integrated new shape based array triangles

[rhayes777]

It's now possible to perform triangle upsampling on Points, Circles, Triangles, Squares and Polygons.

from autoarray.structures.triangles.shape import Polygon, Triangle, Square
from autolens.point.solver.shape_solver import ShapeSolver

solver = ShapeSolver.for_grid(
    grid=grid,
    pixel_scale_precision=0.01,
)

result = solver.solve(
    tracer=tracer
    shape=Circle(
        0.0,
        0.0,
        radius=0.01,
    ),
)

square = Square(
    top=0.0,
    left=0.0,
    bottom=0.1,
    left=0.1,
)

triangle = Triangle(
    (0.0, 0.0), 
    (0.1, 0.0), 
    (0.0, 0.1), 
)

polygon = Polygon([
    (0.0, 0.0), 
    (0.1, 0.0), 
    (0.0, 0.1), 
])

[Jammy2211]

def test_solver_basic(solver):
    result = solver.solve(
        tracer=NullTracer(),
        shape=Circle(
            0.0,
            0.0,
            radius=0.01,
        ),
    )
    assert list(map(tuple, result)) == [
        (-0.0029826688901819823, -0.0078125),
        (-0.0029826688901819823, 0.0078125),
    ]

What exactly does this result mean?

I guess I mean what is the result of a ShapeSolver?

[rhayes777]

def test_solver_basic(solver):
    result = solver.solve(
        tracer=NullTracer(),
        shape=Circle(
            0.0,
            0.0,
            radius=0.01,
        ),
    )
    assert list(map(tuple, result)) == [
        (-0.0029826688901819823, -0.0078125),
        (-0.0029826688901819823, 0.0078125),
    ]

What exactly does this result mean?

I guess I mean what is the result of a ShapeSolver?

Yeah it's the means of the triangles that intersect the circle at that resolution. Not a great assertion.

[Jammy2211]

Does the result contain information on the image-plane coordinates?

For a PointSolver, the solve function returns a Grid2DIrregular of the (y,x) coordinates in the image-plane that map to the source-plane centre.

Intuitively I would expect the ShapeSolver's solve function to return a Grid2DIrregular of all (y,x) coordinates in the image-plane that are at the centres of pixels which map within the shape, maybe? To compute a magnification you would use the ratio of this with the shape's source-plane area.

[rhayes777]

Does the result contain information on the image-plane coordinates?

For a PointSolver, the solve function returns a Grid2DIrregular of the (y,x) coordinates in the image-plane that map to the source-plane centre.

Intuitively I would expect the ShapeSolver's solve function to return a Grid2DIrregular of all (y,x) coordinates in the image-plane that are at the centres of pixels which map within the shape, maybe? To compute a magnification you would use the ratio of this with the shape's source-plane area.

Yeah we should return a richer object I guess. It should certainly be possible to compute the area of the triangles in the image plane and the area of the shape in the source plane then take the ratio

[Jammy2211]

Yeah, if something that makes that simple is included in this PR I have a user I can immediate ask to test the code lol.

[rhayes777]

Yeah, if something that makes that simple is included in this PR I have a user I can immediate ask to test the code lol.

Cool working on something