Jakarta Persistence 3.0，第4.12节写道:

相同抽象模式类型的两个实体当且仅当它们具有相同的主键值时相等。

我看不出为什么Java代码的行为应该有所不同。

If the entity class is in a so called "transient" state, i.e. it's not yet persisted and it has no identifier, then the hashCode/equals methods can not return a value, they ought to blow up, ideally implicitly with a NullPointerException when the method attempts to traverse the ID. Either way, this will effectively stop application code from putting a non-managed entity into a hash-based data structure. In fact, why not go one step further and blow up if the class and identifier are equal, but other important attributes such as the version are unequal (IllegalStateException)! Fail-fast in a deterministic way is always the preferred option.

警告:也要记录下爆发行为。文档本身很重要，但它也希望能够阻止初级开发人员在未来对您的代码做一些愚蠢的事情(他们倾向于压制发生NullPointerException的地方，他们最不关心的是副作用，lol)。

哦，总是使用getClass()而不是instanceof。equals方法要求对称性。如果b等于a，那么a必须等于b。对于子类，instanceof打破了这种关系(a不是b的实例)。

尽管我个人总是使用getClass()，即使在实现非实体类(类型是状态，所以子类添加状态，即使子类是空的或只包含行为)，只有当类是final时，instanceof才会很好。但实体类必须不是最终的(§2.1)，所以我们真的别无选择。

Some folks may not like getClass(), because of the persistence provider's proxy wrapping the object. This might have been a problem in the past, but it really shouldn't be. A provider not returning different proxy classes for different entities, well, I'd say that's not a very smart provider lol. Generally, we shouldn't solve a problem until there is a problem. And, it seems like Hibernate's own documentation doesn't even see it worthwhile mentioning. In fact, they elegantly use getClass() in their own examples (see this).

Lastly, if one has an entity subclass that is an entity, and the inheritance mapping strategy used is not the default ("single table"), but configured to be a "joined subtype", then the primary key in that subclass table will be the same as the superclass table. If the mapping strategy is "table per concrete class", then the primary key may be the same as in the superclass. An entity subclass is very likely to be adding state and therefore just as likely to be logically a different thing. But an equals implementation using instanceof can not necessarily and secondarily rely on the ID only, as we saw may be the same for different entities.

在我看来，instanceof在非final Java类中根本没有位置。对于持久实体来说尤其如此。

如果你想对你的set使用equals()/hashCode()，也就是说同一个实体只能出现一次，那么只有一个选项:选项2。这是因为根据定义，实体的主键永远不会改变(如果有人确实更新了它，它就不再是同一个实体了)

您应该从字面上理解:因为equals()/hashCode()是基于主键的，所以在设置主键之前，您不能使用这些方法。所以你不应该把实体放到集合里，直到它们被赋主键。(是的，uuid和类似的概念可能有助于早期分配主键。)

Now, it's theoretically also possible to achieve that with Option 3, even though so-called "business-keys" have the nasty drawback that they can change: "All you'll have to do is delete the already inserted entities from the set(s), and re-insert them." That is true - but it also means, that in a distributed system, you'll have to make sure, that this is done absolutely everywhere the data has been inserted to (and you'll have to make sure, that the update is performed, before other things occur). You'll need a sophisticated update mechanism, especially if some remote systems aren't currently reachable...

只有当集合中的所有对象都来自同一个Hibernate会话时，才可以使用选项1。Hibernate文档在13.1.3章中非常清楚地说明了这一点。考虑对象同一性:

Within a Session the application can safely use == to compare objects. However, an application that uses == outside of a Session might produce unexpected results. This might occur even in some unexpected places. For example, if you put two detached instances into the same Set, both might have the same database identity (i.e., they represent the same row). JVM identity, however, is by definition not guaranteed for instances in a detached state. The developer has to override the equals() and hashCode() methods in persistent classes and implement their own notion of object equality.

它继续主张选择3:

这里有一个警告:永远不要使用数据库标识符来实现相等。使用由唯一的、通常是不可变的属性组合而成的业务键。如果将瞬态对象持久化，则数据库标识符将更改。如果瞬态实例(通常与分离实例一起)保存在Set中，更改hashcode将破坏Set的契约。

这是真的，如果你

不能提前分配id(例如使用uuid) 当对象处于瞬态时，你肯定想把它们放到集合中。

否则，您可以自由选择选项2。

然后它提到了相对稳定性的需求:

业务键的属性不必像数据库主键那样稳定;只要对象在同一集合中，你就必须保证稳定性。

这是正确的。我所看到的实际问题是:如果你不能保证绝对的稳定性，你如何能够保证“只要对象在同一个集合中”的稳定性。我可以想象一些特殊的情况(比如只在对话中使用集合，然后将其丢弃)，但我会质疑这种方法的一般实用性。

短版:

选项1只能用于单个会话中的对象。 如果可以，使用选项2。(尽早分配PK，因为在分配PK之前你不能在集合中使用对象。) 如果你能保证相对的稳定性，你可以使用选项3。但是要小心。

我试着自己回答这个问题，直到我读了这篇文章，尤其是画了一个，我才完全满意找到的解决方案。我喜欢他懒创建UUID和最佳存储它的方式。

但我想增加更多的灵活性，即惰性创建UUID仅当hashCode()/equals()被访问时，第一次持久化实体与每个解决方案的优点:

Equals()表示“对象指向相同的逻辑实体” 尽可能使用数据库ID，因为为什么我要做两次工作(性能问题) 防止在尚未持久的实体上访问hashCode()/equals()时出现问题，并在它确实被持久后保持相同的行为

我真的很感激对我的混合解决方案的反馈如下

public class MyEntity { @Id() @Column(name = "ID", length = 20, nullable = false, unique = true) @GeneratedValue(strategy = GenerationType.IDENTITY) private Long id = null; @Transient private UUID uuid = null; @Column(name = "UUID_MOST", nullable = true, unique = false, updatable = false) private Long uuidMostSignificantBits = null; @Column(name = "UUID_LEAST", nullable = true, unique = false, updatable = false) private Long uuidLeastSignificantBits = null; @Override public final int hashCode() { return this.getUuid().hashCode(); } @Override public final boolean equals(Object toBeCompared) { if(this == toBeCompared) { return true; } if(toBeCompared == null) { return false; } if(!this.getClass().isInstance(toBeCompared)) { return false; } return this.getUuid().equals(((MyEntity)toBeCompared).getUuid()); } public final UUID getUuid() { // UUID already accessed on this physical object if(this.uuid != null) { return this.uuid; } // UUID one day generated on this entity before it was persisted if(this.uuidMostSignificantBits != null) { this.uuid = new UUID(this.uuidMostSignificantBits, this.uuidLeastSignificantBits); // UUID never generated on this entity before it was persisted } else if(this.getId() != null) { this.uuid = new UUID(this.getId(), this.getId()); // UUID never accessed on this not yet persisted entity } else { this.setUuid(UUID.randomUUID()); } return this.uuid; } private void setUuid(UUID uuid) { if(uuid == null) { return; } // For the one hypothetical case where generated UUID could colude with UUID build from IDs if(uuid.getMostSignificantBits() == uuid.getLeastSignificantBits()) { throw new Exception("UUID: " + this.getUuid() + " format is only for internal use"); } this.uuidMostSignificantBits = uuid.getMostSignificantBits(); this.uuidLeastSignificantBits = uuid.getLeastSignificantBits(); this.uuid = uuid; }

我个人已经在不同的项目中使用了这三种策略。我必须说，选项1在我看来是现实应用中最可行的。以我的经验来看，打破hashCode()/equals()一致性会导致许多疯狂的错误，因为你每次都会遇到这样的情况:在一个实体被添加到一个集合后，相等的结果发生了变化。

但也有更多的选择(也有它们的优点和缺点):

a) hashCode/equals基于一组不可变的、非空的、构造函数赋值的字段

(+)三个标准都有保证

(-)字段值必须可用以创建新实例

(-)如果你必须改变其中一个，处理起来会很复杂

b) hashCode/equals基于应用程序(在构造函数中)分配的主键，而不是JPA

(+)三个标准都有保证

(-)您不能利用简单可靠的ID生成策略，如DB序列

(-)如果在分布式环境(客户端/服务器)或应用服务器集群中创建新实体会很复杂

c) hashCode/equals基于实体的构造函数分配的UUID

(+)三个标准都有保证

(-)生成UUID的开销

(-)可能会有使用两次相同UUID的风险，这取决于所使用的算法(可能由DB上的唯一索引检测到)

这是每个使用Java和JPA的IT系统中的常见问题。痛点不仅仅是实现equals()和hashCode()，它还影响组织引用实体的方式以及其客户机引用同一实体的方式。我已经看够了没有商业钥匙的痛苦，以至于我写了自己的博客来表达我的观点。

简而言之:使用一个简短的、人类可读的、带有有意义前缀的顺序ID作为业务键，该ID生成时不依赖于RAM以外的任何存储。Twitter的雪花就是一个很好的例子。

下面是一个简单的(经过测试的)Scala解决方案。

请注意，此解决方案不属于这3类中的任何一类 在问题中给出。 我所有的实体都是UUIDEntity的子类，所以我遵循 不要重复自己(DRY)原则。 如果需要，可以使UUID生成更精确(通过使用更多 伪随机数)。

Scala代码:

import javax.persistence._
import scala.util.Random

@Entity
@Inheritance(strategy = InheritanceType.TABLE_PER_CLASS)
abstract class UUIDEntity {
  @Id  @GeneratedValue(strategy = GenerationType.TABLE)
  var id:java.lang.Long=null
  var uuid:java.lang.Long=Random.nextLong()
  override def equals(o:Any):Boolean= 
    o match{
      case o : UUIDEntity => o.uuid==uuid
      case _ => false
    }
  override def hashCode() = uuid.hashCode()
}