Spring Core源码分析
基本
本部分从最基本的Spring开始。配置文件:
<?xml version="1.0" encoding="UTF-8"?>
<beans>
<bean class="base.SimpleBean"></bean>
</beans>
启动代码:
public static void main(String[] args) {
ClassPathXmlApplicationContext context = new ClassPathXmlApplicationContext("config.xml");
SimpleBean bean = context.getBean(SimpleBean.class);
bean.send();
context.close();
}
SimpleBean:
public class SimpleBean {
public void send() {
System.out.println("I am send method from SimpleBean!");
}
}
ClassPathXmlApplicationContext
整个继承体系如下:
ResourceLoader代表了加载资源的一种方式,正是策略模式的实现。
构造器源码:
public ClassPathXmlApplicationContext(String[] configLocations, boolean refresh, ApplicationContext parent) {
//null
super(parent);
setConfigLocations(configLocations);
//默认true
if (refresh) {
refresh();
}
}
构造器
首先看父类构造器,沿着继承体系一直向上调用,直到AbstractApplicationContext:
public AbstractApplicationContext(ApplicationContext parent) {
this();
setParent(parent);
}
public AbstractApplicationContext() {
this.resourcePatternResolver = getResourcePatternResolver();
}
getResourcePatternResolver:
protected ResourcePatternResolver getResourcePatternResolver() {
return new PathMatchingResourcePatternResolver(this);
}
PathMatchingResourcePatternResolver支持Ant风格的路径解析。
设置配置文件路径
即AbstractRefreshableConfigApplicationContext.setConfigLocations:
public void setConfigLocations(String... locations) {
if (locations != null) {
Assert.noNullElements(locations, "Config locations must not be null");
this.configLocations = new String[locations.length];
for (int i = 0; i < locations.length; i++) {
this.configLocations[i] = resolvePath(locations[i]).trim();
}
} else {
this.configLocations = null;
}
}
resolvePath:
protected String resolvePath(String path) {
return getEnvironment().resolveRequiredPlaceholders(path);
}
此方法的目的在于将占位符(placeholder)解析成实际的地址。比如可以这么写: new ClassPathXmlApplicationContext("classpath:config.xml");
那么classpath:就是需要被解析的。
getEnvironment方法来自于ConfigurableApplicationContext接口,源码很简单,如果为空就调用createEnvironment创建一个。AbstractApplicationContext.createEnvironment:
protected ConfigurableEnvironment createEnvironment() {
return new StandardEnvironment();
}
Environment接口
继承体系:
Environmen接口代表了当前应用所处的环境。从此接口的方法可以看出,其主要和profile、Property相关。
Profile
Spring Profile特性是从3.1开始的,其主要是为了解决这样一种问题: 线上环境和测试环境使用不同的配置或是数据库或是其它。有了Profile便可以在 不同环境之间无缝切换。Spring容器管理的所有bean都是和一个profile绑定在一起的。使用了Profile的配置文件示例:
<beans profile="develop">
<context:property-placeholder location="classpath*:jdbc-develop.properties"/>
</beans>
<beans profile="production">
<context:property-placeholder location="classpath*:jdbc-production.properties"/>
</beans>
<beans profile="test">
<context:property-placeholder location="classpath*:jdbc-test.properties"/>
</beans>
在启动代码中可以用如下代码设置活跃(当前使用的)Profile:
context.getEnvironment().setActiveProfiles("dev");
当然使用的方式还有很多(比如注解),参考:
Property
这里的Property指的是程序运行时的一些参数,引用注释:
properties files, JVM system properties, system environment variables, JNDI, servlet context parameters, ad-hoc Properties objects,Maps, and so on.
Environment构造器
private final MutablePropertySources propertySources = new MutablePropertySources(this.logger);
public AbstractEnvironment() {
customizePropertySources(this.propertySources);
}
PropertySources接口
继承体系:
此接口实际上是PropertySource的容器,默认的MutablePropertySources实现内部含有一个CopyOnWriteArrayList作为存储载体。
StandardEnvironment.customizePropertySources:
/** System environment property source name: {@value} */
public static final String SYSTEM_ENVIRONMENT_PROPERTY_SOURCE_NAME = "systemEnvironment";
/** JVM system properties property source name: {@value} */
public static final String SYSTEM_PROPERTIES_PROPERTY_SOURCE_NAME = "systemProperties";
@Override
protected void customizePropertySources(MutablePropertySources propertySources) {
propertySources.addLast(new MapPropertySource
(SYSTEM_PROPERTIES_PROPERTY_SOURCE_NAME, getSystemProperties()));
propertySources.addLast(new SystemEnvironmentPropertySource
(SYSTEM_ENVIRONMENT_PROPERTY_SOURCE_NAME, getSystemEnvironment()));
}
PropertySource接口
PropertySource接口代表了键值对的Property来源。继承体系:
AbstractEnvironment.getSystemProperties:
@Override
public Map<String, Object> getSystemProperties() {
try {
return (Map) System.getProperties();
}
catch (AccessControlException ex) {
return (Map) new ReadOnlySystemAttributesMap() {
@Override
protected String getSystemAttribute(String attributeName) {
try {
return System.getProperty(attributeName);
}
catch (AccessControlException ex) {
if (logger.isInfoEnabled()) {
logger.info(format("Caught AccessControlException when accessing system " +
"property [%s]; its value will be returned [null]. Reason: %s",
attributeName, ex.getMessage()));
}
return null;
}
}
};
}
}
这里的实现很有意思,如果安全管理器阻止获取全部的系统属性,那么会尝试获取单个属性的可能性,如果还不行就抛异常了。
getSystemEnvironment方法也是一个套路,不过最终调用的是System.getenv,可以获取jvm和OS的一些版本信息。
路径Placeholder处理
AbstractEnvironment.resolveRequiredPlaceholders:
@Override
public String resolveRequiredPlaceholders(String text) throws IllegalArgumentException {
//text即配置文件路径,比如classpath:config.xml
return this.propertyResolver.resolveRequiredPlaceholders(text);
}
propertyResolver是一个PropertySourcesPropertyResolver对象:
private final ConfigurablePropertyResolver propertyResolver =
new PropertySourcesPropertyResolver(this.propertySources);
PropertyResolver接口
PropertyResolver继承体系(排除Environment分支):
此接口正是用来解析PropertyResource。
解析
AbstractPropertyResolver.resolveRequiredPlaceholders:
@Override
public String resolveRequiredPlaceholders(String text) throws IllegalArgumentException {
if (this.strictHelper == null) {
this.strictHelper = createPlaceholderHelper(false);
}
return doResolvePlaceholders(text, this.strictHelper);
}
private PropertyPlaceholderHelper createPlaceholderHelper(boolean ignoreUnresolvablePlaceholders) {
//三个参数分别是${, }, :
return new PropertyPlaceholderHelper(this.placeholderPrefix, this.placeholderSuffix,
this.valueSeparator, ignoreUnresolvablePlaceholders);
}
doResolvePlaceholders:
private String doResolvePlaceholders(String text, PropertyPlaceholderHelper helper) {
//PlaceholderResolver接口依然是策略模式的体现
return helper.replacePlaceholders(text, new PropertyPlaceholderHelper.PlaceholderResolver() {
@Override
public String resolvePlaceholder(String placeholderName) {
return getPropertyAsRawString(placeholderName);
}
});
}
其实代码执行到这里的时候还没有进行xml配置文件的解析,那么这里的解析placeHolder是什么意思呢,原因在于可以这么写:
System.setProperty("spring", "classpath");
ClassPathXmlApplicationContext context = new ClassPathXmlApplicationContext("${spring}:config.xml");
SimpleBean bean = context.getBean(SimpleBean.class);
这样就可以正确解析。placeholder的替换其实就是字符串操作,这里只说一下正确的属性是怎么来的。实现的关键在于PropertySourcesPropertyResolver.getProperty:
@Override
protected String getPropertyAsRawString(String key) {
return getProperty(key, String.class, false);
}
protected <T> T getProperty(String key, Class<T> targetValueType, boolean resolveNestedPlaceholders) {
if (this.propertySources != null) {
for (PropertySource<?> propertySource : this.propertySources) {
Object value = propertySource.getProperty(key);
return value;
}
}
return null;
}
很明显了,就是从System.getProperty和System.getenv获取,但是由于环境变量是无法自定义的,所以其实此处只能通过System.setProperty指定。
注意,classpath:XXX这种写法的classpath前缀到目前为止还没有被处理。
refresh
Spring bean解析就在此方法,所以单独提出来。
AbstractApplicationContext.refresh:
@Override
public void refresh() throws BeansException, IllegalStateException {
synchronized (this.startupShutdownMonitor) {
// Prepare this context for refreshing.
prepareRefresh();
// Tell the subclass to refresh the internal bean factory.
ConfigurableListableBeanFactory beanFactory = obtainFreshBeanFactory();
// Prepare the bean factory for use in this context.
prepareBeanFactory(beanFactory);
try {
// Allows post-processing of the bean factory in context subclasses.
postProcessBeanFactory(beanFactory);
// Invoke factory processors registered as beans in the context.
invokeBeanFactoryPostProcessors(beanFactory);
// Register bean processors that intercept bean creation.
registerBeanPostProcessors(beanFactory);
// Initialize message source for this context.
initMessageSource();
// Initialize event multicaster for this context.
initApplicationEventMulticaster();
// Initialize other special beans in specific context subclasses.
onRefresh();
// Check for listener beans and register them.
registerListeners();
// Instantiate all remaining (non-lazy-init) singletons.
finishBeanFactoryInitialization(beanFactory);
// Last step: publish corresponding event.
finishRefresh();
} catch (BeansException ex) {
// Destroy already created singletons to avoid dangling resources.
destroyBeans();
// Reset 'active' flag.
cancelRefresh(ex);
// Propagate exception to caller.
throw ex;
} finally {
// Reset common introspection caches in Spring's core, since we
// might not ever need metadata for singleton beans anymore...
resetCommonCaches();
}
}
}
prepareRefresh
protected void prepareRefresh() {
this.startupDate = System.currentTimeMillis();
this.closed.set(false);
this.active.set(true);
// Initialize any placeholder property sources in the context environment
//空实现
initPropertySources();
// Validate that all properties marked as required are resolvable
// see ConfigurablePropertyResolver#setRequiredProperties
getEnvironment().validateRequiredProperties();
// Allow for the collection of early ApplicationEvents,
// to be published once the multicaster is available...
this.earlyApplicationEvents = new LinkedHashSet<ApplicationEvent>();
}
属性校验
AbstractEnvironment.validateRequiredProperties:
@Override
public void validateRequiredProperties() throws MissingRequiredPropertiesException {
this.propertyResolver.validateRequiredProperties();
}
AbstractPropertyResolver.validateRequiredProperties:
@Override
public void validateRequiredProperties() {
MissingRequiredPropertiesException ex = new MissingRequiredPropertiesException();
for (String key : this.requiredProperties) {
if (this.getProperty(key) == null) {
ex.addMissingRequiredProperty(key);
}
}
if (!ex.getMissingRequiredProperties().isEmpty()) {
throw ex;
}
}
requiredProperties是通过setRequiredProperties方法设置的,保存在一个list里面,默认是空的,也就是不需要校验任何属性。
BeanFactory创建
由obtainFreshBeanFactory调用AbstractRefreshableApplicationContext.refreshBeanFactory:
@Override
protected final void refreshBeanFactory() throws BeansException {
//如果已经存在,那么销毁之前的
if (hasBeanFactory()) {
destroyBeans();
closeBeanFactory();
}
//创建了一个DefaultListableBeanFactory对象
DefaultListableBeanFactory beanFactory = createBeanFactory();
beanFactory.setSerializationId(getId());
customizeBeanFactory(beanFactory);
loadBeanDefinitions(beanFactory);
synchronized (this.beanFactoryMonitor) {
this.beanFactory = beanFactory;
}
}
BeanFactory接口
此接口实际上就是Bean容器,其继承体系:
BeanFactory定制
AbstractRefreshableApplicationContext.customizeBeanFactory方法用于给子类提供一个自由配置的机会,默认实现:
protected void customizeBeanFactory(DefaultListableBeanFactory beanFactory) {
if (this.allowBeanDefinitionOverriding != null) {
//默认false,不允许覆盖
beanFactory.setAllowBeanDefinitionOverriding(this.allowBeanDefinitionOverriding);
}
if (this.allowCircularReferences != null) {
//默认false,不允许循环引用
beanFactory.setAllowCircularReferences(this.allowCircularReferences);
}
}
Bean加载
AbstractXmlApplicationContext.loadBeanDefinitions,这个便是核心的bean加载了:
@Override
protected void loadBeanDefinitions(DefaultListableBeanFactory beanFactory) {
// Create a new XmlBeanDefinitionReader for the given BeanFactory.
XmlBeanDefinitionReader beanDefinitionReader = new XmlBeanDefinitionReader(beanFactory);
// Configure the bean definition reader with this context's
// resource loading environment.
beanDefinitionReader.setEnvironment(this.getEnvironment());
beanDefinitionReader.setResourceLoader(this);
beanDefinitionReader.setEntityResolver(new ResourceEntityResolver(this));
// Allow a subclass to provide custom initialization of the reader,
// then proceed with actually loading the bean definitions.
//默认空实现
initBeanDefinitionReader(beanDefinitionReader);
loadBeanDefinitions(beanDefinitionReader);
}
EntityResolver
此处只说明用到的部分继承体系:
EntityResolver接口在org.xml.sax中定义。DelegatingEntityResolver用于schema和dtd的解析。
BeanDefinitionReader
继承体系:
路径解析(Ant)
protected void loadBeanDefinitions(XmlBeanDefinitionReader reader) {
Resource[] configResources = getConfigResources();
if (configResources != null) {
reader.loadBeanDefinitions(configResources);
}
String[] configLocations = getConfigLocations();
//here
if (configLocations != null) {
reader.loadBeanDefinitions(configLocations);
}
}
AbstractBeanDefinitionReader.loadBeanDefinitions:
@Override
public int loadBeanDefinitions(String... locations) throws BeanDefinitionStoreException {
Assert.notNull(locations, "Location array must not be null");
int counter = 0;
for (String location : locations) {
counter += loadBeanDefinitions(location);
}
return counter;
}
之后调用:
//第二个参数为空
public int loadBeanDefinitions(String location, Set<Resource> actualResources) {
ResourceLoader resourceLoader = getResourceLoader();
//参见ResourceLoader类图,ClassPathXmlApplicationContext实现了此接口
if (resourceLoader instanceof ResourcePatternResolver) {
// Resource pattern matching available.
try {
Resource[] resources = ((ResourcePatternResolver) resourceLoader).getResources(location);
int loadCount = loadBeanDefinitions(resources);
if (actualResources != null) {
for (Resource resource : resources) {
actualResources.add(resource);
}
}
return loadCount;
}
catch (IOException ex) {
throw new BeanDefinitionStoreException(
"Could not resolve bean definition resource pattern [" + location + "]", ex);
}
}
else {
// Can only load single resources by absolute URL.
Resource resource = resourceLoader.getResource(location);
int loadCount = loadBeanDefinitions(resource);
if (actualResources != null) {
actualResources.add(resource);
}
return loadCount;
}
}
getResource的实现在AbstractApplicationContext:
@Override
public Resource[] getResources(String locationPattern) throws IOException {
//构造器中初始化,PathMatchingResourcePatternResolver对象
return this.resourcePatternResolver.getResources(locationPattern);
}
PathMatchingResourcePatternResolver是ResourceLoader继承体系的一部分。
@Override
public Resource[] getResources(String locationPattern) throws IOException {
Assert.notNull(locationPattern, "Location pattern must not be null");
//classpath:
if (locationPattern.startsWith(CLASSPATH_ALL_URL_PREFIX)) {
// a class path resource (multiple resources for same name possible)
//matcher是一个AntPathMatcher对象
if (getPathMatcher().isPattern(locationPattern
.substring(CLASSPATH_ALL_URL_PREFIX.length()))) {
// a class path resource pattern
return findPathMatchingResources(locationPattern);
} else {
// all class path resources with the given name
return findAllClassPathResources(locationPattern
.substring(CLASSPATH_ALL_URL_PREFIX.length()));
}
} else {
// Only look for a pattern after a prefix here
// (to not get fooled by a pattern symbol in a strange prefix).
int prefixEnd = locationPattern.indexOf(":") + 1;
if (getPathMatcher().isPattern(locationPattern.substring(prefixEnd))) {
// a file pattern
return findPathMatchingResources(locationPattern);
}
else {
// a single resource with the given name
return new Resource[] {getResourceLoader().getResource(locationPattern)};
}
}
}
isPattern:
@Override
public boolean isPattern(String path) {
return (path.indexOf('*') != -1 || path.indexOf('?') != -1);
}
可以看出配置文件路径是支持ant风格的,也就是可以这么写:
new ClassPathXmlApplicationContext("con*.xml");
具体怎么解析ant风格的就不写了。
配置文件加载
入口方法在AbstractBeanDefinitionReader的217行:
//加载
Resource[] resources = ((ResourcePatternResolver) resourceLoader).getResources(location);
//解析
int loadCount = loadBeanDefinitions(resources);
最终逐个调用XmlBeanDefinitionReader的loadBeanDefinitions方法:
@Override
public int loadBeanDefinitions(Resource resource) {
return loadBeanDefinitions(new EncodedResource(resource));
}
Resource是代表一种资源的接口,其类图:
EncodedResource扮演的其实是一个装饰器的模式,为InputStreamSource添加了字符编码(虽然默认为null)。这样为我们自定义xml配置文件的编码方式提供了机会。
之后关键的源码只有两行:
public int loadBeanDefinitions(EncodedResource encodedResource) throws BeanDefinitionStoreException {
InputStream inputStream = encodedResource.getResource().getInputStream();
InputSource inputSource = new InputSource(inputStream);
return doLoadBeanDefinitions(inputSource, encodedResource.getResource());
}
InputSource是org.xml.sax的类。
doLoadBeanDefinitions:
protected int doLoadBeanDefinitions(InputSource inputSource, Resource resource) {
Document doc = doLoadDocument(inputSource, resource);
return registerBeanDefinitions(doc, resource);
}
doLoadDocument:
protected Document doLoadDocument(InputSource inputSource, Resource resource) {
return this.documentLoader.loadDocument(inputSource, getEntityResolver(), this.errorHandler,
getValidationModeForResource(resource), isNamespaceAware());
}
documentLoader是一个DefaultDocumentLoader对象,此类是DocumentLoader接口的唯一实现。getEntityResolver方法返回ResourceEntityResolver,上面说过了。errorHandler是一个SimpleSaxErrorHandler对象。
校验模型其实就是确定xml文件使用xsd方式还是dtd方式来校验,忘了的话左转度娘。Spring会通过读取xml文件的方式判断应该采用哪种。
NamespaceAware默认false,因为默认配置了校验为true。
DefaultDocumentLoader.loadDocument:
@Override
public Document loadDocument(InputSource inputSource, EntityResolver entityResolver,
ErrorHandler errorHandler, int validationMode, boolean namespaceAware) {
//这里就是老套路了,可以看出,Spring还是使用了dom的方式解析,即一次全部load到内存
DocumentBuilderFactory factory = createDocumentBuilderFactory(validationMode, namespaceAware);
DocumentBuilder builder = createDocumentBuilder(factory, entityResolver, errorHandler);
return builder.parse(inputSource);
}
createDocumentBuilderFactory比较有意思:
protected DocumentBuilderFactory createDocumentBuilderFactory(int validationMode, boolean namespaceAware{
DocumentBuilderFactory factory = DocumentBuilderFactory.newInstance();
factory.setNamespaceAware(namespaceAware);
if (validationMode != XmlValidationModeDetector.VALIDATION_NONE) {
//此方法设为true仅对dtd有效,xsd(schema)无效
factory.setValidating(true);
if (validationMode == XmlValidationModeDetector.VALIDATION_XSD) {
// Enforce namespace aware for XSD...
//开启xsd(schema)支持
factory.setNamespaceAware(true);
//这个也是Java支持Schema的套路,可以问度娘
factory.setAttribute(SCHEMA_LANGUAGE_ATTRIBUTE, XSD_SCHEMA_LANGUAGE);
}
}
return factory;
}
Bean解析
XmlBeanDefinitionReader.registerBeanDefinitions:
public int registerBeanDefinitions(Document doc, Resource resource) {
BeanDefinitionDocumentReader documentReader = createBeanDefinitionDocumentReader();
int countBefore = getRegistry().getBeanDefinitionCount();
documentReader.registerBeanDefinitions(doc, createReaderContext(resource));
return getRegistry().getBeanDefinitionCount() - countBefore;
}
createBeanDefinitionDocumentReader:
protected BeanDefinitionDocumentReader createBeanDefinitionDocumentReader() {
return BeanDefinitionDocumentReader.class.cast
//反射
(BeanUtils.instantiateClass(this.documentReaderClass));
}
documentReaderClass默认是DefaultBeanDefinitionDocumentReader,这其实也是策略模式,通过setter方法可以更换其实现。
注意cast方法,代替了强转。
createReaderContext:
public XmlReaderContext createReaderContext(Resource resource) {
return new XmlReaderContext(resource, this.problemReporter, this.eventListener,
this.sourceExtractor, this, getNamespaceHandlerResolver());
}
problemReporter是一个FailFastProblemReporter对象。
eventListener是EmptyReaderEventListener对象,此类里的方法都是空实现。
sourceExtractor是NullSourceExtractor对象,直接返回空,也是空实现。
getNamespaceHandlerResolver默认返回DefaultNamespaceHandlerResolver对象,用来获取xsd对应的处理器。
XmlReaderContext的作用感觉就是这一堆参数的容器,糅合到一起传给DocumentReader,并美其名为Context。可以看出,Spring中到处都是策略模式,大量操作被抽象成接口。
DefaultBeanDefinitionDocumentReader.registerBeanDefinitions:
@Override
public void registerBeanDefinitions(Document doc, XmlReaderContext readerContext) {
this.readerContext = readerContext;
Element root = doc.getDocumentElement();
doRegisterBeanDefinitions(root);
}
doRegisterBeanDefinitions:
protected void doRegisterBeanDefinitions(Element root) {
BeanDefinitionParserDelegate parent = this.delegate;
this.delegate = createDelegate(getReaderContext(), root, parent);
//默认的命名空间即
//http://www.springframework.org/schema/beans
if (this.delegate.isDefaultNamespace(root)) {
//检查profile属性
String profileSpec = root.getAttribute(PROFILE_ATTRIBUTE);
if (StringUtils.hasText(profileSpec)) {
//profile属性可以以,分割
String[] specifiedProfiles = StringUtils.tokenizeToStringArray(
profileSpec, BeanDefinitionParserDelegate.MULTI_VALUE_ATTRIBUTE_DELIMITERS);
if (!getReaderContext().getEnvironment().acceptsProfiles(specifiedProfiles)) {
return;
}
}
}
preProcessXml(root);
parseBeanDefinitions(root, this.delegate);
postProcessXml(root);
this.delegate = parent;
}
delegate的作用在于处理beans标签的嵌套,其实Spring配置文件是可以写成这样的:
<?xml version="1.0" encoding="UTF-8"?>
<beans>
<bean class="base.SimpleBean"></bean>
<beans>
<bean class="java.lang.Object"></bean>
</beans>
</beans>
xml(schema)的命名空间其实类似于java的报名,命名空间采用URL,比如Spring的是这样:
<?xml version="1.0" encoding="UTF-8"?>
<beans xmlns="http://www.springframework.org/schema/beans"></beans>
xmlns属性就是xml规范定义的用来设置命名空间的。这样设置了之后其实里面的bean元素全名就相当于http://www.springframework.org/schema/beans:bean
,可以有效的防止命名冲突。命名空间可以通过规范定义的org.w3c.dom.Node.getNamespaceURI方法获得。
注意一下profile的检查, AbstractEnvironment.acceptsProfiles:
@Override
public boolean acceptsProfiles(String... profiles) {
Assert.notEmpty(profiles, "Must specify at least one profile");
for (String profile : profiles) {
if (StringUtils.hasLength(profile) && profile.charAt(0) == '!') {
if (!isProfileActive(profile.substring(1))) {
return true;
}
} else if (isProfileActive(profile)) {
return true;
}
}
return false;
}
原理很简单,注意从源码可以看出,profile属性支持!取反。
preProcessXml方法是个空实现,供子类去覆盖,目的在于给子类一个把我们自定义的标签转为Spring标准标签的机会, 想的真周到。
DefaultBeanDefinitionDocumentReader.parseBeanDefinitions:
protected void parseBeanDefinitions(Element root, BeanDefinitionParserDelegate delegate) {
if (delegate.isDefaultNamespace(root)) {
NodeList nl = root.getChildNodes();
for (int i = 0; i < nl.getLength(); i++) {
Node node = nl.item(i);
if (node instanceof Element) {
Element ele = (Element) node;
if (delegate.isDefaultNamespace(ele)) {
parseDefaultElement(ele, delegate);
} else {
delegate.parseCustomElement(ele);
}
}
}
} else {
delegate.parseCustomElement(root);
}
}
可见,对于非默认命名空间的元素交由delegate处理。
默认命名空间解析
即import, alias, bean, 嵌套的beans四种元素。parseDefaultElement:
private void parseDefaultElement(Element ele, BeanDefinitionParserDelegate delegate) {
//"import"
if (delegate.nodeNameEquals(ele, IMPORT_ELEMENT)) {
importBeanDefinitionResource(ele);
}
else if (delegate.nodeNameEquals(ele, ALIAS_ELEMENT)) {
processAliasRegistration(ele);
}
else if (delegate.nodeNameEquals(ele, BEAN_ELEMENT)) {
processBeanDefinition(ele, delegate);
}
else if (delegate.nodeNameEquals(ele, NESTED_BEANS_ELEMENT)) {
// recurse
doRegisterBeanDefinitions(ele);
}
}
import
写法示例:
<import resource="CTIContext.xml" />
<import resource="customerContext.xml" />
importBeanDefinitionResource套路和之前的配置文件加载完全一样,不过注意被import进来的文件是先于当前文件 被解析的。
alias
加入有一个bean名为componentA-dataSource,但是另一个组件想以componentB-dataSource的名字使用,就可以这样定义:
<alias name="componentA-dataSource" alias="componentB-dataSource"/>
processAliasRegistration核心源码:
protected void processAliasRegistration(Element ele) {
String name = ele.getAttribute(NAME_ATTRIBUTE);
String alias = ele.getAttribute(ALIAS_ATTRIBUTE);
getReaderContext().getRegistry().registerAlias(name, alias);
getReaderContext().fireAliasRegistered(name, alias, extractSource(ele));
}
从前面的源码可以发现,registry其实就是DefaultListableBeanFactory,它实现了BeanDefinitionRegistry接口。registerAlias方法的实现在SimpleAliasRegistry:
@Override
public void registerAlias(String name, String alias) {
Assert.hasText(name, "'name' must not be empty");
Assert.hasText(alias, "'alias' must not be empty");
//名字和别名一样
if (alias.equals(name)) {
//ConcurrentHashMap
this.aliasMap.remove(alias);
} else {
String registeredName = this.aliasMap.get(alias);
if (registeredName != null) {
if (registeredName.equals(name)) {
// An existing alias - no need to re-register
return;
}
if (!allowAliasOverriding()) {
throw new IllegalStateException
("Cannot register alias '" + alias + "' for name '" +
name + "': It is already registered for name '" + registeredName + "'.");
}
}
checkForAliasCircle(name, alias);
this.aliasMap.put(alias, name);
}
}
所以别名关系的保存使用Map完成,key为别名,value为本来的名字。
bean
bean节点是Spring最最常见的节点了。
DefaultBeanDefinitionDocumentReader.processBeanDefinition:
protected void processBeanDefinition(Element ele, BeanDefinitionParserDelegate delegate) {
BeanDefinitionHolder bdHolder = delegate.parseBeanDefinitionElement(ele);
if (bdHolder != null) {
bdHolder = delegate.decorateBeanDefinitionIfRequired(ele, bdHolder);
try {
// Register the final decorated instance.
BeanDefinitionReaderUtils.registerBeanDefinition
(bdHolder, getReaderContext().getRegistry());
}
catch (BeanDefinitionStoreException ex) {
getReaderContext().error("Failed to register bean definition with name '" +
bdHolder.getBeanName() + "'", ele, ex);
}
// Send registration event.
getReaderContext().fireComponentRegistered(new BeanComponentDefinition(bdHolder));
}
}
id & name处理
最终调用BeanDefinitionParserDelegate.parseBeanDefinitionElement(Element ele, BeanDefinition containingBean),源码较长,分部分说明。
首先获取到id和name属性,name属性支持配置多个,以逗号分隔,如果没有指定id,那么将以第一个name属性值代替。id必须是唯一的,name属性其实是alias的角色,可以和其它的bean重复,如果name也没有配置,那么其实什么也没做。
String id = ele.getAttribute(ID_ATTRIBUTE);
String nameAttr = ele.getAttribute(NAME_ATTRIBUTE);
List<String> aliases = new ArrayList<String>();
if (StringUtils.hasLength(nameAttr)) {
//按,分隔
String[] nameArr = StringUtils.tokenizeToStringArray
(nameAttr, MULTI_VALUE_ATTRIBUTE_DELIMITERS);
aliases.addAll(Arrays.asList(nameArr));
}
String beanName = id;
if (!StringUtils.hasText(beanName) && !aliases.isEmpty()) {
//name的第一个值作为id
beanName = aliases.remove(0);
}
//默认null
if (containingBean == null) {
//校验id是否已重复,如果重复直接抛异常
//校验是通过内部一个HashSet完成的,出现过的id都会保存进此Set
checkNameUniqueness(beanName, aliases, ele);
}
beanName生成
如果name和id属性都没有指定,那么Spring会自己生成一个, BeanDefinitionParserDelegate.parseBeanDefinitionElement:
beanName = this.readerContext.generateBeanName(beanDefinition);
String beanClassName = beanDefinition.getBeanClassName();
aliases.add(beanClassName);
可见,Spring同时会把类名作为其别名。
最终调用的是BeanDefinitionReaderUtils.generateBeanName:
public static String generateBeanName(
BeanDefinition definition, BeanDefinitionRegistry registry, boolean isInnerBean) {
String generatedBeanName = definition.getBeanClassName();
if (generatedBeanName == null) {
if (definition.getParentName() != null) {
generatedBeanName = definition.getParentName() + "$child";
//工厂方法产生的bean
} else if (definition.getFactoryBeanName() != null) {
generatedBeanName = definition.getFactoryBeanName() + "$created";
}
}
String id = generatedBeanName;
if (isInnerBean) {
// Inner bean: generate identity hashcode suffix.
id = generatedBeanName + GENERATED_BEAN_NAME_SEPARATOR +
ObjectUtils.getIdentityHexString(definition);
} else {
// Top-level bean: use plain class name.
// Increase counter until the id is unique.
int counter = -1;
//用类名#自增的数字命名
while (counter == -1 || registry.containsBeanDefinition(id)) {
counter++;
id = generatedBeanName + GENERATED_BEAN_NAME_SEPARATOR + counter;
}
}
return id;
}
bean解析
还是分部分说明(parseBeanDefinitionElement)。
首先获取到bean的class属性和parent属性,配置了parent之后,当前bean会继承父bean的属性。之后根据class和parent创建BeanDefinition对象。
String className = null;
if (ele.hasAttribute(CLASS_ATTRIBUTE)) {
className = ele.getAttribute(CLASS_ATTRIBUTE).trim();
}
String parent = null;
if (ele.hasAttribute(PARENT_ATTRIBUTE)) {
parent = ele.getAttribute(PARENT_ATTRIBUTE);
}
AbstractBeanDefinition bd = createBeanDefinition(className, parent);
BeanDefinition的创建在BeanDefinitionReaderUtils.createBeanDefinition:
public static AbstractBeanDefinition createBeanDefinition(
String parentName, String className, ClassLoader classLoader) {
GenericBeanDefinition bd = new GenericBeanDefinition();
bd.setParentName(parentName);
if (className != null) {
if (classLoader != null) {
bd.setBeanClass(ClassUtils.forName(className, classLoader));
}
else {
bd.setBeanClassName(className);
}
}
return bd;
}
之后是解析bean的其它属性,其实就是读取其配置,调用相应的setter方法保存在BeanDefinition中:
parseBeanDefinitionAttributes(ele, beanName, containingBean, bd);
之后解析bean的decription子元素:
<bean id="b" name="one, two" class="base.SimpleBean">
<description>SimpleBean</description>
</bean>
就仅仅是个描述。
然后是meta子元素的解析,meta元素在xml配置文件里是这样的:
<bean id="b" name="one, two" class="base.SimpleBean">
<meta key="name" value="skywalker"/>
</bean>
注释上说,这样可以将任意的元数据附到对应的bean definition上。解析过程源码:
public void parseMetaElements(Element ele, BeanMetadataAttributeAccessor attributeAccessor) {
NodeList nl = ele.getChildNodes();
for (int i = 0; i < nl.getLength(); i++) {
Node node = nl.item(i);
if (isCandidateElement(node) && nodeNameEquals(node, META_ELEMENT)) {
Element metaElement = (Element) node;
String key = metaElement.getAttribute(KEY_ATTRIBUTE);
String value = metaElement.getAttribute(VALUE_ATTRIBUTE);
//就是一个key, value的载体,无他
BeanMetadataAttribute attribute = new BeanMetadataAttribute(key, value);
//sourceExtractor默认是NullSourceExtractor,返回的是空
attribute.setSource(extractSource(metaElement));
attributeAccessor.addMetadataAttribute(attribute);
}
}
}
AbstractBeanDefinition继承自BeanMetadataAttributeAccessor类,底层使用了一个LinkedHashMap保存metadata。这个metadata具体是做什么暂时还不知道。
lookup-method解析:
此标签的作用在于当一个bean的某个方法被设置为lookup-method后,每次调用此方法时,都会返回一个新的指定bean的对象。用法示例:
<bean id="apple" class="cn.com.willchen.test.di.Apple" scope="prototype"/>
<!--水果盘-->
<bean id="fruitPlate" class="cn.com.willchen.test.di.FruitPlate">
<lookup-method name="getFruit" bean="apple"/>
</bean>
数据保存在Set中,对应的类是MethodOverrides。可以参考:
Spring - lookup-method方式实现依赖注入
replace-mothod解析:
此标签用于替换bean里面的特定的方法实现,替换者必须实现Spring的MethodReplacer接口,有点像aop的意思。
配置文件示例:
<bean name="replacer" class="springroad.deomo.chap4.MethodReplace" />
<bean name="testBean" class="springroad.deomo.chap4.LookupMethodBean">
<replaced-method name="test" replacer="replacer">
<arg-type match="String" />
</replaced-method>
</bean>
arg-type的作用是指定替换方法的参数类型,因为接口的定义参数都是Object的。参考: SPRING.NET 1.3.2 学习20--方法注入之替换方法注入
解析之后将数据放在ReplaceOverride对象中,里面有一个LinkedList
构造参数(constructor-arg)解析:
作用一目了然,使用示例:
<bean class="base.SimpleBean">
<constructor-arg>
<value type="java.lang.String">Cat</value>
</constructor-arg>
</bean>
type一般不需要指定,除了泛型集合那种。除此之外,constructor-arg还支持name, index, ref等属性,可以具体的指定参数的位置等。构造参数解析后保存在BeanDefinition内部一个ConstructorArgumentValues对象中。如果设置了index属性,那么以Map<Integer, ValueHolder>的形式保存,反之,以List
property解析:
非常常用的标签,用以为bean的属性赋值,支持value和ref两种形式,示例:
<bean class="base.SimpleBean">
<property name="name" value="skywalker" />
</bean>
value和ref属性不能同时出现,如果是ref,那么将其值保存在不可变的RuntimeBeanReference对象中,其实现了BeanReference接口,此接口只有一个getBeanName方法。如果是value,那么将其值保存在TypedStringValue对象中。最终将对象保存在BeanDefinition内部一个MutablePropertyValues对象中(内部以ArrayList实现)。
qualifier解析:
配置示例:
<bean class="base.Student">
<property name="name" value="skywalker"></property>
<property name="age" value="12"></property>
<qualifier type="org.springframework.beans.factory.annotation.Qualifier" value="student" />
</bean>
<bean class="base.Student">
<property name="name" value="seaswalker"></property>
<property name="age" value="15"></property>
<qualifier value="student_2"></qualifier>
</bean>
<bean class="base.SimpleBean" />
SimpleBean部分源码:
@Autowired
@Qualifier("student")
private Student student;
此标签和@Qualifier, @Autowired两个注解一起使用才有作用。@Autowired注解采用按类型查找的方式进行注入,如果找到多个需要类型的bean便会报错,有了@Qualifier标签就可以再按照此注解指定的名称查找。两者结合相当于实现了按类型+名称注入。type属性可以不指定,因为默认就是那个。qualifier标签可以有attribute子元素,比如:
<qualifier type="org.springframework.beans.factory.annotation.Qualifier" value="student">
<attribute key="id" value="1"/>
</qualifier>
貌似是用来在qualifier也区分不开的时候使用。attribute键值对保存在BeanMetadataAttribute对象中。整个qualifier保存在AutowireCandidateQualifier对象中。
Bean装饰
这部分是针对其它schema的属性以及子节点,比如:
<bean class="base.Student" primary="true">
<context:property-override />
</bean>
没见过这种用法,留个坑。
Bean注册
BeanDefinitionReaderUtils.registerBeanDefinition:
public static void registerBeanDefinition(
BeanDefinitionHolder definitionHolder, BeanDefinitionRegistry registry) {
// Register bean definition under primary name.
String beanName = definitionHolder.getBeanName();
registry.registerBeanDefinition(beanName, definitionHolder.getBeanDefinition());
// Register aliases for bean name, if any.
String[] aliases = definitionHolder.getAliases();
if (aliases != null) {
for (String alias : aliases) {
registry.registerAlias(beanName, alias);
}
}
}
registry其实就是DefaultListableBeanFactory对象,registerBeanDefinition方法主要就干了这么两件事:
@Override
public void registerBeanDefinition(String beanName, BeanDefinition beanDefinition) {
this.beanDefinitionMap.put(beanName, beanDefinition);
this.beanDefinitionNames.add(beanName);
}
一个是Map,另一个是List,一目了然。registerAlias方法的实现在其父类SimpleAliasRegistry,就是把键值对放在了一个ConcurrentHashMap里。
ComponentRegistered事件触发:
默认是个空实现,前面说过了。
BeanDefiniton数据结构
BeanDefiniton数据结构如下图:
beans
beans元素的嵌套直接递归调用DefaultBeanDefinitionDocumentReader.parseBeanDefinitions。
其它命名空间解析
入口在DefaultBeanDefinitionDocumentReader.parseBeanDefinitions->BeanDefinitionParserDelegate.parseCustomElement(第二个参数为空):
public BeanDefinition parseCustomElement(Element ele, BeanDefinition containingBd) {
String namespaceUri = getNamespaceURI(ele);
NamespaceHandler handler = this.readerContext.getNamespaceHandlerResolver().resolve(namespaceUri);
return handler.parse(ele, new ParserContext(this.readerContext, this, containingBd));
}
NamespaceHandlerResolver由XmlBeanDefinitionReader初始化,是一个DefaultNamespaceHandlerResolver对象,也是NamespaceHandlerResolver接口的唯一实现。
其resolve方法:
@Override
public NamespaceHandler resolve(String namespaceUri) {
Map<String, Object> handlerMappings = getHandlerMappings();
Object handlerOrClassName = handlerMappings.get(namespaceUri);
if (handlerOrClassName == null) {
return null;
} else if (handlerOrClassName instanceof NamespaceHandler) {
return (NamespaceHandler) handlerOrClassName;
} else {
String className = (String) handlerOrClassName;
Class<?> handlerClass = ClassUtils.forName(className, this.classLoader);
NamespaceHandler namespaceHandler = (NamespaceHandler) BeanUtils.instantiateClass(handlerClass);
namespaceHandler.init();
handlerMappings.put(namespaceUri, namespaceHandler);
return namespaceHandler;
}
}
容易看出,Spring其实使用了一个Map了保存其映射关系,key就是命名空间的uri,value是NamespaceHandler对象或是Class完整名,如果发现是类名,那么用反射的方法进行初始化,如果是NamespaceHandler对象,那么直接返回。
NamespaceHandler映射关系来自于各个Spring jar包下的META-INF/spring.handlers文件,以spring-context包为例:
http\://www.springframework.org/schema/context=org.springframework.context.config.ContextNamespaceHandler
http\://www.springframework.org/schema/jee=org.springframework.ejb.config.JeeNamespaceHandler
http\://www.springframework.org/schema/lang=org.springframework.scripting.config.LangNamespaceHandler
http\://www.springframework.org/schema/task=org.springframework.scheduling.config.TaskNamespaceHandler
http\://www.springframework.org/schema/cache=org.springframework.cache.config.CacheNamespaceHandler
NamespaceHandler继承体系
init
resolve中调用了其init方法,此方法用以向NamespaceHandler对象注册BeanDefinitionParser对象。此接口用以解析顶层(beans下)的非默认命名空间元素,比如<context:annotation-config />
。
所以这样逻辑就很容易理解了: 每种子标签的解析仍是策略模式的体现,init负责向父类NamespaceHandlerSupport注册不同的策略,由父类的NamespaceHandlerSupport.parse方法根据具体的子标签调用相应的策略完成解析的过程。
此部分较为重要,所以重新开始大纲。
BeanFactory数据结构
BeanDefinition在BeanFactory中的主要数据结构如下图:
prepareBeanFactory
此方法负责对BeanFactory进行一些特征的设置工作,"特征"包含这么几个方面:
BeanExpressionResolver
此接口只有一个实现: StandardBeanExpressionResolver。接口只含有一个方法:
Object evaluate(String value, BeanExpressionContext evalContext)
prepareBeanFactory将一个此对象放入BeanFactory:
beanFactory.setBeanExpressionResolver(new StandardBeanExpressionResolver(beanFactory.getBeanClassLoader()));
StandardBeanExpressionResolver对象内部有一个关键的成员: SpelExpressionParser,其整个类图:
这便是Spring3.0开始出现的Spel表达式的解释器。
PropertyEditorRegistrar
此接口用于向Spring注册java.beans.PropertyEditor,只有一个方法:
registerCustomEditors(PropertyEditorRegistry registry)
实现也只有一个: ResourceEditorRegistrar。
在编写xml配置时,我们设置的值都是字符串形式,所以在使用时肯定需要转为我们需要的类型,PropertyEditor接口正是定义了这么个东西。
prepareBeanFactory:
beanFactory.addPropertyEditorRegistrar(new ResourceEditorRegistrar(this, getEnvironment()));
BeanFactory也暴露了registerCustomEditors方法用以添加自定义的转换器,所以这个地方是组合模式的体现。
我们有两种方式可以添加自定义PropertyEditor:
-
通过
context.getBeanFactory().registerCustomEditor
-
通过Spring配置文件:
<bean class="org.springframework.beans.factory.config.CustomEditorConfigurer"> <property name="customEditors"> <map> <entry key="base.Cat" value="base.CatEditor" /> </map> </property> </bean>
参考: 深入理解JavaBean(2):属性编辑器PropertyEditor
环境注入
在Spring中我们自己的bean可以通过实现EnvironmentAware等一系列Aware接口获取到Spring内部的一些对象。prepareBeanFactory:
beanFactory.addBeanPostProcessor(new ApplicationContextAwareProcessor(this));
ApplicationContextAwareProcessor核心的invokeAwareInterfaces方法:
private void invokeAwareInterfaces(Object bean) {
if (bean instanceof Aware) {
if (bean instanceof EnvironmentAware) {
((EnvironmentAware) bean).setEnvironment(this.applicationContext.getEnvironment());
}
if (bean instanceof EmbeddedValueResolverAware) {
((EmbeddedValueResolverAware) bean).setEmbeddedValueResolver(this.embeddedValueResolver);
}
//....
}
}
依赖解析忽略
此部分设置哪些接口在进行依赖注入的时候应该被忽略:
beanFactory.ignoreDependencyInterface(ResourceLoaderAware.class);
beanFactory.ignoreDependencyInterface(ApplicationEventPublisherAware.class);
beanFactory.ignoreDependencyInterface(MessageSourceAware.class);
beanFactory.ignoreDependencyInterface(ApplicationContextAware.class);
beanFactory.ignoreDependencyInterface(EnvironmentAware.class);
bean伪装
有些对象并不在BeanFactory中,但是我们依然想让它们可以被装配,这就需要伪装一下:
beanFactory.registerResolvableDependency(BeanFactory.class, beanFactory);
beanFactory.registerResolvableDependency(ResourceLoader.class, this);
beanFactory.registerResolvableDependency(ApplicationEventPublisher.class, this);
beanFactory.registerResolvableDependency(ApplicationContext.class, this);
伪装关系保存在一个Map<Class<?>, Object>里。
LoadTimeWeaver
如果配置了此bean,那么:
if (beanFactory.containsBean(LOAD_TIME_WEAVER_BEAN_NAME)) {
beanFactory.addBeanPostProcessor(new LoadTimeWeaverAwareProcessor(beanFactory));
// Set a temporary ClassLoader for type matching.
beanFactory.setTempClassLoader(new ContextTypeMatchClassLoader(beanFactory.getBeanClassLoader()));
}
这个东西具体是干什么的在后面context:load-time-weaver中说明。
注册环境
源码:
if (!beanFactory.containsLocalBean(ENVIRONMENT_BEAN_NAME)) {
beanFactory.registerSingleton(ENVIRONMENT_BEAN_NAME, getEnvironment());
}
if (!beanFactory.containsLocalBean(SYSTEM_PROPERTIES_BEAN_NAME)) {
beanFactory.registerSingleton(SYSTEM_PROPERTIES_BEAN_NAME, getEnvironment().getSystemProperties());
}
if (!beanFactory.containsLocalBean(SYSTEM_ENVIRONMENT_BEAN_NAME)) {
beanFactory.registerSingleton(SYSTEM_ENVIRONMENT_BEAN_NAME, getEnvironment().
getSystemEnvironment());
}
containsLocalBean特殊之处在于不会去父BeanFactory寻找。
postProcessBeanFactory
此方法允许子类在所有的bean尚未初始化之前注册BeanPostProcessor。空实现且没有子类覆盖。
invokeBeanFactoryPostProcessors
BeanFactoryPostProcessor接口允许我们在bean正是初始化之前改变其值。此接口只有一个方法:
void postProcessBeanFactory(ConfigurableListableBeanFactory beanFactory);
有两种方式可以向Spring添加此对象:
-
通过代码的方式:
context.addBeanFactoryPostProcessor
-
通过xml配置的方式:
<bean class="base.SimpleBeanFactoryPostProcessor" />
注意此时尚未进行bean的初始化工作,初始化是在后面的finishBeanFactoryInitialization进行的,所以在BeanFactoryPostProcessor对象中获取bean会导致提前初始化。
此方法的关键源码:
protected void invokeBeanFactoryPostProcessors(ConfigurableListableBeanFactory beanFactory) {
PostProcessorRegistrationDelegate.invokeBeanFactoryPostProcessors(beanFactory,
getBeanFactoryPostProcessors());
}
getBeanFactoryPostProcessors获取的就是AbstractApplicationContext的成员beanFactoryPostProcessors(ArrayList),但是很有意思,只有通过context.addBeanFactoryPostProcessor这种方式添加的才会出现在这个List里,所以对于xml配置方式,此List其实没有任何元素。玄机就在PostProcessorRegistrationDelegate里。
核心思想就是使用BeanFactory的getBeanNamesForType方法获取相应的BeanDefinition的name数组,之后逐一调用getBean方法获取到bean(初始化),getBean方法后面再说。
注意此处有一个优先级的概念,如果你的BeanFactoryPostProcessor同时实现了Ordered或者是PriorityOrdered接口,那么会被首先执行。
BeanPostProcessor注册
此部分实质上是在BeanDefinitions中寻找BeanPostProcessor,之后调用BeanFactory.addBeanPostProcessor方法保存在一个List中,注意添加时仍然有优先级的概念,优先级高的在前面。
MessageSource
此接口用以支持Spring国际化。继承体系如下:
AbstractApplicationContext的initMessageSource()方法就是在BeanFactory中查找MessageSource的bean,如果配置了此bean,那么调用getBean方法完成其初始化并将其保存在AbstractApplicationContext内部messageSource成员变量中,用以处理ApplicationContext的getMessage调用,因为从继承体系上来看,ApplicationContext是MessageSource的子类,此处是委托模式的体现。如果没有配置此bean,那么初始化一个DelegatingMessageSource对象,此类是一个空实现,同样用以处理getMessage调用请求。
参考: 学习Spring必学的Java基础知识(8)----国际化信息
事件驱动
此接口代表了Spring的事件驱动(监听器)模式。一个事件驱动包含三部分:
事件
java的所有事件对象一般都是java.util.EventObject的子类,Spring的整个继承体系如下:
发布者
ApplicationEventPublisher
一目了然。
ApplicationEventMulticaster
ApplicationEventPublisher实际上正是将请求委托给ApplicationEventMulticaster来实现的。其继承体系:
监听器
所有的监听器是jdk EventListener的子类,这是一个mark接口。继承体系:
可以看出SmartApplicationListener和GenericApplicationListener是高度相似的,都提供了事件类型检测和顺序机制,而后者是从Spring4.2加入的,Spring官方文档推荐使用后者代替前者。
初始化
前面说过ApplicationEventPublisher是通过委托给ApplicationEventMulticaster实现的,所以refresh方法中完成的是对ApplicationEventMulticaster的初始化:
// Initialize event multicaster for this context.
initApplicationEventMulticaster();
initApplicationEventMulticaster则首先在BeanFactory中寻找ApplicationEventMulticaster的bean,如果找到,那么调用getBean方法将其初始化,如果找不到那么使用SimpleApplicationEventMulticaster。
事件发布
AbstractApplicationContext.publishEvent核心代码:
protected void publishEvent(Object event, ResolvableType eventType) {
getApplicationEventMulticaster().multicastEvent(applicationEvent, eventType);
}
SimpleApplicationEventMulticaster.multicastEvent:
@Override
public void multicastEvent(final ApplicationEvent event, ResolvableType eventType) {
ResolvableType type = (eventType != null ? eventType : resolveDefaultEventType(event));
for (final ApplicationListener<?> listener : getApplicationListeners(event, type)) {
Executor executor = getTaskExecutor();
if (executor != null) {
executor.execute(new Runnable() {
@Override
public void run() {
invokeListener(listener, event);
}
});
} else {
invokeListener(listener, event);
}
}
}
监听器获取
获取当然还是通过beanFactory的getBean来完成的,值得注意的是Spring在此处使用了缓存(ConcurrentHashMap)来加速查找的过程。
同步/异步
可以看出,如果executor不为空,那么监听器的执行实际上是异步的。那么如何配置同步/异步呢?
全局
<task:executor id="multicasterExecutor" pool-size="3"/>
<bean class="org.springframework.context.event.SimpleApplicationEventMulticaster">
<property name="taskExecutor" ref="multicasterExecutor"></property>
</bean>
task schema是Spring从3.0开始加入的,使我们可以不再依赖于Quartz实现定时任务,源码在org.springframework.core.task包下,使用需要引入schema:
xmlns:task="http://www.springframework.org/schema/task"
xsi:schemaLocation="http://www.springframework.org/schema/task http://www.springframework.org/schema/task/spring-task-4.0.xsd"
可以参考: Spring定时任务的几种实现
注解
开启注解支持:
<!-- 开启@AspectJ AOP代理 -->
<aop:aspectj-autoproxy proxy-target-class="true"/>
<!-- 任务调度器 -->
<task:scheduler id="scheduler" pool-size="10"/>
<!-- 任务执行器 -->
<task:executor id="executor" pool-size="10"/>
<!--开启注解调度支持 @Async @Scheduled-->
<task:annotation-driven executor="executor" scheduler="scheduler" proxy-target-class="true"/>
在代码中使用示例:
@Component
public class EmailRegisterListener implements ApplicationListener<RegisterEvent> {
@Async
@Override
public void onApplicationEvent(final RegisterEvent event) {
System.out.println("注册成功,发送确认邮件给:" + ((User)event.getSource()).getUsername());
}
}
参考: 详解Spring事件驱动模型
onRefresh
这又是一个模版方法,允许子类在进行bean初始化之前进行一些定制操作。默认空实现。
ApplicationListener注册
registerListeners方法干的,没什么好说的。
singleton初始化
finishBeanFactoryInitialization:
protected void finishBeanFactoryInitialization(ConfigurableListableBeanFactory beanFactory) {
if (beanFactory.containsBean(CONVERSION_SERVICE_BEAN_NAME) &&
beanFactory.isTypeMatch(CONVERSION_SERVICE_BEAN_NAME, ConversionService.class)) {
beanFactory.setConversionService(
beanFactory.getBean(CONVERSION_SERVICE_BEAN_NAME, ConversionService.class));
}
if (!beanFactory.hasEmbeddedValueResolver()) {
beanFactory.addEmbeddedValueResolver(new StringValueResolver() {
@Override
public String resolveStringValue(String strVal) {
return getEnvironment().resolvePlaceholders(strVal);
}
});
}
String[] weaverAwareNames = beanFactory.getBeanNamesForType
(LoadTimeWeaverAware.class, false, false);
for (String weaverAwareName : weaverAwareNames) {
getBean(weaverAwareName);
}
// Allow for caching all bean definition metadata, not expecting further changes.
beanFactory.freezeConfiguration();
// Instantiate all remaining (non-lazy-init) singletons.
beanFactory.preInstantiateSingletons();
}
分部分说明。
ConversionService
此接口用于类型之间的转换,在Spring里其实就是把配置文件中的String转为其它类型,从3.0开始出现,目的和jdk的PropertyEditor接口是一样的,参考ConfigurableBeanFactory.setConversionService注释:
Specify a Spring 3.0 ConversionService to use for converting
property values, as an alternative to JavaBeans PropertyEditors.
@since 3.0
StringValueResolver
用于解析注解的值。接口只定义了一个方法:
String resolveStringValue(String strVal);
LoadTimeWeaverAware
实现了此接口的bean可以得到LoadTimeWeaver,此处仅仅初始化。
初始化
DefaultListableBeanFactory.preInstantiateSingletons:
@Override
public void preInstantiateSingletons() throws BeansException {
List<String> beanNames = new ArrayList<String>(this.beanDefinitionNames);
for (String beanName : beanNames) {
RootBeanDefinition bd = getMergedLocalBeanDefinition(beanName);
if (!bd.isAbstract() && bd.isSingleton() && !bd.isLazyInit()) {
if (isFactoryBean(beanName)) {
final FactoryBean<?> factory = (FactoryBean<?>) getBean(FACTORY_BEAN_PREFIX
+ beanName);
boolean isEagerInit;
if (System.getSecurityManager() != null && factory instanceof SmartFactoryBean) {
isEagerInit = AccessController.doPrivileged(new PrivilegedAction<Boolean>() {
@Override
public Boolean run() {
return ((SmartFactoryBean<?>) factory).isEagerInit();
}
}, getAccessControlContext());
}
else {
isEagerInit = (factory instanceof SmartFactoryBean &&
((SmartFactoryBean<?>) factory).isEagerInit());
}
if (isEagerInit) {
getBean(beanName);
}
}
else {
getBean(beanName);
}
}
}
// Trigger post-initialization callback for all applicable beans...
for (String beanName : beanNames) {
Object singletonInstance = getSingleton(beanName);
if (singletonInstance instanceof SmartInitializingSingleton) {
final SmartInitializingSingleton smartSingleton =
(SmartInitializingSingleton) singletonInstance;
if (System.getSecurityManager() != null) {
AccessController.doPrivileged(new PrivilegedAction<Object>() {
@Override
public Object run() {
smartSingleton.afterSingletonsInstantiated();
return null;
}
}, getAccessControlContext());
}
else {
smartSingleton.afterSingletonsInstantiated();
}
}
}
}
首先进行Singleton的初始化,其中如果bean是FactoryBean类型(注意,只定义了factory-method属性的普通bean并不是FactoryBean),并且还是SmartFactoryBean类型,那么需要判断是否需要eagerInit(isEagerInit是此接口定义的方法)。
getBean
这里便是bean初始化的核心逻辑。源码比较复杂,分开说。以getBean(String name)为例。AbstractBeanFactory.getBean:
@Override
public Object getBean(String name) throws BeansException {
return doGetBean(name, null, null, false);
}
第二个参数表示bean的Class类型,第三个表示创建bean需要的参数,最后一个表示不需要进行类型检查。
beanName转化
final String beanName = transformedBeanName(name);
这里是将FactoryBean的前缀去掉以及将别名转为真实的名字。
手动注册bean检测
前面注册环境一节说过,Spring其实手动注册了一些单例bean。这一步就是检测是不是这些bean。如果是,那么再检测是不是工厂bean,如果是返回其工厂方法返回的实例,如果不是返回bean本身。
Object sharedInstance = getSingleton(beanName);
if (sharedInstance != null && args == null) {
bean = getObjectForBeanInstance(sharedInstance, name, beanName, null);
}
检查父容器
如果父容器存在并且存在此bean定义,那么交由其父容器初始化:
BeanFactory parentBeanFactory = getParentBeanFactory();
if (parentBeanFactory != null && !containsBeanDefinition(beanName)) {
// Not found -> check parent.
//此方法其实是做了前面beanName转化的逆操作,因为父容器同样会进行转化操作
String nameToLookup = originalBeanName(name);
if (args != null) {
// Delegation to parent with explicit args.
return (T) parentBeanFactory.getBean(nameToLookup, args);
} else {
// No args -> delegate to standard getBean method.
return parentBeanFactory.getBean(nameToLookup, requiredType);
}
}
依赖初始化
bean可以由depends-on属性配置依赖的bean。Spring会首先初始化依赖的bean。
String[] dependsOn = mbd.getDependsOn();
if (dependsOn != null) {
for (String dependsOnBean : dependsOn) {
//检测是否存在循环依赖
if (isDependent(beanName, dependsOnBean)) {
throw new BeanCreationException(mbd.getResourceDescription(), beanName,
"Circular depends-on relationship between '" + beanName + "' and '" + dependsOnBean + "'");
}
registerDependentBean(dependsOnBean, beanName);
getBean(dependsOnBean);
}
}
registerDependentBean进行了依赖关系的注册,这么做的原因是Spring在即进行bean销毁的时候会首先销毁被依赖的bean。依赖关系的保存是通过一个ConcurrentHashMap<String, Set
Singleton初始化
虽然这里大纲是Singleton初始化,但是getBean方法本身是包括所有scope的初始化,在这里一次说明了。
if (mbd.isSingleton()) {
sharedInstance = getSingleton(beanName, new ObjectFactory<Object>() {
@Override
public Object getObject() throws BeansException {
return createBean(beanName, mbd, args);
}
});
bean = getObjectForBeanInstance(sharedInstance, name, beanName, mbd);
}
getSingleton方法
是否存在
首先会检测是否已经存在,如果存在,直接返回:
synchronized (this.singletonObjects) {
Object singletonObject = this.singletonObjects.get(beanName);
}
所有的单例bean都保存在这样的数据结构中: ConcurrentHashMap<String, Object>
。
bean创建
源码位于AbstractAutowireCapableBeanFactory.createBean,主要分为几个部分:
lookup-method检测
此部分用于检测lookup-method标签配置的方法是否存在:
RootBeanDefinition mbdToUse = mbd;
mbdToUse.prepareMethodOverrides();
prepareMethodOverrides:
public void prepareMethodOverrides() throws BeanDefinitionValidationException {
// Check that lookup methods exists.
MethodOverrides methodOverrides = getMethodOverrides();
if (!methodOverrides.isEmpty()) {
Set<MethodOverride> overrides = methodOverrides.getOverrides();
synchronized (overrides) {
for (MethodOverride mo : overrides) {
prepareMethodOverride(mo);
}
}
}
}
prepareMethodOverride:
protected void prepareMethodOverride(MethodOverride mo) {
int count = ClassUtils.getMethodCountForName(getBeanClass(), mo.getMethodName());
if (count == 0) {
throw new BeanDefinitionValidationException(
"Invalid method override: no method with name '" + mo.getMethodName() +
"' on class [" + getBeanClassName() + "]");
} else if (count == 1) {
// Mark override as not overloaded, to avoid the overhead of arg type checking.
mo.setOverloaded(false);
}
}
InstantiationAwareBeanPostProcessor触发
在这里触发的是其postProcessBeforeInitialization和postProcessAfterInstantiation方法。
Object bean = resolveBeforeInstantiation(beanName, mbdToUse);
if (bean != null) {
return bean;
}
Object beanInstance = doCreateBean(beanName, mbdToUse, args);
return beanInstance;
继续:
protected Object resolveBeforeInstantiation(String beanName, RootBeanDefinition mbd) {
Object bean = null;
if (!Boolean.FALSE.equals(mbd.beforeInstantiationResolved)) {
// Make sure bean class is actually resolved at this point.
if (!mbd.isSynthetic() && hasInstantiationAwareBeanPostProcessors()) {
Class<?> targetType = determineTargetType(beanName, mbd);
if (targetType != null) {
bean = applyBeanPostProcessorsBeforeInstantiation(targetType, beanName);
if (bean != null) {
bean = applyBeanPostProcessorsAfterInitialization(bean, beanName);
}
}
}
mbd.beforeInstantiationResolved = (bean != null);
}
return bean;
}
从这里可以看出,如果InstantiationAwareBeanPostProcessor返回的不是空,那么将不会继续执行剩下的Spring初始化流程,此接口用于初始化自定义的bean,主要是在Spring内部使用。
doCreateBean
同样分为几部分。
创建(createBeanInstance)
关键代码:
BeanWrapper instanceWrapper = null;
if (instanceWrapper == null) {
instanceWrapper = createBeanInstance(beanName, mbd, args);
}
createBeanInstance的创建过程又分为以下几种情况:
-
工厂bean:
调用instantiateUsingFactoryMethod方法:
protected BeanWrapper instantiateUsingFactoryMethod( String beanName, RootBeanDefinition mbd, Object[] explicitArgs) { return new ConstructorResolver(this).instantiateUsingFactoryMethod(beanName, mbd, explicitArgs); }
注意,此处的工厂bean指的是配置了factory-bean/factory-method属性的bean,不是实现了FacrotyBean接口的bean。如果没有配置factory-bean属性,那么factory-method指向的方法必须是静态的。此方法主要做了这么几件事:
-
初始化一个BeanWrapperImpl对象。
-
根据设置的参数列表使用反射的方法寻找相应的方法对象。
-
InstantiationStrategy:
bean的初始化在此处又抽成了策略模式,类图:
instantiateUsingFactoryMethod部分源码:
beanInstance = this.beanFactory.getInstantiationStrategy().instantiate( mbd, beanName, this.beanFactory, factoryBean, factoryMethodToUse, argsToUse);
getInstantiationStrategy返回的是CglibSubclassingInstantiationStrategy对象。此处instantiate实现也很简单,就是调用工厂方法的Method对象反射调用其invoke即可得到对象,SimpleInstantiationStrategy.
instantiate核心源码:
@Override public Object instantiate(RootBeanDefinition bd, String beanName, BeanFactory owner, Object factoryBean, final Method factoryMethod, Object... args) { return factoryMethod.invoke(factoryBean, args); }
-
-
构造器自动装配
createBeanInstance部分源码:
// Need to determine the constructor... Constructor<?>[] ctors = determineConstructorsFromBeanPostProcessors(beanClass, beanName); if (ctors != null || mbd.getResolvedAutowireMode() == RootBeanDefinition.AUTOWIRE_CONSTRUCTOR || //配置了<constructor-arg>子元素 mbd.hasConstructorArgumentValues() || !ObjectUtils.isEmpty(args)) { return autowireConstructor(beanName, mbd, ctors, args); }
determineConstructorsFromBeanPostProcessors源码:
protected Constructor<?>[] determineConstructorsFromBeanPostProcessors(Class<?> beanClass, String beanName) { if (beanClass != null && hasInstantiationAwareBeanPostProcessors()) { for (BeanPostProcessor bp : getBeanPostProcessors()) { if (bp instanceof SmartInstantiationAwareBeanPostProcessor) { SmartInstantiationAwareBeanPostProcessor ibp = (SmartInstantiationAwareBeanPostProcessor) bp; Constructor<?>[] ctors = ibp.determineCandidateConstructors(beanClass, beanName); if (ctors != null) { return ctors; } } } } return null; }
可见是由SmartInstantiationAwareBeanPostProcessor决定的,默认是没有配置这种东西的。
之后就是判断bean的自动装配模式,可以通过如下方式配置:
<bean id="student" class="base.Student" primary="true" autowire="default" />
autowire共有以下几种选项:
- no: 默认的,不进行自动装配。在这种情况下,只能通过ref方式引用其它bean。
- byName: 根据bean里面属性的名字在BeanFactory中进行查找并装配。
- byType: 按类型。
- constructor: 以byType的方式查找bean的构造参数列表。
- default: 由父bean决定。
参考: Spring - bean的autowire属性(自动装配)
autowireConstructor调用的是ConstructorResolver.autowireConstructor,此方法主要做了两件事:
-
得到合适的构造器对象。
-
根据构造器参数的类型去BeanFactory查找相应的bean:
入口方法在ConstructorResolver.resolveAutowiredArgument:
protected Object resolveAutowiredArgument( MethodParameter param, String beanName, Set<String> autowiredBeanNames, TypeConverter typeConverter) { return this.beanFactory.resolveDependency( new DependencyDescriptor(param, true), beanName, autowiredBeanNames, typeConverter); }
最终调用的还是CglibSubclassingInstantiationStrategy.instantiate方法,关键源码:
@Override public Object instantiate(RootBeanDefinition bd, String beanName, BeanFactory owner, final Constructor<?> ctor, Object... args) { if (bd.getMethodOverrides().isEmpty()) { //反射调用 return BeanUtils.instantiateClass(ctor, args); } else { return instantiateWithMethodInjection(bd, beanName, owner, ctor, args); } }
可以看出,如果配置了lookup-method标签,得到的实际上是用Cglib生成的目标类的代理子类。
CglibSubclassingInstantiationStrategy.instantiateWithMethodInjection:
@Override protected Object instantiateWithMethodInjection(RootBeanDefinition bd, String beanName, BeanFactory owner,Constructor<?> ctor, Object... args) { // Must generate CGLIB subclass... return new CglibSubclassCreator(bd, owner).instantiate(ctor, args); }
-
默认构造器
一行代码,很简单:
// No special handling: simply use no-arg constructor. return instantiateBean(beanName, mbd);
MergedBeanDefinitionPostProcessor
触发源码:
synchronized (mbd.postProcessingLock) {
if (!mbd.postProcessed) {
applyMergedBeanDefinitionPostProcessors(mbd, beanType, beanName);
mbd.postProcessed = true;
}
}
此接口也是Spring内部使用的,不管它了。
属性解析
入口方法: AbstractAutowireCapableBeanFactory.populateBean,它的作用是: 根据autowire类型进行autowire by name,by type 或者是直接进行设置,简略后的源码:
protected void populateBean(String beanName, RootBeanDefinition mbd, BeanWrapper bw) {
//所有<property>的值
PropertyValues pvs = mbd.getPropertyValues();
if (mbd.getResolvedAutowireMode() == RootBeanDefinition.AUTOWIRE_BY_NAME ||
mbd.getResolvedAutowireMode() == RootBeanDefinition.AUTOWIRE_BY_TYPE) {
MutablePropertyValues newPvs = new MutablePropertyValues(pvs);
// Add property values based on autowire by name if applicable.
if (mbd.getResolvedAutowireMode() == RootBeanDefinition.AUTOWIRE_BY_NAME) {
autowireByName(beanName, mbd, bw, newPvs);
}
// Add property values based on autowire by type if applicable.
if (mbd.getResolvedAutowireMode() == RootBeanDefinition.AUTOWIRE_BY_TYPE) {
autowireByType(beanName, mbd, bw, newPvs);
}
pvs = newPvs;
}
//设值
applyPropertyValues(beanName, mbd, bw, pvs);
}
autowireByName源码:
protected void autowireByName(
String beanName, AbstractBeanDefinition mbd, BeanWrapper bw, MutablePropertyValues pvs) {
//返回所有引用(ref="XXX")的bean名称
String[] propertyNames = unsatisfiedNonSimpleProperties(mbd, bw);
for (String propertyName : propertyNames) {
if (containsBean(propertyName)) {
//从BeanFactory获取
Object bean = getBean(propertyName);
pvs.add(propertyName, bean);
registerDependentBean(propertyName, beanName);
}
}
}
autowireByType也是同样的套路,所以可以得出结论: autowireByName和autowireByType方法只是先获取到引用的bean,真正的设值是在applyPropertyValues中进行的。
属性设置
Spring判断一个属性可不可以被设置(存不存在)是通过java bean的内省操作来完成的,也就是说,属性可以被设置的条件是此属性拥有public的setter方法,并且注入时的属性名应该是setter的名字。
初始化
此处的初始化指的是bean已经构造完成,执行诸如调用其init方法的操作。相关源码:
// Initialize the bean instance.
Object exposedObject = bean;
try {
populateBean(beanName, mbd, instanceWrapper);
if (exposedObject != null) {
exposedObject = initializeBean(beanName, exposedObject, mbd);
}
}
initializeBean:
protected Object initializeBean(final String beanName, final Object bean, RootBeanDefinition mbd) {
if (System.getSecurityManager() != null) {
AccessController.doPrivileged(new PrivilegedAction<Object>() {
@Override
public Object run() {
invokeAwareMethods(beanName, bean);
return null;
}
}, getAccessControlContext());
}
else {
invokeAwareMethods(beanName, bean);
}
Object wrappedBean = bean;
if (mbd == null || !mbd.isSynthetic()) {
wrappedBean = applyBeanPostProcessorsBeforeInitialization(wrappedBean, beanName);
}
invokeInitMethods(beanName, wrappedBean, mbd);
if (mbd == null || !mbd.isSynthetic()) {
wrappedBean = applyBeanPostProcessorsAfterInitialization(wrappedBean, beanName);
}
return wrappedBean;
}
主要的操作步骤一目了然。
-
Aware方法触发:
我们的bean有可能实现了一些XXXAware接口,此处就是负责调用它们:
private void invokeAwareMethods(final String beanName, final Object bean) { if (bean instanceof Aware) { if (bean instanceof BeanNameAware) { ((BeanNameAware) bean).setBeanName(beanName); } if (bean instanceof BeanClassLoaderAware) { ((BeanClassLoaderAware) bean).setBeanClassLoader(getBeanClassLoader()); } if (bean instanceof BeanFactoryAware) { ((BeanFactoryAware) bean).setBeanFactory(AbstractAutowireCapableBeanFactory.this); } } }
-
BeanPostProcessor触发,没什么好说的
-
调用init方法:
在XML配置中,bean可以有一个init-method属性来指定初始化时调用的方法。从原理来说,其实就是一个反射调用。不过注意这里有一个InitializingBean的概念。
此接口只有一个方法:
void afterPropertiesSet() throws Exception;
如果我们的bean实现了此接口,那么此方法会首先被调用。此接口的意义在于: 当此bean的所有属性都被设置(注入)后,给bean一个利用现有属性重新组织或是检查属性的机会。感觉和init方法有些冲突,不过此接口在Spring被广泛使用。
getObjectForBeanInstance
位于AbstractBeanFactory,此方法的目的在于如果bean是FactoryBean,那么返回其工厂方法创建的bean,而不是自身。
Prototype初始化
AbstractBeanFactory.doGetBean相关源码:
else if (mbd.isPrototype()) {
// It's a prototype -> create a new instance.
Object prototypeInstance = null;
try {
beforePrototypeCreation(beanName);
prototypeInstance = createBean(beanName, mbd, args);
}
finally {
afterPrototypeCreation(beanName);
}
bean = getObjectForBeanInstance(prototypeInstance, name, beanName, mbd);
}
beforePrototypeCreation
此方法用于确保在同一时刻只能有一个此bean在初始化。
createBean
和单例的是一样的,不在赘述。
afterPrototypeCreation
和beforePrototypeCreation对应的,你懂的。
总结
可以看出,初始化其实和单例是一样的,只不过单例多了一个是否已经存在的检查。
其它Scope初始化
其它就指的是request、session。此部分源码:
else {
String scopeName = mbd.getScope();
final Scope scope = this.scopes.get(scopeName);
if (scope == null) {
throw new IllegalStateException("No Scope registered for scope name '" + scopeName + "'");
}
Object scopedInstance = scope.get(beanName, new ObjectFactory<Object>() {
@Override
public Object getObject() throws BeansException {
beforePrototypeCreation(beanName);
try {
return createBean(beanName, mbd, args);
}
finally {
afterPrototypeCreation(beanName);
}
}
});
bean = getObjectForBeanInstance(scopedInstance, name, beanName, mbd);
}
scopes是一个LinkedHashMap<String, Scope>,可以调用 ConfigurableBeanFactory定义的registerScope方法注册其值。
Scope接口继承体系:
根据socpe.get的注释,此方法如果找到了叫做beanName的bean,那么返回,如果没有,将调用ObjectFactory创建之。Scope的实现参考类图。