Google Volley框架源码走读
【工匠若水 http://blog.csdn.net/yanbober】 阅读前一篇《Google Volley使用之自定义》 http://blog.csdn.net/yanbober/article/details/45307099
开源项目链接
Volley主页：https://android.googlesource.com/platform/frameworks/volley
Volley仓库：git clone https://android.googlesource.com/platform/frameworks/volley
Volley GitHub Demo：在GitHub主页搜索Volley会有很多，不过建议阅读Android Developer文档。
背景知识
在Volley使用基础那一篇最后一个知识点说到了Volley的请求架构，这里再搬过来说说。
在Android Developer上看到的这幅图：
RequestQueue会维护一个缓存调度线程（cache线程）和一个网络调度线程池（net线程），当一个Request被加到队列中的时候，cache线程会把这个请求进行筛选：如果这个请求的内容可以在缓存中找到，cache线程会亲自解析相应内容，并分发到主线程（UI）。如果缓存中没有，这个request就会被加入到另一个NetworkQueue，所有真正准备进行网络通信的request都在这里，第一个可用的net线程会从NetworkQueue中拿出一个request扔向服务器。当响应数据到的时候，这个net线程会解析原始响应数据，写入缓存，并把解析后的结果返回给主线程。
硬着头皮开始吧
直接这么看好空洞，所以直接把clone的工程导入IDE边看代码边看这个图吧。
还是按照前边的顺序分析吧，使用Volley的第一步首先是通过Volley.newRequestQueue(context)得到RequestQueue队列，那么先看下toolbox下的Volley.java中的这个方法吧。
/** * Creates a default instance of the worker pool and calls {@link RequestQueue#start()} on it. * * @param context A {@link Context} to use for creating the cache dir. * @return A started {@link RequestQueue} instance. */ public static RequestQueue newRequestQueue(Context context) { return newRequestQueue(context, null); } 先看如上注释所示，创建一个默认的worker pool，并且调运RequestQueue的start方法。在这个方法里又调运了该类的另一个连个参数的重载方法，如下所示。
/** * Creates a default instance of the worker pool and calls {@link RequestQueue#start()} on it. * * @param context A {@link Context} to use for creating the cache dir. * @param stack An {@link HttpStack} to use for the network, or null for default. * @return A started {@link RequestQueue} instance. */ public static RequestQueue newRequestQueue(Context context, HttpStack stack) { File cacheDir = new File(context.getCacheDir(), DEFAULT_CACHE_DIR); String userAgent = "volley/0"; try { String packageName = context.getPackageName(); PackageInfo info = context.getPackageManager().getPackageInfo(packageName, 0); userAgent = packageName + "/" + info.versionCode; } catch (PackageManager.NameNotFoundException e) { } if (stack == null) { if (Build.VERSION.SDK_INT >= 9) { stack = new HurlStack(); } else { // Prior to Gingerbread, HttpUrlConnection was unreliable. // See: http://android-developers.blogspot.com/2011/09/androids-http-clients.html stack = new HttpClientStack(AndroidHttpClient.newInstance(userAgent)); } } Network network = new BasicNetwork(stack); RequestQueue queue = new RequestQueue(new DiskBasedCache(cacheDir), network); queue.start(); return queue; } 如上所示，该方法有两个参数，第一个Context是为了拿到当前App的Activity的一些信息，第二个参数在默认的RequestQueue newRequestQueue(Context context)方法中传递为null，也就是说在这段代码中的if (stack == null)会被执行。在这个if中对版本号进行了判断，如果版本号大于等于9就使得HttpStack对象的实例为HurlStack，如果小于9则实例为HttpClientStack。至于这里为何进行版本号判断，实际代码中的(Click Me To See)注释已经说明了。实际上HurlStack类也在toolbox中，他实现了toolbox的HttpStack接口中的HttpResponse performRequest(Request<?> request, Map<String, String> additionalHeaders)方法，其实现过程使用的是HttpURLConnection。而HttpClientStack也在toolbox中，他也实现了toolbox的HttpStack接口的HttpResponse performRequest(Request<?> request, Map<String, String> additionalHeaders)方法，不过其实现过程使用的是HttpClient而已。
其中的userAgent就是App的包名加版本号而已，传入new HttpClientStack(AndroidHttpClient.newInstance(userAgent));作为name TAG使用。
如上HttpStack创建完成之后创建了Network实例。BasicNetwork是Network接口的实现，他们都在toolbox中，BasicNetwork实现了public NetworkResponse performRequest(Request<?> request)方法，其作用是根据传入的HttpStack对象来处理网络请求。紧接着new出一个RequestQueue对象，并调用它的start()方法进行启动，然后将RequestQueue返回。RequestQueue是根目录下的一个类，其作用是一个请求调度队列调度程序的线程池。这样newRequestQueue()的方法就执行结束了。
现在再来看下根目录下RequestQueue队列的start方法，如下所示：
/** * Starts the dispatchers in this queue. */ public void start() { stop(); // Make sure any currently running dispatchers are stopped. // Create the cache dispatcher and start it. mCacheDispatcher = new CacheDispatcher(mCacheQueue, mNetworkQueue, mCache, mDelivery); mCacheDispatcher.start(); // Create network dispatchers (and corresponding threads) up to the pool size. for (int i = 0; i < mDispatchers.length; i++) { NetworkDispatcher networkDispatcher = new NetworkDispatcher(mNetworkQueue, mNetwork, mCache, mDelivery); mDispatchers[i] = networkDispatcher; networkDispatcher.start(); } } 通过注释可以看出来这里在派发队列的事务。先是创建了一个CacheDispatcher的实例，然后调用了它的start()方法，接着在一个for循环里去创建NetworkDispatcher的实例，并分别调用它们的start()方法。这里的CacheDispatcher和NetworkDispatcher都是继承自Thread的，而默认情况下for循环会执行（DEFAULT_NETWORK_THREAD_POOL_SIZE）四次，也就是说当调用了Volley.newRequestQueue(context)之后，就会有五个线程一直在后台运行，不断等待网络请求的到来，其中一个CacheDispatcher是缓存线程，四个NetworkDispatcher是网络请求线程。
按照之前使用Volley可以知道，得到了RequestQueue之后，我们只需要构建出相应的Request，然后调用RequestQueue的add()方法将Request传入就可以完成网络请求操作了。也就是说add()方法的内部是核心代码了。现在看下RequestQueue的add方法，具体如下：
/** * Adds a Request to the dispatch queue. * @param request The request to service * @return The passed-in request */ public <T> Request<T> add(Request<T> request) { // Tag the request as belonging to this queue and add it to the set of current requests. request.setRequestQueue(this); synchronized (mCurrentRequests) { mCurrentRequests.add(request); } // Process requests in the order they are added. request.setSequence(getSequenceNumber()); request.addMarker("add-to-queue"); // If the request is uncacheable, skip the cache queue and go straight to the network. if (!request.shouldCache()) { mNetworkQueue.add(request); return request; } // Insert request into stage if there's already a request with the same cache key in flight. synchronized (mWaitingRequests) { String cacheKey = request.getCacheKey(); if (mWaitingRequests.containsKey(cacheKey)) { // There is already a request in flight. Queue up. Queue<Request<?>> stagedRequests = mWaitingRequests.get(cacheKey); if (stagedRequests == null) { stagedRequests = new LinkedList<Request<?>>(); } stagedRequests.add(request); mWaitingRequests.put(cacheKey, stagedRequests); if (VolleyLog.DEBUG) { VolleyLog.v("Request for cacheKey=%s is in flight, putting on hold.", cacheKey); } } else { // Insert 'null' queue for this cacheKey, indicating there is now a request in // flight. mWaitingRequests.put(cacheKey, null); mCacheQueue.add(request); } return request; } } 可以看到注释所示，添加一个Request到派发队列。Request是所有请求的基类，是一个抽象类。request.setRequestQueue(this);的作用就是将请求Request关联到当前RequestQueue。然后同步操作将当前Request添加到RequestQueue对象的mCurrentRequests HashSet中做记录。通过request.setSequence(getSequenceNumber());得到当前RequestQueue中请求的个数，然后关联到当前Request。request.addMarker("add-to-queue");添加调试的Debug标记。if (!request.shouldCache())判断当前的请求是否可以缓存，如果不能缓存则直接通过mNetworkQueue.add(request);将这条请求加入网络请求队列，然后返回request；如果可以缓存的话则在通过同步操作将这条请求加入缓存队列。在默认情况下，每条请求都是可以缓存的，当然我们也可以调用Request的setShouldCache(false)方法来改变这一默认行为。OK，那么既然默认每条请求都是可以缓存的（shouldCache返回为true），自然就被添加到了缓存队列中，于是一直在后台等待的缓存线程就要开始运行起来了。现在来看下CacheDispatcher中的run()方法，代码如下所示：
@Override public void run() { if (DEBUG) VolleyLog.v("start new dispatcher"); Process.setThreadPriority(Process.THREAD_PRIORITY_BACKGROUND); // Make a blocking call to initialize the cache. mCache.initialize(); while (true) { try { // Get a request from the cache triage queue, blocking until // at least one is available. final Request<?> request = mCacheQueue.take(); request.addMarker("cache-queue-take"); // If the request has been canceled, don't bother dispatching it. if (request.isCanceled()) { request.finish("cache-discard-canceled"); continue; } // Attempt to retrieve this item from cache. Cache.Entry entry = mCache.get(request.getCacheKey()); if (entry == null) { request.addMarker("cache-miss"); // Cache miss; send off to the network dispatcher. mNetworkQueue.put(request); continue; } // If it is completely expired, just send it to the network. if (entry.isExpired()) { request.addMarker("cache-hit-expired"); request.setCacheEntry(entry); mNetworkQueue.put(request); continue; } // We have a cache hit; parse its data for delivery back to the request. request.addMarker("cache-hit"); Response<?> response = request.parseNetworkResponse( new NetworkResponse(entry.data, entry.responseHeaders)); request.addMarker("cache-hit-parsed"); if (!entry.refreshNeeded()) { // Completely unexpired cache hit. Just deliver the response. mDelivery.postResponse(request, response); } else { // Soft-expired cache hit. We can deliver the cached response, // but we need to also send the request to the network for // refreshing. request.addMarker("cache-hit-refresh-needed"); request.setCacheEntry(entry); // Mark the response as intermediate. response.intermediate = true; // Post the intermediate response back to the user and have // the delivery then forward the request along to the network. mDelivery.postResponse(request, response, new Runnable() { @Override public void run() { try { mNetworkQueue.put(request); } catch (InterruptedException e) { // Not much we can do about this. } } }); } } catch (InterruptedException e) { // We may have been interrupted because it was time to quit. if (mQuit) { return; } continue; } } } 首先通过Process.setThreadPriority(Process.THREAD_PRIORITY_BACKGROUND);设置线程优先级，然后通过mCache.initialize(); 初始化缓存块，其中mCache是由Volley.java中的newRequestQueue(Context context, HttpStack stack)方法中实例化传入的，其Cache接口的实现为new DiskBasedCache(cacheDir)，其中cacheDir默认在Volley.java中不设置为/data/data/app-package/cache/volley/。
接下来由while (true)可以发现缓存线程是一直在执行，其中通过mQuit标记进制是否结束线程的操作。mCacheQueue.take()从阻塞队列获取最前面的一个request，没有request就阻塞等待。接着通过mCache.get(request.getCacheKey());尝试从缓存中取出响应结果，如何为空的话则把这条请求加入到网络请求队列中，如果不为空的话再判断该缓存是否已过期，如果已经过期了则同样把这条请求加入到网络请求队列中，否则就认为不需要重发网络请求，直接使用缓存中的数据即可。在这个过程中调运了parseNetworkResponse()方法来对数据进行解析，再往后就是将解析出来的数据进行回调了。现在先来看下Request抽象基类的这部分代码：
/** * Subclasses must implement this to parse the raw network response * and return an appropriate response type. This method will be * called from a worker thread. The response will not be delivered * if you return null. * @param response Response from the network * @return The parsed response, or null in the case of an error */ abstract protected Response<T> parseNetworkResponse(NetworkResponse response);
通过注释可以看到他就是一个解析模块的功能。
上面说了，当调用了Volley.newRequestQueue(context)之后，就会有五个线程一直在后台运行，不断等待网络请求的到来，其中一个CacheDispatcher是缓存线程，四个NetworkDispatcher是网络请求线程。CacheDispatcher的run方法刚才已经大致分析了，解析来看下NetworkDispatcher中是怎么处理网络请求队列的，具体代码如下所示：
@Override public void run() { Process.setThreadPriority(Process.THREAD_PRIORITY_BACKGROUND); while (true) { long startTimeMs = SystemClock.elapsedRealtime(); Request<?> request; try { // Take a request from the queue. request = mQueue.take(); } catch (InterruptedException e) { // We may have been interrupted because it was time to quit. if (mQuit) { return; } continue; } try { request.addMarker("network-queue-take"); // If the request was cancelled already, do not perform the // network request. if (request.isCanceled()) { request.finish("network-discard-cancelled"); continue; } addTrafficStatsTag(request); // Perform the network request. NetworkResponse networkResponse = mNetwork.performRequest(request); request.addMarker("network-http-complete"); // If the server returned 304 AND we delivered a response already, // we're done -- don't deliver a second identical response. if (networkResponse.notModified && request.hasHadResponseDelivered()) { request.finish("not-modified"); continue; } // Parse the response here on the worker thread. Response<?> response = request.parseNetworkResponse(networkResponse); request.addMarker("network-parse-complete"); // Write to cache if applicable. // TODO: Only update cache metadata instead of entire record for 304s. if (request.shouldCache() && response.cacheEntry != null) { mCache.put(request.getCacheKey(), response.cacheEntry); request.addMarker("network-cache-written"); } // Post the response back. request.markDelivered(); mDelivery.postResponse(request, response); } catch (VolleyError volleyError) { volleyError.setNetworkTimeMs(SystemClock.elapsedRealtime() - startTimeMs); parseAndDeliverNetworkError(request, volleyError); } catch (Exception e) { VolleyLog.e(e, "Unhandled exception %s", e.toString()); VolleyError volleyError = new VolleyError(e); volleyError.setNetworkTimeMs(SystemClock.elapsedRealtime() - startTimeMs); mDelivery.postError(request, volleyError); } } } 和CacheDispatcher差不多，如上可以看见一个类似的while(true)循环，说明网络请求线程也是在不断运行的。
如上通过mNetwork.performRequest(request);代码来发送网络请求，而Network是一个接口，这里具体的实现之前已经分析是BasicNetwork，所以先看下它的performRequest()方法，如下所示：
NetWork接口的代码：
public interface Network { /** * Performs the specified request. * @param request Request to process * @return A {@link NetworkResponse} with data and caching metadata; will never be null * @throws VolleyError on errors */ public NetworkResponse performRequest(Request<?> request) throws VolleyError; } 上面说了，就是执行指定的请求。他的BasicNetwork实现子类如下：
@Override public NetworkResponse performRequest(Request<?> request) throws VolleyError { long requestStart = SystemClock.elapsedRealtime(); while (true) { HttpResponse httpResponse = null; byte[] responseContents = null; Map<String, String> responseHeaders = Collections.emptyMap(); try { // Gather headers. Map<String, String> headers = new HashMap<String, String>(); addCacheHeaders(headers, request.getCacheEntry()); httpResponse = mHttpStack.performRequest(request, headers); StatusLine statusLine = httpResponse.getStatusLine(); int statusCode = statusLine.getStatusCode(); responseHeaders = convertHeaders(httpResponse.getAllHeaders()); // Handle cache validation. if (statusCode == HttpStatus.SC_NOT_MODIFIED) { Entry entry = request.getCacheEntry(); if (entry == null) { return new NetworkResponse(HttpStatus.SC_NOT_MODIFIED, null, responseHeaders, true, SystemClock.elapsedRealtime() - requestStart); } // A HTTP 304 response does not have all header fields. We // have to use the header fields from the cache entry plus // the new ones from the response. // http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html#sec10.3.5 entry.responseHeaders.putAll(responseHeaders); return new NetworkResponse(HttpStatus.SC_NOT_MODIFIED, entry.data, entry.responseHeaders, true, SystemClock.elapsedRealtime() - requestStart); } // Some responses such as 204s do not have content. We must check. if (httpResponse.getEntity() != null) { responseContents = entityToBytes(httpResponse.getEntity()); } else { // Add 0 byte response as a way of honestly representing a // no-content request. responseContents = new byte[0]; } // if the request is slow, log it. long requestLifetime = SystemClock.elapsedRealtime() - requestStart; logSlowRequests(requestLifetime, request, responseContents, statusLine); if (statusCode < 200 || statusCode > 299) { throw new IOException(); } return new NetworkResponse(statusCode, responseContents, responseHeaders, false, SystemClock.elapsedRealtime() - requestStart); } catch (SocketTimeoutException e) { attemptRetryOnException("socket", request, new TimeoutError()); } catch (ConnectTimeoutException e) { attemptRetryOnException("connection", request, new TimeoutError()); } catch (MalformedURLException e) { throw new RuntimeException("Bad URL " + request.getUrl(), e); } catch (IOException e) { int statusCode = 0; NetworkResponse networkResponse = null; if (httpResponse != null) { statusCode = httpResponse.getStatusLine().getStatusCode(); } else { throw new NoConnectionError(e); } VolleyLog.e("Unexpected response code %d for %s", statusCode, request.getUrl()); if (responseContents != null) { networkResponse = new NetworkResponse(statusCode, responseContents, responseHeaders, false, SystemClock.elapsedRealtime() - requestStart); if (statusCode == HttpStatus.SC_UNAUTHORIZED || statusCode == HttpStatus.SC_FORBIDDEN) { attemptRetryOnException("auth", request, new AuthFailureError(networkResponse)); } else { // TODO: Only throw ServerError for 5xx status codes. throw new ServerError(networkResponse); } } else { throw new NetworkError(networkResponse); } } } } 这个方法是网络请求的具体实现，也是一个大while循环，其中mHttpStack.performRequest(request, headers);代码中的mHttpStack是Volley的newRequestQueue()方法中创建的实例，前面已经说过，这两个对象的内部实际就是分别使用HttpURLConnection和HttpClient来发送网络请求的，然后把服务器返回的数据组装成一个NetworkResponse对象进行返回。在NetworkDispatcher中收到了NetworkResponse这个返回值后又会调用Request的parseNetworkResponse()方法来解析NetworkResponse中的数据，同时将数据写入到缓存，这个方法的实现是交给Request的子类来完成的，因为不同种类的Request解析的方式也肯定不同。
前面你可以看到在NetWorkDispatcher的run中最后执行了mDelivery.postResponse(request, response);，也就是说在解析完了NetworkResponse中的数据之后，又会调用ExecutorDelivery（ResponseDelivery接口的实现类）的postResponse()方法来回调解析出的数据，具体代码如下所示：
@Override public void postResponse(Request<?> request, Response<?> response, Runnable runnable) { request.markDelivered(); request.addMarker("post-response"); mResponsePoster.execute(new ResponseDeliveryRunnable(request, response, runnable)); } 这里可以看见在mResponsePoster的execute()方法中传入了一个ResponseDeliveryRunnable对象，就可以保证该对象中的run()方法就是在主线程当中运行的了，我们看下run()方法中的代码是什么样的：
/** * A Runnable used for delivering network responses to a listener on the * main thread. */ @SuppressWarnings("rawtypes") private class ResponseDeliveryRunnable implements Runnable { private final Request mRequest; private final Response mResponse; private final Runnable mRunnable; public ResponseDeliveryRunnable(Request request, Response response, Runnable runnable) { mRequest = request; mResponse = response; mRunnable = runnable; } @SuppressWarnings("unchecked") @Override public void run() { // If this request has canceled, finish it and don't deliver. if (mRequest.isCanceled()) { mRequest.finish("canceled-at-delivery"); return; } // Deliver a normal response or error, depending. if (mResponse.isSuccess()) { mRequest.deliverResponse(mResponse.result); } else { mRequest.deliverError(mResponse.error); } // If this is an intermediate response, add a marker, otherwise we're done // and the request can be finished. if (mResponse.intermediate) { mRequest.addMarker("intermediate-response"); } else { mRequest.finish("done"); } // If we have been provided a post-delivery runnable, run it. if (mRunnable != null) { mRunnable.run(); } } } 这段代码里的run方法中可以看到如下一部分细节：
// Deliver a normal response or error, depending. if (mResponse.isSuccess()) { mRequest.deliverResponse(mResponse.result); } else { mRequest.deliverError(mResponse.error); } 这段代码是最核心的，明显可以看到通过mRequest的deliverResponse或者deliverError将反馈发送到回调到UI线程。这也是你重写实现的接口方法。
再来整体看下
现在是该回过头去看背景知识模块了，再看下那幅官方图，对比就明白咋回事了。结合上图和如上分析可以知道：
-
当一个RequestQueue被成功申请后会开启一个CacheDispatcher和4个默认的NetworkDispatcher。
-
CacheDispatcher缓存调度器最为第一层缓冲，开始工作后阻塞的从缓存序列mCacheQueue中取得请求；对于已经取消的请求，标记为跳过并结束这个请求；新的或者过期的请求，直接放入mNetworkQueue中由N个NetworkDispatcher进行处理；已获得缓存信息（网络应答）却没有过期的请求，由Request的parseNetworkResponse进行解析，从而确定此应答是否成功。然后将请求和应答交由Delivery分发者进行处理，如果需要更新缓存那么该请求还会被放入mNetworkQueue中。
-
将请求Request add到RequestQueue后对于不需要缓存的请求（需要额外设置，默认是需要缓存）直接丢入mNetworkQueue交给N个NetworkDispatcher处理；对于需要缓存的，新的请求加到mCacheQueue中给CacheDispatcher处理；需要缓存，但是缓存列表中已经存在了相同URL的请求，放在mWaitingQueue中做暂时处理，等待之前请求完毕后，再重新添加到mCacheQueue中。
-
网络请求调度器NetworkDispatcher作为网络请求真实发生的地方，对消息交给BasicNetwork进行处理，同样的，请求和结果都交由Delivery分发者进行处理。
-
Delivery分发者实际上已经是对网络请求处理的最后一层了，在Delivery对请求处理之前，Request已经对网络应答进行过解析，此时应答成功与否已经设定；而后Delivery根据请求所获得的应答情况做不同处理；若应答成功，则触发deliverResponse方法，最终会触发开发者为Request设定的Listener；若应答失败，则触发deliverError方法，最终会触发开发者为Request设定的ErrorListener；处理完后，一个Request的生命周期就结束了，Delivery会调用Request的finish操作，将其从mRequestQueue中移除，与此同时，如果等待列表中存在相同URL的请求，则会将剩余的层级请求全部丢入mCacheQueue交由CacheDispatcher进行处理。
至此所有搞定。
PPPS一句：通过上面原理分析之后总结发现，推荐整个App全局持有一个RequestQueue的做法，这样会有相对比较高的性能效率。