根据多级列筛选行

Question

我正在尝试根据多层列来筛选行。我不能确定我的数据是否有错误

df[df[[('description','type')]].isin(["Filter1","Filter2"])]

>> TypeError: '<' not supported between instances of 'str' and 'float'

如果我尝试转换为字符串，我得到相同的错误

df.loc[:,[('description','type')]] = df.loc[:,[('description','type')]].astype(str)

>> TypeError: '<' not supported between instances of 'str' and 'float'

数据(包含2列的前5行)：

    description
    type        date
0   Filter 1    1 Mth-7/21
1   Filter 1    1 Yr-7/21
2   Filter 2    10 Yr-7/21
3   Filter 2    15 Yr-7/21
4   Filter 3    2 Yr-7/21
5   Filter 3    3 Mth-7/21

当我认为我所有的列值都是字符串时，我不确定为什么它总是抛出前两个错误

我脑海中的一个简单的例子：

df1 = pd.DataFrame({('description','type'):['a','b','c','d'],
                    ('description','value'):[1,2,3,4]})

df1[df1[('description','type')].isin(['a','b'])]

Answer 1

您的两个筛选器之间存在差异：

df[df[[('description','type')]].isin(["Filter1","Filter2"])]
#    ^------------------------^--- HERE

df1[df1[('description','type')].isin(['a','b'])]

# First case: float appears
>>> df1[df1[[('description','type')]].isin(['a','b'])]
  description
         type value
0           a   NaN
1           b   NaN
2         NaN   NaN
3         NaN   NaN

# Second case
>>> df1[df1[('description','type')].isin(['a','b'])]

  description
         type value
0           a     1
1           b     2