I have a column of my dataframe that is made up of the following:

df['Year] = [2025, 2024, NaN, 2023, 2026, NaN] (these are type float64)

How can I convert these years to something in datetime format? Since there are no months or days included I feel like they have to output as [01-01-2025, 01-01-2021, NaT, 01-01-2023, 01-01-2026, NaT] by default.

But if there was a way to still have the column as [2025, 2024, NaT, 2023, 2026, NaT] then that would work well too.

Using df['Year'] = pd.DatetimeIndex(df['Year']).year just output [1970, 1970, NaN, 1970, 1970, NaN].

Thank you very much.

CodePudding user response：

You can use pandas' to_datetime() and set errors='coerce' to take care of the NaNs (-> NaT)

df['Year'] = pd.to_datetime(df['Year'], format='%Y', errors='coerce')

The output is going to be like 01-01-2025, 01-01-2021 ...

CodePudding user response：

Probably not the most elegant solution but if you convert the column to string and fill the empty with a dummy year (say 1900) you can use parser from dateutil

from dateutil import parser

('01/01/' df['year']).fillna('1900').apply(parser.parse)

Out[67]: 0 2025-01-01 1 2024-01-01 2 1900-07-21 3 2023-01-01 4 2026-01-01 5 1900-07-21