Runing Total sum minus different condition - sql

Question

I looked at some SQL Server running total examples, but I can't manage thing like this.
I have a table where I have columns id, name, type of operation, date, value.
I want to calculate balance for each record. Balance should be calculated like this:
Starting balance must be 0 and then if operation type is IN there will be plus, if operation is OUT there will be minus. Each next record should see previous record balance and then +value or -value depending on operation Type.
This operation should be ordered by date (not Id).
For example, if the table looks like this:
ID Name Op_Type Date Value
1 box Out 2017-05-13 15
2 table In 2017-04-31 65
3 box2 In 2017-05-31 65
then result should look like this
ID Name Op_Type Date Value Balance
2 table In 2017-04-31 65 65
1 box Out 2017-05-13 15 50
3 box2 In 2017-05-31 65 115
result of this code :
select *,
sum(case when Op_Type = 'Out' then -Value else Value end)Over(Order by [Date]) as Balance
From Yourtable
is:
ID Date Type Value Balance
143 2016-12-31 In 980 664.75
89 2016-12-31 Out 300 664.75
90 2016-12-31 Out 80 664.75
But I expect the following result:
ID Date Type Value Balance
143 2016-12-31 In 980 980
89 2016-12-31 Out 300 680
90 2016-12-31 Out 80 600

Answer 1

The problem with answer by Prdp is that SUM(...) OVER (ORDER BY ...) by default uses RANGE option instead of ROW.
This is why you see unexpected results when dates are not unique. This is how the default RANGE option works.
To get results that you expect spell it our explicitly:
SELECT
*
,SUM(CASE WHEN Op_Type = 'Out'
THEN -Value ELSE Value END)
OVER(ORDER BY [Date], Op_Type, ID
ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS Balance
FROM YourTable
ORDER BY [Date], Op_Type, ID;
I also added Op_Type into the ORDER BY to add positive values first in cases when there are several rows with the same date.
I added ID into the ORDER BY to make results stable in cases when there are several rows with the same date.