Using InputEvent

What is it?

Managing input is usually complex, no matter the OS or platform. To ease this a little, a special built-in type is provided, InputEvent. This datatype can be configured to contain several types of input events. Input events travel through the engine and can be received in multiple locations, depending on the purpose.

Here is a quick example, closing your game if the escape key is hit:

func _unhandled_input(event):
    if event is InputEventKey:
        if event.pressed and event.keycode == KEY_ESCAPE:
            get_tree().quit()

However, it is cleaner and more flexible to use the provided InputMap feature, which allows you to define input actions and assign them different keys. This way, you can define multiple keys for the same action (e.g. the keyboard escape key and the start button on a gamepad). You can then more easily change this mapping in the project settings without updating your code, and even build a key mapping feature on top of it to allow your game to change the key mapping at runtime!

You can set up your InputMap under Project > Project Settings > Input Map and then use those actions like this:

func _process(delta):
    if Input.is_action_pressed("ui_right"):
        # Move right.

How does it work?

Every input event is originated from the user/player (though it's possible to generate an InputEvent and feed them back to the engine, which is useful for gestures). The DisplayServer for each platform will read events from the operating system, then feed them to the root Window.

The window's Viewport does quite a lot of stuff with the received input, in order:

../../_images/input_event_flow.webp
  1. If the Viewport is embedding Windows, the Viewport tries to interpret the event in its capability as a Window-Manager (e.g. for resizing or moving Windows).

  2. Next if an embedded Window is focused, the event is sent to that Window and processed in the Windows Viewport and afterwards treated as handled. If no embedded Window is focused, the event is sent to the nodes of the current viewport in the following order.

  3. First of all, the standard Node._input() function will be called in any node that overrides it (and hasn't disabled input processing with Node.set_process_input()). If any function consumes the event, it can call Viewport.set_input_as_handled(), and the event will not spread any more. This ensures that you can filter all events of interest, even before the GUI. For gameplay input, Node._unhandled_input() is generally a better fit, because it allows the GUI to intercept the events.

  4. Second, it will try to feed the input to the GUI, and see if any control can receive it. If so, the Control will be called via the virtual function Control._gui_input() and the signal "gui_input" will be emitted (this function is re-implementable by script by inheriting from it). If the control wants to "consume" the event, it will call Control.accept_event() and the event will not spread any more. Use the Control.mouse_filter property to control whether a Control is notified of mouse events via Control._gui_input() callback, and whether these events are propagated further.

  5. If so far no one consumed the event, the Node._shortcut_input() callback will be called if overridden (and not disabled with Node.set_process_shortcut_input()). This happens only for InputEventKey, InputEventShortcut and InputEventJoypadButton. If any function consumes the event, it can call Viewport.set_input_as_handled(), and the event will not spread any more. The shortcut input callback is ideal for treating events that are intended as shortcuts.

  6. If so far no one consumed the event, the Node._unhandled_key_input() callback will be called if overridden (and not disabled with Node.set_process_unhandled_key_input()). This happens only if the event is a InputEventKey. If any function consumes the event, it can call Viewport.set_input_as_handled(), and the event will not spread any more. The unhandled key input callback is ideal for key events.

  7. If so far no one consumed the event, the Node._unhandled_input() callback will be called if overridden (and not disabled with Node.set_process_unhandled_input()). If any function consumes the event, it can call Viewport.set_input_as_handled(), and the event will not spread any more. The unhandled input callback is ideal for full-screen gameplay events, so they are not received when a GUI is active.

  8. If no one wanted the event so far, and Object Picking is turned on, the event is used for object picking. For the root viewport, this can also be enabled in Project Settings. In the case of a 3D scene if a Camera3D is assigned to the Viewport, a ray to the physics world (in the ray direction from the click) will be cast. If this ray hits an object, it will call the CollisionObject3D._input_event() function in the relevant physics object (bodies receive this callback by default, but areas do not. This can be configured through Area3D properties). In the case of a 2D scene, conceptually the same happens with CollisionObject2D._input_event().

When sending events to its child and descendant nodes, the viewport will do so, as depicted in the following graphic, in a reverse depth-first order, starting with the node at the bottom of the scene tree, and ending at the root node. Excluded from this process are Windows and SubViewports.

../../_images/input_event_scene_flow.png

This order doesn't apply to Control._gui_input(), which uses a different method based on event location or focused Control.

Since Viewports don't send events to other