没有pandas标记，但如果您对此持开放态度，这里有一个关于pandas.concat的 * 命题 *：

#pip install pandas
import pandas as pd

df = pd.read_csv("infile.txt", sep=",") # <- change the sep here to match yours

S1, S2 = 2, 1 #columns to skip at the beginning and while concatenating
N, M = 3, 5 #first range of columns to concatenate

(
    pd.concat([df.astype(str).iloc[:, i+S1:i+M].agg("".join, axis=1)
               for i in range(0, df.shape[1]-M, N+S2)], axis=1)
        .pipe(lambda df_: df_.set_axis(df_.columns.map(lambda x: f"Join {x+1}"), axis=1))
        .to_csv("outfile.csv", sep=",") # <- change the sep here to match yours
)

输出（* 中间体 *）：

Join 1 Join 2  Join 3
0    345    789  111213