Pandas ließ sich an
df.merge(df2, left_on = "doc_id", right_on = "doc_num", how = "left")
Comfortable Cockroach
df.merge(df2, left_on = "doc_id", right_on = "doc_num", how = "left")
In [6]: result = pd.concat(frames, keys=['x', 'y', 'z'])