Crate bevy_sequential_actions

source ·
Expand description

Bevy Sequential Actions

A Bevy library that aims to execute a queue of various actions in a sequential manner. This generally means that one action runs at a time, and when it is done, the next action will start and so on until the queue is empty.

Getting Started

Plugin

In order for everything to work, the SequentialActionsPlugin must be added to your App.

use bevy::prelude::*;
use bevy_sequential_actions::*;

fn main() {
    App::new()
        .add_plugins(DefaultPlugins)
        .add_plugin(SequentialActionsPlugin)
        .run();
}
Modifying Actions

An action is anything that implements the Action trait, and can be added to any Entity that contains the ActionsBundle. An entity with actions is referred to as an agent. See the ModifyActions trait for available methods.

fn setup(mut commands: Commands) {
    let agent = commands.spawn(ActionsBundle::new()).id();
    commands
        .actions(agent)
        .add(action_a)
        .add_parallel(actions![
            action_b,
            action_c
        ])
        .repeat(Repeat::Forever)
        .order(AddOrder::Back)
        .add(action_d)
        // ...
}
Implementing an Action

The Action trait contains two methods:

  • The on_start method which is called when an action is started.
  • The on_stop method which is called when an action is stopped.

In order for the action queue to advance, every action has to somehow signal when they are finished. There are two ways of doing this:

  • Using the ActionFinished component on an agent. By default, a system at the end of the frame will advance the queue if all active actions are finished. This is the typical approach as it composes well with other actions running in parallel.
  • Calling the next method on an agent. This simply advances the queue at the end of the current stage it was called in. Useful for short one-at-a-time actions.

A simple wait action follows.

pub struct WaitAction {
    duration: f32, // Seconds
    current: Option<f32>, // None
}

impl Action for WaitAction {
    fn on_start(&mut self, agent: Entity, world: &mut World, _commands: &mut ActionCommands) {
        // Take current duration (if paused), or use full duration
        let duration = self.current.take().unwrap_or(self.duration);

        // Run the wait system on the agent
        world.entity_mut(agent).insert(Wait(duration));
    }

    fn on_stop(&mut self, agent: Entity, world: &mut World, reason: StopReason) {
        // Remove the wait component from the agent
        let wait = world.entity_mut(agent).take::<Wait>();

        // Store current duration when paused
        if let StopReason::Paused = reason {
            self.current = Some(wait.unwrap().0);
        }
    }
}

#[derive(Component)]
struct Wait(f32);

fn wait_system(mut wait_q: Query<(&mut Wait, &mut ActionFinished)>, time: Res<Time>) {
    for (mut wait, mut finished) in wait_q.iter_mut() {
        wait.0 -= time.delta_seconds();

        // Confirm finished state every frame
        if wait.0 <= 0.0 {
            finished.confirm_and_reset();
        }
    }
}
Warning

One thing to keep in mind is that you should not modify actions using World inside the Action trait. We cannot borrow a mutable action from an agent while also passing a mutable world to it. Since an action is detached from an agent when the trait methods are called, the logic for advancing the action queue will not work properly.

This is why ActionCommands was created, so you can modify actions inside the Action trait in a deferred way.

pub struct SetStateAction<S: States>(S);

impl<S: States> Action for SetStateAction<S> {
    fn on_start(&mut self, agent: Entity, world: &mut World, commands: &mut ActionCommands) {
        // Set state
        world.resource_mut::<NextState<S>>().set(self.0.clone());

        // Bad. The action queue will advance immediately.
        world.actions(agent).next();

        // Good. The action queue will advance a bit later.
        commands.actions(agent).next();

        // Also good. Does the same as above.
        commands.add(move |w: &mut World| {
            w.actions(agent).next();
        });

        // Also good. By default, the action queue will advance at the end of the frame.
        world.get_mut::<ActionFinished>(agent).unwrap().confirm_and_persist();
    }

    fn on_stop(&mut self, _agent: Entity, _world: &mut World, _reason: StopReason) {}
}

Macros

  • actions
    Helper macro for creating a collection of boxed actions.

Structs

Enums

Traits