I have columns amount & assets. Column target should be the cumsum of amount, but the sum should be reset to the current amount if the previous assets was equal to zero.
Sample:
+--------+--------+--------+
| amount | assets | target |
+--------+--------+--------+
| 6 | 10 | 6 |
| 8 | 20 | 14 |
| -1 | 0 | 13 |
| 6 | 1 | 6 |
| -7 | 0 | -1 |
| 2 | 4 | 2 |
| -5 | 7 | -3 |
| 3 | 9 | 0 |
| 7 | 0 | 7 |
| 9 | 2 | 9 |
| 1 | 3 | 10 |
| -4 | 5 | 6 |
+--------+--------+--------+
Use GroupBy.cumsum with groups created by compare column by 0 with shifting Series.shift, processing first NaN and Series.cumsum:
g = df['assets'].eq(0).shift().bfill().cumsum()
#alternative
#g = df['assets'].eq(0).shift(fill_value=0).cumsum()
df['new'] = df.groupby(g)['amount'].cumsum()
print (df)
amount assets target new
0 6 10 6 6
1 8 20 14 14
2 -1 0 13 13
3 6 1 6 6
4 -7 0 -1 -1
5 2 4 2 2
6 -5 7 -3 -3
7 3 9 0 0
8 7 0 7 7
9 9 2 9 9
10 1 3 10 10
11 -4 5 6 6