HttpComponents组件探究 - HttpClient篇

        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的顶级项目了，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