Question

鉴于此代码合并了三个Pandas数据帧：

raw_data = {
        'type': [0, 1, 1],
        'id': ['3', '4', '5'],
        'name_1': ['Alex', 'Amy', 'Allen']}
df_a = pd.DataFrame(raw_data, columns = ['type', 'id', 'name_1' ])
df_a.set_index(['type', 'id'])

raw_datab = {
        'type': [1, 1, 1, 0],
        'id': ['4', '5', '5', '7'],
        'name_2': ['Billy', 'Brian', 'Joe', 'Bryce']}
df_b = pd.DataFrame(raw_datab, columns = ['type', 'id', 'name_2'])
df_b.set_index(['type', 'id'])

raw_datac = {
        'name_3': ['School', 'White', 'Jane', 'Homer'],
        'id': ['4', '6', '5', '5'],
        'type': [1, 1, 1, 1]}
df_c = pd.DataFrame(raw_datac, columns = ['name_3', 'id', 'type' ])
df_c.set_index(['type', 'id'])

dfx = df_a.merge(df_b, how='outer').merge(df_c, how='outer')
print(dfx)

我收到以下回复：

   type id name_1 name_2  name_3
0     0  3   Alex    NaN     NaN
1     1  4    Amy  Billy  School
2     1  5  Allen  Brian    Jane
3     1  5  Allen  Brian   Homer
4     1  5  Allen    Joe    Jane
5     1  5  Allen    Joe   Homer
6     0  7    NaN  Bryce     NaN
7     1  6    NaN    NaN   White

我实际需要的是获得原始顺序中列的串联。例如：

type    id  name_1  type_2  id_2    name_2  name_3  id_3  type_3  
0       3   Alex    0       3       NaN     NaN     3     0     
1       4   Amy     1       4       Billy   School  4     1     
1       5   Allen   1       5       Brian   Jane    5     1     
1       5   Allen   1       5       Brian   Homer   5     1     
1       5   Allen   1       5       Joe     Jane    5     1     
1       5   Allen   1       5       Joe     Homer   5     1     
0       7   NaN     0       7       Bryce   NaN     7     0     
1       6   NaN     1       6       NaN     White   6     1

Pandas可以吗？

Answer 1

我认为你可以这样做：

让我们使用这种语法将列的“副本”添加到每个数据帧的索引中，

[i.set_index([i['type'],i['id']], inplace=True) for i in [df_a, df_b, df_c]]

注意：df.set_index('Col1') vs df.set_index(df['Col1'])，后面的语法在索引中创建Col1的副本，而前者将Col1列移动到索引中。

现在，让我们将数据框合并到索引上，使用how ='outer'，并使用suffixes处理重复列命名：

df_a.merge(df_b, how='outer', suffixes=('_1',''), right_index=True, left_index=True)\
    .merge(df_c, how='outer', suffixes=('_2','_3'), right_index=True, left_index=True)\
    .reset_index()

输出：

   type id  type_1 id_1 name_1  type_2 id_2 name_2  name_3 id_3  type_3
0     0  3     0.0    3   Alex     NaN  NaN    NaN     NaN  NaN     NaN
1     0  7     NaN  NaN    NaN     0.0    7  Bryce     NaN  NaN     NaN
2     1  4     1.0    4    Amy     1.0    4  Billy  School    4     1.0
3     1  5     1.0    5  Allen     1.0    5  Brian    Jane    5     1.0
4     1  5     1.0    5  Allen     1.0    5  Brian   Homer    5     1.0
5     1  5     1.0    5  Allen     1.0    5    Joe    Jane    5     1.0
6     1  5     1.0    5  Allen     1.0    5    Joe   Homer    5     1.0
7     1  6     NaN  NaN    NaN     NaN  NaN    NaN   White    6     1.0

编辑：由于我们正在合并索引，我们可以改为使用join。

df_a.join(df_b,how='outer',lsuffix='_1')
    .join(df_c,how='outer',lsuffix='_2', rsuffix='_3')
    .reset_index()

合并数据帧时获取原始列顺序

1 个答案: