ENH: enable 'on' keyword for groupby(..).resample() · Issue #15021 · pandas-dev/pandas (original) (raw)

Problem description

resample gained a new keyword on to be able to resample based on a column of datetime values instead of the index (which can prevent some use of set_index/reset_index). This feature is however not yet available for resample after a groupby.

Code Sample

In [8]: df = pd.DataFrame({'key': ['A', 'B']*5, 'dates': pd.date_range('2016-01-01', periods=10), 'values': np.random.randn(10)}) 

In [9]: df
Out[9]: 
       dates key    values
0 2016-01-01   A -1.291535
1 2016-01-02   B  0.922945
2 2016-01-03   A -1.039163
3 2016-01-04   B -1.102022
4 2016-01-05   A  1.220347
5 2016-01-06   B  0.265720
6 2016-01-07   A  0.888080
7 2016-01-08   B -0.250236
8 2016-01-09   A  0.485518
9 2016-01-10   B -0.188880

In [11]: df.set_index('dates').groupby('key').resample('D').mean()
Out[11]: 
                  values
key dates               
A   2016-01-01 -1.291535
    2016-01-02       NaN
    2016-01-03 -1.039163
    2016-01-04       NaN
    2016-01-05  1.220347
    2016-01-06       NaN
    2016-01-07  0.888080
    2016-01-08       NaN
    2016-01-09  0.485518
B   2016-01-02  0.922945
    2016-01-03       NaN
    2016-01-04 -1.102022
    2016-01-05       NaN
    2016-01-06  0.265720
    2016-01-07       NaN
    2016-01-08 -0.250236
    2016-01-09       NaN
    2016-01-10 -0.188880

In [12]: df.groupby('key').resample('D', on='dates').mean()
...
TypeError: __init__() got an unexpected keyword argument 'on'

Using pandas master, 0.19.0+289.g1bf94c8