在我们内存或者cache中有两种替换算法来保证内存或者cache中都是“热点”的数据,一个是LRU和LFU。下面会先分析再给出代码!
在这里的get和set操作的时间复杂度都是O(1),因为这两种速度都是很快的,不能要求O(n)。
一、LRU
首先我们搞一个双向链表。
我们定义这个规则:如果有节点被访问(get或者set),那么直接放入到末尾,每一个队列都有一个capacity值表示容量,如果新元素加入,现在达到了容量值,就需要淘汰最近最少使用的,就是头节点。例如:现在node1被访问。
如果现在又来了一个新元素noden+1,超过了容量值。所以:剔除node2,noden+1条件到末尾
这就是最近最少使用的算法,热点数据都在队尾。删除只删除头节点。
基础数据结构Node
public static class Node<V> {
public V value;
public Node<V> pre;
public Node<V> next;
public Node(V value) {
this.value = value;
}
}
双向队列
基本属性和构造函数
public static class DoubleLinkedList<V> {
private Node<V> head;
private Node<V> tail;
public DoubleLinkedList() {
this.head = null;
this.tail = null;
}
下面我们来看方法:
public void addNode(Node<V> newNode) {
if (newNode == null) {
return;
}
if (this.head == null) {
this.head = newNode;
this.tail = newNode;
} else {
this.tail.next = newNode;
newNode.pre = this.tail;
this.tail = newNode;
}
}
添加新节点,如果当前是第一个元素,头指针和尾指针都指向它,如果不是,放在队尾
public void moveNodeToTail(Node<V> node) {
if (this.tail == null) {
return;
}
if (this.head == node) {
this.head = node.next;
this.head.pre = null;
} else {
node.pre.next = node.next;
node.next.pre = node.pre;
}
this.tail.next = node;
node.pre = this.tail;
this.tail = node;
this.tail.next = null;
}
将一个节点移动到队尾,分为该节点是不是头节点,如果是头节点指向下一个,如果不是,就直接拿出来,然后放到队尾。
public Node<V> removeHead() {
if(this.head == null){
return this.head;
}
Node<V> res = this.head;
if(this.head == this.tail){
this.head = null;
this.tail = null;
}else{
this.head = res.next;
res.next = null;
this.head.pre = null;
}
return res;
}
删除头节点并返回。
Cache数据结构
基本属性和构造方法
public static class MyCache<K,V>{
private HashMap<K,Node<V>> keyNodeMap;
private HashMap<Node<V>,K> nodeKeyMap;
private DoubleLinkedList<V> nodeList;
private int capacity;
public MyCache(int capacity){
if(capacity < 1){
throw new RuntimeException();
}
this.nodeKeyMap = new HashMap<>();
this.keyNodeMap = new HashMap<>();
this.nodeList = new DoubleLinkedList<>();
this.capacity = capacity;
}
这里面用到了两个HashMap,为啥?因为hashmap的put和get的时间复杂度都是O(1),如果不用,遍历链表,那么我们的操作就是时间复杂度O(n),所以用到它俩主要是以空间换时间的方式提升效率!!!
get方法,获取value,判断key是否在map里面,如果再然后拿出节点,再移动到末尾。
public V get(K key){
if(this.keyNodeMap.containsKey(key)){
Node<V> res = this.keyNodeMap.get(key);
this.nodeList.moveNodeToTail(res);
return res.value;
}
return null;
}
set方法,判断是否有,如果有取出,移动到末尾,如果没有新建节点,加入,然后判断是否达到了容量,如果达到了容量就剔除头部节点。
public void set(K key,V value){
if(this.keyNodeMap.containsKey(key)){
Node<V> target = keyNodeMap.get(key);
target.value = value;
this.nodeList.moveNodeToTail(target);
}else{
Node<V> node = new Node<>(value);
this.keyNodeMap.put(key,node);
this.nodeKeyMap.put(node,key);
this.nodeList.addNode(node);
if(this.keyNodeMap.size() == this.capacity + 1){
removeMostCache();
}
}
}
private void removeMostCache(){
Node<V> node = this.nodeList.removeHead();
K removeKey = this.nodeKeyMap.get(node);
this.keyNodeMap.remove(removeKey);
this.nodeKeyMap.remove(node);
}
这样就做到了get和set方法的时间复杂度O(1)。
二、LFU
最近最少使用算法,这个算法相比于LRU就有难度了。我们来看思路:两个数据结构。
小节点类型
/**
* 小节点类型
*/
public static class Node {
public Integer key;
public Integer value;
public Integer times;
public Node up;
public Node down;
public Node(int key, int value, int times) {
this.key = key;
this.value = value;
this.times = times;
}
}
大节点类型
/**
* 大节点类型
*/
public static class NodeList {
public Node head;
public Node tail;
public NodeList pre;
public NodeList next;
public NodeList(Node node) {
this.head = node;
this.tail = node;
}
实际上就是一个下面的结构,矩形表示NodeList,圆形表示Node。
我们的数据是字母,上面矩形表示访问(get和set方法)的次数。先解释一下我们的规则:
(1)NodeList只有在下面有节点的时候才会被创建,如果下面的节点被移除,那么删除该NodeList。
(2)NodeList中head和tail分别指向下面圆形双向链表的头跟尾部。他本身也是一个双向链表。
(3)如果现在有一个新节点W被访问。插入到1下面的双向链表的首部
如果此时A被访问:就到2下面的双向链表的首部。
如果A再次被访问:
如果此时2被访问。2下面没有元素了,就删除2,然后c放到3上。
如果此时A再次被访问,那么创建4的NodeList,然后插入A:
如果这个时候来了一个H,但是现在达到了最大容量,就要剔除一个,我们剔除1下面双向链表的尾节点。
这个NodeList的头节点的node双向链表的尾节点就是最近最少使用的节点。以上就是整个LFU的过程,下面我们来看代码!
/**
* 大节点类型
*/
public static class NodeList {
public Node head;
public Node tail;
public NodeList pre;
public NodeList next;
public NodeList(Node node) {
this.head = node;
this.tail = node;
}
/**
* 新到的节点放在头部位置
*/
public void addNodeFromHead(Node newNode) {
newNode.down = this.head;
this.head.up = newNode;
head = newNode;
}
public boolean isEmpty() {
return head == null;
}
/**
* 移除节点到下一个NodeList。
*/
public void deleteNode(Node node) {
if (this.head == this.tail) {
this.head = null;
this.tail = null;
} else {
if (node == head) {
head = node.down;
head.up = null;
} else if (node == tail) {
tail = node.up;
tail.down = null;
} else {
node.up.down = node.down;
node.down.up = node.up;
}
}
node.up = null;
node.down = null;
}
}
在NodeList里面我们提供了三个方法:
(1)addNodeFromHead:将新加入的节点放在头部位置,保证无论进入哪个NodeList都必须保证在头部位置。
(2)isEmpty:判断当前NodeList是否为null。
(3)deleteNode:就是将访问的节点,从该NodeList移除,这个节点可以是头尾节点或者中间节点。
public static class LFUCache {
private int capacity;
private int size;
//key => node
private HashMap<Integer, Node> records;
//存储node的nodelist节点
private HashMap<Node, NodeList> heads;
//nodelist头节点
private NodeList headList;
public LFUCache(int capacity) {
this.capacity = capacity;
this.size = 0;
this.records = new HashMap<>();
this.heads = new HashMap<>();
this.headList = null;
}
这里面有几个属性:
(1)capacity:表示最大容量
(2)size:表示当前大小。
(3)records:表示key和node,这里的key是Integer,value是node类型
(4)heads:表示的是Node的上层NodeList节点是哪个。
(5)headList:表示所有NodeList的双向链表的头一个。
下面看方法:
判断如果有这元素,就更新数据同时times次数+1。然后执行move方法移动到下一个节点。move方法后面会有,如果没在并且达到了容量,就剔除一个节点,然后放入一个新节点
public void set(int key, int value) {
if (records.containsKey(key)) {
Node node = records.get(key);
node.value = value;
node.times++;
NodeList curNodeList = heads.get(node);
//讲目标节点移动到下一个NodeList
move(node, curNodeList);
} else {
//达到容量
if (size == capacity) {
Node node = headList.tail;
headList.deleteNode(node);
//是否删除这个头
modifyHeadList(headList);
records.remove(node.key);
heads.remove(node);
size--;
}
Node node = new Node(key, value, 1);
//刚开始加入第一个元素
if (headList == null) {
headList = new NodeList(node);
} else {
if (headList.head.times.equals(node.times)) {
headList.addNodeFromHead(node);
} else {
NodeList newList = new NodeList(node);
newList.next = headList;
headList.pre = newList;
headList = newList;
}
}
records.put(key, node);
heads.put(node, headList);
size++;
}
}
剔除元素之后,要判断这个NodeList是不是没有node了,入股没有就删除,否则留下。
/**
* 如果当前nodelist为空,就更换headList,否则就不
*/
private boolean modifyHeadList(NodeList nodeList) {
if (nodeList.isEmpty()) {
if (headList == nodeList) {
headList = nodeList.next;
if (headList != null) {
headList.pre = null;
}
} else {
nodeList.pre.next = nodeList.next;
if (nodeList.next != null) {
nodeList.next.pre = nodeList.pre;
}
}
return true;
}
return false;
}
移动节点到新的NodeList
private void move(Node node, NodeList oldNodeList) {
oldNodeList.deleteNode(node);
//拿到前一个节点的NodeList
NodeList preList = modifyHeadList(oldNodeList) ? oldNodeList.pre : oldNodeList;
NodeList nextList = oldNodeList.next;
if (nextList == null) {
NodeList newList = new NodeList(node);
if (preList != null) {
preList.next = newList;
}
newList.pre = preList;
if (headList == null) {
headList = newList;
}
heads.put(node, newList);
} else {
if (nextList.head.times.equals(node.times)) {
nextList.addNodeFromHead(node);
heads.put(node, nextList);
} else {
NodeList newList = new NodeList(node);
if (preList != null) {
preList.next = newList;
}
newList.pre = preList;
newList.next = nextList;
nextList.pre = newList;
if (headList == nextList) {
headList = newList;
}
heads.put(node, newList);
}
}
}
get方法:
public int get(int key) {
if (!records.containsKey(key)) {
return -1;
}
Node node = records.get(key);
node.times++;
NodeList curNodeList = heads.get(node);
move(node, curNodeList);
return node.value;
}
完整代码:
LFU:
public class LFU {
/**
* 小节点类型
*/
public static class Node {
public Integer key;
public Integer value;
public Integer times;
public Node up;
public Node down;
public Node(int key, int value, int times) {
this.key = key;
this.value = value;
this.times = times;
}
}
/**
* 大节点类型
*/
public static class NodeList {
public Node head;
public Node tail;
public NodeList pre;
public NodeList next;
public NodeList(Node node) {
this.head = node;
this.tail = node;
}
/**
* 新到的节点放在头部位置
*/
public void addNodeFromHead(Node newNode) {
newNode.down = this.head;
this.head.up = newNode;
head = newNode;
}
public boolean isEmpty() {
return head == null;
}
/**
* 移除节点到下一个NodeList。
*/
public void deleteNode(Node node) {
if (this.head == this.tail) {
this.head = null;
this.tail = null;
} else {
if (node == head) {
head = node.down;
head.up = null;
} else if (node == tail) {
tail = node.up;
tail.down = null;
} else {
node.up.down = node.down;
node.down.up = node.up;
}
}
node.up = null;
node.down = null;
}
}
public static class LFUCache {
private int capacity;
private int size;
//key => node
private HashMap<Integer, Node> records;
//存储node的nodelist节点
private HashMap<Node, NodeList> heads;
//nodelist头节点
private NodeList headList;
public LFUCache(int capacity) {
this.capacity = capacity;
this.size = 0;
this.records = new HashMap<>();
this.heads = new HashMap<>();
this.headList = null;
}
public void set(int key, int value) {
if (records.containsKey(key)) {
Node node = records.get(key);
node.value = value;
node.times++;
NodeList curNodeList = heads.get(node);
//讲目标节点移动到下一个NodeList
move(node, curNodeList);
} else {
//达到容量
if (size == capacity) {
Node node = headList.tail;
headList.deleteNode(node);
//是否删除这个头
modifyHeadList(headList);
records.remove(node.key);
heads.remove(node);
size--;
}
Node node = new Node(key, value, 1);
//刚开始加入第一个元素
if (headList == null) {
headList = new NodeList(node);
} else {
if (headList.head.times.equals(node.times)) {
headList.addNodeFromHead(node);
} else {
NodeList newList = new NodeList(node);
newList.next = headList;
headList.pre = newList;
headList = newList;
}
}
records.put(key, node);
heads.put(node, headList);
size++;
}
}
public int get(int key) {
if (!records.containsKey(key)) {
return -1;
}
Node node = records.get(key);
node.times++;
NodeList curNodeList = heads.get(node);
move(node, curNodeList);
return node.value;
}
/**
* 如果当前nodelist为空,就更换headList,否则就不
*/
private boolean modifyHeadList(NodeList nodeList) {
if (nodeList.isEmpty()) {
if (headList == nodeList) {
headList = nodeList.next;
if (headList != null) {
headList.pre = null;
}
} else {
nodeList.pre.next = nodeList.next;
if (nodeList.next != null) {
nodeList.next.pre = nodeList.pre;
}
}
return true;
}
return false;
}
private void move(Node node, NodeList oldNodeList) {
oldNodeList.deleteNode(node);
//拿到前一个节点的NodeList
NodeList preList = modifyHeadList(oldNodeList) ? oldNodeList.pre : oldNodeList;
NodeList nextList = oldNodeList.next;
if (nextList == null) {
NodeList newList = new NodeList(node);
if (preList != null) {
preList.next = newList;
}
newList.pre = preList;
if (headList == null) {
headList = newList;
}
heads.put(node, newList);
} else {
if (nextList.head.times.equals(node.times)) {
nextList.addNodeFromHead(node);
heads.put(node, nextList);
} else {
NodeList newList = new NodeList(node);
if (preList != null) {
preList.next = newList;
}
newList.pre = preList;
newList.next = nextList;
nextList.pre = newList;
if (headList == nextList) {
headList = newList;
}
heads.put(node, newList);
}
}
}
}
}
LRU:
package day_05;
import java.util.HashMap;
/**
* @ClassName LRU
* @Description
* @Author 戴书博
* @Date 2020/6/13 20:18
* @Version 1.0
**/
public class LRU {
public static class Node<V> {
public V value;
public Node<V> pre;
public Node<V> next;
public Node(V value) {
this.value = value;
}
}
public static class DoubleLinkedList<V> {
private Node<V> head;
private Node<V> tail;
public DoubleLinkedList() {
this.head = null;
this.tail = null;
}
public void addNode(Node<V> newNode) {
if (newNode == null) {
return;
}
if (this.head == null) {
this.head = newNode;
this.tail = newNode;
} else {
this.tail.next = newNode;
newNode.pre = this.tail;
this.tail = newNode;
}
}
public void moveNodeToTail(Node<V> node) {
if (this.tail == null) {
return;
}
if (this.head == node) {
this.head = node.next;
this.head.pre = null;
} else {
node.pre.next = node.next;
node.next.pre = node.pre;
}
this.tail.next = node;
node.pre = this.tail;
this.tail = node;
this.tail.next = null;
}
public Node<V> removeHead() {
if(this.head == null){
return this.head;
}
Node<V> res = this.head;
if(this.head == this.tail){
this.head = null;
this.tail = null;
}else{
this.head = res.next;
res.next = null;
this.head.pre = null;
}
return res;
}
}
public static class MyCache<K,V>{
private HashMap<K,Node<V>> keyNodeMap;
private HashMap<Node<V>,K> nodeKeyMap;
private DoubleLinkedList<V> nodeList;
private int capacity;
public MyCache(int capacity){
if(capacity < 1){
throw new RuntimeException();
}
this.nodeKeyMap = new HashMap<>();
this.keyNodeMap = new HashMap<>();
this.nodeList = new DoubleLinkedList<>();
this.capacity = capacity;
}
public V get(K key){
if(this.keyNodeMap.containsKey(key)){
Node<V> res = this.keyNodeMap.get(key);
this.nodeList.moveNodeToTail(res);
return res.value;
}
return null;
}
public void set(K key,V value){
if(this.keyNodeMap.containsKey(key)){
Node<V> target = keyNodeMap.get(key);
target.value = value;
this.nodeList.moveNodeToTail(target);
}else{
Node<V> node = new Node<>(value);
this.keyNodeMap.put(key,node);
this.nodeKeyMap.put(node,key);
this.nodeList.addNode(node);
if(this.keyNodeMap.size() == this.capacity + 1){
removeMostCache();
}
}
}
private void removeMostCache(){
Node<V> node = this.nodeList.removeHead();
K removeKey = this.nodeKeyMap.get(node);
this.keyNodeMap.remove(removeKey);
this.nodeKeyMap.remove(node);
}
}
}