Question

I’m trying to group some sets in disjoint sets. For example, if I have these 5 sets:

[[1, 3], [2], [1, 5], [6, 8], [1, 7]]

I want to have this result:

[[2], [6, 8], [1, 3, 5, 7]]

Here is the code:

import java.util.*;

public class SetTest {

public static void main(String[] args) {
    // ---- Initialisation
    Set<Set<Integer>> groups = new LinkedHashSet<Set<Integer>>();
    for (int[] set : new int[][] { {1, 3 }, {2 }, {1, 5 }, {6, 8 }, {1, 7} }) {
        Set<Integer> group = new TreeSet<Integer>();
        for (int i : set) {
            group.add(i);
        }
        groups.add(group);
    }
    System.out.println(groups);
    // ---- Grouping values in disjoint sets
    for (Iterator<Set<Integer>> iterator = groups.iterator(); iterator.hasNext();) {
        Set<Integer> group = iterator.next();
        System.out.println(String.format(" + Checking %20s in \t %s", group, groups));
        for (Set<Integer> other : groups) {
            if (!group.equals(other) && !Collections.disjoint(group, other)) {
                other.addAll(group);
                iterator.remove();
                System.out.println(String.format(" - Removed  %20s -> \t %s", group, groups));
                break;
            }
        }
    }
    System.out.println(groups);
}
}

I am using an iterator over the set, and I want to group 2 sets in one, removing one of them. However, I am having a problem with the Iterator.remove() method.
What this program prints is:

[[1, 3], [2], [1, 5], [6, 8], [1, 7]]
 + Checking               [1, 3] in      [[1, 3], [2], [1, 5], [6, 8], [1, 7]]
 - Removed                [1, 3] ->      [[2], [1, 3, 5], [6, 8], [1, 7]]
 + Checking                  [2] in      [[2], [1, 3, 5], [6, 8], [1, 7]]
 + Checking            [1, 3, 5] in      [[2], [1, 3, 5], [6, 8], [1, 7]]
 - Removed             [1, 3, 5] ->      [[2], [1, 3, 5], [6, 8], [1, 3, 5, 7]]
 + Checking               [6, 8] in      [[2], [1, 3, 5], [6, 8], [1, 3, 5, 7]]
 + Checking         [1, 3, 5, 7] in      [[2], [1, 3, 5], [6, 8], [1, 3, 5, 7]]
 - Removed          [1, 3, 5, 7] ->      [[2], [1, 3, 5, 7], [6, 8], [1, 3, 5, 7]]
[[2], [1, 3, 5, 7], [6, 8], [1, 3, 5, 7]]

The first time, removing [1, 3] works as expected, but the rest of the times, it does not remove the item. I think it’s because I use addAll(), but why is that? Because I don’t make changes in groups; I only change an element inside it(other) – and the reference is the same, right?

Answer 1

An element of a HashSet is supposed to have a stable hashCode, but you are mutating them as you iterate over the outer set. That fails in unpredictable, but documented ways.

In a TreeSet there may also be problems with changing the sort order of the element through mutation.