不再支持 Pandas 在多索引 dataframe 中通過 loc 插入行

Question

我有一個 pandas dataframe 像這樣：

pd.DataFrame.from_dict({(4.0,
  'Net Sales'): {('details',
   pd.Timestamp('2020-04-01 00:00:00'),
   'Apr_FY21'): 1000, ('details',
   pd.Timestamp('2020-05-01 00:00:00'),
   'May_FY21'): 1000, ('details',
   pd.Timestamp('2020-06-01 00:00:00'),
   'Jun_FY21'): 1000},
 (5.0,
  'Margin'): {('details',
   pd.Timestamp('2020-04-01 00:00:00'),
   'Apr_FY21'): 20, ('details',
   pd.Timestamp('2020-05-01 00:00:00'),
   'May_FY21'): 15, ('details',
   pd.Timestamp('2020-06-01 00:00:00'),
   'Jun_FY21'): 10}}).T

我想添加一個計算行，不幸的是，loc 不再支持該行。 什么是替代品？

 df2.loc[(5.1,'Margin %'),'details']=(np.divide(df2.loc[(5,'Margin)'),'details'],(df2.loc[(4,'Net Sales'),'details']))*100).values

KeyError：'不再支持將 list-likes 傳遞給.loc 或帶有任何缺失標簽的 []，請參閱https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#deprecate-loc-reindex -listlike'

Answer 1

將其創建為一個系列，然后將 append 創建為 DataFrame：

s = pd.Series((np.divide(df2.loc[(5.0,'Margin'),'details'],(df2.loc[(4.0,'Net Sales'),'details']))*100).values, 
              index=df2.columns, 
              name = (5.1,'Margin %')
             )
df2 = df2.append(s)

Answer 2

您的陳述中實際上有一個錯字：

df.loc[(5.1,'Margin %'),'details']=(np.divide(df.loc[(5,'Margin)'),'details'],(df.loc[(4,'Net Sales'),'details']))*100).values

-------------------------------------------------- ----------------------------------------------------------^

不需要額外的 ')'

這確實有效：

 df2.loc[(5.1,'Margin %'),'details']=(np.divide(df2.loc[(5,'Margin'),'details'],(df2.loc[(4,'Net Sales'),'details']))*100).values

Output：

                 details                      
              2020-04-01 2020-05-01 2020-06-01
                Apr_FY21   May_FY21   Jun_FY21
4.0 Net Sales     1000.0     1000.0     1000.0
5.0 Margin          20.0       15.0       10.0
5.1 Margin %         2.0        1.5        1.0