计算累积和，而另一列的值保持不变

Question

对于以下df，我想计算列Inst_Dist的累积和，并保存为Cumu_Dist，而WDir_Deg的值保持不变。当WDir_Deg中的值发生变化时，我需要重新启动累积总和。

因此，

index | WDir_Deg | Inst_Dist | Cumu_Dist
0     | 289      | 20        | NaN
1     | 285      | 17        | NaN
2     | 285      | 19        | NaN
3     | 287      | 19        | NaN
4     | 289      | 10        | NaN

变

index | WDir_Deg | Inst_Dist | Cumu_Dist
0     | 289      | 20        | 20
1     | 285      | 17        | 17
2     | 285      | 19        | 36
3     | 287      | 19        | 19
4     | 289      | 10        | 10

我的非惯用（极慢）Python代码如下所示。如果有人可以指导我如何使代码更快和惯用，我真的很感激。

prev_angle = -1
curr_cumu_dist = 0
for curr_ind in df.index:
    curr_angle = df.loc[curr_ind, 'WDir_Deg']
    if prev_angle == curr_angle:
        curr_cumu_dist += df.loc[curr_ind, 'Inst_Dist']
        df.loc[curr_ind, 'Cumu_Dist'] = curr_cumu_dist
    else:
        prev_angle = curr_angle
        curr_cumu_dist = df.loc[curr_ind, 'Inst_Dist']
        df.loc[curr_ind, 'Cumu_Dist'] = curr_cumu_dist

Answer 1

使用helper qazxsw poi与qazxsw poi列比较Series，WDir_Deg和ne不等于连续组并将其传递给shift：

cumsum

详情：

DataFrameGroupBy.cumsum

Answer 2

有点棘手。引用这个问题/答案

s = df['WDir_Deg'].ne(df['WDir_Deg'].shift()).cumsum()
df['Cumu_Dist'] = df.groupby(s)['Inst_Dist'].cumsum()
print (df)
   WDir_Deg  Inst_Dist  Cumu_Dist
0       289         20         20
1       285         17         17
2       285         19         36
3       287         19         19
4       289         10         10

我做了这个解决方案

print (s)
0    1
1    2
2    2
3    3
4    4
Name: WDir_Deg, dtype: int32

哪个回报

Pandas groupby cumulative sum

这使用df['Cumu_Dist'] = df.groupby('WDir_Deg').Inst_Dist.cumsum() 版本

   index  WDir_Deg  Inst_Dist  Cumu_Dist
0      0       285         17         17
1      1       285         19         36
2      2       287         19         19
3      3       289         20         20