Find date differences since inception within group

Question

Is there an efficient way to compute within-group date differences (in days) for a dataframe in the form:

x = pd.DataFrame(
    {'grp':['A','A','A','B','B','B'],
     'dt':pd.DatetimeIndex(['1/1/00 00:00:00','1/2/00','1/3/00','1/2/01','1/3/01','1/5/01'])})
x
Out[1]: 
                   dt grp
0 2000-01-01 00:00:00   A
1 2000-01-02 00:00:00   A
2 2000-01-03 00:00:00   A
3 2001-01-02 00:00:00   B
4 2001-01-03 00:00:00   B
5 2001-01-05 00:00:00   B

So that the result is similar to:

grp    days_since_start
A       0
A       1
A       2
B       0
B       1
B       3

Answer 1

Sure. Group by the group name, take the smallest time in each group, take the difference:

x.set_index('grp') - x.groupby('grp').min()
#        dt
#grp
#A   0 days
#A   1 days
#A   2 days
#B   0 days
#B   1 days
#B   3 days
#Name: dt, dtype: timedelta64[ns]