How to get UI Image screen bounds in a shader?

I need the bottom-left, width and height of a square Image in camera screen-space coordinates (0,0 at bottom-left of display window, 1,1 at top-right). It needs to work with 9-slice images and atlases.

Currently I pass it in as a Material property:

Code (CSharp):

1. Vector3[] worldCorners = new Vector3[4];
2.         image.rectTransform.GetWorldCorners(worldCorners);
3.         for (int i = 0; i < worldCorners.Length; i++) {
4.             Vector3 screenPoint = camera.WorldToScreenPoint(worldCorners[i]);
5.             screenPoint.x = screenPoint.x / Screen.width;
6.             screenPoint.y = screenPoint.y / Screen.height;
7.             worldCorners[i] = screenPoint;
8.         }
9.         Vector3 min = worldCorners[0];
10.         Vector3 max = worldCorners[2];
11.         Vector4 bounds = new Vector4(min.x, min.y, max.x - min.x, max.y - min.y);
12.         image.material.SetVector("_Rect", bounds);

This requires me to Instantiate a copy of the Material for each Image that uses it.

It's also completely broken in Scene view.

There must be a better way! Can I access this info (The Image's RectTransform.rect plus required matrix transforms) from inside the shader?

 

Yep.

Yep.

Nope.

Eh... not really, no.

You can fix the issue for the scene view by setting the value on the material during

onPreRender

or

OnWillRenderObject

.
https://docs.unity3d.com/ScriptReference/Camera-onPreRender.html
https://docs.unity3d.com/ScriptReference/Camera.OnWillRenderObject.html
Both of those give you access to the camera that will be used to render, which you should be using instead of the main camera. This includes the scene view camera. I'd suggest using

OnWillRenderObject

, but I honestly don't know if UI renderer components get that called or not, hence why I listed the other option as well.

To avoid instantiating the material every time, you could write a script that adds the data you need to the vertex data.
https://docs.unity3d.com/2018.3/Documentation/ScriptReference/UI.BaseMeshEffect.html
This is probably more efficient to render, but may not be noticeably faster.

 

Looks like OnWillRenderObject only runs on Renderers, not UI CanvasRenderers. Camera.OnPreRender did the trick though!

Piggybacking shader vars in the uv data also seems like a better system than instantiating a bunch of Materials, and means I'd still have a hope of batching draw calls together. This is a good general workaround for the lack of MaterialPropertyBlocks on UI, thanks!