在Java领域，谈到网络编程，可能大家脑海里第一反应就是MINA，NETTY，GRIZZLY等优秀的开源框架。没错，不过在深入探究这些框架之前，我们需要先从最original的技术探究开始（当然，需要大家先熟悉java.net.*类库）。这里，我要和大家分享一下HttpComponents项目的部分组件特性。HttpClient，想必大家早都接触过了吧。HttpComponents和HttpClient的”血缘“有点像guava和google-collection的关系。目前，HttpComponents已经是Apache的*项目了，它旨在为我们提供一个Http协议相关的Java平台工具集。它的代码组织很精妙，主要分两部分，一部分是核心工具集（包括HttpCore-bio,HttpCore-nio，HttpClient，HttpMIme，HttpCookie等），一部分是扩展工具集（目前主要包括ssl）
HttpClient主要包括Connection management，Status management，Authentication Management三部分。下面给出对它的二次封装，经过了线上的接近半年的验证（这里指的是httpClient 3,httpClient 4还有待检验），可以看做是一个高性能的Client封装吧。感兴趣的朋友可以根据apache的MPM IO模型进行部分参数的调整。
先来段httpClient 4的封装,代码如下:
/**
* @author von gosling 2012-3-2
*/
public class HttpComponentsClientExecutor implements DisposableBean {
private static final int DEFAULT_MAX_TOTAL_CONNECTIONS = 100; private static final int DEFAULT_MAX_CONNECTIONS_PER_ROUTE = 5; //notice IE 6,7,8 private static final int DEFAULT_CONN_TIMEOUT_MILLISECONDS = 5 * 1000; private static final int DEFAULT_READ_TIMEOUT_MILLISECONDS = 60 * 1000; private static final String HTTP_HEADER_CONTENT_ENCODING = "Content-Encoding";
private static final String ENCODING_GZIP = "gzip"; private HttpClient httpClient; /**
* Create a new instance of the HttpComponentsClient with a default
* {@link HttpClient} that uses a default
* {@link org.apache.http.impl.conn.tsccm.ThreadSafeClientConnManager}.
*/
public HttpComponentsClientExecutor() {
SchemeRegistry schemeRegistry = new SchemeRegistry();
schemeRegistry.register(new Scheme("http", 80, PlainSocketFactory.getSocketFactory()));
schemeRegistry.register(new Scheme("https", 443, SSLSocketFactory.getSocketFactory())); ThreadSafeClientConnManager connectionManager = new ThreadSafeClientConnManager(
schemeRegistry);
connectionManager.setMaxTotal(DEFAULT_MAX_TOTAL_CONNECTIONS);
connectionManager.setDefaultMaxPerRoute(DEFAULT_MAX_CONNECTIONS_PER_ROUTE);
this.httpClient = new DefaultHttpClient(connectionManager); setConnectTimeout(DEFAULT_CONN_TIMEOUT_MILLISECONDS);
setReadTimeout(DEFAULT_READ_TIMEOUT_MILLISECONDS);
} /**
* Create a new instance of the HttpComponentsClient with the given
* {@link HttpClient} instance.
*
* @param httpClient the HttpClient instance to use for this request
*/
public HttpComponentsClientExecutor(HttpClient httpClient) {
Validate.notNull(httpClient, "HttpClient must not be null");
//notice: if you want to custom exception recovery mechanism
//you should provide an implementation of the HttpRequestRetryHandler interface.
this.httpClient = httpClient;
} /**
* Set the {@code HttpClient} used by this request.
*/
public void setHttpClient(HttpClient httpClient) {
this.httpClient = httpClient;
} /**
* Return the {@code HttpClient} used by this request.
*/
public HttpClient getHttpClient() {
return this.httpClient;
} /**
* Set the connection timeout for the underlying HttpClient. A timeout value
* of 0 specifies an infinite timeout.
*
* @param timeout the timeout value in milliseconds
*/
public void setConnectTimeout(int timeout) {
Validate.isTrue(timeout >= 0, "Timeout must be a non-negative value");
getHttpClient().getParams().setIntParameter(CoreConnectionPNames.CONNECTION_TIMEOUT,
timeout);
} /**
* Set the socket timeout (SO_TIMEOUT) in milliseconds, which is the timeout
* for waiting for data or, put differently, a maximum period inactivity
* between two consecutive data packets.A timeout value of 0 specifies an
* infinite timeout.
*
* @param timeout the timeout value in milliseconds
*/
public void setReadTimeout(int timeout) {
Validate.isTrue(timeout >= 0, "Timeout must be a non-negative value");
getHttpClient().getParams().setIntParameter(CoreConnectionPNames.SO_TIMEOUT, timeout);
} /**
* Create a Commons HttpMethodBase object for the given HTTP method and URI
* specification.
*
* @param httpMethod the HTTP method
* @param uri the URI
* @return the Commons HttpMethodBase object
*/
protected HttpUriRequest createHttpUriRequest(HttpMethod httpMethod, URI uri) {
switch (httpMethod) {
case GET:
return new HttpGet(uri);
case DELETE:
return new HttpDelete(uri);
case HEAD:
return new HttpHead(uri);
case OPTIONS:
return new HttpOptions(uri);
case POST:
return new HttpPost(uri);
case PUT:
return new HttpPut(uri);
case TRACE:
return new HttpTrace(uri);
default:
throw new IllegalArgumentException("Invalid HTTP method: " + httpMethod);
}
} /**
* Execute the given method on the provided URI.
*
* @param method the HTTP method to execute (GET, POST, etc.)
* @param url the fully-expanded URL to connect to
* @param responseHandler httpClient will automatically take care of
* ensuring release of the connection back to the connection
* manager regardless whether the request execution succeeds or
* causes an exception,if using this response handler
* @return an response object's string representation
* @throws IOException
* @throws ClientProtocolException
*/
public String doExecuteRequest(HttpMethod httpMethod, URI uri,
ResponseHandler<String> responseHandler)
throws ClientProtocolException, IOException {
return httpClient.execute(createHttpUriRequest(httpMethod, uri), responseHandler);
} public InputStream doExecuteRequest(HttpMethod httpMethod, URI uri)
throws ClientProtocolException, IOException {
//1.
HttpUriRequest httpUriRequest = createHttpUriRequest(httpMethod, uri);
//2.
HttpResponse response = httpClient.execute(httpUriRequest);
//3.
validateResponse(response);
//4.
return getResponseBody(response);
} /**
* Validate the given response, throwing an exception if it does not
* correspond to a successful HTTP response.
* <p>
* Default implementation rejects any HTTP status code beyond 2xx, to avoid
* parsing the response body and trying to deserialize from a corrupted
* stream.
*
* @param config the HTTP invoker configuration that specifies the target
* service
* @param response the resulting HttpResponse to validate
* @throws NoHttpResponseException
* @throws java.io.IOException if validation failed
*/
protected void validateResponse(HttpResponse response) throws IOException { StatusLine status = response.getStatusLine();
if (status.getStatusCode() >= 300) {
throw new NoHttpResponseException(
"Did not receive successful HTTP response: status code = "
+ status.getStatusCode() + ", status message = ["
+ status.getReasonPhrase() + "]");
}
} /**
* Extract the response body
* <p>
* The default implementation simply fetches the response body stream. If
* the response is recognized as GZIP response, the InputStream will get
* wrapped in a GZIPInputStream.
*
* @param httpResponse the resulting HttpResponse to read the response body
* from
* @return an InputStream for the response body
* @throws java.io.IOException if thrown by I/O methods
* @see #isGzipResponse
* @see java.util.zip.GZIPInputStream
*/
protected InputStream getResponseBody(HttpResponse httpResponse) throws IOException { if (isGzipResponse(httpResponse)) {
return new GZIPInputStream(httpResponse.getEntity().getContent());
} else {
return httpResponse.getEntity().getContent();
}
} /**
* Determine whether the given response indicates a GZIP response.
* <p>
* The default implementation checks whether the HTTP "Content-Encoding"
* header contains "gzip" (in any casing).
*
* @param httpResponse the resulting HttpResponse to check
* @return whether the given response indicates a GZIP response
*/
protected boolean isGzipResponse(HttpResponse httpResponse) {
Header encodingHeader = httpResponse.getFirstHeader(HTTP_HEADER_CONTENT_ENCODING);
return (encodingHeader != null && encodingHeader.getValue() != null && encodingHeader
.getValue().toLowerCase().contains(ENCODING_GZIP));
} /**
* Shutdown hook that closes the underlying
* {@link org.apache.http.conn.ClientConnectionManager
* ClientConnectionManager}'s connection pool, if any.
*/
public void destroy() {
getHttpClient().getConnectionManager().shutdown();
} enum HttpMethod {
GET,
POST,
HEAD,
OPTIONS,
PUT,
DELETE,
TRACE
}
}
下面是久经考验的httpClient 3的二次封装,如下:
/**
* @author von gosling 2011-12-12
*/
public class HttpClientUtils { private static final Logger log = LoggerFactory
.getLogger(HttpClientUtils.class); private static int timeOut = 100;
private static int retryCount = 1;
private static int connectionTimeout = 100;
private static int maxHostConnections = 32; //根据apache work MPM设置此值
private static int maxTotalConnections = 512; //同上
private static String charsetName = "UTF-8"; public static JSONObject executeMethod(HttpClient httpClient, HttpMethod method) { JSONObject result = new JSONObject();
StopWatch watch = new StopWatch();
int status = -1;
try {
log.info("Execute method({}) begin...", method.getURI()); watch.start();
status = httpClient.executeMethod(method);
watch.stop(); if (status == HttpStatus.SC_OK) {
InputStream inputStream = method.getResponseBodyAsStream();
ByteArrayOutputStream baos = new ByteArrayOutputStream();
IOUtils.copy(inputStream, baos);
String response = new String(baos.toByteArray(), charsetName); log.info("Response is:{}", response); result = JSONObject.parseObject(response);
} else {
log.error("Http request failure! status is {}", status);
}
} catch (SocketTimeoutException e) {
log.error("Request time out!");//只关注请求超时,对于其它两类超时,使用通用异常捕获
} catch (Exception e) {
log.error("Error occur!", e);
} finally {
method.releaseConnection();
log.info("Method {},statusCode {},consuming {} ms", new Object[] { method.getName(),
status, watch.getTime() });
}
return result;
} /**
* @param uri
* @param nameValuePairs
* @return
*/
public static PostMethod createPostMethod(String uri, NameValuePair[] nameValuePairs) {
PostMethod method = new PostMethod(uri);
method.addParameters(nameValuePairs);
method.getParams().setContentCharset(charsetName);
return method;
} /**
* @param uri
* @param nameValuePairs
* @return
*/
public static GetMethod createGetMethod(String uri, NameValuePair[] nameValuePairs) {
GetMethod method = new GetMethod(uri);
List<NameValuePair> list = Lists.newArrayList();
if (nameValuePairs != null) {
Collections.addAll(list, nameValuePairs);
method.setQueryString(list.toArray(new NameValuePair[nameValuePairs.length]));
}
method.getParams().setContentCharset(charsetName);
return method;
} public static HttpClient createHttpClient() {
//1.
HttpClient httpClient = new HttpClient(new MultiThreadedHttpConnectionManager()); //2.
HttpConnectionManagerParams httpConnectionManagerParams = httpClient
.getHttpConnectionManager().getParams();
httpConnectionManagerParams.setConnectionTimeout(connectionTimeout);
httpConnectionManagerParams.setTcpNoDelay(true);//Nagle's algorithm
httpConnectionManagerParams.setSoTimeout(timeOut);
httpConnectionManagerParams.setDefaultMaxConnectionsPerHost(maxHostConnections);
httpConnectionManagerParams.setMaxTotalConnections(maxTotalConnections); //3.
HttpClientParams httpClientParam = httpClient.getParams();
//httpClientParam.setConnectionManagerTimeout(connectionTimeout);//暂且不关注这个超时设置,后面根据性能酌情考虑
httpClientParam.setParameter(HttpMethodParams.RETRY_HANDLER,
new DefaultHttpMethodRetryHandler(retryCount, false));
httpClientParam.setCookiePolicy(CookiePolicy.BROWSER_COMPATIBILITY); return httpClient;
} public static JSONObject doGet(String url, NameValuePair[] params) {
return executeMethod(createHttpClient(), createGetMethod(url, params));
} public static JSONObject doPost(String url, NameValuePair[] params) {
return executeMethod(createHttpClient(), createPostMethod(url, params));
} protected HttpClientUtils() { } public void setTimeOut(int timeOut) {
HttpClientUtils.timeOut = timeOut;
} public static int getTimeOut() {
return timeOut;
} public static int getRetryCount() {
return retryCount;
} public void setRetryCount(int retryCount) {
HttpClientUtils.retryCount = retryCount;
} public static int getConnectionTimeout() {
return connectionTimeout;
} public void setConnectionTimeout(int connectionTimeout) {
HttpClientUtils.connectionTimeout = connectionTimeout;
} public static int getMaxHostConnections() {
return maxHostConnections;
} public void setMaxHostConnections(int maxHostConnections) {
HttpClientUtils.maxHostConnections = maxHostConnections;
} public static int getMaxTotalConnections() {
return maxTotalConnections;
} public void setMaxTotalConnections(int maxTotalConnections) {
HttpClientUtils.maxTotalConnections = maxTotalConnections;
} public static String getCharsetName() {
return charsetName;
} public void setCharsetName(String charsetName) {
HttpClientUtils.charsetName = charsetName;
}
}
好了，有了活生生的代码，我们来总结一下httpClient封装过程中需要注意的一些事项吧。恩，其实更多的是体现在安全，性能上面：
（1）多线程模型，尤其注意finally中collection的释放问题。除此之外，需要考虑池化连接的异常处理，这是我文中提到特别注意的三大异常之一；
（2）Retry机制中对幂等性的处理。尤其是在httpClient4中，put和post操作，未按照http规范行事，需要我们额外注意；
（3）SSL、TLS的定制化处理；
（4）并发标记的处理，这里使用了Concurrency in practice中的并发annotation，有什么用？感兴趣的朋友可以了解下SureLogic（http://www.surelogic.com/concurrency-tools.html），别问我要license，因为俺也不是apache开源社区的developer呀；
（5）拦截器对header的处理；
（6）collection stale check机制；
（7）Cookie specification choose或者是自定义实现；
恩，今天就写到这里吧。感谢大家的阅读，如果哪里有疑问，欢迎留言~
参考文献：
1.http://www.w3.org/Protocols/rfc2616/rfc2616-sec7.html
2.http://hc.apache.org/httpcomponents-client-ga/tutorial/pdf/httpclient-tutorial.pdf