在测试Hibernate的一对多双向关联映射时,碰到很有趣的问题,跟inverse属性直接相关。

1、People.hbm.xml

<hibernate-mapping default-lazy="false"> 
<class name="com.persistent.People" table="people"> 
<id name="id" column="peopleId" unsaved-value="0"> 
<generator class="increment"> 
</generator> 
</id> 
<property name="name" column="name"></property> 
<set name="addresses" cascade="save-update">
<key column="peopleId" not-null="true" />
<one-to-many class="com.persistent.Address"/>
</set> 
</class>
</hibernate-mapping>

2、Address.hbm.xml

<hibernate-mapping>
<class name="com.persistent.Address" table="address"> 
<id name="id" column="addressId" unsaved-value="0">
<generator class="increment">
</generator>
</id> 
<many-to-one name="people" column="peopleId" insert="false" update="false"></many-to-one> 
<property name="addressName" column="addressName"></property> 
<property name="codeNumber" column="codeNumber"></property> 
</class> 
</hibernate-mapping>
3、People.java和Address.java

public class People ...{ 
private long id; 
private String name; 
private Set addresses = new HashSet(); 
...
}

public class Address ...{ 
private long id; 
private People people; 
private String addressName; 
private String codeNumber; 
...
} 

4、数据库结构

people表:{peopleId,name}

address表:{addressId,peopleId,addressName,codeNumber}

5、测试代码

People people = new People(); 
people.setName("linda"); 
Address address = new Address(); 
address.setAddressName("yunnan"); 
address.setCodeNumber("564123"); 
address.setPeople(people); 
people.getAddresses().add(address); 
Session session = HibernateSessionFactory.getSession(); 
session.beginTransaction(); 
session.save(people); 
session.getTransaction().commit(); 

6、运行结果

  上面测试代码运行起来正确:

Hibernate: select max(peopleId) from people
Hibernate: select max(addressId) from address
Hibernate: insert into people (name, peopleId) values (?, ?)
Hibernate: insert into address (addressName, codeNumber, peopleId, addressId) values (?, ?, ?, ?)
Hibernate: update address set peopleId=? where addressId=?

  如果将People.hbm.xml映射改写一下:

<set name="addresses" cascade="save-update" inverse="true">
<key column="peopleId" not-null="true" />
<one-to-many class="com.persistent.Address"/>
</set>

  不同之处在于添加了inverse="true",结果:

Hibernate: select max(peopleId) from people
Hibernate: select max(addressId) from address
Hibernate: insert into people (name, peopleId) values (?, ?)
Hibernate: insert into address (addressName, codeNumber, addressId) values (?, ?, ?)

  可以看到,peopleId并没有写入到关联的address当中,数据库address表中相应记录的peopleId字段为空。

7、分析

  在Hibernate中,术语inverse是反转的意思,在关联关系中,inverse="false"为主控方,由主控方负责维护对象的关联关系。所以上面的映射文件改动之后,address为主控方,people为被控方,但是测试代码只进行了一个保存操作session.save(people),这是针对people的,因此无法正确级联保存address。而原来的映射文件中(虽然没有明确指明,Hibernate默认inverse="false"),people为主控方,因此保存people时它会保证关联的address的正确保存。

  也就是说,Hibernate仅仅按照主控方对象的状态的变化来同步更新数据库。按照原来的映射文件,people.getAddresses().add(address),即主控方对象的状态发生了改变,因此数据库会跟着对象状态的变化来同步更新数据库;而address.setPeople(people),即被控方对象的状态发生了改变,它是不能触发对象和数据库的同步更新的。