efficient mapping of very large dataframes

Question

I have two data frames df1 and df2 with the following columns:

A = [\'id\', \'a_id\', \'c\', \'d\']
B = [\'a_id\', \'e\', \'f\']