深入解析FragmentTransaction的源码实现机制

谈到fragment的使用&＃xff0c;肯定绕不过FragmentTransaction事务&＃xff0c;对fragment的操作必定用到它&＃xff0c;其提供show,hide,add,remove,replace等常用的fragment操作&＃xff0c;最后commit操作&＃xff0c;这么强大的管理类&＃xff0c;它内部是如何实现的呢?为什么可以连续调用多个api&＃xff0c;最后一次要commit操作&＃xff1f;

1、创建FragmentTransaction对象&＃xff1a;

FragmentTransaction ft &＃61; mFragmentManager.beginTransaction();

源码实现&＃xff1a;

&＃64;Overridepublic FragmentTransaction beginTransaction() {return new BackStackRecord(this);}

可以看到new 出一个新的 BackStackRecord对象&＃xff0c;这个BackStackRecord就是api的真正实现者。

我以remove fragment这个操作看下内部实现:

&＃64;Overridepublic FragmentTransaction remove(Fragment fragment) {Op op &＃61; new Op();op.cmd &＃61; OP_REMOVE;op.fragment &＃61; fragment;addOp(op);return this;}

这个Op是什么东东呢&＃xff1f;&＃xff0c;又把op加入到addOp中&＃xff0c;我们不妨先看下这个addOp的实现&＃xff1a;

void addOp(Op op) {// 创建空链表&＃xff0c;头指针和尾指针指向第一个结点if (mHead &＃61;&＃61; null) {mHead &＃61; mTail &＃61; op;} else {// 新结点加入到尾部op.prev &＃61; mTail;mTail.next &＃61; op;mTail &＃61; op;}op.enterAnim &＃61; mEnterAnim;op.exitAnim &＃61; mExitAnim;op.popEnterAnim &＃61; mPopEnterAnim;op.popExitAnim &＃61; mPopExitAnim;mNumOp&＃43;&＃43;;}

这段代码还是比较容易理解的&＃xff0c;其实remove操作&＃xff0c;就是新建一个 链表结点&＃xff0c;结点中保存了当前的操作&＃xff0c;保存了fragment的对象引用&＃xff0c;所以 FragmentTransaction的内部主要实现就是通过链表操作的&＃xff0c;每个链表结点保存了每一个api操作的信息&＃xff0c;好了&＃xff0c;我们看下commit的实现&＃xff1a;

&＃64;Overridepublic int commit() {return commitInternal(false);}

只能commit一次&＃xff0c;否则抛出异常

int commitInternal(boolean allowStateLoss) {if (mCommitted) throw new IllegalStateException("commit already called");if (FragmentManagerImpl.DEBUG) {Log.v(TAG, "Commit: " &＃43; this);LogWriter logw &＃61; new LogWriter(TAG);PrintWriter pw &＃61; new PrintWriter(logw);dump(" ", null, pw, null);}mCommitted &＃61; true;if (mAddToBackStack) {mIndex &＃61; mManager.allocBackStackIndex(this);} else {mIndex &＃61; -1;}mManager.enqueueAction(this, allowStateLoss);return mIndex;}

看下 mManager.enqueueAction(this, allowStateLoss)的实现&＃xff1a;

/*** Adds an action to the queue of pending actions.** &＃64;param action the action to add* &＃64;param allowStateLoss whether to allow loss of state information* &＃64;throws IllegalStateException if the activity has been destroyed*/public void enqueueAction(Runnable action, boolean allowStateLoss) {if (!allowStateLoss) {checkStateLoss();}synchronized (this) {if (mDestroyed || mHost &＃61;&＃61; null) {throw new IllegalStateException("Activity has been destroyed");}if (mPendingActions &＃61;&＃61; null) {mPendingActions &＃61; new ArrayList();}mPendingActions.add(action);if (mPendingActions.size() &＃61;&＃61; 1) {mHost.getHandler().removeCallbacks(mExecCommit);mHost.getHandler().post(mExecCommit);}}}

代码解读&＃xff1a;
1、先去检查 checkStateLoss 状态&＃xff0c;正常commit操作&＃xff0c;是肯定去检查的&＃xff0c;如果是在onSaveInstanceState状态保存之后&＃xff0c;再去commit操作&＃xff0c;肯定会报错。

2、BackStackRecord 类实现了 Runnable,看其run方法。

public void run() {// 省略部分代码// 指向链表头指针Op op &＃61; mHead;// 遍历链表结点while (op !&＃61; null) {//每一个结点的cmd操作switch (op.cmd) {case OP_ADD: {Fragment f &＃61; op.fragment;f.mNextAnim &＃61; op.enterAnim;mManager.addFragment(f, false);}break;case OP_REPLACE: {Fragment f &＃61; op.fragment;int containerId &＃61; f.mContainerId;if (mManager.mAdded !&＃61; null) {for (int i &＃61; mManager.mAdded.size() - 1; i >&＃61; 0; i--) {Fragment old &＃61; mManager.mAdded.get(i);if (FragmentManagerImpl.DEBUG) {Log.v(TAG,"OP_REPLACE: adding&＃61;" &＃43; f &＃43; " old&＃61;" &＃43; old);}if (old.mContainerId &＃61;&＃61; containerId) {if (old &＃61;&＃61; f) {op.fragment &＃61; f &＃61; null;} else {if (op.removed &＃61;&＃61; null) {op.removed &＃61; new ArrayList();}op.removed.add(old);old.mNextAnim &＃61; op.exitAnim;if (mAddToBackStack) {old.mBackStackNesting &＃43;&＃61; 1;if (FragmentManagerImpl.DEBUG) {Log.v(TAG, "Bump nesting of "&＃43; old &＃43; " to " &＃43; old.mBackStackNesting);}}mManager.removeFragment(old, mTransition, mTransitionStyle);}}}}if (f !&＃61; null) {f.mNextAnim &＃61; op.enterAnim;mManager.addFragment(f, false);}}break;case OP_REMOVE: {Fragment f &＃61; op.fragment;f.mNextAnim &＃61; op.exitAnim;mManager.removeFragment(f, mTransition, mTransitionStyle);}break;case OP_HIDE: {Fragment f &＃61; op.fragment;f.mNextAnim &＃61; op.exitAnim;mManager.hideFragment(f, mTransition, mTransitionStyle);}break;case OP_SHOW: {Fragment f &＃61; op.fragment;f.mNextAnim &＃61; op.enterAnim;mManager.showFragment(f, mTransition, mTransitionStyle);}break;case OP_DETACH: {Fragment f &＃61; op.fragment;f.mNextAnim &＃61; op.exitAnim;mManager.detachFragment(f, mTransition, mTransitionStyle);}break;case OP_ATTACH: {Fragment f &＃61; op.fragment;f.mNextAnim &＃61; op.enterAnim;mManager.attachFragment(f, mTransition, mTransitionStyle);}break;default: {throw new IllegalArgumentException("Unknown cmd: " &＃43; op.cmd);}}//指向下一个结点op &＃61; op.next;}mManager.moveToState(mManager.mCurState, mTransition,mTransitionStyle, true);// 是否调用了addBackStack()&＃xff0c;如果加入&＃xff0c;将当前的BackStackRecord加入到栈中if (mAddToBackStack) {mManager.addBackStackState(this);}}

这段代码解读&＃xff1a;

1、每次commit之后&＃xff0c;都会commit一个runnable任务&＃xff0c;run()方法里面 对链表进行遍历操作&＃xff0c;从头结点开始&＃xff0c;依次访问每个结点&＃xff0c;然后读取里面每一个结点的cmd&＃xff0c;执行相应的方法。注意replace,是先remove&＃xff0c;后add。
2、如果调用了addBackStack()方法&＃xff0c;会将当前的任务对象加入到栈中。

void addBackStackState(BackStackRecord state) {if (mBackStack &＃61;&＃61; null) {mBackStack &＃61; new ArrayList();}mBackStack.add(state);reportBackStackChanged();}


可见我的这篇博客 从源码角度解释 fragment 坑&＃xff08;一&＃xff09;

我在想&＃xff0c;FragmentTransaction 事务内部为什么要实现链表操作呢&＃xff1f;

如果单纯的只是commit之前调用几个api&＃xff0c;hide&＃xff0c;show我认为可能没有必要&＃xff0c;它的必要性&＃xff0c;就是在退栈的时候&＃xff0c;能够记住你之前的每个结点的行为&＃xff0c;并且进行相应的反操作&＃xff0c;比如我 先用事务&＃xff0c;hide 1,remove 2,add 3,甚至更多&＃xff0c;退栈的时候&＃xff0c;我们是不是先执行&＃xff0c;remove 3,往往3走到onDestroyview方法&＃xff0c;2 调用 onCreateView&＃xff0c;show 1出来&＃xff0c;如果不用链表的操作的话&＃xff0c;确实不太方便。