How to handle tilemap-based collisions

nphysics
1

I’m making a simple platformer with a tilemap-based world. My levels are laid out like so:

image

For every wall block, I create a Cuboid shape and attach it to the Ground body, as recommended in the guide:

if WALL_TILE_TYPES.contains(&tile.type_) {
    let shape = ShapeHandle::new(Cuboid::new(Vector2::new(
        Tile::SIZE / 2. - 0.01,
        Tile::SIZE / 2. - 0.01,
    )));
    let translation = vector;
    physics.build_collider(
        ColliderDesc::new(shape).translation(translation),
        physics.ground_handle,
        false,
    );
}

My game starts lagging when I have more than 50 blocks, which seems like a completely reasonable amount. Is there a better way to handle tilemap-based collisions? Are there any ways I can optimize this, aside from writing an algorithm that decreases the number of colliders in my world?

2

Hi! It’s surprising that 50 blocks makes the game lag. Are you compiling in release mode? Because debug is much, much, slower than release.

3

You’re absolutely right — release mode runs perfectly. In general, would you say that I’m using nphysics in the correct way for this problem?

4

I’m curious as well to know if that is the correct way to implement tile-based collison using nphysics/ncollide. :slight_smile:

And (assuming your player collision shape is a cuboid too), what if your player collision shape is a sphere instead, would that still work?

5

I have never written a tile-based game myself so I don’t really know what the best-practice are. But I can see two options:

  1. Do as it is done in this thread, with one Cuboid collider per tile.
  2. Or extract the contour of the ground and generate a single Polyline representing this contour.

I’m not sure which is preferred in what situation.