Flutter事件分发源码剖析

概述

不管是原生Android、iOS还是JavaScript，只要是涉及手势交互都会有事件的分发处理。和原生Android、iOS的事件分发的步骤和原理一样，Flutter的事件分发总体也由手势触发、拦截和响应等几个部分构成。Flutter所有事件源头是 hooks.dart文件的_dispatchPointerDataPacket函数，通过拦截屏幕的点击、滑动等各种事件，进而分发给原生代码进行响应（ps：Android事件分发）。

如果你看过了解原生Android、iOS的事件分发机制，那么Flutter的事件分发，其实是在Android和iOS上加了壳，即Flutter的事件分发是在原生Android、iOS的的事件分发上进行包装的（Android - C - Dart，iOS- C -Dart）。其中，C是Flutter的底层engine，负责Flutter上层和原生Android、iOS系统的交互。

事件分发到Dart的入口类是GestureBinding类，此类位于gestures/binding.dart文件中，与手势识别相关的都位于gestures包中，如下图所示。

• converter.dart将物理坐标_dispatchPointerDataPacket收到的物理数据PointerDataPacket转换成PointerEvent， 类似于安卓在ViewRootImpl.java将InputEventReceiver收到的InputEvent转换为MotionEvent。
• recognizer.dart的GestureRecognizer是所有手势识别的基类。
• rendering/binding.dart的RendererBinding类关联了render树和Flutter引擎，等价于安卓的Surface。
• view.dart的RenderView是render树的根节点，等价于安卓的DecorView。

Flutter的事件分发基类是GestureBinding，打开GestureBinding类，它的成员函数包括dispatchEvent、handleEvent和hitTes等，主要是从事件队列里按照先入先出方式处理PointerEvent，源码如下。

mixin GestureBinding on BindingBase implements HitTestable, HitTestDispatcher, HitTestTarget {
  @override
  void initInstances() {
    super.initInstances();
    _instance = this;
    ui.window.onPointerDataPacket = _handlePointerDataPacket;
  }

其中，WidgetsFlutterBinding.ensureInitialized()函数的作用就是初始化各个binging。

Flutter 事件分发

和Android、iOS类似，Flutter的事件分发的入口在runApp函数，相关的代码如下。

void runApp(Widget app) {
  WidgetsFlutterBinding.ensureInitialized()
    ..attachRootWidget(app)
    ..scheduleWarmUpFrame();
}
class WidgetsFlutterBinding extends BindingBase with GestureBinding, ServicesBinding, SchedulerBinding, PaintingBinding, SemanticsBinding, RendererBinding, WidgetsBinding {
  static WidgetsBinding ensureInitialized() {
    if (WidgetsBinding.instance == null)
      WidgetsFlutterBinding();
    return WidgetsBinding.instance;
  }
}
 
void attachRootWidget(Widget rootWidget) {
  _renderViewElement = RenderObjectToWidgetAdapter<RenderBox>(
    container: renderView,
    debugShortDescription: '[root]',
    child: rootWidget
  ).attachToRenderTree(buildOwner, renderViewElement);
}

WidgetsFlutterBinding.ensureInitialized()函数的作用是初始化各个binging。事实上，Flutter 中的 WidgetsFlutterBinding的 Binding可以分为GestureBinding、ServicesBinding、SchedulerBinding、PaintingBinding、SemanticsBinding、RendererBinding、WidgetsBinding 等 7 种 Binding，它们都有自己在功能上的划分。其中，GestureBinding就是处理事件分发的，attachRootWidget就是设置根节点， 可以看到真正的根节点是renderview， 也是Flutter事件分发的起点。

下面我们来重点看一下GestureBinding类。

GestureBinding

和Android事件处理的流程一样，首先，系统会拦截用户的事件，然后在使用GestureBinding的_handlePointerEvent进行事件命中处理。原生事件到达Dart层之后调用的第一个方法是_handlePointerDataPacket，它的源码如下。

 void _handlePointerDataPacket(ui.PointerDataPacket packet) {
    _pendingPointerEvents.addAll(PointerEventConverter.expand(packet.data, window.devicePixelRatio));
 
    if (!locked)
      _flushPointerEventQueue();
  }

_handlePointerDataPacket方法有一个PointerEventConverter类，作用是将原生传来的手势数据全部转化为Dart对应的对象保存数据，然后保存到集合中进行储存。接下来来我们看一下_flushPointerEventQueue方法，源码如下。

void _flushPointerEventQueue() {
    assert(!locked);
    while (_pendingPointerEvents.isNotEmpty)
      _handlePointerEvent(_pendingPointerEvents.removeFirst());
  }

_flushPointerEventQueue方法的作用就是循环处理每个手指的的事件，并进行处理，源码如下。

void _handlePointerEvent(PointerEvent event) {
    assert(!locked);
    HitTestResult hitTestResult;
    //如果是手指按下的话
    if (event is PointerDownEvent || event is PointerSignalEvent) {
      assert(!_hitTests.containsKey(event.pointer));
      hitTestResult = HitTestResult();
      //得到碰撞的控件组
      hitTest(hitTestResult, event.position);
      if (event is PointerDownEvent) {
        _hitTests[event.pointer] = hitTestResult;
      }
      assert(() {
        if (debugPrintHitTestResults)
          debugPrint('$event: $hitTestResult');
        return true;
      }());
    }
    //手指抬起
    else if (event is PointerUpEvent || event is PointerCancelEvent) {
      hitTestResult = _hitTests.remove(event.pointer);
    }
    //缓存点击的事件，接下来发生滑动的时候直接复用原来的碰撞控件组
    else if (event.down) {
      // Because events that occur with the pointer down (like
      // PointerMoveEvents) should be dispatched to the same place that their
      // initial PointerDownEvent was, we want to re-use the path we found when
      // the pointer went down, rather than do hit detection each time we get
      // such an event.
      hitTestResult = _hitTests[event.pointer];
    }
    assert(() {
      if (debugPrintMouseHoverEvents && event is PointerHoverEvent)
        debugPrint('$event');
      return true;
    }());
    if (hitTestResult != null ||
        event is PointerHoverEvent ||
        event is PointerAddedEvent ||
        event is PointerRemovedEvent) {
      dispatchEvent(event, hitTestResult);
    }
  }

这个方法的主要目的就是得到HitTestResult，就是根据按下的坐标位置找出view树中哪些控件在点击的范围内，手指在移动和抬起的时候都复用当前的事件，区别在于不同的手指有不同的索引值。接下来，看一下用户的触摸行为，hitTest首先会进入RendererBinding处理，打开RendererBinding类的hitTest方法，如下所示。

RenderView get renderView => _pipelineOwner.rootNode as RenderView;

void hitTest(HitTestResult result, Offset position) {
    assert(renderView != null);
    renderView.hitTest(result, position: position);
    super.hitTest(result, position);
  }

其中，RenderView可以理解为Flutter 视图树的根View，在Flutter中也叫做Widget ，一个Widget 对应一个Element 。在Flutter中，渲染会三棵树，即Widget 树、Element 树和RenderObject 树。我们进行页面布局分析时，就可以看到它们，如下所示。

关于Widget 树、Element 树和RenderObject 树，可以查看Flutter渲染之Widget、Element 和 RenderObject的介绍。

然后，我们打开renderView.hitTest方法，对应的代码如下所示。

 bool hitTest(HitTestResult result, { Offset position }) {
    if (child != null)
      child.hitTest(BoxHitTestResult.wrap(result), position: position);
    result.add(HitTestEntry(this));
    return true;
  }

可以看到，根视图是先从子view开始放进集合，放完子view再放自己，这和前端JS点击事件冒泡的原理是一样的。并且，只有满足条件子视图才会放到 入RenderBox 的这个方法中。

 bool hitTest(BoxHitTestResult result, { @required Offset position }) {
    //所点击的范围是否在当前控件的范围内
    if (_size.contains(position)) {
    //先添加孩子中的事件后选人
      if (hitTestChildren(result, position: position) || hitTestSelf(position)) {
        result.add(BoxHitTestEntry(this, position));
        return true;
      }
    }
    return false;
  }

接下来，看一下Stack小部件hitTestChildren的实现，源码如下。

  @override
  bool hitTestChildren(BoxHitTestResult result, { Offset position }) {
    return defaultHitTestChildren(result, position: position);
  }

bool defaultHitTestChildren(BoxHitTestResult result, { Offset position }) {
    // the x, y parameters have the top left of the node's box as the origin

    ChildType child = lastChild;
    while (child != null) {
      final ParentDataType childParentData = child.parentData;
      final bool isHit = result.addWithPaintOffset(
        offset: childParentData.offset,
        position: position,
        hitTest: (BoxHitTestResult result, Offset transformed) {
          assert(transformed == position - childParentData.offset);
          return child.hitTest(result, position: transformed);
        },
      );
      if (isHit)
        return true;
      child = childParentData.previousSibling;
    }
    return false;
  }

这个方法的作用就是判断包含Padding的视图是否在点击范围内，如果命中，则阻止其他事件继续冒泡。看到此处，我们大体可以看出，Flutter的事件处理主要是判断点击的坐标知否在控件范围内，如果在范围内直接响应，如果不在继续向上冒泡，并且事件是从叶子开始的，也即Web中的事件冒泡。

完成命中处理后，接下来回到事件处理的主流程，即事件派发dispatchEvent，代码位于gestrues/binding里面，源码如下。

 void dispatchEvent(PointerEvent event, HitTestResult hitTestResult) {
    assert(!locked);
    // No hit test information implies that this is a hover or pointer
    // add/remove event.这种情况出在指针悬停屏幕上方，微微接触或不接触，是手机敏感而言
    if (hitTestResult == null) {
      assert(event is PointerHoverEvent || event is PointerAddedEvent || event is PointerRemovedEvent);
      try {
        pointerRouter.route(event);
      } catch (exception, stack) {
        FlutterError.reportError(FlutterErrorDetailsForPointerEventDispatcher(
          exception: exception,
          stack: stack,
          library: 'gesture library',
          context: ErrorDescription('while dispatching a non-hit-tested pointer event'),
          event: event,
          hitTestEntry: null,
          informationCollector: () sync* {
            yield DiagnosticsProperty<PointerEvent>('Event', event, style: DiagnosticsTreeStyle.errorProperty);
          },
        ));
      }
      return;
    }
 
    for (HitTestEntry entry in hitTestResult.path) {
      try {
        entry.target.handleEvent(event.transformed(entry.transform), entry);
      } catch (exception, stack) {
        FlutterError.reportError(FlutterErrorDetailsForPointerEventDispatcher(
          exception: exception,
          stack: stack,
          library: 'gesture library',
          context: ErrorDescription('while dispatching a pointer event'),
          event: event,
          hitTestEntry: entry,
          informationCollector: () sync* {
            yield DiagnosticsProperty<PointerEvent>('Event', event, style: DiagnosticsTreeStyle.errorProperty);
            yield DiagnosticsProperty<HitTestTarget>('Target', entry.target, style: DiagnosticsTreeStyle.errorProperty);
          },
        ));
      }
    }
  }

此方法最根本的作用是循环事件分发，并以冒泡的形式从底部到分发事件，当事件被命中时，即由当前子节点处理事件，这和Android的事件分发的逻辑是一样的。下面以GestureDetector和Listener来举例事件分发的不同。如果用Listener的话，Listener的组件最终对应的RenderObject是RenderPointerListener，它的监测当前点击是否命中的方法如下。

bool hitTest(BoxHitTestResult result, { Offset position }) {
    bool hitTarget = false;
    if (size.contains(position)) {
      hitTarget = hitTestChildren(result, position: position) || hitTestSelf(position);
      if (hitTarget || behavior == HitTestBehavior.translucent)
        result.add(BoxHitTestEntry(this, position));
    }
    return hitTarget;
  }
 
  @override
  bool hitTestSelf(Offset position) => behavior == HitTestBehavior.opaque;

使用Listener嵌套的子组件默认情况下是命中的，很多子部件例如Text、Image等，它们的hitTestSelf返回True，假如我们为Text嵌套了Listener，那么事件分发的时候设计的代码如下所示。

void handleEvent(PointerEvent event, HitTestEntry entry) {
    assert(debugHandleEvent(event, entry));
    if (onPointerDown != null && event is PointerDownEvent)
      return onPointerDown(event);
    if (onPointerMove != null && event is PointerMoveEvent)
      return onPointerMove(event);
    if (onPointerUp != null && event is PointerUpEvent)
      return onPointerUp(event);
    if (onPointerCancel != null && event is PointerCancelEvent)
      return onPointerCancel(event);
    if (onPointerSignal != null && event is PointerSignalEvent)
      return onPointerSignal(event);
  }

如果使用的是GestureDetector的话，build方法会为我们添加很多处理手势的方法类，如TapGestureRecognizer，通过处理手势识别后，最终返回的是RawGestureDetector，涉及的代码如下。

 final Map<Type, GestureRecognizerFactory> gestures = <Type, GestureRecognizerFactory>{};
 
    if (
      onTapDown != null ||
      onTapUp != null ||
      onTap != null ||
      onTapCancel != null ||
      onSecondaryTapDown != null ||
      onSecondaryTapUp != null ||
      onSecondaryTapCancel != null
    ) {
      gestures[TapGestureRecognizer] = GestureRecognizerFactoryWithHandlers<TapGestureRecognizer>(
        () => TapGestureRecognizer(debugOwner: this),
        (TapGestureRecognizer instance) {
          instance
            ..onTapDown = onTapDown
            ..onTapUp = onTapUp
            ..onTap = onTap
            ..onTapCancel = onTapCancel
            ..onSecondaryTapDown = onSecondaryTapDown
            ..onSecondaryTapUp = onSecondaryTapUp
            ..onSecondaryTapCancel = onSecondaryTapCancel;
        },
      );
    }
 
    if (onDoubleTap != null) {
      gestures[DoubleTapGestureRecognizer] = GestureRecognizerFactoryWithHandlers<DoubleTapGestureRecognizer>(
        () => DoubleTapGestureRecognizer(debugOwner: this),
        (DoubleTapGestureRecognizer instance) {
          instance
            ..onDoubleTap = onDoubleTap;
        },
      );
    }
 
    if (onLongPress != null ||
        onLongPressUp != null ||
        onLongPressStart != null ||
        onLongPressMoveUpdate != null ||
        onLongPressEnd != null) {
      gestures[LongPressGestureRecognizer] = GestureRecognizerFactoryWithHandlers<LongPressGestureRecognizer>(
        () => LongPressGestureRecognizer(debugOwner: this),
        (LongPressGestureRecognizer instance) {
          instance
            ..onLongPress = onLongPress
            ..onLongPressStart = onLongPressStart
            ..onLongPressMoveUpdate = onLongPressMoveUpdate
            ..onLongPressEnd =onLongPressEnd
            ..onLongPressUp = onLongPressUp;
        },
      );
    }
 
    if (onVerticalDragDown != null ||
        onVerticalDragStart != null ||
        onVerticalDragUpdate != null ||
        onVerticalDragEnd != null ||
        onVerticalDragCancel != null) {
      gestures[VerticalDragGestureRecognizer] = GestureRecognizerFactoryWithHandlers<VerticalDragGestureRecognizer>(
        () => VerticalDragGestureRecognizer(debugOwner: this),
        (VerticalDragGestureRecognizer instance) {
          instance
            ..onDown = onVerticalDragDown
            ..onStart = onVerticalDragStart
            ..onUpdate = onVerticalDragUpdate
            ..onEnd = onVerticalDragEnd
            ..onCancel = onVerticalDragCancel
            ..dragStartBehavior = dragStartBehavior;
        },
      );
    }
 
    if (onHorizontalDragDown != null ||
        onHorizontalDragStart != null ||
        onHorizontalDragUpdate != null ||
        onHorizontalDragEnd != null ||
        onHorizontalDragCancel != null) {
      gestures[HorizontalDragGestureRecognizer] = GestureRecognizerFactoryWithHandlers<HorizontalDragGestureRecognizer>(
        () => HorizontalDragGestureRecognizer(debugOwner: this),
        (HorizontalDragGestureRecognizer instance) {
          instance
            ..onDown = onHorizontalDragDown
            ..onStart = onHorizontalDragStart
            ..onUpdate = onHorizontalDragUpdate
            ..onEnd = onHorizontalDragEnd
            ..onCancel = onHorizontalDragCancel
            ..dragStartBehavior = dragStartBehavior;
        },
      );
    }
 
    if (onPanDown != null ||
        onPanStart != null ||
        onPanUpdate != null ||
        onPanEnd != null ||
        onPanCancel != null) {
      gestures[PanGestureRecognizer] = GestureRecognizerFactoryWithHandlers<PanGestureRecognizer>(
        () => PanGestureRecognizer(debugOwner: this),
        (PanGestureRecognizer instance) {
          instance
            ..onDown = onPanDown
            ..onStart = onPanStart
            ..onUpdate = onPanUpdate
            ..onEnd = onPanEnd
            ..onCancel = onPanCancel
            ..dragStartBehavior = dragStartBehavior;
        },
      );
    }
 
    if (onScaleStart != null || onScaleUpdate != null || onScaleEnd != null) {
      gestures[ScaleGestureRecognizer] = GestureRecognizerFactoryWithHandlers<ScaleGestureRecognizer>(
        () => ScaleGestureRecognizer(debugOwner: this),
        (ScaleGestureRecognizer instance) {
          instance
            ..onStart = onScaleStart
            ..onUpdate = onScaleUpdate
            ..onEnd = onScaleEnd;
        },
      );
    }
 
    if (onForcePressStart != null ||
        onForcePressPeak != null ||
        onForcePressUpdate != null ||
        onForcePressEnd != null) {
      gestures[ForcePressGestureRecognizer] = GestureRecognizerFactoryWithHandlers<ForcePressGestureRecognizer>(
        () => ForcePressGestureRecognizer(debugOwner: this),
        (ForcePressGestureRecognizer instance) {
          instance
            ..onStart = onForcePressStart
            ..onPeak = onForcePressPeak
            ..onUpdate = onForcePressUpdate
            ..onEnd = onForcePressEnd;
        },
      );
    }
 
    return RawGestureDetector(
      gestures: gestures,
      behavior: behavior,
      excludeFromSemantics: excludeFromSemantics,
      child: child,
    );

并且，RawGestureDetector默认使用的也是Listener，它注册了手指按下的方法，分发的时候Down事件是sdk默认处理的。

 void _handlePointerDown(PointerDownEvent event) {

    assert(_recognizers != null);
    for (GestureRecognizer recognizer in _recognizers.values)
      recognizer.addPointer(event);
  }

此方法会向Binding路由器中注册那些需要处理的事件，假如我们只声明了点击事件，那么集合中负责添加的GestureRecognizer的实现类就是TapGestureRecognizer，接下来我们看一下addPointer方法。

 void addPointer(PointerDownEvent event) {
    _pointerToKind[event.pointer] = event.kind;
    if (isPointerAllowed(event)) {
      addAllowedPointer(event);
    } else {
      handleNonAllowedPointer(event);
    }
  }

bool isPointerAllowed(PointerDownEvent event) {
    switch (event.buttons) {
      case kPrimaryButton:
        if (onTapDown == null &&
            onTap == null &&
            onTapUp == null &&
            onTapCancel == null)
          return false;
        break;
      case kSecondaryButton:
        if (onSecondaryTapDown == null &&
            onSecondaryTapUp == null &&
            onSecondaryTapCancel == null)
          return false;
        break;
      default:
        return false;
    }
    return super.isPointerAllowed(event);
  }

isPointerAllowed方法的作用就是用来判定当前的手势，默认返回false，如果事件比命中，接下来执行addAllowedPointer方法，如下所示。

void addAllowedPointer(PointerDownEvent event) {
    startTrackingPointer(event.pointer, event.transform);
    if (state == GestureRecognizerState.ready) {
      state = GestureRecognizerState.possible;
      primaryPointer = event.pointer;
      initialPosition = OffsetPair(local: event.localPosition, global: event.position);
      if (deadline != null)
        _timer = Timer(deadline, () => didExceedDeadlineWithEvent(event));
    }
    
void startTrackingPointer(int pointer, [Matrix4 transform]) {
    GestureBinding.instance.pointerRouter.addRoute(pointer, handleEvent, transform);
    _trackedPointers.add(pointer);
    assert(!_entries.containsValue(pointer));
    _entries[pointer] = _addPointerToArena(pointer);
  }

这两个方法的主要作用就是用来将当前的handleEvent方法添加到GestureBinding路由器里面去，而_addPointerToArena是就是添加处理事件的具体逻辑。接下来，我们来看一下GestureBinding里面的handleEvent函数的事件分发逻辑。

void handleEvent(PointerEvent event, HitTestEntry entry) {
 
    pointerRouter.route(event);
    if (event is PointerDownEvent) {
 
      gestureArena.close(event.pointer);
    } else if (event is PointerUpEvent) {
 
      gestureArena.sweep(event.pointer);
    } else if (event is PointerSignalEvent) {
      pointerSignalResolver.resolve(event);
    }
  }
}

如果手指按下的时候GestureRecognizer的handleEvent方法没有决策出到底哪个控件会成为事件的处理者，那么会执行 gestureArena.close()方法，如下所示。

void close(int pointer) {
    final _GestureArena state = _arenas[pointer];
    if (state == null)
      return; // This arena either never existed or has been resolved.
    state.isOpen = false;
    assert(_debugLogDiagnostic(pointer, 'Closing', state));
    _tryToResolveArena(pointer, state);
  }

如果未决策出哪个控件处理事件的时候，state.isOpen此时被标记为false，也即是关闭手势的处理。

void _tryToResolveArena(int pointer, _GestureArena state) {
    assert(_arenas[pointer] == state);
    assert(!state.isOpen);
    if (state.members.length == 1) {
      scheduleMicrotask(() => _resolveByDefault(pointer, state));
    } else if 
 
(state.members.isEmpty) {
      _arenas.remove(pointer);
      assert(_debugLogDiagnostic(pointer, 'Arena empty.'));
    } 
else if (state.eagerWinner != null) {
      assert(_debugLogDiagnostic(pointer, 'Eager winner: ${state.eagerWinner}'));
      _resolveInFavorOf(pointer, state, state.eagerWinner);
    }
  }

如果手势竞争中，有竞争胜出者，则由胜出者执行事件处理，如下所示。

void _resolveInFavorOf(int pointer, _GestureArena state, GestureArenaMember member) {
    assert(state == _arenas[pointer]);
    assert(state != null);
    assert(state.eagerWinner == null || state.eagerWinner == member);
    assert(!state.isOpen);
    _arenas.remove(pointer);
    //其他的命中全部拒绝
    for (GestureArenaMember rejectedMember in state.members) {
      if (rejectedMember != member)
        rejectedMember.rejectGesture(pointer);
    }
    member.acceptGesture(pointer);
  }

如果事件处理中没有具体的事件处理对象，将会默认采用最底层的的叶子节点控件作为事件处理者，也就是说最内层的那个控件将消耗事件。也就是说，如果使用GestureRecognizer来识别手势事件时，最终事件会被最内层的GestureRecognizer消耗，这和Android单个控件消耗事件差不多，所以嵌套滚动总是先滚动内层，先被内层消耗，然后再执行外层。

参考： Flutter 事件分发