在 Python Pandas 中使用 DatetimeIndex 切片 MultIndex 帧时出现 1.1.0 以上的 InvalidIndexError

Question

我的数据包含多个区域的时间线值。我想根据日期切片。

这是我的 MultIndex 数据框，我叫 Bob：

arrays = [[1,1,2,2],
          ['2020-01-06', '2020-01-13','2020-01-06', '2020-01-13']]
df = pd.DataFrame(np.transpose(arrays))
df[1] = pd.to_datetime(df[1])
index = pd.MultiIndex.from_frame(df, names=['zone', 'date'])
bob = pd.Series(np.random.randn(4), index=index)
print(bob)

zone  date      
1     2020-01-06   -0.513744
      2020-01-13    1.367461
2     2020-01-06    0.209916
      2020-01-13    0.397261

现在，我想从单个日期时间索引中获取切片，并使用它来获取 Bob 的切片。 以下代码在 Pandas 1.0.1（可能更旧）中有效，但在 1.1 中中断

print(pd.__version__)
singleIndex = pd.to_datetime(pd.Index(['2020-01-06', '2020-01-13']))
dateSlice = singleIndex[1:]
print(dateSlice)
idx = pd.IndexSlice
print(bob.loc[idx[:,dateSlice]])

1.0.1
DatetimeIndex(['2020-01-13'], dtype='datetime64[ns]', freq=None)
zone  date      
1     2020-01-13    1.367461
2     2020-01-13    0.397261
dtype: float64

在 1.1.0 中

1.1.0
DatetimeIndex(['2020-01-13'], dtype='datetime64[ns]', freq=None)
Traceback (most recent call last):
  File "try.py", line 18, in <module>
    print(bob.loc[idx[:,dateSlice]])
  File "/home/mix/.conda/envs/covidpro/lib/python3.6/site-packages/pandas/core/indexing.py", line 873, in __getitem__
    return self._getitem_tuple(key)
  File "/home/mix/.conda/envs/covidpro/lib/python3.6/site-packages/pandas/core/indexing.py", line 1044, in _getitem_tuple
    return self._getitem_lowerdim(tup)
  File "/home/mix/.conda/envs/covidpro/lib/python3.6/site-packages/pandas/core/indexing.py", line 766, in _getitem_lowerdim
    return self._getitem_nested_tuple(tup)
  File "/home/mix/.conda/envs/covidpro/lib/python3.6/site-packages/pandas/core/indexing.py", line 826, in _getitem_nested_tuple
    result = self._handle_lowerdim_multi_index_axis0(tup)
  File "/home/mix/.conda/envs/covidpro/lib/python3.6/site-packages/pandas/core/indexing.py", line 1066, in _handle_lowerdim_multi_index_axis0
    return self._get_label(tup, axis=axis)
  File "/home/mix/.conda/envs/covidpro/lib/python3.6/site-packages/pandas/core/indexing.py", line 1059, in _get_label
    return self.obj.xs(label, axis=axis)
  File "/home/mix/.conda/envs/covidpro/lib/python3.6/site-packages/pandas/core/generic.py", line 3480, in xs
    loc, new_index = self.index.get_loc_level(key, drop_level=drop_level)
  File "/home/mix/.conda/envs/covidpro/lib/python3.6/site-packages/pandas/core/indexes/multi.py", line 2858, in get_loc_level
    k = self._get_level_indexer(k, level=i)
  File "/home/mix/.conda/envs/covidpro/lib/python3.6/site-packages/pandas/core/indexes/multi.py", line 2965, in _get_level_indexer
    code = self._get_loc_single_level_index(level_index, key)
  File "/home/mix/.conda/envs/covidpro/lib/python3.6/site-packages/pandas/core/indexes/multi.py", line 2634, in _get_loc_single_level_index
    return level_index.get_loc(key)
  File "/home/mix/.conda/envs/covidpro/lib/python3.6/site-packages/pandas/core/indexes/datetimes.py", line 586, in get_loc
    raise InvalidIndexError(key)
pandas.errors.InvalidIndexError: DatetimeIndex(['2020-01-13'], dtype='datetime64[ns]', freq=None)

我承认 MultiIndex 让我非常困惑。如何在新的 pandas 中正确地进行切片？

Answer 1

似乎不再支持，您可以在boolean indexing 中使用Index.get_level_values 和Index.isin 替代：

print(bob[bob.index.get_level_values(1).isin(dateSlice)])
zone  date      
1     2020-01-13   -1.396496
2     2020-01-13   -0.504466
dtype: float64

对于一个字符串值，它的工作方式有点不同：

print(bob.loc[idx[:,'2020-01-13']])
zone
1   -0.200758
2    0.410052
dtype: float64


print(bob.xs('2020-01-13', level=1, drop_level=False))
zone  date      
1     2020-01-13    1.129484
2     2020-01-13    0.185156
dtype: float64