拆分行以在 Pandas Dataframe 中创建具有相同其他行值的新行

Question

我有一个 pandas 数据框，其中一列文本字符串包含多个逗号分隔的值。我想拆分每个字段并仅在逗号数> = 2的情况下为每个条目创建一个新行。例如，a应该变为b：

In [7]: a
Out[7]: 
    var1     var2   var3
0  a,b,c,d     1     X1
1  a,b,c,d     1     X2
2  a,b,c,d     1     X3
3  a,b,c,d     1     
4  e,f,g       2     Y1
5  e,f,g       2     Y2
6  e,f,g       2     
7  h,i         3     Z1
In [8]: b
Out[8]: 
     var1  var2   var3
0    a,d     1     X1
1    b,d     1     X2
3    c,d     1     X3
4    e,g     2     Y1
5    f,g     2     Y2
6    h,i     3     Z1

Answer 1

您可以使用自定义函数：

def custom_split(r):
    if r['var3']:
        s = r['var1']
        i = int(r['var3'][1:])-1
        l = s.split(',')
        return l[i]+','+l[-1]

df['var1'] = df.apply(custom_split, axis=1)
df = df.dropna()

输出：

  var1  var2 var3
0  a,d     1   X1
1  b,d     1   X2
2  c,d     1   X3
4  e,g     2   Y1
5  f,g     2   Y2
7  h,i     3   Z1

Answer 2

df['cc'] = df.groupby('var1')['var1'].cumcount()
df['var1'] = df['var1'].str.split(',')
df['var1'] = df[['cc','var1']].apply(lambda x: x['var1'][x['cc']]+','+x['var1'][-1],axis=1)
df = df.dropna().drop(columns=['cc']).reset_index(drop=True)
df 

Answer 3

您可以通过将逗号上的var1 拆分为列表来实现。 var3 中的整数减 1 可以解释为 var1 中列表中要保留的项目的索引：

import pandas as pd
import io

data = '''    var1     var2   var3
0  a,b,c,d     1     X1
1  a,b,c,d     1     X2
2  a,b,c,d     1     X3
3  a,b,c,d     1     
4  e,f,g       2     Y1
5  e,f,g       2     Y2
6  e,f,g       2     
7  h,i         3     Z1'''

df = pd.read_csv(io.StringIO(data), sep = r'\s\s+', engine='python')
df['var1'] = df["var1"].str.split(',').apply(lambda x: [[i,x[-1]] for i in x[:-1]]) #split the string to list and create combinations of all items with the last item in the list
df = df[df['var3'].notnull()] # drop rows where var3 is None
df['var1'] = df.apply(lambda x: x['var1'][0 if not x['var3'] else int(x['var3'][1:])-1], axis=1) #keep only the element in the list in var1 where the index is the integer in var3 minus 1

输出：

	var1	var2	var3
0	['a', 'd']	1	X1
1	['b', 'd']	1	X2
2	['c', 'd']	1	X3
4	['e', 'g']	2	Y1
5	['f', 'g']	2	Y2
7	['h', 'i']	3	Z1

运行df['var1'] = df['var1'].str.join(',') 将var1 重新转换为字符串。