Regression in DataFrameGroupBy.head()

[danielballan]

For this SO question I isolated a regression in DataFrameGroupBy.head() between 0.12.0 and 0.13.0.

In [1]: df = pd.DataFrame({'A': ['one', 'one', 'two', 'three', 'three', 'one'], 'B': range(6)})

In 0.12.0:

In [2]: df.groupby('A').head()
Out[2]: 
             A  B
A                
one   0    one  0
      1    one  1
      5    one  5
three 3  three  3
      4  three  4
two   2    two  2

which effectively returns the rows of the underlying .obj sorted by group. In 0.13.0:

In [2]: df.groupby('A').head()
Out[2]: 
             A  B
A                
one   0    one  0
      1    one  1
two   2    two  2
three 3  three  3
      4  three  4
one   5    one  5

[6 rows x 2 columns]

which returns the rows of the underlying .obj in their original order. The former is more intuitive.

[jreback]

This in master, looks correct to me and does not include the grouping coumn either)

In [2]: df.groupby('A').head()
Out[2]: 
       A  B
0    one  0
1    one  1
2    two  2
3  three  3
4  three  4
5    one  5

[6 rows x 2 columns]

[danielballan]

As I note in my answer, the existence of head() on a GroupBy is a little confusing. As I think about it, I'm not sure whether it's better to have it show the underlying object or a special re-sorted view.

[danielballan]

OK, I can live with that. Closing.

[jorisvandenbossche]

I don't know if it is worth to bikeshed over, but I think we should then at least update the docstring, so marked it as a DOC issue (https://github.com/pydata/pandas/blob/master/pandas/core/groupby.py#L790)

• clarify that it returns "Returns first n rows of each group in original order."?
• that it is equivalent to .apply(lambda x: x.head(n)) is no longer true (that was the previous behaviour)
• the example uses as_index=False, while the docs say it does ignore as_index

[jreback]

fair enough to make doc string consistent