Error when adding bundle as a single component instead of silently failing

[zenMaya]

When you want to add *Components, you must use .spawn(*Components) and not .spawn((*Components, ..)) the spawning will fail, but there is no indication to the user, that it won't work.

This is also the case when adding .with(*Components) instead of .with_bundle(*Components)

This should definitely tell user, that it's not correct to add bundles this way and not silently fail but compile no problem.

[DJMcNab]

I wonder if it would be possible to cheat a little bit, by making the default Bundles !Sync, maybe by adding a PhantomData<*const ()> field?
Obviously we'd have to change the macros crate to add #[bundle(ignore)].
I'd probably also add #[bundle(optional)] at the same time.
I'm going to do some investigative work around this soon.

[CleanCut]

I wasted an hour of my life on this today (again). Can you spot my error?

struct Something;

fn setup(commands: &mut Commands, asset_server: Res<AssetServer>) {
    commands
        .spawn(Camera2dBundle::default())
        .spawn((Something,))
        .with(Text2dBundle {
            text: Text {
                value: "Some Text".to_string(),
                font: asset_server.load("fonts/FiraSans-Bold.ttf"),
                style: TextStyle {
                    font_size: 60.0,
                    color: Color::WHITE,
                    alignment: TextAlignment {
                        vertical: VerticalAlign::Center,
                        horizontal: HorizontalAlign::Center,
                    },
                },
            },
            ..Default::default()
        });
}

Answer: .with(...) was silently failing. It needed to be .with_bundle(...)

[DJMcNab]

I've had a new idea of how to 'fix' this, which would be to go super simple

fn illegal_component_system<T: Bundle>(q: Query(Entity, With<T>)){
for e in q.iter(){
 //...print message depending on settings
}
}

And then add a method on app to add this system

Possibly also add a field on App which would disable adding these

[cart]

I've had a new idea of how to 'fix' this, which would be to go super simple

Its a bit heavy for my taste to prevent what in most cases is a "newbie error". And would require somehow connecting the World impl and the Schedule impl to ensure a new system is inserted for each Bundle.

Haha its not a full solve, but we could consider un-implementing Bundle entirely for tuples. (ex: the only way to spawn something is through a pre-constructed Bundle derive, otherwise you must use commands.with())

It doesn't solve the commands.with(SomeBundle) problem though.

[alice-i-cecile]

As part of this, trying to access a bundle as a component (commonly as part of a query) should also fail or warn. See #403 for related work on disambiguating components and bundles.

[alice-i-cecile]

Fixed by #2254!