CODEWITHSUNDEEP
pythonpandasdataframecumsum
user3287545
user3287545

Reputation: 2061

Pandas Dataframe cumsum by row in reverse column order?

Is there a more elegant way to achieve this? my current solution based on various stackoverflow answers is as following 

df = pds.DataFrame([[11,12,13,14],[15,16,17,18]], columns = [0,1,2,3])  
print df  
dT = df.T  
dT.reindex(dT.index[::-1]).cumsum().reindex(dT.index).T

Output

df is:
    0   1   2   3  
0  11  12  13  14  
1  15  16  17  18  

after by row reverse cumsum
    0   1   2   3
0   50  39  27  14  
1   66  51  35  18

I have to perform this often on my data (much bigger size also), and try to find out a short/better way to do achieve this.

Thanks

Upvotes: 3

Views: 4868

Answers (1)

BrenBarn
BrenBarn

Reputation: 251408

Here is a slightly more readable alternative:

df[df.columns[::-1]].cumsum(axis=1)[df.columns]

There is no need to transpose your DataFrame; just use the axis=1 argument to cumsum.

Obviously the easiest thing would be to just store your DataFrame columns in the opposite order, but I assume there is some reason why you're not doing that.

Upvotes: 6

Related Questions