潘达斯Multiindex在对Dataframe进行分片后没有受到限制

Question

我需要使用MuliIndex的第一个级别。但是在对Dataframe进行分片后，我得到的index.levels[0]和分片前一样。

[] df['2019-04-26 01:00:00':'2020-01-01 22:00:00'].index.levels[0]

[Out] 
DatetimeIndex(['2019-01-01 01:00:00', '2019-01-01 04:00:00',
               '2019-01-01 07:00:00', '2019-01-01 10:00:00',
               '2019-01-01 13:00:00', '2019-01-01 16:00:00',
               '2019-01-01 19:00:00', '2019-01-01 22:00:00',
               '2019-01-02 01:00:00', '2019-01-02 04:00:00',
               ...
               '2020-02-16 22:00:00', '2020-02-17 01:00:00',
               '2020-02-17 04:00:00', '2020-02-17 07:00:00',
               '2020-02-17 10:00:00', '2020-02-17 13:00:00',
               '2020-02-17 16:00:00', '2020-02-17 19:00:00',
               '2020-02-17 22:00:00', '2020-02-18 01:00:00'],
              dtype='datetime64[ns]', name='Date', length=3307, freq=None)

完整的索引正确地切开了，但我只想要第一层。这是Pandas-1.0.3的bug还是功能？如何才能得到我想要的东西？这个Multiindex已经排序了，我试着给切片后的Dataframe分配另一个变量，比如说 df1 = df['2019-04-26 01:00:00':'2020-01-01 22:00:00'] 和我得到了同样的 df1.index.levels[0]

新增：排序的MultiIndex是。

            ('2019-01-01 01:00:00',   'Зеленга'),
            ('2019-01-01 04:00:00', 'Астрахань'),
            ('2019-01-01 04:00:00',   'Зеленга'),
            ('2019-01-01 07:00:00', 'Астрахань'),
            ('2019-01-01 07:00:00',   'Зеленга'),
            ('2019-01-01 10:00:00', 'Астрахань'),
            ('2019-01-01 10:00:00',   'Зеленга'),
            ('2019-01-01 13:00:00', 'Астрахань'),
            ('2019-01-01 13:00:00',   'Зеленга'),
            ...
            ('2020-02-16 22:00:00',    'Досанг'),
            ('2020-02-17 01:00:00',    'Досанг'),
            ('2020-02-17 04:00:00',    'Досанг'),
            ('2020-02-17 07:00:00',    'Досанг'),
            ('2020-02-17 10:00:00',    'Досанг'),
            ('2020-02-17 13:00:00',    'Досанг'),
            ('2020-02-17 16:00:00',    'Досанг'),
            ('2020-02-17 19:00:00',    'Досанг'),
            ('2020-02-17 22:00:00',    'Досанг'),
            ('2020-02-18 01:00:00',    'Досанг')],
           names=['Дата', 'Location'], length=13185)

然后我想删除没有完整列表的日期的位置。我计算了限制的边界，并对数据框架进行切片。date_frame=(Timestamp('2019-04-26 01:00:00'), Timestamp('2020-01-01 22:00:00'))所以，对于 df[date_range[0]:date_range[1]].index我得到了

            ('2019-04-26 01:00:00',    'Досанг'),
            ('2019-04-26 01:00:00',   'Зеленга'),
            ('2019-04-26 01:00:00',  'Харабали'),
            ('2019-04-26 01:00:00', 'Черный Яр'),
            ('2019-04-26 04:00:00', 'Астрахань'),
            ('2019-04-26 04:00:00',    'Досанг'),
            ('2019-04-26 04:00:00',   'Зеленга'),
            ('2019-04-26 04:00:00',  'Харабали'),
            ('2019-04-26 04:00:00', 'Черный Яр'),
            ...
            ('2020-01-01 19:00:00', 'Астрахань'),
            ('2020-01-01 19:00:00',    'Досанг'),
            ('2020-01-01 19:00:00',   'Зеленга'),
            ('2020-01-01 19:00:00',  'Харабали'),
            ('2020-01-01 19:00:00', 'Черный Яр'),
            ('2020-01-01 22:00:00', 'Астрахань'),
            ('2020-01-01 22:00:00',    'Досанг'),
            ('2020-01-01 22:00:00',   'Зеленга'),
            ('2020-01-01 22:00:00',  'Харабали'),
            ('2020-01-01 22:00:00', 'Черный Яр')],
           names=['Дата', 'Location'], length=10040)

但对于 df[date_range[0]:date_range[1]].index.levels[0] 我总是有

               '2019-01-01 07:00:00', '2019-01-01 10:00:00',
               '2019-01-01 13:00:00', '2019-01-01 16:00:00',
               '2019-01-01 19:00:00', '2019-01-01 22:00:00',
               '2019-01-02 01:00:00', '2019-01-02 04:00:00',
               ...
               '2020-02-16 22:00:00', '2020-02-17 01:00:00',
               '2020-02-17 04:00:00', '2020-02-17 07:00:00',
               '2020-02-17 10:00:00', '2020-02-17 13:00:00',
               '2020-02-17 16:00:00', '2020-02-17 19:00:00',
               '2020-02-17 22:00:00', '2020-02-18 01:00:00'],
              dtype='datetime64[ns]', name='Дата', length=3307, freq=None)

Answer 1

我找到了一个类似的解决方法。

new_idx = df['2019-04-26 01:00:00':'2020-01-01 22:00:00'].reset_index(1).index

但我觉得这不是很优雅