It was aimed at duplicates, but I did not take into account the details - just the general algorithm. I believe you can not do it in one sweep when there are duplicates.