Android技术周报190310期 —— onSaveInstanceState到底做了些什么？

摘要：与调用时机关于调用时机的追踪，我们先来简单看一下一个的启动过程的启动过程与进程间通信机制息息相关，对于进程间通信机制以及的启动过程，我们在后期的文章里会仔细说明，这里就不在做过多说明。

Android技术周报是最近我刚创建的一个文集，这篇文章也是Android技术周报这个文集的第一篇文章（发现自己好久没写文章了），主要目的是记录和分享一些工作中遇到的问题，并对其进行深入研究，对于描述的内容有异议的可以留言，大家一起交流和学习。今天的内容是从源码角度了解Activity中的 onSaveInstanceState 和 onRestoreInstanceState：

protected void onSaveInstanceState(Bundle outState) {
    // 存储窗口视图状态
    outState.putBundle(WINDOW_HIERARCHY_TAG, mWindow.saveHierarchyState());

    outState.putInt(LAST_AUTOFILL_ID, mLastAutofillId);
    // 存储Fragment状态
    Parcelable p = mFragments.saveAllState();
    if (p != null) {
        outState.putParcelable(FRAGMENTS_TAG, p);
    }
    if (mAutoFillResetNeeded) {
        outState.putBoolean(AUTOFILL_RESET_NEEDED, true);
        // 存储Android自动填充的状态，非重点
        getAutofillManager().onSaveInstanceState(outState);
    }
    // 调用ActivityLifeCallbacks的onSaveInstanceState方法进行状态存储
    getApplication().dispatchActivitySaveInstanceState(this, outState);
}

从源码上看，不难发现onSaveInstanceState主要做了以下三个内容：

存储窗口视图状态
存储Fragment状态
调用ActivityLifeCallbacks的onSaveInstanceState方法进行状态存储

通过调用PhoneWindow的saveHierarchyState方法实际上是调用mContentParent的saveHierarchyState方法：

@Override
public Bundle saveHierarchyState() {
    Bundle outState = new Bundle();
    if (mContentParent == null) {
        return outState;
    }

    SparseArray states = new SparseArray();
    mContentParent.saveHierarchyState(states);
    outState.putSparseParcelableArray(VIEWS_TAG, states);
    ....
   return outState;
}

其中mContentParent是Activity的主视图，本质是一个View对象，在Activity#setContentView()方法里进行初始化操作，接下来回到View的saveHierarchyState方法：

public void saveHierarchyState(SparseArray container) {
    dispatchSaveInstanceState(container);
}

protected void dispatchSaveInstanceState(SparseArray container) {
    if (mID != NO_ID && (mViewFlags & SAVE_DISABLED_MASK) == 0) {
        mPrivateFlags &= ~PFLAG_SAVE_STATE_CALLED;
        Parcelable state = onSaveInstanceState();
        if ((mPrivateFlags & PFLAG_SAVE_STATE_CALLED) == 0) {
            throw new IllegalStateException(
                    "Derived class did not call super.onSaveInstanceState()");
        }
        if (state != null) {
            // 以当前View的ID为key值存储起来
            container.put(mID, state);
        }
    }
}

在View的dispatchSaveInstanceState方法里，会将View#onSaveInstanceState()方法返回的Parcelable对象以当前View的ID为key值存储起来，并返回，这也意味着如果当前View没有设置ID，那么将无法进行View的onSaveInstanceState操作：

android:id="@+id/tv_save_state"

View的onSaveInstanceState方法默认返回空状态 BaseSavedState.EMPTY_STATE，可交由子View重写，如TextView#onSaveInstanceState()：

@Override
public Parcelable onSaveInstanceState() {
    Parcelable superState = super.onSaveInstanceState();
    // Save state if we are forced to
    final boolean freezesText = getFreezesText();
    boolean hasSelection = false;
    int start = -1;
    int end = -1;
    if (mText != null) {
        start = getSelectionStart();
        end = getSelectionEnd();
        if (start >= 0 || end >= 0) {
            // Or save state if there is a selection
            hasSelection = true;
        }
    }
    if (freezesText || hasSelection) {
        SavedState ss = new SavedState(superState);
        if (freezesText) {
            if (mText instanceof Spanned) {
                final Spannable sp = new SpannableStringBuilder(mText);
                if (mEditor != null) {
                    removeMisspelledSpans(sp);
                    sp.removeSpan(mEditor.mSuggestionRangeSpan);
                }
                ss.text = sp;
            } else {
                ss.text = mText.toString();
            }
        }
        if (hasSelection) {
            // XXX Should also save the current scroll position!
            ss.selStart = start;
            ss.selEnd = end;
        }
        if (isFocused() && start >= 0 && end >= 0) {
            ss.frozenWithFocus = true;
        }
        ss.error = getError();
        if (mEditor != null) {
            ss.editorState = mEditor.saveInstanceState();
        }
        return ss;
    }
    return superState;
}

接下来进入Fragment状态的保存方法，mFragments.saveAllState()，mFragments是一个FragmentController对象，主要用于提供FragmentManager对象，维护Fragments的生命周期，这里不做过多介绍，所以最终是调用FragmentManager#saveAllState()方法，该方法最终会返回一个FragmentManagerState对象：

Parcelable saveAllState() {
    ...
    // 第一步：收集所有处于active状态的Fragment
    int N = mActive.size();
    FragmentState[] active = new FragmentState[N];
    boolean haveFragments = false;
    for (int i=0; i Fragment.INITIALIZING && fs.mSavedFragmentState == null) {
                fs.mSavedFragmentState = saveFragmentBasicState(f);
                ...
            } else {
                fs.mSavedFragmentState = f.mSavedFragmentState;
            }
        }
    }

    // 第二步：收集已经添加的Fragments的Index
    int[] added = null;
    BackStackState[] backStack = null;
    N = mAdded.size();
    if (N > 0) {
        added = new int[N];
        for (int i=0; i 0) {
            backStack = new BackStackState[N];
            for (int i=0; i
对于第一步的方法最终会调用FragmentManager#saveFragmentBasicState()方法，此方法主要有以下任务：
Bundle saveFragmentBasicState(Fragment f) {
    Bundle result = null;
    ...
   // 调用Fragment#onSaveInstanceState()方法（可由子类重写）
    f.performSaveInstanceState(mStateBundle);
    dispatchOnFragmentSaveInstanceState(f, mStateBundle, false);
    ...
    // 保存View的视图状态，跟上面内容介绍一致
    if (f.mView != null) {
        saveFragmentViewState(f);
    }
    // mSavedViewState为已保存的视图状态
    if (f.mSavedViewState != null) {
        if (result == null) {
            result = new Bundle();
        }
        result.putSparseParcelableArray(
                FragmentManagerImpl.VIEW_STATE_TAG, f.mSavedViewState);
    }
    // 记录Fragment的显示属性
    if (!f.mUserVisibleHint) {
        if (result == null) {
            result = new Bundle();
        }
        // Only add this if it"s not the default value
        result.putBoolean(FragmentManagerImpl.USER_VISIBLE_HINT_TAG, f.mUserVisibleHint);
    }
    return result;
}
Fragment的performSaveInstanceState除了保存自身状态还会保存子Fragment的状态，实现如下：
void performSaveInstanceState(Bundle outState) {
    onSaveInstanceState(outState);
    if (mChildFragmentManager != null) {
        Parcelable p = mChildFragmentManager.saveAllState();
        if (p != null) {
            outState.putParcelable(Activity.FRAGMENTS_TAG, p);
        }
    }
}
最后，看一下Activity#onSaveInstanceState的最后一步：
getApplication().dispatchActivitySaveInstanceState(this, outState);

实际上是遍历已注册的ActivityLifecycleCallback，并调用其onActivitySaveInstanceState()方法：
void dispatchActivitySaveInstanceState(Activity activity, Bundle outState) {
    Object[] callbacks = collectActivityLifecycleCallbacks();
    if (callbacks != null) {
        for (int i=0; i
至此，Activity#onSaveInstanceState的工作已经做完了，总结起来，其时序图如下：

讲解过程省略了不少代码，需要做详细了解的可以进一步查看源码，关于onRestoreInstanceState的方法实现，主要是以上方法的一个逆过程，这里不再做过多的叙述，接下来内容主要讲onSaveInstanceState与onRestoreInstanceState的调用时机。
2. onSaveInstanceState与onRestoreInstanceState调用时机
关于onSaveInstanceState调用时机的追踪，我们先来简单看一下一个Activity的启动过程：

Activity的启动过程与Binder进程间通信机制息息相关，对于Binder进程间通信机制以及Activity的启动过程，我们在后期的文章里会仔细说明，这里就不在做过多说明。Activity的启动过程主要的步骤可以概括如下：
MainActivity向ActivityManagerService发送一个启动SecondActivity的进程间通信请求；
ActivityManagerService保存SecondActivity的相关信息，并向MainActivity发送一个进入中止状态（pause）的进程间通信请求；
MainActivity进入中止状态后，向ActivityManagerService发送一个已进入中止状态的进程间通信请求；
ActivityManagerService发现SecondActivity所在进程不存在，则会发起启动新进程的请求；
当新进程启动完毕后，会向SecondActivity发送一个启动完成的进程间通信请求；
ActivityManagerService将保存的SecondActivity信息发送给新创建的进程，以便其将SecondActivity启动起来

我们主要看第三步，Activity接收ActivityManagerService的进程间通信请求最终都会由该Activity所在的进程对象ActivityThread里的Handler对象去处理：
// ActivityThread.java
private class H extends Handler {
    public void handleMessage(Message msg) {
        if (DEBUG_MESSAGES) Slog.v(TAG, ">>> handling: " + codeToString(msg.what));
        switch (msg.what) {
            ...
            case PAUSE_ACTIVITY: {
                // 通知Activity进入中止状态
                Trace.traceBegin(Trace.TRACE_TAG_ACTIVITY_MANAGER, "activityPause");
                SomeArgs args = (SomeArgs) msg.obj;
                handlePauseActivity((IBinder) args.arg1, false,
                        (args.argi1 & USER_LEAVING) != 0, args.argi2,
                        (args.argi1 & DONT_REPORT) != 0, args.argi3);
                Trace.traceEnd(Trace.TRACE_TAG_ACTIVITY_MANAGER);
            } break;
            ...
        }
}
核心方法为handlePauseActivity：
private void handlePauseActivity(IBinder token, boolean finished,
        boolean userLeaving, int configChanges, boolean dontReport, int seq) {
    ActivityClientRecord r = mActivities.get(token);
    if (DEBUG_ORDER) Slog.d(TAG, "handlePauseActivity " + r + ", seq: " + seq);
    if (!checkAndUpdateLifecycleSeq(seq, r, "pauseActivity")) {
        return;
    }
    if (r != null) {
        ...
        r.activity.mConfigChangeFlags |= configChanges;
        performPauseActivity(token, finished, r.isPreHoneycomb(), "handlePauseActivity");
        ...
    }
}
接着在往下看performPauseActivity：
final Bundle performPauseActivity(IBinder token, boolean finished,
        boolean saveState, String reason) {
    ActivityClientRecord r = mActivities.get(token);
    return r != null ? performPauseActivity(r, finished, saveState, reason) : null;
}

final Bundle performPauseActivity(ActivityClientRecord r, boolean finished,
        boolean saveState, String reason) {
    ...
    if (finished) {
        r.activity.mFinished = true;
    }

    // Next have the activity save its current state and managed dialogs...
    if (!r.activity.mFinished && saveState) {
        callCallActivityOnSaveInstanceState(r);
    }
   ...
}
这时候关键方法已经显而易见了：
private void callCallActivityOnSaveInstanceState(ActivityClientRecord r) {
    r.state = new Bundle();
    r.state.setAllowFds(false);
    if (r.isPersistable()) {
        r.persistentState = new PersistableBundle();
        mInstrumentation.callActivityOnSaveInstanceState(r.activity, r.state,
                r.persistentState);
    } else {
        mInstrumentation.callActivityOnSaveInstanceState(r.activity, r.state);
    }
}

public void callActivityOnSaveInstanceState(Activity activity, Bundle outState,
        PersistableBundle outPersistentState) {
    activity.performSaveInstanceState(outState, outPersistentState);
}

final void performSaveInstanceState(Bundle outState, PersistableBundle outPersistentState) {
    // 这就是我们希望看到的方法
    onSaveInstanceState(outState, outPersistentState);
    saveManagedDialogs(outState);
    storeHasCurrentPermissionRequest(outState);
    if (DEBUG_LIFECYCLE) Slog.v(TAG, "onSaveInstanceState " + this + ": " + outState +
            ", " + outPersistentState);
}

总结起来，onSaveInstanceState会在以下情况下被调用：
当用户按下home键；
长按home键切换应用程序；
按下电源键；
启动一个新的Activity；
屏幕方向切换；
电话打入等情况发生

而onRestoreInstanceState只有在activity确实是被系统回收，重新创建activity的情况下才会被调用，所以主要分为以下两种情况：
a. activity没有被销毁：
onPause -> onSaveInstanceState -> onStop -> onRestart -> onStart -> onResume
b. activity被销毁：
onPause -> onSaveInstanceState -> onStop -> onDestroy -> onCreate -> onStart -> onRestoreInstanceState -> onResume
OK，今天的内容就到这里，下篇文章见 ~