How to make a Datetimeindex not be the index in a dataframe

Question

I’d like to have access to the special methods provided by the Datetimeindex class such as month, day, etc. However I can’t seem to make a series in a dataframe be a Datetimeindex without making it the dataframe’s index. Take the following example:

dates

Out[119]:
         Dates
0     1/1/2012
1     1/2/2012
2     1/3/2012
3     1/4/2012
4     1/5/2012
5     1/6/2012
6     1/7/2012
7     1/8/2012
8     1/9/2012
9    1/10/2012
10  12/31/2012

date_series = pd.DatetimeIndex(dates.Dates)
date_series.month
Out[115]: array([ 1,  1,  1,  1,  1,  1,  1,  1,  1,  1, 12])

dates.Dates = pd.DatetimeIndex(dates.Dates)
dates.Dates.month
AttributeError: 'Series' object has no attribute 'month'

I also tried converting the series to timestamps using pd.to_datetime but it still doesn’t work.

I know I can work around this, but it seems like this functionality should exist?

Asked By: ChrisArmstrong

||

Source

Answer 1

Update: in 0.15 you will have access to a dt attribute for datetimelike methods:

dates.Dates.dt.month

Old post (do use Wes’ solution and not this):

Here’s one (slow!) workaround to do it using apply, not ideal but it works:

In [11]: from pandas.lib import Timestamp

In [12]: df.Dates.apply(lambda x: Timestamp(x).month)
Out[12]: 
0      1
1      1
2      1
3      1
4      1
5      1
6      1
7      1
8      1
9      1
10    12
Name: Dates

It seems like a bug (that you can’t do apply(lambda x: x.month)), perhaps worth adding as an issue on github. As Wes would say: “welcome to hell”.

Answered By: Andy Hayden

Answer 2

For now I would suggest doing pd.DatetimeIndex(dates.Dates).month. I’ve been debating whether to add a bunch of data type-specific attributes to Series that will only work for timestamps, but haven’t done it yet.

Answered By: Wes McKinney

How to make a Datetimeindex not be the index in a dataframe

Question:

Answers: