什么是更好的选择，为什么？该地图用作临时存储。它会将项目保留一段时间，然后刷新到db。
这是我用原子引用实现的伪代码：
public class Service {
private final AtomicReference<Map<String, Entity>> storedEntities = new AtomicReference<>(new HashMap<>());
private final AtomicReference<Map<String, Entity>> newEntities = new AtomicReference<>(new HashMap<>());
private final Dao dao;
public Service(Dao dao) {
this.dao = dao;
}
@Transactional
@Async
public CompletableFuture<Void> save() {
Map<String, Entity> map = newEntities.getAndSet(new HashMap<>());
return dao.saveAsync(map.values());
}
@Transactional(readOnly = true)
@Async
public CompletableFuture<Map<String, Entity>> readAll() {
return dao.getAllAsync().thenApply(map -> {
storedEntities.set(map);
return map;
});
}
@Scheduled(cron = "${cron}")
public void refreshNow() {
save();
readAll();
}
public void addNewentity(Entity entity) {
newEntities.getAndUpdate(map -> {
map.put(entity.getHash(), entity);
return map;
});
}
public AtomicReference<List<Entity>> getStoredEntities() {
return storedEntities.get().values();
}
public AtomicReference<List<Entity>> getNewEntities() {
return newEntities.get().values();
}
}
正如我所说，我只需要保留数据一段时间，然后通过cron将数据刷新到db。我对什么是更好的方法感兴趣-AR与CHM？
答案 0 :(得分：2)
首先，我假设您需要跨多个线程共享对某些资源（临时存储映射）的访问权限。
简短答案：
使用ConcurrentHashMap。
较长的答案：
如果您期望通过使用AtomicReference来获得一致的或线程安全的地图视图，或者您的操作将自动执行，那么您很可能是错误的-但是，由于您尚未提供您用法的示例，我不能完全确定。
尽管不能忽略，但性能不应成为您关注的主要问题-相反，您应确保程序具有按预期正确执行的能力，然后寻求提高其性能。
如果有多个线程正在读取和/或写入临时存储映射，则保持映射的内部状态一致和正确很重要；您可以通过以atomic和线程安全的方式实施操作来实现这一点，在此程度上，ConcurrentHashMap或由SynchronizedMap包装的映射都可以实现这一点。但是，上述每种方法都以不同的粒度实现了确保这一点的手段，并且由于ConcurrentHashMap采取了专门的（乐观）方法，与包装后的地图的朴素（悲观）方法相反，ConcurrentHashMap较少容易受到资源争用的影响，并且可能在两者中表现更好。
前进
我在下面的这篇文章中提供了三种机制的实现；除了Java API文档外，还可以浏览以下代码。
import java.io.PrintStream;
import java.text.DecimalFormat;
import java.util.Collections;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.AtomicReference;
import java.util.function.BiFunction;
import java.util.stream.Collectors;
import java.util.stream.IntStream;
public class ConcurrencyExample {
interface TemporaryStorage<K, V> {
V compute(K key, BiFunction<? super K, ? super V, ? extends V> remapper);
V put(K key, V value);
V get(K key);
void clear();
@FunctionalInterface
interface UnitTest<K, V> {
void test(TemporaryStorage<K, V> store, K key);
}
}
static class ConcurrentHashMapTS<K, V> implements TemporaryStorage<K, V> {
private final Map<K, V> map = new ConcurrentHashMap<>();
@Override
public V compute(K key, BiFunction<? super K, ? super V, ? extends V> remapper) {
return map.compute(key, remapper);
}
@Override
public V put(K key, V value) {
return map.put(key, value);
}
@Override
public V get(K key) {
return map.get(key);
}
@Override
public void clear() {
map.clear();
}
}
static class AtomicReferenceHashMapTS<K, V> implements TemporaryStorage<K, V> {
private final AtomicReference<Map<K, V>> map = new AtomicReference<>(new HashMap<>());
@Override
public V compute(K key, BiFunction<? super K, ? super V, ? extends V> remapper) {
return map.get().compute(key, remapper);
}
@Override
public V put(K key, V value) {
return map.get().put(key, value);
}
@Override
public V get(K key) {
return map.get().get(key);
}
@Override
public void clear() {
map.get().clear();
}
}
static class MonitorLockedHashMapTS<K, V> implements TemporaryStorage<K, V> {
private final Map<K, V> map = new HashMap<>();
private final Object mutex = new Object(); // could use the map as the mutex
@Override
public V compute(K key, BiFunction<? super K, ? super V, ? extends V> remapper) {
synchronized (mutex) {
return map.compute(key, remapper);
}
}
@Override
public V put(K key, V value) {
synchronized (mutex) {
return map.put(key, value);
}
}
@Override
public V get(K key) {
synchronized (mutex) {
return map.get(key);
}
}
@Override
public void clear() {
synchronized (mutex) {
map.clear();
}
}
}
static class WrappedHashMapTS<K, V> implements TemporaryStorage<K, V> {
private final Map<K, V> map = Collections.synchronizedMap(new HashMap<>());
@Override
public V compute(K key, BiFunction<? super K, ? super V, ? extends V> remapper) {
return map.compute(key, remapper);
}
@Override
public V put(K key, V value) {
return map.put(key, value);
}
@Override
public V get(K key) {
return map.get(key);
}
@Override
public void clear() {
map.clear();
}
}
static class AtomicUnitTest implements TemporaryStorage.UnitTest<Integer, Integer> {
@Override
public void test(TemporaryStorage<Integer, Integer> store, Integer key) {
store.compute(key, (k, v) -> (v == null ? 0 : v) + 1);
}
}
static class UnsafeUnitTest implements TemporaryStorage.UnitTest<Integer, Integer> {
@Override
public void test(TemporaryStorage<Integer, Integer> store, Integer key) {
Integer value = store.get(key);
store.put(key, (value == null ? 0 : value) + 1);
}
}
public static class TestRunner {
public static void main(String... args) throws InterruptedException {
final int iterations = 1_000;
final List<Integer> keys = IntStream.rangeClosed(1, iterations).boxed().collect(Collectors.toList());
final int expected = iterations;
for (int batch = 1; batch <= 5; batch++) {
System.out.println(String.format("--- START BATCH %d ---", batch));
test(System.out, new ConcurrentHashMapTS<>(), new AtomicUnitTest(), keys, expected, iterations);
test(System.out, new ConcurrentHashMapTS<>(), new UnsafeUnitTest(), keys, expected, iterations);
test(System.out, new AtomicReferenceHashMapTS<>(), new AtomicUnitTest(), keys, expected, iterations);
test(System.out, new AtomicReferenceHashMapTS<>(), new UnsafeUnitTest(), keys, expected, iterations);
test(System.out, new MonitorLockedHashMapTS<>(), new AtomicUnitTest(), keys, expected, iterations);
test(System.out, new MonitorLockedHashMapTS<>(), new UnsafeUnitTest(), keys, expected, iterations);
test(System.out, new WrappedHashMapTS<>(), new AtomicUnitTest(), keys, expected, iterations);
test(System.out, new WrappedHashMapTS<>(), new UnsafeUnitTest(), keys, expected, iterations);
System.out.println(String.format("--- END BATCH %d ---", batch));
System.out.println();
}
}
private static <K, V> void test(PrintStream printer, TemporaryStorage<K, V> store, TemporaryStorage.UnitTest<K, V> work, List<K> keys, V expected, int iterations) throws InterruptedException {
test(printer, store, work, keys, expected, iterations, Runtime.getRuntime().availableProcessors() * 4);
}
private static <K, V> void test(PrintStream printer, TemporaryStorage<K, V> store, TemporaryStorage.UnitTest<K, V> work, List<K> keys, V expected, int iterations, int parallelism) throws InterruptedException {
final ExecutorService workers = Executors.newFixedThreadPool(parallelism);
final long start = System.currentTimeMillis();
for (K key : keys) {
for (int iteration = 1; iteration <= iterations; iteration++) {
workers.execute(() -> {
try {
work.test(store, key);
} catch (Exception e) {
//e.printStackTrace(); //thrown by the AtomicReference<Map<K, V>> implementation
}
});
}
}
workers.shutdown();
workers.awaitTermination(Long.MAX_VALUE, TimeUnit.DAYS);
final long finish = System.currentTimeMillis();
final DecimalFormat formatter = new DecimalFormat("###,###");
final long correct = keys.stream().filter(key -> expected.equals(store.get(key))).count();
printer.println(String.format("Store '%s' performed %s iterations of %s across %s threads in %sms. Accuracy: %d / %d (%4.2f percent)", store.getClass().getSimpleName(), formatter.format(iterations), work.getClass().getSimpleName(), formatter.format(parallelism), formatter.format(finish - start), correct, keys.size(), ((double) correct / keys.size()) * 100));
}
}
}