合并两个具有重叠 MultiIndex 列的 DataFrame

Posted 2023-03-12

【中文标题】合并两个具有重叠 MultiIndex 列的 DataFrame

【英文标题】：Merge two DataFrames with overlapping MultiIndex columns

【发布时间】：2021-05-29 19:09:58

【问题描述】：

我正在尝试找到一种简单的方法将两个 MultiIndex 数据框合并在一起，以添加新列并合并现有列。例如，如果我有两个数据框

d1_columns = pd.MultiIndex.from_product([["A", "B",], ["1", "2"]])
d1_index = pd.date_range("2020-01-01", "2020-01-5", freq="D")
d1 = pd.DataFrame(random.rand(5, 4), columns=d1_columns, index=d1_index)

print(d1)


                 A                   B          
                   1         2         1         2
2020-01-01  0.381909  0.487480  0.389250  0.853449
2020-01-02  0.752374  0.508806  0.491892  0.786918
2020-01-03  0.019655  0.537763  0.263242  0.378302
2020-01-04  0.460276  0.227113  0.423472  0.514639
2020-01-05  0.046673  0.864282  0.223340  0.929049

d2_columns = pd.MultiIndex.from_product([["B", "C"], ["1", "2"]])
d2_index = pd.date_range("2020-01-03", "2020-01-7", freq="D")
d2 = pd.DataFrame(random.rand(5, 4), columns=d2_columns, index=d2_index)

print(d2)

                   B                   C          
                   1         2         1         2
2020-01-03  0.495979  0.888207  0.776861  0.531693
2020-01-04  0.408030  0.545351  0.452913  0.768284
2020-01-05  0.374996  0.593571  0.925979  0.398629
2020-01-06  0.085565  0.845354  0.792325  0.501057
2020-01-07  0.780985  0.390948  0.731769  0.488155

如果我想合并它们，我会将重叠的列分开，而新列可以正常工作：

df = d1.merge(d2, left_index=True, right_index=True, how="outer")

print(df)

                   A                 B_x                 B_y            \
                   1         2         1         2         1         2   
2020-01-01  0.381909  0.487480  0.389250  0.853449       NaN       NaN   
2020-01-02  0.752374  0.508806  0.491892  0.786918       NaN       NaN   
2020-01-03  0.019655  0.537763  0.263242  0.378302  0.495979  0.888207   
2020-01-04  0.460276  0.227113  0.423472  0.514639  0.408030  0.545351   
2020-01-05  0.046673  0.864282  0.223340  0.929049  0.374996  0.593571   
2020-01-06       NaN       NaN       NaN       NaN  0.085565  0.845354   
2020-01-07       NaN       NaN       NaN       NaN  0.780985  0.390948   

                   C            
                   1         2  
2020-01-01       NaN       NaN  
2020-01-02       NaN       NaN  
2020-01-03  0.776861  0.531693  
2020-01-04  0.452913  0.768284  
2020-01-05  0.925979  0.398629  
2020-01-06  0.792325  0.501057  
2020-01-07  0.731769  0.488155 

有没有一种简单的方法可以合并重叠的列，以便将新数据添加到现有列中（是否覆盖以前的数据无关紧要），所以输出看起来像这样？

                   A                 B                     C          
                   1         2         1         2         1         2
2020-01-01  0.633182  0.335651  0.072520  0.578472       NaN       NaN
2020-01-02  0.785482  0.562421  0.658556  0.557171       NaN       NaN
2020-01-03  0.755049  0.575611  0.592934  0.735094  0.647117  0.306296
2020-01-04  0.035943  0.792211  0.002617  0.159366  0.320691  0.825184
2020-01-05  0.932623  0.643129  0.778002  0.581527  0.718405  0.289289
2020-01-06       NaN       NaN  0.085565  0.845354  0.012412  0.960234
2020-01-07       NaN       NaN  0.780985  0.390948  0.444406  0.210821

谢谢

【参考方案1】：

看来你想要

df = d1.combine_first(d2)

或

df = d2.combine_first(d1)

取决于应首选哪个帧的值。