HttpMessageConverter主要是用于Http请求和响应的转换:
请求参数的反序列化: Controller中定义的接口参数类型是T, 但是前端传递的是json格式数据，需要借助它反序列化响应的序列化: Controller中定义的接口返回类型是T，但是最终请求响应是json格式(content-type: application/json)，需要借助它序列化调用远程请求响应的反序列化: 远程调用接口返回的响应body是json string类型的，但是我们最终能得到T，需要借助它反序列化
1. 扫描阶段
这部分主要是借助Spring boot的自动装配特性来进行MessageConverter注册，具体而言就是org.springframework.boot.autoconfigure.http.HttpMessageConvertersAutoConfiguration。从顶部@Import可以看出，默认会尝试加载Jackson、Gson等转换器。
@Import({ JacksonHttpMessageConvertersConfiguration.class, GsonHttpMessageConvertersConfiguration.class,
JsonbHttpMessageConvertersConfiguration.class })
public class HttpMessageConvertersAutoConfiguration {
@Bean
@ConditionalOnMissingBean
public HttpMessageConverters messageConverters(ObjectProvider<HttpMessageConverter<?>> converters) {
return new HttpMessageConverters(converters.orderedStream().collect(Collectors.toList()));
}
}
@Configuration(proxyBeanMethods = false)
class JacksonHttpMessageConvertersConfiguration {
@Configuration(proxyBeanMethods = false)
@ConditionalOnClass(ObjectMapper.class)
@ConditionalOnBean(ObjectMapper.class)
@ConditionalOnProperty(name = HttpMessageConvertersAutoConfiguration.PREFERRED_MAPPER_PROPERTY,
havingValue = "jackson", matchIfMissing = true)
static class MappingJackson2HttpMessageConverterConfiguration {
@Bean
@ConditionalOnMissingBean(value = MappingJackson2HttpMessageConverter.class,
ignoredType = {
"org.springframework.hateoas.server.mvc.TypeConstrainedMappingJackson2HttpMessageConverter",
"org.springframework.data.rest.webmvc.alps.AlpsJsonHttpMessageConverter" })
MappingJackson2HttpMessageConverter mappingJackson2HttpMessageConverter(ObjectMapper objectMapper) {
return new MappingJackson2HttpMessageConverter(objectMapper);
}
}
}
在初始化阶段会去扫描HttpMessageConverter实现类Bean，如StringHttpMessageConverter、MappingJackson2HttpMessageConverter等。
这里更为关键的是自动装配会暴露很多HttpMessageConverter类型的Bean，那么Spring boot是如何整合它们以及确保调用的顺序的呢？
public class HttpMessageConvertersAutoConfiguration {
@Bean
@ConditionalOnMissingBean
// class org.springframework.beans.factory.support.DefaultListableBeanFactory$DependencyObjectProvider
// 初始化的并没有实际 去处理 HttpMessageConverter， 注意底下的stream
public HttpMessageConverters messageConverters(ObjectProvider<HttpMessageConverter<?>> converters) {
return new HttpMessageConverters(converters.orderedStream().collect(Collectors.toList()));
}
}
答案就是包装在HttpMessageConverters中的时候，会通过ObjectProvider进行排序，也就是上面的orderedStream对应的逻辑。 如果是PriorityOrdered排在最前面，然后是Ordered Bean，优先级相同任意排列，最后是没有设置优先级的。具体逻辑参考org.springframework.core.OrderComparator
2. HttpMessageConverter的调用
@RequestMapping(value = "/api", method = RequestMethod.POST)
public BaseResultData<List<JavaModel>> queryModelList(@RequestBody JavaModel request) throws Exception {
}
{
"id": "1234",
"modelSource": "web"
}
2.1 请求Body的反序列化，典型的如json string to T
针对上面的示例接口以及示例请求Body，在org.springframework.web.servlet.mvc.method.annotation.AbstractMessageConverterMethodArgumentResolver.readWithMessageConverters方法中，会去按照注册的HttpMessageConverter尝试去反解响应，如果可以，则转换相应的Model，也就是上面的JavaModel。
/**
源码注释的这个地方也解释的很清楚了，就是基于InputMessage解析成对应类型的 方法参数
* Create the method argument value of the expected parameter type by reading from the given HttpInputMessage.
* @param <T> the expected type of the argument value to be created
* @param inputMessage the HTTP input message representing the current request
*/
@Nullable
protected <T> Object readWithMessageConverters(HttpInputMessage inputMessage, MethodParameter parameter,
Type targetType) throws IOException, HttpMediaTypeNotSupportedException, HttpMessageNotReadableException {
Object body = NO_VALUE;
message = new EmptyBodyCheckingHttpInputMessage(inputMessage);
for (HttpMessageConverter<?> converter : this.messageConverters) {
// 是否可以解析
if (genericConverter != null ? genericConverter.canRead(targetType, contextClass, contentType) :
(targetClass != null && converter.canRead(targetClass, contentType))) {
if (message.hasBody()) {
HttpInputMessage msgToUse =
getAdvice().beforeBodyRead(message, parameter, targetType, converterType);
body = (genericConverter != null ? genericConverter.read(targetType, contextClass, msgToUse) :
((HttpMessageConverter<T>) converter).read(targetClass, msgToUse));
body = getAdvice().afterBodyRead(body, msgToUse, parameter, targetType, converterType);
}
}
}
}
是否可以解析的核心就是，支持的MediaType是否包括当前请求header中对应的content-type。
2.2 响应的序列化 典型的如T to json string
当要返回响应的时候，需要经历一个反向过程，也就是序列化过程。org.springframework.web.servlet.mvc.method.annotation.AbstractMessageConverterMethodProcessor.writeWithMessageConverters方法中，会去按照注册的HttpMessageConverter尝试去序列化相应的model，也就是上面的JavaModel - BaseResultData<List<JavaModel>>。
/**
Writes the given return type to the given output message.
* @param value the value to write to the output message
* @param returnType the type of the value
* @param inputMessage the input messages. Used to inspect the {@code Accept} header.
* @param outputMessage the output message to write to
*/
protected <T> void writeWithMessageConverters(@Nullable T value, MethodParameter returnType,
ServletServerHttpRequest inputMessage, ServletServerHttpResponse outputMessage)
throws IOException, HttpMediaTypeNotAcceptableException, HttpMessageNotWritableException {
for (HttpMessageConverter<?> converter : this.messageConverters) {
if (genericConverter != null ?
((GenericHttpMessageConverter) converter).canWrite(targetType, valueType, selectedMediaType) :
converter.canWrite(valueType, selectedMediaType)) {
genericConverter.write(body, targetType, selectedMediaType, outputMessage);
}
}
}
2.3 远程调用的响应转换
org.springframework.web.client.HttpMessageConverterExtractor实现了该功能，底层还是借助注册的HttpMessageConverter实现。
/*
* Response extractor that uses the given {@linkplain HttpMessageConverter entity converters}
* to convert the response into a type {@code T}.
*/
public class HttpMessageConverterExtractor<T> implements ResponseExtractor<T> {
public T extractData(ClientHttpResponse response) throws IOException {
for (HttpMessageConverter<?> messageConverter : this.messageConverters) {
...
if (genericMessageConverter.canRead(this.responseType, null, contentType)) {
genericMessageConverter.read
}
...
}
}
}