跳表(Skip List)是一种概率性数据结构,它通过在普通有序链表的基础上增加多级索引层来实现快速的查找、插入和删除操作。跳表的效率可以与平衡树相媲美,其操作的时间复杂度也是O(log n),但跳表的结构更简单,更易于实现。
跳表通过简单的随机化过程来避免复杂的重平衡操作,使得它成为一种既高效又易于实现的数据结构选项。
import java.util.Random;
class SkipListNode {
int value;
SkipListNode[] forward; // 指向不同层的指针数组
public SkipListNode(int value, int level) {
this.value = value;
this.forward = new SkipListNode[level + 1];
}
}
public class SkipList {
private static final float P = 0.5f;
private static final int MAX_LEVEL = 16;
private SkipListNode head;
private int level;
private Random random;
public SkipList() {
level = 0;
head = new SkipListNode(0, MAX_LEVEL);
random = new Random();
}
// 随机生成节点的层数
private int randomLevel() {
int lvl = 1;
while (random.nextFloat() < P && lvl < MAX_LEVEL) {
lvl++;
}
return lvl;
}
// 插入节点
public void insert(int value) {
int lvl = randomLevel();
SkipListNode newNode = new SkipListNode(value, lvl);
SkipListNode current = head;
SkipListNode[] update = new SkipListNode[MAX_LEVEL + 1];
for (int i = level; i >= 0; i--) {
while (current.forward[i] != null && current.forward[i].value < value) {
current = current.forward[i];
}
update[i] = current;
}
for (int i = 0; i <= lvl; i++) {
newNode.forward[i] = update[i].forward[i];
update[i].forward[i] = newNode;
}
if (lvl > level) {
level = lvl;
}
}
// 查找节点
public boolean search(int value) {
SkipListNode current = head;
for (int i = level; i >= 0; i--) {
while (current.forward[i] != null && current.forward[i].value < value) {
current = current.forward[i];
}
}
current = current.forward[0];
return current != null && current.value == value;
}
// 删除节点
public void delete(int value) {
SkipListNode[] update = new SkipListNode[MAX_LEVEL + 1];
SkipListNode current = head;
for (int i = level; i >= 0; i--) {
while (current.forward[i] != null && current.forward[i].value < value) {
current = current.forward[i];
}
update[i] = current;
}
current = current.forward[0];
if (current.value == value) {
for (int i = 0; i <= level; i++) {
if (update[i].forward[i] != current) break;
update[i].forward[i] = current.forward[i];
}
while (level > 0 && head.forward[level] == null) {
level--;
}
}
}
// 打印跳表的内容
public void display() {
System.out.println("SkipList: ");
for (int i = 0; i <= level; i++) {
SkipListNode node = head.forward[i];
System.out.print("Level " + i + ": ");
while (node != null) {
System.out.print(node.value + " ");
node = node.forward[i];
}
System.out.println();
}
}
}
// 使用示例
public class Main {
public static void main(String[] args) {
SkipList list = new SkipList();
list.insert(3);
list.insert(6);
list.insert(7);
list.insert(9);
list.insert(12);
list.insert(19);
list.insert(17);
list.display();
System.out.println("Searching 6: " + list.search(6));
System.out.println("Searching 15: " + list.search(15));
list.delete(6);
System.out.println("After deleting 6: ");
list.display();
}
}
这段代码首先定义了SkipListNode类,它是跳表节点的结构,包括节点值和一个数组forward,数组中每个元素是对应层级的下一个节点的引用。SkipList类实现了跳表,包括初始化、插入、查找、删除和打印跳表的方法。
假设我们有一个跳表,它当前的状态如下,其中每一行代表一个层级(层级0是最底层,包含所有元素):
层级3:1 --------------------------------> 9 层级2:1 ------------> 5 ------------> 9 层级1:1 ----> 3 ----> 5 ----> 7 ----> 9 层级0:1 -> 2 -> 3 -> 4 -> 5 -> 6 -> 7 -> 9
现在,我们想要插入一个新的节点,值为8,并假设通过随机过程,决定新节点8将出现在层级0、1和2上(不出现在层级3上)。下面是插入过程的步骤:
从跳表的最高层(在这个例子中是层级3)开始寻找,直到找到比插入值小的最大节点。因为8不会被插入到层级3,我们直接从层级2开始:
插入8后,跳表变为:
层级3:1 --------------------------------> 9 层级2:1 ------------> 5 -------> 8 ----> 9 层级1:1 ----> 3 ----> 5 ----> 7 -> 8 -> 9 层级0:1 -> 2 -> 3 -> 4 -> 5 -> 6 -> 7 -> 8 -> 9
在这个例子中,新插入的节点8并没有增加跳表的总层数,因此不需要调整。
通过这个例子,你可以看到插入过程如何在每一层找到正确的插入位置,并更新指针来维护跳表的结构。这个过程确保了跳表的搜索效率,使得搜索、插入和删除操作的时间复杂度都为O(log n)。