Trimming a sorted set

Question

I have a SortedSet (specifically a TreeSet) containing updates. An update is something like an SVN commit, Facebook wall post, new Trac ticket, etc. I'm storing these in a SortedSet because:

Sorted: The updates need to be sorted by date, descending.
Set: When fetching the latest updates from an update source, I'll usually receive updates that are already in the set.

Now, after a while the set will grow really huge, so I'd like to remove anything but the first X items from the set (because others won't be displayed anyway). How can I do this, since it's not a List?

score 5 · Accepted Answer · answered Nov 28 '10 at 13:36

5

While(mySet.size() > limit) {
  mySet.remove(mySet.last());
}

answered Nov 28 '10 at 13:36

sje397

41,293
8
87
103

I don't know about the performance of that compared to my workaround, but it doesn't really matter. This is extremely easy to read – Bart van Heukelom Nov 28 '10 at 13:43
This doesn't do atomically. I need similar thing but done atomically – Wins Aug 25 '15 at 08:16

score 3 · Answer 2 · edited Oct 23 '18 at 21:31

3

The accepted answer in not really efficient O(ln n) because it uses remove function. Here is a better one since pollLast() is O(1) and size is O(1). Complexity will be only affected by trimmed size.

While(mySet.size() > limit) {
  mySet.pollLast();
}

edited Oct 23 '18 at 21:31

Alexander I.

2,380
3
17
42

answered Oct 23 '18 at 20:21

Wojtek Sierka

31
1

score 1 · Answer 3 · answered Nov 28 '10 at 13:51

The solution here should depend on the fact whether you need the "extra" data in future. If you need your solution based on additional list is ok. If not, I'd suggest the following:

Create your own sorted set that extends the java.util.SortedSet and overrides its add() method. This method should do nothing after certain limit. Alternatively you can create "wrapper" Set that holds payload set and delegates all methods except add(). The add() method should delegate its call only if size of payload set is less than the predefined limit. This is how FixedSizeSortedMap of jakarta collection framework works, so you can just use it.

The thing is, I don't want add to do nothing. New items will be added, and old ones should be removed. — Bart van Heukelom, Nov 28 '10 at 13:52

Nestor Milyaev · Answer 4 · 2016-10-19T16:07:19.703

0

Here's a working solution method for Java, given a TreeSet results, and a variable size specifying the size of the resulting set:

void setLimit(Set<T> resutls, int size) {
    List<T> items = new ArrayList<T>();
    items.addAll(resutls);
    resutls.clear();
    int trim = size>items.size() ? items.size() : size;
    resutls.addAll(items.subList(0,trim));
    // return results; // optionally, if required
}

edited Oct 19 '16 at 16:07

answered Oct 19 '16 at 15:47

Nestor Milyaev

5,845
2
35
51

score 0 · Answer 5 · answered Nov 28 '10 at 13:35

My own workaround is:

        List<Update> trimmed = new ArrayList<Update>(20);
        int i = 0;
        for (Update u : updates) {
            trimmed.add(u);
            i++;
            if (i > 20) break;
        }
        updates = new TreeSet<Update>(trimmed);

Trimming a sorted set

5 Answers5