android中的一个层次状态机：HierarchicalStateMachine（A2DP）

介绍的仅是代码跟踪，不是结构描述。感觉结构的话就是通过HierarchicalStateMachine.HsmHandler.StateInfo来维持的吧。子状态上面是父状态，于是形成了类似树一样的结构，每个字节点都只有一个父节点，于是便可以通过类似迷宫算法的逆向遍历来取得状态分支了。
源自跟踪A2dp的连接方法connectSink：在类BluetoothService，

view plainprint?

1. public boolean connectSink(String address) { BluetoothDeviceProfileState state = mDeviceProfileState.get(address); if (state != null) { Message msg = new Message(); msg.arg1 = BluetoothDeviceProfileState.CONNECT_A2DP_OUTGOING; msg.obj = state; mA2dpProfileState.sendMessage(msg); return true; } return false; }

1.
public boolean connectSink(String address) {
BluetoothDeviceProfileState state = mDeviceProfileState.get(address);
if (state != null) {
Message msg = new Message();
msg.arg1 = BluetoothDeviceProfileState.CONNECT_A2DP_OUTGOING;
msg.obj = state;
mA2dpProfileState.sendMessage(msg);
return true;
}
return false;
}

mA2dpProfileState是BluetoothProfileState的实例，而BluetoothProfileState继承自HierarchicalStateMachine。

下面看一下HierarchicalStateMachine类:

这个类只有几个表示名字或状态的常数，除此之外还有一个名为HsmHandler的内部类实例，这个类是Handler的子类，里面有大堆的方法和参数，而HierarchicalStateMachine的真正成员方法里面都是调用这个handler里面的方法。所以此状态机的实现都在这个 handler的类中，因此只需跟踪这个子类的代码即可。

众所周知，直接看代码不如分析流程，所以现在回到主题，跟踪一下

view plainprint?

2.mA2dpProfileState.sendMessage(msg);

2.mA2dpProfileState.sendMessage(msg);

这段代码。

mA2dpProfileState初始化在BluetoothService中：

view plainprint?

3.mA2dpProfileState = new BluetoothProfileState(mContext, BluetoothProfileState.A2DP); //... mA2dpProfileState.start();

3.mA2dpProfileState = new BluetoothProfileState(mContext, BluetoothProfileState.A2DP);
//...
mA2dpProfileState.start();

类BluetoothProfileState构造方法：



view plainprint?

4.public BluetoothProfileState(Context context, int profile) { //... addState(mStableState); addState(mPendingCommandState); setInitialState(mStableState); //... }

4.public BluetoothProfileState(Context context, int profile) {
//...
addState(mStableState);
addState(mPendingCommandState);
setInitialState(mStableState);
//...
}

addState()是调用的HierarchicalStateMachine中的方法，实际调用：HsmHandler类的addState方法：



view plainprint?

5.private final StateInfo addState(HierarchicalState state, HierarchicalState parent){ //... stateInfo = new StateInfo(); mStateInfo.put(state, stateInfo); //... }

5.private final StateInfo addState(HierarchicalState state, HierarchicalState parent){
//...
stateInfo = new StateInfo();
mStateInfo.put(state, stateInfo);
//...
}

就是把一个状态放在mStateInfo中。我们可以想象，每一个状态机构造的不同state都放到这个HashMap中以供选择了。

在HsmHandler类：

view plainprint?

6.private HashMap<HierarchicalState, StateInfo> mStateInfo = new HashMap<HierarchicalState, StateInfo>();

6.private HashMap<HierarchicalState, StateInfo> mStateInfo = new HashMap<HierarchicalState, StateInfo>();

另外一个方法：

view plainprint?

7.setInitialState(HierarchicalState initialState){mInitialState = initialState;}

7.setInitialState(HierarchicalState initialState){mInitialState = initialState;}

这个方法把mInitialState置为传入的状态。

所以，BluetoothProfileState的初始化，做了两件事情：

a>把mStableState和mPendingCommandState这两个HierarchicalState加入到mStateInfo中。

b>把mInitialState设为mStableState。

下面回到 代码3 的mA2dpProfileState.start();

调用的就是HierarchicalStateMachine的start方法，实际调用HsmHandler的方法completeConstruction：

view plainprint?

8.private final void completeConstruction() {
if (mDbg) Log.d(TAG, "completeConstruction: E");

/**
* Determine the maximum depth of the state hierarchy
* so we can allocate the state stacks.
*/
int maxDepth = 0;
for (StateInfo si : mStateInfo.values()) {
int depth = 0;
for (StateInfo i = si; i != null; depth++) {
i = i.parentStateInfo;
}
if (maxDepth < depth) {
maxDepth = depth;
}
}
if (mDbg) Log.d(TAG, "completeConstruction: maxDepth=" + maxDepth);

mStateStack = new StateInfo[maxDepth];
mTempStateStack = new StateInfo[maxDepth];
setupInitialStateStack();

/**
* Construction is complete call all enter methods
* starting at the first entry.
*/
mIsConstructionCompleted = true;
mMsg = obtainMessage(HSM_INIT_CMD);
invokeEnterMethods(0);

/**
* Perform any transitions requested by the enter methods
*/
performTransitions();

if (mDbg) Log.d(TAG, "completeConstruction: X");
}

8.private final void completeConstruction() {
if (mDbg) Log.d(TAG, "completeConstruction: E");

/**
* Determine the maximum depth of the state hierarchy
* so we can allocate the state stacks.
*/
int maxDepth = 0;
for (StateInfo si : mStateInfo.values()) {
int depth = 0;
for (StateInfo i = si; i != null; depth++) {
i = i.parentStateInfo;
}
if (maxDepth < depth) {
maxDepth = depth;
}
}
if (mDbg) Log.d(TAG, "completeConstruction: maxDepth=" + maxDepth);

mStateStack = new StateInfo[maxDepth];
mTempStateStack = new StateInfo[maxDepth];
setupInitialStateStack();

/**
* Construction is complete call all enter methods
* starting at the first entry.
*/
mIsConstructionCompleted = true;
mMsg = obtainMessage(HSM_INIT_CMD);
invokeEnterMethods(0);

/**
* Perform any transitions requested by the enter methods
*/
performTransitions();

if (mDbg) Log.d(TAG, "completeConstruction: X");
}

首先，从mStateInfo中查找长度最长的（状态最多的那条分支），该长度作为数组mStateStack的大小。而本例中所有状态都没有父状态所以最大长度应为1.

mStateStack这个东西就是一个数组，里面装的是某个状态（StateInfo）（这个状态在状态机初始化的时候取值为mInitialState）

的分支（状态的父状态或者子状态的父状态或者子状态 ..and so on）。

setupInitialStateStack().//这个方法显然就是从mStateInfo中取出某个状态然后把这个状态的分支放在了mStateStack数组中了。

invokeEnterMethods(0);//这个方法就是调用当前mStateStack中所有状态的enter方法并且把active设为true。

performTransitions();//由于初始状态mDestState为null，所以此时不做任何处理。

然后初始化完成，等待接收新状态了。

此时完成的内容是：

a>把mInitialState即mStableState放到了mStateStack中。

b>调用mStateStack中所有状态的enter方法。本例中则调用类：BluetoothProfileState.StableState的enter方法：

view plainprint?

9.mPendingDevice = null;

9.mPendingDevice = null;

现在BluetoothProfileState这个状态机初始化完了。开始接收消息了。为了跟上面的初始化过程衔接起来，假设我们A2dp的连接请求是初始化之后的第一次请求。回到代码1:

首先构造了一个Message，然后调用状态机的SendMessage方法。回到BluetoothProfileState中并未找到该方法，

说明没有覆盖，则调用的是HierarchicalStateMachine的相同方法：

类HierarchicalStateMachine：

view plainprint?

10.public final void sendMessage(Message msg) {
mHsmHandler.sendMessage(msg);
}

10.public final void sendMessage(Message msg) {
mHsmHandler.sendMessage(msg);
}

调用了Handler的sendMessage，原来是Handler机制，所以，该消息由类HsmHandler的handleMessage()方法来处理此消息：

类：HsmHandler



view plainprint?

11.public final void handleMessage(Message msg) {//msg.what = null
if (mDbg) Log.d(TAG, "handleMessage: E msg.what=" + msg.what);

/** Save the current message */
mMsg = msg;

/**
* Check that construction was completed
*/
if (!mIsConstructionCompleted) {
Log.e(TAG, "The start method not called, ignore msg: " + msg);
return;
}

/**
* Process the message abiding by the hierarchical semantics
* and perform any requested transitions.
*/
processMsg(msg);
performTransitions();

if (mDbg) Log.d(TAG, "handleMessage: X");
}

11.public final void handleMessage(Message msg) {//msg.what = null
if (mDbg) Log.d(TAG, "handleMessage: E msg.what=" + msg.what);

/** Save the current message */
mMsg = msg;

/**
* Check that construction was completed
*/
if (!mIsConstructionCompleted) {
Log.e(TAG, "The start method not called, ignore msg: " + msg);
return;
}

/**
* Process the message abiding by the hierarchical semantics
* and perform any requested transitions.
*/
processMsg(msg);
performTransitions();

if (mDbg) Log.d(TAG, "handleMessage: X");
}

首先设置mMsg为传入msg，调用processMsg();
类 HsmHandler:

view plainprint?

12.private final void processMsg(Message msg) {
StateInfo curStateInfo = mStateStack[mStateStackTopIndex];
if (mDbg) {
Log.d(TAG, "processMsg: " + curStateInfo.state.getName());
}
while (!curStateInfo.state.processMessage(msg)) {//it is true, skip this scope
/**
* Not processed
*/
curStateInfo = curStateInfo.parentStateInfo;
if (curStateInfo == null) {
/**
* No parents left so it's not handled
*/
mHsm.unhandledMessage(msg);
if (isQuit(msg)) {
transitionTo(mQuittingState);
}
break;
}
if (mDbg) {
Log.d(TAG, "processMsg: " + curStateInfo.state.getName());
}
}

/**
* Record that we processed the message
*/
if (curStateInfo != null) {
HierarchicalState orgState = mStateStack[mStateStackTopIndex].state;
mProcessedMessages.add(msg, curStateInfo.state, orgState);
} else {
mProcessedMessages.add(msg, null, null);
}
}

12.private final void processMsg(Message msg) {
StateInfo curStateInfo = mStateStack[mStateStackTopIndex];
if (mDbg) {
Log.d(TAG, "processMsg: " + curStateInfo.state.getName());
}
while (!curStateInfo.state.processMessage(msg)) {//it is true, skip this scope
/**
* Not processed
*/
curStateInfo = curStateInfo.parentStateInfo;
if (curStateInfo == null) {
/**
* No parents left so it's not handled
*/
mHsm.unhandledMessage(msg);
if (isQuit(msg)) {
transitionTo(mQuittingState);
}
break;
}
if (mDbg) {
Log.d(TAG, "processMsg: " + curStateInfo.state.getName());
}
}

/**
* Record that we processed the message
*/
if (curStateInfo != null) {
HierarchicalState orgState = mStateStack[mStateStackTopIndex].state;
mProcessedMessages.add(msg, curStateInfo.state, orgState);
} else {
mProcessedMessages.add(msg, null, null);
}
}

看这句话：curStateInfo.state.processMessage(msg)；curStateInfo是mStateStack中的栈顶元素，从刚才初始化中我们知道，此时mStateStack中装的是mStableState。

因此调用BluetoothProfileState.StableState.processMessage(msg);
类BluetoothProfileState.StableState:



view plainprint?

13.protected boolean processMessage(Message msg) {
if (msg.what != TRANSITION_TO_STABLE) {
transitionTo(mPendingCommandState);//it is the mDestState in HierarchicalStateMachine,the initState is mStableState
}
return true;
}

13.protected
boolean processMessage(Message msg) {
if (msg.what != TRANSITION_TO_STABLE) {
transitionTo(mPendingCommandState);//it is the
mDestState in HierarchicalStateMachine,the initState is mStableState
}
return true;
}

从代码1中知道，msg.what显然不是TRANSITION_TO_STABLE,所以调用了

transitionTo(mPendingCommandState),transitionTo方法在

HierarchicalStateMachine.HsmHandler类中;
把mDestState这个成员变量设为mPendingCommandState。完了。没有了，可能是忘了。

回到代码11，发现还有一个方法没有执行呢。刚才我们处理的是processMsg方法。那么接下来，方法：HierarchicalStateMachine.HsmHandler.performTransitions():

view plainprint?

14.private void performTransitions() { /** * If transitionTo has been called, exit and then enter * the appropriate states. We loop on this to allow * enter and exit methods to use transitionTo. */ HierarchicalState destState = null; while (mDestState != null) { if (mDbg) Log.d(TAG, "handleMessage: new destination call exit"); /** * Save mDestState locally and set to null * to know if enter/exit use transitionTo. */ destState = mDestState; mDestState = null; /** * Determine the states to exit and enter and return the * common ancestor state of the enter/exit states. Then * invoke the exit methods then the enter methods. */ StateInfo commonStateInfo = setupTempStateStackWithStatesToEnter(destState); invokeExitMethods(commonStateInfo); int stateStackEnteringIndex = moveTempStateStackToStateStack(); invokeEnterMethods(stateStackEnteringIndex); /** * Since we have transitioned to a new state we need to have * any deferred messages moved to the front of the message queue * so they will be processed before any other messages in the * message queue. */ moveDeferredMessageAtFrontOfQueue(); }

14.private void performTransitions() {
/**
* If transitionTo has been called, exit and then enter
* the appropriate states. We loop on this to allow
* enter and exit methods to use transitionTo.
*/
HierarchicalState destState = null;
while (mDestState != null) {
if (mDbg) Log.d(TAG, "handleMessage: new destination call exit");

/**
* Save mDestState locally and set to null
* to know if enter/exit use transitionTo.
*/
destState = mDestState;
mDestState = null;

/**
* Determine the states to exit and enter and return the
* common ancestor state of the enter/exit states. Then
* invoke the exit methods then the enter methods.
*/
StateInfo commonStateInfo = setupTempStateStackWithStatesToEnter(destState);
invokeExitMethods(commonStateInfo);
int stateStackEnteringIndex = moveTempStateStackToStateStack();
invokeEnterMethods(stateStackEnteringIndex);

/**
* Since we have transitioned to a new state we need to have
* any deferred messages moved to the front of the message queue
* so they will be processed before any other messages in the
* message queue.
*/
moveDeferredMessageAtFrontOfQueue();
}

记得刚才在状态机初始化的时候见过这个方法，不过好像因为有些原因没有任何处理，是什么原因呢，mDestState为空？
[b]那么这一次这条原因则不成立，因为刚才我们通过transitionTo方法把mDestState已经变成了mPendingCommandState。[/b]

[b]这个方法做了什么呢？

>把mDestState赋给临时状态，然后置空；

>调用setupTempStateStackWithStatesToEnter(destState)；把destState的状态分支都放到临时状态栈：mTempStateStack中，[/b]

[b]刚才忘记说了，这个临时栈跟mStateStack结构相同，每次有不同时序的处理的时候起到辅助作用，[/b]

[b]由于在本例中作用不明显未做深究；

>然后mStateStack中的所有状态调用exit方法，active置为false；

>调用mStateStack中所有状态的enter方法。本例中即调用：BluetoothProfileState.PendingCommandState.enter()：[/b]

view plainprint?

15.protected void enter() { log("Entering PendingCommandState State"); dispatchMessage(getCurrentMessage());// log(getCurrentMessage()); }

15.protected void enter() {
log("Entering PendingCommandState State");
dispatchMessage(getCurrentMessage());//
log(getCurrentMessage());
}

getCurrentMessage得到的就是mMsg即 代码11 中传入的消息，也就是本次调用要处理的消息。然后调用BluetoothProfileState.PendingCommandState.dispatchMessage()：



view plainprint?

16.private void dispatchMessage(Message msg) { BluetoothDeviceProfileState deviceProfileMgr = (BluetoothDeviceProfileState)msg.obj; int cmd = msg.arg1; if (mPendingDevice == null || mPendingDevice.equals(deviceProfileMgr.getDevice())) { mPendingDevice = deviceProfileMgr.getDevice(); deviceProfileMgr.sendMessage(cmd); } else { Message deferMsg = new Message(); deferMsg.arg1 = cmd; deferMsg.obj = deviceProfileMgr; deferMessage(deferMsg); } }

16.private void dispatchMessage(Message msg) {
BluetoothDeviceProfileState deviceProfileMgr =
(BluetoothDeviceProfileState)msg.obj;
int cmd = msg.arg1;
if (mPendingDevice == null || mPendingDevice.equals(deviceProfileMgr.getDevice())) {
mPendingDevice = deviceProfileMgr.getDevice();
deviceProfileMgr.sendMessage(cmd);
} else {
Message deferMsg = new Message();
deferMsg.arg1 = cmd;
deferMsg.obj = deviceProfileMgr;
deferMessage(deferMsg);
}
}

首先，BluetoothDeviceProfileState deviceProfileMgr；回到代码1 查看msg.obj发现这个变量的值为

mDeviceProfileState.get(address);

address则为要连接音频设备的mac，mDeviceProfileState是一个hashMap储存了不同设备mac地址及设备协议状态。

它的初始化及赋值什么的都在BluetoothService中，这个hashMap里的值来自两个方面：

一是在蓝牙打开时把已经配对的设备，加到里面；

二是配对成功后，把相应设备加到这个hashMap中。

因此，每个设备对应了一个deviceProfileMgr，即BluetoothDeviceProfileState，这也是个状态机，先放在这里，我们等会再看。

接下来判断：if (mPendingDevice == null || mPendingDevice.equals(deviceProfileMgr.getDevice()))

哦，记得之前把mPengingDevice置为空过，在 代码 9中：

[b]判断条件符合mPendingDevice置为当前device，然后deviceProfileMgr.sendMessage(cmd);[/b]

然后，没了。

只剩下了 deviceProfileMgr.sendMessage(cmd);

不过这个形式挺眼熟，之前好想见到过，见 代码1.

mA2dpProfileState.sendMessage(msg);

只不过mA2dpProfileState换成了deviceProfileMgr，状态机由BluetoothProfileState换成了BluetoothDeviceProfileState。

其他的真的一点变化都没有，于是也就不用分析了。跟上面一样。

追到最后发现，其实connectSink方法最后调用：

BluetoothA2dpService.connectSinkInternal(BluetoothDevice device)



-------------------------------------------------

总结：状态机发送一个msg，通过父类HierarchicalStateMachine的成员handler中的handleMessage处理，

然后调用processMsg，实际调用的是mStateStack中存储的state的processMessage方法，然后调用performTransitions方法，

检测mDestState是否为空，不为空则将当前mStateStack的状态全部调用exit方法且avtive置为false，

然后把mDestState的状态分支都放进mStateStack中，调用这些状态的enter方法，且把active置为true。

更简单一点就是，状态机发送msg--->mStateStack--当前状态分支的exit方法--->调用mDestState状态分支的enter方法

-->其中msg里面的参数，会存在mMsg中可以通过getCurrentMsg来取得之。