Count sub-sequence by other key on PostgreSQL w/o locking table or making table for numbering?

Question

Is there any way to manage sub-sequence with composite primary key table, rather than using "LOCK TABLE and get max(id)" or making table