Android高工面试被问OKHTTP内核解析，我慌了！

基本使用

从使用方法出发，首先是怎么使用，其次是我们使用的功能在内部是如何实现的.建议大家下载 OkHttp 源码之后，跟着本文，过一遍源码。

官方博客栗子：http://square.github.io/okhttp/#examples

1. OkHttpClient client = new OkHttpClient();

2. String run(String url) throws IOException {

3. Request request = new Request.Builder()

4. .url(url)

5. .build();

6. Response response = client.newCall(request).execute();

7. return response.body().string();

8. }

Request、Response、Call 基本概念

上面的代码中涉及到几个常用的类：Request、Response和Call。下面分别介绍：

Request

每一个HTTP请求包含一个URL、一个方法（GET或POST或其他）、一些HTTP头。请求还可能包含一个特定内容类型的数据类的主体部分。

Response

响应是对请求的回复，包含状态码、HTTP头和主体部分。

Call

OkHttp使用Call抽象出一个满足请求的模型，尽管中间可能会有多个请求或响应。执行Call有两种方式，同步或异步

第一步：创建 OkHttpClient对象,进行源码分析：

1. OkHttpClient client = new OkHttpClient();

通过okhttp源码分析,直接创建的 OkHttpClient对象并且默认构造builder对象进行初始化

1. public class OkHttpClient implements Cloneable, Call.Factory, WebSocket.Factory {

2. public OkHttpClient() {

3. this(new Builder());

4. }

5. OkHttpClient(Builder builder) {

6. this.dispatcher = builder.dispatcher;

7. this.proxy = builder.proxy;

8. this.protocols = builder.protocols;

9. this.connectionSpecs = builder.connectionSpecs;

10. this.interceptors = Util.immutableList(builder.interceptors);

11. this.networkInterceptors = Util.immutableList(builder.networkInterceptors);

12. this.eventListenerFactory = builder.eventListenerFactory;

13. this.proxySelector = builder.proxySelector;

14. this.cookieJar = builder.cookieJar;

15. this.cache = builder.cache;

16. this.internalCache = builder.internalCache;

17. this.socketFactory = builder.socketFactory;

18. boolean isTLS = false;

19. ......

20. this.hostnameVerifier = builder.hostnameVerifier;

21. this.certificatePinner = builder.certificatePinner.withCertificateChainCleaner(

22. certificateChainCleaner);

23. this.proxyAuthenticator = builder.proxyAuthenticator;

24. this.authenticator = builder.authenticator;

25. this.connectionPool = builder.connectionPool;

26. this.dns = builder.dns;

27. this.followSslRedirects = builder.followSslRedirects;

28. this.followRedirects = builder.followRedirects;

29. this.retryOnConnectionFailure = builder.retryOnConnectionFailure;

30. this.connectTimeout = builder.connectTimeout;

31. this.readTimeout = builder.readTimeout;

32. this.writeTimeout = builder.writeTimeout;

33. this.pingInterval = builder.pingInterval;

34. }

35. }

第二步：接下来发起 HTTP 请求

1. Request request = new Request.Builder().url("url").build();

2. okHttpClient.newCall(request).enqueue(new Callback() {

3. @Override

4. public void onFailure(Call call, IOException e) {

5. }

6. @Override

7. public void onResponse(Call call, Response response) throws IOException {

8. }

9. });

第二步：代码流程分析：

1. Request request = new Request.Builder().url("url").build();

初始化构建者模式和请求对象，并且用URL替换Web套接字URL。

1. public final class Request {

2. public Builder() {

3. this.method = "GET";

4. this.headers = new Headers.Builder();

5. }

6. public Builder url(String url) {

7. ......

8. // Silently replace web socket URLs with HTTP URLs.

9. if (url.regionMatches(true, 0, "ws:", 0, 3)) {

10. url = "http:" + url.substring(3);

11. } else if (url.regionMatches(true, 0, "wss:", 0, 4)) {

12. url = "https:" + url.substring(4);

13. }

14. HttpUrl parsed = HttpUrl.parse(url);

15. ......

16. return url(parsed);

17. }

18. public Request build() {

19. ......

20. return new Request(this);

21. }

22. }

第三步：方法解析：

1. okHttpClient.newCall(request).enqueue(new Callback() {

2. @Override

3. public void onFailure(Call call, IOException e) {

4. }

5. @Override

6. public void onResponse(Call call, Response response) throws IOException {

7. }

8. });

源码分析：

1. public class OkHttpClient implements Cloneable, Call.Factory, WebSocket.Factory {

2. @Override

3. public Call newCall(Request request) {

4. return new RealCall(this, request, false /* for web socket */);

5. }

6. }

RealCall实现了Call.Factory接口创建了一个RealCall的实例，而RealCall是Call接口的实现。

异步请求的执行流程

1. final class RealCall implements Call {
2. @Override
3. public void enqueue(Callback responseCallback) {
4. synchronized (this) {
5. if (executed) throw new IllegalStateException("Already Executed");
6. executed = true;
7. }
8. captureCallStackTrace();
9. client.dispatcher().enqueue(new AsyncCall(responseCallback));
10. }
11. }

由以上源码得知：

1） 检查这个 call 是否已经被执行了，每个 call 只能被执行一次，如果想要一个完全一样的 call，可以利用 call#clone 方法进行克隆。

2）利用 client.dispatcher().enqueue(this) 来进行实际执行，dispatcher 是刚才看到的 OkHttpClient.Builder 的成员之一

3）AsyncCall是RealCall的一个内部类并且继承NamedRunnable，那么首先看NamedRunnable类是什么样的，如下：

1. public abstract class NamedRunnable implements Runnable {

2. ......

3. @Override

4. public final void run() {

5. ......

6. try {

7. execute();

8. }

9. ......

10. }

11. protected abstract void execute();

12. }

可以看到NamedRunnable实现了Runnbale接口并且是个抽象类，其抽象方法是execute()，该方法是在run方法中被调用的，这也就意味着NamedRunnable是一个任务，并且其子类应该实现execute方法。下面再看AsyncCall的实现：

1. final class AsyncCall extends NamedRunnable {

2. private final Callback responseCallback;

3. AsyncCall(Callback responseCallback) {

4. super("OkHttp %s", redactedUrl());

5. this.responseCallback = responseCallback;

6. }

7. ......

8. final class RealCall implements Call {

9. @Override protected void execute() {

10. boolean signalledCallback = false;

11. try {

12. Response response = getResponseWithInterceptorChain();

13. if (retryAndFollowUpInterceptor.isCanceled()) {

14. signalledCallback = true;

15. responseCallback.onFailure(RealCall.this, new IOException("Canceled"));

16. } else {

17. signalledCallback = true;

18. responseCallback.onResponse(RealCall.this, response);

19. }

20. } catch (IOException e) {

21. ......

22. responseCallback.onFailure(RealCall.this, e);

23. } finally {

24. client.dispatcher().finished(this);

25. }

26. }

AsyncCall实现了execute方法，首先是调用getResponseWithInterceptorChain()方法获取响应，然后获取成功后，就调用回调的onReponse方法，如果失败，就调用回调的onFailure方法。最后，调用Dispatcher的finished方法。

关键代码：

responseCallback.onFailure(RealCall.this, new IOException(“Canceled”));

和

responseCallback.onResponse(RealCall.this, response);

走完这两句代码会进行回调到刚刚我们初始化Okhttp的地方,如下：

1. okHttpClient.newCall(request).enqueue(new Callback() {

2. @Override

3. public void onFailure(Call call, IOException e) {

4. }

5. @Override

6. public void onResponse(Call call, Response response) throws IOException {

7. }

8. });

核心重点类Dispatcher线程池介绍

1. public final class Dispatcher {

2. /** 最大并发请求数为64 */

3. private int maxRequests = 64;

4. /** 每个主机最大请求数为5 */

5. private int maxRequestsPerHost = 5;

6. /** 线程池 */

7. private ExecutorService executorService;

8. /** 准备执行的请求 */

9. private final Deque<AsyncCall> readyAsyncCalls = new ArrayDeque<>();

10. /** 正在执行的异步请求，包含已经取消但未执行完的请求 */

11. private final Deque<AsyncCall> runningAsyncCalls = new ArrayDeque<>();

12. /** 正在执行的同步请求，包含已经取消单未执行完的请求 */

13. private final Deque<RealCall> runningSyncCalls = new ArrayDeque<>();

在OkHttp，使用如下构造了单例线程池

1. public synchronized ExecutorService executorService() {
2. if (executorService == null) {
3. executorService = new ThreadPoolExecutor(0, Integer.MAX_VALUE, 60, TimeUnit.SECONDS,
4. new SynchronousQueue<Runnable>(), Util.threadFactory("OkHttp Dispatcher", false));
5. }
6. return executorService;
7. }

构造一个线程池ExecutorService：

1. executorService = new ThreadPoolExecutor(
2. //corePoolSize 最小并发线程数,如果是0的话，空闲一段时间后所有线程将全部被销毁
3. 0,
4. //maximumPoolSize: 最大线程数，当任务进来时可以扩充的线程最大值，当大于了这个值就会根据丢弃处理机制来处理
5. Integer.MAX_VALUE,
6. //keepAliveTime: 当线程数大于corePoolSize时，多余的空闲线程的最大存活时间
7. 60,
8. //单位秒
9. TimeUnit.SECONDS,
10. //工作队列,先进先出
11. new SynchronousQueue<Runnable>(),
12. //单个线程的工厂
13. Util.threadFactory("OkHttp Dispatcher", false));

可以看出，在Okhttp中，构建了一个核心为[0, Integer.MAX_VALUE]的线程池，它不保留任何最小线程数，随时创建更多的线程数，当线程空闲时只能活60秒，它使用了一个不存储元素的阻塞工作队列，一个叫做”OkHttp Dispatcher”的线程工厂。

也就是说，在实际运行中，当收到10个并发请求时，线程池会创建十个线程，当工作完成后，线程池会在60s后相继关闭所有线程。

1. synchronized void enqueue(AsyncCall call) {
2. if (runningAsyncCalls.size() < maxRequests && runningCallsForHost(call) < maxRequestsPerHost) {
3. runningAsyncCalls.add(call);
4. executorService().execute(call);
5. } else {
6. readyAsyncCalls.add(call);
7. }
8. }

从上述源码分析，如果当前还能执行一个并发请求，则加入 runningAsyncCalls ，立即执行，否则加入 readyAsyncCalls 队列。

Dispatcher线程池总结

1）调度线程池Disptcher实现了高并发，低阻塞的实现 2）采用Deque作为缓存，先进先出的顺序执行 3）任务在try/finally中调用了finished函数，控制任务队列的执行顺序，而不是采用锁，减少了编码复杂性提高性能

这里是分析OkHttp源码，并不详细讲线程池原理，如对线程池不了解请参考如下链接

点我，线程池原理，在文章性能优化最后有视频对线程池原理讲解

1. try {
2. Response response = getResponseWithInterceptorChain();
3. if (retryAndFollowUpInterceptor.isCanceled()) {
4. signalledCallback = true;
5. responseCallback.onFailure(RealCall.this, new IOException("Canceled"));
6. } else {
7. signalledCallback = true;
8. responseCallback.onResponse(RealCall.this, response);
9. }
10. } finally {
11. client.dispatcher().finished(this);
12. }

当任务执行完成后，无论是否有异常，finally代码段总会被执行，也就是会调用Dispatcher的finished函数

1. void finished(AsyncCall call) {
2. finished(runningAsyncCalls, call, true);
3. }

从上面的代码可以看出，第一个参数传入的是正在运行的异步队列，第三个参数为true，下面再看有是三个参数的finished方法：

1. private <T> void finished(Deque<T> calls, T call, boolean promoteCalls) {

2. int runningCallsCount;

3. Runnable idleCallback;

4. synchronized (this) {

5. if (!calls.remove(call)) throw new AssertionError("Call wasn't in-flight!");

6. if (promoteCalls) promoteCalls();

7. runningCallsCount = runningCallsCount();

8. idleCallback = this.idleCallback;

9. }

10. if (runningCallsCount == 0 && idleCallback != null) {

11. idleCallback.run();

12. }

13. }

打开源码，发现它将正在运行的任务Call从队列runningAsyncCalls中移除后，获取运行数量判断是否进入了Idle状态,接着执行promoteCalls()函数,下面是promoteCalls()方法：

1. private void promoteCalls() {

2. if (runningAsyncCalls.size() >= maxRequests) return; // Already running max capacity.

3. if (readyAsyncCalls.isEmpty()) return; // No ready calls to promote.

4. for (Iterator<AsyncCall> i = readyAsyncCalls.iterator(); i.hasNext(); ) {

5. AsyncCall call = i.next();

6. if (runningCallsForHost(call) < maxRequestsPerHost) {

7. i.remove();

8. runningAsyncCalls.add(call);

9. executorService().execute(call);

10. }

11. if (runningAsyncCalls.size() >= maxRequests) return; // Reached max capacity.

12. }

13. }

主要就是遍历等待队列，并且需要满足同一主机的请求小于maxRequestsPerHost时，就移到运行队列中并交给线程池运行。就主动的把缓存队列向前走了一步，而没有使用互斥锁等复杂编码

核心重点getResponseWithInterceptorChain方法

1. Response getResponseWithInterceptorChain() throws IOException {

2. // Build a full stack of interceptors.

3. List<Interceptor> interceptors = new ArrayList<>();

4. interceptors.addAll(client.interceptors());

5. interceptors.add(retryAndFollowUpInterceptor);

6. interceptors.add(new BridgeInterceptor(client.cookieJar()));

7. interceptors.add(new CacheInterceptor(client.internalCache()));

8. interceptors.add(new ConnectInterceptor(client));

9. if (!forWebSocket) {

10. interceptors.addAll(client.networkInterceptors());

11. }

12. interceptors.add(new CallServerInterceptor(forWebSocket));

13. Interceptor.Chain chain = new RealInterceptorChain(

14. interceptors, null, null, null, 0, originalRequest);

15. return chain.proceed(originalRequest);

16. }

1）在配置 OkHttpClient 时设置的 interceptors； 2）负责失败重试以及重定向的 RetryAndFollowUpInterceptor； 3）负责把用户构造的请求转换为发送到服务器的请求、把服务器返回的响应转换为用户友好的响应的 BridgeInterceptor； 4）负责读取缓存直接返回、更新缓存的 CacheInterceptor； 5）负责和服务器建立连接的 ConnectInterceptor； 6）配置 OkHttpClient 时设置的 networkInterceptors； 7）负责向服务器发送请求数据、从服务器读取响应数据的 CallServerInterceptor。

OkHttp的这种拦截器链采用的是责任链模式，这样的好处是将请求的发送和处理分开，并且可以动态添加中间的处理方实现对请求的处理、短路等操作。

从上述源码得知，不管okhttp有多少拦截器最后都会走，如下方法：

1. Interceptor.Chain chain = new RealInterceptorChain(
2. interceptors, null, null, null, 0, originalRequest);
3. return chain.proceed(originalRequest);

从方法名字基本可以猜到是干嘛的，调用 chain.proceed(originalRequest); 将request传递进来，从拦截器链里拿到返回结果。那么拦截器Interceptor是干嘛的，Chain是干嘛的呢？继续往下看RealInterceptorChain

RealInterceptorChain类

下面是RealInterceptorChain的定义，该类实现了Chain接口，在getResponseWithInterceptorChain调用时好几个参数都传的null。

1. public final class RealInterceptorChain implements Interceptor.Chain {

2. public RealInterceptorChain(List<Interceptor> interceptors, StreamAllocation streamAllocation,

3. HttpCodec httpCodec, RealConnection connection, int index, Request request) {

4. this.interceptors = interceptors;

5. this.connection = connection;

6. this.streamAllocation = streamAllocation;

7. this.httpCodec = httpCodec;

8. this.index = index;

9. this.request = request;

10. }

11. ......

12. @Override

13. public Response proceed(Request request) throws IOException {

14. return proceed(request, streamAllocation, httpCodec, connection);

15. }

16. public Response proceed(Request request, StreamAllocation streamAllocation, HttpCodec httpCodec,

17. RealConnection connection) throws IOException {

18. if (index >= interceptors.size()) throw new AssertionError();

19. calls++;

20. ......

21. // Call the next interceptor in the chain.

22. RealInterceptorChain next = new RealInterceptorChain(

23. interceptors, streamAllocation, httpCodec, connection, index + 1, request);

24. Interceptor interceptor = interceptors.get(index);

25. Response response = interceptor.intercept(next);

26. ......

27. return response;

28. }

29. protected abstract void execute();

30. }

主要看proceed方法，proceed方法中判断index（此时为0）是否大于或者等于client.interceptors(List )的大小。由于httpStream为null，所以首先创建next拦截器链，主需要把索引置为index+1即可；然后获取第一个拦截器，调用其intercept方法。

Interceptor 代码如下：

1. public interface Interceptor {

2. Response intercept(Chain chain) throws IOException;

3. interface Chain {

4. Request request();

5. Response proceed(Request request) throws IOException;

6. Connection connection();

7. }

8. }

BridgeInterceptor

BridgeInterceptor从用户的请求构建网络请求，然后提交给网络，最后从网络响应中提取出用户响应。从最上面的图可以看出，BridgeInterceptor实现了适配的功能。下面是其intercept方法：

1. public final class BridgeInterceptor implements Interceptor {

2. ......

3. @Override

4. public Response intercept(Chain chain) throws IOException {

5. Request userRequest = chain.request();

6. Request.Builder requestBuilder = userRequest.newBuilder();

7. RequestBody body = userRequest.body();

8. //如果存在请求主体部分，那么需要添加Content-Type、Content-Length首部

9. if (body != null) {

10. MediaType contentType = body.contentType();

11. if (contentType != null) {

12. requestBuilder.header("Content-Type", contentType.toString());

13. }

14. long contentLength = body.contentLength();

15. if (contentLength != -1) {

16. requestBuilder.header("Content-Length", Long.toString(contentLength));

17. requestBuilder.removeHeader("Transfer-Encoding");

18. } else {

19. requestBuilder.header("Transfer-Encoding", "chunked");

20. requestBuilder.removeHeader("Content-Length");

21. }

22. }

23. if (userRequest.header("Host") == null) {

24. requestBuilder.header("Host", hostHeader(userRequest.url(), false));

25. }

26. if (userRequest.header("Connection") == null) {

27. requestBuilder.header("Connection", "Keep-Alive");

28. }

29. // If we add an "Accept-Encoding: gzip" header field we're responsible for also decompressing

30. // the transfer stream.

31. boolean transparentGzip = false;

32. if (userRequest.header("Accept-Encoding") == null && userRequest.header("Range") == null) {

33. transparentGzip = true;

34. requestBuilder.header("Accept-Encoding", "gzip");

35. }

36. List<Cookie> cookies = cookieJar.loadForRequest(userRequest.url());

37. if (!cookies.isEmpty()) {

38. requestBuilder.header("Cookie", cookieHeader(cookies));

39. }

40. if (userRequest.header("User-Agent") == null) {

41. requestBuilder.header("User-Agent", Version.userAgent());

42. }

43. Response networkResponse = chain.proceed(requestBuilder.build());

44. HttpHeaders.receiveHeaders(cookieJar, userRequest.url(), networkResponse.headers());

45. Response.Builder responseBuilder = networkResponse.newBuilder()

46. .request(userRequest);

47. if (transparentGzip

48. && "gzip".equalsIgnoreCase(networkResponse.header("Content-Encoding"))

49. && HttpHeaders.hasBody(networkResponse)) {

50. GzipSource responseBody = new GzipSource(networkResponse.body().source());

51. Headers strippedHeaders = networkResponse.headers().newBuilder()

52. .removeAll("Content-Encoding")

53. .removeAll("Content-Length")

54. .build();

55. responseBuilder.headers(strippedHeaders);

56. responseBuilder.body(new RealResponseBody(strippedHeaders, Okio.buffer(responseBody)));

57. }

58. return responseBuilder.build();

59. }

60. /** Returns a 'Cookie' HTTP request header with all cookies, like {@code a=b; c=d}. */

61. private String cookieHeader(List<Cookie> cookies) {

62. StringBuilder cookieHeader = new StringBuilder();

63. for (int i = 0, size = cookies.size(); i < size; i++) {

64. if (i > 0) {

65. cookieHeader.append("; ");

66. }

67. Cookie cookie = cookies.get(i);

68. cookieHeader.append(cookie.name()).append('=').append(cookie.value());

69. }

70. return cookieHeader.toString();

71. }

72. }

从上面的代码可以看出，首先获取原请求，然后在请求中添加头，比如Host、Connection、Accept-Encoding参数等，然后根据看是否需要填充Cookie，在对原始请求做出处理后，使用chain的procced方法得到响应，接下来对响应做处理得到用户响应，最后返回响应。接下来再看下一个拦截器ConnectInterceptor的处理。

1. public final class ConnectInterceptor implements Interceptor {

2. ......

3. @Override

4. public Response intercept(Chain chain) throws IOException {

5. RealInterceptorChain realChain = (RealInterceptorChain) chain;

6. Request request = realChain.request();

7. StreamAllocation streamAllocation = realChain.streamAllocation();

8. // We need the network to satisfy this request. Possibly for validating a conditional GET.

9. boolean doExtensiveHealthChecks = !request.method().equals("GET");

10. HttpCodec httpCodec = streamAllocation.newStream(client, doExtensiveHealthChecks);

11. RealConnection connection = streamAllocation.connection();

12. return realChain.proceed(request, streamAllocation, httpCodec, connection);

13. }

14. }

实际上建立连接就是创建了一个 HttpCodec 对象，它利用 Okio 对 Socket 的读写操作进行封装，Okio 以后有机会再进行分析，现在让我们对它们保持一个简单地认识：它对 java.io 和 java.nio 进行了封装，让我们更便捷高效的进行 IO 操作。

CallServerInterceptor

CallServerInterceptor是拦截器链中最后一个拦截器，负责将网络请求提交给服务器。它的intercept方法实现如下：

1. @Override

2. public Response intercept(Chain chain) throws IOException {

3. RealInterceptorChain realChain = (RealInterceptorChain) chain;

4. HttpCodec httpCodec = realChain.httpStream();

5. StreamAllocation streamAllocation = realChain.streamAllocation();

6. RealConnection connection = (RealConnection) realChain.connection();

7. Request request = realChain.request();

8. long sentRequestMillis = System.currentTimeMillis();

9. httpCodec.writeRequestHeaders(request);

10. Response.Builder responseBuilder = null;

11. if (HttpMethod.permitsRequestBody(request.method()) && request.body() != null) {

12. // If there's a "Expect: 100-continue" header on the request, wait for a "HTTP/1.1 100

13. // Continue" response before transmitting the request body. If we don't get that, return what

14. // we did get (such as a 4xx response) without ever transmitting the request body.

15. if ("100-continue".equalsIgnoreCase(request.header("Expect"))) {

16. httpCodec.flushRequest();

17. responseBuilder = httpCodec.readResponseHeaders(true);

18. }

19. if (responseBuilder == null) {

20. // Write the request body if the "Expect: 100-continue" expectation was met.

21. Sink requestBodyOut = httpCodec.createRequestBody(request, request.body().contentLength());

22. BufferedSink bufferedRequestBody = Okio.buffer(requestBodyOut);

23. request.body().writeTo(bufferedRequestBody);

24. bufferedRequestBody.close();

25. } else if (!connection.isMultiplexed()) {

26. // If the "Expect: 100-continue" expectation wasn't met, prevent the HTTP/1 connection from

27. // being reused. Otherwise we're still obligated to transmit the request body to leave the

28. // connection in a consistent state.

29. streamAllocation.noNewStreams();

30. }

31. }

32. httpCodec.finishRequest();

33. if (responseBuilder == null) {

34. responseBuilder = httpCodec.readResponseHeaders(false);

35. }

36. Response response = responseBuilder

37. .request(request)

38. .handshake(streamAllocation.connection().handshake())

39. .sentRequestAtMillis(sentRequestMillis)

40. .receivedResponseAtMillis(System.currentTimeMillis())

41. .build();

42. int code = response.code();

43. if (forWebSocket && code == 101) {

44. // Connection is upgrading, but we need to ensure interceptors see a non-null response body.

45. response = response.newBuilder()

46. .body(Util.EMPTY_RESPONSE)

47. .build();

48. } else {

49. response = response.newBuilder()

50. .body(httpCodec.openResponseBody(response))

51. .build();

52. }

53. if ("close".equalsIgnoreCase(response.request().header("Connection"))

54. || "close".equalsIgnoreCase(response.header("Connection"))) {

55. streamAllocation.noNewStreams();

56. }

57. if ((code == 204 || code == 205) && response.body().contentLength() > 0) {

58. throw new ProtocolException(

59. "HTTP " + code + " had non-zero Content-Length: " + response.body().contentLength());

60. }

61. return response;

62. }

从上面的代码中可以看出，首先获取HttpStream对象，然后调用writeRequestHeaders方法写入请求的头部，然后判断是否需要写入请求的body部分，最后调用finishRequest()方法将所有数据刷新给底层的Socket，接下来尝试调用readResponseHeaders()方法读取响应的头部，然后再调用openResponseBody()方法得到响应的body部分，最后返回响应。

最后总结

OkHttp的底层是通过Java的Socket发送HTTP请求与接受响应的(这也好理解，HTTP就是基于TCP协议的)，但是OkHttp实现了连接池的概念，即对于同一主机的多个请求，其实可以公用一个Socket连接，而不是每次发送完HTTP请求就关闭底层的Socket，这样就实现了连接池的概念。而OkHttp对Socket的读写操作使用的OkIo库进行了一层封装。

本文在开源项目：https://github.com/Android-Alvin/Android-LearningNotes 中已收录，里面包含不同方向的自学编程路线、面试题集合/面经、及系列技术文章等，资源持续更新中...