python - 将新的嵌套级别值添加到 MultiIndex DataFrame

Question

如何将另一个级别值添加到初始化为特定值(例如无)的 MultiIndex。很难用文字来描述如何添加新的值级别，最好用图形来描述:

df_before
          a   b   c   d
l1  l2                 
bar one  24  13   8   9
    two  11  30   7  23
baz one  21  31  12  30
    two   2   5  19  24
foo one  15  18   3  16
    two   2  24  28  11
qux one  23   9   6  12
    two  29  28  11  21

df_after
             a   b   c   d
l1  l2                 
bar one     24  13   8   9
    two     11  30   7  23
    new     None  None   None  None
baz one     21  31  12  30
    two     2   5  19  24
    new     None  None   None  None
foo one     15  18   3  16
    two     2  24  28  11
    new     None  None   None  None
qux one     23   9   6  12
    two     29  28  11  21
    new     None  None   None  None

注意:我的 DataFrame 确实具有三个级别，因此可以推广到更多级别的解决方案将不胜感激。我最好的尝试是获取旧级别的唯一值，附加一个新值并设置新级别，但它没有产生我想要的结果

# this is a failed attempt of what I wanted to do
new_level_values = [*list(df.index.get_level_values(2).unique()), "new"]
df.index = df.index.set_levels(levels=new_level_values, level=2)
df

Answer 1

使用DataFrame.reindex将新值附加到第二级并通过 MultiIndex.from_product 重新创建 MultiIndex :

v = [*list(df.index.levels[1]), "new"]

df = df.reindex(pd.MultiIndex.from_product([df.index.levels[0],v]))
print (df)
            a     b     c     d
l1                             
bar one  24.0  13.0   8.0   9.0
    two  11.0  30.0   7.0  23.0
    new   NaN   NaN   NaN   NaN
baz one  21.0  31.0  12.0  30.0
    two   2.0   5.0  19.0  24.0
    new   NaN   NaN   NaN   NaN
foo one  15.0  18.0   3.0  16.0
    two   2.0  24.0  28.0  11.0
    new   NaN   NaN   NaN   NaN
qux one  23.0   9.0   6.0  12.0
    two  29.0  28.0  11.0  21.0
    new   NaN   NaN   NaN   NaN