Replace value with a condition from 2 columns using pandas

问题

I have a pandas data-frame like as shown below

df1_new = pd.DataFrame({'person_id': [1, 2, 3, 4, 5],
                        'start_date': ['07/23/2377', '05/29/2477', '02/03/2177', '7/27/2277', '7/13/2077'],
                        'start_datetime': ['07/23/2377  12:00:00', '05/29/2477  04:00:00', '02/03/2177  02:00:00', '7/27/2277  05:00:00', '7/13/2077  12:00:00'],
                        'end_date': ['07/25/2377', '06/09/2477', '02/05/2177', '01/01/2000', '01/01/2000'],
                        'end_datetime': ['07/25/2377 02:00:00', '06/09/2477 04:00:00', '02/05/2177 01:00:00', '01/01/2000 00:00:00', '01/01/2000 00:00:00'],
                        'Type' :['IP','IP','OP','OP','IP']})

What I would like to do is

if ((end_date contains 2000 or end_datetime contains 2000) and (type == IP)) then
     end_date = start_date + 2 days
     end_datetime = start_datetime + 2 days

else ((if end_date contains 2000 or end_datetime contains 2000) and (type == OP)) then 
     end_date = start_date
     end_datetime = start_datetime

This is what I tried but it isn't resulting in accurate output

 df['end_date'] = df['start_date'].apply(lambda x: df['start_date'] + pd.DateOffset(days=2) if (x == 'OP' and x == '01/01/2000') else df['start_date'])
 df['end_datetime'] = df['start_datetime'].apply(lambda x: df['start_datetime'] + pd.DateOffset(days=2) if (x == 'OP' and x == '01/01/2000') else df['start_datetime'])

I expect my output to be like as shown below

回答1:

Here is an example. See comments I think you'll understand the basic approach.

from copy import deepcopy
from datetime import datetime
import pandas as pd
from dateutil.relativedelta import relativedelta


df = pd.DataFrame.from_dict({
    'person_id': [1, 2, 3, 4, 5],
    'start_date': ['07/23/2377', '05/29/2477', '02/03/2177', '7/27/2277', '7/13/2077'],
    'start_datetime': ['07/23/2377  12:00:00', '05/29/2477  04:00:00', '02/03/2177  02:00:00', '7/27/2277  05:00:00', '7/13/2077  12:00:00'],
    'end_date': ['07/25/2377', '06/09/2477', '02/05/2177', '01/01/2000', '01/01/2000'],
    'end_datetime': ['07/25/2377 02:00:00', '06/09/2477 04:00:00', '02/05/2177 01:00:00', '01/01/2000 00:00:00', '01/01/2000 00:00:00'],
    'type': ['IP', 'IP', 'OP', 'OP', 'IP']
})


def calculate_days(x):
    # datetime object from string
    x['end_date'] = datetime.strptime(x['end_date'], '%m/%d/%Y')
    x['start_date'] = datetime.strptime(x['start_date'], '%m/%d/%Y')
    x['end_datetime'] = datetime.strptime(x['end_datetime'], '%m/%d/%Y %H:%M:%S')
    x['start_datetime'] = datetime.strptime(x['start_datetime'], '%m/%d/%Y %H:%M:%S')
    # you need only 2000 year...
    if not (x['end_date'].year == 2000 or x['end_datetime'] == 2000):
        return x

    # type conditions and calculations...
    if x['type'] == 'IP':
        x['end_date'] = x['start_date'] + relativedelta(days=2)
        x['end_datetime'] = x['start_datetime'] + relativedelta(days=2)
    elif x['type'] == 'OP':
        x['end_date'] = deepcopy(x['start_date'])
        x['end_datetime'] = deepcopy(x['start_datetime'])
    return x


# apply our custom function
df = df.apply(calculate_days, axis=1)
print(df.head())
#   person_id           start_date  ...         end_datetime type
# 0          1  2377-07-23 00:00:00  ...  2377-07-25 02:00:00   IP
# 1          2  2477-05-29 00:00:00  ...  2477-06-09 04:00:00   IP
# 2          3  2177-02-03 00:00:00  ...  2177-02-05 01:00:00   OP
# 3          4  2277-07-27 00:00:00  ...  2277-07-27 05:00:00   OP
# 4          5  2077-07-13 00:00:00  ...  2077-07-15 12:00:00   IP
# [5 rows x 6 columns]

Hope this helps.

来源：https://stackoverflow.com/questions/61170220/replace-value-with-a-condition-from-2-columns-using-pandas