使用 Python/Pandas 将两行数据合并到一行中

Question

I have a dataframe like this:

   ID   A1    A2    A3    A4                                      
0  01  100   101   103   104
1  01  501   502   503   504
2  01  701   702   703   704
3  02  1001  1002  1003  1004
4  03  2001  2002  2003  2004
5  03  5001  5002  5003  5004

I need the rows belonging to the same ID to be merged in a single row, the merged dataframe will be like this

   ID   A1    A2    A3    A4    B1    B2    B3     B4     C1   C2    C3    C4                                                   
0  01  101   102   103   104   501   502    503    504    701  702   703   704 
1  02  1001  2001  1003  1004  
2  03  2001  2002  2003  2004  5001  5002   5003   5004

I tried using np.random.permutation, np.roll etc but unable to get the desired result. The count of rows in my original data set is in thousands so loops and creating individual data sets and then merging is not helping

Answer 1

unstacked = df.unstack() gives you the first step:

A1  0    1001
    1    5001
    2    7001
A2  0    1002
    1    5002
    2    7002
A3  0    1003
    1    5003
    2    7003
A4  0    1004
    1    5004
    2    7004

Then you can extract the two "levels" of the index:

colname = unstacked.index.get_level_values(0) # A1,A1,A1,A2,...
rownum = unstacked.index.get_level_values(1) # 0,1,2,0,...

Then convert them to the desired format:

idxchr = (rownum + ord('A')).map(chr) # A,B,C,A,...
idxnum = colname.str[1] # 1,1,1,2,...

And finally, overwrite the unstacked index:

unstacked.index = idxchr + idxnum

Result:

A1    1001
B1    5001
C1    7001
A2    1002
B2    5002
C2    7002
A3    1003
B3    5003
C3    7003
A4    1004
B4    5004
C4    7004

Edit: You edited your question while I wrote this answer, now you may need to enhance it a bit to work for the new example input you've posted.

Answer 2

This is how you do it:

import pandas as pd


def widen(x):
    num_rows = len(x)
    num_cols = len(x.columns)

    new_index = [
        chr(ord('A') + row_number) + str(col_number + 1)
        for row_number in range(num_rows)
        for col_number in range(num_cols)
    ]

    return pd.Series(x.loc[:, 'A1':].unstack().values, index=new_index)

res = df.groupby('ID').apply(widen).unstack()

The output would be:

        A1      A2      A3      A4      B1  ...      B4     C1     C2     C3     C4
ID                                          ...                                    
1    100.0   501.0   701.0   101.0   502.0  ...   503.0  703.0  104.0  504.0  704.0
2   1001.0  1002.0  1003.0  1004.0     NaN  ...     NaN    NaN    NaN    NaN    NaN
3   2001.0  5001.0  2002.0  5002.0  2003.0  ...  5004.0    NaN    NaN    NaN    NaN

caveat : this will only work assuming each ID won't have more than 26 rows.