How to transform nested sets efficiently into a object hierarchy

Question

Suppose I have a list of nodes which represent a nested set hierachy (examples are in pseudo c#).

class Node
{
    public decimal left;
    public decimal right;
    public decimal id;
    public void AddChild(Node child) {...}
    ...
}

List<Node> nodes = GetFlatNodesWithoutChildrenSetFromDatabase();

The fields left, right and id get filled, since these values are stored in some database.

What is an efficient way to transform this flat list into a hierachy, that means filling in the appropriate children nodes for each parent node?

One way is just to find all ancestors of each node, sort them to find the parent node and add the child node to that node.

foreach (var n in nodes)
{
    var parent = nodes.Where(i => i.left < n.left && i.right > n.right).OrderBy(i => i.right - n.right).FirstOrDefault();
    if (parent != null)
        parent.AddChild(n);
}

But this is rather inefficient.

Is there a better (that means faster) approach?

EDIT

Possible solution (as suggested by Chris):

var stack = new Stack<Node>(nodes.Take(1));
foreach (var n in nodes.Skip(1))
{
    while (stack.Peek().right < n.left)
        stack.Pop();

    stack.Peek().addChild(n);
    stack.Push(n);
}

Nodes have to be ordered by left.

Yeah, I'm finding it hard to work out what you are doing with left and right to get the parent node there... — Chris, Jun 01 '12 at 14:54
left and right are numbers describing the position of the node in a nested set hierachy http://en.wikipedia.org/wiki/Nested_set_model — sloth, Jun 01 '12 at 14:55
Right, so the problem is essentially: you have a list of `(leftindex, rightindex, key)` triples and you want a tree of `(leftpointer, rightpointer, key)` nodes? — Fred Foo, Jun 01 '12 at 15:14

score 3 · Accepted Answer · answered Jun 01 '12 at 15:11

The method I might think about exploring is to order by left and then you can just iterate through once.

You "open" a node when you get to its left and stick it on a stack.

When you get to a new node to process you check if the node on the top of the stack should be closed by determining if its right is less than the new nodes left. If it is you remove it from the stack (closing it) and you have processed all its children. You then do the check for the new top of the stack til you find one that is still open. You then add the current node as a child to the node on the top of the stack and that node is then opened (so it goes on the top of the stack).

The diagram on the wikipedia page you linked (http://en.wikipedia.org/wiki/Nested_set_model) is what inspired me to this.

Nested Set Diagram

My algorithm basically travels down the line in the middle and whenever you enter one of the sets is what I refer to as opening and leaving a set is closing. Clearly the most recent set you have opened but not closed will be on the top of the stack and thus where you put the children.

I think the complexity of this should be O(nlogn) due to the sort. The rest of it is just O(n).

nice solution, since sorting is pretty cheap most of the time. — sloth, Jun 04 '12 at 07:05

Mironor · Answer 2 · 2014-11-01T18:11:04.097

I know that the question is quite old (I didn't find any other question/information on the topic) and I don't know "pseudo C#", but just in case some of you straggle with a recursive algorithm for nested sets list => tree algorithm, here is what I came to (in scala):

def retrieveUserFolderTree(user: User): Future[List[Folder]] = {
  // Get a list of user's folders orderred by left asc
  val dbFoldersPromise = folderDAO.findUserFolders(user)

  dbFoldersPromise.map {
    case rootFolder :: tail => generateChildren(0, rootFolder.right, tail)
    case Nil => Nil
  }
}

private def generateChildren(currentLeft: Int, currentRight: Int, dbFolders: Seq[DBFolder]): List[Folder] = {
  dbFolders match {
    case dbFolder :: tail if dbFolder.left > currentRight => Nil
    case dbFolder :: tail if dbFolder.left > currentLeft => Folder(dbFolder.id, dbFolder.name, generateChildren(dbFolder.left, dbFolder.right, tail)) :: generateChildren(dbFolder.right, currentRight, tail)
    case dbFolder :: tail => generateChildren(currentLeft, currentRight, tail)
    case Nil => Nil
  }
}

Hope this will help someone.

score 0 · Answer 3 · answered May 03 '16 at 15:41

Maybe I missed a step somewhere but when I was working on this using the logic above I ended up with a couple elements on the stack that were duplicated. They were in the tree as expected but additionally they were also on the top of the stack above the root node. I had to add a little loop at the end to clean up the stack.

        var stack = new Stack<DvrNode>(nodes.Take(1));
        foreach (var n in nodes.Skip(1))
        {
            while (stack.Peek().Right < n.Left)
                stack.Pop();

            ((List<DvrNode>)stack.Peek().Children).Add(n);
            stack.Push(n);
        }

        while (stack.Peek().Left != 1)
            stack.Pop();

How to transform nested sets efficiently into a object hierarchy

3 Answers3