Grouby select multiple columns

[hayd]

Is this supported?

g[['X', 'Y']]  # do groupby stuff with just these columns

http://stackoverflow.com/q/22139053/1240268

[hayd]

This should probably raise NotImplemented ?

[naught101]

@hayd, thanks for the follow up. For reference:

df = pandas.DataFrame({"Dummy":[1,2]*6, "X":[1,3,7]*4, 
                       "Y":[2,3,4]*4, "group":["A","B"]*6})
df[['X', 'Y']].head(1)
   X  Y
0  1  2
[1 rows x 2 columns]

df[:,['X', 'Y']].head(1)
TypeError: unhashable type: 'slice'

df.loc[:,['X', 'Y']].head(1)
   X  Y
0  1  2
[1 rows x 2 columns]

df.groupby('group')[['X', 'Y']].head(1)
         Dummy  X  Y group
group                     
A     0      1  1  2     A
B     1      2  3  3     B
[2 rows x 4 columns]

df.groupby('group').loc[:,['X', 'Y']].head(1)
AttributeError: Cannot access attribute 'loc' of 'DataFrameGroupBy' objects, try using the 'apply' method

[jreback]

This works, except in head/tail.
jreback@92e5c50
Current

In [1]: df = DataFrame([[1, 2], [1, 4], [5, 6]], columns=['A', 'B'])

In [2]: df
Out[2]: 
   A  B
0  1  2
1  1  4
2  5  6

[3 rows x 2 columns]

In [4]: df.groupby('A',as_index=True).head(1)
Out[4]: 
     A  B
A        
1 0  1  2
5 2  5  6

[2 rows x 2 columns]

In [5]: df.groupby('A',as_index=False).head(1)
Out[5]: 
   A  B
0  1  2
2  5  6

[2 rows x 2 columns]

master (with my change)

In [1]: df = DataFrame([[1, 2], [1, 4], [5, 6]], columns=['A', 'B'])

In [2]: df
Out[2]: 
   A  B
0  1  2
1  1  4
2  5  6

[3 rows x 2 columns]

In [3]: df.groupby('A',as_index=True).head(1)
Out[3]: 
     B
A     
1 0  2
5 2  6

[2 rows x 1 columns]

In [4]: df.groupby('A',as_index=False).head(1)
Out[4]: 
   B
0  2
2  6

[2 rows x 1 columns]

I think master is wrong here ?

[TomAugspurger]

Same issue? #5264

[jreback]

@TomAugspurger actually might be the same....let's close this on and consolidate.

The fix is pretty trivial, but a couple of tests are 'wrong' (that's why I put them up). So needs to be carefully gone over.

[jreback]

consolidating issue to #5264

[hayd]

@jreback I really want to kill this index behaviour of head/tail, it should act like filter. There is an issue somewhere, maybe should do it sooner rather than later #5755

[hayd]

@jreback master is wrong there!

[hayd]

will put up change (breaking) to make head/tail act like filter (regardless of as_index), the reason it's not is historical IMO (from when it was .apply(head) )... I don't think people ever want it ?