将两个 pandas 列合并为索引，创建新列，并将其列名作为值

Question

I have a df:

import pandas as pd
df = pd.DataFrame({"A": [1, 3, 7, 10], "B": [2, 5, 8, 11], "C": list("WXYZ") })  
print(df)

>>>    A   B  C
>>>0   1   2  W
>>>1   3   5  X
>>>2   7   8  Y
>>>3  10  11  Z

I want to combine now the two columns A and B as the index, keep C , and create a new column vals that represents the former A and B columns:

     vals  C
         
1       A  W
2       B  W
3       A  X
5       B  X
7       A  Y
8       B  Y
10      A  Z
11      B  Z

I tried several versions with stack , pivot , melt without success. I could limp through the finishing line with:

import numpy as np
n = len(df.A)
arr = np.zeros((2, 2*n))
arr[0, :n] = df.A
arr[0, n:] = df.B
arr[1, :n] = 0
arr[1, n:] = 1

new_df = pd.DataFrame(arr.T, columns=["ind", "vals"], dtype="int").set_index("ind")
new_df["C"] = np.tile(df.C, 2)
new_df.sort_index(inplace=True)
print(new_df)

>>>     vals  C
>>>ind         
>>>1       0  W
>>>2       1  W
>>>3       0  X
>>>5       1  X
>>>7       0  Y
>>>8       1  Y
>>>10      0  Z
>>>11      1  Z

which not only looks horrible but also has several drawbacks (dtype changes and such). I bet there is a better pandas solution to this problem.

Answer 1

It's indeed melt :

df.melt('C').set_index('value')

Output:

       C variable
value            
1      W        A
3      X        A
7      Y        A
10     Z        A
2      W        B
5      X        B
8      Y        B
11     Z        B