How to detect collisions on the canvaslayer in Godot 4.x

[Daedalus]

Hi all! I’m trying to get this card sorting Godot 4.x project to work on a canvas layer; I want the cards to stay in the foreground on the UI while the camera pans across a scene in the background. But I just can’t get it to work.

Once the CardHand node (which contains the individual cards) is a child of a CanvasLayer node, Godot stops detecting collisions. Originally I thought this might have to do with the fact that the project uses intersect_point and not intersect_point_on_canvas but that method appears to have been removed in Godot 4.0 and I can’t find what, if anything, replaced it.

Is there a proper way to adapt a Godot 4.x project that uses intersect_point to work on the canvas layer? More specifically, how can I get this particular project working on a canvas layer? Thanks for any insight, I’m stumped!

[pennyloafers]

Your links appear to be incorrect or broken.

[pennyloafers]

Anyway the new function is called intersect_point and encapsulates the parameters in the query class

[Daedalus]

Thanks for following up, and sorry about the links, they should work now.

The original project already makes use of intersect_point and works as intended so long as it isn’t a child of a CanvasLayer. However, my goal is for the cards to be a part of the UI and thus on a CanvasLayer; when I make such a change it breaks the raycasting and I cannot figure out why.

There are no warnings or errors; it just simply performs differently if it’s a child of a Canvas Layer.

Image #1: I can sort Card1 between Card2 and Card 3 because the code functioned properly (detecting via raycasting that Card1 was dropped between two and three).

Image #2: Cards cannot be sorted once it is a child of a CanvasLayer; no collisions are detected and thus you cannot re-order the cards.

I know I’m missing something, but I can’t find any info on what I’m supposed to change between these two scenarios–from what I can tell intersect_point should function the same regardless but it’s not. Again, thanks for any help, it’s greatly appreciated!

[pennyloafers]

Every view port has a built in canvas layers, that’s why 2d works without adding a canvaslayer. When you add a canvas layer you probably need to adjust the layer index, as I’m sure it probably messes with ray casting if multiple layers are on the same index. (This is just a hunch)

Or if you did change the layer index, or it defaulted to another index, then you should also adjust the query property to which layer it’s targeting.

[Daedalus]

Ah, that was precisely the issue–I didn’t set/update the canvas instance ID (and did not realize the base viewport already counts as a canvas layer). Works as intended now. Thanks so much pennyloafers!!

In case it’s useful to anyone as clueless as me, here’s the updated code:

Task<Array<Dictionary>> raycastQuery = GetRaycastQuery(new PhysicsPointQueryParameters2D()
{
	Position = GetGlobalMousePosition(),
	CanvasInstanceId = hudNode.GetInstanceId(),
	CollideWithAreas = true,
	CollideWithBodies = false
});