login counter

import pandas as pd
 
 
# Sample dataset
test_df = pd.DataFrame({'user_id': ['A', 'B', 'C', 'A', 'D', 'E', 'A', 'B']
                        , 'app_open_time':['2021-06-20 23:40:32', '2021-06-20 23:50:32', '2021-06-21 23:40:32'
                                           , '2021-06-20 23:45:32', '2021-06-23 15:20:10', '2021-06-23 15:25:10'
                                           , '2021-06-20 23:53:10', '2021-06-21 12:45:20' ]})
 
# The final dataframe should be created based on the following constraints:
# Datetime values within 5 minutes(<= 5 minutes) of each other should be counted as 1 for each user id
# Datetime values should be compared only within an user id instead of across user ids
# Intended output is given below
final_df = pd.DataFrame({'user_id': ['A', 'B', 'C', 'D', 'E']
                        , 'app_opens': [2, 2, 1, 1, 1]})

df = test_df.copy()

df['app_open_time'] = pd.to_datetime(df['app_open_time'])
def calc_unique_logins(series):
    diffs = series.diff().dt.seconds.fillna(0)
    counter = 1
    placeholder = 0
    for offset in diffs:
        if offset - placeholder > 5 * 60: # 5 min
            counter += 1
            placeholder = offset
    return counter
    

df.groupby('user_id')['app_open_time'].apply(calc_unique_logins)