ENH: replace method for more time-related type

[GYHHAHA]

Now we can only implement replace on Timestamp, I think TimedeltaIndex, Timedelta, DatetimeIndex and even Series with datetime64[ns] or timedelta64[ns] all need replace method (works like replace on Timestamp). Is there any plan to do this? (Maybe in pandas 2.0) Thanks!

[jreback]

pls show exactly what you mean

replace is a Timestamp method (inherited from datetime.datetime)

and a Series / index method

[GYHHAHA]

>>>pd.Timestamp('20200101').replace(day=5)
Timestamp('2020-01-05 00:00:00')

Here replace change the day to 5.

>>>ts = pd.date_range('20200101','20200501',freq='MS').to_series()
>>>ts
2020-01-01   2020-01-01
2020-02-01   2020-02-01
2020-03-01   2020-03-01
2020-04-01   2020-04-01
2020-05-01   2020-05-01
Freq: MS, dtype: datetime64[ns]

I want to get this after use replace.

>>>ts.replace(day=5)
2020-01-01   2020-01-05
2020-02-01   2020-02-05
2020-03-01   2020-03-05
2020-04-01   2020-04-05
2020-05-01   2020-05-05
Freq: MS, dtype: datetime64[ns]

Equal to ts.apply(lambda x:x.replace(day=5)).

>>>pd.date_range('20200101','20200501',freq='MS')
DatetimeIndex(['2020-01-01', '2020-02-01', '2020-03-01', '2020-04-01',
               '2020-05-01'],
              dtype='datetime64[ns]', freq='MS')
>>>pd.date_range('20200101','20200501',freq='MS').replace(day=5)
DatetimeIndex(['2020-01-05', '2020-02-05', '2020-03-05', '2020-04-05',
               '2020-05-05'],
              dtype='datetime64[ns]', freq='MS')

Also for the index type.

[jreback]

try .dt.replace

[GYHHAHA]

>>>pd.date_range('20200101','20200501',freq='MS').to_series().dt.replace(day=5)
AttributeError: 'DatetimeProperties' object has no attribute 'replace'

this raises error

[jreback]

ok i suppose

though to be honest .replace is not very useful and has a lot of tz implications

much better to simply use offsets

[GYHHAHA]

I see.

But since in here DateOffset arguments that do not end in s is suggested to be deprecated in the future, how to do this (.replace(day=5)) by offset object?

We do this by ts + pd.DateOffset(day=5) formerly.

[hornstrandir]

Is this still under discussion? I would also find this method very helpful.