Calculate login duration by month

This is my initial dataframe:

How do I get the following result in pandas?

Do I firstly have to split it into a login and a logout dataframe and then joining it?

You could use a custom function to split the months:

def split_months(start, end):
    return (pd.Series([start, end, *pd.date_range(start, end, freq='M')])
              .drop_duplicates().sort_values(ignore_index=True)
              .to_frame(name='date')
              .assign(TimeInDays=lambda d: d['date'].diff(),
                      Month=lambda d: d.pop('date').dt.month_name()
                     )
              .iloc[1:]
           )

tmp = df.sort_values(by=['ID', 'Datetime'])

out = (tmp
  .groupby(['ID', df['Event'].eq('Login').cumsum()])
  .apply(lambda g: split_months(g['Datetime'].min(), g['Datetime'].max()))
  .reset_index('ID')
)

NB. if you have several years, you might want to use Month=lambda d: d.pop('date').dt.to_period('M') to avoid ambiguity. If you want an integer for the number of days, use TimeInDays=lambda d: d['date'].diff().dt.days.

Output:

   ID TimeInDays  Month
0   1    21 days  March
1   1     5 days  April
2   1     4 days    May
3   2     2 days  April