设计模式学习心得<享元模式 Flyweight>

享元模式（Flyweight Pattern）主要用于减少创建对象的数量，以减少内存占用和提高性能。这种类型的设计模式属于结构型模式，它提供了减少对象数量从而改善应用所需的对象结构的方式。

享元模式尝试重用现有的同类对象，如果未找到匹配的对象，则创建新对象。

概述

---

• 意图
  运用共享技术有效地支持大量细粒度的对象。

• 主要解决
  在有大量对象时，有可能会造成内存溢出，我们把其中共同的部分抽象出来，如果有相同的业务请求，直接返回在内存中已有的对象，避免重新创建。

• 何时使用

1. 系统中有大量对象。
2. 这些对象消耗大量内存。
3. 这些对象的状态大部分可以外部化。
4. 这些对象可以按照内蕴状态分为很多组，当把外蕴对象从对象中剔除出来时，每一组对象都可以用一个对象来代替。
5. 系统不依赖于这些对象身份，这些对象是不可分辨的。

• 如何解决
  用唯一标识码判断，如果在内存中有，则返回这个唯一标识码所标识的对象。

• 关键代码
  用 HashMap 存储这些对象。

• 应用实例

1. JAVA 中的 String，如果有则返回，如果没有则创建一个字符串保存在字符串缓存池里面。
2. 数据库的数据池。
3. hibernate-validate自定义了基于弱引用的缓存容器ConcurrentReferenceHashMap

• 优点
  大大减少对象的创建，降低系统的内存，使效率提高。

• 缺点
  提高了系统的复杂度，需要分离出外部状态和内部状态，而且外部状态具有固有化的性质，不应该随着内部状态的变化而变化，否则会造成系统的混乱。

• 使用场景

1. 系统有大量相似对象。
2. 需要缓冲池的场景。

• 注意事项

1. 注意划分外部状态和内部状态，否则可能会引起线程安全问题。
2. 这些类必须有一个工厂对象加以控制。

实现

---

摘抄hibernate-validate源码

容器：ConcurrentReferenceHashMap

调用摘抄

/**
 * This manager is in charge of providing all constraint related meta data
 * required by the validation engine.
 * <p>
 * Actual retrieval of meta data is delegated to {@link MetaDataProvider}
 * implementations which load meta-data based e.g. based on annotations or XML.
 * </p>
 * <p>
 * For performance reasons a cache is used which stores all meta data once
 * loaded for repeated retrieval. Upon initialization this cache is populated
 * with meta data provided by the given <i>eager</i> providers. If the cache
 * doesn't contain the meta data for a requested type it will be retrieved on
 * demand using the annotation based provider.
 * </p>
 *
 * @author Gunnar Morling
 * @author Chris Beckey &lt;cbeckey@paypal.com&gt;
 * @author Guillaume Smet
*/
public class BeanMetaDataManager {
	/**
	 * The default initial capacity for this cache.
	 */
	private static final int DEFAULT_INITIAL_CAPACITY = 16;

	/**
	 * The default load factor for this cache.
	 */
	private static final float DEFAULT_LOAD_FACTOR = 0.75f;

	/**
	 * The default concurrency level for this cache.
	 */
	private static final int DEFAULT_CONCURRENCY_LEVEL = 16;

	/**
	 * Additional metadata providers used for meta data retrieval if
	 * the XML and/or programmatic configuration is used.
	 */
	@Immutable
	private final List<MetaDataProvider> metaDataProviders;

	/**
	 * Helper for builtin constraints and their validator implementations
	 */
	private final ConstraintHelper constraintHelper;

	/**
	 * Used for resolving generic type information.
	 */
	private final TypeResolutionHelper typeResolutionHelper;

	/**
	 * The {@link ValueExtractor} manager.
	 */
	private final ValueExtractorManager valueExtractorManager;

	private final ExecutableParameterNameProvider parameterNameProvider;

	/**
	 * Used to cache the constraint meta data for validated entities
	 */
	private final ConcurrentReferenceHashMap<Class<?>, BeanMetaData<?>> beanMetaDataCache;

	/**
	 * Used for resolving type parameters. Thread-safe.
	 */
	private final ExecutableHelper executableHelper;

	private final ValidationOrderGenerator validationOrderGenerator;

	/**
	 * the three properties in this field affect the invocation of rules associated to section 4.5.5
	 * of the specification.  By default they are all false, if true they allow
	 * for relaxation of the Liskov Substitution Principal.
	 */
	private final MethodValidationConfiguration methodValidationConfiguration;

	public BeanMetaDataManager(ConstraintHelper constraintHelper,
			ExecutableHelper executableHelper,
			TypeResolutionHelper typeResolutionHelper,
			ExecutableParameterNameProvider parameterNameProvider,
			ValueExtractorManager valueExtractorManager,
			ValidationOrderGenerator validationOrderGenerator,
			List<MetaDataProvider> optionalMetaDataProviders,
			MethodValidationConfiguration methodValidationConfiguration) {
		this.constraintHelper = constraintHelper;
		this.executableHelper = executableHelper;
		this.typeResolutionHelper = typeResolutionHelper;
		this.valueExtractorManager = valueExtractorManager;
		this.parameterNameProvider = parameterNameProvider;
		this.validationOrderGenerator = validationOrderGenerator;

		this.methodValidationConfiguration = methodValidationConfiguration;

		this.beanMetaDataCache = new ConcurrentReferenceHashMap<>(
				DEFAULT_INITIAL_CAPACITY,
				DEFAULT_LOAD_FACTOR,
				DEFAULT_CONCURRENCY_LEVEL,
				SOFT,
				SOFT,
				EnumSet.of( IDENTITY_COMPARISONS )
		);

		AnnotationProcessingOptions annotationProcessingOptions = getAnnotationProcessingOptionsFromNonDefaultProviders( optionalMetaDataProviders );
		AnnotationMetaDataProvider defaultProvider = new AnnotationMetaDataProvider(
				constraintHelper,
				typeResolutionHelper,
				valueExtractorManager,
				annotationProcessingOptions
		);
		List<MetaDataProvider> tmpMetaDataProviders = new ArrayList<>( optionalMetaDataProviders.size() + 1 );
		// We add the annotation based metadata provider at the first position so that the entire metadata model is assembled
		// first.
		// The other optional metadata providers will then contribute their additional metadata to the preexisting model.
		// This helps to mitigate issues like HV-1450.
		tmpMetaDataProviders.add( defaultProvider );
		tmpMetaDataProviders.addAll( optionalMetaDataProviders );

		this.metaDataProviders = CollectionHelper.toImmutableList( tmpMetaDataProviders );
	}

	@SuppressWarnings("unchecked")
	public <T> BeanMetaData<T> getBeanMetaData(Class<T> beanClass) {
		Contracts.assertNotNull( beanClass, MESSAGES.beanTypeCannotBeNull() );

		BeanMetaData<T> beanMetaData = (BeanMetaData<T>) beanMetaDataCache.computeIfAbsent( beanClass,
				bc -> createBeanMetaData( bc ) );

		return beanMetaData;
	}

	public void clear() {
		beanMetaDataCache.clear();
	}

	public int numberOfCachedBeanMetaDataInstances() {
		return beanMetaDataCache.size();
	}
}