在多索引数据帧中，groupby之后的.columns.levels [1]给出整个数据帧的列

Question

Lets say I have three data frames that I concatenate horizontally with the help of multi-index:

df1 = pd.DataFrame(data=np.random.randint(0, 100, (4, 5)), columns=list('ABCDE'))
df2 = pd.DataFrame(data=np.random.randint(0, 100, (4, 5)),columns=list('AGHIJ'))
df3 = pd.DataFrame(data=np.random.randint(0, 100, (4, 5)),columns=list('ALMNP'))
dfs = []
dfs.append(df1)
dfs.append(df2)
dfs.append(df3)
result = pd.concat(dfs, axis=1, keys=range(len(dfs)))

if I groupby the by the first index I should get my first dataframe and If I look at the list of its columns it should be ABCD but it is not the case.

print(result.groupby(axis=1, level=0).get_group(0).columns.levels[1])

gives me all the columns of df1, df2 and df3

Answer 1

I will use get_level_values , since levels keep all category of original dataframe columns

result.groupby(axis=1, level=0).get_group(0).columns.get_level_values(1)
Out[1296]: Index(['A', 'B', 'C', 'D', 'E'], dtype='object')

Answer 2

This is an issue with unused levels . When you have a MultiIndex, the levels are still there, just unused, so you can remove them if needed:

result.groupby(axis=1, level=0).get_group(0).columns.remove_unused_levels().levels[1]
#Index(['A', 'B', 'C', 'D', 'E'], dtype='object')

---

To see that everything is still there, look at the columns. There are still 13 values for the first level, but this group only references the first 5.

print(result.groupby(axis=1, level=0).get_group(0).columns)
#MultiIndex(levels=[[0, 1, 2], ['A', 'B', 'C', 'D', 'E', 'G', 'H', 'I', 'J', 'L', 'M', 'N', 'P']],
#           codes=[[0, 0, 0, 0, 0], [0, 1, 2, 3, 4]])