"DataFrame object has no attribute dt" error-CodePudding

I have the following dataframe with some dates (datetime64[ns]):

            A6U20      A6Z20      A6H21      A6M21      A6U21      A6Z21
expiry 2020-09-14 2020-12-14 2021-03-15 2021-06-14 2021-09-13 2021-12-13

I want to create another dataframe with the number of days between each column and the preceding one. I tried the following:

df2 = (df - df.shift(axis=1))

which resulted in:

       A6U20   A6Z20   A6H21   A6M21   A6U21   A6Z21
expiry   NaT 91 days 91 days 91 days 91 days 91 days

which is fine, except that I don't need the "days" suffix at the end of each timedelta value. After reading some similar questions, I tried the following command:

df2 = (df - df.shift(axis=1)).dt.days

which resulted in the following error: 'DataFrame' object has no attribute 'dt'.

What am I doing wrong?

CodePudding user response：

The dt attribute is for objects of type Series, not DataFrame.

If you just want the number of days instead of a pandas.Timedelta object, you can do this:

df2 = (df - df.shift(axis=1))
df2.loc['expiry',:] = df2.loc['expiry',:].dt.days

Output:

       A6U20 A6Z20 A6H21 A6M21 A6U21 A6Z21
expiry   NaT  91.0  91.0  91.0  91.0  91.0

If you would like the NaT to be replaced with 0 days, you can use fillna():

df2 = (df - df.shift(axis=1)).fillna(pd.Timedelta(0))
df2.loc['expiry',:] = df2.loc['expiry',:].dt.days

Output:

       A6U20 A6Z20 A6H21 A6M21 A6U21 A6Z21
expiry     0    91    91    91    91    91

CodePudding user response：

I figured out dt works only on series and not on dataframes. The following apply function did the trick:

df2 = df2.apply(lambda x : x.dt.days)