在某些条件下与上面的行组合

Question

表格格式（空单元格为空，列为：字段、维度）

field | dimension
-----------------
a     | 
b     | abc
e     | efg
      | xyz
r     | abc
      | def
      | xyz

所需格式：

field | dimension
-----------------
a     | [nan]
b     | [abc]
e     | [efg, xyz]
r     | [abc, def, xyz]

我试过了：

df.dimension = [df.dimension]

并且打算在字段中找到每个空单元格的索引，并与上面的行合并。然而，我得到了——

值

错误：值的长度与索引的长度不匹配。

我也认为一定有比我接近它的方式更好的方法。提前致谢

Answer 1

使用：

df =(df.groupby(df['field'].ffill())['dimension']
       .apply(lambda x: np.nan if x.isnull().all() else list(x))
       .reset_index())
print (df)
  field        dimension
0     a              NaN
1     b            [abc]
2     e       [efg, xyz]
3     r  [abc, def, xyz]

---

df = (df[df['dimension'].notnull()].groupby(df['field'].ffill())['dimension']
                                  .apply(list)
                                  .reindex(pd.unique(df['field'].dropna()))
                                  .reset_index())
print (df)
  field        dimension
0     a              NaN
1     b            [abc]
2     e       [efg, xyz]
3     r  [abc, def, xyz]

但如果在列表中NaN没问题：

df =(df.groupby(df['field'].ffill())['dimension']
       .apply(list)
       .reset_index())
print (df)
  field        dimension
0     a            [nan]
1     b            [abc]
2     e       [efg, xyz]
3     r  [abc, def, xyz]

Answer 2

让我们试试：

df['field'] = df['field'].ffill()
df_out = df.groupby('field')['dimension'].apply(list).reset_index()

输出：

  field        dimension
0     a            [nan]
1     b            [abc]
2     e       [efg, xyz]
3     r  [abc, def, xyz]