Question

我有一个数据框df_in定义为：

import pandas as pd

dic_in = {'A': ['ff','rr' ,'nn' ,'qq','tt' ,'pp','uu'],
      'B1': ['33',r'\N','39' ,'22',r'\N','56','90'],
      'C1': ['44',r'\N','74' ,'34',r'\N','89','99'],
      'B2': ['33','63' ,r'\N','22','71' ,'56','90'],
      'C2': ['44','85' ,r'\N','34','52' ,'89','99']}
df_in = pd.DataFrame(dic_in,columns=['A','B1','C1','B2','C2'])

如果我在控制台上打印它，它看起来像这样：

In [28]:df_in
Out[28]: 
    A  B1  C1  B2  C2
0  ff  33  44  33  44
1  rr  \N  \N  63  85
2  nn  39  74  \N  \N
3  qq  22  34  22  34
4  tt  \N  \N  71  52
5  pp  56  89  56  89
6  uu  90  99  90  99

我想要做的是调查列B1和C1的每一行：如果通用行在两列中都包含\N，则需要将其替换为分别为B2和C2的内容。以这种方式，输出（df_out）应如下所示：

In [28]:df_in                In[30]:df_out
Out[28]:                     Out[30]:
    A  B1  C1  B2  C2            A   B   C
0  ff  33  44  33  44        0  ff  33  44
1  rr  \N  \N  63  85 -----> 1  rr  63  85
2  nn  39  74  \N  \N -----> 2  nn  39  74
3  qq  22  34  22  34        3  qq  22  34
4  tt  \N  \N  71  52 -----> 4  tt  71  52
5  pp  56  89  56  89        5  pp  56  89
6  uu  90  99  90  99        6  uu  90  99

我能够使用以下代码行实现目标：

df_out = pd.DataFrame()
for index, row in df_in.iterrows():
    if row['B1']!=r'\N' and row['C1']!=r'\N':
        dic = {'A': [row['A']], 'B': [row['B1']], 'C': [row['C1']]}
        df_out = pd.concat([df_out,pd.DataFrame(dic)], ignore_index=True)
    else:
        dic = {'A': [row['A']], 'B': [row['B2']], 'C': [row['C2']]}
        df_out = pd.concat([df_out,pd.DataFrame(dic)], ignore_index=True)

你能否告诉我一个聪明的方法来实现这样的结果？

Answer 1

您可以先replace \N至NaN，然后combine_first或fillna：

df_out = df_in.replace({'\\N': np.nan})
df_out['B']= df_out.B1.combine_first(df_out.B2)
df_out['C'] = df_out.C1.combine_first(df_out.C2)
df_out = df_out[['A','B','C']]
print (df_out)
    A   B   C
0  ff  33  44
1  rr  63  85
2  nn  39  74
3  qq  22  34
4  tt  71  52
5  pp  56  89
6  uu  90  99

如果需要将带有B1的子集C1添加到带有B2的{{1}}：

C2

Answer 2

这是另一种解决方案，您可以在需要替换值时明确说明情况：

import pandas as pd

dic_in = {'A': ['ff','rr' ,'nn' ,'qq','tt' ,'pp','uu'],
      'B1': ['33',r'\N','39' ,'22',r'\N','56','90'],
      'C1': ['44',r'\N','74' ,'34',r'\N','89','99'],
      'B2': ['33','63' ,r'\N','22','71' ,'56','90'],
      'C2': ['44','85' ,r'\N','34','52' ,'89','99']}

df_in = pd.DataFrame(dic_in,columns=['A','B1','C1','B2','C2'])

df_out = pd.DataFrame(df_in['A'])

def substitute(row):
   return row[0] if row[0]!='\N' else row[1]

df_out['B'] = df_in[['B1', 'B2']].apply(substitute, axis = 1)
df_out['C'] = df_in[['C1', 'C2']].apply(substitute, axis = 1)

df_out
Out[35]: 
    A   B   C
0  ff  33  44
1  rr  63  85
2  nn  39  74
3  qq  22  34
4  tt  71  52
5  pp  56  89
6  uu  90  99

用Pandas替换数据框中不同列的值

2 个答案: