在域模型中,类与类之间除了关联关系和聚集关系,还可以存在继承关系,在下图所示的域模型中,Deparment类和Employee类之间为一对多的双向关联关系,Employee类有两个子类:Skiller类和Sales类。由于Java只允许一个类最多有一个直接的父类,因此Employee类、 Skiller类和Sales类构成了一棵继承关系树。
在面向对象的范畴中,还存在多态的概念,多态建立在继承关系的基础上。简单地理解,多态是指当一个Java应用变量被声明为Employee类时,这个变量实际上既可以引用Employee类自己的实例,Skiller类的实例,也可以引用Sales类的实例。Department类的getEmps()方法通过HibernateAPI从数据库中检索出所有Employee对象。getEmps()方法返回的集合既可以包含Employee类自己的实例,Skiller类的实例,也可以引用Sales类的实例。,这种查询被称为多态查询。数据库表之间并不存在继承关系,那么如何把域模型的继承关系映射到关系数据模型中呢?hibernate有以下三种映射方式:
继承关系树的根类对应一个表:对关系数据模型进行非常规设计,在数据库表中加入额外的区分子类型的字段。通过这种方式,可以使关系数据模型支持继承关系和多态。
继承关系树的每个类对应一个表(子类与父类通过外键关联):在关系数据模型中用外键参照关系来表示继承关系。
继承关系树的每个具体类对应一个表:关系数据模型完全不支持域模型中的继承关系和多态。
1.继承关系树的根类对应一个表employee(整个继承树一张表):
employee的表结构如下所示:
mysql> desc employee;
+------------+--------------+------+-----+---------+----------------+
| Field | Type | Null | Key | Default | Extra |
+------------+--------------+------+-----+---------+----------------+
| id | int(11) | NO | PRI | NULL | auto_increment |
| type | int(11) | NO | | NULL | |
| name | varchar(255) | YES | UNI | NULL | |
| depart_id | int(11) | YES | MUL | NULL | |
| skill | varchar(255) | YES | | NULL | |
| saleAmount | int(11) | YES | | NULL | |
+------------+--------------+------+-----+---------+----------------+
实体类Department和Employee请参看我前面的文章,Skiller和Sales分别如下所示:
- package com.reiyen.hibernate.domain;
- public class Skiller extends Employee {
- private String skill;
- //setter和getter方法
- }
- package com.reiyen.hibernate.domain;
- public class Sales extends Employee {
- private int saleAmount;
- //setter和getter方法
- }
Employee.hbm.xml映射文件如下:
- <?xml version="1.0"?>
- <!DOCTYPE hibernate-mapping PUBLIC
- "-//Hibernate/Hibernate Mapping DTD 3.0//EN"
- "http://hibernate.sourceforge.net/hibernate-mapping-3.0.dtd">
- <hibernate-mapping package="com.reiyen.hibernate.domain">
- <class name="Employee" discriminator-value="0">
- <id name="id">
- <generator class="native" />
- </id>
- <!--discriminator(鉴别器):缺省类型为string,这里指定为int类型 -->
- <discriminator column="type" type="int"></discriminator>
- <property name="name" unique="true"/>
- <!-- name="department" 这个名称必须与Employee中的属性名一致. 设置了column="depart_id",默认它会去department中找id与depart_id值相等的对象.如果要找name的值与depart_id相等的对象,则可以设置property-ref="name" -->
- <many-to-one name="department" column="depart_id" />
- <subclass name="Skiller" discriminator-value="1">
- <property name="skill" />
- </subclass>
- <subclass name="Sales" discriminator-value="2">
- <property name="saleAmount" />
- </subclass>
- </class>
- </hibernate-mapping>
测试类如下:
- public class Many2One {
- public static void main(String[] args) {
- add();
- query(1);//1
- }
- static void query(int empId) {
- Session s = null;
- Transaction tx = null;
- try {
- s = HibernateUtil.getSession();
- tx = s.beginTransaction();
- Employee emp = (Employee) s.get(Employee.class, empId);//2
- System.out.println(emp.getClass());
- tx.commit();
- } finally {
- if (s != null)
- s.close();
- }
- }
- static void add() {
- Session s = null;
- Transaction tx = null;
- try {
- Department depart = new Department();
- depart.setName("department name");
- Employee employee1 = new Employee();
- employee1.setDepartment(depart); //1 对象模型:建立两个对象的关联
- employee1.setName("employee1 name1");
- Skiller employee2 = new Skiller();
- employee2.setDepartment(depart); //2 对象模型:建立两个对象的关联
- employee2.setName("employee2 name2");
- employee2.setSkill("j2se");
- Sales employee3 = new Sales();
- employee3.setDepartment(depart); //2 对象模型:建立两个对象的关联
- employee3.setName("employee3 name3");
- employee3.setSaleAmount(1000);
- s = HibernateUtil.getSession();
- tx = s.beginTransaction();
- s.save(depart);
- s.save(employee1);
- s.save(employee2);
- s.save(employee3);
- tx.commit();
- } finally {
- if (s != null)
- s.close();
- }
- }
- }
程序运行后,控制台打印信息如下所示:
Hibernate: insert into Department (name) values (?)
Hibernate: insert into Employee (name, depart_id, type) values (?, ?, 0)
Hibernate: insert into Employee (name, depart_id, skill, type) values (?, ?, ?, 1)
Hibernate: insert into Employee (name, depart_id, saleAmount, type) values (?, ?, ?, 2)
Hibernate: select employee0_.id as id1_0_, employee0_.name as name1_0_,employee0_.depart_id as depart4_1_0_, employee0_.skill as skill1_0_,employee0_.saleAmount as saleAmount1_0_, employee0_.type as type1_0_from Employee employee0_ where employee0_.id=?
class com.reiyen.hibernate.domain.Employee
employee表中记录如下所示:
mysql> select * from employee;
+----+------+-----------------+-----------+-------+------------+
| id | type | name | depart_id | skill | saleAmount |
+----+------+-----------------+-----------+-------+------------+
| 1 | 0 | employee1 name1 | 1 | NULL | NULL |
| 2 | 1 | employee2 name2 | 1 | j2se | NULL |
| 3 | 2 | employee3 name3 | 1 | NULL | 1000 |
+----+------+-----------------+-----------+-------+------------+
3 rows in set (0.00 sec)
将测试代码中注释为1的语句改成:
- query(2);
再运行,控制台打印的class如下所示(因为hibernate支持多态查询): class com.reiyen.hibernate.domain.Skiller
打印的查询语句还是如上面所示的没有改变。
在上面修改的基础上,再将测试代码中注释为2的语句改成:
- Employee emp = (Employee) s.get(Skiller.class, empId);
再运行,则控制台打印的查询语句为:
Hibernate: select skiller0_.id as id1_0_,skiller0_.name as name1_0_, skiller0_.depart_id as depart4_1_0_,skiller0_.skill as skill1_0_ from Employee skiller0_ where skiller0_.id=? and skiller0_.type=1
class com.reiyen.hibernate.domain.Skiller
优点:操作效率高
缺点:如果说给employee增加子类的话,必须修改表结构,给表结构增加一个字段;同时表中对应子类的字段不能有非空约束.
2.继承关系树的每子类对应一个表(joined-subclass),表结构如下所示:
修改Employee.hbm.xml映射文件如下所示:
- <?xml version="1.0"?>
- <!DOCTYPE hibernate-mapping PUBLIC
- "-//Hibernate/Hibernate Mapping DTD 3.0//EN"
- "http://hibernate.sourceforge.net/hibernate-mapping-3.0.dtd">
- <hibernate-mapping package="com.reiyen.hibernate.domain">
- <class name="Employee">
- <id name="id">
- <generator class="native" />
- </id>
- <property name="name" unique="true"/>
- <!-- name="department" 这个名称必须与Employee中的属性名一致. 设置了column="depart_id",默认它会去department中找id与depart_id值相等的对象.如果要找name的值与depart_id相等的对象,则可以设置property-ref="name" -->
- <many-to-one name="department" column="depart_id" />
- <joined-subclass name="Skiller" table="skiller">
- <key column="employee_id" />
- <property name="skill" />
- </joined-subclass>
- <joined-subclass name="Sales" table="sales">
- <key column="employee_id" />
- <property name="saleAmount" column="sale_amount" />
- </joined-subclass>
- </class>
- </hibernate-mapping>
测试类不变,只是将测试代码中注释为1的语句改成:
- query(2);
则控制台打印的信息如下所示:
Hibernate: insert into Department (name) values (?)
Hibernate: insert into Employee (name, depart_id) values (?, ?)
Hibernate: insert into Employee (name, depart_id) values (?, ?)
Hibernate: insert into skiller (skill, employee_id) values (?, ?)
Hibernate: insert into Employee (name, depart_id) values (?, ?)
Hibernate: insert into sales (sale_amount, employee_id) values (?, ?)
Hibernate: select employee0_.id as id1_0_, employee0_.name as name1_0_, employee0_.depart_id as depart3_1_0_, employee0_1_.skill as skill2_0_, employee0_2_.sale_amount as sale2_3_0_, case when employee0_1_.employee_id is not null then 1 when employee0_2_.employee_id is not null then 2 when employee0_.id is not null then 0 end as clazz_0_ from Employee employee0_ left outer join skiller employee0_1_ on employee0_.id=employee0_1_.employee_id left outer join sales employee0_2_ on employee0_.id=employee0_2_.employee_id where employee0_.id=?
class com.reiyen.hibernate.domain.Skiller
从打印的SQL语句可以看出,此时,如果保存的是Employee对象的子类的实例的话,则要在两张表中保存记录;如果查询的是子类对象的话,是三张表关联在一起进行查询。
在上面修改的基础上,再将测试代码中注释为2的语句改成:
- Employee emp = (Employee) s.get(Skiller.class, empId);
再运行,则控制台打印的查询语句为:
Hibernate: select skiller0_.employee_id as id1_0_, skiller0_1_.name as name1_0_, skiller0_1_.depart_id as depart3_1_0_, skiller0_.skill as skill2_0_ from skiller skiller0_ inner join Employee skiller0_1_ on skiller0_.employee_id=skiller0_1_.id where skiller0_.employee_id=?
此时只关联两张表查询。
数据库中表记录如下所示:
mysql> select * from employee;
+----+-----------------+-----------+
| id | name | depart_id |
+----+-----------------+-----------+
| 1 | employee1 name1 | 1 |
| 2 | employee2 name2 | 1 |
| 3 | employee3 name3 | 1 |
+----+-----------------+-----------+
3 rows in set (0.00 sec)
mysql> select * from skiller;
+-------------+-------+
| employee_id | skill |
+-------------+-------+
| 2 | j2se |
+-------------+-------+
1 row in set (0.00 sec)
mysql> select * from sales;
+-------------+-------------+
| employee_id | sale_amount |
+-------------+-------------+
| 3 | 1000 |
+-------------+-------------+
1 row in set (0.00 sec)
3.混合使用,假设如果Sales的属性很多,而Skiller的属性很少,这时可以混使用“一个类继承体系一张表”和“每个子类一张表”,表结构如下所示:
Employee.hbm.xml映射文件如下所示:
- <?xml version="1.0"?>
- <!DOCTYPE hibernate-mapping PUBLIC
- "-//Hibernate/Hibernate Mapping DTD 3.0//EN"
- "http://hibernate.sourceforge.net/hibernate-mapping-3.0.dtd">
- <hibernate-mapping package="com.reiyen.hibernate.domain">
- <class name="Employee" discriminator-value="0">
- <id name="id">
- <generator class="native" />
- </id>
- <discriminator column="type" type="int" />
- <property name="name" unique="true" />
- <!-- name="department" 这个名称必须与Employee中的属性名一致. 设置了column="depart_id",默认它会去department中找id与depart_id值相等的对象.如果要找name的值与depart_id相等的对象,则可以设置property-ref="name" -->
- <many-to-one name="department" column="depart_id" />
- <!--如果discriminator-value没有显式的给定值的话,则与name属性的值保持一致,即为Skiller -->
- <subclass name="Skiller" discriminator-value="1">
- <property name="skill" />
- </subclass>
- <subclass name="Sales" discriminator-value="2">
- <join table="sales">
- <key column="employee_id" />
- <property name="saleAmount" column="sale_amount" />
- </join>
- </subclass>
- </class>
- </hibernate-mapping>
此时测试类不变,只是将测试代码中注释为1的语句改成:
- query(2);
然后在上面的基础上运行原程序,则控制台会打印出如下异常信息:
Hibernate: insert into Department (name) values (?)
Hibernate: insert into Employee (name, depart_id, type) values (?, ?, 0)
Exception in thread "main" org.hibernate.exception.SQLGrammarException: could not insert: [com.reiyen.hibernate.domain.Employee]
Caused by: com.mysql.jdbc.exceptions.MySQLSyntaxErrorException: Unknown column 'type' in 'field list'
这是因为我在hibernate.cfg.xml配置文件中配置了此项:
- <property name="hbm2ddl.auto">create</property>
所以在此次程序运行时,会删除数据库中的employee表,sales表,而employee表中有skiller表的外键关联,所以不能删除employee数据表了,所以抛出了上面的异常。此时你再查看数据库表,如下所示 :
mysql> select * from employee;
+----+-----------------+-----------+
| id | name | depart_id |
+----+-----------------+-----------+
| 1 | employee1 name1 | 1 |
| 2 | employee2 name2 | 1 |
| 3 | employee3 name3 | 1 |
+----+-----------------+-----------+
3 rows in set (0.00 sec)
mysql> select * from skiller;
+-------------+-------+
| employee_id | skill |
+-------------+-------+
| 2 | j2se |
+-------------+-------+
1 row in set (0.00 sec)
mysql> select * from sales;
Empty set (0.00 sec)
所以得先手动删除skiller数据表,然后再来运行程序:
如果Employee.hbm.xml配置文件中
- <subclass name="Skiller" >
不配置discriminator-value="1",则会抛出如下异常:
java.lang.ExceptionInInitializerError
Caused by: org.hibernate.MappingException: Could not format discriminator value to SQL string
因为如果discriminator-value没有显式的给定值的话,则与name属性的值保持一致,即为Skiller ,所以会抛出如上异常!
4.继承关系树的每个具体类对应一个表(union-subclass)
表结构如下所示:
Employee.hbm.xml映射文件如下:
- <?xml version="1.0"?>
- <!DOCTYPE hibernate-mapping PUBLIC
- "-//Hibernate/Hibernate Mapping DTD 3.0//EN"
- "http://hibernate.sourceforge.net/hibernate-mapping-3.0.dtd">
- <hibernate-mapping package="com.reiyen.hibernate.domain">
- <class name="Employee">
- <id name="id">
- <generator class="hilo" />
- </id>
- <property name="name" unique="true" />
- <many-to-one name="department" column="depart_id" />
- <union-subclass name="Skiller" table="skiller">
- <property name="skill" />
- </union-subclass>
- <union-subclass name="Sales" table="sales">
- <property name="saleAmount" column="sale_amount" />
- </union-subclass>
- </class>
- </hibernate-mapping>
此时主键增长不能再是:
- <generator class="native" />
因为如果使用native的话三张表会产生相同的id值,这样当根据id查询Employee时就会出错了。所以如果你配置成native时会抛出如下异常(因为Employee实体类中id对应 的是int了,所以在此使用hilo(高低位)主键生成方式):
org.hibernate.MappingException: Cannot use identity column key generation with <union-subclass> mapping for: com.reiyen.hibernate.domain.Skiller
运行测试程序后,此时控制台打印信息如下所示:
Hibernate: insert into Department (name) values (?)
Hibernate: insert into Employee (name, depart_id, id) values (?, ?, ?)
Hibernate: insert into skiller (name, depart_id, skill, id) values (?, ?, ?, ?)
Hibernate: insert into sales (name, depart_id, sale_amount, id) values (?, ?, ?, ?)
Hibernate: select employee0_.id as id1_0_, employee0_.name as name1_0_, employee0_.depart_id as depart3_1_0_, employee0_.skill as skill2_0_, employee0_.sale_amount as sale1_3_0_, employee0_.clazz_ as clazz_0_ from ( select id, null as sale_amount, depart_id, null as skill, name, 0 as clazz_ from Employee union select id, null as sale_amount, depart_id, skill, name, 1 as clazz_ from skiller union select id, sale_amount, depart_id, null as skill, name, 2 as clazz_ from sales ) employee0_ where employee0_.id=?
class com.reiyen.hibernate.domain.Skiller
执行查询时,首先使用子查询,在子查询中使用union将三张表的结果全成一张表,然后再在合成的结果集中进行查询。
如果Employee是一个抽象类,你不想在数据表中对应相应的数据表,则可以设置abstract="true".如下所示:
- <class name="Employee" abstract="true" >
此外,如果继承关系中有接口,可以把它当作抽象类对待。
三种映射方式的比较和选择
为了方便说明为三种方式按顺序标号为[1]整个继承树一张表;[2]每子类对应一个表(joined-subclass);[4]每个具体类对应一个表(union-subclass)。
1、复杂度:
[1]简单;
[2]表较多且之间有外键约束;
[4]包含重复字段;
2、查询性能:
[1]效率高;
[2]需要表内连接或左外连接;
[4]若查询父类需查所有子类表;
3、可维护性:
[1]只需修改一个表;
[2]若某个类属性变化只修改这个类对应的表;
[4]若父类属性变化需要修改所有子类对应的表;
综上,选择时,可以参考以下原则:
1、子类属性不是非常多时,优先考虑[1],因为其性能最佳。
2、子类属性非常多,且对性能要求不是很严格时,优先考虑[2]