pandas groupby will not do anything, including as_index if it detects the apply function doesn't do anything?

Pandas groupby will be 'smart' if it knows it doesn't need to do anything?

have a df

test = pd.DataFrame([{'a': 1, 'b': 15}, {'a':1, 'b': 14}])

if we do

test.groupby('a').apply(lambda x: x.iloc[:-1])

Then pandas will process everything,

a	b
a			
1	0	1	15

but if we just naively do nothing:

test.groupby('a').apply(lambda x: x)

returns:

a	b
0	1	15
1	1	14

the return is the df itself - Normally we should expect the index would be changed to 'a', as the default value for 'as_index' is True. but it doesn't.

Yes, pandas is "smart" in the way it processes the output, but not the way you think it is.

The decision to add the grouping keys depends on the shape and index of the output. Pandas doesn't detect what the function is doing to the values.

Let's use pandas 1.5.3 and just add 1:

test.groupby('a').apply(lambda x: x+1)

We get a FutureWarning:

FutureWarning: Not prepending group keys to the result index of transform-like apply. In the future, the group keys will be included in the index, regardless of whether the applied function returns a like-indexed object.
To preserve the previous behavior, use

	>>> .groupby(..., group_keys=False)

To adopt the future behavior and silence this warning, use 

	>>> .groupby(..., group_keys=True)
  test.groupby('a').apply(lambda x: x+1)

This clearly explains that if the function keeps the index unchanged, then the grouping keys is not added to the index.

If you want to have the key use an explicit:

test.groupby('a', group_keys=True).apply(lambda x: x+1)

Output:

     a   b
a         
1 0  2  16
  1  2  15