[FIXED] np.timedelta - Unit M is not supported error after updating packages version

Issue

The following code worked fine:

import pandas as pd
import numpy as np
df = pd.DataFrame({'from_date': ['01-10-2003', '15-12-2004', '01-08-2012', '09-07-2001'],
                  'to_date': ['01-11-2003', '15-12-2006', '01-09-2012', '09-12-2001']})
df['from_date'] = pd.to_datetime(df['from_date'], format='%d-%m-%Y')
df['to_date'] = pd.to_datetime(df['to_date'], format='%d-%m-%Y')
df['Months'] = (df['to_date']-df['from_date'])/np.timedelta64(1, 'M')
df

After packages update it gives the following error:

ValueError: Unit M is not supported. Only unambiguous timedelta values durations are supported. Allowed units are 'W', 'D', 'h', 'm', 's', 'ms', 'us', 'ns'

Pandas = 2.1.4, Numpy = 1.26.2

How can I now calculate month difference between two dates?

Solution

If you want to compute an exact difference, convert to_period:

df['Months_period'] = (df['to_date'].dt.to_period('M')
                       .sub(df['from_date'].dt.to_period('M'))
                       .apply(lambda x: x.n)
                      )

Output (compared to using 30 days an approximation for a month):

   from_date    to_date  Months_30D  Months_period
0 2003-10-01 2003-11-01    1.033333              1
1 2004-12-15 2006-12-15   24.333333             24
2 2012-08-01 2012-09-01    1.033333              1
3 2001-07-09 2001-12-09    5.100000              5
4 2000-01-01 2023-01-01  280.033333            276

Answered By - mozway