熊猫在多索引级别上的groupby：group_keys

Question

I have a DataFrame whose columns are MultiIndex. I want to groupby one level of the columns and use apply to perform a transformations.

Goal: I want that the DataFrame passed to the function using apply does not have the keys of groupby in the index.

From the docs it looks like this is what group_keys is doing, but it seems to have no effect:

import numpy as np
import pandas as pd

data = {'A': pd.DataFrame(np.random.randn(100, 5)),
        'B': pd.DataFrame(np.random.randn(100, 5)),
        'C': pd.DataFrame(np.random.randn(100, 5))}

data = pd.concat(data, axis=1, names=['feat_1', 'feat_2'])

def foo(df):
    print(df.columns)
    return df.sum(1)

My attempt:

result = data.groupby(level=['feat_1'], axis=1, group_keys=False).apply(foo)

This is what is printed on screen:

MultiIndex(levels=[['A', 'B', 'C'], [0, 1, 2, 3, 4]],
           labels=[[0, 0, 0, 0, 0], [0, 1, 2, 3, 4]],
           names=['feat_1', 'feat_2'])
MultiIndex(levels=[['A', 'B', 'C'], [0, 1, 2, 3, 4]],
           labels=[[1, 1, 1, 1, 1], [0, 1, 2, 3, 4]],
           names=['feat_1', 'feat_2'])
MultiIndex(levels=[['A', 'B', 'C'], [0, 1, 2, 3, 4]],
           labels=[[2, 2, 2, 2, 2], [0, 1, 2, 3, 4]],
           names=['feat_1', 'feat_2'])

Desired output of print: I would like the function foo to receive a dataframe with only feat_2 as columns, given I am using group_keys=False

[0, 1, 2, 3, 4]
[0, 1, 2, 3, 4]
[0, 1, 2, 3, 4]

Am I missing something from the documentation? Or how can I achieve what I want (possibly without modifying the function foo)?

Note: I am using pandas 0.20.3 on Python 3

Answer 1

Rather than grouping, how about:

for feat1 in data.columns.levels[0]:
    print(list(data.columns.levels[1]))

[0, 1, 2, 3, 4]
[0, 1, 2, 3, 4]
[0, 1, 2, 3, 4]