前言:
對於一個Android研發而言,親身體會就是不管在平時開發或者面試的時候,Handler消息機制毋庸置疑都是一個必備的知識點,所以這邊留一份個人筆記,如有分析不對的地方,還望指出!
目錄:
1、如何分析Handler源碼
2、源碼大致流程:消息的入隊與出隊
3、從大致流程進入細化分析
3.1、Handler、Looper、MessageQueue三者之間的關系
3.2、Handler、Looper、MessageQueue之間的協作
總結圖1:Handler在子線程中發送消息,消息會被添加到MessageQueue消息隊列中,再來由Handler所處的當前線程的Looper來不斷的輪詢MessageQueue以獲取出隊消息,最后調用dispatchMessage進行消息傳遞給handleMessage進行處理
1、如何分析源碼
眾所皆知的Android源碼的有很多,涉及到一個類或者多個類,一個類中又有很多代碼,所以這邊最簡單的分析方式就是回歸到Handler的使用中來,也就是如何使用Handler
1.1、實例一個Handler對象(主線程)
1.2、在子線程中使用Handler發送一個消息,如:handler.sendEmptyMessage(1)
1.3、消息發送出之后(執行1.2步驟),消息最終會被轉發到我們new出來的Handler中的handleMessage方法進行處理(子線程消息發送到主線程中處理)
以上3個步驟即為我們對Handler的基本使用方式,所以,我們可以以發送消息的時機,作為源碼分析的切入點,並留下一個疑問:
問題1:子線程發送的消息為什么是在主線程中接收的呢?
2、源碼大致流程:消息的入隊與出隊
2.1、消息發送:sendMessage(Message msg) \ sendEmptyMessage(int what) \ postDelayed(Runnable r, long delayMillis) 等等
2.2、消息及發送時間處理:sendMessageAtTime(Message msg, long uptimeMillis)
2.3、消息隊列添加:enqueueMessage(MessageQueue queue, Message msg, long uptimeMillis)
2.4、消息出隊:到這里,既然有消息添加到隊列中的流程,而且我們最終都會獲得相應的消息返回,那么消息是如何出隊的呢?帶着這個疑問,我們最終在MessageQueue 消息隊列中找到一個函數名稱為 next() 的函數。
問題2:這個next()函數 是在什么時候調用的呢?
在Handler源碼上,以消息發送作為分析切入點來查看,如2.1羅列的幾種消息發送方式,我們都可以很清楚的發現,消息都是調用了sendMessageDelayed(Message msg, long delayMillis),最終調用到sendMessageAtTime(Message msg, long uptimeMillis),然后在該方法里面調用了enqueueMessage(MessageQueue queue, Message msg, long uptimeMillis),到這里,不管從方法名稱還是局部變量的名稱來看,這邊都出現了一個隊列的信息,所以可以知道Handler的消息發送最終是在sendMessageAtTime里面調用了MessageQueue.enqueueMessage()對消息進行隊列添加,然后調用了MessageQueue.next()進行消息輪詢並返回Message結果。
3、從大致流程進入細化分析
3.1、Handler、Looper、MessageQueue三者之間的 關系圖2 如下:
在分析到第2步的sendMessageAtTime結束時,我們這邊引出了一個消息隊列的內容:MessageQueue queue = mQueue
1 /** 2 * Enqueue a message at the front of the message queue, to be processed on 3 * the next iteration of the message loop. You will receive it in 4 * {@link #handleMessage}, in the thread attached to this handler. 5 * <b>This method is only for use in very special circumstances -- it 6 * can easily starve the message queue, cause ordering problems, or have 7 * other unexpected side-effects.</b> 8 * 9 * @return Returns true if the message was successfully placed in to the 10 * message queue. Returns false on failure, usually because the 11 * looper processing the message queue is exiting. 12 */ 13 public final boolean sendMessageAtFrontOfQueue(Message msg) { 14 MessageQueue queue = mQueue; 15 if (queue == null) { 16 RuntimeException e = new RuntimeException( 17 this + " sendMessageAtTime() called with no mQueue"); 18 Log.w("Looper", e.getMessage(), e); 19 return false; 20 } 21 return enqueueMessage(queue, msg, 0); 22 } 23 24 private boolean enqueueMessage(MessageQueue queue, Message msg, long uptimeMillis) { 25 msg.target = this; 26 if (mAsynchronous) { 27 msg.setAsynchronous(true); 28 } 29 return queue.enqueueMessage(msg, uptimeMillis); 30 }
問題:mQueue是什么東西?這個mQueue是怎么來的?所以我們在Handler的構造方法中找到了它的初始化位置
1 /** 2 * Use the {@link Looper} for the current thread with the specified callback interface 3 * and set whether the handler should be asynchronous. 4 * 5 * Handlers are synchronous by default unless this constructor is used to make 6 * one that is strictly asynchronous. 7 * 8 * Asynchronous messages represent interrupts or events that do not require global ordering 9 * with respect to synchronous messages. Asynchronous messages are not subject to 10 * the synchronization barriers introduced by {@link MessageQueue#enqueueSyncBarrier(long)}. 11 * 12 * @param callback The callback interface in which to handle messages, or null. 13 * @param async If true, the handler calls {@link Message#setAsynchronous(boolean)} for 14 * each {@link Message} that is sent to it or {@link Runnable} that is posted to it. 15 * 16 * @hide 17 */ 18 public Handler(Callback callback, boolean async) { 19 if (FIND_POTENTIAL_LEAKS) { 20 final Class<? extends Handler> klass = getClass(); 21 if ((klass.isAnonymousClass() || klass.isMemberClass() || klass.isLocalClass()) && 22 (klass.getModifiers() & Modifier.STATIC) == 0) { 23 Log.w(TAG, "The following Handler class should be static or leaks might occur: " + 24 klass.getCanonicalName()); 25 } 26 } 27 28 mLooper = Looper.myLooper(); 29 if (mLooper == null) { 30 throw new RuntimeException( 31 "Can't create handler inside thread that has not called Looper.prepare()"); 32 } 33 mQueue = mLooper.mQueue; 34 mCallback = callback; 35 mAsynchronous = async; 36 }
到此,從上面的兩段源碼,且帶着第2點中,next()被調用的時機問題,我們引出了os/Handler中的兩個成員變量
final Looper mLooper; final MessageQueue mQueue;
MessageQueue 對象是從Looper中獲得的,也就是說mQueue是在Looper中實例化的,所以很明顯,Handler中的消息隊列MessageQueue 是從輪詢器Looper中獲得的。
那么問題來了:為什么消息隊列要在輪詢器中進行實例化,請看以下源碼
1 /** 2 * Return the {@link MessageQueue} object associated with the current 3 * thread. This must be called from a thread running a Looper, or a 4 * NullPointerException will be thrown. 5 */ 6 public static @NonNull MessageQueue myQueue() { 7 return myLooper().mQueue; 8 } 9 10 private Looper(boolean quitAllowed) { 11 mQueue = new MessageQueue(quitAllowed); 12 mThread = Thread.currentThread(); 13 }
MessageQueue在Looper中進行實例化,也就是說一個Looper就有一個MessageQueue,屬於綁定關系,從而得出一個Looper只能輪詢一個消息隊列
所以可得出如關系圖2中Handler、Looper、MessageQueue三者的關系:Handler中持有Looper和MessageQueue,Looper中持有MessageQueue,而且Handler中的MessageQueue來自於Looper中的MessageQueue。
Handler是使用時通過New實例化出來的,MessageQueue是在Looper中進行實例的,那么這個Looper是如何實例化的?所以這邊我們將引出 ActivityThread。而ActivityThread是什么東西呢?這邊就稍微介紹一下:
安卓應用程序作為一個控制類程序,跟Java程序類似,都是有一個入口的,而這個入口就是ActivityThread的main函數:
1 public static void main(String[] args) { 2 Trace.traceBegin(Trace.TRACE_TAG_ACTIVITY_MANAGER, "ActivityThreadMain"); 3 SamplingProfilerIntegration.start(); 4 5 // CloseGuard defaults to true and can be quite spammy. We 6 // disable it here, but selectively enable it later (via 7 // StrictMode) on debug builds, but using DropBox, not logs. 8 CloseGuard.setEnabled(false); 9 10 Environment.initForCurrentUser(); 11 12 // Set the reporter for event logging in libcore 13 EventLogger.setReporter(new EventLoggingReporter()); 14 15 // Make sure TrustedCertificateStore looks in the right place for CA certificates 16 final File configDir = Environment.getUserConfigDirectory(UserHandle.myUserId()); 17 TrustedCertificateStore.setDefaultUserDirectory(configDir); 18 19 Process.setArgV0("<pre-initialized>"); 20 21 Looper.prepareMainLooper(); 22 23 ActivityThread thread = new ActivityThread(); 24 thread.attach(false); 25 26 if (sMainThreadHandler == null) { 27 sMainThreadHandler = thread.getHandler(); 28 } 29 30 if (false) { 31 Looper.myLooper().setMessageLogging(new 32 LogPrinter(Log.DEBUG, "ActivityThread")); 33 } 34 35 // End of event ActivityThreadMain. 36 Trace.traceEnd(Trace.TRACE_TAG_ACTIVITY_MANAGER); 37 Looper.loop(); 38 39 throw new RuntimeException("Main thread loop unexpectedly exited"); 40 }
以上的main函數代碼中,我們還需要意識到兩個問題:
1.我們之所以可以在Activity用Handler handler=new Handler()直接創建出來就默認綁定到主線程了,是因為上面的代碼為我們做了綁定主線程的Looper的事情,
2.主線程的Looper是不能在程序中調用退出的,最后一句代碼看到沒,如果調用的話,就會拋出異常,退出主線程的循環是框架層在調用退出應用程序的時候才調用的
1 /** 2 * Initialize the current thread as a looper, marking it as an 3 * application's main looper. The main looper for your application 4 * is created by the Android environment, so you should never need 5 * to call this function yourself. See also: {@link #prepare()} 6 */ 7 public static void prepareMainLooper() { 8 prepare(false); 9 synchronized (Looper.class) { 10 if (sMainLooper != null) { 11 throw new IllegalStateException("The main Looper has already been prepared."); 12 } 13 sMainLooper = myLooper(); 14 } 15 } 16 17 18 private static void prepare(boolean quitAllowed) { 19 if (sThreadLocal.get() != null) { 20 throw new RuntimeException("Only one Looper may be created per thread"); 21 } 22 sThreadLocal.set(new Looper(quitAllowed)); 23 }
在ActivityThread的main中調用了 Looper.prepareMainLooper() -> prepare(false) -> sThreadLocal.set(new Looper(quitAllowed)), 這么一來,是不是執行到了上面的Looper構造函數中了?到這里,細心的人會發現這么一個問題:
問題3:Looper被實例化出來之后並沒有直接返回,而是被set到了ThreadLocal中?
Handler與Looper是成對出現的,一個子線程發送消息,一個主線程接收消息,那么這邊就涉及到了多線程,線程之間的通訊,是需要保證數據的安全,即數據隔離,所以使用到了ThreadLocal進行線程管理:如A線程在獲取數據時只能獲取A線程所控制的數據,而不能去獲取到B線程中對應的數據,否則就會引起數據不同步,比如A線程數據被B線程數據所覆蓋之類的問題,同時也驗證了一個線程只能關聯一個Looper對象。
所以問題3解決了。最后這個main的結尾,調用了 Looper.loop(); 進行輪詢消息隊列! 是不是很完美了?
3.2、Handler、Looper、MessageQueue之間的協作
通過前面的源碼分析,我們已經知道了消息是如果添加到消息隊列了。我們再來看消息的出隊分析。
以下在Looper輪詢器中的loop()中我們看到這樣一句代碼:Message msg = queue.next(); // might block
1 /** 2 * Run the message queue in this thread. Be sure to call 3 * {@link #quit()} to end the loop. 4 */ 5 public static void loop() { 6 final Looper me = myLooper(); 7 if (me == null) { 8 throw new RuntimeException("No Looper; Looper.prepare() wasn't called on this thread."); 9 } 10 final MessageQueue queue = me.mQueue; 11 12 // Make sure the identity of this thread is that of the local process, 13 // and keep track of what that identity token actually is. 14 Binder.clearCallingIdentity(); 15 final long ident = Binder.clearCallingIdentity(); 16 17 for (;;) { 18 Message msg = queue.next(); // might block 19 if (msg == null) { 20 // No message indicates that the message queue is quitting. 21 return; 22 } 23 24 // This must be in a local variable, in case a UI event sets the logger 25 final Printer logging = me.mLogging; 26 if (logging != null) { 27 logging.println(">>>>> Dispatching to " + msg.target + " " + 28 msg.callback + ": " + msg.what); 29 } 30 31 final long traceTag = me.mTraceTag; 32 if (traceTag != 0 && Trace.isTagEnabled(traceTag)) { 33 Trace.traceBegin(traceTag, msg.target.getTraceName(msg)); 34 } 35 try { 36 msg.target.dispatchMessage(msg); 37 } finally { 38 if (traceTag != 0) { 39 Trace.traceEnd(traceTag); 40 } 41 } 42 43 if (logging != null) { 44 logging.println("<<<<< Finished to " + msg.target + " " + msg.callback); 45 } 46 47 // Make sure that during the course of dispatching the 48 // identity of the thread wasn't corrupted. 49 final long newIdent = Binder.clearCallingIdentity(); 50 if (ident != newIdent) { 51 Log.wtf(TAG, "Thread identity changed from 0x" 52 + Long.toHexString(ident) + " to 0x" 53 + Long.toHexString(newIdent) + " while dispatching to " 54 + msg.target.getClass().getName() + " " 55 + msg.callback + " what=" + msg.what); 56 } 57 58 msg.recycleUnchecked(); 59 } 60 }
所以通過以上代碼,我們可以知道消息的出隊,是在Looper這個輪詢器中的loop()函數通過死循環的方式:for (;;),不斷的通過隊列的next()方法中拿到消息:queue.next(),並且如果隊列消息為null了,就跳出該輪詢。所以問題2是不是已經解決了?
在出隊過程中,也就是MessageQueue消息隊列中的next()函數中,我們可以知道next()返回的是一個Message消息對象,從函數注釋上來看:當輪詢器 loop 輪詢的時候會返回一條消息,且從代碼for (;;)循環的代碼中可以看出,是在這里不斷的拿到消息隊列並返回下一條消息,到這里,我們需要注意的是因為這個消息是可以循環使用的,而且我們可以看到這樣一個native函數調用:nativePollOnce(ptr, nextPollTimeoutMillis);所以我們可以得出消息的循環使用內存是通過C++來維護完成的(這邊因為對native沒有深入研究,所以pass這塊!)
1 Message next() { 2 // Return here if the message loop has already quit and been disposed. 3 // This can happen if the application tries to restart a looper after quit 4 // which is not supported. 5 final long ptr = mPtr; 6 if (ptr == 0) { 7 return null; 8 } 9 10 int pendingIdleHandlerCount = -1; // -1 only during first iteration 11 int nextPollTimeoutMillis = 0; 12 for (;;) { 13 if (nextPollTimeoutMillis != 0) { 14 Binder.flushPendingCommands(); 15 } 16 17 nativePollOnce(ptr, nextPollTimeoutMillis); 18 19 synchronized (this) { 20 // Try to retrieve the next message. Return if found. 21 final long now = SystemClock.uptimeMillis(); 22 Message prevMsg = null; 23 Message msg = mMessages; 24 if (msg != null && msg.target == null) { 25 // Stalled by a barrier. Find the next asynchronous message in the queue. 26 do { 27 prevMsg = msg; 28 msg = msg.next; 29 } while (msg != null && !msg.isAsynchronous()); 30 } 31 if (msg != null) { 32 if (now < msg.when) { 33 // Next message is not ready. Set a timeout to wake up when it is ready. 34 nextPollTimeoutMillis = (int) Math.min(msg.when - now, Integer.MAX_VALUE); 35 } else { 36 // Got a message. 37 mBlocked = false; 38 if (prevMsg != null) { 39 prevMsg.next = msg.next; 40 } else { 41 mMessages = msg.next; 42 } 43 msg.next = null; 44 if (DEBUG) Log.v(TAG, "Returning message: " + msg); 45 msg.markInUse(); 46 return msg; 47 } 48 } else { 49 // No more messages. 50 nextPollTimeoutMillis = -1; 51 } 52 53 // Process the quit message now that all pending messages have been handled. 54 if (mQuitting) { 55 dispose(); 56 return null; 57 } 58 59 // If first time idle, then get the number of idlers to run. 60 // Idle handles only run if the queue is empty or if the first message 61 // in the queue (possibly a barrier) is due to be handled in the future. 62 if (pendingIdleHandlerCount < 0 63 && (mMessages == null || now < mMessages.when)) { 64 pendingIdleHandlerCount = mIdleHandlers.size(); 65 } 66 if (pendingIdleHandlerCount <= 0) { 67 // No idle handlers to run. Loop and wait some more. 68 mBlocked = true; 69 continue; 70 } 71 72 if (mPendingIdleHandlers == null) { 73 mPendingIdleHandlers = new IdleHandler[Math.max(pendingIdleHandlerCount, 4)]; 74 } 75 mPendingIdleHandlers = mIdleHandlers.toArray(mPendingIdleHandlers); 76 } 77 78 // Run the idle handlers. 79 // We only ever reach this code block during the first iteration. 80 for (int i = 0; i < pendingIdleHandlerCount; i++) { 81 final IdleHandler idler = mPendingIdleHandlers[i]; 82 mPendingIdleHandlers[i] = null; // release the reference to the handler 83 84 boolean keep = false; 85 try { 86 keep = idler.queueIdle(); 87 } catch (Throwable t) { 88 Log.wtf(TAG, "IdleHandler threw exception", t); 89 } 90 91 if (!keep) { 92 synchronized (this) { 93 mIdleHandlers.remove(idler); 94 } 95 } 96 } 97 98 // Reset the idle handler count to 0 so we do not run them again. 99 pendingIdleHandlerCount = 0; 100 101 // While calling an idle handler, a new message could have been delivered 102 // so go back and look again for a pending message without waiting. 103 nextPollTimeoutMillis = 0; 104 } 105 }
到這邊,如果所有的分析及源碼查看都看懂的話,我們就已經掌握了在整個Handler消息機制中,是如何從一個消息的發送,進行了怎么樣的世界環游,最終如何回到了Handler的handleMessage中的!
分析到這里為止,如果還沒蒙圈的人會發現,我們在前面提出過的幾個問題都解決了,那么問題1呢?
子線程發送的消息為什么是在主線程中接收的呢?
其實我們在前面也已經有提及到了該問題,就是為什么在ActivityThread的main中實例化的Looper對象是被set到了ThreadLocal中。
在java中,main是不是主線程呢?不需要解釋了吧。看代碼:當前線程中Looper的獲取方式
1 /** 2 * Return the Looper object associated with the current thread. Returns 3 * null if the calling thread is not associated with a Looper. 4 */ 5 public static @Nullable Looper myLooper() { 6 return sThreadLocal.get(); 7 }
程序一開始在ActivityThread在執行main函數時實例化Looper,然后保存到了ThreadLocal中。而我們在主線程中new了一個Handler,那么Handler默認對應的Looper就是主線程的Looper:通過以上代碼,從ThreadLocal管理中獲取出當前線程(主線程)對應的Looper對象,所以對應的Looper自然也是主線程的Looper,明白了嗎?
所以主線程的Looper在輪詢出消息隊列MessageQueue中的消息時,就是出於主線程中,這樣問題1是不是就清楚了。