Can't backward through a transformation matrix

[simoncozens]

I'm guessing this project is pretty dead given the lack of updates, but filing this in case it bites someone else the way it bit me: if you're using a transformation matrix (the shape_to_canvas parameter to pydiffvg.ShapeGroup), gradients don't get transformed using the matrix. An obvious symptom of this is: if you flip the coordinate space upside down using a matrix, the updates to the shape will push it down when they should push it up and vice versa.

e.g. replacing

diffvg/apps/single_curve.py

Lines 24 to 25 in 85802a7

	path_group = pydiffvg.ShapeGroup(shape_ids = torch.tensor([0]),
	fill_color = torch.tensor([0.3, 0.6, 0.3, 1.0]))

with

transform = torch.tensor(
    [
        [1.0, 0.0, 0.0],
        [0.0, -1.0, 256.0],
        [0.0, 0.0, 1.0],
    ]
)
path_group = pydiffvg.ShapeGroup(
    shape_ids=torch.tensor([0]),
    fill_color=torch.tensor([0.3, 0.6, 0.3, 1.0]),
    shape_to_canvas=transform,
)

loss will increase, and the final result will look like this: