问题

Need to drop a sub-column of multi-index data frame created from pivot table

Need to drop a sub-column only at specific columns(month) dynamically

I have a dataframe created from pivot table and need to drop a sub-column at specific columns dynamically...
if todays date is less than 15 i need to drop the sub-column Bill1 for all the months except Sep-19(current month)
if todays date is greater than 15, it should drop the sub-column Bill1 for all the months except Oct-19(next month)

data_frame1 = pd.pivot_table(data_frame, index=['PC', 'Geo', 'Comp'], values=['Bill1', 'Bill2'], columns=['Month'], fill_value=0)
data_frame1 = data_frame1.swaplevel(0,1, axis=1).sort_index(axis=1)
tuples = [(a.strftime('%b-%y'), b) if a!= 'All' else (a,b) for a,b in data_frame1.columns]
data_frame1.columns = pd.MultiIndex.from_tuples(tuples)

output:

              Sep-19             OCT-19        Nov-19
             Bill1 Bill2      Bill1 Bill2     Bill1 Bill2     
PC Geo Comp
A  Ind   OS   1     1.28        1    1.28      1    1.28

desired Output:
if todays date is less than 15

               Sep-19          OCT-19        Nov-19
              Bill1  Bill2       Bill2         Bill2     
PC Geo Comp
A  Ind   OS    1      1.28        1.28          1.28

if todays date is greater than 15

               Sep-19       OCT-19            Nov-19
                Bill2     Bill1  Bill2        Bill2     
PC Geo Comp
A  Ind   OS     1.28        1     1.28         1.28

回答1:

Use:

#convert first level for datetimes and to month periods
level0 = pd.to_datetime(df.columns.get_level_values(0), format='%b-%y').to_period('m')
#get second level
level1 = df.columns.get_level_values(1)
print (level0)
PeriodIndex(['2019-09', '2019-09', '2019-10', '2019-10', '2019-11', '2019-11'],
             dtype='period[M]', freq='M')

print (level1)
Index(['Bill1', 'Bill2', 'Bill1', 'Bill2', 'Bill1', 'Bill2'], dtype='object')

#test for next 15 days
#dat = pd.to_datetime('2019-09-20')
#get today timestamp
dat = pd.to_datetime('now')
print (dat)

#convert timestamp to period
today_per = dat.to_period('m')

#compare day and filter
if dat.day < 15:
    df = df.loc[:, (level0 == today_per) | (level1 != 'Bill1')]
else:
    #test with add 1 month to today period
    df = df.loc[:, (level0 == today_per + 1) | (level1 != 'Bill1')]
print (df)
         Sep-19       Oct-19 Nov-19
          Bill1 Bill2  Bill2  Bill2
A Ind OS      1  1.28   1.28   1.28

Test next month:

#convert first level for datetimes and to month periods
level0 = pd.to_datetime(df.columns.get_level_values(0), format='%b-%y').to_period('m')
#get second level
level1 = df.columns.get_level_values(1)
print (level0)
PeriodIndex(['2019-09', '2019-09', '2019-10', '2019-10', '2019-11', '2019-11'],
             dtype='period[M]', freq='M')

print (level1)
Index(['Bill1', 'Bill2', 'Bill1', 'Bill2', 'Bill1', 'Bill2'], dtype='object')

#test for next 15 days
dat = pd.to_datetime('2019-09-20')
#get today timestamp
#dat = pd.to_datetime('now')
print (dat)

#convert timestamp to period
today_per = dat.to_period('m')

#compare day and filter
if dat.day < 15:
    df = df.loc[:, (level0 == today_per) | (level1 != 'Bill1')]
else:
    #test with add 1 month to today period
    df = df.loc[:, (level0 == today_per + 1) | (level1 != 'Bill1')]
print (df)
         Sep-19 Oct-19       Nov-19
          Bill2  Bill1 Bill2  Bill2
A Ind OS   1.28      1  1.28   1.28

来源：https://stackoverflow.com/questions/57783172/how-to-drop-values-column-of-pivot-table-dataframe

标签

python-3.x