SQL query to calculate table based on previous rows

Question

I need a query to transform a table from tab A to Table B: I used windowing function like Lag but it couldn't help table 1:

Concat_date_hour	Type	Value	Next_Movement
2021092414	D	16923	63
2021092415	D	0	149
2021092415	D	0	-56
2021092415	D	0	-131
2021092415	D	0	-79

Output table:

Concat_date_hour	Type	Actual Value	New value
2021092414	D	16923	16986
2021092415	D	16986	17135
2021092415	D	17135	17079
2021092415	D	17079	16948
2021092415	D	16948	16869

PS type can be different from D

Answer 1

As an ordering column is missing in the source table it will be created on the fly. Other parts of the query depend on that rn arbitrary ordering column.

select Concat_date_hour, Type,
  --rn, Value, Next_Movement, 
  rt - Next_Movement actual, rt [new value]
from (
  select *, sum(Value + Next_Movement) over (partition by Type order by Concat_date_hour, rn) rt
  from (
     select  *, row_number() over(partition by Type, Concat_date_hour order by Concat_date_hour) rn
     from tbl ) t
) t
order by Type, Concat_date_hour, rn;

db<>fiddle Note the extra row added to the original sample data.