销毁二叉树中的节点

Question

Suppose I have a certain binary tree. I used new to create nodes, but now I want to delete them in the destructor:

void BinaryTree::recursiveDestructor(Node *& noeud){
    if (node != 0) {
        recursiveDestrutor(node->leftTree);
        recursiveDestructor(node->rightTree);
        delete node;
        node = 0;
    }
}

And I use the method in the destructor

BinaryTree::~BinaryTree(){
    recursiveDestructor(root);
}

Is there a way to just use a while or a for loop in the destructor instead of calling recursively a method inside the destructor? It would be faster using a loop and avoid using a method inside the destructor. If I can't use a loop, is there a way to improve the performance?

Answer 1

Why do you need a recursive dtor? Just do

Node::~Node()
{
   delete left;  // assuming left and right are both an instance of Node
   delete right;
}

BinaryTree::~BinaryTree() {
   delete root;
}

The dtor will be called down the whole tree working through all the nodes.

Answer 2

If your tree nodes store their parent it is possible to find the following node from any position and you could do a while loop on that but it will take more time than the recursive form (because the call to next() will require more than just a single node step half the time where the recursive-descent never takes more than a single step to find the next node to free).

Answer 3

You can breadth-delete a tree with a std::queue . Note: this is hammered out online with no syntax checking, but hopefully you get the idea:

BinaryTree::~BinaryTree()
{
    if (!root)
        return;

    std::queue<Node*> q;
    q.push(root);
    while (!q.empty())
    {
        Node *p = q.front();
        q.pop();
        if (p->left)
            q.push(p->left);
        if (p->right)
            q.push(p->right);
        delete p;
    }
}

Note: it is imperative that the nodes are not child-lifetime managed. Ie a node cannot auto-delete it's children on destruction. Doing so would invoke a double-delete and a boatload of undefined behavior.