This document explains how to listen for, and respond to, gestures in Flutter. Examples of gestures include taps, drags, and scaling.
The gesture system in Flutter has two separate layers. The first layer has raw pointer events that describe the location and movement of pointers (for example, touches, mice, and styli) across the screen. The second layer has gestures that describe semantic actions that consist of one or more pointer movements.
Pointers represent raw data about the user’s interaction with the device’s screen. There are four types of pointer events:
- The pointer has contacted the screen at a particular location.
- The pointer has moved from one location on the screen to another.
- The pointer has stopped contacting the screen.
- Input from this pointer is no longer directed towards this app.
On pointer down, the framework does a hit test on your app to determine which widget exists at the location where the pointer contacted the screen. The pointer down event (and subsequent events for that pointer) are then dispatched to the innermost widget found by the hit test. From there, the events bubble up the tree and are dispatched to all the widgets on the path from the innermost widget to the root of the tree. There is no mechanism for canceling or stopping pointer events from being dispatched further.
To listen to pointer events directly from the widgets layer, use a
Listener widget. However, generally,
consider using gestures (as discussed below) instead.
Gestures represent semantic actions (for example, tap, drag, and scale) that are recognized from multiple individual pointer events, potentially even multiple individual pointers. Gestures can dispatch multiple events, corresponding to the lifecycle of the gesture (for example, drag start, drag update, and drag end):
- A pointer that might cause a tap has contacted the screen at a particular location.
- A pointer that triggers a tap has stopped contacting the screen at a particular location.
- The pointer that previously triggered the
onTapDownhas also triggered
onTapUpwhich ends up causing a tap.
- The pointer that previously triggered the
onTapDownwon’t end up causing a tap.
- The user has tapped the screen at the same location twice in quick succession.
- A pointer has remained in contact with the screen at the same location for a long period of time.
- A pointer has contacted the screen and might begin to move vertically.
- A pointer that is in contact with the screen and moving vertically has moved in the vertical direction.
- A pointer that was previously in contact with the screen and moving vertically is no longer in contact with the screen and was moving at a specific velocity when it stopped contacting the screen.
- A pointer has contacted the screen and might begin to move horizontally.
- A pointer that is in contact with the screen and moving horizontally has moved in the horizontal direction.
- A pointer that was previously in contact with the screen and moving horizontally is no longer in contact with the screen and was moving at a specific velocity when it stopped contacting the screen.
- A pointer has contacted the screen and might begin to move
horizontally or vertically. This callback crashes if
- A pointer that is in contact with the screen and is moving
in the vertical or horizontal direction. This callback
- A pointer that was previously in contact with screen
is no longer in contact with the screen and is moving
at a specific velocity when it stopped contacting the screen.
This callback crashes if
Adding gesture detection to widgets
To listen to gestures from the widgets layer,
If you’re using Material Components,
many of those widgets already respond to taps or gestures.
respond to presses (taps), and
responds to swipes to trigger scrolling.
If you aren’t using those widgets, but you want the
“ink splash” effect on a tap, you can use
At a given location on screen, there might be multiple gesture detectors. For example:
ListTilehas a tap recognizer that responds to the entire
ListTile, and a nested one around a trailing icon button. The screen rect of the trailing icon is now covered by two gesture recognizers that need to negotiate for who handles the gesture if it turns out to be a tap.
- A single
GestureDetectorcovers a screen area configured to handle multiple gestures, such as a long press and a tap. The
taprecognizer must now negotiate with the
long pressrecognizer when the user touches that part of the screen. Depending on what happens next with that pointer, one of the two recognizers receives the gesture, or neither receives the gesture if the user performs something that’s neither a tap nor a long press.
All of these gesture detectors listen to the stream
of pointer events as they flow past and attempt to recognize
specific gestures. The
GestureDetector widget decides
which gestures to attempt to recognize based on which of its
callbacks are non-null.
When there is more than one gesture recognizer for a given pointer on the screen, the framework disambiguates which gesture the user intends by having each recognizer join the gesture arena. The gesture arena determines which gesture wins using the following rules:
At any time, a recognizer can eliminate itself and leave the arena. If there’s only one recognizer left in the arena, that recognizer wins.
At any time, a recognizer can declare itself the winner, causing all of the remaining recognizers to lose.
For example, when disambiguating horizontal and vertical dragging, both recognizers enter the arena when they receive the pointer down event. The recognizers observe the pointer move events. If the user moves the pointer more than a certain number of logical pixels horizontally, the horizontal recognizer declares the win and the gesture is interpreted as a horizontal drag. Similarly, if the user moves more than a certain number of logical pixels vertically, the vertical recognizer declares itself the winner.
The gesture arena is beneficial when there is only a horizontal (or vertical) drag recognizer. In that case, there is only one recognizer in the arena and the horizontal drag is recognized immediately, which means the first pixel of horizontal movement can be treated as a drag and the user won’t need to wait for further gesture disambiguation.