Pandas Dataframe Creating a New Column Based on Other Columns

Question

I have a dataframe(df) as following:

    Customer    Date_Begin  Date_End    Product
0   1   2017-01-01  2020-01-01  UsagePeriod
1   1   2017-01-02  2018-01-02  Token
2   1   2017-01-06  2018-01-06  Token
3   1   2018-12-01  2019-12-01  Token
4   1   2019-06-01  2020-06-01  Token
5   1   2019-12-21  2022-12-21  UsagePeriod
6   1   2020-01-31  2021-01-31  Token
7   1   2021-06-30  2022-06-30  Token
8   1   2021-09-30  2022-09-30  Token
9   2   2019-06-01  2022-06-01  UsagePeriod
10  2   2019-06-01  2020-06-01  Token
11  3   2019-06-01  2022-06-01  UsagePeriod
12  3   2019-06-01  2020-06-01  Token
13  3   2020-06-01  2021-06-01  Token
14  4   2017-01-01  2020-01-01  UsagePeriod
15  4   2017-02-15  2018-02-15  Token
16  4   2019-12-15  2020-12-15  Token

There are some rules to be applied to reach the desired output dataframe:

1. Each customer has a UsagePeriod where (s)he can buy tokens. Normally the tokens are expired in a year(at most) and the UsagePeriod is always 3 years from the Date_Begin. Date_End column gives the expected dates of expiry.
2. A customer would not buy a new token unless the one (s)he has expired. Meaning that (s)he has run out of that token if (s)he is buying a new one. Assuming that the customer bought the new token on the expiry of the previous one, I want to open an additional column containing all the expiry dates.
3. The tokens can only be used in the defined UsagePeriod of the customer unless the customer buys a new Usage period. The output dataframe that I want to reach is:

    Customer    Date_Begin  Date_End    Product Expiry
0   1   2017-01-01  2020-01-01  UsagePeriod 2020-01-01
1   1   2017-01-02  2018-01-02  Token   2017-01-06
2   1   2017-01-06  2018-01-06  Token   2018-01-06
3   1   2018-12-01  2019-12-01  Token   2019-06-01
4   1   2019-06-01  2020-06-01  Token   2019-12-21
5   1   2019-12-21  2022-12-21  UsagePeriod 2022-12-21
6   1   2020-01-31  2021-01-31  Token   2021-01-31
7   1   2021-06-30  2022-06-30  Token   2021-09-30
8   1   2021-09-30  2022-09-30  Token   2022-09-30
9   2   2019-06-01  2022-06-01  UsagePeriod 2022-06-01
10  2   2019-06-01  2020-06-01  Token   2020-06-01
11  3   2019-06-01  2022-06-01  UsagePeriod 2022-06-01
12  3   2019-06-01  2020-06-01  Token   2020-06-01
13  3   2020-06-01  2021-06-01  Token   2021-06-01
14  4   2017-01-01  2020-01-01  UsagePeriod 2020-01-01
15  4   2017-02-15  2018-02-15  Token   2018-02-15
16  4   2019-12-15  2020-12-15  Token   2020-01-01
            

The code that I currently have is:

def myFunc(df):
                                                                         
for i in range(0,len(df)-1): #loop through each row of dataframe
    if df.Product.loc[i] == "UsagePeriod": #When product equals "UsagePeriod"
      df['expiricy'].loc[i] = df['Date_Begin'].loc[i]  + pd.offsets.DateOffset(years=3); #update new column
    elif df.Product.loc[i] == "Token": #When product equals "Token"
      if df.Customer.loc[i] == df.Customer.loc[i+1]: #customer changed?
        if abs(df.Date_Begin.loc[i+1]-df.Date_Begin.loc[i])>= pd.Timedelta(364, 'd'): #has year increased by more than 1(year)? product = product
          df['expiricy'].loc[i] = df['Date_Begin'].loc[i]  + pd.offsets.DateOffset(years=1); #update new column
        elif abs(df.Date_Begin.loc[i+1]-df.Date_Begin.loc[i]) < pd.Timedelta(364, 'd'): #When less than a year?
          df['expiricy'].loc[i] = df['Date_Begin'].loc[i+1];

I am struggling with defining the tokens end period in the UsagePeriod. Any help would be appreciated.

Answer 1

First convert the date columns to datetime objects. You will need this for valid comparisons.

df["Date_Begin"] = pd.to_datetime(df["Date_Begin"])
df["Date_End"] = pd.to_datetime(df["Date_End"])

All the rows with Token as product are shifted backwards in groups of customers:

df["Expiry"] = df.loc[df["Product"].eq("Token"), "Date_Begin"].groupby(df["Customer"]).shift(-1)

The results:

    Customer Date_Begin   Date_End      Product     Expiry
0          1 2017-01-01 2020-01-01  UsagePeriod        NaT
1          1 2017-01-02 2018-01-02        Token 2017-01-06
2          1 2017-01-06 2018-01-06        Token 2018-12-01
3          1 2018-12-01 2019-12-01        Token 2019-06-01
4          1 2019-06-01 2020-06-01        Token 2020-01-31
5          1 2019-12-21 2022-12-21  UsagePeriod        NaT
6          1 2020-01-31 2021-01-31        Token 2021-06-30
7          1 2021-06-30 2022-06-30        Token 2021-09-30
8          1 2021-09-30 2022-09-30        Token        NaT
9          2 2019-06-01 2022-06-01  UsagePeriod        NaT
10         2 2019-06-01 2020-06-01        Token        NaT
11         3 2019-06-01 2022-06-01  UsagePeriod        NaT
12         3 2019-06-01 2020-06-01        Token 2020-06-01
13         3 2020-06-01 2021-06-01        Token        NaT
14         4 2017-01-01 2020-01-01  UsagePeriod        NaT
15         4 2017-02-15 2018-02-15        Token 2019-12-15
16         4 2019-12-15 2020-12-15        Token        NaT

Then we compare Date_End with the current Expiry. if Date_End is sooner we use date end else Expiry.

df["Expiry"] = df["Expiry"].where(df["Expiry"].lt(df["Date_End"]), df["Date_End"])

This also sets the values where the product is Usage Period to Date_End values.

Customer Date_Begin   Date_End      Product     Expiry
0          1 2017-01-01 2020-01-01  UsagePeriod 2020-01-01
1          1 2017-01-02 2018-01-02        Token 2017-01-06
2          1 2017-01-06 2018-01-06        Token 2018-01-06
3          1 2018-12-01 2019-12-01        Token 2019-06-01
4          1 2019-06-01 2020-06-01        Token 2020-01-31
5          1 2019-12-21 2022-12-21  UsagePeriod 2022-12-21
6          1 2020-01-31 2021-01-31        Token 2021-01-31
7          1 2021-06-30 2022-06-30        Token 2021-09-30
8          1 2021-09-30 2022-09-30        Token 2022-09-30
9          2 2019-06-01 2022-06-01  UsagePeriod 2022-06-01
10         2 2019-06-01 2020-06-01        Token 2020-06-01
11         3 2019-06-01 2022-06-01  UsagePeriod 2022-06-01
12         3 2019-06-01 2020-06-01        Token 2020-06-01
13         3 2020-06-01 2021-06-01        Token 2021-06-01
14         4 2017-01-01 2020-01-01  UsagePeriod 2020-01-01
15         4 2017-02-15 2018-02-15        Token 2018-02-15
16         4 2019-12-15 2020-12-15        Token 2020-12-15