在搜索系统中，如何缓存搜索最频繁的1000个搜索结果？自定制的精准短文本搜索服务项目代码（https://github.com/ysc/short-text-search/blob/master/src/main/java/org/apdplat/search/utils/ConcurrentLRUCache.java）。
本文利用了ConcurrentHashMap和AtomicLong实现了线程安全且支持高并发的最频繁访问驻留缓存算法，除了缓存功能，还提供了缓存状态查询接口，非常实用。
import java.util.Collections;
import java.util.Map;
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.concurrent.atomic.AtomicLong;
/**
* 最频繁访问驻留缓存算法
* Created by ysc on 7/18/16.
*/
public class ConcurrentLRUCache {
private int maxCacheSize;
private Map> cache = new ConcurrentHashMap<>();
private AtomicLong totalEvictCount = new AtomicLong();
private AtomicLong hitCount = new AtomicLong();
private AtomicLong notHitCount = new AtomicLong();
public ConcurrentLRUCache(int maxCacheSize) {
cache = new ConcurrentHashMap<>(maxCacheSize, 1, 10);
this.maxCacheSize = maxCacheSize;
}
public String getStatus(){
StringBuilder status = new StringBuilder();
long total = hitCount.get()+notHitCount.get();
status.append("最大缓存数量: ").append(maxCacheSize).append("\n")
.append("当前缓存数量: ").append(getActualCacheSize()).append("\n")
.append("驱逐缓存次数: ").append(totalEvictCount.get()).append("\n")
.append("命中缓存次数: ").append(hitCount.get()).append("\n")
.append("未命中缓存次数: ").append(notHitCount.get()).append("\n")
.append("缓存命中比例: ").append(total == 0 ? 0 : hitCount.get()/(float)total*100).append(" %\n");
return status.toString();
}
public String getKeyAndHitCount(){
StringBuilder status = new StringBuilder();
AtomicInteger i = new AtomicInteger();
cache.entrySet().stream().sorted((a,b)->b.getValue().getCount()-a.getValue().getCount()).forEach(entry->status.append(i.incrementAndGet()).append("\t").append(entry.getKey()).append("\t").append(entry.getValue().getCount()).append("\n"));
return status.toString();
}
public int getMaxCacheSize() {
return maxCacheSize;
}
public int getActualCacheSize() {
return cache.size();
}
public Map> getCache() {
return Collections.unmodifiableMap(cache);
}
public AtomicLong getTotalEvictCount() {
return totalEvictCount;
}
public long getHitCount() {
return hitCount.longValue();
}
public long getNotHitCount() {
return notHitCount.longValue();
}
public void put(K key, V value){
if(cache.size() >= maxCacheSize){
// evict count
int evictCount = (int)(maxCacheSize*0.1);
if(evictCount < 1){
evictCount = 1;
}
totalEvictCount.addAndGet(evictCount);
cache.entrySet().stream().sorted((a,b)->a.getValue().getCount()-b.getValue().getCount()).limit(evictCount).forEach(entry->cache.remove(entry.getKey()));
return;
}
cache.put(key, new CacheItem(value, new AtomicInteger()));
}
public V get(K key){
CacheItem item = cache.get(key);
if(item != null){
item.hit();
hitCount.incrementAndGet();
return item.getValue();
}
notHitCount.incrementAndGet();
return null;
}
private static class CacheItem{
private V value;
private AtomicInteger count;
public CacheItem(V value, AtomicInteger count) {
this.value = value;
this.count = count;
}
public void hit(){
this.count.incrementAndGet();
}
public V getValue() {
return value;
}
public int getCount() {
return count.get();
}
}
public static void main(String[] args) {
ConcurrentLRUCache cache = new ConcurrentLRUCache<>(5);
for(int i=0; i<9; i++){
cache.put(i, i);
if(i%2==0){
for(int j=0; j
cache.get(i);
}
}
}
System.out.println(cache.getStatus());
System.out.println(cache.getKeyAndHitCount());
}
}
