Pandas 如果为正，则将负值与先前值相加

Question

It is bit hard to put my problem in words. I have a dataframe with positive and negative values.

2012-01-01    58.0
2012-06-01     8.0
2012-07-01    10.0
2013-01-01    50.0
2013-02-01    -6.0
2013-03-01    -8.0
2013-04-01    20.0
2013-07-01     3.0
2013-12-01     0.0
2014-02-01    88.0
2014-03-01   -40.0

I want to sum a negative value in a row with the previous row value if it is positive until no negatives are left.

For example: the final list should be: [58,8,10, 50+(-6-8),20.0, 3.0, 0.0, 88+(-40)]

2012-01-01    58.0
2012-06-01     8.0
2012-07-01    10.0
2013-01-01    36.0
2013-04-01    20.0
2013-07-01     3.0
2013-12-01     0.0
2014-02-01    48.0

The dataframe is huge so i would really prefer a pandas solution.

Answer 1

You can identify the negative blocks with cumsum , and use that for groupby:

(df.groupby(df['value'].ge(0).cumsum(), as_index=False)
   .agg({'date':'first','value':'sum'})
)

Output:

         date  value
0  2012-01-01   58.0
1  2012-06-01    8.0
2  2012-07-01   10.0
3  2013-01-01   36.0
4  2013-04-01   20.0
5  2013-07-01    3.0
6  2013-12-01    0.0
7  2014-02-01   48.0