Spring 中 @Primary 注解的原理是什么?

1. 问题分析

当我们使用 Spring 的时候,有时候会遇到下面这种情况。

假设我有 A、B 两个类,在 A 中注入 B,如下:

1
2
3
4
5
@Component
public class A {
@Autowired
B b;
}

至于 B,则在配置类中存在多个实例:

1
2
3
4
5
6
7
8
9
10
11
12
13
@Configuration
@ComponentScan
public class JavaConfig {
@Bean("b1")
B b1() {
return new B();
}

@Bean("b2")
B b2() {
return new B();
}
}

这样的项目启动之后,必然会抛出如下异常:

当然,对于这样的问题,相信有经验的同学都知道该怎么解决:

  1. 可以使用 @Resource 注解,使用该注解时指定具体的 Bean 名称即可。
  2. 在 @Autowired 注解之上,再多加一个 @Qualifier(“b1”) 注解,通过该注解去指定要加载的 Bean 名称。
1
2
3
4
5
6
@Component
public class A {
@Autowired
@Qualifier("b1")
B b;
}
  1. 在多个 B 对象的某一个之上,添加 @Primary 注解,表示当存在重复的 B 对象时,优先使用哪一个。
1
2
3
4
5
6
7
8
9
10
11
12
13
14
@Configuration
@ComponentScan
public class JavaConfig {
@Bean("b1")
@Primary
B b1() {
return new B();
}

@Bean("b2")
B b2() {
return new B();
}
}

除了这三个,还有没有其他办法呢?必须有!!!在 Spring 中 @Qualifier 注解还能这么用? 一文中,松哥还和大家扩展了 @Qualifier 注解的其他用法,感兴趣的小伙伴不要错过哦。

这里三个方法,其中 @Resource 是 JSR 中提供的注解,我这里先不展开,松哥后面专门再来和大家聊 @Resource 注解的注入原理。今天我主要是想和小伙伴们分享一下后面两种方案的实现原理。

2. 源码解析

本文基于前面@Autowired 到底是怎么把变量注入进来的?一文展开,所以如果还没看过改文章的小伙伴,建议先去阅读一下,这有助于更好的理解本文。

2.1 doResolveDependency

@Autowired 到底是怎么把变量注入进来的?的 3.3 小节中,我们提到,给 A 注入 B 的时候,会调用到 doResolveDependency 方法,我们再来看下该方法:

DefaultListableBeanFactory#doResolveDependency:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
@Nullable
public Object doResolveDependency(DependencyDescriptor descriptor, @Nullable String beanName,
@Nullable Set<String> autowiredBeanNames, @Nullable TypeConverter typeConverter) throws BeansException {
//...
Map<String, Object> matchingBeans = findAutowireCandidates(beanName, type, descriptor);
if (matchingBeans.isEmpty()) {
if (isRequired(descriptor)) {
raiseNoMatchingBeanFound(type, descriptor.getResolvableType(), descriptor);
}
return null;
}
if (matchingBeans.size() > 1) {
autowiredBeanName = determineAutowireCandidate(matchingBeans, descriptor);
if (autowiredBeanName == null) {
if (isRequired(descriptor) || !indicatesMultipleBeans(type)) {
return descriptor.resolveNotUnique(descriptor.getResolvableType(), matchingBeans);
}
else {
// In case of an optional Collection/Map, silently ignore a non-unique case:
// possibly it was meant to be an empty collection of multiple regular beans
// (before 4.3 in particular when we didn't even look for collection beans).
return null;
}
}
instanceCandidate = matchingBeans.get(autowiredBeanName);
//...
}

在这个方法中,首先调用了 findAutowireCandidates 方法去找到所有满足条件的 Class。Map 中的 key 就是 Bean 的名称,value 则是一个 Class,此时还没有实例化。

如果我们是通过 @Qualifier 注解来解决问题的,那么问题就在 findAutowireCandidates 方法中被解决了。这个在前面的文章 Spring 中 @Qualifier 注解还能这么用? 中已经和小伙伴们聊过了。

如果 @Qualifier 注解没把问题解决掉,就会导致最终查询到的 matchingBeans 的数量大于 1,那么就会进入到接下来的 if 环节中,通过 determineAutowireCandidate 方法进一步确定到底使用哪一个 Bean,@Primary 注解的处理,就在该方法中完成。

2.2 determineAutowireCandidate

DefaultListableBeanFactory#determineAutowireCandidate

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
@Nullable
protected String determineAutowireCandidate(Map<String, Object> candidates, DependencyDescriptor descriptor) {
Class<?> requiredType = descriptor.getDependencyType();
String primaryCandidate = determinePrimaryCandidate(candidates, requiredType);
if (primaryCandidate != null) {
return primaryCandidate;
}
String priorityCandidate = determineHighestPriorityCandidate(candidates, requiredType);
if (priorityCandidate != null) {
return priorityCandidate;
}
// Fallback
for (Map.Entry<String, Object> entry : candidates.entrySet()) {
String candidateName = entry.getKey();
Object beanInstance = entry.getValue();
if ((beanInstance != null && this.resolvableDependencies.containsValue(beanInstance)) ||
matchesBeanName(candidateName, descriptor.getDependencyName())) {
return candidateName;
}
}
return null;
}

这个方法里边一共做了三种尝试:

  1. 第一个尝试就是调用 determinePrimaryCandidate 方法去确定最佳候选 Bean,这个方法本质上就是通过 @Primary 注解找到最佳 BeanName。
  2. 如果第一步没有找到最佳 BeanName,那么接下来会调用 determineHighestPriorityCandidate 方法去查找最佳 Bean,该方法本质上是通过查找 JSR-330 中的 @Priority 注解,来确定 Bean 的优先级。
  3. 如果前两步都没找到合适的 BeanName,那么接下来这个 for 循环则是通过 Bean 的名称进行匹配了,即 A 类中变量的名称和目标 Bean 的名称是否匹配,如果能匹配上,那也可以。这也就是我么常说的 @Autowired 注解先按照类型去匹配,如果类型匹配不上,就会按照名称去匹配。

上面大致介绍了这个方法的执行思路,接下来我们就来看一下执行细节。

2.2.1 determinePrimaryCandidate

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
@Nullable
protected String determinePrimaryCandidate(Map<String, Object> candidates, Class<?> requiredType) {
String primaryBeanName = null;
for (Map.Entry<String, Object> entry : candidates.entrySet()) {
String candidateBeanName = entry.getKey();
Object beanInstance = entry.getValue();
if (isPrimary(candidateBeanName, beanInstance)) {
if (primaryBeanName != null) {
boolean candidateLocal = containsBeanDefinition(candidateBeanName);
boolean primaryLocal = containsBeanDefinition(primaryBeanName);
if (candidateLocal && primaryLocal) {
throw new NoUniqueBeanDefinitionException(requiredType, candidates.size(),
"more than one 'primary' bean found among candidates: " + candidates.keySet());
}
else if (candidateLocal) {
primaryBeanName = candidateBeanName;
}
}
else {
primaryBeanName = candidateBeanName;
}
}
}
return primaryBeanName;
}
protected boolean isPrimary(String beanName, Object beanInstance) {
String transformedBeanName = transformedBeanName(beanName);
if (containsBeanDefinition(transformedBeanName)) {
return getMergedLocalBeanDefinition(transformedBeanName).isPrimary();
}
return (getParentBeanFactory() instanceof DefaultListableBeanFactory parent &&
parent.isPrimary(transformedBeanName, beanInstance));
}

我们来看下这个方法的执行逻辑。

参数 candidates 中保存了所有符合条件的 BeanDefinition,参数 key 就是 Bean 的名称,Value 则是对应的 BeanDefinition。现在就去遍历 candidates,在遍历的时候,调用 isPrimary 方法去判断这个 BeanDefinition 上是否含有 @Primary 注解,isPrimary 方法的逻辑比较简单,我就不啰嗦了,该方法中涉及到 getMergedLocalBeanDefinition 方法去父容器中查找两个细节,这个松哥在之前的文章中也都和大家聊过了(Spring BeanDefinition:父子关系解密Spring 中的父子容器是咋回事?)。

在查找的过程中,如果有满足条件的 BeanName,则赋值给 primaryBeanName 变量然后返回,如果存在多个满足条件的 BeanName,那就抛出 NoUniqueBeanDefinitionException 异常。

2.2.2 determineHighestPriorityCandidate

要理解 determineHighestPriorityCandidate 方法,得先了解 @Priority 注解的用法。考虑到有的小伙伴可能还不熟悉 @Priority 注解,我这里也跟大家稍微说两句。

@Priority 注解作用有点类似于 @Order,可以用来指定一个 Bean 的优先级,这是 JSR 中提供的注解,所以如果想使用这个注解,需要先添加依赖:

1
2
3
4
5
<dependency>
<groupId>jakarta.annotation</groupId>
<artifactId>jakarta.annotation-api</artifactId>
<version>2.1.1</version>
</dependency>

然后在类上添加该注解,像下面这样:

1
2
3
4
5
6
7
8
9
10
public interface IBService {
}
@Component
@Priority(100)
public class BServiceImpl1 implements IBService{
}
@Component
@Priority(101)
public class BServiceImpl2 implements IBService{
}

@Priority 注解中的数字表示优先级,数字越大优先级越小。将来在 A 中注入 IBService 时,就会优先查找优先级高的 Bean。虽然 @Priority 注解可以加在类上,也可以加在方法上,但是在具体实践中,加在方法上这个注解并不会生效,只能加在类上面。至于原因,大家看完接下来的源码分析就懂了。

现在我们再来看下 determineHighestPriorityCandidate 方法:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
@Nullable
protected String determineHighestPriorityCandidate(Map<String, Object> candidates, Class<?> requiredType) {
String highestPriorityBeanName = null;
Integer highestPriority = null;
for (Map.Entry<String, Object> entry : candidates.entrySet()) {
String candidateBeanName = entry.getKey();
Object beanInstance = entry.getValue();
if (beanInstance != null) {
Integer candidatePriority = getPriority(beanInstance);
if (candidatePriority != null) {
if (highestPriorityBeanName != null) {
if (candidatePriority.equals(highestPriority)) {
throw new NoUniqueBeanDefinitionException(requiredType, candidates.size(),
"Multiple beans found with the same priority ('" + highestPriority +
"') among candidates: " + candidates.keySet());
}
else if (candidatePriority < highestPriority) {
highestPriorityBeanName = candidateBeanName;
highestPriority = candidatePriority;
}
}
else {
highestPriorityBeanName = candidateBeanName;
highestPriority = candidatePriority;
}
}
}
}
return highestPriorityBeanName;
}

determineHighestPriorityCandidate 方法的整体处理思路跟 determinePrimaryCandidate 方法特别像,不同的是 determinePrimaryCandidate 方法处理的是 @Primary 注解,而 determineHighestPriorityCandidate 方法处理的是 @Priority 注解。

determineHighestPriorityCandidate 方法也是遍历 candidates,然后调用 getPriority 方法获取到具体的优先级的值。然后根据这个具体的数字选定一个合适的 beanName 返回,如果存在多个优先级相同的 bean,那么就会抛出 NoUniqueBeanDefinitionException 异常。

最后再来看下 getPriority 方法,几经辗转之后,该方法会调用到 AnnotationAwareOrderComparator#getPriority 方法:

1
2
3
4
5
6
7
8
9
10
11
12
@Override
@Nullable
public Integer getPriority(Object obj) {
if (obj instanceof Class<?> clazz) {
return OrderUtils.getPriority(clazz);
}
Integer priority = OrderUtils.getPriority(obj.getClass());
if (priority == null && obj instanceof DecoratingProxy decoratingProxy) {
return getPriority(decoratingProxy.getDecoratedClass());
}
return priority;
}

可以看到,这里最终就是调用 OrderUtils.getPriority 方法去查找参数 clazz 上的 @Priority 注解,并找到注解上对应的值返回。OrderUtils.getPriority 在执行的时候,参数时 clazz,即只会查找 clazz 上的 注解,并不会查找方法上的注解,因此前面我说 @Priority 注解要加在类上才有效。

2.2.3 按名称匹配

最后我们再来看下按照名字去匹配的逻辑:

1
2
3
4
5
6
7
8
9
10
11
12
13
// Fallback
for (Map.Entry<String, Object> entry : candidates.entrySet()) {
String candidateName = entry.getKey();
Object beanInstance = entry.getValue();
if ((beanInstance != null && this.resolvableDependencies.containsValue(beanInstance)) ||
matchesBeanName(candidateName, descriptor.getDependencyName())) {
return candidateName;
}
}
protected boolean matchesBeanName(String beanName, @Nullable String candidateName) {
return (candidateName != null &&
(candidateName.equals(beanName) || ObjectUtils.containsElement(getAliases(beanName), candidateName)));
}

可以看到,这里也是遍历 candidates 集合,然后调用 matchesBeanName 方法,在该方法中,会去判断候选的 BeanName 和需要注入的变量名(descriptor.getDependencyName())是否相等,如果相等,就直接返回即可。即下面这种代码不需要额外的注解是可以运行不会报错的:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
@Component
public class AService {
@Autowired
B b1;

}
@Configuration
@ComponentScan
public class JavaConfig {

@Bean
public B b1() {
return new B();
}

@Bean
B b2() {
return new B();
}

}

3. 小结

好啦,经过上面的分析,现在小伙伴们明白了 @Primary 注解的完整处理逻辑了吧~本文结合@Autowired 到底是怎么把变量注入进来的?Spring 中 @Qualifier 注解还能这么用? 一起食用效果更好哦!

# Spring

喜欢这篇文章吗?扫码关注公众号【江南一点雨】【江南一点雨】专注于 SPRING BOOT+微服务以及前后端分离技术,每天推送原创技术干货,关注后回复 JAVA,领取松哥为你精心准备的 JAVA 干货!

评论

Your browser is out-of-date!

Update your browser to view this website correctly. Update my browser now

×