Does bevy Tranform changes an object anchor point?

[frail kestrel]

This is my camera code:

use crate::{player::Player, world::MyMainWorld};
use bevy::prelude::*;
use bevy_voxel_world::prelude::*;

const CAMERA_DISTANCE: f32 = 1.0;
const HORIZONTAL_ANGLE: f32 = 45.0_f32.to_radians();
const VERTICAL_ANGLE: f32 = 35.0_f32.to_radians();

pub fn spawn_isometric_camera(
    mut commands: Commands,
    player_query: Query<&Transform, With<Player>>,
) {
    let player_position = player_query.single().unwrap().translation;

    let offset_x = CAMERA_DISTANCE * HORIZONTAL_ANGLE.cos() * VERTICAL_ANGLE.cos();
    let offset_y = CAMERA_DISTANCE * VERTICAL_ANGLE.sin();
    let offset_z = CAMERA_DISTANCE * HORIZONTAL_ANGLE.sin() * VERTICAL_ANGLE.cos();

    let camera_position = player_position + Vec3::new(offset_x, offset_y, offset_z);

    // Create transform looking at the player
    let transform =
        Transform::from_translation(camera_position).looking_at(player_position, Vec3::Y);

    commands.spawn((
        Camera3d::default(),
        Projection::Orthographic(OrthographicProjection {
            scale: 0.05, // zoom level
            near: -35.0, // Increase this (more negative = see more behind)
            far: 50.0,
            ..OrthographicProjection::default_3d()
        }),
        transform,
        VoxelWorldCamera::<MyMainWorld>::default(),
    ));
}

pub fn move_camera_towards_player(
    player_query: Query<&Transform, With<Player>>,
    mut camera_query: Query<&mut Transform, (With<Camera3d>, Without<Player>)>,
) {
    let player_position = player_query.single().unwrap().translation;
    let mut camera = camera_query.single_mut().unwrap();

    camera.translation.x = player_position.x;
    camera.translation.z = player_position.z;
}
``` In my mind, this should not work, this should position the camera on top of my player (on the first movement key stroke) and face forward, I should not be able to see my player, but I see it. Does the Transform thing changes the camera anchor point or something like that?

This is my main.rs file. I just load stuff sync:

fn main() {
    App::new()
        .add_plugins(DefaultPlugins)
        .add_plugins(VoxelWorldPlugin::with_config(world::MyMainWorld))
        .add_systems(
            Startup,
            (
                player::spawn_player,
                camera::spawn_isometric_camera,
                world::setup,
                world::create_voxel_scene,
            )
                .chain(),
        )
        .add_systems(FixedUpdate, player::bind_controls)
        .add_systems(Update, camera::move_camera_towards_player)
        .run();
}

[tawdry osprey]

orthographic cameras are quite different than usual perspective cameras and can be unintuitive

orthographic cameras view an arbitrary volume and their "anchor point" is usually within that volume

it would be true that you couldn't see the player if you changed near to be zero or positive

[frail kestrel]

If I reconstructed my ortographic camera with a perpective + fov = 0, would I avoid this problem? I feel like I don't know how to control this camera and this frustrates me

[tawdry osprey]

what is the problem you are trying to solve?

so far I understand you to have said "this shouldn't work but actually it works. why?"

[frail kestrel]

The camera right now is perfect, it works as I aim, but I'm unable to understand completely why it works. I got the correct near and far values on trial and error, I don't understand it.

[tawdry osprey]

the near and far values need to be set so they enclose the area around the origin point that you want to be visible.

that's all.

the origin of the camera as moved by its Transform, that is

[frail kestrel]

But what does this values represent? a radius?

[tawdry osprey]

the volume visible to an orthographic camera is a box

it has six faces, left right top bottom near far

the near and far values position where the near and far faces of the box go

[frail kestrel]

OOoooo

[tawdry osprey]

the main reason it is unintuitive to set them is that they make no visual difference until you make them too close (clipping the scene) or too distant (loss of depth buffer precision, Z-fighting)

[frail kestrel]

It's a pyramid like cube

[tawdry osprey]

no

a perspective camera has a frustum (truncated pyramid)

in an orthographic camera, instead of sloping sides making a pyramid, you have parallel sides making a box

[frail kestrel]

Awesome, I understand it now

you are pretty good at explaining it

This image on the docs would be convenient I think

[tawdry osprey]

you can sort of think of an orthographic camera as being like a perspective camera where the view point is infinitely far away, but we don't set them up that way because infinities make it harder to get the math right, so that's more of an abstract perspective than a practical one

[frail kestrel]

but we don't set them up that way because infinities make it harder to get the math right This is something I'm gonna need to understand later.

Those are two distances from the camera

loss of depth buffer precision
Do you have any link showing this problem?

[tawdry osprey]

https://en.wikipedia.org/wiki/Z-fighting

[frail kestrel]

Thank you, appreciate it a lot. You spared me a good time of stress and research

Now even the translation makes sense. The "origin" being considered on these vectors and distances is the player center. As when I set near = 0 the player is cut in half: