How to get coordinates relative to scene's origin?

Krumelur

In my game setup I have a class "Cursor" which derives from "Sprite". It's a child of a ConvasLayer called "UILayer" which holds UI elements that's should remain static (and can be clicked). Depending on mouse movement, click, and touch events, the cursor's "Position" is updated. If I print the cursor's position, I can see that 0/0 is always top left of my gaming window and bottom right is 479/269, independent of viewport settings.

My goal is to find out what tile on the TileMap was clicked. I'm using my tilemap's world_to_map() function, however, I cannot get the input coordinates right, because obviously the screen coordinates won't work.

In the example below, the tilemap's 0/0 is where the cursor is pointing to. In the canvas layer's coordinates that's 240/50. So for the given map, I'd have to subtract 240 on X and 50 on Y to get my coordinates. But other maps may have a different origin.

How can I achieve my goal to convert from screen coordinates to cell position on the tilemap while taking into account the camera position (child of the player character)?

cybereality

I believe you can call:

get_global_mouse_position()

Krumelur

I could, yes. But I don't want to use this. I don't want the level to get input directly from the mouse because my UI provides other ways of moving the cursor around. Also, I already have the global position and I can pass it down to the current level or room. So that's not the problem.

Krumelur

OK, now I found this to work but still wondering if it's correct (C#):

var vpTransform = GetViewportTransform();
var localPos = screenPos - vpTransform.origin;
var tilePos = _floorMap.WorldToMap(localPos);

If this is correct, can someone explain why? :-) I'd love to learn.

cybereality

Yeah, that could work. It basically translates the local mouse screen coordinates to world coordinates, and then converts that to map coordinates. So that is a valid method.

Krumelur

I read a bit more and found that in order to take other transformations into account, I should use this:

var localPos = (GetViewportTransform() * GetGlobalTransform()).XformInv(screenPos);

However, I'm getting a warning that XformInv() is deprectaed and I should use the * operator. How do I fix this warning? Replacing XformInv with a multiplication operator yields different results, so there must be more to it.

TwistedTwigleg

I think you can use GetGlobalTransform().Inverse().Xform(screenPos) to invert a Transform and fix the warning.

Krumelur

XformInv is marked as "obsolete" and Xform is marked "deprecated". Trading one warning for another :-)

I mean it clearly tells me to use a a multiplication but I don't know how:

'Transform2D.XformInv(Vector2)' is obsolete: 'XformInv is deprecated. Use the multiplication operator (Vector2 * Transform2D) instead

Krumelur

Ha, just as I had sent the above post, I found it :-)

If you look at the Godot source it becomes clearer.

It (of course) matters whether you do "Vector Transform" or "Transform Vector". So the right operator is:

public static Vector2 operator *(Vector2 vector, Transform2D transform)

which means, this is how to get rid of XformInv() and the warning:

var localPos = screenPos * (GetViewportTransform() * GetGlobalTransform());

cybereality

Can you explain again why you can't use:

get_global_mouse_position()

This seems like the simple and obvious solution.

cybereality

And you already know what the position is because you are displaying a hand cursor on the screen.

Krumelur

Because the pointer isn’t necessarily controlled by a mouse or touch screen.

cybereality

Okay, I see. It's just a simple matter of addition and subtraction of the positions then (you find the viewport/screen position of the cursor and then subtract the position of the TileMap). It works with a Camera object as well.

func _input(event):
	if event.is_action_pressed("click"):
		var screen_pos = get_viewport().get_mouse_position()
		var world_pos = screen_pos - world.position
		change_tile(world_pos)
		
func change_tile(pos):
	var tile_pos = world.world_to_map(pos)
	world.set_cellv(tile_pos, -1)

And if you are using a Camera:

func _input(event):
	if event.is_action_pressed("click"):
		var screen_pos = get_viewport().get_mouse_position()
		var world_pos = screen_pos - world.position + camera.position + camera.offset - get_viewport().size * 0.5
		change_tile(world_pos)

If you don't want to use the mouse position, you can find the viewport/screen position of any object like this:

var screen_pos = cursor.get_global_transform_with_canvas()

Just tested it and it works.

cybereality

@Krumelur said: Because the pointer isn’t necessarily controlled by a mouse or touch screen.

You can fake a mouse position using warp_mouse_position. This allows you to move the mouse with the keyboard or controller (or by code, whatever) and then get_global_mouse_position() will return your fake position and not the actual mouse position.

Input.warp_mouse_position(fake_mouse_pos)