二进制搜索树递归析构函数

Binary search tree recursive destructor

本文关键字:析构函数 递归 搜索树 二进制      更新时间:2023-10-16

我检查过其他类似的主题,但似乎没有任何帮助。

我正试图为这个特定的BST实现编写一个析构函数。每个节点都包含一个指向父节点的指针,一个指向左节点的指针、一个指向右节点的指针和一个指向其包含的值的指针。这就是课程开始的样子:

#ifndef BST_H
#define BST_H
#include <iostream>
template <typename T>
class BST{
    private:
        BST<T> *left;
        BST<T> *right;
        BST<T> *parent;
        T *value;
    public:
        BST() {
            this->parent = NULL;
            this->left = NULL;
            this->right = NULL;
            this->value = NULL;
        }
        ~BST() {
            removeRecursively(this);
        }
        void removeRecursively(BST<T>* node) {
            if (node->left != NULL)
                removeRecursively(node->left);
            if (node->right != NULL)
                removeRecursively(node->right);
            if (node->left == NULL && node->right == NULL) {
                if (node->parent->left == node)
                    node->parent->left = NULL;
                if (node->parent->right == node)
                    node->parent->right = NULL;
                node->parent = NULL;
                node->value = NULL;
                delete node->value;
                delete node;
            }
        }
        void add(T value) {
            if (this->value == NULL) { // root-only case
                this->value = new T;
                *(this->value) = value;
            }
            else {
                if (value < *(this->value)) {
                    if (this->left != NULL) // has left child
                        this->left->add(value);
                    else { // has no left child
                        this->left = new BST<T>;
                        this->left->value = new T;
                        this->left->parent = this;
                        *(this->left->value) = value;
                    }
                }
                else {
                    if (this->right != NULL) // has right child
                        this->right->add(value);
                    else { // has no right child
                        this->right = new BST<T>;
                        this->right->value = new T;
                        this->right->parent = this;
                        *(this->right->value) = value;
                    }
                }
            }
        }
        void inOrderDisplay() {
            if (this->left != NULL)
                this->left->inOrderDisplay();
            std::cout << *(this->value) << " ";
            if (this->right != NULL)
                this->right->inOrderDisplay();
        }
        BST<T>* search(T value) {
            if (*(this->value) == value)
                return this;
            else if (this->left != NULL && value < *(this->value))
                return this->left->search(value);
            else if (this->right != NULL && value > *(this->value))
                return this->right->search(value);
            else
                return NULL;
        }
        BST<T>* remove(T value) {
            BST<T>* node = search(value);
            if (node != NULL) {
                if (node->left == NULL && node->right == NULL) { // leaf node
                    delete node->value;
                    if (node->parent->left == node) // is left child
                        node->parent->left = NULL;
                    else // is right child
                        node->parent->right = NULL;
                    delete node;
                }
                // 1 child nodes
                if (node->left != NULL && node->right == NULL) { // has left child
                    if (node->parent->left == node) // is left child
                        node->parent->left = node->left;
                    else // is right child
                        node->parent->right = node->left;
                    delete node->value;
                    node->parent = NULL;
                    delete node;
                }
                if (node->left == NULL && node->right != NULL) { // has right child
                    if (node->parent->left == node) // is left child
                        node->parent->left = node->right;
                    else // is right child
                        node->parent->right = node->right;
                    delete node->value;
                    node->parent = NULL;
                    delete node;
                }
                // 2 children nodes
                if (node->left != NULL && node->right != NULL) {
                    T aux;
                    BST<T>* auxNode = node->right;
                    while (auxNode->left != NULL)
                    auxNode = auxNode->left;
                    aux = *(auxNode->value);
                    if (auxNode->right != NULL) {
                        *(auxNode->value) = *(auxNode->right->value);
                        auxNode->right->parent = NULL;
                        auxNode->right->value = NULL;
                        auxNode->right = NULL;
                        delete auxNode->right;
                    }
                    else {
                        if (auxNode->parent->left == auxNode) // is left child
                            auxNode->parent->left = NULL;
                        else // is right child
                            auxNode->parent->right = NULL;
                        auxNode->value = NULL;
                        delete auxNode;
                    }
                    *(node->value) = aux; 
                }
            }
            return this;
        }
};
#endif // BST_H

BST类的使用如下:

BST<int>* root = new BST<int>();
root->add(5);
root->add(2);
root->add(-17);
root->inOrderDisplay();
root->remove(5);

我提到所有的方法都能正常工作(我决定不发布它们,因为它们不是这个问题的主题)。问题是,当我用Valgrind运行我的测试文件时,它检测到一些内存泄漏,我确信它们的发生是因为缺乏合适的析构函数(上面的析构因子会产生分段错误)。

编辑:我添加了其他方法的代码

谢谢!

你的基本设计有点破了,至少IMO是这样。也就是说,如果你愿意跳过足够多的障碍,你可能会让它发挥作用,但即使充其量,它也可能总是有点笨拙。

首先,我将为树中的节点定义一个单独的类。然后,树本身将保存一个指向树根的指针,并定义树的大部分接口。

template <class T>
class Tree {
    struct node {
        node *parent;
        node *left;
        node *right;
        T *value;
        node(T *value) 
            : parent(nullptr), left(nullptr), right(nullptr), value(new T(value))
        {             
        }
        ~node() { 
            delete(left);
            delete(right);
            delete(value);
        }           
    } *root;
    Tree() : root(nullptr) {}
    ~Tree() { 
        delete(root);
    }
};

销毁不必是显式递归的。delete(left)(例如)将为左边的子节点(如果有)调用dtor,然后为它的子节点调用dtor等等。当我们到达一个叶节点时,我们将得到(相当于)delete nullptr;,它被定义为什么都不做,停止递归。

您应该交换行:

            node->value = NULL;
            delete node->value;

像这样:

            delete node->value;
            node->value = NULL;

如果您首先指定NULL,则不会删除任何内容。

最难找到的bug是代码中那些你决定不看的bug,因为你确信它们不相关。

无论如何,你的实现有一个明显的问题:在删除了一堆其他东西之后,你基本上就是在做

class foo
{
    ~foo() { delete this; }
};

将RAII与unique_ptr一起使用可以避免以下问题:

template <typename T>
class BST{
public:
    BST() = default;
    ~BST() = default;
    BST(const BST&) = delete;
    BST& operator =(const BST&) = delete;
    BST(BST&&) = delete;
    BST& operator =(BST&&) = delete;
private:
    std::unique_ptr<BST<T>> left;
    std::unique_ptr<BST<T>> right;
    BST<T>* parent = nullptr;
    std::unique_ptr<T> value;
};