ZIm/crates/gpui/docs/key_dispatch.md

Key Dispatch

GPUI is designed for keyboard-first interactivity.

To expose functionality to the mouse, you render a button with a click handler.

To expose functionality to the keyboard, you bind an action in a key context.

Actions are similar to framework-level events like MouseDown, KeyDown, etc, but you can define them yourself:

mod menu {
    #[gpui::action]
    struct MoveUp;

    #[gpui::action]
    struct MoveDown;
}

Actions are frequently unit structs, for which we have a macro. The above could also be written:

mod menu {
    actions!(gpui, [MoveUp, MoveDown]);
}

Actions can also be more complex types:

mod menu {
    #[gpui::action]
    struct Move {
        direction: Direction,
        select: bool,
    }
}

To bind actions, chain on_action on to your element:

impl Render for Menu {
    fn render(&mut self, window: &mut Window, cx: &mut ModelContext<Self>) -> impl Component {
        div()
            .on_action(|this: &mut Menu, move: &MoveUp, window: &mut Window, cx: &mut ModelContext<Menu>| {
                // ...
            })
            .on_action(|this, move: &MoveDown, cx| {
                // ...
            })
            .children(unimplemented!())
    }
}

In order to bind keys to actions, you need to declare a key context for part of the element tree by calling key_context.

impl Render for Menu {
    fn render(&mut self, window: &mut Window, cx: &mut ModelContext<Self>) -> impl Component {
        div()
            .key_context("menu")
            .on_action(|this: &mut Menu, move: &MoveUp, window: &mut Window, cx: &mut ModelContext<Menu>| {
                // ...
            })
            .on_action(|this, move: &MoveDown, cx| {
                // ...
            })
            .children(unimplemented!())
    }
}

Now you can target your context in the keymap. Note how actions are identified in the keymap by their fully-qualified type name.

{
  "context": "menu",
  "bindings": {
    "up": "menu::MoveUp",
    "down": "menu::MoveDown"
  }
}

If you had opted for the more complex type definition, you'd provide the serialized representation of the action alongside the name:

{
  "context": "menu",
  "bindings": {
    "up": ["menu::Move", {direction: "up", select: false}]
    "down": ["menu::Move", {direction: "down", select: false}]
    "shift-up": ["menu::Move", {direction: "up", select: true}]
    "shift-down": ["menu::Move", {direction: "down", select: true}]
  }
}