df.groupby() does not allow to use name of index at level 0 when not using a MultiIndex

[floux]

In [24]: df = pd.DataFrame({
   ....:         'exp' : ['A']*3 + ['B']*3,
   ....:         'var1' : range(6),
   ....:         })

In [25]: df = df.set_index(['exp'])

In [26]: df.index.names
Out[27]: ['exp']


In [28]: df
Out[28]:
     var1
exp
A       0
A       1
A       2
B       3
B       4
B       5

In [29]: try:
   ....:             df.groupby(level='exp').size()
   ....: except ValueError as e:
   ....:             print('index can not be accessed by name')
   ....:             print(e)
   ....:
index can not be accessed by name
level > 0 only valid with MultiIndex
# at this point I would argue that the level 'exp' is level 0,
# and therefore this should work.

In [29]: df1 = df.groupby(level=0).size()

In [30]: df1
Out[30]:
exp
A      3
B      3
dtype: int64

[hayd]

Good find. I've put a fix together for this (coming shortly).

Note: although in python 2 'exp' > 0 in py3 this raises an error.

[pepicello]

Many years later, I have this issue with pandas 0.25.1, but with a different error. Small reproducible example:

pd.DataFrame([[1, 2], [3, 4]], columns=pd.Index(['a', 'b'], name='a')).groupby(level='a', axis=1)

which returns a ValueError: level name a is not the name of the index.

Is this behaviour expected? If it is, it would still be very handy to have a way to do this quickly on index names without needing to check if it is a MultiIndex before.