【Hibernate 8】Hibernate的调优方法:抓取策略

时间:2022-01-24 11:29:38

在上一篇博客中,介绍了Hibernate的缓存机制。合理的配置缓存,可以极大程度上优化Hibernate的性能。这篇博客,介绍另外一个调优方式:抓取策略。

一、什么是抓取策略

抓取策略(fetching strategy):当应用程序需要在关联关系间进行导航的时候,Hibernate如何获取关联对象的策略。抓取策略可以在O / R映射的元数据中声明,也可以在特定的HQL或条件查询(Criteria Query)中重载声明。

二、Hibernate的抓取策略

在Hibernate中,主要包括四种抓取策略:

2.1,连接抓取(Join fetching)

Hibernate通过在select语句中使用join等关键字,来获得对象的关联实例或者关联集合

2.2,查询抓取(Select fetching)

Hibernate通过发送另外一条select语句抓取当前对象的关联实体或集合。除非是显示的指定lazy=“false”禁止延迟抓取,否则只有当你真正访问关联关系的时候,才会执行第二条select语句。

2.3,子查询抓取(Subselect fetching)

Hibernate通过另外发送一条SELECT 语句抓取在前面查询到(或者抓取到)的所有实体对象的关联集合。除非你显式的指定lazy="false" 禁止延迟抓取(lazy fetching),否则只有当你真正访问关联关系的时候,才会执行第二条select语句。

2.4,批量抓取(Batch fetching)

是Hibernate对查询抓取的优化方案, 通过指定一个主键或外键列表,Hibernate使用单条SELECT语句获取一批对象实例或集合。

二、详解Hibernate的抓取策略

首先,在Hibernate.cfg.xml中配置抓取策略的数据量(非必须,比如批量更新,可以在用到批量更新的时候配置):

<span style="font-family:KaiTi_GB2312;font-size:18px;"> <!-- 批量更新配置 -->
<property name="hibernate.jdbc.batch_size">30</property> <!-- 批量抓取参数的配置 -->
<property name="hibernate.jdbc.fetch_size">50</property></span>

2.1,连接抓取

<span style="font-family:KaiTi_GB2312;font-size:18px;"><many-to-one name="classes" column="classesid" fetch="join" not-null="true"/></span>

fetch="join",hibernate会通过select语句使用外连接来加载其关联实体或集合,不过,此时的lazy失效

测试方法代码如查询抓取中的testFetch1()。

控制台打印:

Hibernate:
select student0_.id as id0_0_, classes1_.id as id1_1_, student0_.name as name0_0_, student0_.classesid as classesid0_0_, classes1_.name as name1_1_
from t_student student0_ inner join t_classes classes1_ on student0_.classesid=classes1_.id
where student0_.id=1

student.name=班级0的学生0

classes.name=班级0

2.2,查询抓取

<span style="font-family:KaiTi_GB2312;font-size:18px;"><many-to-one name="classes" column="classesid" fetch="select"/></span>

fetch="select",另外发送一条select语句抓取当前对象关联实体或集合

<span style="font-family:KaiTi_GB2312;font-size:18px;">public void testFetch1() {
Session session = null;
try {
session = HibernateUtils.getSession();
session.beginTransaction(); Student student = (Student)session.load(Student.class, 1);
System.out.println("student.name=" + student.getName()); Classes classes = student.getClasses();
System.out.println("classes.name=" + classes.getName()); session.getTransaction().commit();
}catch(Exception e) {
e.printStackTrace();
session.getTransaction().rollback();
}finally {
HibernateUtils.closeSession(session);
}
}</span>

控制台打印:

Hibernate:
select student0_.id as id0_0_, student0_.name as name0_0_, student0_.classesid as classesid0_0_ from t_student student0_
where student0_.id=?

student.name=班级0的学生0

Hibernate: select classes0_.id as id1_0_, classes0_.name as name1_0_
from t_classes classes0_
where classes0_.id=?

classes.name=班级0

默认的启用了延迟加载,所以,仅仅当执行Classes classes=student.getClasses()时,才发出了另一条select语句。

2.3,子查询抓取

<span style="font-family:KaiTi_GB2312;font-size:18px;"><set name="students" order-by="id" inverse="true" cascade="all" fetch="subselect">
<key column="classesid"/>
<one-to-many class="com.bjpowernode.hibernate.Student"/>
</set></span>

fetch="subselect",另外发送一条select语句抓取在前面查询到的所有实体对象的关联集合

<span style="font-family:KaiTi_GB2312;font-size:18px;">public void testFetch1() {
Session session = null;
try {
session = HibernateUtils.getSession();
session.beginTransaction(); List classesList = session.createQuery("select c from Classes c where c.id in(1, 2, 3)").list();
for (Iterator iter1 = classesList.iterator(); iter1.hasNext();) {
Classes classes = (Classes) iter1.next();
System.out.println(classes.getName());
for (Iterator iter = classes.getStudents().iterator(); iter.hasNext();) {
Student student = (Student) iter.next();
System.out.println(student.getName());
}
}
session.getTransaction().commit();
} catch (Exception e) {
e.printStackTrace();
session.getTransaction().rollback();
} finally {
HibernateUtils.closeSession(session);
}
}</span>

控制台打印:

Hibernate: select classes0_.id as id1_, classes0_.name as name1_ from t_classes classes0_ where classes0_.id in (1 , 2 , 3)

班级0

Hibernate: select students0_.classesid as classesid1_, students0_.id as id1_, students0_.id as id0_0_, students0_.name as name0_0_, students0_.classesid as classesid0_0_
from t_student students0_
where students0_.classesid in (select classes0_.id from t_classes classes0_
where classes0_.id in (1 , 2 , 3)) order by students0_.id

班级0的学生0

班级0的学生1

班级0的学生2

班级0的学生3

班级0的学生4

班级0的学生5

班级0的学生6

班级0的学生7

班级0的学生8

班级0的学生9

班级1

班级1的学生0

班级1的学生1

班级1的学生2

班级1的学生3

班级1的学生4

班级1的学生5

班级1的学生6

班级1的学生7

班级1的学生8

班级1的学生9

班级2

班级2的学生0

班级2的学生1

班级2的学生2

班级2的学生3

班级2的学生4

班级2的学生5

班级2的学生6

班级2的学生7

班级2的学生8

班级2的学生9

2.4,批量抓取

单端代理的批量抓取:Hibernate可以充分有效的使用批量抓取,也就是说,如果仅一个访问代理(或集合),那么Hibernate将不载入其他未实例化的代理。 批量抓取是延迟查询抓取的优化方案,你可以在两种批量抓取方案之间进行选择:在类级别和集合级别。

如果:你在一个Session中载入了25个 Cat实例,每个Cat实例都拥有一个引用成员owner, 其指向Person,而Person类是代理,同时lazy="true"。 如果你必须遍历整个cats集合,对每个元素调用getOwner()方法,Hibernate将会默认的执行25次SELECT查询, 得到其owner的代理对象。这时,你可以通过在映射文件的Person属性,显式声明batch-size,改变其行为: 

<class name="Person" batch-size="10">...</class>随之,Hibernate将只需要执行三次查询,分别为10、10、 5。

实例分析:

<span style="font-family:KaiTi_GB2312;font-size:18px;"><class name="com.angel.hibernate.Classes" table="t_classes" batch-size="10"></span>

测试类:

<span style="font-family:KaiTi_GB2312;font-size:18px;">public void testFetch1() {
Session session = null;
try {
session = HibernateUtils.getSession();
session.beginTransaction();
List students = session
.createQuery("select s from Student s where s.id in(:ids)")
.setParameterList("ids",new Object[] { 1, 11, 21, 31, 41, 51, 61, 71, 81,91 }).list();
for (Iterator iter = students.iterator(); iter.hasNext();) {
Student student = (Student) iter.next();
System.out.println(student.getName());
System.out.println(student.getClasses().getName());
}
session.getTransaction().commit();
} catch (Exception e) {
e.printStackTrace();
session.getTransaction().rollback();
} finally {
HibernateUtils.closeSession(session);
}
}</span>

控制台输出结果:

Hibernate: select student0_.id as id0_, student0_.name as name0_, student0_.classesid as classesid0_ from t_student student0_ where student0_.id in (? , ? , ? , ? , ? , ? , ? , ? , ? , ?)

班级0的学生0

Hibernate: select classes0_.id as id1_0_, classes0_.name as name1_0_ from t_classes classes0_ where classes0_.id in (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)

班级0

班级1的学生0

班级1

班级2的学生0

班级2

班级3的学生0

班级3

班级4的学生0

班级4

班级5的学生0

班级5

班级6的学生0

班级6

班级7的学生0

班级7

班级8的学生0

班级8

班级9的学生0

班级9

如果不配置批量抓取batch-size="10",那么控制台打印的结果为(出现了N+1问题,所以,配置批量抓取,也可以解决N+1问题):

Hibernate: select student0_.id as id0_, student0_.name as name0_, student0_.classesid as classesid0_ from t_student student0_ where student0_.id in (? , ? , ? , ? , ? , ? , ? , ? , ? , ?)

班级0的学生0

Hibernate: select classes0_.id as id1_0_, classes0_.name as name1_0_ from t_classes classes0_ where classes0_.id=?

班级0

班级1的学生0

Hibernate: select classes0_.id as id1_0_, classes0_.name as name1_0_ from t_classes classes0_ where classes0_.id=?

班级1

班级2的学生0

Hibernate: select classes0_.id as id1_0_, classes0_.name as name1_0_ from t_classes classes0_ where classes0_.id=?

班级2

班级3的学生0

Hibernate: select classes0_.id as id1_0_, classes0_.name as name1_0_ from t_classes classes0_ where classes0_.id=?

班级3

班级4的学生0

Hibernate: select classes0_.id as id1_0_, classes0_.name as name1_0_ from t_classes classes0_ where classes0_.id=?

班级4

班级5的学生0

Hibernate: select classes0_.id as id1_0_, classes0_.name as name1_0_ from t_classes classes0_ where classes0_.id=?

班级5

班级6的学生0

Hibernate: select classes0_.id as id1_0_, classes0_.name as name1_0_ from t_classes classes0_ where classes0_.id=?

班级6

班级7的学生0

Hibernate: select classes0_.id as id1_0_, classes0_.name as name1_0_ from t_classes classes0_ where classes0_.id=?

班级7

班级8的学生0

Hibernate: select classes0_.id as id1_0_, classes0_.name as name1_0_ from t_classes classes0_ where classes0_.id=?

班级8

班级9的学生0

Hibernate: select classes0_.id as id1_0_, classes0_.name as name1_0_ from t_classes classes0_ where classes0_.id=?

班级9

三、总结

抓取策略和缓存机制一样,都是Hibernate调优的重要手段:我们通过Hibernate的抓取策略,减少发出sql的次数,将其数据存入缓存,再次减少开合数据的次数。但并不是每一个应用程序都必须用上这些东西。具体的使用,还需要根据业务的需求,酌情而定。下面一篇博客,将介绍锁的概念:悲观锁和乐观锁