python - 从数据帧中的当前索引反向搜索

Question

问题陈述:
我有一个像这样的数据框:

   Description      Type      x      y  z
0       Branch  Actuated      0      0  0
1      Forward  Actuated   7.07   7.07  0
2    Backwards  Actuated   7.07  -2.93  0
3      Forward  Actuated  17.07  -2.93  0
4    Backwards  Actuated     10    -10  0
5      Forward  Actuated  17.07 -17.07  0
6          EOL  Actuated   7.07 -17.07  0
7      Forward  Actuated  -7.07  -7.07  0
8    Backwards  Actuated  -7.07   2.93  0
9      Forward  Actuated -17.07   2.93  0
10   Backwards  Actuated    -10     10  0
11     Forward  Actuated -17.07  17.07  0
12         EOL  Actuated  -7.07  17.07  0
13     Forward  Actuated     -0     10  0
14     Forward  Actuated     -0     20  0
15     Forward  Actuated     10      0  0

当遇到 EOL 行时，算法需要反向查找第一个 Branch 行并将其插入到 EOL ro 之后，如下所示:

   Description      Type      x      y  z
0       Branch  Actuated      0      0  0
1      Forward  Actuated   7.07   7.07  0
2    Backwards  Actuated   7.07  -2.93  0
3      Forward  Actuated  17.07  -2.93  0
4    Backwards  Actuated     10    -10  0
5      Forward  Actuated  17.07 -17.07  0
6          EOL  Actuated   7.07 -17.07  0
0       Branch  Actuated      0      0  0 <--
7      Forward  Actuated  -7.07  -7.07  0
8    Backwards  Actuated  -7.07   2.93  0
9      Forward  Actuated -17.07   2.93  0
10   Backwards  Actuated    -10     10  0
11     Forward  Actuated -17.07  17.07  0
12         EOL  Actuated  -7.07  17.07  0
0       Branch  Actuated      0      0  0 <--
13     Forward  Actuated     -0     10  0
14     Forward  Actuated     -0     20  0
15     Forward  Actuated     10      0  0

注意:反向搜索应该基于原始数据帧而不是增强数据帧。

问题:如何高效(资源宽松且快速)完成此任务？

Answer 1

使用merge_asof要获取最后匹配的行，请通过 concat 连接在一起, DataFrame.sort_index最后DataFrame.reset_index使用 drop=True 来防止重复的索引值:

df1 = df[df['Description'] == 'Branch']
print (df1)
  Description      Type    x    y  z
0      Branch  Actuated  0.0  0.0  0

df2 = df[df['Description'] == 'EOL']
print (df2)
   Description      Type     x      y  z
6          EOL  Actuated  7.07 -17.07  0
12         EOL  Actuated -7.07  17.07  0

df3=pd.merge_asof(df2,df1,left_index=True,right_index=True,suffixes=('_',''))[df1.columns]
print (df3)
   Description      Type    x    y  z
6       Branch  Actuated  0.0  0.0  0
12      Branch  Actuated  0.0  0.0  0

df = pd.concat([df, df3]).sort_index().reset_index(drop=True)

<小时/>

print (df)
   Description      Type      x      y  z
0       Branch  Actuated   0.00   0.00  0
1      Forward  Actuated   7.07   7.07  0
2    Backwards  Actuated   7.07  -2.93  0
3      Forward  Actuated  17.07  -2.93  0
4    Backwards  Actuated  10.00 -10.00  0
5      Forward  Actuated  17.07 -17.07  0
6          EOL  Actuated   7.07 -17.07  0
7       Branch  Actuated   0.00   0.00  0
8      Forward  Actuated  -7.07  -7.07  0
9    Backwards  Actuated  -7.07   2.93  0
10     Forward  Actuated -17.07   2.93  0
11   Backwards  Actuated -10.00  10.00  0
12     Forward  Actuated -17.07  17.07  0
13         EOL  Actuated  -7.07  17.07  0
14      Branch  Actuated   0.00   0.00  0
15     Forward  Actuated  -0.00  10.00  0
16     Forward  Actuated  -0.00  20.00  0
17     Forward  Actuated  10.00   0.00  0