BUG: make sure that we are passing thru kwargs to groupby

[jreback]

BUG: allow timedelta64 to work in groupby with numeric_only=False

closes #5724

In [6]:         df = pd.DataFrame(
   ...:              {'group': [1, 1, 2],
   ...:               'int': [1, 2, 3],
   ...:               'float': [4., 5., 6.],
   ...:               'string': list('abc'),
   ...:               'category_string': pd.Series(list('abc')).astype('category'),
   ...:               'category_int': [7, 8, 9],
   ...:               'datetime': pd.date_range('20130101', periods=3),
   ...:               'datetimetz': pd.date_range('20130101',
   ...:                                           periods=3,
   ...:                                           tz='US/Eastern'),
   ...:               'timedelta': pd.timedelta_range('1 s', periods=3, freq='s')},
   ...:              columns=['group', 'int', 'float', 'string',
   ...:                       'category_string', 'category_int',
   ...:                       'datetime', 'datetimetz',
   ...:                       'timedelta'])

In [7]: df
Out[7]: 
   group  int  float string category_string  category_int   datetime                datetimetz  timedelta
0      1    1    4.0      a               a             7 2013-01-01 2013-01-01 00:00:00-05:00   00:00:01
1      1    2    5.0      b               b             8 2013-01-02 2013-01-02 00:00:00-05:00   00:00:02
2      2    3    6.0      c               c             9 2013-01-03 2013-01-03 00:00:00-05:00   00:00:03

# same as in master
In [8]: df.groupby('group').mean()
Out[8]: 
       int  float  category_int
group                          
1      1.5    4.5           7.5
2      3.0    6.0           9.0

# works
In [9]: df.groupby('group').mean(numeric_only=False)
Out[9]: 
       int  float  category_int            datetime                datetimetz       timedelta
group                                                                                        
1      1.5    4.5           7.5 2013-01-01 12:00:00 2013-01-01 12:00:00-05:00 00:00:01.500000
2      3.0    6.0           9.0 2013-01-03 00:00:00 2013-01-03 00:00:00-05:00        00:00:03

# same as in master
In [10]: df.groupby('group').sum()
Out[10]: 
       int  float  category_int
group                          
1        3    9.0            15
2        3    6.0             9

# works
In [11]: df.groupby('group').sum(numeric_only=False)
Out[11]: 
       int  float string  category_int  timedelta
group                                            
1        3    9.0     ab            15   00:00:03
2        3    6.0      c             9   00:00:03

# same as in master
In [13]: df.groupby('group').min()
Out[13]: 
       int  float string  category_int   datetime                datetimetz  timedelta
group                                                                                 
1        1    4.0      a             7 2013-01-01 2013-01-01 00:00:00-05:00   00:00:01
2        3    6.0      c             9 2013-01-03 2013-01-03 00:00:00-05:00   00:00:03

[codecov-io]

Current coverage is 84.73% (diff: 64.61%)

Merging #15054 into master will decrease coverage by 0.01%

@@             master     #15054   diff @@
==========================================
  Files           145        145          
  Lines         51232      51270    +38   
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
+ Hits          43420      43445    +25   
- Misses         7812       7825    +13   
  Partials          0          0          

Powered by Codecov. Last update 0fe491d...768fce1

[jorisvandenbossche]

Shouldn't we also explicitly test that groupby().mean() now works on a timedelta column? (as I understand this was also what was reported in the issue)

[jorisvandenbossche]

According to the coverage report, this error is never reached. Is it needed for the categorical you added in the test?

[jreback]

so finally fixed this. was a PITA!

passing numeric_only=False simply did not work before at all. Now datetimes / timedeltas (and even strings work if the function supports it).

[jreback]

@jorisvandenbossche