Is there a way to cumsum value based on the flag and add the value to other flag consecutive rows without iteration in pandas

Question

I’m trying to cumsum ‘value’ for ‘flag’==2, and add those cummulative values to consecutive rows of ‘flag’==1 and dropped flag 2 rows.

Input

Index_Date   flag  value  
 ========================
 2020-01-31     1   10   
 2020-02-01     2    5   
 2020-02-02     1   10   
 2020-02-03     2    5  
 2020-03-04     1   20

Desired output

Index_Date   flag  value  
========================
2020-01-31     1   10   
2020-02-02     1   15  
2020-03-04     1   30

achieved the above result by iteration.

import pandas as pd

df = df.sort_values(by=['Index_Date'])

adjustments = 0
for index, row in df.iterrows():    
    if row['flag'] == 2:
        adjustments += row['value']
    else:
        row['value'] += adjustments

df.drop(df[df['flag'] == 2].index, inplace = True)

Is there a way to achieve this without iteration?
Thanks.

Asked By: Dhaaz

||

Source

Answer 1

Try:

df['value'] += np.where(df['flag'] == 2, df['value'], 0).cumsum()
df = df[df['flag']==1]

Output:

   Index_Date  flag  value
0  2020-01-31     1     10
2  2020-02-02     1     15
4  2020-03-04     1     30

Answered By: luigigi

Answer 2

def function1(dd:pd.DataFrame):
    return dd.tail(1).assign(value=dd.value.sum())

col1=df1.flag.shift().eq(1).cumsum()
df1.groupby(col1,group_keys=False).apply(function1)

Output:

   Index_Date  flag  value
0  2020-01-31     1     10
2  2020-02-02     1     15
4  2020-03-04     1     30

Answered By: G.G

Is there a way to cumsum value based on the flag and add the value to other flag consecutive rows without iteration in pandas

Question:

Answers: