DEC'
END AS myMONTH
from (
SELECT
case
when r.growth is not null then r.growth
when r.growth is null then 0
END as Tgrowth,
t.mon_num AS Month
FROM
(SELECT 1 mon_num UNION SELECT 2 UNION SELECT 3 UNION SELECT 4
UNION SELECT 5 UNION SELECT 6 UNION SELECT 7 UNION SELECT 8
UNION SELECT 9 UNION SELECT 10 UNION SELECT 11 UNION SELECT 12) t
LEFT JOIN Reports r ON t.mon_num = r.themonth
AND r.theYear = 1993
GROUP BY r.growth , Month
ORDER BY t.mon_num ASC ) t1;

Answer 2

Since you are running MySQL 8.0, I would recommend a recursive query to generate the dates, and then window functions and aggregation.
If you want the whole 1993 year:
with dates as (
select '1993-01-01' dt
union all
select dt + interval 1 month from dates where dt < '1993-12-01'
)
select
date_format(d.dt, '%b') mymonth,
coalesce(sum(started), 0) - coalesce(sum(ended), 0) growth,
sum(coalesce(sum(started), 0) - coalesce(sum(ended), 0)) over(order by d.dt) emp_count
from dates d
left join reports r on r.theDate >= d.dt and r.theDate < d.dt + interval 1 month
group by d.dt
order by d.dt
This assumes that theDate is stored as a date datatype and not a string (else, you would need to convert it first, using str_to_date()).
This also takes in account the possibility that the table may contain several rows for a given month. If that's not the case, then there is no need for aggregation:
with dates as (
select '1993-01-01' dt
union all
select dt + interval 1 month from dates where dt < '1993-12-01'
)
select
date_format(d.dt, '%b') mymonth,
coalesce(started, 0) - coalesce(ended, 0) growth,
sum(coalesce(started, 0) - coalesce(ended, 0)) over(order by d.dt) emp_count
from dates d
left join reports r on r.theDate >= d.dt and r.theDate < d.dt + interval 1 month
order by d.dt